Beyond the Frontier: 2011 Frontier Nursing University Alumni Magazine
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS...
Transcript of OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS...
OPTIMAL ALLOCATION OF
WATER IN VILLAGE IRRIGATION
SYSTEMS OF SRI LANKA
Mohottala Gedara Kularatne
M.Sc. in Socio-Economic Information for Natural Resources Management, ITC,
University of Twente, The Netherlands, B.A. (Economics) Hons, University of
Peradeniya, Sri Lanka
Principal Supervisor: Associate Professor Clevo Wilson
Associate Supervisor: Professor Tim Robinson
Associate Supervisor: Professor Sean Pascoe
Submitted in fulfilment of the requirements for the degree of
Doctor of Philosophy (Research)
QUT Business School
Queensland University of Technology
Gardens Point Campus, Brisbane, Australia
May 2011
i
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA i
Keywords
Culture-based fisheries, equi-marginal principle, imposing theoretical consistency,
marginal value product, optimal allocation of water, rice farming, Sri Lanka,
stochastic frontier production function, technical efficiency, village irrigation
systems.
ii
ii OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA
Abstract
This PhD study examines whether water allocation becomes more productive
when it is re-allocated from „low‟ to „high‟ efficient alternative uses in village
irrigation systems (VISs) in Sri Lanka. Reservoir-based agriculture is a collective
farming economic activity, which inter-sectoral allocation of water is assumed to be
inefficient due to market imperfections and weak user rights. Furthermore, the
available literature shows that a „head-tail syndrome‟ is the most common issue for
intra-sectoral water management in „irrigation‟ agriculture. This research analyses
the issue of water allocation by using primary data collected from two surveys of 460
rice farmers and 325 fish farming groups in two administrative districts in Sri Lanka.
Technical efficiency estimates are undertaken for both rice farming and culture-
based fisheries (CBF) production. The equi-marginal principle is applied for inter
and intra-sectoral allocation of water. Welfare benefits of water re-allocation are
measured through consumer surplus estimation.
Based on these analyses, the overall findings of the thesis can be summarised
as follows. The estimated mean technical efficiency (MTE) for rice farming is 73%.
For CBF production, the estimated MTE is 33%. The technical efficiency
distribution is skewed to the left for rice farming, while it skewed to the right for
CBF production. The results show that technical efficiency of rice farming can be
improved by formalising transferability of land ownership and, therefore, water user
rights by enhancing the institutional capacity of Farmer Organisations (FOs). Other
effective tools for improving technical efficiency of CBF production are
strengthening group stability of CBF farmers, improving the accessibility of official
consultation, and attracting independent investments.
iii
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA iii
Inter-sectoral optimal allocation shows that the estimated inefficient volume of
water in rice farming, which can be re-allocated for CBF production, is 32%. With
the application of successive policy instruments (e.g., a community transferable
quota system and promoting CBF activities), there is potential for a threefold
increase in marginal value product (MVP) of total reservoir water in VISs. The
existing intra-sectoral inefficient volume of water use in tail-end fields and head-end
fields can potentially be removed by reducing water use by 10% and 23%
respectively and re-allocating this to middle fields. This re-allocation may enable a
twofold increase in MVP of water used in rice farming without reducing the existing
rice output, but will require developing irrigation practices to facilitate this re-
allocation.
Finally, the total productivity of reservoir water can be increased by
responsible village level institutions and primary level stakeholders (i.e., co-
management) sharing responsibility of water management, while allowing market
forces to guide the efficient re-allocation decisions. This PhD has demonstrated that
instead of farmers allocating water between uses haphazardly, they can now base
their decisions on efficient water use with a view to increasing water productivity.
Such an approach, no doubt will enhance farmer incomes and community welfare.
iv
iv OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA
Table of Contents
Keywords ................................................................................................................................................. i
Abstract ................................................................................................................................................... ii
Table of Contents ................................................................................................................................... iv
List of Figures ....................................................................................................................................... vii
List of Tables ...................................................................................................................................... viii
List of Abbreviations ............................................................................................................................. ix
Statement of Original Authorship .......................................................................................................... xi
Acknowledgments ................................................................................................................................. xii
Dedication ............................................................................................................................................. xv
CHAPTER 1: INTRODUCTION ....................................................................................................... 1
1.1 Overview ..................................................................................................................................... 1
1.2 Context ......................................................................................................................................... 4
1.3 General problem .......................................................................................................................... 6
1.4 Objectives of the thesis ................................................................................................................ 9
1.5 Research questions ..................................................................................................................... 10
1.6 Data analysis .............................................................................................................................. 12
1.7 Contribution to knowledge ........................................................................................................ 13
1.8 Thesis outline ............................................................................................................................. 14
CHAPTER 2: ALLOCATION OF WATER RESOURCES IN SRI LANKA .............................. 17
2.1 Introduction................................................................................................................................ 17
2.2 Water resource in Sri Lanka ...................................................................................................... 17
2.3 Water resources management and allocation ............................................................................. 20 2.3.1 Use of water resources in reservoir-based agriculture and related issues ....................... 23 2.3.2 Volume of water used for competing water demands .................................................... 25 2.3.3 Technical limitation of water allocation ......................................................................... 27
2.4 Reservoir water as a commodity ................................................................................................ 29 2.4.1 Missing markets for reservoir water allocation .............................................................. 29
2.5 Reservoir water as a common property issue of non-market solution ....................................... 31 2.5.1 Issues in non-market solutions for reservoir water allocation ........................................ 35
2.6 Chapter summary ....................................................................................................................... 36
CHAPTER 3: PRODUCTION FUNCTIONS AND OPTIMAL ALLOCATION OF WATER . 37
3.1 Introduction................................................................................................................................ 37
3.2 Theoretical overview of stochastic production frontier and analytical framwork ...................... 38 3.2.1 Frontier production functions ......................................................................................... 38 3.2.2 Technical efficiency and technical inefficiency ............................................................. 44 3.2.3 Selection of the functional forms and theoretical consistency ........................................ 47 3.2.4 Estimation of theoretical consistency ............................................................................. 50 3.2.5 Simple three step procedure for imposing monotonicity ................................................ 50 3.2.6 Estimation of technical efficiency .................................................................................. 51
v
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA v
3.2.7 Estimation of technical inefficiency ............................................................................... 52
3.3 Estimation of optimal allocation of water .................................................................................. 53 3.3.1 Marginal value product (MVP), equi-marginal principle ............................................... 53 3.3.2 MVP and technical efficiency (how derived from SPF) ................................................. 58 3.3.3 Estimation of inter-sectoral optimal allocation of water ................................................. 59 3.3.4 Estimation of intra-sectoral optimal allocation of water ................................................. 61
3.4 Estimation of consumer surplus of water re-allocation .............................................................. 62
3.5 Chapter summary ....................................................................................................................... 64
CHAPTER 4: DATA COLLECTION AND MODEL DEFINITION ........................................... 65
4.1 Introduction ................................................................................................................................ 65
4.2 Data ............................................................................................................................................ 65
4.3 Study areas ................................................................................................................................. 66
4.4 Sample selection methods .......................................................................................................... 67
4.5 Selected sample .......................................................................................................................... 67 4.5.1 Rice farmer study ............................................................................................................ 67 4.5.2 CBF farmer study ........................................................................................................... 68
4.6 Data collection method .............................................................................................................. 69 4.6.1 Rice farmer survey .......................................................................................................... 69 4.6.2 CBF farmer survey ......................................................................................................... 70
4.7 Model definition......................................................................................................................... 70
4.8 Chapter summary ....................................................................................................................... 75
CHAPTER 5: EFFICIENT WATER USAGE IN VILLAGE IRRIGATION SYSTEMS FOR
RICE FARMING ................................................................................................................................ 77
5.1 Introduction; ............................................................................................................................... 77
5.2 Rice production .......................................................................................................................... 77
5.3 Literature review ........................................................................................................................ 79
5.4 Empirical model ......................................................................................................................... 84
5.5 Results ........................................................................................................................................ 86
5.6 Discussion .................................................................................................................................. 93
5.7 Chapter summary ....................................................................................................................... 99
CHAPTER 6: EFFICIENT WATER USAGE IN VILLAGE IRRIGATION SYSTEMS FOR
CULTURE-BASED FISHERIES PRODUCTION ........................................................................ 101
6.1 Introduction .............................................................................................................................. 101
6.2 CBF production ........................................................................................................................ 101
6.3 Literature review ...................................................................................................................... 105
6.4 Empirical model ....................................................................................................................... 107
6.5 Results ...................................................................................................................................... 109
6.6 Discussion ................................................................................................................................ 116
6.7 Chapter summary ..................................................................................................................... 122
CHAPTER 7: INTER-SECTORAL OPTIMAL ALLOCATION OF WATER ......................... 125
7.1 Introduction .............................................................................................................................. 125
7.2 Inter- sectoral water allocation ................................................................................................. 125
vi
vi OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA
7.3 The current water allocation system ........................................................................................ 127
7.4 Optimal allocation of water ..................................................................................................... 129
7.5 Empirical model ....................................................................................................................... 133
7.6 Results ..................................................................................................................................... 134
7.7 Discussion ................................................................................................................................ 137
7.8 Chapter summary ..................................................................................................................... 140
CHAPTER 8: INTRA-SECTORAL OPTIMAL ALLOCATION OF WATER ......................... 141
8.1 Introduction.............................................................................................................................. 141
8.2 Intra-sectoral water allocation.................................................................................................. 141
8.3 Literature review ...................................................................................................................... 143
8.4 Empirical models and results ................................................................................................... 147
8.5 Results ..................................................................................................................................... 148
8.6 Discussion ................................................................................................................................ 151
8.7 Chapter summary ..................................................................................................................... 156
CHAPTER 9: RESERVOIR WATER RE-ALLOCATION AND COMMUNITY WELFARE157
9.1 Introduction.............................................................................................................................. 157
9.2 Reservoir water re-allocation ................................................................................................... 157
9.3 Literature review ...................................................................................................................... 159
9.4 Results and estimation of potential gains from water re-allocation ......................................... 161
9.5 Discussion: Issues associated with reservoir water re-allocation ............................................. 164 9.5.1 Establishing water user rights ....................................................................................... 165 9.5.2 Internalising CBF externalities ..................................................................................... 167 9.5.3 Co-managment as a mechanism for water re-allocation ............................................... 169
9.6 Chapter summary ..................................................................................................................... 173
CHAPTER 10: CONCLUDING REMARKS ............................................................................. 175
10.1 Conclusions.............................................................................................................................. 175
10.2 Summary, key findings and discussions .................................................................................. 175
10.3 Policy implications .................................................................................................................. 179
10.4 Limitations and future direction of research ............................................................................ 185
BIBLIOGRAPHY ............................................................................................................................. 187
APPENDICES ................................................................................................................................... 205 Appendix A ............................................................................................................................. 205
Appendix B ............................................................................................................................. 209
Appendix C ............................................................................................................................. 211
Appendix D ............................................................................................................................. 213
Appendix E ............................................................................................................................. 221
Appendix F ............................................................................................................................. 229
Appendix G ............................................................................................................................. 235
Appendix H ............................................................................................................................. 253
Appendix I ............................................................................................................................. 257
Appendix J ............................................................................................................................. 261
vii
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA vii
List of Figures
Figure 1.1. Identification of the research problem. ................................................................................. 8
Figure 1.2.The water allocation problem between competing uses. ....................................................... 9
Figure 1.3. The chapter outline of the thesis. ........................................................................................ 15
Figure 2.1. Method of water allocation in village reservoirs................................................................. 25
Figure 2.2. Semantic diagram of intra-sectoral water allocation. ......................................................... 26
Figure 2.3. Graphical presentation of land and water relationship. ...................................................... 27
Figure 3.1. Overall analytical framework. ............................................................................................ 38
Figure 3.2. Simple isoquant diagram of input-orientated TE measures. ............................................... 45
Figure 3.3. Rice-water frontier production function. ............................................................................ 46
Figure 3.4. Concavity and monotonicity properties of a production function. ..................................... 49
Figure 3.5. Non-monotonic production frontier with non-monotonic interval. .................................... 49
Figure 3.6. Efficient level of inter-sectoral allocation of water.............................. .............................. 54
Figure 3.7. Illustration of current and optimal water allocation in rice and CBF production. .............. 55
Figure 3.8. Determining the optimal distance of water allocation................................ ........................ 56
Figure 3.9. Inter-sector water re-allocation. .......................................................................................... 63
Figure 5.1. Frequency distribution of TE estimates .............................................................................. 93
Figure 6.1. Frequency distribution of TE estimates ............................................................................ 115
Figure 7.1. Measuring water levels in village reservoirs........................................... .......................... 128
Figure 7.2. MVP of water for CBF and rice production in VISs ........................................................ 136
Figure 8.1. Relationship between declining rice output and distance from water source.. ................. 143
Figure 9.1. Farmers‟ welfare benefits of reservoir water re-allocation. .............................................. 162
Figure 9.2. Co-management settings for reservoir-based agriculture in VISs .................................... 172
viii
viii OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA
List of Tables
Table 2.1 Basic economic characteristics of reservoir water as a commodity ..................................... 29
Table 2.2 A trichotomy of resource user regimes ................................................................................. 33
Table 4.1 Number of reservoirs used for CBF in the selected districts ................................................. 66
Table 4.2 The breakdown of the total sample ....................................................................................... 68
Table 4.3 Description of variables of the inefficiency model ................................................................ 73
Table 4.4 Description of variables of the inefficiency model ................................................................ 75
Table 5.1 Summary statistics of variables involved in the stochastic frontier model ............................ 87
Table 5.2 Initial maximum likelihood estimates (unrestricted frontier estimation) .............................. 88
Table 5.3 Performances of monotonicity and quasi-concavity ............................................................. 89
Table 5.4 Minimum distance estimation ............................................................................................... 90
Table 5.5 Final stochastic frontier model ............................................................................................. 91
Table 5.6 Inefficiency model ................................................................................................................. 92
Table 6.1 Incorporation of agricultural and CBF activities in village reservoirs .............................. 104
Table 6.2 Summary statistics of variables involved in the SFM for CBF production ......................... 110
Table 6.3 Initial maximum likelihood estimates (unrestricted frontier estimation) ............................ 112
Table 6.4 Performances of monotonicity and quasi-concavity ........................................................... 113
Table 6.5 Minimum distance estimation ............................................................................................. 113
Table 6.6 Final stochastic frontier ...................................................................................................... 114
Table 6.7 Inefficiency model ............................................................................................................... 114
Table 6.8. Mean TE of selected South and South Asian countries ...................................................... 116
Table 7.1 Inter-sectoral optimal allocation and shadow value of water ............................................. 135
Table 8.1 Sectoral average production and TE levels ........................................................................ 149
Table 8.2 Estimated technical inefficiency model for sectoral rice production .................................. 150
Table 8.3 The optimal intra-sector allocation of water ...................................................................... 151
Table 9.1 Analysis of demand shifting due to water re-llocation ........................................................ 163
Table 9.2 Consumer surpluses for rice and CBF production with water re-allocation ...................... 164
Table 10.1 Decision-making of kanna meetings in the framework of co-management strategy ......... 182
ix
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA ix
List of Abbreviations
ADB Asian Development Bank
AEO Aquaculture Extension Officers
ARPAs Agriculture Research and Production Assistants
CBF Culture-Based Fisheries
CTQS Community Transferable Quota System
DAD Department of Agrarian Development
DADCs District Agrarian Development Commissioners
DEA Data Envelopment Analysis
DMU Decision-Making Unit
ADOs Agrarian Development Officers
DS Divisional Secretary
DSDs Divisional Secretary Divisions
DvACs Divisional Agricultural Committees
EE Economic Efficiency
FAO Food and Agricultural Organisation
FGS Fast Growing Species
FOs Farmers‟ Organisations
HEFs Head-end Fields
HKARTI Hector Kobbakaduwa Agrarian Research and Training Institute
ID Irrigation Department
ITQ Individual Transferable Quota
LKR Sri Lankan Rupees
MFs Middle Fields
M/ha Metre per Hectare
x
x OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA
MCMC Markov Chain Monte Carlo
MPP Marginal Physical Product
MTE Mean Technical Efficiency
MVP Marginal Value Product
NAQDA National Aquaculture Development Authority
SGFs Small Groups of Farmers
TEFs Tail-end Fields
TE Technical Efficiency
UWA User-based Water Allocation
VISs Village Irrigation Systems
WUA Water User Association
WUAs Water User Associations
xi
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA xi
Statement of Original Authorship
The work contained in this thesis has not been previously submitted to meet
requirements for an award at this or any other higher education institution. To the
best of my knowledge and belief, the thesis contains no material previously
published or written by another person except where due reference is made.
Signature: _________________________
Date: _________________________
xii
xii OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA
Acknowledgments
This thesis is the ultimate result of the collective efforts of numerous
individuals and support received from many institutions. My primary
acknowledgment must be to my supervisory team, namely Associate Professor Clevo
Wilson, Professor Tim Robinson and Professor Sean Pascoe for their supervision,
guidance and advice. In particular, my special thanks go to Associate Professor
Wilson, for his friendly, untiring efforts and passion to make this PhD a success. I
also sincerely thank Professor Robinson for creating a motivating environment and
for the productive comments at various stages of my PhD. In short, I am indebted to
the indefatigable assistance provided by Professor Pascoe, Clevo Wilson and Dr.
Wattage who kept me on the PhD track, bringing me from the University of
Portsmouth UK to QUT. They are ultimately deserving of much more credit than I
could possibly give. I am also grateful to Professor Stan Hurn for his compassionate
support and assistance extended to me throughout my PhD.
I also wish to thank the panel members of my PhD confirmation seminar and
the final seminar for their constructive comments. In particular Dr. Mark McGovern,
panel chair of the final seminar and anonymous external examiners. All the staff
members of the School of Economics and Finance are acknowledged for their
interactive suggestions and support. Administrative support from the staff of the
QUT Business School, especially Trina Robbie, Thu Nguyen, Carol O‟Brien, Patrea
Sullivan, Lynne Eddy, Michelle Smith, Katalina Mok, Brian Cordwell, Maria Lucey,
Takae Warwick, Carly, Stone, Angela Feltcher and Lloyd Marken are highly
appreciated. The library staff at QUT, especially Janet Baker and staff at
International Student Services, too, deserves special praise. I also wish to
acknowledge the financial support of the Faculty of Business, QUT to pursue my
PhD. I also appreciate the many discussions and support I had from my friends and
colleagues, especially, Dr. Maria Leichtfried, Dr. Shyama Ratnasiri, Dr. Wasantha
Athukorala, Dr. Renuka Ganegoda, Dr. Ben Drakeford, Dr. Paolo Accadia, Dr. Jonas
Lindberg, Dr. Wasantha Welianga, Axelsson, Marcus, Dave, Tony, Marco, Prasad,
Muditha and Suresh. Dr. Jonathan Bader, Kerrie Petersen and Jeanette Berman are
acknowledged for their effort on improving the manuscript of the thesis.
xiii
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA xiii
I gratefully acknowledge the financial support of the Sri Lankan Presidential
Scholarship Fund and the University of Kelaniya, which helped me to initiate my
graduate studies at Portsmouth University, UK. All staff members and friends in the
University of Kelaniya, especially the Department of Economics, Faculty of Social
Science, are acknowledged for their administrative support and encouragement.
Especially, I wish to thank Professor K. Karunathilake and Dr. Sena Ratnayake for
their encouragement.
I am very grateful to Professor U.S. Amarasinghe who introduced me to
international academia and motivated my research interest on fisheries and natural
resources. I appreciate very much his encouragement and guidance, which has been a
driving force behind my academic career. Professor Mrs. Amarasinghe and Professor
Gunathilake Herath are also acknowledged for their direction and guidance.
The Department of Agrarian Development in Sri Lanka facilitated me to
complete the primary data collection. My sincere thanks go to Mr. Dharmasekara,
Commissioner (Human Resources) and Mr Prabhath Vitharana at the Colombo head
office and District Commissioners (Kurunagala), Mr Bandara and Mr. Amanugama
(Anuradhapura) for their prompt action coordinating the fieldwork. I am also deeply
appreciative of all the support received from the Divisional Officers. Importantly, I
am grateful to Mr. Volter Pradeep Sumith for his support and friendship and his staff,
especially Mr. Kulawansa and Vidane Mahathaya. I also wish to thank the core team
of my enumerators in Galgamuwa and other government officers for their hard work,
most often under extremely demanding circumstances. I also wish to thank some of
the officers of the National Aquaculture Development Authority (NAQDA) officers
in Kurunagala and Anuradhapura Districts for their willingness to assist my research,
despite NAQDA authorities restricting permission to obtain district level support for
data collection.
Parents and teachers are a guiding light. I am deeply grateful to my beloved
late mother for her invaluable contribution throughout my life. I also owe a debt of
gratitude to my father. I also acknowledge, my brothers and sisters for their
inspiration and encouragement. Late Professor W.M. Thilakaratne is also
respectfully acknowledged. In completing my PhD, I am fulfilling what he expected
of me.
xiv
xiv OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA
Mr. Abyaratne (Class teacher of my primary school), Mr. Wiamalsena (Class
teacher of my high School) Prof. Dharmasena and Mr. Gamini de Silva are
acknowledged for all their assistance. Among the others, late Rev. Galatha
Gnanakusala, Prof. Kamal Karunanayake, and Anders Narman, Brother- in-law, Ravi
Abyaratne are acknowledged for encouraging me to continue my higher studies.
Finally, my deepest appreciation goes to my beloved wife, Devika for her
immense contribution to my academic life, apart from being a caring mother to my
daughter, Nipuni (“Chuuti ammi”) and son, Charith (“Shan kotiya”). I very much
appreciate their understanding, patience and encouragement throughout my
postgraduate studies.
xv
OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS OF SRI LANKA xv
Dedication
This thesis is dedicated to the wisdom of our ancestors who began building the
reservoirs of Sri Lanka in the fourth century BC.
“It may appear to be such a simple matter
to raise a long bank of earth in order to
hold back a certain quantity of rain water
for bathing purposes or for watering an
adjoining rice field after the rains have
ceased, that any people living in hot
countries where the rains are only seasonal
and are followed by several almost
rainless months might be expected to be
struck by the idea of making these little
reservoirs for themselves” – Sir Henry
Parker (Ancient Ceylon, 1909).
“Without a general persuasion that the
work was one of paramount necessity and
that all would participate in its benefits” –
Richard Leslie Brohier (Ancient irrigation
works in Ceylon, 1934).
Chapter 1: Introduction 1
Chapter 1: Introduction
1.1 OVERVIEW
The global scarcity of water raises two important questions: (i) to what extent
can water resources be used efficiently, equitably and sustainably and (ii) what are
the possible ways and means by which water scarcity can be alleviated or managed
to meet multiple uses. The answers to these questions will enable decision-makers to
design appropriate water development policies as well as allocation strategies and
regimes. A critical issue of water use is ensuring that producers consider the
consequences of their decisions and how that could lead to the depletion of water
resources. A solution to this issue is to either allow the relevant institutions to
allocate resources systematically or leave the problem of resource allocation to the
market to determine the allocation based on the largest benefits (Bostock et al.,
2010).
Water management represents one of the most fundamental challenges facing
South Asia in the 21st century. Water transfers between competing sectors have
received much attention in the western world, while intra sectoral water management
is given higher priority through much of Asia: from China to Viet Nam, the
Philippines, Indonesia, and India and Sri Lanka. However, neither the analysis of
water re-allocation, nor studies on the impacts of the re-allocation have been
undertaken. The nature of water transfer processes has important implications for the
degree to which third-party effects are considered, the types of compensation
provided, and the public response to such water allocations (Kashaigili, 2002;
Meinzen-Dick, 2006). The diversity of water rights agreements between users and
the livelihood strategies adopted by the affected communities make these issues even
more complicated.
The urgency of addressing this issue is reflected in terms of decreasing per
capita availability of water resources in South and Central Asia. For example, the
availability of water dropped by almost 70% between 1950 and 1995 (ADB, 2001).
Half of Asia‟s projected population (4.2 billion) is expected to live in urban areas by
2025, resulting in severe pressure on already strained water resources in the region.
2
2 Chapter 1: Introduction
The domestic and industrial water demands in Asia are projected to grow at rates
ranging from 70 to 345 percent between 1995 and 2025 (ADB, 2001).
Decreasing water availability and accessibility as well as competing demands
for water among different sectors is also likely to lead to inter-sectoral and intra-
sectoral water allocation conflicts. These key issues have been well documented and
are being given priority in different parts of the world (Appasamy, 2004).
The main drivers responsible for increasing the demand for water use (and
hence water allocation) are growing population, expanding urbanisation, ineffective
and conflicting government policies, overlapping and often contradictory legislation
as well as policies and declining motivation for traditional collective action norms. In
addition, policies on investments, agricultural subsidies, and foreign direct
investments, which directly target water, have often contributed to the growing
demand and re-allocation of water issues and problems (Dixit, 1997; Meinzen-Dick
& Claudia, 2006). To address these issues, three main formal and informal water
allocation mechanisms have been discussed in the relevant literature: (i)
administrative re-allocations (a form of regulatory approach); (ii) market-based
transfers; and (iii) user-based water allocation based on negotiation with the user
communities (Meinzen-Dick & Jackson, 1996; Dinar et al., 1997; Dudu & Chumi,
2008).
Administrative re-allocation often occurs in large irrigation systems such as
rivers, lakes and perennial reservoirs, which are managed by central governmental
agencies. At least historically, this is the case. The main feature of this allocation
method is that water is treated inherently as public property and hence allocations are
undertaken accordingly. Under the collective negotiation mechanism, water is
allocated through decisions either between existing water users and the state or
between the old and new users themselves. Market-based water re-allocation
involves selling water directly to buyers for agricultural or non-agricultural uses.
In a market-based system, the main determinant of water allocation is its price,
which should reflect the economic value of water. At least theoretically to make such
a system work, an accurate estimate of the economic value of irrigation water is a
prerequisite (Ward & Michelsen, 2002). When the market system works, markets
allocate more water to sectors yielding the greatest returns. However, it should be
3
Chapter 1: Introduction 3
noted that market institutions that allocate irrigation water are lacking in many
countries, especially in developing countries.
In many Asian countries, water ownership, allocation and water rights are not
well established (Dennis & Arriens, 2005). This issue is important because reservoir
water is treated as a common property resource. In such situations it is important to
consider the value of water and its alternative uses so that it can enable re-allocation
decisions (Kadigi et al., 2004). Therefore, the development of a water allocation
model for reservoir water use is needed to cater to competitive demand (Dudu &
Chumi, 2008) especially, where water rights have not yet been established (Dennis &
Arriens, 2005). However, increasing scarcity and competition between users are
significant determinants (Meinzen-Dick & Bakker, 2001) in the area of water
allocation in small-scale irrigation systems. Therefore, the need to develop an
optimal water allocation model taking into consideration the full economic and social
returns of all water users is significant (Meinzen-Dick & Jackson, 1996). However,
little work has been undertaken to demonstrate the potential magnitude of the
economic gains of water users in village irrigation systems1 (VISs) in Sri Lanka.
The VISs are distributed over the entire low rainfall regions of the country
(See, Figure A1 in Appendix A). Historically, the rural lifestyle in Sri Lanka has
been based on “water culture” that has the concept of “one tank - one village”
(Siriweera, 1994). These small-scale water conservation systems are generally
referred to as VISs (Figure A2 in Appendix A) with paddy fields. According to
Department of Agrarian Development (DAD), of the 12,005 VISs recorded in the
country (DAD, 2000), 10,094 of them are in working condition (See Table A1 in
Appendix A). In the context of reservoir-based agriculture, farmer households face a
trade-off between income risks and expected profit when decisions are made in
relation to water allocation under weak institutions (Mendola, 2007) or missing
markets. The behavioural assumption of a firm is to receive maximum profit in the
production process (Varian, 1992) which is not readily applicable to VISs due to the
lack of property rights of users.
1 Minor reservoirs which have less than 80 hectares of command area, and are managed by the
respective FOs are defined as village irrigation systems (DAD, 2000).
4
4 Chapter 1: Introduction
The aim of this PhD study is to estimate the optimal allocation of water
between two competing uses by estimating the marginal value product (MVP) of
water use by each sector in VISs of Sri Lanka. Therefore, in this thesis, it is argued
that water allocation becomes more productive when the use of water changes from
„low‟ to „high‟ efficient alternatives. This thesis argues that such efficient allocation
of water is necessary to increase the total productivity of all resources (De Silva,
2003).
1.2 CONTEXT
A multitude of reservoirs have been constructed in Sri Lanka primarily to
irrigate paddy fields which are widely distributed in the low rainfall regions (See,
Figure A1 in Appendix A). The reservoir density in Sri Lanka is about 2.7 hectares
per every km2 of land area (Fernando, 1993). These reservoirs represent
approximately 74.8% of the inland water surface area of the country (NSF, 2000).
Based on the capacity and the functions, the reservoirs can be categorised into four
types: (i) large (major) reservoirs, (ii) medium sized reservoirs, (iii) minor perennial
reservoirs and (iv) minor non-perennial reservoirs. These minor non-perennial
reservoir systems are also referred to as VISs. VISs are dependent entirely on
monsoonal rainfall (See, Figure A1 in Appendix A) and they are not randomly
located, but organised in a distinctly cascading2 manner (Udawattage, 1985;
Panabokke, 2001).
As a tradition, a community meeting is held at the beginning of each cropping
season (which is a major event during the year) to discuss reservoir water
management and allocation3. During this meeting, planning of agricultural activities
takes place and collective decisions are made that cannot be changed by a single or a
few individuals. Farmers who own a plot of land in the reservoir command area with
2 A cascade can be defined as a “connected series of tanks organised within the meso-catchments of
the dry zone landscape, storing, conveying and utilising water from an ephemeral rivulet” (Panabokke
et al., 2001, p.14).
3 This is called kanna meeting. In addition to the water distribution, there also needs to be agreement
about the timing of water issues since once the tank sluice is opened all receive water. Traditionally
the most important date is when the water will be first issued since this is when land preparation will
begin. There must also be agreement about the date of first sowing, the type of rice to be sown, the
date for harvest and for draining the field. Various combinations of government, farmer and hereditary
leaders have been involved in these timing decisions. In addition, so called lucky or auspicious days
are generally preferred (Leach, 1961).
5
Chapter 1: Introduction 5
or without the membership of the Farmers Organisations (FOs)4 have access to water
use for rice farming5 (DAD, 2000). The quantity of water received by an individual
farmer (or paddy fields) depends on the time which it takes to irrigate his plot of
cultivated land. This is because water is supplied via a single unprotected canal that
traverses the block from upper fields to lower fields.
The institutional mechanisms of water allocation in village reservoirs facilitate
collective decision-making which is based on shared cultivation. A direct market
price for the amount of water used by individual farmers does not exist. Therefore, a
market mechanism in reservoir water does not necessarily work. The main factor
responsible for the market failure of reservoir water allocation is the inability to
identify the target group of reservoir water users (non-excludability). Non-exclusion
of water users leads to an overuse of water. This implies that there is less water
available for other users. The main weakness of these organisations is that it would
be less effective for inter-sectoral and intra-sectoral water allocation because they do
not include all sectors of users when they make water allocation decisions (Meinzen-
Dick, 1996; Dinar, 1997). Therefore, one of the pertinent unsolved problems in
reservoir-based agriculture in Sri Lanka is that the total volume of reservoir water is
not being allocated efficiently among multiple uses (i.e., irrigation, domestic use,
fisheries, livestock and cottage industries) and users (groups of farmers in the head-
end, middle and tail end farmers and cattle owners).
The use of water for rice farming by an individual farmer can have an impact
on the volume of water used by other farmers. If FOs decide to increase the residual
volume of water6 which is used for other purposes, it implies a reduction in the
volume of water for rice farming. The decision to reduce the volume of water made
4 FOs were established under the Agrarian Services Act (No 58 of 1979, No 4 of 1991) and the
Agrarian Development Act of 2000. FOs encourage farmers to undertake various agricultural
activities that enhance their members‟ living conditions. They include, amongst others, preparing
agricultural plans, engaging in marketing, accessing formal credit facilities and receiving government
subsidies, maintaining minor irrigation systems and intervening in farmers‟ conflicts. Therefore, FOs
play a strong role in the village reservoir-based agricultural system.
5 The common term which is used for rice farming in Sri Lanka is paddy cultivation. However, in this
thesis these two terms are used interchangeably.
6 Residual volume of water is defined as the „remaining volume of water in the reservoir after water
has been released for rice farming at time (t). This volume represents any point between Wa to W*.
(in Figure 2.3) The volume of residual water left behind in the reservoir depends on the volume of
water used for rice farming. Here, we assume that loss of water due to evaporation, seepage is
minimal.
6
6 Chapter 1: Introduction
available for rice farming will also reduce the extent of paddy land cultivated during
the irrigation season.
There is no village tradition to exclude those who use residual volumes of
water for multiple purposes in the reservoirs (Siriweera, 1994). Out of the competing
uses, culture-based fisheries (CBF) are currently being given priority due to the
commercial value of fish production. CBF are a form of aquaculture which is
practised in inland waters. In situations where CBF has been popular among farmers
as an additional source of income, the demand for „residual‟ water has increased.
Under these circumstances, farmers have to use water for rice farming more
efficiently in order to maintain a „residual‟ volume of water for other competing
demands.
In this context, the main problem associated with reservoir-based VISs are inter
and intra-sectoral water allocation. Therefore, the main issue of allocating water
across multiple uses needs to be addressed. Farmers have private property rights over
individual holdings7. However, farmers cannot transfer their water user rights to any
other productive alternatives because water allocations are made by FOs based on
collective decisions with priority given for rice farming. For this reason, the needs of
CBF are not considered by the FOs, and as a result, water is always under- allocated
for CBF. Therefore, there is a trade-off between the use of water for rice farming and
other competing uses such as CBF because the existing allocation mechanism for
residual water fails to achieve the maximum social benefits.
1.3 GENERAL PROBLEM
CBF involves non-consumptive use of irrigation water. Water use in CBF
production has tended to increase the demand for water in village reservoirs. The
interest in CBF activities by rural communities, including those who have been
fishing wild stock earlier, is growing.
The existing situation of water use and related issues in VISs, in Sri Lanka is
illustrated in Figure 1.1. Data related to Figure 1.1 were extracted from a previous
socio-economic survey of Australian Centre for International Agricultural Research
7 The plot of land, which belongs to an individual, is from a number of scattered small parcels
separated by shallow bunds (Daleus et. al., 1988 and1989; Mahendrarajah & Warr, 1991).
7
Chapter 1: Introduction 7
(ACIAR) project (FIS/2001/030) entitled „Management strategies for enhanced
fisheries production in Sri Lanka and Australian lakes and reservoirs‟. This study
was conducted out in five administrative districts (i.e., Anuradhapura, Kurunegala,
Hambantota, Monaragala, and Ratnapura) in Sri Lanka (See Figure A1in Appendix
A). These districts have a high density of village reservoirs that represent different
social and economic characteristics. In total 500 preliminary questionnaires based on
basic biological, social, economic and market related criteria were distributed among
FOs in the five districts through the coordination of DAD and Aquaculture Extension
Officers (AEO) of the National Aquaculture Development Authority (NAQDA)8.
The FOs were requested to indicate their interest in CBF production and related
existing issues for CBF production. Over 400 completed questionnaires were
retrieved and short-listed according to simple criteria such as reservoir size (<20 ha),
water retention time (6 to 11 months), accessibility, available infrastructure, market
status and willingness to participate in culture-based fisheries. Forty-seven village
reservoirs were randomly selected for the in depth survey. These farmer communities
were located within 46 village reservoirs in 29 Divisional Secretariat Divisions
(Kularatne et al., 2008). The survey results are summarised in Figure 1.1. It clearly
shows the issues related to water allocation in VISs in Sri Lanka.
The majority of landowners are not necessarily members of FOs, but use
water for rice farming. On the other hand ownership of reservoirs is a complicated
issue. There is no clear understanding among the villagers as to who owns the
reservoirs and the reservoir water. However, according to the ACIAR (2001) survey
many farmers (27%) believe that the reservoirs belong to the government. Hence, the
public notion is that all villagers can access reservoir water. Therefore, the main
problem of reservoir-based agricultural production is the inability to identify a group
of people who have well defined property rights to access water. In total 69% of the
interviewed farmers showed their willingness to start CBF activities, including those
who were fishing wild stock. However, previous experience has shown that CBF
have not been productive due to fish poaching. ACIAR survey results for example,
show that fish poaching occurs at a rate of about 59% in Sri Lanka (Jayasinghe &
Amarasinghe, 2007).
8 NAQDA was established under the Parliamentary Act No. 53 of 1998 and amendment act No. 145
of 2006 and is responsible for the development of inland fisheries and aquaculture in the country.
8
8 Chapter 1: Introduction
Source: Compiled by Author.
Figure 1.1. Identification of the research problem.
Fish poaching is assumed to be the group instability of solving water
allocation between the different users and lack of water user rights to avoid free
riders. The other main deterrent is institutional instability which, amongst other
factors, prevents farmers from investing money in CBF. The lack of property rights
among water users generates external costs among competing water users (e.g., water
user disagreements between different users).
9
Chapter 1: Introduction 9
Disagreements between rice farmers and fish farmers are common. Therefore,
two forms of inefficiencies can be identified: (i) those associated with allocating
water between the two uses (i.e., water use between rice farming and CBF) and (ii)
inefficiencies in allocating water among users for the same activity (e.g., rice
farmers). Due to the lack of proper water allocation system between rice and fish
farming, farmers realise that they do not receive the maximum net benefits from the
reservoir based agriculture. The allocation issue from the farmers‟ point of view is
shown in Figure 1.2.
Source: Compiled by Author.
Figure 1.2.The water allocation problem between competing uses.
1.4 OBJECTIVES OF THE THESIS
The main objective of this PhD research is to determine the optimal allocation
of water in VISs in Sri Lanka between rice production and CBF. This involves the
estimation of the MVP of water in each use. In addition, as a part of MVP analysis,
the technical efficiency (TE) in both rice farming and CBF production are also
estimated. The thesis aims to determine the optimal water use through inter and intra-
sectoral allocation of water. The overall objective is to maximise returns from
reservoir-based agricultural production.
The specific objectives of the study are:
10
10 Chapter 1: Introduction
1. To examine why the existing use of water is efficient for some users (both rice
farming and CBF production), while not for others.
2. To estimate the optimal level of rice and CBF production given the limited
availability of water resource.
3. To estimate the volume of water that can be saved by more efficient rice
production; and
4. To estimate the changes to the total benefits to villagers from re-allocating water
to CBF production.
1.5 RESEARCH QUESTIONS
The primary research question is whether the TE of existing allocations of
water can be improved and if so, how they can be optimally reallocated in the VISs
in Sri Lanka. This research question entails four specific questions, discussed in
detail below.
Research question 1: Why is the existing use of water among some farmers (both
rice and CBF production) more efficient for some and less for others?
This question requires estimating the TE of both rice and CBF farmers at the
current level of production. This involves examining the factors influencing TE. The
specific characteristics of input variables of the models are likely to have an impact
on TE of production. The most common characteristics found in the literature are
farmers‟ age and education level, years of experience, land ownership, farm size,
extension services, technology, infrastructure and institutions (ownership and user
rights). Lower levels of flexibility in water allocation and land ownership, as well as
poor irrigation management practices, failures in collective action and the location of
paddy fields in the command areas are also likely to have an influence on production
efficiency of reservoir-based rice production.
Past experience of CBF activities has shown that some reservoirs have been
successful with CBF while others were unsuccessful in terms of CBF output.
Institutional capacity and governance are important driving factors in the efficiency
of a firm (Estache & Kouassi, 2002). In addition to institutional factors, water
allocation issues are also likely to have an impact on efficient CBF production.
Technical factors and random factors which influence efficiency will be estimated
separately.
11
Chapter 1: Introduction 11
Research question 2. What is the optimal level of rice and CBF production given
available water resources?
Reservoir-based agricultural water allocation is assumed to be inefficient due
to market imperfections which are likely to be represented by the absence of a
market price and weak property rights for water. As mentioned earlier, there are two
prioritised economic activities using village reservoir water: (i) rice farming and (ii)
CBF production. As mentioned earlier, FOs decide water distribution as a collective
agreement at the first cropping meeting in the beginning of each cropping year.
Reservoir-based agriculture is a collective action-based economic activity. The
decision-making on water allocation and preparing the cropping calendar are a group
activity (i.e., selection of seeds). However, individuals are able to decide on the
quantity of inputs used (e.g., seeds, labour, fertiliser and pesticides) in rice farming.
There are two decision-making units (DMUs) in rice farming: (i) FOs are a reservoir
level DMU and (ii) individual farmers are decision makers at the field level. On the
other hand, CBF is entirely a group-based activity. All decisions made in CBF
production involve collective agreements. The DMU in CBF production is an
individual reservoir community. Therefore, research question two investigates the
optimal level of water allocation between rice farming and CBF production. This is
estimated at the frontier level of production and the current level of production at the
existing level of TE. This study aims to estimate the amount of water that could be
saved from one sector and be re-allocated for other sectoral needs.
Research question 3. How much water is over utilised („wasted‟) through inefficient
rice production?
Without a market price for water, it is not possible for institutions to allocate
water efficiently (Wade, 1982). The lack of physical infrastructure and the volume of
water used by head-end farmers are likely to have a negative impact on the volume
available for tail-end farmers (Chakravorty & Roumasset, 1991). The estimation of
the optimal water use in individual plots of land (farms) will enable identification of
heterogeneity in production. From the available literature, an output difference is
present in rice fields, with head-end fields (HEFs) having higher yields than tail-end
fields (TEFs) (Daleus et al., 1989; Charavorty & Roumasset, 1991). Such a „head-tail
syndrome‟ is the most common water management issue in „irrigation‟ agriculture
(Sengupta et al., 2001). This is because the quantity of water received by individual
12
12 Chapter 1: Introduction
farms decreases with the distance of the water source even though the volume of
water released from reservoirs increases (Charavorty & Roumasset, 1991).
Inefficient sectoral water use has an impact on the residual volume of water available
in the reservoirs for use in activities such as CBF.
Research question 4. Does re-allocation of water increase farmers‟ welfare and if so
what does it mean for the potential expansion of CBF?
Several objectives in food production are likely to be achieved if water is
allocated efficiently. Ensuring food security is one of the social objectives that can be
achieved by re-allocation. Re-allocation of inefficient volumes of water in rice
farming to other alternative uses, which have a higher economic value, is desirable in
increasing water productivity (Molle & Berkoff, 2009). According to Molle &
Berkoff (2009), water is often used in economically less efficient, low return (usually
agricultural) uses. Therefore, re-allocation of such water to more efficient, high
return (non-agriculture) uses is likely to increase total economic welfare. The
economic gains of re-allocating water are measured by estimating consumers‟
welfare among competing water users. Therefore, this research question is aimed at
estimating the economic benefits of water re-allocation in VISs.
1.6 DATA ANALYSIS
The translog production frontier is used in this thesis to estimate the relative
technical efficiencies. However, the monotonicity condition has been found to be
very important in the analysis of stochastic production frontier (SPF). It has been
found that some of the estimated SPF violates theoretical consistency (Sauer et al.,
2006). When the production frontier is not theoretically consistent, the efficiency
estimates of individual firms are inaccurate. Translog production frontiers in this
thesis were estimated using a simple three step procedure which was introduced by
Henningsen and Henning (2009) for imposing theoretical consistency. Therefore, the
estimated models are theoretically consistent and the predictions based on the
inefficiency models are accurate. The stochastic frontiers are estimated using
“FRONTIER 4.1” software (Coelli, 1996)9 and “R” packages “Frontier” (Coelli &
9“FRONTIER 4.1” software can be downloaded at:
http://www//uq.edu.au/economics/cepa/software/CROB2005.zip.
13
Chapter 1: Introduction 13
Henningsen, 2009). The source files used for this analysis are shown in appendix D,
E, and G).
1.7 CONTRIBUTION TO KNOWLEDGE
This thesis contributes to existing knowledge through the development of a
unique data set on rice and CBF production in Sri Lanka, and by utilising this data
set to determine an optimal allocation of water. While the application is based on
existing methods, the approach developed to use these methods (determining the
optimal allocation of a resource) is a further contribution to new knowledge. This
type of work has not been attempted before.
The field survey data were used to determine production functions relating to
water use in rice and CBF production in Sri Lanka as well as factors determining
heterogeneity in production. The latter were estimated by inefficiency models that
considered individual characteristics to estimate the productivity. These analyses had
not been undertaken previously for rice production in Sri Lanka and CBF production
in Asia.
Although there are some studies on the TE of inland aquaculture in the Asian
region (Dey et al., 2005), this research is the first efficiency estimation study in Sri
Lankan aquaculture. Moreover, the estimation of the spatial (sectoral) allocation of
rice farming in this thesis is the first of its kind.
Further, the way in which translog production frontiers are generally used has
been shown to lack theoretical consistency, which could lead to inappropriate policy-
making decisions (Sauer, et al., 2006). Estimations of stochastic frontier models in
this thesis differ from previous studies because here they follow the simple three step
procedure for imposing the monotonicity conditions as advocated by Henningsen &
Henning (2009). Therefore, the methodology applied in this thesis will help
researchers make more accurately estimates of translog production frontiers,
especially, in Asian and African countries where small scale irrigation is widely
distributed.
In addition, the justification of this study is that there is no adequate
mechanism for optimal water allocation in small-scale irrigations. Application of
MVP of water is not currently used for the estimation of optimal allocation of water
in such systems. Chakravorty & Roumasset (1991) have demonstrated a theoretical
14
14 Chapter 1: Introduction
model for intra-sectoral optimal allocation but they have not empirically estimated it.
The most important contribution of the research is that this research fills this
literature gap.
1.8 THESIS OUTLINE
In a broader sense, the thesis discusses water allocation issues in small-scale
irrigation systems known as VISs. The thesis consists of ten chapters as shown in
Figure 1.3.
The discussion in Chapter 2 provides background information on water
allocation issues in VISs in Sri Lanka. A theoretical overview of production
functions and analytical methods of optimal allocation of water is discussed in
Chapter 3. Furthermore, this chapter provides details of selection of the functional
forms and theoretical consistency of SPF. Chapter 4 discusses the research design
and the definition of models.
The remaining chapters of the thesis are organised as follows. Chapters 5 and 6
analyse existing water user efficiency in rice and CBF production respectively. The
two main water allocation issues, which are analysed in the thesis, are discussed in
Chapter 7 and Chapter 8. They relate to inter-sectoral water allocation between rice
farming and CBF production and intra-sectoral water allocation between head-end,
middle and tail-end sectors of the reservoirs in the command area. This corresponds
to research questions two and three. Chapter 9 estimates the welfare benefits of water
re-allocation. The final chapter concludes the research by presenting the policy
implications, the scope of potential applications of the study and directions for future
research.
15
Chapter 1: Introduction 15
Figure 1.3. The chapter outline of the thesis.
In each chapter the appropriate literature is also reviewed (rather than
undertaking a separate review chapter). Section 5.3 of Chapter 5 presents the
literature related to TE of rice farming. Individual characteristics of the inputs (land,
water, labour, pesticides, and fertiliser and power) as well farmers‟ age and education
level, years of experience, land ownership, farm size, extension services, technology,
institutions (ownership and user rights) are discussed as factors influencing the TE of
rice farming. In this section, how the literature relates to the individual characteristics
Research question 4 Welfare effects of water
re-allocation
Research questions
2&3
Optimal allocation of
water
Research questions 1
Technical efficiency of
water uses
CHAPTER 2
Allocation of water
resources in Sri Lanka
CHAPTER 3
Production functions
and optimal allocation
of water
CHAPTER 1
General problem
CHAPTER 5
Technical efficiency of
rice farming
CHAPTER 6
Technical efficiency of
CBF
CHAPTER: 7
Inter- sectoral
CHAPTER 8
Intra-sectoral
CHAPTER 9
Welfare effects
CHAPTER 10
Concluding remarks, policy implications and further research
Method 3
* aTNB TNB
* aTNB TNB
F R
Method 2
MVP = MVPR F
Method 1
- y x v ui i i i
CHAPTER 4
Data collection and
model definitions
16
16 Chapter 1: Introduction
of water use is considered. Therefore, the relevant literature on land-based
aquaculture is examined in Section 6.3 of Chapter 6, with respect to CBF production
in Sri Lanka. Previous research in Asia and Africa are reviewed due to similarities of
aquaculture systems and are compared with the current study. Efficiency objectives,
rather than the equity aspects, are observed from the review of the literature in
Section 7.4 in Chapter 7. The, literature, especially related to re-allocating water
from low to higher valued uses (Molden et el., 2010) through the equi-marginal
principle (Gopalakrishnan, 1967) is also discussed. The theoretical and empirical
evidence which relate to intra-sectoral water allocation are reviewed in Chapter 8,
while Chapter 9 presents the literature related to the welfare effects of water re-
allocation.
Chapter 2: Allocation of water resources in Sri Lanka 17
Chapter 2: Allocation of water resources in
Sri Lanka
2.1 INTRODUCTION
This chapter discusses reservoir based agriculture and issues relating to VISs in
Sri Lanka. Firstly, water resources allocation in irrigated agriculture in Sri Lanka will
be explained. The focus is then narrowed down to issues relating to water allocation
between competing demands and technical limitations of capacity of VISs. Reservoir
water can also be considered as a commodity and common property resource.
Finally, issues of non-market solutions and property rights are discussed.
2.2 WATER RESOURCE IN SRI LANKA
Over a span of two thousand years, a multitude of reservoirs have been
constructed in Sri Lanka primarily to irrigate paddy fields. Construction of these
reservoirs has enabled rainfall to be widely distributed in low rainfall regions (See,
Figure A1 in Appendix A). Reservoir density in Sri Lanka is about 2.7 ha per km2 of
land area of the country (Fernando, 1993). These reservoirs represent 74.8% of the
inland water surface area of the country (NSF, 2000). There are four types of
reservoirs, categorised on their capacity and functions: (i) Large reservoirs are used
only for hydroelectric power generation. Six large reservoirs were constructed during
the last 30 years for hydro power generation under the Mahaweli river water
diversion scheme covering 21,747 ha of land area. (ii) 72 ancient medium sized
reservoirs covering 70,850 ha of land area provide water for irrigation and power
generation. (iii) 160 minor perennial reservoirs, covering 17,001 ha of land area, do
not directly discharge water for cultivation, however they convey irrigation water
(Costa & De Silva, 1995). (iv) Approximately 10,000 operational VISs covering
23.1% (39,271 ha) of the total surface of land water area have been designed for
multiple uses ( De Silva, 1988; Fernando, 1993).
VISs in Sri Lanka depend entirely on direct rainfall and runoff water from their
own catchment areas. Therefore, they are positioned where distinct cascades exist
18
18 Chapter 2: Allocation of water resources in Sri Lanka
either in well-defined small cascades or in meso-catchment basins (Udawattage,
1985Panabokke, 2001). A cascade can be defined as a “connected series of tanks10
organised within the meso-catchments of the dry zone landscape, storing, conveying
and utilising water from an ephemeral rivulet” (Panabokke et al., 2001, p.14).
Drainage from paddy fields in the upper parts of the cascade flows into a
downstream reservoir for re-use. (See Figure B1. in Appendix B).
The distribution of man-made reservoirs in the country is also based on
monsoonal patterns (See Figure A1 in Appendix A). Reservoir density is highest in
districts located in low rainfall regions. The low rainfall regions (dry zone) of Sri
Lanka are located within the lowest peneplain of the island and covers approximately
66% of total land area. This area accounts for 33% of the country‟s population.
Current irrigation withdrawals in these districts account for over 75% of reservoir
water (Samad, 2005).
Reservoirs and canals are entirely the result of human intervention to ensure
inland water security in Sri Lanka. As in most other Asian countries, irrigation
agriculture is widespread and is an important common economic activity. In low
rainfall regions, reservoir water is the main source of water for reservoir-based rice
farming, which accounts for approximately 90% of the total water used in about 0.6
million ha of cultivated land (DAD, 2000).
Ancient inscriptions indicate that reservoir water has long been considered a
measure of wealth of people in Sri Lanka. The economy, initially based upon
subsistence agriculture, became a dual economy with the introduction of plantation
agriculture in the 1830s. The focus of the post independence (since 1948) domestic
agricultural policy has been national self-sufficiency in rice production (Pain, 1986).
Successive governments have promoted the expansion of the paddy sector through
new irrigation settlement schemes, fertiliser and pesticide subsidies, investment in
research and extension, and other support services. The subsistence agricultural
sector has undergone a dramatic technological transformation after the introduction
of green revolution technology in Sri Lanka in the 1960s (Pain, 1986).
Sri Lanka‟s economy has experienced significant structural changes since
1977, after adopting market-oriented open economic policies. These changes in the
10
In the early literature reservoirs were referred to „Tank”.
Chapter 2: Allocation of water resources in Sri Lanka 19
economy have had an influence on domestic agriculture as well. For example, the
agricultural workforce declined from 52% in 1977 to 36% in 2000 (Karunagoda,
2004). Furthermore, despite the area expansion of cultivatable land under the
Mahaweli river diversion scheme, the farm size of small-scale agriculture decreased
from 1.97 ac (0.8 ha) in 1982 to 1.2 ac (0.5 ha) in 2002. This is largely attributed to
land fragmentation due to population increase. Furthermore, wage rates in the
agricultural sector have remained low with relatively high levels of rural
unemployment (Karunagoda, 2004).
Due to a number of reasons, rice production under VISs has been declining
since 1977. This could be partly attributed to cheaper imports resulting in village
reservoir-based producers receiving low prices for their paddy output. Although rice
production using village reservoir water is less profitable to village farmers, they
continue to engage in rice cultivation in some form. Furthermore, there are other
socio-economic benefits arising from village reservoirs such as provision of water for
bathing, household washing, livestock rearing and fisheries.
Most of these reservoirs have increasingly been used for fish production in
recent times. Traditionally, fish production from inland reservoirs was based on
indigenous species whereas the extent of commercial-scale inland fisheries was
limited until a few decades ago. However, with the introduction of government
assistance (on a small-scale) for the development of inland fisheries in the 1950s,
commercial scale fish production has increased. The pioneering work of Mendis
(1965), reinvestigated by Rosenthal (1979) and Oglesby (1981), recommended the
development of CBF in village reservoirs in the late 1970s and early 1980s. CBF is
essentially a farming practice conducted in small water bodies (generally less than
100 ha), which cannot support a subsistence fishery due to inadequate „natural
recruitment‟ (Amarasinghe & Nauyen, 2009 ). Since the introduction of CBF,
attention has been focussed on the development of CBF in village reservoirs with
successive governments supporting this approach to increase fish production in the
country11
. In recent years, inland fish production has been approximately 14% of the
total fish production. (NAQDA, 2008). As the rural population grows, the demand
for fresh water fish especially among the poorer households has increased (Dey &
11
Annual inland fish production in 2008 was approximately 44,490 (Mt) compared to 20,266 (Mt) in
1980.
20
20 Chapter 2: Allocation of water resources in Sri Lanka
Garcia, 2008). Increased CBF production (De Silva, 2003) could maximise reservoir-
based community welfare. However, the economic efficiency (EE) of community-
based water allocation in rural agricultural systems has changed with agricultural
modernisation (Mahendrarajah & Warr, 1991). Therefore, competition for limited
inland water resources essentially requires an efficient water allocation system to
sustain competing water demands (Dugan et al., 2006), especially among the
multiple users (Meinzen-Dick & Bakker, 2001; Dennis & Arriens, 2005).
2.3 WATER RESOURCES MANAGEMENT AND ALLOCATION
To address these issues, three main formal and informal water allocation
mechanisms have been discussed in the relevant literature: (i) administrative re-
allocations (a form of regulatory approach), (ii) market-based transfers, (iii) user-
based water allocation based on negotiating with the user communities (Meinzen-
Dick & Jackson, 1996; Dinar, et al., 1997; Dudu & Chumi, 2008).
Administrative re-allocation often occurs in large irrigation systems such as
rivers, lakes and perennial reservoirs, managed by central governmental agencies. At
least historically, this is the case. The main feature of this allocation method is that
water is treated inherently as public property and hence allocations are undertaken
accordingly. Market-based water re-allocation involves selling water directly to
buyers for agriculture or non-agricultural uses. Water markets, giving compensation
to those who receive less water, presuppose strong recognition of private water
rights. Under the collective negotiation mechanism, water is allocated according to
decisions made either between existing water users and the state or between the old
and new users themselves. In a market-based system, the main determinant of water
allocation is its price, which should reflect the economic value of water.
Theoretically, to make such a system work, an accurate estimate of the economic
value of irrigation water is a prerequisite (Ward & Michelsen, 2002). When the
market system works, markets allocate more water to sectors yielding the greatest
returns. However, it should be noted that market institutions that allocate irrigation
water are lacking in many countries, especially in developing countries. Little work
has been undertaken to demonstrate the potential magnitude of the economic gains of
water users in VISs in Sri Lanka.
Chapter 2: Allocation of water resources in Sri Lanka 21
The management of water VISs has a long history. “Water law” dates back to
1856 under the British colonial administration. The irrigation Ordinance (No.32) was
the first enacted to both legalise customary irrigation practices and legislate the
conditions of water extraction, especially for rice farming (Samad, 2005). The
Irrigation Department (ID) and the Mahaweli Authority manage the major irrigation
systems having over 400 hectares of command area. The Provincial Councils
administer the medium-scale irrigation schemes, with 80-400 hectares of command
area. Village reservoirs which have less than 80 hectares of command area are
managed by their respective FOs with the technical guidance of Provincial Irrigation
Departments. Members of FOs have well defined property rights in relation to the
use of reservoir water for agriculture (especially for rice farming). In practice,
however, user rights are not clearly defined in relation to the use of reservoir water
for CBF activities.
The Agrarian Development Officer of DAD coordinates FOs in each village
with the help of Agrarian Research and Production Assistants (ARPAs). ARPAs are
the village level government technical officers responsible for agricultural research
and production (See Figure B2 in Annex B). Monthly meetings of Divisional
Agriculture Committees (DvAC) presided by the Divisional Secretary12
(DS) are
held to discuss the issues relating to water management in village reservoirs.
The typical village paddy field is a single block of land lying immediately
below the reservoir bund (See Figure B3 in Appendix B). Distributions of water from
one plot to the other take place through unprotected canals. Water is supplied via a
single unprotected canal that traverses the block from upper fields to lower fields.
Once the sluice gate of the reservoir is opened by one of the farmers (who is
appointed by the FO), the exact quantity of water the first farmer receives, is not
known to him. The quantity of water received by an individual farmer depends on the
time it takes to irrigate his plot of cultivated land. Under the share cropping system
(Bethma13
), the FO decides the extent of the total land area to be cultivated and
12
Divisional level government administrative officer.
13
This is an ancient practice designed to minimise conveyance losses and to conserve the available
irrigation water (Bandara, 1999). A suitable sized portion of the field is selected and the rest is
abandoned. The selected portion is divided into an equal number of shares. Therefore, the person
whose land is selected does not get a larger allotment than the others. Each bethma arrangement is
binding only for one crop, and when it has been removed, reverts to their original position. Quite
22
22 Chapter 2: Allocation of water resources in Sri Lanka
irrigated in a particular irrigation season. This can involve reducing the cropping
intensity.
The Kanna meeting (a community meeting held at the beginning of the
cropping season) of the FOs discusses reservoir water management in the village14
.
At this meeting, planning of agricultural activities takes place and collective
decisions are made that cannot be changed by individuals until the end of the
cultivation season, unless there are special circumstances. ADOs, village level
technical officers of ADA and NAQDA also attend these meetings. Therefore, there
are a number of levels involved in making decisions over the management of village
reservoirs (See Figure B2 in Appendix B). There are also legal provisions for various
rural development activities through FOs, under the Agrarian Development Act No
46 of 2000, which include provisions for the development of CBF in village
reservoirs. AEO of NAQDA is also invited to attend the monthly meetings of DvAC.
Existing practices of CBF activities in most instances are performed through
small groups of farmers (SGFs) in FOs. Consequently, strategies for stocking,
protection and harvesting the stock are decided collectively. The members arrive at
agreements on sharing CBF profits between fish farmers and agricultural farmers.
Levies paid by SGFs (generally about 5% of profit) to FOs are normally used for
village rehabilitation work.
There are also monthly meetings held by Divisional Secretaries, with the aqua-
culturists or regional (AEO) of the Ministry of Fisheries and Aquatic Resources and
other heads of relevant departments and organisations pertaining to agricultural
development in the district. This committee is called the “District Agriculture
Committee” and is presided by the DS. In reality, CBF is still not a high priority area
for the DvACs, in spite of the legal provision in the Agrarian Development Act No
46 of 2000 for CBF development. Therefore, one of the unsolved problems in
often, the paddy tract selected for bethma lies close to the reservoir bund or irrigation ditch, therefore
helping).
14
In addition to water distribution, there needs to be an agreement on the timing of water issues as
once the tank sluice is opened, all receive water. Traditionally the most important date was when the
water would first be issued as this was when land preparation began. There must also be agreement on
the date of first sowing, type of rice to be sown, and the date for harvest and draining the field.
Various combinations of government, farmer and hereditary leaders have been involved in these
timing decisions. In addition, so called lucky or auspicious days are generally preferred (Leach, 1961).
Chapter 2: Allocation of water resources in Sri Lanka 23
reservoir-based agriculture in Sri Lanka is that water is not being allocated efficiently
among multiple uses (i.e., irrigation, domestic, fisheries, livestock and cottage
industries) and users (groups of farmers in the head-end, middle and tail end,
fishermen and cattle owners).
2.3.1 USE OF WATER RESOURCES IN RESERVOIR-BASED AGRICULTURE AND
RELATED ISSUES
The greatest challenge in irrigated agriculture is to use inputs efficiently (e.g.,
water) in order to „grow more crops with less water‟ (Khan et al., 2006). User-based
water allocation systems (UWA) are currently practiced by FOs in VISs. A water use
association (WUA) is an administrative body formed by the beneficiaries of the
water resources (i.e., FOs in Sri Lanka). This community-based organisation may or
may not be represented by government officials and institutions in the management
board. Beneficiaries can be represented directly or indirectly and their aim is to
maximise profits. In some cases, however, the beneficiaries can be a non-profit
organisation (Dudu & Chumi, 2008). Some case studies show that water user
associations (WUAs) operate efficiently (Schoengold & Zilberman, 2005). However,
Samad (2005) states that user-based water allocation systems are inefficient in Sri
Lanka due to two major reasons: (i) failures in state-sponsored field level institutions
(i.e., FOs); and (ii) failed implementations of irrigation fees due to high transaction
costs (costs incurred due to negotiation, enforcement, and exchange of property
rights). Therefore, village reservoir water allocation is likely to be inefficient in Sri
Lanka due to the absence of a water market and well defined water user rights.
The total volume of reservoir water can be categorised into two sub
categories based on the type of water uses. They are (i) water used for rice farming
and (ii) the „residual‟ volume of reservoir water used for other competing water
demands. The water level in the VISs is subject to evaporation losses. By the end of
August each year, most of these water bodies dry-up completely (Mahendrarajah &
Warr, 1991). The water received from monsoonal rain (during the high rainfall
season), is allocated mainly for rice farming. The „residual‟ volume of reservoir
water is used for other purposes. Among them, CBF has been given priority due to
the commercial value of fish production. Water use for rice farming by an individual
farmer can have an impact on the volume of water used by other farmers. Illegal use
of reservoir water for rice farming is not possible as water is allocated and managed
24
24 Chapter 2: Allocation of water resources in Sri Lanka
under a common agreement, by FOs. Due to the inefficiency of field level
institutions (Samad, 2005) and absence of water pricing, there is no efficient
allocation of water from the „head-end‟, middle-end and „tail-end‟ fields, despite
observed output heterogeneity in paddy fields of the command area in VISs (Daleus
et al., 1988, 1989).
As mentioned earlier the main commercial use of „residual‟ reservoir water is
for CBF. Water used for CBF increases „congestion‟ of other water uses (i.e.,
domestic use, wild stock fishing, animal husbandry and other domestic cottage
industries such as brick-making). The „residual‟ volume of water has characteristics
of being rival in consumption and non-excludable. Therefore, reservoir water can be
considered an impure public good (Bailey, 1995). Non-exclusion of water users leads
to the overuse of water where some stakeholders receive more water than others.
One important aspect in conservation and management of reservoir water use is
obtaining a better understanding of the needs of various stakeholders (Heltberg,
2000).
The main traditional water allocation objectives in Sri Lanka (Mahendrarajah
& Warr, 1991) were avoiding conflicts among water users and EE. Reservoir water
has high productivity due to the multiple uses among various agricultural and non-
agricultural activities (Phengphaengsy & Okudaira, 2008). However, farmers
measure the value of irrigation water by only taking into account the value of total
output of main agricultural activities (i.e., rice harvest). This measure undervalues
reservoir water, since it does not account for the user value of water for fisheries,
domestic use, animal husbandry and other domestic small-scale industries such as
brick-making.
The current user-based water allocation by FOs is mainly aimed at allocating
water for rice farming. This is because paddy cultivation, largely, provides rural food
security. Furthermore, alternative commercial uses of water are not well developed in
these areas. Water for CBF was not prioritised until the 1990s (Amarasinghe &
Nguyen, 2009) even though CBF was introduced on a large-scale into village
reservoirs in the mid1980s. Since then, CBF has become popular among farmers as
an alternative income generating economic activity (Amarasinghe & Nguyen, 2009).
The marginal productivity of reservoir water used in agriculture has increased with
the technological transformation of agriculture (i.e., green revolution) since the
Chapter 2: Allocation of water resources in Sri Lanka 25
1960s (Mahendrarajah & Warr, 1991). Furthermore, the marginal productivity of
residual water of reservoirs for CBF has also increased with the introduction of stock
enhancement strategies in CBF in the mid 1980s (Amarasinghe & Nguyen, 2009).
Therefore, the allocation of reservoir water between users has become a crucial issue
in maximising reservoir-based agricultural and CBF production.
Figure 2.1. Method of water allocation in village reservoirs. Adapted from “Missing
markets for storage and the potential economic cost of expanding the spatial
scope of water trade by D. Brennan, 2008, Agricultural and Resource
Economics, 52(4)p. 473.
Figure 2.1, shows that the water available in the irrigation season (t) entirely
depends on the quantity of inflow from rain. This is because the residual volume of
water carried forward from the previous irrigation season (t-1) is zero. At present, the
available water is used for rice farming and the residual volume of water (dead
storage) is used for other competing water demands.
2.3.2 VOLUME OF WATER USED FOR COMPETING WATER DEMANDS
As mentioned earlier, the residual reservoir water is used for multiple
purposes15
. However, well-defined groups for the use of residual water in reservoirs
do not exist. Therefore, the residual water in the reservoirs can be considered an open
15
During the low rainfall season, village reservoirs are the only source of water for domestic (home)
use and for farm animals. Alternatively, villages have to use either wells or nearby rivers and streams.
Most of the reservoirs dry-up for a few months, unless they receive unexpected rain. Therefore,
villagers have to face a few months of severe water shortages.
26
26 Chapter 2: Allocation of water resources in Sri Lanka
access resource. Access to residual water is not restricted and, therefore, a non-
excludability feature is associated with this resource. Residual water is used as a
common pool resource so there is no special or quantitative rivalry. However, there
are externalities that affect the use of the water (i.e., water is polluted after the fish
harvest, human bathing and washing clothes). In the case of irrigation water supply,
water for individual paddy fields (farms) is first channelled to head-end farmers and
then to the tail-end farmers of the command area. The volume of water used for rice
farming by one farmer in the head-field (W1) may reduce the volume of water
available for use by another farmer in the middle-field (W2) or tail-end field (W3)
(See Figure 2.2)16
.
In the case of the reservoir-based irrigation systems in Sri Lanka a direct
market price for the amount of water used by individual farmers does not exist.
Farmers who own a plot of land in the reservoir command area, with or without the
membership of the FO, have a right to use water for agriculture (DAD, 2000). The
only cost to individual farmers is paying the head farmer of each reservoir for his
services of controlling water between the fields at the end of the particular season.
These payments are usually non-monetary in nature (e.g., based on an agreed
quantity of rice). However such transactions are not always properly implemented.
Source: Compiled by Author.
Figure 2.2. Semantic diagram of intra-sectoral water allocation.
16
According to estimates of the Department of Agrarian Development of Sri Lanka, water
requirement for one hectare of paddy cultivation under a village reservoir system is 0.9 metres during
the major (Maha) cultivation season, assuming an expected rainfall of 22 inches. Similarly, the
estimated volume of water required for the minor (yala) cultivation season is 1.3 metres (DAD, 2009).
Chapter 2: Allocation of water resources in Sri Lanka 27
During periods of water shortages in the area, other villagers may also use the
residual volume of water for domestic purposes. There is no village tradition to
exclude domestic water users from these reservoirs in Sri Lanka (Siriweera, 1994).
2.3.3 TECHNICAL LIMITATION OF WATER ALLOCATION
The extent of paddy fields to be cultivated during the irrigation season (t) is
decided by the respective FO based on the volume of water (Wa to W*) available for
rice farming and CBF. Dead storage (W0 to Wa) is only available for other competing
uses including CBF as shown in Figure 2.3.
The vertical axis of Figure 2.3 represents the total volume of water received
from monsoonal rain. This is indicated by (W0 to W*) during the inflow season, t. W0
is the reservoir bed and Wa is the water level at the sluice gate of the reservoir (the
point at which the water reservoir water can be released to paddy fields). This is
known as ‟dead storage‟ in the reservoir. The maximum capacity of the reservoirs is
indicated by W*
in Figure 2.3. The total volume of water, which is technically
available for use, is Wa to W*. The horizontal axis (L0 to L
*) represents the land
command area to be cultivated in the reservoir.
Source: Compiled by Author.
Figure 2.3. Graphical presentation of land and water relationship.
28
28 Chapter 2: Allocation of water resources in Sri Lanka
Assuming that the volume of water in the reservoir at full supply level is W*,
farmers may decide to cultivate the full extent of the land (up to L*) in the command
area. The farmers presume that, if the reservoir is at full capacity then it has enough
water to supply the entire extent of cultivatable paddy land. If the volume of water is
below the full supply level (between Wa and W*), the water allocation decision
would be to cultivate paddy land less than L*. The second option is called „share
cropping‟ or locally termed the “bethma” system (Bandara, 1999). A change in the
extent of land cultivated during the irrigation season (t) is dependent on the potential
volume of water during the inflow season (t).
There are two technical factors that influence the level of water use. They are
(i) limits placed on the maximum level of reservoir water stored (W*), and (ii) limits
on the total cultivatable land (L*). This is due to reservoirs being located as cascade
systems. Therefore, these two important technical limitations decide the maximum
water available for use and the extent of paddy land that can be cultivated during the
irrigation season (t). If FOs decide to increase the residual volume of water, water for
rice farming is reduced. The decision to reduce the volume of water made available
for rice farming also reduces the extent of paddy land cultivated during the irrigation
season. Therefore, land is a variable input relating to the volume of water available.
In situations where CBF has been popular among farmers as an additional
source of income, the demand for „residual‟ water has increased. Under these
circumstances, farmers have to use water for rice farming more efficiently in order to
maintain a „residual‟ volume of water for other competing demands. Therefore, there
is a trade-off between the use of water for rice farming and other competing uses
such as CBF.
As members of FOs, farmers have equal rights to use water for agriculture
(DAD, 2000). Farmers who have land ownership in the command area are entitled to
use water for rice farming even without membership of FOs. As a result, farmers
cannot be excluded from using water for agricultural purposes. However, the volume
of water used for individual farms cannot be decided by the farmers themselves
unless there is no established FO in the village. Individual share of water depends on
the extent of land that will be cultivated during the irrigation season (t). With the
development of CBF as a commercial practice, water inadequacy has become an
issue especially when there are other competing demands.
Chapter 2: Allocation of water resources in Sri Lanka 29
2.4 RESERVOIR WATER AS A COMMODITY
The total volume of reservoir water can be categorised into three groups based
on the requirements of water users. They are the total volume of water, volume of
water used for rice farming and the volume of water used for competing water
demands. As a commodity, they have different characteristics.
Table 2.1
Basic economic characteristics of reservoir water as a commodity
Source: Hanley et al., 1997.
The basic characteristics of reservoir water-market relationships are
excludability and rivalry (subtractability) in reservoir water uses. This is shown in
Table 2.1. Excludability implies the possibility of excluding specific water users
from using the water. Subtractability refers to water being used by one user leads to
subtractions from other users. (Hanley et al., 1997; Grafton et al., 2000).
Nevertheless, use of the total volume of reservoir water as a whole is non-excludable,
however one farmer‟s consumption of the reservoir water does affect other farmers‟
consumption.
Therefore, reservoir water cannot be considered a public good. The situation
having either non-rivalry or non-excludability, or substantial elements of both are
identified as impure public goods in the literature (Bailey, 1995).
2.4.1 MISSING MARKETS FOR RESERVOIR WATER ALLOCATION
Inefficient production processes fail to maximise profit. Inefficient allocations
of resources emerge with the malfunctioning of the market mechanism (Li & Ng,
1995). Markets can provide a flexible mechanism to allocate water. The main role of
30
30 Chapter 2: Allocation of water resources in Sri Lanka
market prices is to facilitate efficient allocation of scarce water resources among
competing uses and users (Johansson et al., 2002). It must be noted that the market
does not necessarily work in such a manner in relation to community-managed
systems. However, this conclusion depends on restrictive assumptions as observed
by various researchers based on many different characteristics of the water market
(Stiglitz, 2002).
Market failure takes place when the market mechanism is incapable of
allocating scarce resources to generate the greatest social welfare. When prices do
not communicate society‟s desires and constraints accurately, markets fail for many
environmental goods (e.g., water) and services (Hanley et al., 1997). Furthermore,
inefficient outcomes are generated whenever firms or individuals allocate resources
for consumption or production having an external effect other than price (Grafton et
al., 2000). Johansson et al. (2002) identified the public good nature of water
including incomplete information, externalities, implementation costs of large
irrigation projects, returns to scale, and equity concerns are among the most
important reasons for market failures. Furthermore, Hanley et al. (1997) examined
five interrelated cases for market failure: externalities, non-convexities, non-
exclusion, non-rival consumption and asymmetric information.
Incomplete information about the availability, accessibility and the security of
water is one of the problems for missing irrigation water markets. The most common
reasons of asymmetric information are due to the nature of water supply. Water users
do not have exact information about the quality, quantity and timing of supply, as
supply of water is determined by climatic conditions. However, in the case of
reservoirs in Sri Lanka, decisions made by FOs to allocate water for rice farming are
based on their experience, rather than having accurate information about the quantity
and timing of water received (which is determined by monsoonal rains).
Another reason that underlines the market failure for irrigation water is
externalities (Hanley et al., 1997) generated by the public good nature of reservoir
water. Provision of irrigation water may generate both negative and positive
externalities. Furthermore, high implementation costs of large irrigation projects can
cause market failure (Hanley et al., 1997). In most cases, building irrigation
infrastructure is not economically feasible for the private sector. This is also
applicable to reservoir water in Sri Lanka.
Chapter 2: Allocation of water resources in Sri Lanka 31
The main factor responsible for market failure of reservoir water allocation is
the inability to identify the target group of reservoir water users (non-excludability).
Non-exclusion of water users leads to overuse of water. This implies less water for
other users (Hanley et al., 1997). Property rights can be defined as the socially
accepted rights of farmers or groups of farmers to exploit a harvest for their benefit
with at least a partial right to exclude other individuals in agriculture (Heltberg,
2002). Resource allocation is often complicated due to the divergence between
private and social efficiency. Therefore, the definition of water property rights and
the enforcement rules in the context of developing country systems (Ferguson, 1992)
is not clear.
Furthermore, failures in establishing and or enforcing property rights is one of
the four main underlying factors (market, government and population growth failure)
that lead to environmental degradation. Therefore, conservation and management
require better understanding of property rights (Heltberg, 2002). Buyers and sellers
can freely undertake market transactions when property rights are well-defined
(Hanley et al., 1997). Specialisation and accumulation of capital are vital
components of economic growth. Strong property rights are a fundamental
requirement for specialisation and capital accumulation (Arnason, 2005).
Specialisation increases trading of goods. Trade in turn requires property rights.
Therefore, trade is dependent on the transferability of property rights.
2.5 RESERVOIR WATER AS A COMMON PROPERTY ISSUE OF NON-
MARKET SOLUTION
In general, a common property resource refers to where exclusion is difficult
and some degree of rivalry exists (Berkes, 1989). However, there is lack of clarity,
between open access and common property resources. „Tragedy of the Commons‟
(Hardin, 1968) actually refers to open access resources. Therefore, Stevenson (1991)
referred to this as the Tragedy of Open Access. Resources used by multiple users
without rules governing their use will be overexploited (Costanza et al., 1997).
Hardin's (1968) work symbolises the expected degradation of the environment when
many individuals use scarce resources, Today this perception of common property is
recognised as having no basis in reality (Hanna, 1990), as social scientists have
observed that not all common property resources are subject to such a 'tragedy' and
they are not overexploited.
32
32 Chapter 2: Allocation of water resources in Sri Lanka
The above observation suggests why a form of asset ownership of particular
significance to the rural poor is communal. Hardin's (1968) observation is on the fate
of common property resources. Common property resources erode because people
free-ride off others. However, "tragedy of the commons" is not necessarily a suitable
terminology for geographically localised common-property resources, such as
irrigation water and local forests, threshing grounds, grazing fields, and inland and
small coastal fisheries. It is now known that typically, the local commons are not
open for use by all. They are not "open access" resources; in most cases they are
open only to those having customary rights, through kinship ties, and community
membership. It has been known for some time that the users themselves (Dasgupta,
1998) can, in principle, manage the local commons efficiently.
From a theoretical point of view, optimal use of state property is possible, but
in practice, weak enforcement of the government rules and the provision of subsidies
(without defined property rights) may lead to a de facto open access situation. Many
attempts at state control of natural resource use have failed. Therefore, the only
remaining management regime is known as common property or collective
management. Common property management has the potential to optimise the use of
natural resources. Unlike private property, the externalities are internalised if all the
individuals potentially affected by the resource use are members of the management
group. Therefore, in principle, resource use can be controlled by collective use to
ensure that its use continues until marginal private benefits equals marginal social
costs. Common property resources can also be expected to have limited controls to
entry. Indeed, Stevenson (1991) goes further and states that common property
resources must also have limitations on how much each user of the resource can
extract. Consequently, Stevenson summarises that private, common and open access
resources use regimes based on group and extraction limitations.
However, common property does share some aspects of open access in certain
situations (e.g., fishery), whereby users impose negative externalities on one another.
Common property through limitations on entry and extraction by users constrains the
negative externality to a non destructive level. In practice, the restriction on
extraction may be more varied and complex than simple physical limits on the
quantity extracted (See Table 2.2).
Chapter 2: Allocation of water resources in Sri Lanka 33
Table 2.2
A trichotomy of resource user regimes
Property rights Private property Common
property
Open access
Limited user Unlimited user
Group limitation
Extraction
limitation
One person
individual
decision
Members only
limited by rules
Members only
Unlimited
Open to
anyone
Unlimited
Source: Stevenson, 1991.
According to Stevenson (1991) common property management must have
seven characteristics:
1. The resource unit has boundaries which are well defined by physical,
biological and social parameters.
2. There is a well delineated group of users, who are distinct from persons
excluded from resource use.
3. Multiple users participate in resource extraction (i.e., there are at least two in
the group).
4. Explicit or implicit rules exist among users regarding their resource rights and
their duties to one another about resource extraction17.
5. Users share joint, non-exclusive entitlement to the situ or fugitive resource
prior to its capture or use.
6. Users compete for the resource, and thereby impose negative externalities on
one another.
7. A well delineated group of rights holders exists, which may or may not
coincide with the group of users.
17
There are also rules about the distribution of fish in VIS. A farmer can fish anytime, except for two
periods at the end of the irrigation season, when the water level is at its lowest (about 2-3 feet), and so
fishing is very easy. At this point, the Chief Farmer (Vel vidane) erects a pole in the reservoir showing
that individual fishing must cease. There are strong taboos against breaking this rule and it is
supported by government regulations. As soon as farmers agree on a date for fishing, it is set by the
vel vidane. Fishing is reserved primarily for landholders, but they can introduce their friends and
relatives as assistants. The fishing party of about 30 drive the fish towards the bund in the corner
where they can be scooped out into baskets. This lasts for three days. At the end of each day, the fish
are counted. The fishermen can keep two thirds and one third goes into the land pile, which is shared
according to land holdings (Leach, 1961). This system cannot be practiced with CBF production due
to the CBF farming system.
34
34 Chapter 2: Allocation of water resources in Sri Lanka
However, as suggested by Bromley (1991), the management group (the
owners) has rights to exclude non-members, and non-members have duties to abide
by exclusion of the common property resources. Individual members of the
management group (the co-owners) have rights and duties with respect to using and
maintaining resources. According to Bromley (1991), there is one main difference
between private and common property. Private property has only one owner whereas
common property has more than one owner. This is the system that prevails in VISs
used for CBF. In VISs, rights of access to water include criteria based on land, crop
share and membership of village (Gardiner et al., 1994). It must be mentioned here
that since access to land is relatively difficult, irrigation facilities will experience less
conflict with outsiders than fisheries. This has been a common problem in VISs
(Leach, 1961). In many common property regimes, free-riding can become an issue.
The works of Wade (1982), Ostrom (1990) and Barland & Platteau (1996)
analysed the effect on local communities managing and governing common pool
resources. According to Wade (1987), effective rules of restraint on access and use
are unlikely to last when there are many users. Most of Ostrom's principles focus on
local institutions, or on relationships within the local context. Baland and Platteau
(1996) suggested that small user groups, a location close to the resource,
homogeneity among group members, effective enforcement mechanisms, and past
experiences of cooperation are some of the analytical factors necessary to achieve
cooperation. In addition, they also highlight the importance of external aid and strong
leadership. However, they all conclude that members of small local groups can
design institutional arrangements to help manage the sustainability of resources.
Agricultural farmers maximise their private benefits ignoring other competing
water uses in the reservoirs. As a result, existing markets for residual water fail to
achieve maximum social benefits. Understanding the value of water and its
competing demands is an essential condition to make decisions on water
management and allocation. In many Asian countries, water ownership, allocation
and water rights are not major concerns (Dennis & Arriens, 2005). This situation is
rather crucial where people use deficit (residual) irrigation water as a common
property resource with multiple uses. Competition for limited inland water requires
developing a water allocation model to sustain competitive demand where water
rights have not yet been established (Dennis & Arriens, 2005; Dugan et al., 2006).
Chapter 2: Allocation of water resources in Sri Lanka 35
Many developing countries have begun to decentralise policies and decision-
making which included development activities, public services and the environment
(Agarawal & Ostrom, 2001). On the other hand, central government management of
water and aquatic resources (e.g., fisheries) often lacks the capacity to enforce
property rights and regulate resource use (Ahmed et al., 2004). In addition to
institutional arrangements, market power for allocating property rights through
transferable property rights is also discussed in the literature (Hahn, 1984; Wingard,
2000). In general, high transaction costs imply that property rights are improperly
specified (Grafton et al., 2000). Wade (1987) argued that community rights were an
effective way of internalising external costs imposed on others by resource users.
Community rights make mutually obligatory pacts as to what takes place in village
communities in Sri Lanka.
2.5.1 ISSUES IN NON-MARKET SOLUTIONS FOR RESERVOIR WATER ALLOCATION
A user-based water allocation system is common in most community managed
inland waters in Asia (Sriweera, 1994; Meinzen-Dick et al., 1996; Dinar et al., 1997).
Therefore, institutional mechanisms of water allocation in village reservoirs facilitate
the collective decision-making based on shared cultivation. WUA of water allocation
does not entirely depend on the market system or the public administrative body.
However, the main weakness of these organisations is that water allocation would be
less effective for inter-sectoral and intra-sectoral water allocation because WUAs do
not include all sectors of users when they make water allocation decisions (Meinzen-
Dick et al., 1996; Dinar et al., 1997). Rice farmers in Sri Lanka have equal user
rights and use water for agricultural purposes, even though farmers downstream
receive less water due to allocation problems.
In the context of reservoir-based agriculture, farmer households face a trade-off
between income risks and expected profit when decisions are made in relation to
water allocation under weak institutions (Mendola, 2007) or missing markets. The
behavioural assumption of a firm is to receive maximum profit in the production
process (Varian, 1992) which is not readily applicable to VISs.
This thesis argued that water allocation becomes more productive when the use
of water changes from „low‟ to „high‟ efficient alternatives. Such efficient allocation
of water is necessary to increase the total productivity of all resources (De Silva,
2003).
36
36 Chapter 2: Allocation of water resources in Sri Lanka
2.6 CHAPTER SUMMARY
Man made inland water resources are the main source of irrigation for reservoir
based agriculture in Sri Lanka. Especially in VISs, user-based water allocation is
given higher priority for rice farming. However, there is competition between water
users due to the introduction of CBF production in VISs. This has resulted in
allocating water among the competing water demands, i.e., between rice farming and
CBF production. However, there is no market price for reservoir water. Furthermore,
well defined user rights for residual volumes of water have not been established.
Given the issues mentioned, there is a growing need to increase reservoir water
productivity. This can be achieved by allocating water between competing users
optimally.
Chapter 3: Production functions and optimal allocation of water 37
Chapter 3: Production functions and
optimal allocation of water
3.1 INTRODUCTION
This chapter describes the theoretical overview and the analytical framework of
the study. The detailed theoretical overview of the study includes a discussion about
stochastic production function and technical inefficiency models that are used to
estimate TE of rice farming and the CBF production. Selection of the functional form
and the importance of imposing theoretical consistency of the selected functions are
then discussed. Analytical methods are presented after the theoretical discussion of
each section. There are three main analytical methods employed in this thesis (i)
Stochastic Production Frontier (SPF) function estimation, (ii) equi-marginal
condition and (iii) consumer surplus estimation. First, stochastic production frontiers
are estimated from primary data and used to estimate the MVP of existing water uses
(rice and CBF farming). The frontier models include an explicit inefficiency model
to determine the factors affecting TE. The estimates of MVP are adjusted to take into
account differences in TE. Second, inter and intra sectoral optimal allocation of water
is estimated based on the equi-marginal condition. Finally, welfare effects of water
re-allocation are estimated by predicting consumer surplus of water uses.
The overall analysis of the thesis is based on a „static water allocation problem‟
(Grafton et al., 2004), which occurs when there is competing demand for a fixed
quantity of water18
. This analytical framework is diagrammatically shown in Figure
3.1.
18
It assumes that reservoirs are at full supply level and that rainfall has no considerable effect on the
level of water available in the reservoirs.
38
38 Chapter 3: Production functions and optimal allocation of water
Source: Compiled by Author.
Figure 3.1. Overall analytical framework.
As a whole, the analysis of the thesis leads to a static water allocation model.
Following Figure 3.1, TE of water use is estimated for rice farming and CBF
production in Chapters 5 & 6. Inefficiency of reservoir water use is identified
between rice farming and the CBF production and between paddy fields (intra-
sector). Therefore, the estimation of TE of these two water uses is followed by the
analysis of inter-sectoral and intra-sectoral allocation. The welfare effect of water re-
allocation is estimated in Chapter 9. The next section provides a detail discussion of
the theoretical overview and the analytical framework.
3.2 THEORETICAL OVERVIEW OF STOCHASTIC PRODUCTION
FRONTIER AND ANALYTICAL FRAMWORK
3.2.1 FRONTIER PRODUCTION FUNCTIONS
Production technology can be discussed in relation to production, costs, profits
and revenue functions. In this study, only the production function approach is used.
Attention on measuring farm efficiency started with Farrell‟s (1957) pioneering
39
Chapter 3: Production functions and optimal allocation of water 39
explanation of efficiency measurement of the production frontier. Several approaches
to estimate efficiency have been discussed. Generally, parametric (Stochastic
Frontier Approach - SFA) or non-parametric (Data Envelopment Analysis - DEA)
methods are used to estimate frontier functions in efficiency studies. For the present
study, the SFA is employed to estimate frontier production functions19
for rice and
CBF production.
Production is the process of transforming inputs into outputs in the form of
either intermediate or final consumer goods. Kumbhakar & Lovell (2000) define a
production frontier as the maximum output attainable from a given level of inputs
and technology. The production frontier describes the current state of technology of a
particular firm, shown in Equation 3.1.
( )Y f X (3.1)
Y is an output and X is a vector of inputs. A production frontier is used to define the
relationship between an input and an output to show the maximum output that can be
achieved from each input or, alternatively, by representing the minimum input used
to produce a given level of output using the current level of technology.
Deterministic frontier functions can be estimated using techniques such as
deterministic non-parametric frontier, deterministic parametric frontier and
deterministic statistical frontier. However, this discussion will be limited to the
parametric approach used in the PhD study. Forsund et al. (1980) stated that there are
two main advantages of the parametric frontier approach. First, the ability to specify
the frontier in a simple mathematical form and second, the relaxation of constant
returns to scale assumption of the production function.
A general production frontier can be written as:
( ; ).y f x TEi i i
, 0 1TE (3.2)
where, iy is an observed scalar output of the producer i, i=1…N, ix is a vector of
inputs used by producer i, ( ; )if x is the production frontier, and is vector of
technology parameters to be estimated. N represents the total number of production
19
Efficiency measurement literature commonly uses the term frontier to describe the function giving
the maximum technologically feasible output (Coelli et al., 2005).
40
40 Chapter 3: Production functions and optimal allocation of water
units. iTE means TE of the i-th production unit which can be defined as the ratio of
observed output to maximum feasible output, given as:
,( ; )
yiTE
i f xi
(3.3)
When 1TEi
the i-th firm produces the maximum feasible output while 1iTE
provides a measure of the underperformance of the observed output from the
maximum feasible output.
The deterministic part of Equation 3.2 is given by ( ; )f xi
. Entire shortfalls of
observed output iy explained in Equation 2.3 from the maximum feasible output
( ; )f xi
is attributed to technical inefficiency. The deterministic part of the
production frontier ignores random shocks that can affect the production process
which is outside the control of the producer.
Stochastic production frontier
To include such random shocks into the model, Aigner, Lovell & Schmidt (1977)
and Meeusen & Van den Broeck (1977) independently proposed a specification of a
stochastic production frontier to incorporate producer specific random shocks into
the deterministic frontier. The random shocks may affect the production process due
to factors such as weather changes, economic adversities, or plain luck of the
producers (Kumbhakar & Lovell, 2000). This implies that each producer faces a
different shock, however it is usually assumed that the shocks are random and
described by a common distribution with the random shock component exp{ }vi .
The stochastic frontier production with two error terms can be modelled as:
( ; ).exp( )y f x v ui i i i
(3.4)
where:
Yi is the production of the i-th farmer (i=1, 2, 3...n),
xi is a (l x k) vector of functions of input quantities applied by the i-th farmer;
β is a (k x l) vector of unknown parameters to be estimated,
vi
is are random variables assumed to be independently and identically
distributed 2( , )N Ov
and independent of ui
s.
41
Chapter 3: Production functions and optimal allocation of water 41
ui
s are non-negative random variables, associated with technical inefficiency
in production assumed to be independently and identically distributed and
truncations (at zero) of the normal distribution with mean, Ziδ and variance
2 2(| [ , ] |)u i uN Z. Zi is a (l x m) vector of farm specific variables associated with
technical inefficiency, and δ is a (m x l) vector of unknown parameters to be
estimated (Sharma and Leung, 1998).
,( ; ).exp{ }
ii
i i
yTE
f x v (3.5)
In a stochastic frontier model the observed output iy achieves its potential
value of [ ( ; ).exp{ }f x vi i
] if, 1iTE . Otherwise, 1iTE provides a measure of the
under performance of the observed output from maximum feasible outputs in a
random shock (environmental characteristics) expressed by exp{ }vi
. The
environmental characteristics are allowed to vary among the individual producers
(Kumbhakar & Lovell, 2000).
The basic stochastic production frontier is estimated as:
ln (ln ) 1,..., 1,...,it it itY f x v u i N t T (3.6)
where lnYit is the output of firm i in time period t, x is a vector of explanatory
variables, vit is estimate of statistical noise and uit, is the estimated technical
inefficiency of firm i. Both vit and uit
are assumed to be independent and identically
distributed (i.i.d) with variance of 2
v and
2u
respectively. Several distribution
assumptions may be used (and tested) for the inefficiencies distribution. These
include a normal distribution truncated at zero, uit ~ 2[ (0, )]uN (Aigner, Lovell &
Schmidt, 1977), and a half normal distribution truncated at zero, uit ~ 2[ (0, )]uN
(Jondrow et al., 1982). The first inefficiency model was proposed by Battese &
Coelli (1992) , in which uit is defined as a time variant component (Uit= Uiexp[η(t-
T)]), where T is the terminal time period (i.e., ui,t = ui when t = T). Estimation of
reasons behind TE between firm and the industry had been undertaken as a two-stage
estimation procedure. However, Battese & Coelli (1995) proposed a one-step
procedure for estimation of parameters of an inefficiency model along with the
42
42 Chapter 3: Production functions and optimal allocation of water
parameters of the production frontier. This model defined the inefficiency as a
function of the firm-specific factors such as u z w where z is the vector of firm
-specific variables, is the associated matrix of coefficients and w is a matrix of
i.i.d. random error terms.
Output and input data are normalised in the estimation of the rice and CBF
production functions. When the data were normalised, such that
_
ln(X) and ln(Y) = 0 , the coefficient on the input levels directly report the
elasticity of the mean. The normalisation of input and output data enables the
interpretation of coefficients directly as partial output elasticities. It also allows
estimation of efficiency gains to be made while ignoring the interaction terms in the
translog model.
Production functions in the form of polynomial expressions have been used to
estimate optimal allocation of water (Gulati & Murty, 1979). Stochastic production
frontiers have been estimated to determine input oriented TE of irrigation water use
(Karagiannis et al., 2003). However they have not linked TE with MVP in order to
explain optimal allocation issues. The next section shows how TE and MVP could be
linked.
The stochastic frontier (Aigner et al., 1977; Meeusen & van den Broeck, 1977)
is considered more appropriate in agricultural applications, especially in developing
countries, where data are likely to be heavily influenced by the measurement errors
and the effects of random factors such as weather and diseases.
Technical inefficiency models
Inefficiency models developed by Battese and Coelli (1995) are used to
identify factors that influence rice and CBF production assuming that inefficient
factors may have an impact on water re-allocation. Battese and Coelli (1995) define
the inefficiency model as:
U Z Wi i i
(3.7)
iU is technical inefficiency effect, iZ is a (1 )m vector of explanatory variables
associated with technical inefficiency of producers, is an ( 1)m vector of unknown
coefficients. iW are unobserved random variables, assumed to be identically
43
Chapter 3: Production functions and optimal allocation of water 43
distributed, obtained by truncation of normal distribution with mean zero and
unknown variance 2 , such that iU are non-negative (Battese & Coelli, 1995). The
TE of the i-th sample farm, denoted by TEi is given by:
TEi = exp (-Ui) = Yi/ƒ (Xiβ) exp (Vi) = Yi/Yi* (3.8)
where Yi*= ƒ(X iβ) exp (Vi) is the farm specific stochastic frontier. If Yi is equal to Yi*
then TEi = 1 reflects 100% efficiency. The difference between Yi and Yi* is embedded
in Ui (Dey et al., 1999). If Ui = 0, implying that production lies on the stochastic
frontier, the farm obtains its maximum attainable output given its level of input. If Ui
< 0, production lies below the frontier. This is an indication of inefficiency. The
efficiencies are estimated using a predictor that is based on the conditional
expectation of exp (-U) (Coelli, 1994; Battese & Coelli, 1993). In the process, the
variance parameters 2
u and 2
v are expressed in terms of the parameterisation:
2 2( / )u (3.9)
2 2 2( )u v (3.10)
The value of γ ranges from 0 to 1 with values close to 1 indicating that random
components of the inefficiency effects make an insignificant contribution to the
analysis of the production system (Coelli & Battese, 1996).
The main justification for selecting the stochastic frontier production function
for estimating production relationships of reservoir-based agriculture is that
agricultural systems largely depend on bi-annual monsoonal rainfall. Agriculture
(rice farming) and CBF are highly sensitive to random factors such as weather, water
deficiency and pest infestations. When the production function is stochastic, the
noise (v) component represents random shocks unknown to the firm. This is an
important factor to consider. When the production process is not instantaneous (e.g.,
in the case of agriculture, fishery, dairy production) random effects on output cannot
be identified before the inputs are allocated in production (Kumbhakar, 1987). The
village agricultural economy in Sri Lanka consists of groups of farmers allocating
resources. Real economies are more complex, however principles leading to the
efficient allocation of resources are the same or a version of the equi-marginal
principle is still relevant.
44
44 Chapter 3: Production functions and optimal allocation of water
The SFA reduces reliance on the measurement of a single efficient firm which
is often the problem in other methods such as Corrected Ordinary Least Squares and
DEA. However, accounting for stochastic errors requires additional specification of a
probability function for distribution of the error and distribution of inefficiencies
(e.g., half-normal and truncated normal) depending on the assumptions imposed.
Another drawback to this method is that even if there are no errors in efficiency
measurements, there is a danger that some inefficiency may be wrongly regarded as
noise.
A number of studies pertaining to agricultural efficiency have used the
stochastic frontier technique as this method is able to take into account „random
noise‟. Of the 30 studies reviewed by Bravo-Ureta et al. (1993, 2007) in 14
developing countries, 12 studies used the stochastic production frontier approach
either using cross sectional or panel data. These studies revealed that education,
experience, accessibility to credit, extension services, and confidence in using
technology are farmer specific factors which influence TE. These papers are
examined in the literature review.
Battese & Broca (1997) have tested three different inefficiency models using
translog and Cobb-Douglas stochastic frontiers. These three inefficiency models are:
the time varying inefficiency model proposed by Battese and Coelli (1992), the
inefficiency effects model proposed by Battese and Coelli (1992) for panel data and
the non-neutral frontier model proposed by Huang and Liu (1994). Battese and Broca
(1997) highlighted possible differences in model formulations and frontier function
specifications in empirical applications. They recommend simpler model
specifications to estimate inefficiency effects.
3.2.2 TECHNICAL EFFICIENCY AND TECHNICAL INEFFICIENCY
A measure of TE of the i-th firm production unit can be defined as the ratio of
observed output to maximum feasible output. The basic model generally used to
measure TE following Kalirajan & Shand (1999) can be written as:
iTE = */i iy y = actual output/maximum possible output
Actual output is „observable‟ but the maximum possible output is not. Various
methods using different assumptions have been suggested in the literature to estimate
maximum possible output and TE. These methods are deterministic, stochastic and
45
Chapter 3: Production functions and optimal allocation of water 45
Bayesian (Kalirajan & Shand, 1999). Farrell (1957) introduced the deterministic
approach to measure EE of a firm. The stochastic frontier approach was first used in
1977 by Aigner et al. (1977) and Meeusen & van den Broeck (1977). This is the
most suitable approach for estimating production frontiers where rice and CBF
production appear to be inefficient due to water allocation issues and random effects
in reservoir-based agriculture.
TE is a component of EE. EE can be defined as the capacity of a firm to
produce a predetermined quantity of output at a minimum cost for a given level of
technology (Farrell, 1957). EE of a firm is basically divided into two components:
TE and allocative efficiency (AE). According to Farrell (1957), AE refers to the
ability to produce a given level of output using cost-minimising input ratios. TE is
associated with the ability to produce on the frontier isoquant. Alternatively,
technical inefficiency is related to the deviation from the frontier isoquant. In
general, TE is defined as „the ability of a firm to obtain maximal output from a given
set of inputs vector‟ [output-oriented TE] or „the ability to minimise input use in the
production of a given output vector‟ [input-oriented measures] (Kumbhakar &
Lovell, 2000; Coelli et al., 2005).
Input-oriented technical measure
Farrell (1957) assumes two factors of production (X1 and X2) are used to
produce a single output under constant returns to scale. Furthermore, he assumes that
the efficient production function is known. The isoquant R‟-R‟ in Figure 3.2 shows
the various combinations of X1 and X2 that a technically efficient firm might use to
produce a given unit of outputs.
Source: Coelli et al., 2005.
Figure 3.2. Simple isoquant diagram of input-orientated TE measures.
46
46 Chapter 3: Production functions and optimal allocation of water
If a firm uses quantities of inputs defined by point L1 to produce a unit of
output, the technical inefficiency of that firm is represented by the distance L1 L2.
The output at point L2 can be produced by proportionally reducing the quantity of
inputs from point L1 to L2 where the firm is technically efficient (See Figure 3.2).
This thesis used an output-oriented TE to investigate the issues of water
allocation in village reservoirs in Sri Lanka because minimising input uses such as
land in order to expand output in the industry has no meaning. Therefore, the aim is
to estimate water allocation efficiency from a given set of inputs such as land and
water. In the next section, the output-oriented technical measure of water use in rice
farming and CBF production are discussed in detail.
Output-oriented technical measure
Production is technically efficient when the maximum possible output is
generated using a given set of inputs. This is shown in Figure 3.3. It is assumed that
f(x) is the technically efficient output. If farmers use input (water) defined by point
W‟ (Figure 3.3) to produce a unit of output of rice Y1, the technical inefficiency
(output oriented measure) of that farmer/farm is represented by the distance Y1/Y2.
Furthermore, the output of rice at point P can be expanded without altering the inputs
(i.e., land and water).
Source: Coelli et al., 2005.
Figure 3.3. Rice-water frontier production function.
47
Chapter 3: Production functions and optimal allocation of water 47
3.2.3 SELECTION OF THE FUNCTIONAL FORMS AND THEORETICAL CONSISTENCY
A common question raised in estimating the relationship between observable
or unobservable variables is the use of only a priori information not specific to the
particular data set (Lau, 1986). Choice of functional forms requires investigating
several characteristics which will suit the data set. In this section, the selection
criteria of the functional forms are presented. In the second part, theoretical
consistency of the selected functional forms is discussed. Finally, estimating
theoretical consistencies using a three step procedure is examined. (Henningsen &
Henning, 2009).
Selection of functional forms
Economic theories do not provide a priori guidance for the selection of
algebraic relationship of the variables. Nevertheless, Lau (1986) has broadly
categorised selection criterion into five groups. This helps to study the problem of ex
ante choice of functional forms when the correct functional form is unknown. The
five groups are theoretical consistency; domain of applicability, flexibility,
computational facility; and factual conformity.
Theoretical consistency means that the selected algebraic functional form
should be capable enough to explain the theoretical properties of the particular
economic relationship. The most common usage of domain of applicability is the set
of values of explanatory variables over the functional form that satisfies all the
requirements of the theoretical consistency. Flexibility is an important criterion when
selecting functional forms (Griffin et al., 1987). Flexibility means the capacity of the
selected functional form to approximate stochastic effects. However, estimated
functions should show theoretically consistent behaviour through suitable choice of
the parameters (i.e., not impose abstract assumptions about the behaviour).
Generalised Leontief and translog functional forms are found to be most commonly
used in the literature (Diewert & Wales, 1987).
Computational facility can have one or more properties among linearity in
parameters: (i) explicit representativeness (ii), uniformity and (iii) parsimony. The
property of explicit representativeness makes it easy to manipulate and estimate the
values of different quantities of output and their derivatives with respect to the
explanatory variables. Uniformity means that if the functional form relates to a
48
48 Chapter 3: Production functions and optimal allocation of water
complete system, the different functions in the same system should have the same
algebraic form but different parameters. Parsimony refers to the number of
parameters in the functional form which should be the minimum possible number
required to achieve a given desired degree of flexibility. The last category of Lau‟s
(1986) criteria for the selection of functional form is factual conformity. This implies
consistency of the functional form with known empirical facts.
Griffin et al. (1987) based on a comprehensive review of the literature. They
also listed twelve choice criteria to decide how one functional form is better or more
appropriate than another. They have categorised these criteria into four categories
according to maintained hypotheses, estimation, data and application. One of the
important factors of selected functional forms is its theoretical consistency.
Theoretical consistency
The functional relationships between the input and output described by the
production frontier20
has several properties: non-negativity, weak essentiality, non-
decreasing in x (monotonicity) and concave in x (concavity); although these are not
exhaustive and not universally maintained (Coelli et al., 1999).
The values of x and q, represented on the horizontal and vertical axes are
non-negative and are finite real numbers (Figure 3.4). Therefore, the function
satisfies the non-negativity property condition. The weak essentiality property
describes that positive output is impossible unless at least one input is used. The
function shown in Figure 3.4 shows that the productions function is positive from the
origin to G region. The region from G to R violates the monotonicity property and
region 0-D violates the concavity property. According to microeconomic theory, the
production function should be monotonically increasing in all inputs. With respect to
a (single output) production function, monotonicity requires positive (or at least non-
negative) marginal products with respect to all inputs ( 0i
y
x).
20
The term „frontier‟ means that the function providing the maximum output is technically feasible
(Coelli et al., 1999).
49
Chapter 3: Production functions and optimal allocation of water 49
Source: Coelli, et al., 2005.
Figure 3.4. Concavity and monotonicity properties of a production function.
When the production frontier is not monotonically increasing, the efficiency
estimates of individual firms are inaccurate. Firm A is below the production frontier
in the non-monotone production frontier shown in Figure 3.5 and is therefore
inefficient. Theoretically, Firm B is efficient as it is on the production frontier,
however, Firm B uses more inputs to produce the same output produced by Firm A.
The same problem can occur when there are some non-monotonic intervals between
the data points (i.e., a-b in Figure 3.5). Between both data points in “A” use is
increasing while output quantity is decreasing.
.
Source: Henningsen & Henning, 2009.
Figure 3.5. Non-monotonic production frontier with non-monotonic interval.
50
50 Chapter 3: Production functions and optimal allocation of water
The translog production frontier will be used in order to estimate relative
technical efficiencies of rice and CBF farmers in this study. The translog production
function is the best investigated flexible functional form and is widely used in
efficiency estimation. However, theoretical consistency is not always achieved
globally (Sauer et al., 2006) and this is a problem. In the next section, the three step
procedure proposed by Henningsen & Henning (2009) for imposing regional
monotonocity on translog stochastic production frontiers is explained in detail.
3.2.4 ESTIMATION OF THEORETICAL CONSISTENCY
The micro economic argument is that the production of a firm increases with
respect to increases in all inputs. Theoretical consistency is significant in efficiency
frontier analysis (Sauer & Hockmann, 2005) which is particularly important for
estimating individual firm level efficiency.
There are two approaches found in relation to imposing monotonicity in SFA
in the literature. The first approach, based on maximum likelihood (ML) estimation,
estimates a translog production frontier under monotonicity and quasi-concavity
restriction (Bokusheva & Hockman, 2006). Bokusheva and Hockman (2006) applied
these restrictions locally at the sample mean which was not sufficient for obtaining
globally applicable efficiency estimates. Another problem is that the maximisation of
the likelihood function under constraints is complex and the algorithms used for
optimisation frequently have convergence problems. Several approaches use the
Bayesian Markov Chain Monte Carlo (MCMC) method (O‟Donnell & Coelli, 2005).
Nevertheless, this method is highly sophisticated and requires advanced skills in
econometrics (Heninigsen & Henning, 2009). Therefore, the MCMC method is not
widely used by applied researchers.
3.2.5 SIMPLE THREE STEP PROCEDURE FOR IMPOSING MONOTONICITY
A two–step approach for imposing monotonicity introduced by Kobel et al.
(2003) was extended to a three-step procedure by Henningsen and Henning (2009).
This is discussed briefly below:
Step 1.
Estimate an unrestricted stochastic production frontier:
51
Chapter 3: Production functions and optimal allocation of water 51
ln ln ( , ) - (3.11)
'[ ] (3.12)
y f x b u v
E u z
Step 2.
Obtain restricted β parameters by minimum distance estimation where:
^ ^ ^^ ^10 0 0
^0
arg min( ) ( ) (3.13)
. . ( , ) 0 , (3.14)is t f x i x
Step 3.
Determine the efficiency estimates of the firms and the effect of the variables
explaining technical inefficiency based on a theoretically consistent production
frontier. Then estimate the stochastic production frontier model. That is:
0 0
0 1
0 ' 0
ln ln - (3.15)
[ ] (3.16)
y y f u v
E u z
whereby the only “input variable” is the “frontier output” of each firm calculated
with the parameters of the restricted model (Henningsen & Henning, 2009). The
constant and parameter variables of the minimum distance estimation are selected by
the 1 parameter to give the final parameter estimation.
3.2.6 ESTIMATION OF TECHNICAL EFFICIENCY
Specification of the stochastic production frontier for rice
The basic stochastic production frontier for farmer i in one cropping season can be
stated as:
i i iy x v u (3.17)
where y is an ( 1)n column vector of per hectare rice output, x is an ( )n k
matrix of inputs used in rice production, except the first column which takes the
value 1 to represent intercept terms. is an ( 1)n column vector of production
parameters to be estimated. u is an ( 1)n column vector of random variable in
which iu is the difference between the ith
farmer‟s practice and best practice
52
52 Chapter 3: Production functions and optimal allocation of water
technique giving the maximum yield, given the ith
farmer‟s level of inputs ijx . This
represents the farmer specific variability and iu is either zero or negative. riv is an
( 1)n vector representing a random error term either positive, negative or zero.
Both riv and riu are assumed to be independent and identically distributed (i.i.d) with
variance of 2
v and
2
u respectively. From Equation 3.17 the ith
farmer‟s
maximum output for its specific level of inputs is represented by ix β, providing it uses
the best practice technique (i.e., u 0)i and the influence of random factors on
production is negligible (v 0)i . The advantage of stochastic frontier estimation is
that the relative variability of u and v can be separately identified. The variance ratio
parameters (γ), relates to the variability of u to total variability (ζ2) (Battese & Corra,
1977) as follows:
2 2/u
(3.18)
where: 2 2 2 0 1andu v .
3.2.7 ESTIMATION OF TECHNICAL INEFFICIENCY
Following Battese and Coelli (1995), the technical inefficiency effects, iU , is
estimated for rice farming in order to identify factors that influence technical
inefficiency of rice production assuming that inefficient factors can have an impact
on water re-allocation. This is shown as follows:
i iU Z (3.19)
where, iU is a random variable that is assumed to account for technical inefficiency
in production and is assumed to be independently distributed as truncation (at zero)
of the half normal distribution 2( (0, ))uN iZ is a (1 )m vector of explanatory
variables associated with technical inefficiency of producers, is an ( 1)m vector of
unknown coefficients. The TE of production for the ith
farmer (TEi) is defined as:
( ; ) ( )exp( )
( ; ) ( )
i i i
i i
i i
f X V UTE U
f X V (3.20)
Predicting TE is based on the conditional expectation of expression (3.20), given the
model assumptions. However, accuracy of interpretation of the TE score depends on
53
Chapter 3: Production functions and optimal allocation of water 53
the theoretical consistency of the estimated model (Sauer et al., 2006). This is
discussed in detail in the next section.
3.3 ESTIMATION OF OPTIMAL ALLOCATION OF WATER
There are two approaches used in the literature on water resource systems
analysis to estimate marginal water value: (i) simulation approach and (ii)
mathematical programming approach. In this thesis, the mathematical programming
approach was used for derivation of optimal allocation of water. In this approach,
optimisation of an economic objective function was performed subject to constraints.
With this approach, the marginal value of water represents the Lagrange multiplier.
The Lagrange multiplier shows the change in the objective function due to a change
in the constraints. These multipliers also represent the shadow prices that correspond
to what market prices would be if such methods existed (Tilmant et al., 2008). With
this shadow price, the optimal volumes of water usage can be estimated for both inter
sectors (i.e., between rice farming and CBF) and intra-sectors (i.e., within rice
farming).
3.3.1 MARGINAL VALUE PRODUCT (MVP), EQUI-MARGINAL PRINCIPLE
The theoretical underpinnings of optimal resource allocation can be analysed
by examining the input side of production technology. This involves an allocation of
variable inputs among competing uses.
Limited resources can be allocated considering the equal MVP among several
uses with knowledge of the production function and the unit price of output of each
use (Doll & Orazem, 1984). In the case of reservoir water allocation, limited water is
equally allocated among competing uses. Formally:
=... MVP MVP MVPWA WB WN (3.21)
where, WAMVP is the MVP of water used for product A, WBMVP is the MVP of water
used for product B and N is the number of users under consideration (Freebairn,
2003).
It was assumed that a fixed volume of water, W (a static allocation problem) is
allocated across competing uses and competing users (Grafton et al., 2004) engaged
in rice farming and CBF production. As a result, the optimum water allocation
between irrigation and CBF can be stated as:
54
54 Chapter 3: Production functions and optimal allocation of water
MVP MVPr f
(3.22)
Water allocation between rice farming and CBF is illustrated in Figure 3.6. The
horizontal axis shows the total volume of water available for use in rice farming and
CBF during irrigation season, (t). The vertical axis on the left depicts MVP of water
( rMVP ) used for rice farming during the irrigation season (t) and the vertical axis on
the right axis depicts the MVP of CBF (fMVP ) during the same irrigation season.
“W*” is the optimum level of water allocation.
Figure 3.6. Efficient level of inter-sectoral allocation of water. Adapted from
“Missing markets for storage and the potential economic cost of expanding
the spatial scope of water trade by D. Brennan, 2008, Agricultural and
Resource Economics, 52(4)p. 473.
From the stochastic production frontier model shown in Equation 3.17, the
marginal physical product (MPP) for both outputs can be derived, given by:
ri ri riMPP = y / x (3.23)
fi fi fiMPP = y / x (3.24)
The MPP of rice farming indicates that the amount of additional output yield riy
will be available if an additional amount of input rix is applied to rice production.
Similarly /fi fiy x represents the additional unit of output of CBF which can be
generated by using an additional unit of inputs. Consequently, the MVP for rice and
55
Chapter 3: Production functions and optimal allocation of water 55
CBF production can be derived by multiplying the output price, whereby P is
assumed to be a fixed price of output:
ri r riMVP = P (MPP ) (3.25)
fi f fiMVP = P (MPP ) (3.26)
The hypothesis here is that under a co-operative water allocation system the
marginal value products are unequally distributed across the different uses due to
inefficient water allocation.
Inter-sector water allocation for rice farming
In order to achieve optimal efficiency in water allocation, the MVP of water
used for rice farming should be equivalent to the MVP of water used for CBF (See
Figure 3.7).
Source: Compiled by Author.
Figure 3.7. Illustration of current and optimal water allocation in rice and CBF
production.
Intra-sector water allocation for rice farming
A one-period (i.e., an agricultural year) model of water use is analysed in this
thesis. A fixed amount of water, Z, is supplied from village irrigation to a canal. N
homogeneous farmers are spaced sequentially and equidistantly on the canal. Each
farmer‟s field is served by one outlet that takes off from the canal and leads directly
to his field. The ith
farmer withdraws a quantity of water wi from the overall system
supply only after the i-1th
farmer directly upstream to them. In this analysis, a
cooperative outcome was considered where net revenue was maximised across the
entire command area, given the irrigation constraint to the system as a whole.
56
56 Chapter 3: Production functions and optimal allocation of water
The diMVP curves can be derived as a function of water sources for farmers at
different locations. The term iwd is defined as the volume of water which farmers in
different locations in the command area receive from the canal (See Figure 3.8).
Figure 3.8. Determining the optimal distance of water allocation. Adapted from
“Efficient spatial allocation of irrigation water,” by U. Chakravorty and J.
Roumasset, (1991) American Journal of Agricultural Economics, 73(1)p.168.
It was assumed that the estimated relationship between rice output ( Riy ) and
distance iwd would be negative ( / 0Ri iy wd ). As pointed out, earlier reservoir-
based rice farming is essentially a rain-fed practice. Suppiah (1985) found four types
of relationships (i.e., positive, negative, no relationships and complex) between
rainfall and rice production in Sri Lanka. This finding allows us to hypothesise that
similar patterns can be observed between distance from the reservoir and rice output.
Therefore, it was assumed that one of the following two or both may hold for the
second derivative of the estimated relationship between output ( )iy and distance
( )iwd .
1. 2 2/ 0RiMVP wdi
,
2. 2 2/ 0RiMVP wdi
,
The MVP function for water received from paddy fields can be derived as
( / )Ri RiMVP p y dwi
where p is the competitive market price for rice.
57
Chapter 3: Production functions and optimal allocation of water 57
The optimal level of production is determined by setting the MVP of water
used in rice farming ( RiMVP ) to the marginal cost of water used in rice farming
( )RiMC . Similarly, the marginal benefit of water used for CBF production is
Fi RiMVP MC where RiMC is considered the shadow price of water used in rice
farming and CBF production. Therefore, when Ri RiMVP MC , the present optimal
water allocation model is assumed to be efficient.
The rule of optimal allocation between users was then derived by maximising
producer surplus subject to a total water constraint as shown in Equation 3.27 below:
Maximize ,01
W WRnMVP dw MC dwi iRi Rioi
(3.27)
subject to,
,1
nW W W
R Fi (3.27.a)
where i represents the i-th farmer and i =1, 2,…, n; W21
is the total volume of water
at the reservoir and FW is the residual volume of water after irrigation. RiMC is the
total short-term marginal cost function (Chakravorty & Roumasset, 1991) since
reservoir water is at a fixed capacity during the cropping season. RiMC is considered
water used for CBF or opportunity cost of water used in rice farming in one cropping
season. The following Lagrangian can be maximised:
L = { ( )}101
W WR nnVMP dW MC dW W W Wi iRi Ri R Fo ii (3.28)
with respect to the decision variables RW and FW , where is the usual Lagrangian
multiplier. The first order conditions give us:
MVP
Ri, (3.29)
21
The total volume of water = 0.9*total land cultivated (ha) plus the residual volume of water after
irrigation. According to estimates of the Department of Agrarian Services in Sri Lanka, the total
volume of water required to cultivate a hectare of paddy land is 0.9 metres for a single cropping
season. This calculation is based on an expected rainfall of 22 inches.
58
58 Chapter 3: Production functions and optimal allocation of water
( )MC V
Ri, (3.30)
and (2.18) and (2.19) give:
MVP MCRi Ri
(3.31)
where, is the shadow price of reservoir water. Equations (3.29) and (3.30) equate
the shadow price, to the MVP of water used for rice farming in the reservoir for
each farmer and the short run marginal cost at optimal system capacity. Finally, the
equations show the equilibrium conditions for optimal allocation of water among
users. It was estimated that the optimal distance ( *
iw d ) at which level per unit of
water is maximised by:
'( )MVP pf y MC
Ri Ri Ri (3.32)
3.3.2 MVP AND TECHNICAL EFFICIENCY (HOW DERIVED FROM SPF)
Farm-level efficiency and optimum usage of inputs can be measured by
estimating the production function. Optimal resource allocation can be measured by
deriving the MVP of each resource uses and equating the MVP of each other.
Furthermore, the MVP of each input is compared to the marginal factor cost (MFC).
Inequality of MVP and MFC shows that inputs are being used inefficiently (Husain,
1999).
It was assumed that the production relationship can be estimated as:
i i i i iLny = β lnw + v - u (3.33)
The marginal product can be derived from the production function utilising the
relationship between the production elasticity and marginal product (i.e., elasticity is
equal to the marginal product divided by the average product). This can be shown as:
ln
ln
y y w
w w y and therefore,
ln
ln
y y y
x w w (3.34)
The frontier marginal value product _______
( )MVP is equal to:
59
Chapter 3: Production functions and optimal allocation of water 59
__ __ln y y
MVP =P* *lnw w
YP
W
(3.35)
where y denotes the frontier level of production. As a result, the relationship
between TE of an existing level of production ( iMVP ) can be stated as:
______- uMVP e MVP
i , since
__- uy e y
i (3.36)
3.3.3 ESTIMATION OF INTER-SECTORAL OPTIMAL ALLOCATION OF WATER
The total benefits function of reservoir water use was optimised as follows:
(3.37)
S.T. W
where,
( ) (3.37. )
( ) (3.37. )
MaxT P Y P YR R F F
W WR F
Y f W aR R
Y f W bF F
The Lagranginan under joint maximization is:
( - - ) (3.38)T P Y P Y W W WR R F F R F
The Kuhn-Tucker (necessary first-order) conditions are:
- 0 (3.39)YT RP
RW WR R
- 0 (3.40)YT FP
FW WF F
- - 0 (3.41)T
W W WR F
60
60 Chapter 3: Production functions and optimal allocation of water
Solve for maximum use of W and W :
(3.39) - 0, (3.42)
(3.40) - 0 (3.43)
R F
Y YR RP P MVP
R R RW WR R
Y YF FP P MVP
F F FW WF F
Therefore,
, ( shadow value of water)
Then,
(3.44)
(3.41) - - 0 (3.45)R
Y YR FP P
R FW WR F
MVP MVPR F
W W W W W WR F F
as the expression 3.35 has demonstrated that:
R
RR 1 5 Ri
R
R
R
. (3.46)w
where:
P = Average market price for paddy/kg.
lnYε = Input elasticity of water used for rice production = = β +2β lnw
lnW
Y = Sample mean of rice production
w = Volume of
YRMVP P
R R RR
water allocated for rice farming
61
Chapter 3: Production functions and optimal allocation of water 61
Similarly,
F
F
F
F
YFMVP =P ε . , (3.47)
F F F wF
P = Average market price for CBF/kg
lnYFε = Input elasticity of water use for CBF production = = β +2β lnw
1 4 FilnWF
Y = Sample mean of CBF production
w = Volume of wat
where
er allocate for CBF production.
Then,
P .ε .YR R Rλ = MVP = (3.48)
R wR
P .ε .YF F Fλ= MVP = (3.49)
F wF
The optimal level of water (W ) allocation condition is:
MVP MVPR F
Therefore, optimal inter-sectoral allocation is estimated by:
. . . .P Y P Y
R R R F F F
w wR F (3.50)
3.3.4 ESTIMATION OF INTRA-SECTORAL OPTIMAL ALLOCATION OF WATER
It is assumed that the production relationship is as follows:
2lnY = + lnw + lnw
H 0 1 Hi 5 Hiv uiH iH (3.51)
2lnY = + lnw + lnw
M 0 1 Mi 5 Miv uiM iM (3.52)
2ln ln ln
0 1 5Y w w v uT Ti Ti iT iT (3.53)
where, H, M and T denote HEFs, , middle fields (MFs) and TEFs respectively.
62
62 Chapter 3: Production functions and optimal allocation of water
The total volume of water allocated for rice farming as estimated in part 1
Chapter 5 is 2.56779 Metres22
/ha. This volume of water is assumed to be distributed
among the three sectors defined as:
R H M TW = W + W + W (3.54)
Where:
RW = Total water HEFs allocated for rice farming at the existing level of TE of rice
farming, HW = HEFs, MW = MFs and TW = TEFs.
Then the total benefit function for intra-sectoral allocation of water for rice farming
is:
(3.55)
. . (3.55. )
MaxT P Y P Y P YR H R M R T
S T W W W W aH M T R
To solve the total benefit function the same procedure is followed for intra-sectoral
optimal use of water:
(3.56)MVP MVP MVPWH WM WT
Therefore, the shadow value of water is equal to the MVP of each sector (when they
are equal).
3.4 ESTIMATION OF CONSUMER SURPLUS OF WATER RE-
ALLOCATION
The economic gains of re-allocating water were measured by estimating consumer‟s
surplus among competing water users. In the context of water, consumer surplus is
the net benefits of water use to farmers after they have paid for their water. The price
of reservoir water was estimated from the MVP of water used. The allocation of
water in village irrigations was assumed to be sub optimal when water usage is
inefficient and markets are not present. Two conditions were established for effective
water re-allocation between rice farming and CBF production at the optimal and
existing levels of TE in production:
22
1 metre = 10,000 cubic metres
63
Chapter 3: Production functions and optimal allocation of water 63
1. * aTNB TNB (3.57)
2. TNB TNBF R (3.58)
Condition one is that the total net benefits of reservoir water use at the frontier
level of production (TNB*) should be greater than or equal to the total net benefits of
reservoir water use at the existing level of production (TNBa). Condition two
specifies that total benefits of water use at the frontier level of production for CBF
*
F(TMVP ) should be greater than or equal to the total benefits of water use at the
existing level of TE in production *
R(TMVP ) .
Source: Compiled by Author.
Figure 3.9. Inter-sector water re-allocation.
These two re-allocation conditions are further demonstrated in Figure 3.9.
MVP curves which are represented by Ra and F
a show production levels of rice and
CBF at the existing level of TE respectively. R is the optimal allocation of water
whereby a
RW and a
FW are the volumes of water optimally allocated for rice farming
and CBF production. The area under the two curves is consumer surplus of water
demand for rice farmers and CBF farmers. Then:
64
64 Chapter 3: Production functions and optimal allocation of water
TNBa = (R
a+R+λ
a) + (F
a+R+ λ
a) (3.59)
Similarly, the MVP curves which are represented by R* and F* indicate the frontier
level of rice and CBF production respectively. F is the optimal allocation of water
whereby *
RW and *
FW are the optimal volumes of water allocation for rice farming
and CBF production. The area covered by R*, F and λ* is consumer surplus for
water demand for rice farming. The area covered by F*, F and λ* is the consumer
surplus of water demand for CBF production. Then:
TNB* = (R*+F+λ*) +(F*+F+ λ*) (3.60)
3.5 CHAPTER SUMMARY
The review of the empirical and theoretical background in this chapter
provided a comprehensive overview on water allocation issues and model
specification, estimation and interpretation of results to be dealt with in the analytical
phase of the study. The estimation of consistent production functions, encountered in
translog production functions, was also discussed in order to further improve the
theoretical and model specification in the thesis. The production frontier model can
be applied under different scenarios and assumptions according to the specificity of
the rice farming and CBF production. In this thesis, the SPF model was selected due
to the features of property characteristics included in the model. Furthermore, this
chapter also examined some estimation problems in theoretical consistency of
translog frontier models. Finally, the simple three step procedure imposing
monotonicity and quasi-concavity introduced by Henninsen and Henning (2009) was
explained.
Chapter 4: Data collection and model definition 65
Chapter 4: Data collection and model
definition
4.1 INTRODUCTION
This chapter describes the overall design of the PhD research. The research is
dependent on primary data. Therefore, two farmer surveys were conducted in two
administrative districts, Kurunagala and Anuradhapura, in Sri Lanka from October
2009 to March 2010. One survey is called the rice farmer survey and the other survey
is called the CBF farmer survey. In total, 460 rice farmers and 325 CBF farmers were
selected from the sample districts using multi-stage cluster sampling methods. Face
to face interviewing methods were used for data collection, using pre-tested
questionnaires. Individual farmers of the rice farmer survey and the CBF farmer
groups in the CBF farmer survey were considered a unit for analysis. This chapter
also provides details of models and the data collected. Importantly, to estimate
individual volumes of water use in rice farming and CBF production, an electronic
database provided by DAD (2000) was used.
4.2 DATA
To answer the research questions, why current inter- and intra-sector water
allocation is inefficient and how water can be optimally re-allocated in VISs in Sri
Lanka, rice production data of the individual farmers and reservoir level data are
necessary. The responsible government body where secondary data can be obtained
is the DAD in Sri Lanka. However, DAD has no field level databases on rice farming
which can be used in this study. In Sri Lanka, a reservoir database on village
irrigation schemes is available instead. This electronic database includes all reservoir
characteristics (i.e., water depth, dam length, and maximum water depth), the number
of farmers and size of the command areas for individual reservoirs.
The original research plan was to obtain secondary data for the CBF survey
from the NAQDA. However, during the course of this study, NAQDA authorities did
not provide access to their official database for independent research. Therefore, the
PhD research is based on primary data except for the DAD secondary data. Two
research reports published by Hector Kobbakaduwa Agrarian Research and Training
66
66 Chapter 4: Data collection and model definition
Institute (HKARTI) and National Science Foundation (NSF) in Sri Lanka were also
used as secondary sources. These two researches reports (Aheeyar et al., 2005;
Thiruchelvam, 2003) discuss the cost of rice production in two districts, Kurunagala
(2005 December) and Anuradhapura (2003 August). The next section documents the
data collected from the field survey. .
4.3 STUDY AREAS
Kurunegala and Anuradhapura districts were selected because they have the
highest density of reservoirs in the country. According to DAD (2000), there are
10,094 village reservoirs currently being used for rice production. Kurunagala
district has 4,192 working reservoirs, the highest reservoir distribution in Sri Lanka
(DAD, 2000). Anuradhapura district is second to the Kurunagala District (See Table
A1 in Appendix A). The study area for the CBF farmer field survey was also
essentially Kurunagala and Anuradhapura districts, as high numbers of reservoirs in
these two districts have been used for CBF production in the country. This is shown
in Table 4.1.
Table 4.1
Number of reservoirs used for CBF in the selected districts
Districts Total working
reservoirs
Number of reservoirs used for CBF activities in
three culture cycles from 2006 to 2009
2008/09 2007/08 2006/07 Total
Anuradapura
Kurunegala
Total (Island)
2,333
4,192
10,094
52 (2.3%)
54 (1.24%)
375 (3.7%)
54 (2.3%)
52 (1.24%)
321 (3.2%)
63 (2.7%)
59 (1.4%)
472 (4.7%)
169
165
1168
Sources: Anon., 2009(a); DAD, 2000 and field data collected by the author.
Data on CBF production was collected in 22 Divisional Secretary Divisions
(DSDs) in the Kurunagala district. In the Anuradhapura district, 29 DSDs were
covered in the CBF farmer survey. Kurunagala district has 42% of all reservoirs on
the Island. These reservoirs are distributed among 30 DSDs. A total of 127 reservoirs
are used for rice farming in the Galgamuwa Divisional Secretary Division (DSD). 29
reservoirs in the Galgamuwa DSD were used for CBF in 2009 (NAQDA, 2008). 607
67
Chapter 4: Data collection and model definition 67
farmer households are involved in paddy cultivation, covering 1,225 ha using the
village reservoirs in the Galgamuwa DSD (ADO‟s Annual report, Galgamuwa,
2009).
The rice farmer study was conducted on 14 selected rice-farming villages, each
of which has its own reservoirs in the Galgamuwa DSD. Of the DSDs in the
Kurunagala district, the Galgamuwa DSD had the highest density of reservoirs used
for rice farming and CBF production in the 2008/09 principal agricultural season. As
these two districts are adjacent, they are therefore, homogeneous in morphology,
climate, vegetation and all other social and economic aspects. The two districts with
high reservoir densities are located in low rainfall regions. The low rainfall regions
(so called dry zones) of Sri Lanka are located within the lowest peneplain of the
island and cover approximately 66% of the total land area. This area is inhabited by
33% of the country‟s population. Current irrigation withdrawals for rice production
in these districts account for over 75% of reservoir capacity (Samad, 2005).
4.4 SAMPLE SELECTION METHODS
The multi-stage cluster sampling method (Cochran, 1960) was used for sample
selection. Each stage represents the number of reservoirs, based on an administrative
hierarchy from national level to village level as:
Stage 1; Districts
Stage 2; DSDs/ Agrarian Development Divisions
Stage 3; Grama Niladhari Divisions23
/ village reservoirs
The number of reservoirs and economic activities (rice farming and the CBF
activities) were equally taken into consideration in each stage of the sampling
method. The next sections provide details of the two studies (rice and CBF farmer
studies) and the data collection method.
4.5 SELECTED SAMPLE
4.5.1 RICE FARMER STUDY
A rice farmer is the unit of analysis in the rice farmer survey. An individual
farmer was selected from fourteen villages24
in the Galgamuwa DSD to complete the
23
Local level government administrative unit.
68
68 Chapter 4: Data collection and model definition
survey. The list of farmers in the FOs was used as a sample frame. In total, 460
farmers were interviewed. The total sample represented 76% of the total farmers in
the study area.
For the analysis of intra-sector water allocation issues in Chapter 7, the total
sample was divided into three sub-samples based on the location of the paddy fields
in the command area. The first part (denoted as HEFs are the 1000 metres from the
reservoir dam. The second 1000 metres located next to the Head-end is called MFs.
The final 1000 metres of the command area is the Tail-end (T). The breakdown of
the 460 farmers is shown below:
Table 4.2
The breakdown of the total sample
Location Distance from the dam (Metres) Number of farmers
HEFs
MFs
TEFs
Total
Less than 1000
1000 to 2000
Above 2000
160
152
148
460
Source: Compiled by Author
The estimated total reservoir capacity was 5.421(metres/hectare)25
. However,
based on the collective decision of the FOs, 62.5% of the total capacity of the
reservoir is allocated for rice farming. This is approximately 3.3881 metres/hectare
according to our estimation.
4.5.2 CBF FARMER STUDY
A group of fish farmers from each reservoir engaged in CBF production was
considered as a sample unit for the CBF survey. CBF is essentially a group activity.
For this reason, individual performance of CBF activity is unlikely. Furthermore, as
the CBF industry is not well established in all village reservoirs of Sri Lanka, CBF
24
Arthikulama, Dabagahawewa, Gallawa wewa, Gojaragama, Iddamalpitiya, Kallanchiya,
Madawachchiya, Makalanegama, Molewa, Nochchiya, Pahala konwewa, Pahala saviyagama,
Ussankuutiya wewa and Walpothu wewa.
25
10000 cubic metres = 1metre/hectare
69
Chapter 4: Data collection and model definition 69
activities are not continuing annually. Therefore, CBF production data were collected
during several different culture cycles from 2006 to 2009. In Kurunegala and
Anuradhapura districts, there were 165 and 169 reservoirs respectively (a total of
334) where CBF activities had been carried out during the three culture cycles. Data
were collected from 325 CBF farmer groups consisting of 165 and 160 reservoirs
respectively. This represents 29% of the total reservoirs (1,168) used for CBF
production in the country over the last three culture cycles (See Table 4.1). Nine
reservoirs used for CBF production in the Anuradhapura district were not sampled
due to the unavailability of an adequate number of farmers in the village during the
survey (See Table C1 in Appendix C).
4.6 DATA COLLECTION METHOD
The purpose of the surveys was to collect rice and CBF production data from
selected farmers in the sample. In this context, the most appropriate data collection
method was face-to-face interviews with selected rice farmers using a pre-tested
questionnaire. The ethical clearance committee of the QUT Business School,
Queensland University of Technology approved the questionnaires that were used for
the surveys (See Appendix J).
4.6.1 RICE FARMER SURVEY
The rice farmer study was conducted to identify inefficient water uses in rice
production and to investigate the possibility of optimal inter-and intra-sector
allocation of water. Two surveys conducted earlier by HKARTI (Aheeyar et al.,
2005) and NSF, Sri Lanka (Thiruchelvam, 2003) reported the costs of rice farming in
the Kurunagala and Anuradhapura districts. This information was used to develop
the survey questions. Two graduates from each village were trained as enumerators
for the survey. While training the enumerators, the questionnaire was pre-tested on
10 individual rice farmers. After the pre-test, the questionnaire was modified. The
final questionnaire was used in face-to-face interviews. The survey was conducted
from November 2009 to January 2010. Enumerators visited each farmer‟s house to
interview them. In each village, ARPAs and the president of the FOs facilitated the
collection of data.
70
70 Chapter 4: Data collection and model definition
4.6.2 CBF FARMER SURVEY
Through the CBF farmer survey, TE and factors influencing CBF production
were measured and the MVP of water used for CBF production was estimated. Due
to the geographical distribution of the 325 village irrigation systems selected from
the two districts and time constraints, the sample survey was organised as follows:
Step 1: Two special one-day workshops were organised for all ADOs of the
Agrarian Services Development Divisions (ASDDs) where CBF
production was carried out during the last three culture cycles in the two
districts. In these meetings, the purpose of the survey and the
questionnaire were discussed comprehensively with the Divisional Officers.
Step 2: Another two-day workshop was organised by all ADOs in their ASDD for
ARPAs working in the respective villages in the two districts. Similarly, in
this meeting, the purpose of the survey and the questionnaire were discussed
in detail with ARPAs. The following day, ARPAs were trained to interview
CBF farmers.
The CBF farmer survey was organised as a group discussion. Officials
(president, secretary and treasury) and a few members of FOs essentially represented
the group interview. ARPAs worked as enumerators of the survey. Districts Agrarian
Development Commissioners (DADC) from the two districts organised meetings
with ADOs. DADC also helped organise and train ARPAs for the surveys in their
divisions. All ARPAs corresponded with each other during the survey. The CBF
farmer survey was completed within 4 months, from December 2009 to March 2010.
4.7 MODEL DEFINITION
Rice production model
The main explanatory variables used in the stochastic rice production frontier
model, were discussed in Section 2.2.1 in Chapter 2. All variables used in the model
are expected to have a positive impact on production.
1. Water (metres/hectare)
There is no proper water measuring system under VISs in relation to the way
data were calculated. Each farmer draws water from an unmetered sluice gate to the
field along the canal. Since the amount of water used is unmetered, the following
71
Chapter 4: Data collection and model definition 71
approach was used to measure the amount of water used. An individual share of
water used by each farmer was measured using the following equation:
wri = lit/Li*Cr*Ri
where:
wri = Individual share of water use by ith
farmer
lit = Land size cultivated by the ith
farmer in season t
Lt = Total land cultivated at the command area of the ith
cropping season
Cr = Reservoir capacity of full supply level
Ri = Percentage of water use in rice farming.
This last coefficient is assumed constant for all reservoirs and has been
estimated by FOs as 0.625 based on the existing water allocation.
Reservoirs are distinct by their size. No two reservoirs are the same.
Therefore, each reservoir‟s capacity is different. Individual reservoir capacity is
calculated using the following formula by Department of Agrarian Development,
Department of the Ministry of Agriculture and Agrarian Services (MAAS) for
reservoir capacity estimation:
C = 0.4 × D × WSAri i i
where:
Cri = ith
Reservoir capacity at full supply level
0.4 = coefficient,
D i= maximum water height (Ft) of ith
reservoir
WSAi = water spread area (acres) during full capacity level
2. Labour (man days)
Labour is defined as all forms of physical labour (own, hired, shared, and
family) used for agricultural activities. A man-day counts for approximately 8.5
working hours per day.
72
72 Chapter 4: Data collection and model definition
3. Mechanical Power (minutes)
Mechanical power is used for land preparation, threshing paddy and
transporting the harvest. Mechanical power is measured by time (minutes) used for
each activity.
4. Irrigation time (minutes)
Irrigation times of the paddy fields depend on various factors such as location
of the command area, soil type, evaporation and farmers‟ water conservation
practices. It has been measured as irrigating time of an individual paddy field by
minutes ceteris paribus. Individual share of water used by farmers correlates highly
with the individual cultivated land. Therefore, “land” is not an explanatory variable
in the model. Addition of “irrigation time” as an explanatory variable avoids multi co
linearity of the model. Irrigating time has been estimated based on a 24 hour per day
calculation.
5. Pesticides
Pesticides include insecticides and weedicides. Some farmers used
insecticides while others used weedicides and some used both. Therefore, in the main
model, pesticides were defined as the use of insecticides and weedicides. Use of
insecticides and weedicides were included in the inefficiency model as dummy
variables. Use of pesticides was measured by millilitres.
Use of fertiliser was one of the most important variables excluded from the
main model. Every farmer must use chemical fertiliser under the government-
subsidised programme. Under this programme, DAD provides sufficient quantities of
chemical fertiliser to every farmer who is involved in cultivation.
Technical Inefficiency model
The variables used in the technical inefficiency model for rice production (See
Table 4.2) were expected to have both positive and negative signs. These variables
broadly fall into two categories; farmer-specific variables and farm-specific
variables.
73
Chapter 4: Data collection and model definition 73
Table 4.3
Description of variables of the inefficiency model
Variables Sign Description
Farmers‟ age (years)
Farmers‟ education (years)
Participatory rate for FO activities (%)
Membership of FO2 (0 or 1)
Paddy field location ,Head-end (0 or 1)
Paddy field location, Middle (0 or 1)
Locational water sharing issue (0 or 1)
Paddy field ownership (0 or 1)
Use of insecticides (0 or 1)
Use of weedicides (0 or 1)
Success of field level water mgt.
(+)
(-)
(-)
(-)
(-)
(-)
(+)
(+)
(-)
(-)
(-)
Age of the farmer
Number of schooling years
Rate of farmers involved in collective action
Member of a FO (Dummy variable)
Paddy field located at the HEFs (Dummy variable)
Paddy field located at the MFs(Dummy variable)
Issue prevailing with water sharing among the fields
Whole ownership of paddy field (Dummy variable)
Insecticide used (Dummy variable)
weedicides used(Dummy variable)
Individual of field level water management
Most of the variables in the inefficiency model provide additional information
of the inputs in the production function as part of the frontier. Farmers‟ age,
education, participation rates for FO activities and FO membership are related to
labour. To be a FO member and entitled to water use for irrigation systems, land
ownership is a prerequisite. Therefore, FO membership and land ownership are
indirectly linked to water use. The pre assumption that individual volumes of water
are gradually decreasing with increasing distances to the water source (Chakravorty
& Roumasset, 1991). Therefore, difference in efficiency due to location may result
measurement in the water use. Water sharing between paddy fields as well as
different water users is directly linked with the water variable in the model. Use of
insecticides and weedicides are characteristics of the variable used in the main
model. However, it has also been found that the use of weedicides relates to the
quantity of water used for rice farming. The last variable of the inefficiency model is
the individual farmers‟ assessment of their success of the field level water
management level.
74
74 Chapter 4: Data collection and model definition
CBF models
The main explanatory variables, their expected signs and descriptions in the
CBF stochastic production frontier are given below:
1. Water (metres/hectare)
The amount of water available for CBF activities is given by:
(1 )w C Rfi r i
where:
Cri = ith
Reservoir capacity of the full supply level (defined previously)
Ri = Percentage of water use in rice farming.
wfi = volume of water available for CBF activities at the end of the
production cycle
Stocking density of fish fingerlings is estimated at 50% of the full supply level
of the reservoir‟s capacity because reservoir capacity varies over the season
(Wijenayake et al., 2005). However, under the current water allocation system by
FOs, allocated water for CBF is estimated as 0.375 out of the reservoir capacity Wfi =
Individual share of water used by i-th
reservoir for CBF production.
2. Group labour (man days)
Labour is defined as all forms of physical labour (own, hired, share, and
family) and labour use for CBF activities is considered as group labour. Labour
allocation depends on group decisions, determined by group size. Individual labour
in CBF activities is minimal.
3. Number of fish fingerlings seeded
This variable represents the total number of fish fingerlings stocked in the
reservoirs. Two types of fish species are stocked in reservoirs, Indian carp and
Chinese carp. Some species are fast growing while others grow at a slower rate
(Wijenayake et al., 2005; http://www.fao.org/fishery/culturedspecies/search/en).
75
Chapter 4: Data collection and model definition 75
Technical inefficiency model
Descriptions of the variables used in the technical inefficiency model for CBF
are shown in Table 4.4. The expected signs associated with these variables (negative
or positive) are also indicated in Table 4.4.
Table 4.4
Description of variables of the inefficiency model
Variables sign Description
Group stability
Time spent on meeting
officials
No rain water risk for CBF
Subsidised fingerlings supply
No of cattle and buffalos
Slow growing fingerlings
Fast growing fingerlings
Number of months of water use
for other uses
(-)
(-)
(+)
(+)
(-)
(+)
(-)
(-)
Continuation of CBF activities with the same group in the
following year (Dummy variable)
Visiting time of government officials to provide extension
services (hours)
Yearly adequate rain water availability for CBF (Dummy
variable)
Fingerling or money received from third party to invest CBF
(Dummy variable)
Number of cattle and water buffalos grazing or living in the
reservoir catchment
Mrigal (Cirrhinus mrigala Hamilton), rohu (Labeo rohita
Hamilton), Nile tilapia (Oreochromis niloticus L.) and the
other species considered as slow growing species.
Common carp (Cyprinus carpio L.), bighead carp
(Hypophthalmichthys nobilis ) and catla (Catla catla
Hamilton)
(http://www.fao.org/fishery/culturedspecies/search/en).
Number of months whereby water is used for other uses
4.8 CHAPTER SUMMARY
The details of research design and model description were presented in this
chapter. In particular, it outlined the most suitable methodology for data collection.
The research mainly depended on primary data due to the unavailability and
inaccessibility of secondary data relating to the research questions. However,
individual volumes of water used by the rice farmers and the CBF farmers were
estimated using the electronic database of DAD in Sri Lanka. Suitable variables for
the models were identified in previous research and pretesting stages of the
questionnaires.
Chapter 5: Efficient water usage in village irrigation systems for rice farming 77
Chapter 5: Efficient water usage in village
irrigation systems for rice
farming
5.1 INTRODUCTION;
The main objective of this chapter is to investigate why the existing use of
water is efficient for some rice farmers while it is not for others. This question is
addressed by estimating theoretically consistent stochastic production frontier for
rice farming in the VISs. The estimated model specifies five input variables: water,
labour, mechanical power, irrigating time of paddy fields and use of pesticides. The
inefficiency model is specified by 11 characteristics of the input variables (a detailed
description of the variables are provided in Section 4.7). Individual volumes of water
and the success of field level water management are two variables, which have not
been tested in previous studies with respect to the TE of rice farming. This chapter
also provides an overview of rice production and reviews existing literature on the
TE of rice farming. In Sections 4.4 and 4.5 descriptions of the estimated model and
results are presented. The results are then discussed and relevant policy implications
are suggested in Section 4.7.
5.2 RICE PRODUCTION
After land, irrigation is the main input factor for rice production. The projected
global rice consumption in 2020 will increase by 35% from the 1995 level. At the
same time, water availability for agricultural purposes over this period is expected to
drop from 72% to 62% globally and 87% to 73% in developing countries (Rosegrant
et al., 1997). Rice is a staple food for three out of five people in the world. The vast
majority of monsoonal Asia are rice-consumers (Farmer, 1977). The Agricultural
sector is still dominant in providing employment, generating Gross National Product
(GNP), alleviating permanent and temporary poverty (Hussain & Hanjra, 2004) and
reducing malnutrition. Irrigation development and management plays an important
role in agriculture development in Asia. One of the important targets of agricultural
development in Asian countries is to adopt strategies to achieve food self-sufficiency
(Sampath, 1992). Moreover, about 50% of the total fresh water resources in these
78
78 Chapter 5: Efficient water usage in village irrigation systems for rice farming
countries are used to cultivate rice (Barker & Tuong, 2001). The economy of Sri
Lanka has diversified with the structural adjustment programme since 1977, however
“rice” remains a key economic sector. Over 800,000 farming families are involved in
rice production, contributing about 5% to the Gross Domestic Product (Anon, 2009).
Irrigation systems ranging from very large reservoirs to the village irrigations
throughout the country are used for paddy production. However, the VISs play a
main role in areas which are unable to access water from the main irrigation systems
in the low rainfall region. Increase in competition between multiple uses has further
aggravated the problem of “water scarcity”.
Agriculture in Sri Lanka consists of cultivation of rice (paddy) and field crops,
livestock, bee keeping and inland fisheries (Agrarian Development Act, No 46 of
2000). This study focuses on the village reservoir-based agriculture in Sri Lanka. The
majority of these reservoirs are less than 100 Hectares in surface area and distributed
across the undulating landscape of the low rainfall regions. The total extent of VISs
in Sri Lanka covers 39,271 Hectares (Mendis, 1977). This is approximately 23.1% of
total surface water area in the low rainfall regions of the country (De Silva, 1988;
Fernando, 1993). However, agricultural activities are practised in the entire low
rainfall regions of the country where VISs are widely located.
Water in the VISs is used for cultivation of crops (mainly paddy), livestock
farming, aquaculture, brick-making, and domestic purposes. Therefore, the village
reservoirs are a common pool resource with multiple uses (Meinzen-Dick & Bakker,
2001). Management of VISs are undertaken by FO which has been established for
each village. The farmers who are landowners in the command area of the system
can obtain membership in a FO. The water allocation for each plot of paddy field is
not determined by its size. Farmers receive an allocation of water based on a
communal agreement. This type of water allocation mechanism is called „user-based
water allocation‟ (Dudu & Chumi, 2008; Dinar et al., 1997). The irrigation network
operates on a rotational delivery system based on decisions made at the FO‟s first
meeting of the year (the kanna meeting). Decisions are also made at this meeting
about periodic clearing of the drainage channels and maintenance of the reservoir
systems which are to be undertaken by individual farmers.
Rice farming is the main source of employment for the majority of people
living in villages during the two main agricultural seasons. The season of cultivation
79
Chapter 5: Efficient water usage in village irrigation systems for rice farming 79
(maha) is the first agricultural season, usually running from October to March. This
coincides with the North-Eastern monsoon giving favourable rainfall for paddy
cultivation. The second season, a subsidiary period of cultivation (yala), runs from
April to September falling into the low rainfall season26
. This season relies on the
South-Western monsoon (Zubair, 2002). Between these two labour intensive
agricultural seasons (See Table 6.1), farmers have sufficient time to organise other
economic activities such as CBF. However, existing water allocation between rice
farming and other uses is not well organised by the FOs.
Rice farming is given higher priority than other agricultural activities since rice
is a staple food and provides food security for rural farmers. Rice farming accounts
for 16% of total land area of the country and 800,000 farmers and their families are
directly involved in rice farming (Anon, 2009). In total, 70% of rice farmers have
small land holdings of less than a hectare (De Silva et al., 2007). The contribution of
rice production to the GDP is well below the potential rice production for Sri Lanka.
Furthermore, TE of rice farming in Asia is still very low compared to other rice
producing countries in the world (Bravo-Ureta et al., 2007). According to Bravo-
Ureta et al (2007), inefficiency of rice farming is mainly attributed to inefficient
water resource allocation. Some of the significant factors which have been identified
in relation to TE in agriculture are: environmental and geographical characteristics
(Squires & Tabor, 1991), infrastructure (Huang et al., 1986), and human capital
(Kalirajan & Flinn, 1983). The next section reviews agricultural related literature on
production efficiency.
5.3 LITERATURE REVIEW
A number of studies have been undertaken to address the various aspects of
TE. These studies have estimated TE of different production sectors (i.e., agriculture,
dairy, railway, health) and crops (i.e., rice, sugarcane, wheat, maize) in developed
and less developed countries (Bravo-Ureta et al., 2007). TE also measures various
functional forms (i.e., linear, CES, Cobb-Douglas) using parametric or non-
parametric estimation methods (Battese & Corra, 1977; Griffin, et al., 1987;
Kalirajan & Shand, 1999). Furthermore, current technical inefficiency measures
26
There is no major difference between summer and winter seasons in Sri Lanka. These two seasons
are based on the average rainfall. However, maha season is the main agricultural season in Sri Lanka.
80
80 Chapter 5: Efficient water usage in village irrigation systems for rice farming
involve first stage specification and estimation independently proposed in stochastic
frontier production functions by Aigner et al. (1977) and Meeusen & van den Broeck
(1977). In this chapter, a review of relevant literature directly related to the research
questions in this PhD thesis will be discussed.
The most practical parametric methods of TE estimation are based on
stochastic frontier production function models, which have been applied across a
large number of empirical studies in agricultural economics (Belloumi & Matoussi,
2006). This method facilitates the estimation of the magnitude of random effects on
TE, beyond the control of producers. The functional forms of estimation of
production and theoretical consistency (See Chapter 2.2) of the models (especially,
the monotonicity condition) are equally important for policy decisions based on the
results of the inefficiency models (Sauer et al., 2006).
There are various factors that can have an impact on the TE of crop production.
Key variables explaining TE in agriculture can be broadly divided into two
categories. Land, water, labour, pesticides, fertiliser, and power are considered main
input variables. The farm and farmers‟ specific characteristics such as farmers‟ age
and education level, years of experience, land ownership, farm size, extension
services, technology, and institutions (ownership and user rights) are factors which
influence TE. These factors were reviewed and compared to previous studies. Merits
and drawbacks of each factor are discussed in the next section.
Land is the foremost input in agriculture. And other land augmenting input is
water, However, Water is used as an explanatory variable in very few studies. Yao
and Liu (1998) found that the most appropriate way to increase grain output in China
was to increase land productivity. Land productivity can increase by using more
inputs (i.e., fertiliser, irrigation) in the short term. However, increasing inputs on a
fixed extent of land is subjected to the law of diminishing returns. Therefore, Yao
and Liu (1998) suggested improvement of TE as a long-term solution for increasing
grain production in China. Furthermore, Yao and Liu (1988) found that irrigation
water was significant (1% level) variable in their stochastic frontier production
function, studying grain (rice, wheat and maize) production. However, they
measured water as a ratio of irrigated area to the total cultivated area for grain
production. This ratio was used as a proxy variable to the water which does not
measure the exact volume of water used in the production. hectares of irrigated land
81
Chapter 5: Efficient water usage in village irrigation systems for rice farming 81
was used by Belloumi and Matoussi (2006) in their estimated stochastic frontier
function for date production in Tunisia. Results of Belloumi and Matoussis‟ (2006)
study revealed that increasing irrigation water by 10% made it possible to increase
date production by only 2.23%. Nevertheless, measuring water by the hectares of
irrigated land no longer estimates the volume of water used. Sharma et al. (2001)
carried out a similar study for rice production in Tarai, Nepal. Sharma‟s (2001) study
revealed that increasing irrigation by 10% increased rice production by only 1.2%.
They measured water by the number of irrigations (the number of times) of water
released to the farm from the main water source. However, Sharma et al. (2001) were
not confident with their method for measuring water. They suggested that the actual
volume of water (such as cm3 or m
3) used by individual farms should be used to
estimate the effects of water supply in agriculture.
The specific characteristics of the main inputs of the production affect
technical inefficiency effects among farmers. Resource allocation could become
inefficient due to underemployed attributes of inputs. The presence of technical
inefficiency also leads to a reduction in output which in turn demands more inputs
(Kumbhakar, 1987). In order to estimate inefficiencies of a farm, Battese and Coelli
(1995) developed an inefficiency model. Battese and Coelli (1995) tested 14 Indian
paddy farmers over a ten-year period using panel data. The null hypothesis tested
was that inefficiency effects were not stochastic or they were independent from
farmer specific attributes. Battese and Coelli‟s (1995) results rejected the null
hypothesis. However, results of the inefficiency model were interpreted in two
different ways: (i) A positive coefficient described the negative effect on TE and, (ii)
negative coefficient of the model explained positive effects on TE. However, the
theoretically inconsistent translog estimated models did not facilitate accurate
interpretation of TE.
Farmers‟ age is one of the influential characteristics of agricultural labour on
TE of agricultural production. Several studies suggested that farmers‟ age has a
negative effect on TE (Battese & Coelli, 1995; Wadud & White, 2000;
Thiruchelvam, 2002; 2003; Al-hassan, 2008) including rice farming in VISs in Sri
Lanka. This means that older farmers are more inefficient than younger farmers.
Older farmers were not always willing to adopt better practices, whereas younger
farmers were more motivated to use better agricultural production practices.
82
82 Chapter 5: Efficient water usage in village irrigation systems for rice farming
Nevertheless, in some other cases, the older farmers were more technically efficient
than the younger farmers (Villano & Fleming, 2006; Khan al el., 2010). Furthermore,
Aheeyar et al. (2005) found that farmers age is a significant (at 1% level) and
positively related variable on TE of rice farming in VIS. This would be possible if
older farmers had more experience and knowledge of the production activities and
were more reliable in performing production tasks. Battese and Broca (1997) found
that farmers‟ age had positive and negative influences on TE with respect to the
special variation of farm locations in Pakistan. However, this inefficiency effect
cannot be correctly predicted because their estimated stochastic production frontier
was theoretically inconsistent due to the violation of monotonicity conditions of
labour.
Many studies considered education and experience as important characteristics
of labour, used for agricultural production. Formal education does not directly focus
on rice farming practices in most of the agricultural dominated countries. Generally,
education enhances human capital while experience improves efficiency. According
to Villano & Fleming (2006), a higher level of education results in lower
inefficiency, especially in farm management. Kalirajan (1981) found significant yield
variations among paddy fields due to years of experience and education levels of rice
farmers in India. Studies by Battese and Coelli, (1995), Sharif & Dar, (1996),
Aheeyar et al. (2005) and confirm this. Khan et al. (2010) revealed that education
and experience had different TE results with respect to the two varieties of rice in
Bangladesh. Estimated coefficients for education and experience had insignificant
positive effects on TE for Aman rice production, whereas it significantly (5% level)
improved TE for Boro rice production. This could be due to inadequate information
and training which should be provided through the agricultural extension services.
The extension services for new technological applications is the other factor for
yield variations in rice farming (Kalirajan, 1981; Pitt & Lee, 1981 and Kalirajan &
Flinn, 1983). Karagiannis et al. (2003) revealed that modern technology (e.g., green
revolution technology), education and extension services were positively associated
with irrigation water efficiency whereas farming intensity, chemical use, and the
percentage of rental land were negatively associated with irrigation water efficiency.
However, these results cannot be compared to a production system of VISs due to the
different irrigation infrastructures. The hypothesis that enhancement of irrigation
83
Chapter 5: Efficient water usage in village irrigation systems for rice farming 83
infrastructures can have a positive impact on high level of TE was examined by
Kalirajan and Shand (1986) in Malaysia. They found that TE of rice farmers with and
without irrigation facilities were significantly not different from each other. The
results demonstrated that new technology, compared to traditional technologies, had
not increased TE. However, the use of new technological advancement was
constrained by farm size.
Here and elsewhere Huang et al. (1986) found that small farms had higher
returns to scale (0.92) than that of large farms (0.84) in Punjab and Haryana States,
India. However, the difference in economic efficiency (EE) between large and small
farms was only 4%. In addition, Pitt & Lee (1981) found that large and newly
established farms were more efficient than smaller and older farms. Furthermore,
Kumbhakar (1994) found that the inadequate use of inputs like chemical fertiliser,
organic manure, labour, and bullock power had greater impact on TE of rice
production in West Bengal, India. Both the farm level production risk and technical
inefficiency in rice farming have been simultaneously investigated in the Philippines
by Villano and Fleming (2006). According to Villano and Fleming (2006), mean area
under cultivation, labour and quantity of fertiliser used influenced the output.
The other aspects of TE of a firm are institutional capacity and governance
(Estache & Kouassi, 2002). The introduction of a management system for resource
allocation with poorly defined property rights may generate externalities that impose
indirect costs or benefits to water users, leading to an inefficient allocation of water
resources (Heaney & Beare, 2001). Thiruchelvam (2002) investigated forms of
institutional arrangement in agricultural production in Sri Lanka. Thiruchelvam
(2002) emphasised that group efficiency of farmers under FOs should be enhanced in
order to increase the efficient resource allocation. Lack of property rights is a
common issue in water allocation efficiency because individual volumes of water use
are not correctly estimated using estimated models as input variables. However,
developing a suitable water allocation policy will remain a challenge without proper
understanding of the impact of institutions on the TE of resource use. Therefore,
property rights are considered a strong component of any water allocation
mechanism for efficiency and for achieving equity (Meinzen-Dick & Bakker, 2001).
In conclusion, the theoretical consistency (monotonicity condition) was not
fulfilled for estimated models (i.e., Thiruchelvam, 2003, Aheeyar et al., 2005).
84
84 Chapter 5: Efficient water usage in village irrigation systems for rice farming
Therefore, some of the estimated results of the literature may not be interpreted
correctly with respect to inefficiency effects because the estimated results of the
stochastic model were theoretically inconsistent (Sauer et al., 2006). Therefore,
estimating theoretically consistent frontier models is an essential requirement to
correctly predict TE effects. Secondly, correct measurement of individual volumes of
water used for rice farming is necessary to understand the marginal effect of input
use in the production. Finally, individual performance of water management at the
field level is a crucial variable in estimating technical inefficiency in rice farming.
5.4 EMPIRICAL MODEL
This chapter addresses two main shortcomings in the previous research: (1)
individual water allocation, (2) the theoretical consistency of the model and
inappropriate sample size. Clear evidence is yet to be provided to show how
individual water allocation contributes to efficiency in rice production; it has been
found that some translog stochastic production frontier estimates in the literature
have not fulfilled the theoretical consistency (Sauer, et al., 2006). Therefore,
inaccurate estimations have the potential to lead to inappropriate policy-making
decisions (Sauer et al., 2006). This study differs from previous research
methodologies: firstly, by integrating the individual volume of water as an
explanatory variable in the frontier production function and secondly, estimating the
translog stochastic production frontier by following the three step simple procedure
for imposing monotonicity conditions proposed by Henningsen and Henning (2009).
The main objective of the research was to investigate TE and to discover
factors influencing technical inefficiency of rice farming in VISs in Sri Lanka. In Sri
Lanka, there have only been two attempts to estimate the TE of rice production for
VISs (Thiruchelvalm, 2003; Aheeyar et al., 2005) Thiruchelvalm (2003) and
Aheeyar et al. (2005) estimated TE in Anuradhpura and Kurunagala districts. Their
sample size constituted 40 and 50 rice farmers respectively. In this study, primary
data were collected from 460 farmers in the Galgamuwa DSD in the Kurunagala
district, Sri Lanka. Multi-stage cluster sampling method was used for sample
selection. Each stage represented the number of reservoirs, based on an
administrative hierarchy from national level to village level. A rice farmer was the
unit of analysis in the rice farmer survey and 460 farmers were interviewed. This
represented 76% of the total farmers of the study area. Data were collected in person
85
Chapter 5: Efficient water usage in village irrigation systems for rice farming 85
by interviewing selected rice farmers using pre-tested questionnaires. Two university
graduates who were trained as enumerators assisted in data collection from each
village participating in the survey. The survey was undertaken from November 2009
to January 2010 (See Chapter 4).
Empirical models
The general translog functional form used to estimate the production frontier
can be expressed as:
5 5 51ln ln ln ln -
, 0 , , , , , , ,21 1 1
Y x x x v ui t i i k i k i k i l i t i t
i i k (5.1)
where Y is the rice output of farmer i in period t and i,k,tx is the agriculture inputs
(k,l) to the production process. riv and riu follow the same definition in Equation
3.4. During the literature review, discussions with farmers during the pretesting
period and data collection phase, several input were identified. Input variables i( )x
included explanatory variables, described in Section 4.7. Several other specifications
of the models were estimated using different inputs (i.e., land, fertiliser). This
specification was chosen as the final variables were found to be significant and not
highly correlated.
1rx = Individual share of water use by the i-th farmer (cubic metre).
2rx = labour (total operational man days in rice farming)
3rx= Mechanical power (minutes)
4rx= Irrigating time/total time for irrigating
5rx= Pesticides
i = the parameters to be estimated.
Furthermore, u is assumed to be non-negative and has truncated half normal
distribution. The vector v is normally distributed as (0, 2
v). Following Battese and
Coelli (1995), the mean of farmers-specific TE (Ui) is defined as:
11
0
1
i j ij
j
U Z (5.2)
86
86 Chapter 5: Efficient water usage in village irrigation systems for rice farming
where:
Z1 = Farmers‟ age (years)
Z2 = Farmers‟ education level (years)
Z3 = Participatory rate for FO* activities (%)
Z4 = Membership of FO (Dummy; 1 = yes, 0, otherwise)
Z5 = Paddy field location (Dummy; 1 = located at head-end, 0, otherwise)
Z6 = Paddy field location, (Dummy; 1 = located at the middle, 0, otherwise)
Z7 = Locational water sharing issue (Dummy; 1 = yes, 0, otherwise)
Z8 = Paddy field ownership (1 = own land; 0, otherwise)
Z9 = Use of insecticides (Dummy; 1 = yes, 0, otherwise)
Z10 = Use of weedicides (Dummy; 1 = yes, 0, otherwise)
Z11= Success of field level water management
5.5 RESULTS
This chapter focused on TE and the factors influencing technical (in)efficiency
of rice production for VISs. The variations of the rice production were explained in
terms of four inputs, water, labour, mechanical power, and irrigating time (See
Section 4.7). The volume of water use by the individual farmers was one of the input
variables, which was the most appropriate measure of estimation of the production
frontier in rice farming. Summary statistics of the output and input variables together
with various farm and farmer-specific variables included in the frontier model are
presented in Table 5.1. The labour for rice farming was estimated as eight and half
hours per day for the whole cropping season. Mechanical power, used for land
preparation, harvesting, thresh paddy and transport was estimated as the total minutes
for the investigated cropping season.
The individual shares of water were estimated as a standard quantity by the
DAD in Sri Lanka (i.e., the total estimated volume of water required for a hectare in
the main (maha) season is 0.9). The variation of the individual shares of water was
estimated in the sample by including the individual irrigating time per cropping
season. Therefore, area of cultivation (land) was not used as an input variable in the
model as it is correlated perfectly with water use. Therefore, the inclusion of water
and land variables may lead to a problem of multicollinearity in the frontier model.
The use of fertiliser has been imposed as a law by the government. The government
has provided fertiliser for a subsidised price since 2006. Therefore, farmers must use
87
Chapter 5: Efficient water usage in village irrigation systems for rice farming 87
the recommended quantity of fertiliser for their paddy field. However, farmers who
cultivate as a tenant or under another agreement with the landowner are not entitled
to receive fertiliser from the government fertiliser subsidy programme. As a result,
land and fertiliser were excluded from the model.
Table 5.1
Summary statistics of variables involved in the stochastic frontier model
Variables Mean Std. Dve. Min Max
Yield 1183.8260 904.2277 44 5100
Water (Metres/ha) 0.0957 0.1069 0.0141 1
Labour (man days) 49.1648 61.8065 4.0000 560
Power (min) 323.4685 260.7954 15.0000 1520
Irrigating time of fields (min) 2185.7670 2152.6910 120 17280
Pesticides (ml) 728.4652 703.0407 50 5600
Age of farmers 49.1848 12.6344 19 90
Ln water -2.7109 0.8012 -4.2638 -0.1409
Ln water *Ln water 7.9895 4.0649 0.0198 18.1800
Level of education 8.0848 3.2819 2 13
Participation rate FO activities 80.6457 23.5615 4 100
Membership of FO 0.8630 0.3442 0 1
Field location (head-end) 0.3478 0.4768 0 1
Field location (Middle) 0.3304 0.4709 0 1
Locational water issue 0.3761 0.4849 0 1
Land ownership 0.6413 0.4801 0 1
Use of insecticides 0.7283 0.4453 0 1
Use of weedicides 0.9391 0.2394 0 1
Success of field level water mgt 77.6457 25.3991 0 100
Step one is to estimate unrestricted stochastic frontier using maximum
likelihood estimates. The results are presented in Table 5.2.
88
88 Chapter 5: Efficient water usage in village irrigation systems for rice farming
Table 5.2
Initial maximum likelihood estimates (unrestricted frontier estimation)
Variables β Estimates Std.Error Pr(>|z|)
Constant β0 0.2754 0.0615 0.0000***
Ln water (cubic metre) β1 0.3108 0.0350 0.0000***
Ln labour (man days) β2 0.1620 0.0298 0.0000***
Ln power(min) β3 0.1631 0.0364 0.0000***
Ln irrigating time of fields(min) β4 0.0521 0.0270 0.0538*
Ln pesticides β5 0.1100 0.0300 0.0002***
Ln water x Ln water β6 0.1503 0.0563 0.0077
Ln water x Ln labour β7 -0.0445 0.0406 0.2731
Ln water x Ln power β8 -0.0578 0.0347 0.0955
Ln water x Ln irrigating time Β9 0.0073 0.0312 0.8142
Ln water x Ln pesticides β10 -0.0072 0.0379 0.8488
Ln labour x Ln labour β11 0.0418 0.0593 0.4803
Ln labour x Ln power β12 0.0808 0.0489 0.0983
Ln labour x Ln irrigating time β13 -0.0099 0.0352 0.7776
Ln labour x Ln pesticides β14 -0.0121 0.0367 0.7408
Ln power x Ln power β15 0.1445 0.0468 0.0020
Ln power x Ln irrigating time β16 0.0345 0.0354 0.3291
Ln power x Ln pesticides β17 -0.0798 0.0398 0.0449
Ln irrigating time x Ln irrigating time β18 -0.0754 0.0464 0.1044
Ln irrigating time x Ln pesticides β19 0.0069 0.0334 0.8369
Ln pesticides x Ln pesticides β20 0.1237 0.0523 0.0181
Age of farmer (yrs) δ1 0.0029 0.0066 0.6604
Farmer‟s education level ( yrs of schooling) δ2 0.0018 0.0310 0.9529
Participation rate for FO activities (%) δ3 -0.0138 0.0077 0.0739*
FO membership (1= yes, 0, otherwise) δ4 -0.6316 0.2750 0.0216**
Field location (1= head-end, 0,otherwise) δ5 0.3194 0.2382 0.1801
Field location (1= middle, 0, otherwise) δ6 0.6528 0.3368 0.0526*
Water sharing issues (1= Yes, 0 = no) δ7 0.9940 0.3908 0.0110**
Land ownership (1= own; 0, other) δ8 0.4529 0.3248 0.1632
Use of insecticides (1= yes 0.other) δ9 1.2353 0.4979 0.0131**
Use of weedicide (1= yes 0.other) δ10 -1.0495 0.5906 0.0756*
Success of field level water mgt δ11 -0.0115 0.0056 0.0396**
sigmaSq 2 2 2( )v u 2 0.7860 0.2756 0.0043***
gamma27
2 2( / )u 0.8524 0.0516 0.0000***
Notes: significance at * 10%, **5%, ***1%.
27
If γ = 1, it implies that all deviations from the frontier is due to technical inefficiency. But this result
is only an approximation of the contribution of inefficiency to total variance of ui is proportional but
not exactly equal to2
uσ (Coelli et al., 1998).
89
Chapter 5: Efficient water usage in village irrigation systems for rice farming 89
The production frontier was estimated by a simple three step procedure
discussed in Section 3.2.4, in order to maintain the theoretical consistency especially
imposing monotonicity conditions. The sources file of the estimation and results of
the three step estimation are shown in Appendix D and Tables D1. The objective of
using the three step procedure to estimate the model is to impose monotonicity
conditions on the production function in order to maintain theoretical consistency.
The β, γ, ζ2, and δ coefficients are defined in Sections 3.2.6 and 3.2.7
respectively. The monotonicity condition fulfilled only 59.1% while quasi-concavity
only 0.4% out of the total observations for all variables at the first step. The
coefficients of inputs are significant at the 5% level or lower. While the participation
rate for FOs activities, FO membership, use of weedicides and field level water
management practices had significant (10% or lower) positive influences on TE, all
other variables of the inefficiency model had no positive effects on TE. Farmers‟ age
and their education have no significant influences on TE. As discussed in Section
3.2.4, inconsistency of the estimated model misleads the inefficiency scores.
Therefore, the magnitudes of monotonicity and quasi-concavity conditions, fulfilled
in the initial and final steps, were estimated. The results are presented in Table 5.3.
Table 5.3
Performances of monotonicity and quasi-concavity
Variables Initial (%) Adjusted (%)
Water
Labour
Power
Irrigating time
Pesticides
Quasi-concavity
100
98
88.7
81.3
86.1
0.4
100
100
100
100
100
84.6
Water has satisfied the monotonicity conditions of the estimated model
whereas all other variables were violated. However, at the final estimation of the
three step procedure all variables were theoretically consistent. Quasi-concavity
condition of the final stochastic frontier model was satisfied only by 84.6%, however
it dramatically improved from the initial stage to the final estimation of the model.
Imposing monotonicity was mainly focused whereas quasi-concavity was not
considered a necessary condition for estimating a consistent translog frontier model
(Henningsen & Henning, 2009).
90
90 Chapter 5: Efficient water usage in village irrigation systems for rice farming
Step two is to obtain restricted β parameters by minimising the overall
deviation from the original production set taking into consideration the variance –
covariance matrix. The minimum distance estimations are shown in Table 5.4. Many
coefficients did not change considerably (See column “difference” in Table
5.4).However, many changes were less than one times the standard error of the first-
step estimation (See column “Diff/std. Error” in Table 5.4).
Table 5.4
Minimum distance estimation
Variables Coefficient Difference* Diff/Std.
Error
Adjusted
coefficient
Constant 0
0β 0.2918 -0.0164 -0.2667 0.2866
Ln water (Metre per hectare/ ) 0
1β 0.3227 -0.0119 -0.3400 0.3231
Ln labour (man days) 0
2β 0.1622 -0.0002 -0.0067 0.1624
Ln power (minutes) 0
3β 0.1339 0.0292 0.8022 0.1340
Ln irrigating time (minutes) 0
4β 0.0586 -0.0065 -0.2407 0.0587
Ln pesticides (ml) 0
5β 0.1147 -0.0047 -0.1567 0.1148
Ln water x Ln water 0
6β 0.1659 -0.0156 -0.2771 0.1661
Ln water x Ln labour 0
7β -0.0161 -0.0284 -0.6995 -0.0161
Ln water x Ln power 0
8β -0.0332 -0.0246 -0.7089 -0.0333
Ln water x Ln irrigating time 0
9β 0.0066 0.0007 0.0224 0.0066
Ln water x Ln pesticides 0
10β -0.0113 0.0041 0.1082 -0.0113
Ln labour x Ln labour 0
11β 0.0509 -0.0091 -0.1535 0.0510
Ln labour x Ln power 0
12β 0.0136 0.0672 1.3742 0.0136
Ln labour x Ln irrigating time 0
13β -0.0117 0.0018 0.0511 -0.0117
Ln labour x Ln pesticides 0
14β -0.0064 -0.0057 -0.1553 -0.0064
Ln power x Ln power 0
15β 0.0226 0.1219 2.6047 0.0226
Ln power x Ln irrigating time 0
16β 0.0163 0.0182 0.5141 0.0163
Ln power x Ln pesticides 0
17β -0.0050 -0.0748 -1.8794 -0.0050
Ln irrigating time x Ln irrigating time 0
18β -0.0196 -0.0558 -1.2026 -0.0196
Ln irrigating time x Ln pesticides 0
19β 0.0003 0.0066 0.1976 0.0003
Ln pesticides x Ln pesticides 0
20β 0.0474 0.0763 1.4589 0.0474
The restricted coefficients of the estimated model, after adjusting the
production frontier is shown in the last column (“Adjusted coefficients”) of Table
5.4. These coefficients are used to interpret the estimated stochastic frontier models.
91
Chapter 5: Efficient water usage in village irrigation systems for rice farming 91
The coefficients of the stochastic frontier represent output elasticities relating to the
inputs used. The estimate of output elasticity for rice production with respect to the
individual volume of water use for rice production was highly significant (at 1%
level). An increase of 10 % in water usage can increase the rice production by 3.2 %.
As shown in Table 5.3, monotonicity condition is satisfied by 100% at the final stage
of the stochastic frontier estimation, shown in Table 5.5.
Table 5.5
Final stochastic frontier model
Variables Estimates Std. Error Pr(>|z|)
Intercept -0.0055 0.0622 92.93%
lcFitted 1.0012 0.0455 0.00%
sigmaSq 2( ) 0.6445 0.2132 0.00%
gamma ( ) 0.7947 0.0737 0.00%
As a result of increased theoretical consistency of the observations, estimated
coefficients of the restricted model can differ from the unrestricted model. However,
in this estimation, the coefficient of the intercept (α0) is virtually zero and the
coefficient of the “frontier output” (α1) is virtually one. Therefore, the coefficients of
the adjusted and non-adjusted restricted production frontier are almost the same (See
“Coefficient and “Adjusted coefficient” in Table 5.4). The estimated total error
variance (ζ2) is 64% and the proportion of variance of technical inefficiency in the
total error variance (γ) is 79% (See Table 5.5).
The estimated results of the inefficiency model are shown in Table 5.6. In
inefficiency models, positive coefficients indicate that the corresponding variable has
a negative effect on the TE. Factors with negative signs have positive effects on TE.
The farmers‟ age had no significant effects on TE. However, the education level
positively relates to TE but is not a statistically significant variable even at 10%
level. There are five factors of the inefficiency model shown to decrease technical
inefficiency, (and increase efficiency) namely, farmers‟ education, participation in
common activities28
(collective actions29
), membership of the FO, use of weedicides
and success of individual field level water management.
28
Farmers have four collective responsibilities:
92
92 Chapter 5: Efficient water usage in village irrigation systems for rice farming
Table 5.6
Inefficiency model
Variables α Estimates Std. Error Pr(>|z|)
Age of farmer (yrs) δ1 0.0047 0.0061 0.4469
Farmer‟s education level (yrs of schooling) δ2 -0.0060 0.0243 0.8039
Participation rate for FO activities (%) δ3 -0.0121 0.0062 0.0531*
FO membership (1= yes, 0, otherwise) δ4 -0.5929 0.2652 0.0254**
Field location (1= head-end, 0,otherwise) δ5 0.3410 0.2503 0.1731
Field location (1= middle, 0, otherwise) δ6 0.5976 0.2881 0.0381**
Water sharing issues (1= Yes, 0 = no) δ7 0.9149 0.3370 0.0066***
Land ownership (1= own; 0, other) δ8 0.4594 0.2914 0.1149
Use of insecticides (1= yes 0.other) δ9 1.0500 0.4366 0.0162**
Use of weedicides (1= yes 0.other) δ10 -0.8458 0.4769 0.0761*
Success of field level water mgt δ11 -0.0096 0.0045 0.0321**
The membership of FO is highly significant amongst the variables having a
positive impact on TE. On the other hand, farmers‟ age, individual locations of the
paddy fields of the command area, water sharing issues, landownership and use of
insecticides have negative effects on TE. As hypothesised, the water sharing issue is
the most significant (at 1% level) and most influential factor of technical
inefficiency. The frequency distribution of TE is shown in Figure 5.1.
(i) Each farmer shall carry out a certain amount of repair work on the bund annually,
proportional to the amount of his land holdings.
(ii) Each farmer shall maintain in good condition any irrigation ditches going past or through
the land where he works.
(iii) Each farmer shall build and maintain a portion of the main field fence and opposite ends
of any strips where he works.
(iv) Each farmer shall take his turn to sit up all night in one of the field huts to ward off wild
animals, liable to attack the field during harvest time.
29
The benefits of collective actions are threefold: increased profitability, greater social equity and
reduced social conflict. Increased profitability was due to greater water reliability, guarding against
wild animals and shared fish catch. Leach concluded that ownership of purana (old) lands was very
profitable in Pul Eliya although no figures were given (Leach, 1961)
93
Chapter 5: Efficient water usage in village irrigation systems for rice farming 93
Figure 5.1. Frequency distribution of TE estimates
TE distribution is skewed to the left. Mean TE of rice farming in village
irrigations is 0.73. Mean TE estimated for stochastic production frontier for Asia is
0.72 (Bravo-Ureta et al., 2007). The range of TE varies from 0.08 to 0.92. Therefore,
rice production in village irrigation can be increased by 27% with the present state of
technology if technical inefficiency is removed completely.
5.6 DISCUSSION
The period from 1000 to 1300 A.D. was seen as the flowering of Sri Lankan
culture in the low rainfall region, based on systematically organised irrigation
systems focused on increased rice production (Pain, 1986). Since then, the main
objectives of the irrigation investments of successive ruling parties were to achieve
EE in the use of scarce irrigation water both at the point in time (static) and over a
period of time (dynamic) and to attain equitable distribution of benefits to water
users (Sampath, 1992). However, poor irrigation performance can have a negative
effect on yield (Marikar et al., 1992). Therefore, it is important to improve water user
efficiency in irrigated agriculture to increase the output per unit of water used.
The estimated stochastic frontier production function model presents a number
of important features in relation to the performance of rice production and their
specific characteristics in the VISs in Sri Lanka. All estimated first-order adjusted
coefficients in the translog model (See Table 5.4) fall between zero and one,
94
94 Chapter 5: Efficient water usage in village irrigation systems for rice farming
satisfying monotonicity conditions (Sauer et al., 2006; Villano & Fleming, 2006;
Henningsen & Henning, 2009). The function coefficient is approximately 0.79
indicating all marginal products are positive, diminishing the mean of inputs (Binam
et al., 2004). The estimated first-order coefficients for all inputs (water, labour,
mechanical power, irrigating time and pesticide) are significant at the 5% level.
Furthermore, the γ-parameter relating to the variance of technical inefficiency in the
model is estimated to be 0.79. This result indicates that TE effects are a significant
component of the total variability of rice output in the VISs.
The estimated results of this study show that a 10% increase in individual
volume of water use increases output of rice by 3.2 % (See Table 5.4). This was
estimated irrespective of field location in the command area. Measuring this amount
of water used by individual farmers through the number of irrigation30
will lead to
poor estimates due to the number of irrigations not considering paddy fields located
at the tail end receiving less water due to conveyance losses (Sharma, et al., 2001).
Sharma et al. (2001) estimated irrigation requirements for two irrigation systems in
Nepal, using number of irrigation releases as a measure of water use. According to
their results, output elasticity with respect to number of irrigation is 0.078. These
results are not comparable to the estimated results in this study due to different
measures of irrigation estimation. Irrigation is also measured on the ratio of irrigated
area to the total cultivated area for grain (rice, wheat and maize) production in China
(Yao & Liu, 1998). Yao & Liu (1998) found that irrigation water was significant (at
0.01% level) with 0.101 output elasticity. In order to calculate the variation of the
individual volume of water used by each farmer, the irrigating time for the individual
paddy field was added to the model. Output elasticity with respect to the irrigating
time was 0.0587 with 5% significance level. Irrigating time can be controlled by
various environmental characteristics (Daleus et al., 1988; De Silva, et al., 2007)
such as soil types, evaporation and some farmer specific factors (i.e., farmers‟ water
management practices). However, results of the previous study in the Kurunagala
district found that about 32% of farmers received inadequate water during the main
season and 25% of farmers perceived that yield reduction was due to water shortage.
The estimated results with respect to individual volume of water and irrigating time
30
Number of times water is released to the paddy field from the reservoir. Water is released once a
week in village irrigations (See Section 7.5 for more detail).
95
Chapter 5: Efficient water usage in village irrigation systems for rice farming 95
need to be discussed further with locational water sharing issues. These two variables
were not included in the model developed by Aheeyar et al. (2005) in the same study
area. Nevertheless, Aheeyar et al. (2005) estimated labour and power in rice
production. Labour can be substituted with animal power for land preparation,
threshing the paddy and transport. In some cases, lack of labour caused some plots to
be left fallow (Ulluwishewa, 1991). At present, there is a growing trend to substitute
labour with mechanical power in rice farming mainly in land preparation, threshing
the paddy, cleaning the paddy and transportation. The output elasticity of labour
(significant at 10% level) in rice farming in VISs in the Kurunagala district was
0.1422 (Aheeyar et al., 2005). The model estimated shows that output elasticity is
positive with respect to labour although the volume of water is not significantly
different to the previous estimates of Aheeyar et al. (2005). The estimated output
elasticity with respect to mechanical power is 0.134 (significant at 1% level) and has
a positive effect on rice production (Tadesse & Krisnamoorthy, 1997). The highest
labour cost (Rs. 3843.20 per ha) of rice farming for VISs was reported in the
Kurunagala district in 2005, but comparatively, the cost of mechanical power per ha
(Rs. 1489.20) was less than the labour cost (Aheeyar et al., 2005). Another reason for
mechanical power substitution is a drop in traditional forms of labour involved in
rice farming. The village farmers used two types of collective labour sharing, attam31
and kaiya32
. Labour sharing (collective actions) is gradually disappearing due to
changes of biophysical, environmental and socio-economic backgrounds
(Abeyaratne & Perera, 1984; Aheeyar, 2001) as a result of agricultural
modernisation33
(Ulluwishewa, 1991; Thilakaratne et al., 1997). Furthermore, this
31
attam is the traditional term for exchange labour. On the village level, when work is to be done,
neighbors communicate and mutually agree to do the work. For example, when farmer A harvests his
paddy land, the five neighboring paddy land owners provide labour. Then, when one of the five
neighbors harvest their paddy land, it is farmer A‟s responsibility to participate. Likewise, farmer A
has to participate in harvesting the other four farmers‟ land as well. If any of the farmers fail to
provide labour as mutually agreed, he must hire a person to participate on behalf of him.
32
The kaiya labour is similar to attam labour however it is a completely voluntary. For example,
farmer A formally invites his fellow farmers to participate in harvesting his paddy land. Whoever is
willing to help farmer A, harvests the paddy land on an agreed day. All the participants receive a
specially prepared meal which is also called “kaiya” from farmer A on the day of harvesting.
33
Decline of the traditional common property management system has been caused by state sponsored
efforts to promote land expansion in the reservoir command area to allow for population expansion.
96
96 Chapter 5: Efficient water usage in village irrigation systems for rice farming
situation illustrates, as hypothesised by Arthur Lewis (1954), that zero MVP of
agricultural labour does not exist and there is no surplus labour in the informal
sector. However, labour is essential for some activities of rice farming, such as
seeding (broadcasting and transplanting), spraying pesticides and other chemicals,
and preparing bund coverings for small plots of land called liyadda in order to
maintain water levels and watering.
Aheeyar et al. (2005) found that using agro-chemicals had a negative impact on
rice production for VISs. When all farmers start work in their fields at the same time,
integrated pest management is possible. In this study the use of pesticides had
positive effects on the output.
Factors influencing technical efficiency of rice farming for village irrigation
systems
Mean TE estimated for stochastic production frontier for agricultural
production in developing countries is 0.72 (Bravo-Ureta et al., 2007). Thaim et al.,
(2001) estimated the mean TE of rice farming for most Asian countries to be68%.
Furthermore, mean TE estimated for agricultural production in 167 countries
including developed and developing countries reported 77.3% for all stochastic
frontier models (Bravo-Ureta et al., 2007). These results suggest that it is possible to
increase agricultural output without additional inputs and existing technology.
The estimated coefficients in the inefficiency model were used to explain the
factors affecting efficiency. The factors considered in the estimation of TE of farmers
and their estimated coefficients (inefficiency model) are shown in Table 5.6. In total,
eleven variables were included in the model. Most of them (55%) were found to
enhance technical inefficiency.
The coefficient of farmers‟ education is expected to have a negative sign
(decreases inefficiency or increases efficiency) because the educational attainment of
farm manager is a proxy for human capital (Kalirajan, 1981; Coelli & Battese, 1996;
Aheeyar et al., 2005; Villano & Fleming, 2006), simply because general education in
school may not be directly relevant to farming (particularly specific type of crop) or
training for agriculture (Lindara et al., 2006). In this study, the coefficient was
negative but not significant, suggesting that productivity is not related to farmers‟
level of schooling.
97
Chapter 5: Efficient water usage in village irrigation systems for rice farming 97
Two more important findings were made in this study. Having membership of
a respective FO and the percentage of participation in FO activities positively
affected TE (Thiruchelvam, 2002) and both were significant at the 5% and 10%
level respectively. The positive influence of efficiency of these two variables
indicates the importance of cooperative or collective action in rice farming village
irrigation through the cooperative water management system. The Lerma-Chapala
Basin in Mexico, proved that small-scale farmer managed water harvesting irrigation
systems are more productive in terms of agricultural productivity than the
government administrated irrigation systems (Scott & Ochoa, 2001). Farmers are
more likely to be positive about water allocation and management systems designed
by themselves than by others (Bardhan, 2000). Furthermore, grass root level
collective action of farmers is considered a potential alternative to improve farmer
community‟s welfare (Ayer, 1997).
According to the power vested by the Agrarian Development Act (2000),
people directly or indirectly involved in agriculture or agricultural related activities
can be members of the FO for particular VISs. The FO is the key village level
institution responsible for management and decision-making in respect to cultivation
and irrigation in VISs. The Kanna meeting is the major meeting that discusses
cultivation and water management issues. At this meeting, agricultural activities are
planned and collective decisions are made that cannot be changed by individuals
until the end of the cultivation season, unless there are special circumstances.
Therefore, participation in these activities contributes immensely to the increase of
TE. Generally, the average participation in FO activities in these two districts is 38%
(relatively low) due to the lack of accountability and transparency of the functions of
the FOs (Thiruchelvam, 2010). Other factors responsible for lack of participation in
irrigation management are farmers‟ attitudes toward participation in irrigation
management and extension services of water management, family size, the problem
perception, dependence level of water source and farmers‟ education (Khalkheili &
Zamani, 2009).
Use of weedicides is another factor which increases TE. Another method of
weed control used by village farmers is to maintain ample coverage of water (Daleus
98
98 Chapter 5: Efficient water usage in village irrigation systems for rice farming
et al., 1989). However, at present, this mechanism cannot be practised due to limited
supply of water34
.
There are five factors in the model that are significant and positively influence
technical inefficiency (i.e., reduce efficiencies). Farmers‟ age is not significantly
related to TE. Ownership of paddy fields belongs to the elderly farmers in the VISs
as ownerships were transferred to successive generations through inheritance
(Codrington, 1938). However, at the operational level in the field, those who are
working in the paddy fields are farmers younger than 50 years old. Therefore,
influence of age is not statistically significant although positively relates to technical
inefficiency. These results can be compared to the previous two researches
conducted in the Kurunagala (Aheeyar et al., 2005) and Anuradhapura
(Thiruchelvam, 2003) districts. In both cases, farmers‟ age was positive with
technical inefficiency and the average age of the farmers was over 55. The other
study conducted in rice production for two blocks of main irrigation in the
Anuradhapura district by Thiruchelvam (2002) reported similar results. Similarly, in
the case of central Luzon in the Philippines, farmers‟ age has caused an increase in
technical inefficiency, because older farmers were not willing to adopt better
practices (Villano & Fleming, 2006).
Water sharing from the head-end to the tail-end is another factor influencing
TE. The results from this study suggest the tail-end farmers are the most efficient and
the middle farmers are the least efficient. A case study conducted in the
Anuradhapura district in Sri Lanka by Daleus et al. (1988) found that there was
decreasing yield with increasing distance from the reservoir. However, they have
assumed that yield variation was attributed to management problems. Therefore, it
cannot be concluded that water allocation issues cause yield variation. Sharma et al.
(2001) found that there was a variation of TE from head end to tail-end in two
irrigation systems called Pithuwa (farmer managed irrigation system) and Khageri
(government managed irrigation system). From these two studies, it can be
concluded that intra-sectoral water allocation issues can be anticipated to affect both
yield and TE variation from the HEFs to the TEFs of the command area. In the case
of main irrigation, there is evidence of TE variation from the head-end to the tail-end
34
Dharmasena (1994) found that inadequacy of water storage of VIS was a consequence of
sedimentation due to upland cultivation.
99
Chapter 5: Efficient water usage in village irrigation systems for rice farming 99
(Ekanayake & Jayasoriya, 1987). This situation would always be acceptable when
water supply from the reservoir to the paddy fields is through a single outer canal. In
order to minimise this issue, in the traditional VISs, there were three outlet canals to
the three sections of the command area. Due to the expansion of the area of land,
farmers had to remove these original settings from most VISs.
As previously mentioned in this chapter, ownership of the paddy land belongs
to the old farmers of the village. They do not transfer their land right to their
subordinates (most probably to the owners‟ family) until their death. However, most
landowners are not active farmers. Sometimes, there are absentee landowners. As a
result, lands are cultivated by others who do not have land ownership. In this sample,
only 64% of farmers had land ownership. Tenant farmers were more efficient than
landowners, because they have to pay an agreed amount either in cash or share of the
harvest, to the landowner as land rent. Therefore, he has to efficiently manage his
land to achieve maximum yield. Similar results were found in the Anuradhapura
district where land ownership was negatively related to TE (Thiruchelvam, 2003).
Another reason for negative TE is land ownership of rice farming in VISs is that
traditional farmers can have more than one plot of land for one command area or
smaller plots of land from other nearby VISs. Therefore, lack of motivation from
these farmers is not favourable for TE. Thiruchelvam (2003) revealed that farmers
who cultivated paddy land during the main irrigation had 90% of paddy land
ownership. Thiruchelvam‟s (2003) estimates showed paddy land ownership caused a
significant (at 10% level) decrease in technical inefficiency.
5.7 CHAPTER SUMMARY
The empirical estimates of TE in rice farming for VISs were proven to be useful.
With respect to water resource allocation, it was important for policy makers to know
how far agricultural production could be expected to increase its output by simply
increasing its TE without altering further resources, given the technology involved.
This chapter discussed the results of estimated TE measures derived from 460
sample rice farmers in the Kurangala district in Sri Lanka. The translog stochastic
production frontier estimate followed a simple three step procedure imposing
theoretical consistency. The output elasticities with respect to all inputs were positive
and statistically significant. Technical inefficiency was represented by 79% of the
total variance of the model. The estimated mean TE in this study was 73%, although
100
100 Chapter 5: Efficient water usage in village irrigation systems for rice farming
the distribution was skewed. Most farmers had high TE whereas some had very low
scores. This indicated that output can improve by 28% without altering the inputs
and without changing existing technology used in rice farming.
Two ideas were proposed to improve TE. Firstly, formalising transferability
of land ownership and therefore water user rights. Secondly, enhancing institutional
capacities of FOs to solve intra-sectoral water sharing issues. Promoting multiple
uses of reservoir water for subsistence and commercially important economic
activities will further increase total productivity of water for VISs.
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 101
Chapter 6: Efficient water usage in village
irrigation systems for culture-
based fisheries production
6.1 INTRODUCTION
To make CBF production a success a broad set of biological, economic, social,
and institutional management aspects should be addressed (Lorenzen, 2008). This
chapter discusses the TE and other factors influencing inefficient use of village
reservoir water for CBF production in an economic, social, and institutional context.
CBF in VISs in Sri Lanka is different from most other Asian aquaculture systems
with respect to the use of inputs in the production. The input variables of the
estimated production function are limited to only three variables: water, labour and
total fingerlings stocked. Nine characteristics of the input variables are specified in
the estimated inefficiency model of the stochastic production frontier.
The chapter then provides a general introduction of CBF production and
further extends this to a discussion of CBF development in Sri Lanka. The literature
review in Section 5.3 identified the knowledge gap within the existing CBF research.
It shows there are no previous studies on estimation of TE of CBF in Sri Lanka and
nor any tested species growth performance on TE in the literature. In Sections 5.4
and 5.5 the estimated empirical models are demonstrated and the results presented.
Section 5.6 discusses the results and Section 5.7 suggests possible policy
implications for the improvement of TE.
6.2 CBF PRODUCTION
Asia is the epicentre of aquaculture production and is the highest consumer of
freshwater fish. Global aquaculture is growing rapidly, with Asia contributing 88.9%
of the world‟s freshwater fish supply, with China being the largest producer. The
remainder is produced in Africa (1.8%), America (4.6%), Europe (4.4%) and
Oceania (0.3%) (Bostock et al., 2010).
The aquaculture production is operated in five types of aquaculture systems in
the world. They are: freshwater ponds and tanks, freshwater cages, coastal ponds and
102
102 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
tanks and coastal cage and farms, marine molluscs and aquatic plants. In total, 56%
of the value of aquaculture production in the world is produced in freshwater ponds
and tanks (Bostock et al., 2010). Freshwater fish production is dominated by carp
species with more than 100 countries in the world producing more than 1 million
tonnes of common carp species in 2008 (Bostock et al., 2010). Particularly, CBF in
Sri Lanka is based on a combination of Chinese and Indian major carp species. They
are rohu, mrigal, common carp, bighead carp, and silver carp, and the exotic cichlid
species, [Oreochromis niloticus and Oreochromis mossambicus] (De Silva, 2003).
Fish is the main source of animal protein for rural communities. Freshwater
fish production accounts for between 15% to 53% of the total animal protein intake
in most Asian countries such as Bangladesh, China, India, Indonesia, the Philippines,
Thailand, and Vietnam (Dey et al., 2005). The per capita fish consumption and the
type of fish species consumed are generally determined by the economic status of
households. Furthermore, per capita fish consumption is comparatively higher in
rural areas than in urban areas (Dey et al., 2005)35
. In the Asian region, the share of
fish expenditure to the total food expenditure is 11%, while 69% of the fish
consumers prefer fresh water fish, both high and low valued. Only 29% of the total
population favours marine fish. In general, the price elasticity for freshwater fish is
slightly higher (1.08) compared to that of marine fish (0.98).
The favourable market demand and higher prices tend to motivate CBF
production in Sri Lanka (Amarasinghe & Nguyen, 2009). Therefore, market demand
for CBF production is no longer a valid constraint on increasing CBF production. In
Sri Lanka, CBF takes place in the existing, operational, VISs of reservoir more than
10,000 in number. These reservoirs are highly productive for CBF (De Silva, et al.,
2003; Jayasinghe et al., 2005; 2005a; Amarasinghe & Nguyen, 2009). The CBF
development activities in VISs in the 1980s were not successful due to biological
productivity-related problems such as the non-availability of effective means to
35
Dey et al. (2005) conducted a survey of 5931 households in selected Asian countries and
summarised their findings on fish consumption as follows:
1. Fish consumption depends on income classes and the location
2. Low income groups expend more for fish of the food budget while high income groups
consume more fish
3. Rural fish consumption is higher than urban fish consumption
4. Fish consumption of fish producers is higher than that of the non-producers
5. Low-value fresh water fish is favoured by poor consumers.
103
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 103
select suitable reservoirs and lack of a guaranteed supply of fingerlings for stocking
(De Silva, 2003). Furthermore, weak institutional linkages, lack of legislation and
poorly planned social mobilisation procedures also resulted in CBF activities being
unsustainable. Although some of these constraints, especially at the grassroots level,
have been dealt with through the concerted efforts of fisheries biologists, barriers at
the institutional level (i.e., water allocations and water user rights) and infrastructure
(especially, communication and accurate information) still exist.
Development of CBF production in Sri Lanka
Sri Lanka has traditionally developed various management practices for
sustainable utilisation of fisheries resources in VISs (Ulluwishewa, 1995). Mendis
(1965) was the first to identify the possibility of CBF development in small village
reservoirs in Sri Lanka. In 1963, eight reservoirs in Polonnaruwa administrative
district were stocked with juvenile Chanos chanos and O. mossambicus (Anon.,
1964). Indrasena (1965) and Fernando & Ellepola (1969) as early as the 1960s
conducted CBF trials in several village reservoirs. Initial funding by Food and
Agricultural Organisation (FAO) and United Nations Development Programme
(UNDP) (Chakrabarty & Samaranayake, 1983) and subsequent financing by the
Asian Development Bank (ADB) facilitated the development of CBF in seasonal
reservoirs in the 1980s (Thayaparan, 1982). Chandrasoma & Kumarasiri (1986) have
reported that in 15 village reservoirs, fish output ranged from 220 to 2300 kg ha-1
(mean 892 kg ha-1
) within a single culture cycle. Therefore, village reservoirs have a
large potential for the development of CBF (Mendis, 1977; De Silva, 2003).
CBF is essentially a fisheries enhancement strategy (Lorenzen, 2008). CBF
developments in village reservoirs involve stocking of fingerlings after the inter-
monsoonal rainy season in December/January and harvesting the stocked fish during
the dry season from August-September (See Table 6.1). However, until recently, CBF
has not been well received by reservoir-based agricultural systems in spite of its
potential for increased fish production and enhancing rural livelihood (De Silva,
2003).
104
104 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
Table 6.1
Incorporation of agricultural and CBF activities in village reservoirs
The development of inland fisheries and aquaculture has been given high
priority as fish is a cheap source of animal protein for rural low income communities.
CBF generates income and is a source of additional employment to rural farmers.
Since the early 1980s, the emphasis on the development of CBF in village reservoirs
has increased. For instance, the development of CBF has been included in national
development plans (Thayaparan, 1982; Chakrabarty & Samaranayake, 1983). It has
been identified that CBF has had a significant influence on local institutions and rural
livelihoods in recent years (Amarasinghe & Nguyen, 2009). However, the main
drawback for the development of reservoir-based CBF production is likely to be
issues relating to reservoir water allocation.
Members of FOs, as discussed in Chapter 2, have well defined property rights for
reservoir water use for agriculture by the power vested in the Agrarian Development
Act of 2000. However, user rights of water for CBF are not well-defined under any
of the available legislation. Even in the 1998 (Act 53) and subsequent amendment in
2006 (Act No. 145) legislation, which established NAQDA, there are no sufficient
legal provisions to facilitate CBF or aquaculture development in VISs of the country.
The CBF activities practised in Sri Lanka are considerably different to those
practised in some other Asian countries. In Sri Lanka, CBF is somewhat similar to
extensive aquaculture in reservoirs, which are man-made water bodies where
ecological conditions are different from natural inland water bodies. Some countries
have used natural inland water bodies for CBF and aquaculture (e.g., Oxbow lakes in
105
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 105
Bangladesh, Taal Lake in the Philippines). In Sri Lanka, the main role of fish
breeding is fulfilled by government-breeding centres, which undertake fingerling
rearing and distribution. Nevertheless, in some other countries, private dealers
dominate this sector. Most countries supply either individual family labour or hired
labour for aquaculture, but CBF production in Sri Lanka is mainly operated with
group labour that is based on collective agreements. Consequently, in Sri Lanka,
CBF is organised under FOs rather than individuals. In some countries there is a
well-defined system of property rights for aquaculture farmers (e.g., Nigeria)
whereas Sri Lankan CBF farmers use common pool water resources for the CBF
production. Additionally, no CBF depends on supplementary fish feeding (De Silva,
2003) that aquaculture systems in other parts of the world heavily use. Pond size is
adjustable in some countries; however, reservoir size in Sri Lanka is fixed. The only
possibility to increase the level of reservoir water is to increase the level of water
user efficiency of other uses such as rice farming. Where the same water source is
used for multiple activities, water allocation among them becomes important.
6.3 LITERATURE REVIEW
Inland water availability is one of the factors necessary for the development of
inland fresh water aquaculture. Also, as Bostock, et al. (2010) points out, it is
necessary to put in place a medium term strategy to increase output to create new
environments, intensifying and improving efficiency. CBF is considered to have a
very high potential for the enhancement of aquaculture production that is produced in
competition with other water uses in inland areas. Therefore, in this context the
relevant literature on land-based aquaculture is examined.
Aquaculture systems in Asia operate either as an extensive farm, intensive farm
or mixed system of stocked Chinese and Indian carp species. Most of the intensive
aquaculture systems are more technically efficient than the extensive farming
systems in some South Asian and South East Asian countries, and China (Sharma &
Leung, 2000). However, differences in efficiency levels are based on the various
farm-specific and country-specific factors.
The survey conducted by Sharma & Leung (2000) in the four Asian countries
namely Nepal, India, Bangladesh and Pakistan revealed that the adoption of
recommended fish species, water and feed management would be critical for best
106
106 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
performance of fish production. In addition to intensification (Linuma et al., 1999;
Dey, et al., 2005; Kareem et al., 2009), integrated rice-fish culture (Saikia & Das,
2008) and efficient resource allocation of resources (Alam et al., 2008;
Phengphaengsy & Okudaraia, 2008) have been found as other important means of
increasing TE in some countries (e.g., Malaysia, Vietnam, Thailand, the Philippines,
China and Nigeria). Furthermore, TE and the resulting increased productivity is
constrained by human capital (education and training), basic infrastructure (roads),
easy access to fingerlings, and security of property rights (Dey et al., 2005).
The authority of supplying fish fingerlings from government breeding centres
certifies the quality of fingerlings. If there is an insufficient quantity of fingerlings
produced by the government breeding centres to meet the fingerling demand, other
sources of fingerlings such as rural farming systems are approached. Middlemen
(private dealers) who intervene by providing other sources of fingerlings cannot
guarantee the quality of the seed that supplies fish fingerlings in different sizes and in
a different species mix (Singh et al., 2009). In West Tripura District, India, the size
of pond, labour and other inputs like lime have been found to be potential factors
increasing fish production (Singh et al., 2009).
Kareem et al. (2009) examined the TE of natural and artificial pond culture
systems in Nigeria using the stochastic production analysis. They showed that TE of
earthen ponds was slightly higher (89%) than concrete ponds (88%). They found that
pond area, quantity of inputs (lime, labour), the quality of fingerlings and other
material were significant factors influencing TE of both concrete and earthen ponds.
According to these researchers, intensification of aquaculture practices increased
technical and price efficiency. The major suppliers of fingerlings in this study were
private dealers (79%).
Resource allocation in carp production in India has been estimated by using a
stochastic frontier production function (Sharma & Leung, 2000). According to this
study, a higher inefficient allocation of resources occurred in extensive carp farming
than in semi-intensive carp farming. Sharma & Leung (2000) concluded that
optimum resource allocation could increase the production of semi-intensive farms
from 3.4 Mt ha -1
to 4.1 Mt ha -1
while the production of extensive carp farms could
be increased from 1.3 Mt ha -1
to 1.9 Mt ha -1
. They have shown that stocking ponds
107
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 107
with certain fish species, improving water and feed management and improving
monitoring practices could enhance efficiency.
Saikia and Das (2008) found that integrated rice-fish farmers produced
approximately 500 kg per hectare per season without adding any supplementary feed
to the fish stock in their rice fields. This has resulted in a 65.8% increase in economic
returns per annum. Rice-fish integrated field systems are successful where the use of
pesticides and fertiliser is minimal. This system has totally disappeared in Sri Lanka
due to the heavy use of chemicals in rice farming since the 1970s (Fernando, 1993).
Therefore, CBF is the most appropriate aquaculture system for inland waters. In
addition to the above-mentioned factors that increase efficiency in production, there
are contributory factors that result in inefficiency. In the next section, the literatures
related to such contributory factors are examined.
It has been found that homogeneity of high level of education (at least
secondary schooling) and leadership qualities of small farming communities have a
positive influence on their attitudes towards CBF development (Kularatne et al.,
2009). Larger groups are less likely to contribute to collective action than smaller
groups (Oliver & Marwell, 1988).
An investigation of the fish farming industry in Nigeria (Kareem et al., 2008)
found that there was a relationship among factors of efficient allocation of resources,
level of experience and formal level of education of farmers. The estimated
inefficiency model showed that „experience‟ of the farmer was significant at 1%
level indicating that fish farming experience was an important factor in increasing
efficiency.
Similarly, Alam et al. (2008) studied efficient resource allocation in a prawn-
carp poly-culture system in Bangladesh. DEA was employed to estimate efficiency.
The results showed that 50% of prawn-carp farms were at full efficiency level with
only 9% cost efficient. They also found that labour, fingerlings and feed were
inefficiently allocated. They recommended adjustments in actual input allocation in
order to increase the level of efficiency.
6.4 EMPIRICAL MODEL
The main objective of this chapter, as mentioned in the introduction, is to
investigate TE and factors influencing technical inefficiency of CBF in VISs in Sri
108
108 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
Lanka. Primary data were used for this purpose that was examined in Chapter 3. As
discussed in Chapter 3, Kurunagala and Anuradhapura districts in Sri Lanka were
selected since they have the highest number of reservoirs that are used for CBF
production. The two districts are adjacent districts and, as such, are homogeneous in
morphology, climate, vegetation and all other social and economic aspects. Multi-
stage cluster sampling method (Cochran, 1960) was used for sample selection. Each
stage represented the number of reservoirs, based on an administrative hierarchy
from national level to village level.
The group of fish farmers of each reservoir engaged in CBF production was
considered as the sample unit of the CBF survey. CBF production data were
collected from several different culture cycles from 2006 to 2009. In two districts,
there were a total of 334 reservoirs (165 reservoirs in Kurunegala and 169 in
Anuradhapura) where CBF activities had been carried out during the three respective
culture cycles. The CBF farmer survey was organised as a group discussion. During
the interview, Officials (president, secretary and treasury) and the few members of
the FO essentially represented the group. ARPAs worked as enumerators of this
survey. All ARPAs were communicating over the telephone during the survey for
any clarification of the survey. The CBF farmer survey was completed within 4
months from December 2009 to March 2010.
Empirical models
The general translog functional form (Christensen et al., 1973) which is used
to estimate a production frontier can be expressed as:
3 3 31ln ln ln ln -
, 0 , , , , , , ,21 1 1
Y x x x v ui t i i k i k i k i l i t i t
i i k (6.1)
where Y is the CBF output of reservoir i for a culture cycle (period t) and i,k,x are
the inputs (k, l) used in the production process. riv and riu are as previously defined
in Equation 3.4. The explanatory variables discussed in Section 4.7 are as follows:
1rx = Water (individual share of water used by the i-th
reservoir for CBF is
estimated as 0.375 out of the total reservoir capacity measured by metres ha.).
2rx
= Labour (man days for a culture cycle)
109
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 109
3rx = Total number of fish fingerlings seeded
i are parameters to be estimated.
Following Battese and Coelli (1995), the mean of farmers‟ specific TE (Ui) is
defined as:
8
0
1
i j ij
j
u Z (6.2)
where: Z1 = Group stability for solving water disputes (1= yes, 0, otherwise)
Z2 = Time spent for meeting officials (hours)
Z3 = Risk on rain water adequacy/risk (Dummy; 1=yes, 0, otherwise)
Z4 = Subsidised fingerling supply (Dummy; 1=yes, 0, otherwise)
Z5 = Number of cattle and buffalos grazing in the catchment
Z6 = Slow growing fish fingerlings (Dummy, 1=yes, 0, otherwise)36
Z7 = Fast growing fish fingerlings (Dummy, 1=yes, 0, otherwise)
Z8 = Number of months of water used for other uses
6.5 RESULTS
The summary statistics of the input and output variables together with reservoir
and farmer-specific variables included in the technical inefficiency model are
presented in Table 6.2. The detailed description and the calculation methods of
individual volumes of water used for CBF production were discussed in Section 4.7.
In Sri Lanka, a limited number of inputs are used in CBF activities compared with
other Asian countries (De Silva, 2003). This is because CBF activities are conducted
in existing water bodies and do not utilise supplementary feeding. The labour used
for CBF production was estimated as the number of man-days actively involved in
CBF related activities in one culture cycle. All activities of CBF production were
undertaken as a group. Stocking of fish fingerlings, protecting the fish harvest from
theft and the harvesting were identified as the three major labour intensive factors of
CBF production.
It can be noted that the VISs dry-up during some months of the year. Hence
they do not harbour rich indigenous fish communities (Amarasinghe, 2008).
36
An individual reservoir may have both fast and slow growing species of the same time. That is it is
not an “either-or” situation.
110
110 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
Therefore, two types of fish species are used to stock reservoirs. They are Indian carp
and Chinese carp species. Common carp (Cyprinus carpio), bighead carp
(Aristichtthya nobilis) and catla (Catla catla) are considered as fast growing fish
species where as mrigal (Cirrhinus mrigala), rohu (Labeo rohita), Nile tilapia
(Oreochromis niloticus) and the other species are considered as slow growing species
(Wijenayake et al., 2005). Fish farmers prefer stocking fast growing fish species.
Total fish fingerlings stocked were categorised into two, based on their growth rate.
This was included in the inefficiency model as dummy variables (See footnote 36).
There are two ways to express the stocking density of fingerlings. That is, as the
number (i.e., numbers or /hectares) and as the weight (i.e., kg/ha) of fingerlings (De
Silva et al., 2007). This study used the first method, and the stocking density varied
from 58 to 20,000 fingerlings per ha in the sampled reservoirs.
Table 6.2
Summary statistics of variables involved in the SFM for CBF production
Variables Mean Std. Dev Minimum Maximum
Output (kg) 2715.48 3739.899 18 20000
Individual volume of water 2.03913 1.8876 0.074009 9.62116
Labour (man days) 30 38 2 164
Total fish fingerlings 13165.04 11806.69 1000 91500
Ln water 0.24826 1 -2.6036 2.2640
Ln water *Ln water 1.14874 1.3 0.0000 6.7786
Group stability 0.4154 0.5 0 1
Time spend to meet officials 17.5262 19.8 0 96
Rain water risk for CBF 0.4369 0.5 0 1
Subsidised culture cycle 0.6646 0.5 0 1
Number of cattle and buffalos 185 232.8 0 1300
Slow growing fish fingerlings 0.5538 0.5 0 1
Fast growing fish fingerlings 0.9200 0.3 0 1
Number of months water use for other 5.2831 3.6354 0 12
Single-person aquaculture committees were found only in a few CBF activities
and the majority of reservoirs had SGFs for CBF activities (Amarasinghe & Nauyen,
2009). Therefore, the nature of the group‟s stability on solving water disputes can
have a considerable impact on TE. The government provides extension services for
111
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 111
agriculture. Farmers consult the extension officers of NAQDA and DAD regularly
when organising agricultural activities. It has been revealed that cost of information
is very low (i.e., transaction costs) of CBF production in Sri Lanka (Senaratne &
Karunanayake, 2006). It was anticipated that the time spent in consulting officials
has a positive effect on TE in CBF production. However, on average the survey data
show that farmers spent 17.5 hours to meet government officials, especially officers
of NAQDA.
Agricultural activities in Sri Lanka are highly subsidised. For instance, since
2005, the government has subsidised fertiliser for rice farming. The supply of
fingerlings is also subsidised in CBF production. Therefore, the impact of the
subsidies on TE for CBF production was also investigated.
Feeding is not undertaken but instead CBF relies on run-off containing
materials into the reservoirs. This is because animal husbandry practices in the
catchment areas have a positive impact on nutrient loading of the reservoirs (De
Silva et al., 2007). As introduced species into the reservoirs are mainly herbivorous
(Amarasinghe & Nauyen, 2009) this is investigated as a factor influencing TE in
CBF production. It has been shown that the number of animals (cattle and water
buffalos) living in the reservoir catchment has a positive relationship with the CBF
production (Rabbani et al., 2004; Jayasinghe & Amarasinghe, 2007). Phengphaengsy
& Okudaira (2008) have also shown that there was a positive relationship between
multiple uses of water and water productivity in VISs. Therefore, the number of
months of water use for other purposes was included in the model.
The CBF-water frontier production function was estimated following a simple
three steps procedure as discussed in Section 3.2.4 for ensure the theoretical
consistency. The results are presented in Table 6.3. Furthermore, the source file of
the estimation and results of the model are shown in Appendix E.
112
112 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
Table 6.3
Initial maximum likelihood estimates (unrestricted frontier estimation)
Variables β Estimates Std. Error Pr(>|z|)
Constant β0 1.2474 0.2417 0.0000***
Ln water β1 0.4514 0.0721 0.0000***
Ln labour β2 -0.0586 0.0858 0.4945
Ln no. of total fish fingerlings β3 0.2850 0.0973 0.0034**
Ln water x Ln water β4 0.3984 0.1260 0.0016**
Ln water x Ln labour β5 0.0401 0.0677 0.5539
Ln water x Ln totalf β6 -0.1969 0.0987 0.0461*
Ln labour x Ln labour β7 0.0812 0.1399 0.5616
Ln labour x Ln totalf β8 0.0050 0.1000 0.9599
Ln totalf x Ln totalf Β9 0.1186 0.1657 0.4743
Group stability on solving water disputes β10 -0.3684 0.3321 0.2672
Time spent to meet officials β11 0.0171 0.0069 0.0135**
Rain water risk for CBF β12 0.3530 0.3145 0.2617
Supply of subsidised fingerlings β13 0.7982 0.3295 0.0154**
Number of cattle and buffalos β14 -0.0011 0.0008 0.1677
Slow growing fish fingerlings β15 -0.1561 0.3333 0.6395
Fast growing fish fingerlings β16 0.4044 0.4739 0.3935
Number of months of water used for other β17 -0.0366 0.0426 0.3904
sigmaSq 2 2 2( )v u 2 2.6905 0.5199 0.0000***
gamma 2 2( / )u
0.7976 0.0761 0.0000***
Notes: significance at * 10%, **5%, ***1%.
The β, γ, and ζ2, are defined in Section 3.2.6, and the δ are defined in Section
3.2.7. At the first step of estimation, the monotonicity condition and quasi-concavity
are fulfilled only 22.2% and 2.2% out of the total observations for all input variables.
The coefficients of water and total fingerlings are significant at the 1% level and
labour is not a significant variable even at the 10% level. Group stability on solving
water disputes and number of cattle and buffalos grazing in the catchment have a
significant (10% level) positive influence on TE. Although, stocking of slow growth
fingerlings and number of months which water use for other uses are positively relate
with TE they are not significant variables. All the other variables of the inefficiency
model are significant (10% level or lower) but they have no positive effects on TE.
In Section 3.2.4, it was pointed out that inconsistency of the estimated model can
mislead the coefficients of inefficiency variables. Therefore, monotonicity and quasi-
concavity conditions that are fulfilled in initial and final steps of estimation were
examined. The results are presented in Table 6.4.
113
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 113
Table 6.4
Performances of monotonicity and quasi-concavity
Variables Initial (%) Adjusted (%)
Water
Labour
Total fingerlings
Quasiconcavity
84.3
25.8
94.2
2.2
100
100
100
92.9
None of the variables are monotonically increased at the initial stage of the
estimated production frontier for CBF production. However, at the final steps, all
variables were theoretically consistent. The quasi-concavity of the model was
fulfilled by 92.9% in the final stochastic frontier estimation. Therefore, the estimated
model is theoretically consistent. The model estimates for minimum distance at the
second steps to obtain restricted coefficients for the inputs variables. The results are
shown in Table 6.5.
Table 6.5
Minimum distance estimation
Variables 0β Coefficients Distance Diff/Std.Error Adjusted
coefficients
Constant 0
0β 1.4894 -0.242 -1.0012 1.5025
Ln water 0
1β 0.4470 0.0044 0.0610 0.4466
Ln labour 0
2β 0.0046 -0.0632 -0.7366 0.0046
Ln total fish fingerlings 0
3β 0.2656 0.0194 0.1994 0.2654
Ln water x Ln water 0
4β 0.1648 0.2336 1.8540 0.1647
Ln water x Ln labour 0
5β 0.0016 0.0385 0.5687 0.0016
Ln water x Ln totalf 0
6β -0.0909 -0.106 -1.0740 -0.0909
Ln labour x Ln labour 0
7β -0.0001 0.0813 0.5811 -0.0001
Ln labour x Ln totalf 0
8β 0.0004 0.0046 0.0460 0.0004
Ln totalf x Ln totalf 0
9β 0.0703 0.0483 0.2915 0.0703
The last column (“adjusted coefficients”) of Table 6.5 shows the restricted
coefficients of the estimated model after adjusting the production frontier. These
coefficients are used to interpret the estimated stochastic frontier models. The
coefficients of the stochastic frontier represent output elasticities relating to the
114
114 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
inputs used. An increase of 10% in water usage can increase the CBF production by
4.5% percent. As shown in the Table 6.4, monotonicity condition is satisfied by
100% at the final stage of the stochastic frontier estimation that is shown in Table
6.6. The coefficient of the intercept (α0) is virtually zero and “frontier output” (α1) is
virtually one. Hence, the coefficients of the adjusted and non-adjusted restricted
production frontier are the same (See “coefficient and “adjusted coefficient” in Table
5.5). The proportion of variance of technical in efficiency in the total error variance
(γ) is approximately 82% (See Table 6.6).
Table 6.6
Final stochastic frontier
Variables Estimates Std. Error Pr(>|z|)
Intercept 0.0143 0.2709 0.9580
lcFitted 0.9992 0.1208 0.0000
sigmaSq 2( )2 2 2( )v u
2.7172 0.4822 0.0000
gamma ( ) 2 2( / )u 0.8150 0.0642 0.0000
The imposition of the monotonicity property increased the total error variance
from 2.69 to 2.71. The proportion of the variance of technical inefficiency in the total
error variance has not changed considerably. Estimated results of the inefficiency
model are shown in Table 6.7.
Table 6.7
Inefficiency model
Variables α Estimates Std. Error Pr(>|z|)
Group stability on solving water disputes δ1 -0.3862 0.3249 0.2345
Time spend to meet officials δ2 0.0166 0.0066 0.0117
Rain water risk for CBF δ3 0.3188 0.2947 0.2794
Supply of subsidised fingerlings δ4 0.8909 0.3140 0.0045
No of cattle and buffalos δ5 -0.0012 0.0007 0.1088
Slow growth fish fingerlings δ6 -0.1651 0.3021 0.5848
Fast growing fish fingerlings δ7 0.5506 0.4406 0.2114
Number of months of water used for other uses δ8 -0.0408 0.0409 0.3184
115
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 115
Group stability, number of cattle and water buffalos in the catchment were the
most significant (significant at 10% level) factors that have positively influenced TE.
Slow growth Fish fingerlings had a negative effect while fast growing fish
fingerlings had a positive influence on technical inefficiency. The number of months
of water use for other purposes (multiple use of water) had a positive relationship
with TE.
The time spent meeting officials (i.e., fisheries extension officers), the
expectation of receiving adequate rain water to the reservoir, supply of subsidised
fingerling for CBF or investment for CBF activity in a respective culture cycle from
a third party aid (i.e., government, local level politicians) are negatively influenced
on TE (See Table 6.7).
The mean TE of CBF production in village systems is 0.33. Figure 6.1 shows
that the majority of the farmers are technically inefficient. In other words, TE
distribution is skewed to the right (See Figure 6.1).
Figure 6.1. Frequency distribution of TE estimates
The range of TE varied from 0.01 to 0.79. Therefore, CBF production in VISs
can be increased approximately threefold with the present state of technology, if the
technical inefficiency is removed completely.
116
116 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
6.6 DISCUSSION
Promoting multiple uses of water in village reservoirs for various agricultural
activities is likely to increase water productivity (Phengphaengsy & Okudaira, 2008).
The main objective of the construction of village irrigation systems in Sri Lanka was
to harvest rainwater in the low rainfall regions of the country to undertake rice
farming. Since the 1980s, there has been a growing trend of releasing fish into these
reservoirs. This involves stocking of hatchery–reared fingerlings, especially those
carp species capable of feeding and growing on the natural productivity of the
reservoirs (Ryther, 1981). This CBF activity adds a new dimension in increasing
water productivity VISs. However, few studies have been conducted on the
efficiency of fish farming in the Asian region (Dey et al., 2000) and this is the first
attempt to estimate the TE of CBF production in Sri Lanka.
From the results of the analysis, the current average TE of CBF in Sri Lankan
irrigation systems is only 33%, which is considerably lower than the regional mean
TE (57%) for extensive aquaculture systems in Asia (Sharma & Leung, 2010). Mean
TE of extensive aquaculture systems of some selected South and South-East Asian
countries are shown in Table 6.8.
Table 6.8.
Mean TE of selected South and South Asian countries
Country Mean source
Pakistan
Malaysia
Philippines
Nepal
India
Bangladesh
0.56
0.42
0.83
0.59
0.50
0.46
Sharma (1999)
Limuma et al. (1999).
Dey et al. (2000).
Sharma & Leung (2000)
Sharma & Leung (2000)
Sharma & Leung (2000)
In this study, the possible reasons for this low level of TE of CBF production
in Sri Lanka were investigated, with a view to identifying appropriate remedial
measures.
The inputs used in CBF in Sri Lanka are limited compared to other Asian
countries (i.e., Bangladesh, India, Viet Nam, and Nepal) where CBF is practised. In
117
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 117
Sri Lanka, the existing water bodies are used for CBF instead of ponds. There is no
need to allocate land for the construction of water ponds. Supplementary feeding
using fish feed, oil cake or rice bran (Singh et al., 2009) is not undertaken and
fertilisation of water to enhance growth of natural food like the addition of cow dung
(Singh et al., 2009) is not practised. Similarly, CBF practices in Sri Lankan village
reservoirs do not involve water quality enhancement using lime (Rubbani et al.,
2004; Kareem et al., 2009), urea (Rubbani et al., 2004) or chemical fertilisers
(Sharma & Leung, 1998; Singh et al., 2009)37
. Such measures are not needed in Sri
Lanka as the reservoir water is supplemented with allochthonous nutrients by the
livestock grazing within the catchments that contribute a large amount of nitrogen
and phosphorus through their faecal matter (Jayasinghe & Amarasinghe, 2007).
Similar means of supply of nutrient inputs are reported elsewhere in the literature
(Nash & Halliwell, 1999; Bravo-Ureta et al., 2003; Jennings et al., 2003).
As shown in Table 6.5, the volume of water used for CBF is a highly
influential factor of CBF production. The output elasticity with respect to water level
in the reservoir is 0.45. This indicates that a 10% increase of residual water for CBF
production increases the output by 4.5%. There is no possibility of increasing the
capacity of these reservoirs as the reservoir has been constructed in a cascade
system38
, as discussed in Chapter 2. It is not possible to increase capacity either by
connecting with the main irrigation systems39
or by relying on monsoonal rain.
Changes in reservoir capacity through monsoonal rain are completely random.
Therefore, the only practical possibilities of increasing the residual volume in VISs
are through the efficient use of water in rice farming.
The output elasticity with respect to labour involvement on CBF is positive in
general (Sharma & Leung, 1998; Linuma et al., 1999; Dey et al., 2000; Kareem et
al., 2009; Sing et al., 2009), but it is not a significant input in the case of CBF in Sri
37
From the point of view of biodiversity, conservation and environmental protection the CBF in
village irrigations is considered as an eco-friendly development strategy (De Silva 2003).
38
Only one example can be found in southern Sri Lanka where some of the village reservoirs are
connected with the main irrigation system “Malala-Mauara project”.These village reservoirs are minor
perennial reservoirs.
39
One example can be found in southern Sri Lanka where some of the village reservoirs are connected
with the main irrigation system “Malala-Mauara project”. These village reservoirs are minor perennial
reservoirs.
118
118 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
Lanka. There are two reasons for the lack of significant labour involvement to be in
CBF in the context of Sri Lanka. Firstly, there is only limited labour requirements
needed for the three phases of CBF. That is, stocking the fingerlings, protecting the
fish from poachers and harvesting. Since labour is not used for feeding, care or
adding fertiliser, an increase in labour input would not result in higher production.
Secondly, farmers work as a group keep as it was decides on the allocation of labour
for each activity. The performance of the group‟s labour depends on group stability.
This will be further discussed in this chapter. Excess supply of labour by the group
may result in no change in output, or a significant relationship may not be apparent
in increasing technical efficiency.
Output elasticity of total fish fingerlings is 0.27 in the model. This indicates
that a 10% increase in total fish fingerlings can only increase CBF production
approximately by 3%. However, there are two factors that have to be taken into
account in fish stocking: stocking density and the water capacity of the village
irrigation systems. Recommended stocking density of 2000 fingerlings ha-1
has been
suggested for achieving an average yield of 750-1000kg ha-1
(Chandrasoma &
Kumarasiri, 1986). For estimation of SD, 50% of reservoir area at full supply level
(FSL) can be considered as the effective area. This is because reservoir capacity
varies from FSL (See Figure 7.1) during the rainy season to almost zero during the
low rainfall season (Wijenayake et al., 2005). With these constraints, the only
possibility is to change the species combination instead of increasing the total
stocking density. Fingerling production for aquaculture is still managed by
government breeding centres. The breeding centres grow post-larvae up to the fry
stage for approximately one month duration. The rural farmers grow them up to the
size of fingerlings, for approximately two months. A certain quantity of fish
fingerlings is also produced at government fish breeding centres. Fish farmers
purchase these fingerlings from rural farmers and/or government breeding centres.
The private sector does not yet play a role in fish breeding due to various constraints
(i.e., technical know how). As discussed earlier, Indian and Chinese carp species are
stocked into VISs. Depending on their growth status, fingerlings are categorised into
two groups; FGS and slow growing species (SGS). However, most of the efficiency
studies have not considered such partial influences on TE of different species
combinations (Sharma & Leung, 1998; Linuma, 1999; Kareem et al., 2008).
119
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 119
Determinants of technical inefficiency
The results of the inefficiency model are shown in Table 6.7. Inefficiency can
occur due to random and non-random reasons (Aigner et al., 1977; Meeusen & van
den Broeck, 1977). The Gamma (γ) value of the estimated model (82%) shows that
technical inefficiency effects deviate from the frontier due to technical inefficiency
(See Table 6.6).
Amongst the factors considered, there are four that positively affect TE. They
are group stability, number of cattle and water buffalos grazing in the catchment
area, stock of slow growing fish species and the number of months of water used by
other users. The group stability for solving water disputes as mentioned earlier, is
farmers‟ willingness to continue CBF activities for the next culture cycle with the
same group members. This is one way of solving water disputes. This measure
addresses the issue of how much group members agree on collective decisions. Most
importantly, collective agreements in protecting fish from the fish poachers until the
final harvest significantly influences TE. Collective decisions of such communities
are dependent on homogeneity of group characteristics (Kularatne et al., 2009).
However, group stability can be determined by various social and economic factors
(age of farmers, education, income and employment).
Another significant positive factor impacting on TE is the number of animals
(cattle and water buffalos) living in the catchment areas. Cow dung or other manure
is used as fish feed (input) in most Asian countries (Nepal, India, Bangladesh and
Pakistan) in their culture ponds (Dey et al., 2000; Sharma & Leung, 2000; Singh, et
al., 2009). Adding cow dung results in enhanced biological productivity and thereby
increased aquaculture production. It has been reported that there is a positive
correlation between cattle/buffalos density and the fish yield (Jayasinghe &
Amarasinghe, 2007). Therefore, the number of cattle and buffalos grazing in the
watershed area was included in the inefficiency model as a proxy for the amount of
animal manure entering the reservoirs. As expected, the estimated coefficient relating
to the number of cattle and water buffalos was significant and positively influenced
TE.
Jayasinghe et al. (2005) argue that the difference in stocked fish fingerlings
combination with respect to their growth rates may not be the main reason for the
120
120 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
fish yield variations. However, analysis suggests that fish species with a slow growth
rate have a positive influence on TE. Jayasinghe et al. (2005) demonstrate that the
particular two districts were limnologically less productive than other districts in the
low rainfall zone (e.g., Monaragala and Hambantota) of the country. Therefore, it is
possible that FGS may have a growth feed issue in the two districts. This result is
consisted with the biological research findings of Wijenayake et al. (2005), from
which they demonstrated instantaneous mortality rates for common carp (Cyprinus
carpio), bighead carp (Hypophthalmichthys nobilis40
and catla (Catla catla). These
species (which have been considered as FGS in this study) were higher than other
species.
Water in VISs have multiple uses (Renwick, 2001 (a)). Reservoirs that are
located close to the village may have more alternative uses than the reservoirs
located far from the village. Therefore, the marginal value of water should be higher
in the reservoirs that are located close to the village. As described previously, 59% of
the reservoirs had fish poaching problems due to the open access nature of reservoirs.
On the other hand, costs of enforcement and monitoring are 78.6% the total
transaction cost of FO organised CBF production (Senaratne & Karunanayake,
2006). It is presumed that an increase in the number of months of other water uses of
the reservoir may increase enforcement and monitoring costs and technical
inefficiency. This situation is ultimately the result of the absence of well-defined
property rights which are linked with the spatial patterns of economic activities
(Otsuki, 2002).
As evident from the present analysis there are factors that influence TE both
positively and negatively. The key influential factor on technical inefficiency is
subsidisation of the supply of fingerlings for CBF activities. In addition, the time
spent meeting government officials is significant at the 5% significance level. The
other two factors, which positively influence technical inefficiency, are risk of
receiving adequate rainwater and fast growing fish fingerlings. Clearly, receiving
adequate rainwater is beyond the control of CBF farmers and this might have a
significant influence on the commencement of CBF activities. Most indigenous
species that naturally breed in the reservoirs are carnivores (i.e., snakehead, goby,
40
See http://www.fishbase.org/Summary/SpeciesSummary.php?ID=275&AT=bighead+carp.
121
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 121
and climbing perch). They can be a threat to introduced herbivorous carp species.
However, breeding of non-stocked fish species in these reservoirs is considerably
low, as reservoirs dry-up completely during a few months of the year. Therefore, this
factor is excluded from the inefficiency model.
Fingerling supply is subsidised by various sources such as non-government
organisations, regional and local level government authorities, as well as direct
government subsidy programmes. As the main cost item in CBF production most of
these subsidies are provided to farmers as a support to reduce the cost of fingerlings.
One of the theoretically possible impacts of providing subsidies on production is that
the level of input used is affected due to the changes to the cost of inputs. (Zhu &
Lansink, 2010). Subsidies can have a positive effect on income and TE if they are
provided to farmers who have well defined property rights of their economic activity.
In CBF production, due to subsidised fingerling supply, farmers are pushed towards
a situation of „free-to-all‟ hence leading to an open access tragedy as discussed by
Hardin (1968). In Anuradhapura, 66% of CBF farmers have considered the problem
of free-to-all (poaching and other problem from the villagers) as a constraint to CBF
development (Senaratne & Karunanayake, 2006). The FOs are given the power to
organise all agriculture-related activities by a government act. When all villagers are
members of an FO in a given village, the rights of villagers to use reservoir water and
the other resources are almost similar to their rights that have evolved historically.
Therefore, providing subsidies to CBF with ill-defined property rights leads to
technical inefficiency. This aspect will be further discussed in Chapter 8 with respect
to internalising the re-allocation of water.
The time spent on consulting government fisheries officials (i.e., fisheries
extension officer of NAQDA) for extension services are the second crucial factor
influencing inefficiency in CBF production41
. The cost of time searching for
information is part of managerial transaction costs (Furubotn & Richter, 2005).
Senaratne and Karunanayake (2006) have estimated that the information cost is 8.6%
of the total cost of CBF production that has been organised by FO, while it is 5.2%
41
Difficulties of meeting these officers are due to a shortage of officers. For example, there are only
nine Fisheries Extension Officers (FEO) for Kurunagala District and five FEO for Anuradhapura
District are available to inspect in total 6525 irrigation systems but all of them are not used for CBF.
122
122 Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production
for SGFs in VISs in the Anuradhapura District. Cost of information was estimated as
wage cost per working day that farmers have spent in meeting government officials.
Technical efficiencies
The estimated technical efficiencies for overall farmers in the sample ranged
from 0.01 to 0.79, with mean efficiency of 0.33. The frequency distributions of the
estimated TE levels are shown in Figure 6.1. The mean TE of CBF in Sri Lanka is
low compared to the other Asian countries. Mean TE of extensive aquaculture in
Nepal is 69%, which is lower than that of the intensive farming with the reported
mean efficiency of 77% (Sharma & Leung, 1998). In addition, tilapia production in
the Philippines has an 83% of mean efficiency and the estimated mean efficiency in
Thailand is 72% (Dey et al., 2000; 2005). However, a recent study shows that mean
efficiency in Indian fresh water aquaculture is 67% which is even less than that of
Nepal (Dey et al., 2005, Singh et al., 2009). A low mean efficiency (42%) is also
reported from Malaysia where 70% farmers are below 50% of mean TE (Linuma et
al., 1999). However, regional mean TE where extensive aquaculture takes place in
South Asia is 52% while intensive farming recorded a 75% of regional efficiency
level (Sharma & Leung, 2000).
Based on the above efficiency figures CBF production in Sri Lanka is operating
well below the mean TE levels of the Asian region. However, it is not possible to
generalise the mean efficiency situation in a regional context because of
discrepancies in the production systems.
6.7 CHAPTER SUMMARY
This is the first study of this nature and has been conducted to estimate the TE
of CBF production in Sri Lanka. The estimated TE of CBF in VISs are only 33%.
The results show that TE in Sri Lanka is the lowest compared to other Asian
countries. Basically the effect of random factors on TE of CBF production is
expected due to water use for CBF production which is entirely dependent on
monsoonal rainfall in the VISs. However, the proportion of variance of technical
inefficiency in the total error variance (γ) is 0.82.
When CBF production operates at full efficiency level without altering the
existing level of inputs use, there is a possibility to increase production by threefold.
Conversely, such production can be estimated using few inputs. In order to achieve
123
Chapter 6: Efficient water usage in village irrigation systems for culture-based fisheries production 123
these efficiency gains, it is important to strengthen group stability in order to solve
water disputes, improve official consultation, promote a mechanism to encourage
independent investments in CBF without depending on subsidies, and finally to
ensure water user rights are well defined.
.
125
Chapter 7: Inter-sectoral optimal allocation of water 125
Chapter 7: Inter-sectoral optimal allocation
of water
7.1 INTRODUCTION
The analyses of Chapters 4 and 5 show that water sharing issues and group stability
of solving water disputes are two important factors that have an influence on the TE
of rice farming and CBF production respectively. This chapter details the estimated
inter sector optimal allocation of water. That is, the optimal allocation of water for
rice farming and CBF. We first estimated the “frontier level” optimal water
allocation between rice and CBF. For this, we estimated a frontier production
function at frontier level for rice and fish separately. Then by equating the marginal
value products for each production, we estimated the optimal water allocation
between rice farming and CBF at the frontier level. Secondly, following the same
approach, we estimated the optimal water allocation between rice farming and CBF
at the “current” level of TE. By “current” we mean the existing level of TE (details
of the method of estimation have been documented in Appendix F). Molden et al.
2010 suggested that re-allocation of water from low to higher valued uses as one of
the strategies for increasing water use efficiency. Water allocation mechanisms are
part of the collective action of reservoir-based agriculture in Sri Lanka. This chapter
discusses group collective demand linked with the MVP of water uses in terms of the
optimal allocation for rice farming and CBF production.
7.2 INTER- SECTORAL WATER ALLOCATION
Irrigation development aimed at enhancing the physical structure of irrigation
systems has been a major development strategy of Sri Lanka. However, the
aggregate statistics show very pronounced temporal and spatial aspects of water
scarcity in the country (Samad, 2005) due to the bimodal pattern of rainfall. In the
beginning of the investment in irrigation development, the TE aspects were not given
inadequate attention (Thiruchelvam, 2010). Based on the rainfall pattern, the country
is geographically divided into a high rainfall region and low rainfall region42
42
In the literature, these two zones are called the wet zone and the dry zone.
126
126 Chapter 7: Inter-sectoral optimal allocation of water
(Kularatne et al., 2009). The high rainfall region, comprising about one-third of the
land area of the country, receives an average of 2,350 mm of annual rainfall
distributed over the two seasons. In the low rainfall region, the average annual
rainfall is only 1,450 mm. Most districts in the low rainfall region are likely to face
severe seasonal or year-round absolute water scarcity at the current level of irrigation
efficiency by 2025 (Samad, 2005). The farmers of these districts use approximately
75% of water available for irrigation. Consequently, there is no doubt that a suitable
water allocation mechanism is required for sustainable management of reservoir
water.
Water allocation and rights to access water are addressed by a number of
legislative enactments that have been developed over many years in response to a
variety of water allocation issues. The more important legislation and its key
provisions are summarised below to provide an overview of the situation relating to
the legal dimension of water institutions in the country43
. The village council was the
earliest known institution that engaged in water allocation rights. In 1815, the British
rulers abolished village councils but they were re-established in 1856. The Irrigation
Ordinance (No. 32) was first enacted in 1856 by the British colonial administration
to both legalise customary irrigation practices and to prescribe the conditions for
water extraction, particularly for rice cultivation. Notably, this ordinance does not
mandate a planning system nor does it address important issues such as inter-sectoral
allocation.
Appointment of an irrigation headman by the British administration was the
first turning point of collective management into state control in 1856 (Leach, 1961).
Following political independence in 1948, cultivation committees were appointed
under the paddy land act. However, these committees were abolished and replaced
by appointed officers nominated by Members of Parliament in 1977. The Agrarian
Services Act of 1979 and subsequent amendments were related to regulations
governing the land tenure systems of paddy land and the management of minor
43
The most important changes are:
a. the establishment of Cultivation Committees in 1958;
b. the establishment of Agricultural Production Committees and Agrarian Service Centres in
1972/1973;
c. the provision for Cultivation Officers in 1979; and
d. the provision for Farmer Organisations 1991.
127
Chapter 7: Inter-sectoral optimal allocation of water 127
irrigation schemes. The latter function was transferred subsequently to provincial
councils. At present, the Act provides legal recognition to FOs, stipulates the
responsibilities of the FOs including the levying of water fees, and confers the
authority on DAD to support the activities of FOs. However, the existing legislation
does not adequately address Sri Lanka‟s current and anticipated water resources
management needs. One of the major shortcomings is that existing laws do not
provide a logical basis for inter-sectoral water allocation.
The FOs are in charge of managing all minor irrigation systems. Since 1979,
the respective FOs have managed VISs which consist of less than 80 hectares of the
command area. The Agrarian Services Act (Amendment) of 1991 and the 1994
amendment to the Irrigation Ordinance gave substantial authority over irrigation to
FOs. Previously this authority had been held by public officials. This includes
obtaining bank loans, delivery of water to farmers and engaging in supplying farm
inputs and marketing farm produce. There are legal provisions for various rural
development activities through FOs, under the Agrarian Development Act 46 of
2000, which includes provisions for the development of CBF in village reservoirs.
However, water user rights are still not clearly defined for CBF activities.
7.3 THE CURRENT WATER ALLOCATION SYSTEM
The existing water allocation mechanism (See given example below) in village
reservoirs is very similar to the user-based water allocation system (Meinzen-Dick &
Jackson, 1996; Dinaret al., 1997; Dudu & Chumi, 2008). A user-based water
allocation system entails collective decision-making by users who hold rights (de-
facto or de-jure) over the water. These decisions could be influenced by factors such
as local norms, authority and capacity of the particular water users‟ associations at
the tertiary level. Therefore, UWA is flexible enough to be re-adjusted from one
season to another to meet changing local needs. However, in practice, UWA is often
mainly concerned with the efficiency of water use in terms of maximising output of a
particular product, while it ignores potential alternatives (Meinzen-Dick & Jackson,
1996).
128
128 Chapter 7: Inter-sectoral optimal allocation of water
How the current water allocation systems operate - an example
FOs begin their planning with the assumption that the reservoir capacity is in full
supply level and that no rainwater has been received during the cropping season. The
maximum depth of water in the reservoir is eight feet. The total duration of the
cropping period is 16 weeks and the total period of water supply is 14 weeks. No
water is supplied for 2 weeks before harvest. During the first 2 weeks of the cropping
season, water is supplied for land preparation. This is expected to reduce reservoir
capacity by one foot (0.30 metres) of the full supply level. Three days after seeding,
the sluice gate is opened for watering of all paddy fields. Thereafter, once a week,
water is supplied to the paddy fields for a period of 14 weeks. After the 14th
week, the
sluice gate is closed. During this period, FOs are expected to drain three feet
(approximately one metre) of water reservoir capacity. There is an additional foot
(0.30 metres) of water remaining in the reservoir for an extra supply for the paddy
fields if necessary, as well as for other uses. Basically, three feet (approximately one
metre) of water is maintained in every reservoir as „dead storage‟. This water cannot
be released from the normal sluice gates as discussed in .
Figure 2.3. Graphical presentation of land and water relationship. if there is an urgent
necessity, there is a middle sluice gate that can be used for releasing the dead storage
water (Personal communication, M.G. Haramanis, a FO president. December
15,2009). A typical reservoir with water and sluice gates is shown in Figure 7.1.
Figure 7.1. Measuring water levels in village reservoirs. Adapted from “Rains,
droughts and dreams of prosperity,” by P. Van der Molden, 2001, p.93.
129
Chapter 7: Inter-sectoral optimal allocation of water 129
During the cropping seasons (especially in the Yala season) that commences
with inadequate reservoir capacity due to the low rainfall, farmers work under the
ancient practice of a share cultivation system (Bethma44
). In the share cropping
cultivation, the FO decides how many hectare of land area will be cultivated and
irrigated in a particular season. This involves reducing the cropping intensity of the
paddy fields.
The demand for reservoir water increased with the CBF activities.
Consequently, the opportunity cost of water use in rice faming increased. In this
context, opportunity cost is defined as the value of water in its highest alternative
use. When the opportunity costs are high, conflicts among users may arise. Therefore
optimal inter-sector and intra-sector allocation choices have to be made. Therefore,
re-allocation of reservoir water has become a crucial issue in maximising agricultural
and CBF production. Therefore, one of the researchable questions in reservoir-based
agriculture is to quantify the inefficient allocation of water among multiple purposes
(irrigation, domestic, and fishing, livestock and cottage industries).
7.4 OPTIMAL ALLOCATION OF WATER
There are three main strategies which can be used to increase the net value of
water used in agriculture: (i) increasing yield (ii) changing from low to high value
crops and (iii) re-allocating water from low to higher valued uses (Molden et el.,
2010). This chapter discusses re-allocating water from low to higher valued uses.
Water allocation systems can be differentiated by the extent to which they fulfil
efficiency and equity objectives.
In this study, efficiency objectives rather than the equity aspects are observed.
The estimation of efficiency is necessary to achieve social objectives such as
resolving conflicts and assessing the opportunity cost of use in alternative uses
44
There are two main forms of Bethma. First as described by Leach (1961), each farmer decide to
cultivate only a percentages of their field area (such as one third), thus allowing more water for those
further from the reservoir. In the second, from a portion of head-end or part of the middle of the
command area of suitable size is selected and the rest is abandoned. The main determinant for this
selection is the availability of water in the reservoir. The selected portion is divided into an equal
number of shares. The person whose land is thus selected does not get a larger allotment than others
do. Each Bethma arrangement is binding only for one crop, and when it has been removed, reverts to
their original position. Quite often, the paddy tract selected for Bethma lies close to the reservoir bund
or to the irrigation ditch, thus helping to minimise conveyance losses and to conserve the available
irrigation water (Bandara, 1999).
130
130 Chapter 7: Inter-sectoral optimal allocation of water
(Young, 1991). Furthermore, an analysis of efficiency provides a useful guidance to
identify factors influencing inefficient allocation and to identify a way to enhance
total economic benefits of irrigation production (Turner et al., 2004). Fairness of
allocation of water among the multiple users, regardless of efficiency of water uses,
is paramount in water allocation. However, with regard to efficiency aspects,
marginal benefits from the use of water should be equal across the sectors (users) in
order to maximise social welfare (Dinar et al., 1997). Additionally, Dinar et al.
(1977) suggested that flexibility in allocation of supplies (inter and intra sectors) and
the security of tenure for water uses are important determinants in achieving
efficiency.
Two fundamental resource allocation methods can be found in the literature.
They are the (i) concept of proportionality embodied in the law of diminishing
returns, and the (ii) equi-marginal principle (Gopalakrishnan, 1967). In this chapter,
the equi-marginal principle is employed to value irrigation water in VISs in Sri
Lanka.
Water has a value but not a market price. Water is a classic example of a non-
market resource, which performs many functions and has important socio-economic
values even when used as a tradable commodity with the absence of a general market
price (Turner et al., 2004). However, information on the economic value of water
enables decision makers to choose between alternative uses of water (development,
conservation and allocation), especially, when there are competing demands and
water shortages (Ward & Michelsen, 2002). The value of water is likely to be site-
specific and each case deals with its own unique issues (i.e., irrigation, industrial
uses, hydroelectric power generation and domestic use). In the case of value of water
for irrigation, it is a measure of the net economic contribution of water to the value of
agricultural production (Young, 1996). Young (1996) suggests that the use of a
shadow price45
for water is possible in the absence of market prices.
The optimal allocation of water for irrigation maximises net benefits to society
in the short run. The social marginal value of irrigation water across different groups
of users should be the same; if not, re-allocation and co-managing water among uses
45
“the value used in economic analysis when the market price is in some way an inadequate measure
of economic value” (Young, 1996 p.xv).
131
Chapter 7: Inter-sectoral optimal allocation of water 131
from lower to higher value uses within and between sectors could increase the net
social benefits (Sampath, 1992; Molden, et al., 2010). This study uses a similar
definition: an efficient (optimal) allocation maximises net benefits in the equalisation
of MVP from the use of water across sectors to maximise social welfare (Johansson
et al., 2002). The estimation of the optimum level provides useful guidance for
understanding the causes of inefficient allocation (Turner et al., 2004). The EE of
irrigation water needs to be clearly differentiated from the various technical
definitions of efficiencies with respect to irrigation. Two types of efficient allocation
can be identified which are related to maximising water productivity. They are: (i)
optimal allocation of water is that which maximises social welfare that is gained
from the available water, and (ii) economically efficient allocation maximises the
value of water across all sectors of the economy, because water productivity can be
increased by promoting multiple uses of water (Phengphaengsy, 2008). The latter can
be defined as water productivity, i.e., the amount of food produced per unit of water
used or net return for a unit of water used (Wichelns, 2002; Molden et al., 2010).
There are various methods that may be used to measure the value of water
(Young, 1997; Long, 1991). The production function approach is the most
appropriate valuation method to value water in the short run, especially when there is
a water shortage situation (Long, 1991). In the linear production function, there is a
constant relationship between inputs and output and, consequently, marginal product
is constant. Therefore, a logarithmic production function has an advantage over the
linear production function in that diminishing marginal returns may exist. The
estimation of such functions is useful in decision-making for the optimal allocation
of water because it estimates MVP, where marginal costs equal marginal returns
(Long, 1991). MVP as a shadow price of water (Turner et al., 2004) can measure
water. This value represents the additional value of product due to the use of an
additional unit of water (Tilmant et al., 2008). MVP measures can be used to increase
the productivity of water by re-allocating it to uses that are more productive.
Junna et al. (2006) suggested that water re-allocation from the agricultural
sector to other sectors, based on the sectoral marginal value, would reduce the
income of the poorest households and hence increase the disparity between rich and
poor households. To minimise this disparity, they suggested a 30% reduction in inter
sectoral re-allocation based on MVP while promoting intra-sectoral re-allocation
132
132 Chapter 7: Inter-sectoral optimal allocation of water
based on efficiency. Irrigation water in most cases is allocated between agriculture
and fisheries. Improper allocation of surface water creates unequal performance of
fisheries production (Nguyen Khoa et al., 2005; Shankar et al., 2005). Shankar
(2005) examined the effect of surface water abstraction for rice irrigation on
floodplain fish production in Bangladesh. He estimated that every hectare of surface
water-based irrigation for rice reduced total fish catch by 272 kg. Also, Nguyen Khoa
et al. (2005) obtained similar results by analysing the impact of 10 weir and 10 dam
irrigation sites in Laos. This literature clearly demonstrates that there is a significant
relationship between water allocation and fish production efficiency.
Brooks & Harris (2008) provided estimates of the magnitude of efficiency
gains from water markets operating on a weekly basis in three trading zones in
Australia. Results indicated that a substantial gain in EE could be achieved by re-
allocating water from low to high value uses, which could further improve water
productivity.
According to Tuong and Bouman (2002), water productivity in rice culture is
twofold: (i) the volume of water required to prepare the land, and (ii) the volume of
water needed from rice growth period. As suggested, water productivity can be
increased by (i) reducing the water inputs that do not contribute to the yield
formation, or (ii) increasing rice yield. The first strategy suggested by Tuong and
Bouman (2002) reduces the inefficient volume of water used in rice farming and the
second implies increasing yield adopting new technology.
As Lallana et al. (2001) and Le Moigne et al. (1997) have shown, there are
advantages and disadvantages of different water allocation mechanisms46
. The main
disadvantage of user-based water allocation, which is examined in this chapter, is
that it can be limited in their effectiveness for inter-sectoral allocation of water, as
decision makers do not include all users. However, it is very hard to find micro level
optimal water allocation models to help understand the optimal allocation issues.
Decision-making on water allocation or rights transfer, incentives or disincentives to
adopt efficient irrigation technologies are also difficult to locate.
46
There are four main water allocation mechanisms: marginal cost pricing, public administrative
allocation, water markets and user-based water allocation. Le Moigne et al. (1997) considered only the
first three allocation mechanisms.
133
Chapter 7: Inter-sectoral optimal allocation of water 133
Reservoir-based agriculture is a collective economic activity. The decision-
making on activities such as water allocation, selection of seeds and preparing a
cropping calendar are a group activity. Therefore, the demand for water can be
identified as the groups‟ collective demand. However, individuals are able to decide
on the quantity of other inputs used (e.g., seeds, labour, fertiliser, insecticides and
herbicides) in rice farming. There are two decision-making units in rice farming,
FOs, who operate at a reservoir level DMU, and individual farmers. On the other
hand, CBF is entirely a group-based activity. All the decisions in CBF production
involve collective agreements. Therefore, the DMU in CBF is for an individual
reservoir that also is impacted by groups‟ collective demand for water. However, this
can vary among individual reservoirs, as group sizes differ.
Poor management of irrigation systems has resulted in a high degree of water
misuse while poor maintenance has led to huge conveyance loses (Dennis & Arriens,
2005). Even though rice farmers have property rights to their lands, they do not have
user rights over the collective demand for water. As a result, the individual farmers
are not in a position to transfer their rights to use water to another party for a
different purpose.
7.5 EMPIRICAL MODEL
The main objective of this chapter, as mentioned in the introduction, is to
investigate optimal inter-sectoral water allocation in VISs, in order to maximise the
return from reservoir-based agricultural production. The research explores why
current inter-sector water allocation is inefficient and how they can be optimally
reallocated. The research is dependent mainly on primary data due to the absence of
a reservoir level secondary database. Multi-stage cluster sampling methods as
discussed in Chapter 3 (Cochran, 1960) were used for sample selection. Each stage
represents the number of reservoirs, based on an administrative hierarchy from
national level to village level. Two administrative districts, namely Anuradhapura
and Kurunagala, were selected as study areas, because these two districts have the
highest number of village reservoirs in the country. As a whole, 460 rice farmers
have been interviewed. The sample represented 76% of the total farmers of the study
area. The total sample of these two districts included 325 CBF farmer groups. Of the
village of Anuradhapura has 165 reservoirs and Kurunegala has 160 reservoirs. This
represents 29% of the total 1168 reservoirs used for CBF production in the country
134
134 Chapter 7: Inter-sectoral optimal allocation of water
over the last three culture cycles. Nine reservoirs used for CBF production in the
Anuradhapura district were not sampled due to the insufficient number of farmers for
the group discussion in the village during the sample survey. Data collection
occurred through direct interviews with selected rice farmers and CBF farmer groups
using pre-tested questionnaires. The CBF farmer survey was organised as a group
discussion. Both farmer surveys were conducted from December 2009 to March
2010.
The general translog functional forms that have been used to estimate the rice
frontier production function in Section 5.4 and CBF frontier production function in
Section 6.4 are used to derive MVP in this section.
When there is no direct
valuation for irrigation water, then indirect methods of valuing should be used. This
study uses the MVP of water to estimate the shadow value of water (Tuner et al.,
2004) in rice farming and CBF production.
Then the optimal allocation condition holds (See Section 2.3.4 for more details)
when:
MVP MVP
R F
7.6 RESULTS
The production frontier approach for estimating MVP is the most appropriate
tool to be used for short run water allocation issues. Briefly, this approach estimates
the relationship between the volume of water use and output, while other factors of
production are assumed at the average level. Based on the estimated parameters of
the production frontiers (See Tables 5.4 and 6.5), estimated rice-water (lnYR) and
CBF-water (lnYF) frontier production functions are shown in Equations 6.1 and 6.2.
2
RlnY = 0.2866 + 0.3231lnw + 0.1661lnw (7.1)
2
FlnY = 1.5025 + 0.4466lnw + 0.1647lnw (7.2)
lnYR and lnYF represent the natural logarithm of rice output and CBF production.
Natural logarithms of individual volumes of water used for rice and CBF production
are represented by lnwRi and lnwFi
respectively.
Not all farms are efficient. The estimated mean TE for rice production is 0.73
and for CBF production is 0.33. The production functions that have been estimated at
135
Chapter 7: Inter-sectoral optimal allocation of water 135
frontier level transform into current levels of input use by mean level of TE.
Estimated production function at the Mean TE47
for rice and CBF are:
2
rlny = 0.2092+0.3231lnw+0.1661lnw (7.3)
2
rlny = 0.4958 + 0.4466lnw + 0.1647lnw (7.4)
As illustrated in Figure 3.7, the initial assumption was that MVP of water use
for CBF production would be higher than the MVP of water used for rice production
given the current level of TE. This pre-assumption was derived due to comparatively
higher market price for CBF production. The results of inter-sectoral water allocation
are shown in Table 7.1 at the frontier and the current level of shadow value of water
(See detailed estimation of optimal water allocation is demonstrated in Appendix F).
Table 7.1
Inter-sectoral optimal allocation and shadow value of water
Water allocation levels Allocation
conditions
Shadow value of water (λ), inter-sectoral allocation
of water at frontier and current level of production
Water for rice
(M/ha)
Water for
CBF (M/ha)
Shadow value
(LKR48
/Mha)
Actual allocation
Optimal given efficiency
Optimal frontier level
WR >WF
mvpr = mvpf
MVPR=MVPF
3.3881
4.2338
2.3100
2.0329
1.1872
3.111
-
20660
71055
The effects of optimal allocation at given
efficiency from actual allocation on inter-
sectoral allocation
Increased
by 25%
Decreased
by 42%
-
The effects of optimal allocation from
actual allocation to frontier level of
production on inter-sectoral allocation of
water and to the shadow value
Volume of water
can be decreased
by 32%
Increase
by 53%
A threefold
increase in MVP
Notes: Estimated mean capacity of VIS = 5.421M/ha
Mean TE for rice farming = 0.73
Mean TE for CBF production = 0.33
LKR = Sri Lankan Rupees/Currency
47
2
r
2
f
lny = MTE(lnY ) = 0.73(0.2866+0.3231lnw+0.1661lnw )
lny MTE(lnY ) 0.33(1.5025+0.4466lnw+0.1647w )
R
F
48 Exchange rate AU$ 1 = LKR 100
136
136 Chapter 7: Inter-sectoral optimal allocation of water
The estimated mean capacity of VIS is 5.421 M/ha. The actual allocation of
water is decided by FOs. Assuming reservoir capacity is at the full supply level,
62.5% (3.3881 M/ha) of water is allocated for rice farming, while the rest is used for
other purposes including CBF. The volume of water used for rice farming at the
optimal allocation of given TE is 4.2338M/ha. This means that the actual allocation
needs to be increased by 25% for rice farming. Therefore, the volume of water used
for CBF should be decreased by 42%. This is because actual allocation is an ad hoc
decision of FOs. However, as shown as in Figure 7.2, there is a huge potential to
increase MVP of CBF production at the level of frontier production. The estimation
shows that the effect of optimal allocation from actual level to the frontier level of
production would increase total water productivity three fold. For this to occur, water
would need to be reallocated by reducing 32% of the actual allocation. Such
inefficient volumes of water can be reallocated for CBF production by 53%.
Source: Compiled by Author
Figure 7.2. MVP of water for CBF and rice production in VISs
The estimated shadow value of water (per M/ha) at the given level of TE is
LKR 20660 (approximately AU$ 206) per M/ha. This can be increased
137
Chapter 7: Inter-sectoral optimal allocation of water 137
approximately by five times (up to LKR 71055 per M/ha) by removing the technical
inefficiency of rice farming. This situation is shown in Figure 7.2.
Empirically estimated MVP of water (per M/ha) use for CBF is higher than the
MVP of rice at the frontier level of production ( F R ). However due to the low
level of TE, MVP for CBF production is lower than the MVP of rice production at
the estimated level of TE.
7.7 DISCUSSION
The aim is to achieve optimal efficiency of water allocation: the MVP of water
used for rice farming should be equivalent to the MVP of that used for CBF
(Freebairn, 2003). The best combination of resource allocation can be called the
“optimal” or “efficient” allocation. It is clear that factors influencing efficiency
contribute to the optimal allocation and vice versa. Accordingly, every decision on
allocation is considered to have an input and output relationship (“production
function”) to achieve a technically feasible output. Allocation can be made for multi-
purposes such as irrigation, fishery, and domestic uses or between multi-users such
as farmers, fishermen and dairy farmers. Furthermore, water can also be re-allocated
within the same sector. For instance, for rice farming if the current allocation system
is inefficient (Meinzen-Dick & Bakker, 2001).
The expectation is that both rice and CBF farmers utilise water as much as
possible to maximise their profits. It can be argued that allocation of water for rice
and CBF farming might be more beneficial than the present practice in which water
is used only for rice farming. As shown in Figure 7.2, at the current level of TE,
MVP of water that is used for rice farming is lower than the MVP of water use for
CBF. With the appropriate measures for removing the technical inefficiency of water
use, if the water demand can be allocated for both rice farming and CBF, MVP of
water can be dramatically changed. At the frontier level of production, MVP of water
used for CBF is higher than the value of water used for rice farming. Eventually,
more water could be allocated for CBF and less water for rice farming if there is a
free water-right system. Similar findings were reported in South Africa by Farolfi &
Perret, (2002), who found that the productivity of water in the mining sector was far
higher than the smallholders‟ irrigation. They suggested that if a free water-right
market were actually implemented, such unbalanced willingness to pay would result
138
138 Chapter 7: Inter-sectoral optimal allocation of water
in the total transfer of water rights allocated to the smallholding irrigation sector
towards the mining sector. However, they did not calculate water productivity that
was used for purposes other than mining and irrigation agriculture.
The ongoing water allocation mechanism favours rice farming. TE in rice
production in the Kurunegala district was recorded as 0.69 (Aheeyar et al., 2005) in
2005 and in the Anuradhpura district it was 0.65 in 2001/2002 (Thiruchelvam, 2003).
Rice farming for VISs have been a well-established economic activity, while CBF
has been a growing economic sector for most VISs. However, water plays a main
role in the production process.
In the broader sense, subsidising fingerling supply and transaction costs (i.e.,
time spending to meet officials) were significant factors that influenced the low level
of TE of CBF production. It is a fact that providing subsidies without well-
established user rights of water may lead to alienate the direct beneficiaries of the
CBF production. Due to a lack of property rights, farmers cannot transfer their water
user rights to other types of uses or users. The existing system of water user rights
does not facilitate inter-sectoral transfers of water and existing policies are not
implemented successfully to inter-sector water transfer. Water markets have not been
common either (Hearne, 1995). One reason why particular inter-sectoral transfers of
water are not common in VISs are that the less dramatic transfer of irrigated land
with its irrigation water to other uses is lacking. These results are in line with the
study by Kulindwa (2000) in Tanzania. He revealed two main reasons for the
inefficient allocation of water: (i) the restriction on water transfers which prevents
water to be re-allocated to the highest value use, and (ii) charging inefficiently low
prices for water. In the case of reservoir water in Sri Lanka, there is no absolute price
for irrigation water.
However, successive mechanisms that could increase TE (up to the frontier
level) of rice production would lead to increased MVP of water. With the increase of
TE, there will be a need to re-allocate water with a proper allocation mechanism.
There is an opportunity to increase the MVP of reservoir water approximately by
three times through the efficient use of water for rice farming. According to the
estimation in this thesis, inefficient use of water in rice farming (at the frontier level
of production) is approximately 32%. This percentage could be used for other uses.
Efficient use of water for irrigation enables users to increase the residual volume of
139
Chapter 7: Inter-sectoral optimal allocation of water 139
water. Consequently, the ultimate effect of efficient use of water in rice farming is
increased residual water in the reservoir.
The promotion of collective action for members of the farmer organisations is a
significant approach for influencing efficient water uses in the VISs. Kashaigili et al.
(2003) have revealed four factors as the major constraints and potential for achieving
efficient systems of allocating water resources to different uses and users in
Tanzania. The constraints that they identified are: (a) the lack of active community
involvement in management of water resources, (b) conflicting institutions and weak
institutional capacities in terms of both regulations and protection, (c) lack of data
and information to inform policy and strategies for balanced water allocation, and (d)
inadequate funds for operation, maintenance and expansion of water supply systems.
These aspects are substantiated by the present analysis.
Market based water allocation methods are associated with the value of water
because markets enable the observation of human behaviour, in particular,
observation about the actual choices made by stakeholders with their scarce
resources (Morris, 2006).
Failure to identify policy instruments for the best allocation of water
significantly affects rural agricultural production and income. As Rosegrant (1997)
suggests, the institutional, legal and environmental reforms must empower the users
to make their own allocation decisions. Even though users are empowered, the
multiple water users are vulnerable to unfavourable decisions on water allocation,
unless they are well organised (Bakker & Matsuno, 2001). The possible result of
such unfavourable decisions is that it could lead to inefficiency, in-equitability and
un-sustainability (Renwick, 2001). Sectoral “allocation stress” is seen as resulting
from the unequal share of and inefficient use of water in the agricultural sector
(Molle & Berkoff, 2009). If re-allocation of water is aimed at achieving optimal
allocation, it is necessary to identify in advance the exact water requirements for
each area, namely rice and CBF production. Optimal allocation of water guarantees
aggregate benefits for users. Howell (2001) stated that through efficient use of
irrigated water, the output per unit of water used can be increased by minimising
losses from less efficient uses and re-allocate water into „high priority‟ sectors.
However, this market-based re-allocating would not address the issue of equitable
140
140 Chapter 7: Inter-sectoral optimal allocation of water
allocation of reservoir water in the context of benefits sharing of reservoir-based
agriculture.
7.8 CHAPTER SUMMARY
Agricultural use has a low marginal value for water (Juana et al., 2006). Re-
allocation of water from this sector to others, based on the sectoral marginal values
of the resource, has the potential to increase the income of the poorest households.
The chapter showed agriculture‟s marginal returns from using water in VISs are not
as high as in CBF production. Rice farming plays a major role in sustaining the
livelihoods of farmer households in the villages. It has forward and backward
linkages in the economy, which are not captured in the direct impact analysis
(Delgado et al., 1998). Therefore, any water re-allocation strategy that significantly
alters the production structure in this sector will be transmitted to the most vulnerable
population in the economy. However, the transfer of water from agriculture to other
sectors on the basis of marginal value will at least promote income generation for the
most vulnerable households.
Furthermore, re-allocation of water from agriculture to CBF production
increases the MVP of reservoir water. Therefore, the results favour the
implementation of inter-sectoral water re-allocation based on TE and support the
recommendation that the institution of user rights and inter-sectoral transfer of rights
could be a workable policy for promoting CBF production.
141
Chapter 8: Intra-sectoral optimal allocation of water 141
Chapter 8: Intra-sectoral optimal allocation
of water
8.1 INTRODUCTION
The main objective of this section is to investigate optimal intra-sector water
allocation in VISs in order to maximise returns from reservoir-based rice production.
The research explores why current intra-sector water allocation is inefficient and how
it can be optimally reallocated. Chapter 6 discussed the inter-sectoral water
allocation mechanism as a part of the issues that have arisen out of the analyses of
Chapters 4 and 5. This chapter will further analyse intra-sectoral optimal allocation
issues in VISs. That is, optimal allocation of water between head-end, middle and
TEFs. The same methodology is followed as for the estimation of optimal allocation
of water based on the frontier level and existing level of production. In this study, we
separate the total command area into three parts. The first part is considered as 1000
metres from the reservoir dam denoted as H. The second 1000 metres are called M.
The final 1000 metres of the command area is the tail-end (T) of the command
area49
.
8.2 INTRA-SECTORAL WATER ALLOCATION
The common water related issues such as water ownership, allocation and
water rights are not dealt with adequately in many Asian countries (Dennis &
Arriens, 2005). Nevertheless, these issues are vital because people use most water
bodies as a common pool resource with multiple uses. In the case of reservoir water
allocation, estimating the value of water and identifying its alternative uses are
essential for making re-allocation decisions (Kadigi et al., 2004). Reservoir water
essentially requires developing such a water allocation model to cater to competitive
demand (Dugan et al., 2006), especially where water rights have not yet been
established (Dennis & Arriens, 2005). Increasing water scarcity and competition
49
Traditionally, the command area of a VIS is divided into interconnected three main fields.
Therefore, the fields near to the reservoir (HEFs) is called ”Udapotha or Mulpotha”, the MFs area are
called “peralapotha” and the fields further down (TEFs) are called “ Aswaddumpotha” (Bandara,
2007).
142
142 Chapter 8: Intra-sectoral optimal allocation of water
among the multiple uses (Meinzen-Dick & Bakker, 2001) are important aspects in
the area of water allocation. Therefore, there is a challenge to develop an optimal
water allocation model, taking into consideration the full economic and social returns
to all water users (Meinzen-Dick and Jackson 1996). However, the potential
magnitude of the economic gains of spatial water re-allocation is yet to be
understood in small-scale irrigation systems (Mahendrarajah & Warr, 1991) for
better policy-making.
Spatial50
variations in agricultural production are a significant factor to
consider in measuring the level of efficiency, because the level of productivity across
agricultural farms could vary with respect to the physical location of the individual
plots. Such productivity differences can be due to various factors. For example, soils
have heterogeneous characteristics (Bell & Irwin, 2002). Therefore, local
productivity variations can be expected even within short distances (Florax et al.,
2002). Moreover, the „head-tail dilemma‟ has been identified as one of the most
common water allocation problems in irrigation water management (Sengupta et al.,
2001) that leads to productivity differences. A conveyance loss is also one of the
factors responsible for less water received by tail-end farmers (Chakravorty &
Roumasset, 1991).
The common hypothesis is that water productivity51
, and hence water user
efficiency52
, tends to be reduced with the distance from the water source (reservoir)
due to the conveyance losses, even though the volume of water released from
reservoirs increases (Chakravorty & Roumasset, 1991). This is shown in Figure 8.1.
50
Most researchers have examined intra-sector allocation of water under the term of “spatial
allocation”. See, for example Brumbelow & Georgakakos, (2007); Jinfen, (2004); Essafi, (1997);
Chakravorty & Roumasset, (1991); Chakravorty et al., (1993).
51
Water productivity is the net return for a unit of water used (Molden et al., 2010).
52
Irrigation specialists have used the term „water user efficiency‟ to show how effectively water is
delivered to crops and to indicate the amount of water wasted.
143
Chapter 8: Intra-sectoral optimal allocation of water 143
Figure 8.1. Relationship between declining rice output and distance from water
source. Adapted from “Efficient spatial allocation of irrigation water” by
Chakravorty & Roumasset, 1991, p.167, and “introduction to the special
issue on spatial analysis for agricultural economists” by Nelson, 2002.
Molden et al. (2010) suggested four primary methods of increasing the water
productivity. They are:
a. Increasing the productivity per unit of evapotranspiration (ET) at plant-
field and farm-scale.
b. Minimising non-productive depletion of water flows by reducing water
flows to sinks, minimising salinity and discharging polluted water to sink.
c. Improving management of existing irrigation facilities and reusing return
flows by controlling, diverting and storing drainage flows.
d. Re-allocating and co-managing water among users by re-allocating water
from lower value to higher value uses within and between sectors.
8.3 LITERATURE REVIEW
One of the common problems of water allocation from irrigation systems is
that the tail-end farmers receive insufficient water while head-end farmers over-
irrigate (use without proper controlling mechanism) their fields (Daleus et al., 1988;
Chakravorty et al., 1995; Chakravorty & Roumasset, 1991; Wichelns, 2002).
However, this can vary with different irrigation systems. Under an asymmetrical
system, water allocation becomes inefficient because distance from the main water
source to a particular field plays a vital role in irrigation (Van der Zaag, 2007) due to
conveyance losses (Chakravorty & Roumasset, 1991; Chakravorty et al., 1995).
144
144 Chapter 8: Intra-sectoral optimal allocation of water
Chakravorty & Roumasset (1991) presented a comprehensive spatial water
allocation model. They have shown that there is a negative relationship between
yield and field located distance from the water sources. According to them, the
volume of water applied should fall with the distance from the water source due to
conveyance losses. However, their theoretical model was not empirically estimated.
They have shown that with an optimal allocation, the value of the marginal product
of water used at the source would be equal across the farmers, but in practice the on-
farm value of marginal product of water was unequal across the farmers and would
rise with distance from the source. (Jinfen et al., 2004). Jinfen et al. (2004) discussed
the inequality of MVP of the different areas (marginal revenue) as part of spatial
optimisation. According to these researchers, when there is equal marginal revenue
in all areas, the economic benefit of water for all the intake sectors reaches its
optimum level. In addition, they demonstrated a way of estimating economic benefits
of water allocation between different sectors.
Brumbelow & Georgakakos (2007) investigated spatial distribution of water
with five different hypothetical allocation scenarios which were based on efficiency,
equity and security for the Lake Victoria basin in East Africa. They used common
optimisation techniques (dynamic programming) and adapted them to a multilevel
irrigation allocation. They concluded that equity objectives could lead to different
patterns of spatial water allocation and crop production depending on the political
and social scenarios defined.
Similar analysis can be found in terms of addressing spatial allocation issues by
estimating crop-water production functions for different sectors (Chakrabarty &
Samaranayake, 1983; Chakravorty & Roumasset, 1991; Essafi, 1997; Salman et al.
2001; Jinfen, 2004; Brumbelow, 2007). Salman et al.(2001) have investigated
physical water allocation issues at the macro level in Jordan. The estimated inter-
seasonal agricultural water allocation model (SAWAS) was a linear optimising
model of agriculture. They used data on available land, volume of water required per
unit land area for different crops, and the revenue generated growing these crops in
different locations. The applicability of this model is that it can be used to examine
the effect of water allocation between crops and different locations because of
changes in the output, price and water restrictions in the particular locations.
145
Chapter 8: Intra-sectoral optimal allocation of water 145
Another study of smallholder irrigation schemes in Eastern Zimbabwe has
found that their inefficiency can be largely attributed to unreliable and inadequate
water delivery (Pazvakawambwa & Van der Zaag, 2001). Pazvakawambwa & Van
der Zaag (2001) found that, in the case of maize production, the greater the distance
from the reservoir, the lower the yield. They estimated that for each metre away from
the upstream plot, maize yields decreased by 2.1 kg ha-1
. As a result of improper
allocation of water resources, large areas of TEFs in Eastern Zimbabwe were found
to be not cultivated during the low rainfall season due to water shortages (Ferguson,
1992). Therefore, there is output heterogeneity in respect to the distance between the
reservoir and fields.
There are only four studies reported in the context of Sri Lanka, directly linked
with the analysis of intra-sector allocation of water. Ekanayake & Jayasooriya (1987)
examined technical and allocative efficiency and water allocation in large irrigation
systems. Daleus et al. (1989, 1988) and Mahandararajha & Warr (1991) studied
water allocation and management issues in VISs.
Ekanayake & Jayasooriya (1987) estimated technical and allocative efficiency
of 124 Sri Lankan rice farmers. They found that even in larger irrigation systems,
water allocation issues between head-end and TEFs existed. They measured firm-
specific TE as well. Their results showed that there was a high level of technical
inefficiency (50%) in the tail end rice fields due to inadequate and non-timely supply
of water due to the distance from the water source and the other related effects such
as conveyance losses, evaporation and poor maintenance of the field canals.
Although the distance between water source and the fields is comparatively short in
VISs, it is important to investigate spatial allocation of water in relation to village
reservoirs.
The main objective of the study of Daleus et al. (1988) was to analyse yield
variation of rice production under a user-based water allocation mechanism. The
results of the linear regression analysis indicated a positive relationship between
yield and water coverage and hence the increase in the total efficiency of water use.
Furthermore, Daleus et al. (1988) noted that TEFs need less volume of water
compared to the middle and HEFs due to higher groundwater levels and high clay
content of the soil. The other important finding in this study was that farmers holding
land in the HEFs received lower yields than other locations. The key findings of this
146
146 Chapter 8: Intra-sectoral optimal allocation of water
study are that there is a positive relationship between water use and yield, but with a
spatial yield variation.
The main objective of the study by Daleus et al. (1989) was to analyse spatial
water allocation and rice yield variations in VISs. They found that both under-
irrigation and over-irrigation could have inefficient effects on the total output.
Distribution of water between fields (land) located close to the reservoirs was three
times higher than the downstream fields (Daleus et al., 1989). This study estimated
that mean agricultural yields decreased from about 4,176 to 718 kg/ha over a
distance of 300 metres. The study also estimated that a strong negative relationship
exists between yield and the distance from the reservoir. However, Daleus et al.
(1989) also stated that variation of the yield was due to factors other than water such
as water coverage, soil conditions, use of fertiliser, and poor pest and disease
management.
Mahendrarajah & Warr (1991) studied inter-temporal water allocation in VISs
in Sri Lanka. The main hypothesis of this study was that even though the adoption
HYVs in rice technology increased rice output, it has made the traditional inter-
temporal water management system less efficient. Mahendrarajah & Warr (1991)
stressed that there is a possibility of increasing HYV yields by one fourth with an
improved water allocation system.
Those who are close to the main canal (those at the head of the distributaries
canal) have ample access to water, and have to perform very little maintenance.
Those who are closer to the tail of distributional canals, on the contrary, have
uncertainty in accessing to water, and are potentially faced with performing a great
deal of maintenance to keep the distributaries working (Chambers, 1988; Hunt,
1989). Those who are at the head of the system have very little incentive to support
maintenance efforts below them on the distributaries. Further, if demand for water
below them increases, their own access to water will be impaired. For those at the
tail, on the other hand, there is very little incentive to perform the maintenance
because even if the maintenance is performed they have no advantage over those at
the head to release appropriate amounts of water. Uphoff (1985), reviewing an
irrigation system in Gal Oya, Sri Lanka, found that incentives for head-enders could
increase yields, but these would only occur if the supply of inputs were increased.
147
Chapter 8: Intra-sectoral optimal allocation of water 147
Collective management needs to have ways to increase total water availability.
Therefore reservoir management is much easier than canal management.
Chakravorty and Roumasset (1991) developed models for intra-sector
irrigation water allocation based on MVP that applied a similar approach to this
study to analyse optimal intra-sector water allocation among three different sectors of
the command area instead of among the individual farmers.
8.4 EMPIRICAL MODELS AND RESULTS
This section describes the intra-sector water allocation issue or optimal water
allocation among head-end paddy fields (H), middle paddy fields (M) and the tail-
end paddy fields (T). The research depends mainly on sample data collected in the
field survey.
The rice farmer study was conducted in 14 selected rice-farming villages53
each of which has its own reservoirs in Galgamuwa DSD of Kuraunagala district in
Sri Lanka. As documented in Chapter 3 multi-stage cluster sampling method
(Cochran, 1960) was used for sample selection. In total 460 farmers were
interviewed, representing 76% of the total farmers of the study area. The total sample
was divided into three sub-samples based on the location of the paddy fields in the
command area, as mentioned previously. Of the total sample of 460 farmers, 160
farmers were from HEFs. 152 farmers were from the MFs and 148 farmers were
from the TEFs. A rice farmer in a particular sector was the unit of analysis in the
survey. Data were collected in person using a pre-tested questionnaire.
As mentioned in chapter 5 the general translog functional form that was used
to estimate the rice production frontier can be expressed as:
5 5 51(H) lnY =β + β lnx + β lnx lnx +v -u (8.1)
i,t 0 i i,k i,k i,k i,l, i,t i,t2i=1 1 1i k
5 5 51
(M) lnY =β + β lnx + β lnx lnx +v -u (8.2)i,t 0 i i,k i,k i,k i,l, i,t i,t2i=1 1 1i k
53
They were Arthikulama, Dabagahawewe, Gallawa wewe, Gojaragama, Iddamalpitiya, Kallanchiya,
Madawachchiya, Makalanegama, Molewa, Nochchiya, Pahala konwewe, Pahala saviyagama,
Ussankuutiyawewe and Walpothuwewe.
148
148 Chapter 8: Intra-sectoral optimal allocation of water
5 5 51(T) lnY =β + β lnx + β lnx lnx +v -u (8.3)
i,t 0 i i,k i,k i,k i,l, i,t i,t2i=1 1 1i k
where Y is the rice output of farmer i in period t and i,k,tx are the agriculture inputs
(k,l) to the production process. riv and riu are as previously defined in Equation 3.4.
The variable includes in the sectoral rice frontier production functions are described
in Section 5.4. Furthermore, sectoral inefficiency models are specified following
Battese and Coelli (1995):
9
0
1
(3.6) (8.4)i j ij
j
U Z
where i is represented H.M and T. Variables of the inefficiency models are described
in Section 5.4. However, two variables of sectoral inefficiency models on paddy
fields locations are excluded
Furthermore, the following condition holds for optimal intra-sector water allocation
as is documented in Section 3.3.4
MVP =MVP =MVPH M T
8.5 RESULTS
The main five input variables (water, labour, power, irrigating time and
pesticides) are theoretically consistent of the estimated sectoral frontier production
functions (See source file and Tables G1, G2 and G3 in Appendix G). The estimated
empirical translog functions monotonically increase by 83% for the HEFs, 87% for
the MFs and 21% for the TEFs. Re-estimated rice-water frontier production
functions for the HEFs, MFs, and the TEFs are shown in Equations 7.4 - 7.6 and
include all other inputs at mean value, since data are logged and normalised to a
mean of 1, such that ln( ) = 0X .
2lnY = 0.2511 + 0.3180lnw + 0.1659lnw (8.4)H
R
2lnY = 0.6800 + 0.2816lnw + 0.0089lnw (8.5)M
R
2lnY = 0.2285 + 0.3999lnw + 0.1372lnw (8.6)T
R
149
Chapter 8: Intra-sectoral optimal allocation of water 149
where, lnY H
R, lnYM
Rand lnYT
R are natural logarithms of rice outputs and natural
logarithms of sectoral volume of water used for rice production represented by lnw
in HEFs, MFs and TEFs respectively. The mean production and technical
efficiencies of the three locations of the command areas are shown in Table 8.1.
Table 8.1
Sectoral average production and TE levels
Location Sectoral average
production (Kg)
Mean TE (%) Output elasticity
(frontier level)
HEFs 1078 74 0.32
MFs 1076 55 0.28
TEFs 1409 80 0.40
The average production was similar in the HEFs and the MFs, but the TE
varied by 20%. Both the highest average production and the highest TE were
reported from TEFs. The MFs were less technically efficient than the other two
sectors (See Table 8.1). The output elasticity for water is inelastic for all fields (See
Table 8.1).
2lny = 0.1858 + 0.3180lnw + 0.1659lnwH
r (8.7)
2lny = 0.3740 + 0.28161lnw + 0.0089lnwM
r (8.8)
2lny =0.1828 + 0.3999lnw + 0.1372lnwT
r (8.9)
The production functions that were estimated at the frontier level were
transformed into the existing level of production using the mean level of TE (i.e.,
lny = ln(x) + u ). The estimated frontier production functions at existing level of TE
for the three locations are shown in Equation 7.7 to 7.9.
150
150 Chapter 8: Intra-sectoral optimal allocation of water
Table 8.2
Estimated technical inefficiency model for sectoral rice production
Inefficiency variables of the models HEFs MFs TEFs
Age of farmer 0.0047 0.0206** 0.0036
Farmer‟s education level -0.0060 0.0241 0.0572*
Participation rate for FO activities -0.0121* -0.0045 -0.0115
FO membership -0.5929** -0.3872* -0.7195*
Water sharing issues 0.9149** 0.3924* 1.4232*
Land ownership 0.4594* -0.0142 -0.1194
Use of insecticides 1.0500** 0.6648* 2.8940
Use of weedicides -0.8458* 0.0826 -3.4103
Success of field level water mgt -0.0096** -0.0117** -0.0072
Notes: significance at * 10%, **5%, ***1%.
The factors that influence sectoral technical inefficiencies are presented in
Table 8.2 where a separate model uses are re-estimated for each sector following the
same method used to estimate frontier production functions for rice farming and CBF
production. Table 8.2 shows only the results of the re-estimated sectoral inefficiency
models.
The FO membership and sectoral water sharing issues are common and
significant (at least at 10% level) factors, with the expected sign for TE in the three
sectors. Use of insecticides in HEFs and MFs decreases the TE while success of
water management practises at the field level increases TE. However, these two
factors are not significant for the TEFs. Participation in collective activities increases
TE on HEFs but it has no significant impact on the other two sectors. The farmers
who have land in the head-end area are more involved in clearing canals in the
beginning of the cropping season. However, farmers in the other two sectors do not
bother too much because water flows to their fields due to the morphological settings
of the command area. A key factor affecting technical inefficiency in the MFs is
151
Chapter 8: Intra-sectoral optimal allocation of water 151
farmers‟ age, while education positively affects TEFs. Although land ownership
appears to decrease TE in the HEFs, it is not significant for the other two sectors.
The intra-sector allocation of water has been estimated for two different levels
of production: at the potential (frontier) level of production and the existing level of
TE as shown in Table 8.3. There is no method to approximate the actual sectoral
allocation of water uses.
These results show that inefficiency of water use in TEFs and the HEFs at the
frontier level of production. The inefficient volume of water in head-end and TEFs
(approximately 10% and 23% respectively) can be re-allocated for middle-fields
approximately by 63%. As a result of removing intra-sectoral inefficiency a twofold
increase can be expected in MVP of reservoir water (See Table 8.3).
Table 8.3
The optimal intra-sector allocation of water
Production levels Shadow
value of
water (M/ha)
Intra-sector water allocation (M/ha)
Head-end Middle Tail- end
Frontier level
Existing efficiency level
159350
78350
0.9551
1.0637
2.0163
0.7552
0.6816
1.5692
Impact of increased TE from existing
level to frontier level on water
allocation and shadow value of water
Can be
increased by
twofold
Can be
decreased
by 10%
Can be
increased
by 63%
Can be
decreased
by 23%
8.6 DISCUSSION
As in most industries, heterogeneity in farmer efficiency units results in actual
output being different from potential production. Hence, the production function
approach is more appropriate for estimating optimal water allocation since it also
allows the estimation of inefficiency in production. The results of inter-sectoral
allocation estimates in Chapter 6 revealed that inefficient volumes of water used in
rice farming could be removed by 53% at the frontier level of production. In this
152
152 Chapter 8: Intra-sectoral optimal allocation of water
section, it was further investigated how this inefficient volume of water spatially
(from head-end to tail-end) would be distributed in the command area. Investigating
optimal allocation at the potential level of production and deriving existing level of
allocation rules would inform the productive water use of VISs based on the
locational MVP of reservoir water.
Water markets are usually absent or ineffective. There are only two types of
payments made by farmers during the cropping season, but it cannot be considered as
price paid for water. Firstly, every farmer gives 11 kg of rice per acre to the head
farmer who is responsible for water allocation and management. This always has to
be paid in rice and is not a nominal value. Secondly, every farmer pays LKR 15 per
acre (equal to Au$ 0.15) to DAD through ARPAs. This is called Akkra badda (acre
tax or tax for land). However, this is not defined as a water tax and it is a negligible
amount compared to the revenue derived from rice farming. Therefore, the value of
water cannot be directly derived from market activities. Economists have proposed
various valuation techniques to determine the value of water specially to address
allocation issues. The shadow price is one of the non-market water valuation
methods dealing with the short-run allocation problems (Tilmant et al., 2008).
Traditional rice farming systems, which have sought to optimise the management
and use of internal inputs (i.e., on-farm resources) and to minimise the use of off-
farm inputs such as fertiliser and pesticides, have changed with the introduction of
modern rice farming practices. This modern farming system is evident by the huge
capital investment in farming and use of new technology, economies of scale and
HYVs that require extensive external inputs such as pesticides and fertilisers
(Bandara, 2007).
When water is used as a variable input in production, the MVP can measure the
value of water; that is the value of an additional quantity of product due to the use of
an additional unit of water. The importance of estimating MVP is to guide the re-
allocation of water to more productive uses (Tilmant et al., 2008).
Water is said to be inelastic: a given percentage change in volume of water
does not result in a greater percentage change in output. This is shows for the
estimated elasticity for sectoral water demand (See table 8.1). The TEFs and HEFs
are recorded to have the highest TE and average product per hectare and also more
elastic than the MFs. There is only a 6% difference of TE between the HEFs and
153
Chapter 8: Intra-sectoral optimal allocation of water 153
TEFs. Therefore, MFs are least efficient, inelastic and produce lowest average
production. Estimated average production in this thesis contradicts the findings of the
study carried out in three VISs in the Anuradhapura district of Sri Lanka by Daleus et
al. (1989). Their case study has analysed the allocation of water to individual plots in
a village reservoir to compare the spatial and temporal variation in water coverage
with the variation in yield. Daleus et al. (1989) found that the duration of water
coverage explained variations in rice yield for the middle and lower parts, whereas
the relationship between water coverage and yield was weak in the upper part of the
rice tracts. According to Daleus et al. (1989), in general, there was a decreasing yield
as distance from the tank increased (from 4,176 to 718 kg/ha over a distance of 300
m). At the same time, they found that there was a large variation in yields within
each sector that could be attributed to management problems. Furthermore, they
found that land fragmentation was relatively modest, but the average yield for the
farmer decreased when fragmentation increased. They also found that the
relationship between water coverage and yield is positive and significant. Daleus et
al. (1989) showed a positive relationship between yield and the water coverage for
the middle and the TEFs. This study also reveals a positive relationship between
water coverage and water retention period and therefore, a positive relationship with
average production.
There are two factors that influence overuse of water in tail end farms and
impact on higher average production: (i) the length of water retention period and (ii)
soil fertility. Normally, once a week, water is supplied to the fields from the
reservoir. According to the farmers, water retention days in the three sections of the
command area are different. In the HEFs, water is retained for 2 days. In the MFs
water is retained for two to three days, while in the TEFs water is retained for four
days. This is dependent on the soil type of the head and middle part of the command
area, slope and the water management practices of individual farmers. Therefore, the
water retention period in tail end field is higher than in the other two sectors. Also,
there is improved soil fertility in TEFs with the inundation from the downstream
reservoir. There are several reasons for soil quality variation of rice fields in the
head-end to the TEFs. The variations in productivity of a command area are
attributed to soil fertility even though water management practices are equally
practiced throughout the command area. The farmers observe this trend through their
154
154 Chapter 8: Intra-sectoral optimal allocation of water
experience. Furthermore, some studies too revealed that soil fertility was higher in
TEFs than the HEFs.
There were three main factors creating higher soil fertility in the TEFs. The
first was sedimentation of organic material as this area was temporarily covered with
water from the reservoir below. Submergence of TEFs frequently happens during the
high rainfall season with rains filling the reservoir located just below (Daleus et al.,
1988). The second factor was higher ground water level and a higher clay content of
the soil (Daleus et al., 1988) due to siltation during the inundated period of the
downstream. This happens because of the physical distribution of the VISs (See
Figure B1 in Appendix B). A soil with higher clay content accumulates more soil
organic matter. According to Daleus et al. (1988), the most fertile part of the
command area is the TEFs of the command area. The third factor was the effect of
grazing by cattle and water buffalos in the command area (Seniviratne et al., 1994).
Most of these animals mainly graze in the tail end part of the command area due to
two reasons. The first reason is that the original layout of the VISs (See Figure A.2 in
Appendix A) allowed buffalos to wallow, in the fallow period of the paddy field next
to the TEFs (Ulluwishewa, 1991)54
. Water buffaloes grazing also controlled annual
weeds in the succeeding rice crop and enhanced the release of mineral nitrogen in
organic forms to the soil (Seneviratne et al., 1994). The second reason for the
animals to live in the tail end part of the command area is that there is a very short
distance to drinking water from the reservoir located below.
TEFs are more productive than the other two sectors. These estimates are not
robust with the study carried out by Daleus et al. (1988). According to them, the
HEFs are more productive than the other parts of the command area. However, they
also found that the soil fertility in the head-end is poorer than the TEFs as the clay
content increased with the distance from the reservoir towards the TEFs. This
difference has been estimated as 30 % (Daleus et al., 1988). This means that the sand
fraction is higher in the HEfs that affects the surface water runoff. Light material
54
“For the purpose of buffalo wallowing, a pool was maintained at the lowest point of the paddy tract.
This pool was a permanent body of water and a wide range of fish species lived there. At the onset of
the rains, the fish in the buffalo wallow migrated into the newly formed streams to the flooded paddy
fields and established colonies” (Ulluwishewa, 1991).
155
Chapter 8: Intra-sectoral optimal allocation of water 155
(i.e., clay) is transported along the drainage system leaving the heavier sand near the
area of the water source (Panabokke, 1958).
In the present study, it was found that farmers perceived the salinity problem as
a main reason for the low TE of the HEFs. This is as a result of the changes of the
original layout of the VISs. Land located immediately next to the reservoir dam,
which is known as “kattakaduwa”55
, has been encroached by the farmers over the
decades. The farmers have added this land area to their paddy fields as a result of the
disintegration of land due to increased population. The main function of the land
below the reservoir (i.e., kattakaduwa) is meant to prevent salt and ferric ions
entering into the paddy fields through seepage. This thesis shows that the low TE and
average production of the MFs is mainly because of water shortage. This area also
has a water retention problem.
One of the significant common factors (significant at 10% level) associated
with the technical inefficiency in the three sectors is water sharing (See Table 8.2).
Problems related to water sharing have become important sectoral water
management issues. It can be concluded that the effects of random factors such as the
water retention period and soil fertility, have a significant contribution to increase TE
of TEFs. The calculated marginal value of water use for rice farming at the optimal
allocation of given TE level is LKR 78350 (approximately AU$ 783) per M/ha. At
this point 1.5692 M/ha is the highest volume of water received for TEFs. The lowest
volume of water (0.7552M/ha) is allocated for MFs and 1.0637 M/ha of water is
allocated for the HEFs. In order to achieve production of the frontier level, water
should be reallocated. When technical inefficiency is zero, the marginal value of
reservoir water can be increased to LKR 159350 (AU$ 1593) per M/ha. This is a
twofold increase in MVP of total water used for rice farming (See Table 8.3) with
zero level of technical inefficiency. In practice, water allocation for different degrees
55
Kattakaduwa consist of three microclimate environments: water hole, wetland and dry upland.
Therefore, this area is suitable for growing a wide variety of plants. The water hole is called
yathuruwala wich is designed to minimise the dam seepage by raising the ground water table.
Farmers plant inedible plants along the dam to strengthen the stability of the dam. Kattakaduwa
appears to be the village garden where people utilise various parts of vegetation for purposes such as
fuel wood, medicine, timber, fencing material for housing construction and farm implements, food,
fruits and vegetables. They harvest raw materials from this vegetation for cottage industries (Peris et
al., 2008).
156
156 Chapter 8: Intra-sectoral optimal allocation of water
of TE can be estimated depending on the performance of policy instruments
implemented for the increase in TE.
8.7 CHAPTER SUMMARY
Overall productivity of reservoir water can be increased by intra-sector re-
allocation. For this purpose, it is necessary to identify inefficient sectors of the
command areas. In this study, it has been revealed that HEFs and MFs are
technically less efficient than the TEFs. The most common factor for the intra-sector
production inefficiency is the water sharing issue between the sectors.
The water is allocated on the basis of collective agreement, and all farmers
must have FO membership to contribute to the collective action organised by the
FOs. On the other hand, as it has been revealed, an understanding of soil fertility and
environmental services of VISs (especially services provided by kattakaduwa) and
the water retention period of each sector is necessary. This can be communicated to
farmers through formal or informal farmers‟ education. As such, this dilemma has to
be solved through collective action under the umbrella of FOs in order to enhance
sectoral water productivity
Historically, the sharecropping system has been instrumental for allocating
land either from head-end or MFs. This is entirely due to water constraints. However,
based on the results of this study, this traditional method of allocation cannot be
recommended.
.
157
Chapter 9: Reservoir water re-allocation and community welfare 157
Chapter 9: Reservoir water re-allocation
and community welfare
9.1 INTRODUCTION
The previous chapters of the thesis examined TE of water usage in rice farming
and CBF. Furthermore, inter-sectoral and intra-sectoral optimal allocation of water
was estimated in Chapter 7. This chapter focuses on how to put water re-allocation
recommendations from chapters 6 and 7 into practice, in order to improve the net
benefits for farmers. Through this analysis and discussion, policy makers will have
the necessary information to decide whether water re-allocation will increase the
total water productivity of VISs. Theoretically, this has been found by estimating the
consumer surplus for water demand from rice and CBF farmers. Nevertheless, the
terminology used here is consumer welfare or consumer net benefits. To avoid any
misinterpretation arising from surplus terminology (Griffin, 2006), note that
consumer surplus for water demand is described here using the terms consumer
welfare or consumer net benefits. The first part of the chapter discusses the net
benefit estimation and the remaining sections discuss internalising potential
externalities arising from re-allocation of water.
9.2 RESERVOIR WATER RE-ALLOCATION
Despite agricultural contribution to food security, income and livelihoods, the
agricultural sector is responsible for withdrawing water approximately 70% of all the
global fresh water for farming (Peris et al., 2008). In agriculture, water is allocated
for on-storage economic activities (i.e., fishery) and off-storage economic activities
(i.e., crop production). When allocation of water is non-profitable in mono-cropping,
farmers can engage in multi-crop production (Peris et al., 2008). Peris et al. (2008)
found that in rice-fish integrated systems, the farmers produce 500 kg per hectare per
one cropping season without adding any supplementary feed to the fish stock in their
rice fields. This gives 65.8% economic return per annum from the rice-fish integrated
fields. Increasing water user efficiency by incorporating multiple uses of fields is
beneficial for a number of reasons. Rice-fish integrated field systems are successful
where use of pesticides and fertiliser are minimal. The main benefits of rice-fish
158
158 Chapter 9: Reservoir water re-allocation and community welfare
farming are related to environmental sustainability, system bio-diversity, farm
diversification and household nutrition (Peris et al., 2008). However, due to the use
of chemicals in rice farming, rice-fish integrated field systems are not practised in Sri
Lanka. Furthermore, cultural reasons, such as the Buddhist philosophy which views
the rearing and killing of animals as not culturally acceptable, also prevent the
establishment of rice-fish integrated field systems. The introduction of CBF activities
is a stock enhancement activity with technology innovation in the fisheries sector
which tends to increase the marginal productivity of water. The same amount of
water that is used for rice farming could be utilised to generate more profitable non-
crop economic activities such as CBF. In practice, allocation of more water for rice
may be accepted by society. However, allocating more water for CBF production is
not a socially optimum answer in water re-allocation.
Efficient water allocation has several objectives. First, efficiency and equity of
water allocation can be considered. To do this, property rights, transaction costs and
water accessibility are used as determinants to compare forms of water allocation
(Peris et al., 2008). Ensuring food security is a social objective of water allocation
that can also be prioritised. Allocation of efficient volumes of water for use in rice
farming means moving the water for use in areas with higher economic value.
According to Molle & Berkoff (2009), water is often used in economically less
efficient, low return uses (usually agricultural). Re-allocation of water to more
efficient, high return (non-agricultural) uses would increase the total economic
welfare.
To achieve the objectives related to efficient water allocation it is important to
understand how to make decisions about water management and allocation in its
alternative uses (Peris et al., 2008). In this study, the value of water used for rice
farming and CBF development has been derived from MVP by estimating frontier
production functions, which is one of the non-economic valuation methods of
irrigation water (Peris et al., 2008). This estimation method is commonly used in
areas where water rights and the water price have not yet been established (Peris et
al., 2008). As a whole, if users cannot utilise the total water supplied by the physical
environment, then there is a need to select the right mechanism for water
management. This can be done either through demand management of water (such as
pricing, technology restrictions and water use regulations) or through supply
159
Chapter 9: Reservoir water re-allocation and community welfare 159
enhancement strategies (such as efficient structures and appropriately designed
rules). However, through supply enhancement strategies new water cannot be
materialised (Griffin, 2006). As discussed in Chapter 2 imposing water pricing was
not a successful strategy for demand management of reservoir water in Sri Lanka
(Samad, 2005). Therefore, re-allocation of water for more efficient alternatives,
within the existing institutional framework, should be implemented when possible.
9.3 LITERATURE REVIEW
EE is concerned with the amount of wealth that can be created by a given
resource base (Dennis & Arriens, 2005). A behavioural assumption of a firm is to
receive maximum profit, while minimising the cost, which depends on the action
taken by the firm (Varian, 1992). Decision making on the allocation of resources is
one of the most important actions of the planning stage of a firm. Collective
decisions (cooperative decisions) taken by groups may have an impact on individual
profit maximisation. This situation is much more crucial with common pool water
resource allocation. In the context of rural agriculture, the investment of peasant
households has trade-offs between income risks and the expected profit when they
make allocation decisions under weak or missing institutions (Mendola, 2007). Well
established collective decision making processes should consider the actual value of
the available water in order to generate high returns.
Productivity changes in water aim to increase the incentives of holding more
water in order to allocate it for other more productive uses. Clearly, water allocation
changes may decrease the quantity of water used for agriculture. However, the
reduction of water in one sector becomes an increase for another sector. For these
reasons we refer to social cost as well.
Failures of efficient resource allocation in production or in the market
mechanism generate positive or negative external effects. “External effects” is a
confused, concept in economics and it has arisen with the absence of well-defined
property rights (Verhoef, 1999). Nevertheless, Demsetz (1967) explained that
property rights are used as a primary function to accomplish internalisation of
externalities. Furthermore, there is a possibility to solve the external problems when
transaction costs are sufficiently small (Coase, 1960). Furubotn (1972) has examined
property rights analysis as a new and meaningful way to look at economic problems.
160
160 Chapter 9: Reservoir water re-allocation and community welfare
Further analysis of property rights by Swanson (2003) has also highlighted that
conservation objectives are affected by poorly defined property rights. Externalities
have both efficiency and equity aspects. Nevertheless, there is no direct mechanism
to measure the difference between the two goals of efficient resource allocation and
equitable distribution of the benefits (Verhoef, 1999). Arnason (2008) demonstrated
that a theoretically, a mixture of taxes and subsidies for the implementation of
property rights could minimise the social externalities in the fisheries sector.
Many studies of fisheries problems under various property right regimes have
revealed that a lack of property rights and the inability to find solutions to introduce
these rights were the main causes for external problems (Arnason, 2008). In this
study, the production of CBF is not generally valued in the market system. Village
societies like to produce an output that people are willing to put a price on
(desirables) or, they expect compensation to leave them with an equitable
distribution among individuals (Gough, 1957). Lack of property rights causes
externalities and the market system is only efficient if there are no externalities
(Debreu, 1959). According to Chou (2002), social capital has mutual links with
human capital and financial development. Absence of social capital in a situation
with poorly defined property rights leads to resource depletion in both private and
communal property regimes (Katz, 2000). Collective action, property rights, local
institutions, poverty and natural resources management are interconnected (Heltberg,
2000). Their implications are technology adoption, economic growth, food security,
poverty reduction, and environmental sustainability (Meinzen-Dick & Gregorio,
2004). Many developing countries have begun to decentralise policies and decision-
making related to the development, public services, and the environment (Agarawal
& Ostrom, 2001). Nevertheless, central government management of water and
aquatic resources (e.g., fisheries) often lacks the capacity to enforce property rights
and regulations on resource use (Ahmed et al., 2004). In addition to institutional
arrangements, market power for allocation of property rights through transferable
property rights is discussed in the literature (Hahn, 1984). Wingard (2000) suggests
that transferable quotas to the community minimise social impacts and internalise
externalities rather than transfer to the individuals. Suitable water allocation policy
reforms remain poorly understood. Furthermore, because of increasing competition
161
Chapter 9: Reservoir water re-allocation and community welfare 161
for water use, water allocation has to be treated in an integrated manner, considering
all purposes of water uses (Swanson, 2003).
9.4 RESULTS AND ESTIMATION OF POTENTIAL GAINS FROM
WATER RE-ALLOCATION
The water demand functions derived in Chapter 6 for rice and CBF are
employed for benefit calculations in this chapter. Most water-related benefit
estimations are based on water demand functions. Griffin (2006) demonstrated four
primary mechanisms used for estimation of policy changes. They are price rationing,
quantity rationing, supply shifting and demand shifting. The potential increase in the
water price is discussed under price rationing mechanism (Griffin, 2006)). Some
regulatory policy can impose limiting water demand (i.e., irrigating a low value crop)
but in this mechanism water rates can remain unchanged, while imposing quantity
rationing. The other basic policy mechanism is supply change. Supply shifting policy
occurs by manipulating water supply through the various factors. The fourth and
final policy mechanism is demand shifting. This motivates shifts or rotations of the
water demand curve, but impact of price rationing and quantity rationing policies
movements occur along the demand curve. An excellent example of a demand-
reducing policy in irrigation is providing low interest loans for advanced irrigation
technologies (Griffin, 2006). Demand increasing policies are less common due to
water scarcity, but in a situation like the addition of new agricultural land,
commercial enterprises, population growth, economic development, demand
increases naturally even without policy.
In the re-allocation of reservoir water, for efficient alternatives to materialise as
a policy, maximum net benefits (welfare) to the society have to be estimated. Hence,
the empirical approach to policy analysis is to measure the monetary values of
efficient allocation compared to the monetary value of proposed new costs. For this,
the change in net benefits for rice farming and CBF production has to be calculated.
If the aggregate net benefits are positive, then the water re-allocation can be accepted
as a useful policy for increasing water productivity of VISs. The condition applied
for efficiency-enhancing policy is ΣΔNB > 0 (Griffin, 2006). In connection with
welfare effects of reservoir water re-allocation two conditions are measured as:
162
162 Chapter 9: Reservoir water re-allocation and community welfare
* aTNB TNB (8.1)
TNB TNBF R
(8.2)
These two conditions indicate that the total net benefit (TNB) of reservoir
water use at the frontier level (TNB*) of production is higher than or equal to the
TNB received at the existing level of TE (TNBa). Further, the total net benefit of
water use for CBF at the frontier level ( TNBF) of production is higher than or equal
to the TNB received from water used for rice farming ( TNBR) at the frontier level of
production.
Water re-allocation in VISs can be estimated under the policy option of
demand shifting. Existing demand for water shifts with re-allocation decisions (more
details are provided in Section 3.4). Removing inefficient use of water in rice
farming is the main factor for the demand shift. Consequently, MVP of water is
increased by three times at the optimal allocation of water in the frontier level of
production. This huge increase is due to the relative price between rice and CBF
fish56
.
Source: Compiled by Author.
Figure 9.1. Farmers‟ welfare benefits of reservoir water re-allocation.
Figures 9.1.a and 9.1.b show the welfare effects of reservoir water re-allocation
of rice and CBF production respectively. And the area which represents the welfare
56
Average prices for paddy and fish are LKR 30.00 and 100.00 per kg respectively.
163
Chapter 9: Reservoir water re-allocation and community welfare 163
effects of the existing level and the frontier level of production are presented in Table
9.1.
Table 9.1
Analysis of demand shifting due to water re-llocation
Farmers‟ welfare Rice
farming
CBF
production
Total welfare
Farmer welfare at existing production levels
Farmer welfare at frontier level
Total farmer welfare
A
B
A+B
C
D
C+D
A+C
B+D
(A+C)+(B+D)
Net welfare effect of water re-allocation (A+C)+(B+D)
Demand increases due to water re-allocation changes the volume of water used
in two ways. The welfare effects that existed before re-allocation of rice farming are
shown in area A of Figure 9.1.a. Area B shows post re-allocation welfare effects at
the frontier level (See Table 9.1). In the context of rice farming, water demand
decreases by approximately 70% at the frontier level of production. This is because
of inelastic demand for water at the frontier level. This means that inefficient volume
of water is one of the determinants of the elasticity of water demand for rice farming.
The illustrative Figure 8.1.b is associated with CBF production. The areas C
and D show the before re-allocation and post re-allocation welfare effects of water
(See table 9.1). With the increase of water demand, the volume of water is increased
by approximately 32%. This is because the residual volume of water is increased
with optimal water allocation (re-allocation) in the reservoirs. Therefore, removing
inefficient usage of water in rice farming increases the volume of water which can be
used for CBF production. This means that farmers‟ TNB increases by LKR 21553
per M/ha of water used for reservoir based agriculture. This effect is shown in Table
9.2 which illustrates the details of estimation of community welfare (See calculations
and Figures I8.2 and I83 in Appendix I).
164
164 Chapter 9: Reservoir water re-allocation and community welfare
Table 9.2
Consumer surpluses for rice and CBF production with water re-allocation
Production types Consumer surplus for water demand Changes of consumer
surplus with water re-
allocation Existing level Frontier level
Rice farming
CBF production
Total surplus
38756
-20318
18438
-26712
29828
3115
12043
9510
21553
With the re-allocation of water, net MVP is positive. This estimation is shown
in both existing and frontier levels of production.
9.5 DISCUSSION: ISSUES ASSOCIATED WITH RESERVOIR WATER
RE-ALLOCATION
According to the analysis of total benefits of water re-allocation, it is possible to
make four possible conclusions:
(i) Increases in TE of current water use are essential in order to save water
in the VISs.
(ii) The total MVP (benefits) of a reservoir can be increased by five times.
Consequently, farmers‟ welfare is increased.
(iii) Increasing the total reservoir water productivity and farmers‟ welfare
are mainly attributed to the marginal value water productivity of CBF
production. Therefore, promoting CBF activities is an incentive to
efficient use of water in VISs.
(iv) The MFs are the most inefficient sector of the command area while
TEFs are more efficient. The tail-end sector is more suitable for
sharecropping than the head-end sector.
Clearly, water must be re-allocated between rice farming and CBF production
in order to achieve higher level of reservoir water productivity. Zhou et al. (2009)
have revealed that water re-allocation also has impacts on crop production and
farmers‟ income in the larger irrigation system. They further revealed that water re-
allocation from upstream to downstream areas has reduced agricultural water supply
and the area irrigated. There are two key issues which are associated with the water
165
Chapter 9: Reservoir water re-allocation and community welfare 165
re-allocation: (i) establishing water user rights among the farmers (rice and CBF) and
(ii) the establishment of a mechanism to internalise CBF externalities, which are
generated by the unequal distribution of the benefits that arise from CBF production.
These two factors are discussed in detail in the next two subsections.
9.5.1 ESTABLISHING WATER USER RIGHTS
The interdisciplinary nature of problems associated with water resource use
needs be integrated into an environmental, technical, social, economic and legal
framework. However, introducing any management system for water resources with
poorly defined property rights is likely to generate externalities which impose
indirect costs or benefits to water users and the environment, leading to an inefficient
allocation (Heaney & Beare, 2001).
Many developing countries have begun to decentralise policies and decision-
making related to the development of public services and the environment (Agarwal
& Ostrom, 2001). Nevertheless, water and aquatic resources (e.g., fisheries) managed
by central governments often lack the capacity to enforce property rights and
regulations on resource use (Ahmed et al., 2004). In addition to institutional
arrangements, market power for the allocation of property rights through transferable
property rights is discussed in the literature (Hahn, 1984). Wingard (2000) suggests
that transferable quotas to the community minimises social impacts and internalises
externalities rather than transferring them to private individuals. Suitable water
allocation policy reforms remain poorly understood. Furthermore, because of
increasing competition for water use, water allocation has to be treated in an
integrated manner, considering all purposes of water uses (Renwick, 2001).
The subject of water rights is receiving increasing attention from policy
makers due to the growing understanding that ill-defined water user rights impairs
efficient use because it creates high transaction costs (information search costs,
negotiation and monitoring) on decision making on water use (Wichelns, 2004). The
main costs of collective decision-making reviewed in the economic literature are the
so called transaction costs. Transaction costs are those costs of collective agreement
decisions or the costs of making decisions. One of the determinants of the transaction
cost is the group size which is involved in decision making. There is a large amount
of literature that discusses the effect of group size on net benefits to the group. The
early literature (Olsen, 1962) argues that small groups are less likely to be suitable.
166
166 Chapter 9: Reservoir water re-allocation and community welfare
By contrast, one of the disadvantages of large groups is the difficulty of reaching any
agreement. Hence large groups are less likely to contribute to collective decision
making than small groups (Oliver, 1998). In the case of CBF production in a VISs it
has been found that CBF activities organised by small groups have a positive
relationship with the fish yield (Kularatne et al., 2009) and such groups are the most
successful in providing benefits to participants (Senaratne & Karunanayake, 2006).
Senaratne and Karunanayake (2006) further revealed that large groups have higher
information costs (9%), but lower enforcement and monitoring costs (78%)
compared to small groups (90%) in CBF production. In the case of a single private
owner, the transaction costs are assumed to be zero. CBF activities under private
owners are minimal in VISs because of water sharing issues. However, reservoir
water is a common pool resource, where more than one user is involved, so the
transaction costs are likely to be positive (Senanayake & Karunanayake, 2006). Low
transaction costs have been linked to less conflict ridden groups, where agreement is
naturally easier to reach. Access exclusion costs are the costs of preventing outsiders
from using the resource. In principle, it could be argued that access exclusion costs
are likely to be the same for different types of management regimes. However, in
CBF production, access exclusion costs of FOs in large groups are less than small
groups (Senanayake & Karunanayake, 2006). Nevertheless, it could be argued that
for a fixed size of a resource, a larger group implies more individuals are involved in
monitoring, so exclusion costs may be lower with common pool resources. Similar
arguments arise with regard to enforcing rules about how group members or
“insiders” use the resource. A second cause of the decline of VISs management is the
declining productivity compared to alternative income sources. This arises when the
total economic gains from collective management are less than the costs. A case
study in South Africa revealed that small-scale farmers are prepared to pay a higher
price for improvement of water right systems while lower institutional trust and
income levels lead to lower willingness to pay (Speelman et al., 2010). Similarly,
FOs with medium sized groups of farmers (30-40 members) and economically
homogeneous members are better for irrigation water management (Thiruchelvam,
2010).
167
Chapter 9: Reservoir water re-allocation and community welfare 167
9.5.2 INTERNALISING CBF EXTERNALITIES
One of the main outcomes of the welfare effects of the inter-sectoral allocation
of reservoir water is increasing village community welfare mainly attributed to
increasing CBF production. The recent trend of CBF development in Sri Lanka can
be identified as transformation of a common pool resource (village reservoir) into
private property (for a small group of farmers). With subsidies for CBF activities
(i.e., subsidised fingering supply), reservoirs are facing problems linked to tragedy of
the commons documented by Hardin (1968).
In frontier level CBF production, a technically efficient solution has been
estimated. However, this estimate may not be enough to argue that on a frontier level
production is the most socially efficient solution. This is simply because of the
unequal distribution of the CBF benefits among the other water users. The farmers
who have no access to CBF production may receive neither private benefits nor
compensation for the cost of water allocation for CBF production. Some of the costs
arising from CBF development are a combination of other water uses (especially
domestic use: bathing, washing clothes in the water deficit period). A key aspect of
CBF development is capitalisation, which can lead to overcapitalisation with
increasing profit margins of CBF farmers. However, application of an individual
fishing quota system (IFQ) or individual transferable quota (ITQ) system57
on CBF
resources allocation may not be practical (Wingard, 2000; Arnoson, 2009) as the
reservoir water is a common pool regime. Therefore, rather than allocating a
transferable quota to individuals, allocating them into communities may capture the
benefits of CBF, while minimising the social impact and internalising externalities of
CBF production. In the next section, details will be provided on the applicability of
community transferable quota systems (CTQs) rather than allocating CBF activities
individually or to a selected small group of farmers (Wingard, 2000).
As a whole, society will benefit when resources are used efficiently. With
overcapitalisation, resources tend to get wasted due to overuse. Therefore, property
rights are considered as the best way to achieve the most efficient use of the
resources. Private property rights on resources ensure that the benefit of investment
57
An ITQ is quota shares that are individual is allocated as a privilege of the total annual fish catch.
Quota shares determine how the total annual fish catch is to be subdivided among individual
fishermen. ITQs are usually allocated to individuals and groups of fishermen during some designated
period of time. ITQ shares can be transferred to other parties (Wingard, 2002).
168
168 Chapter 9: Reservoir water re-allocation and community welfare
will be received by the investor. Some economists (e.g., Arnason, 2005; 2009) argue
that the ITQs must generate economically optimal results, but it is a self-centred
utility maximising Homo economicus practice described in neoclassical economic
theory (Wingard, 2000). Especially in the case of CBF production, the allocation of
an ITQ system makes entry into the fishery more difficult: some reservoirs
accommodate all farmers in the FO in the CBF group in a particular culture cycle.
There should be a mechanism which fulfils sustained participation of communities in
CBF activities, which will minimise adverse economic impacts on such
communities. For this reason, CTQs could accomplish many of the economic and
biological goals, while minimising negative social impacts (Wingard, 2000).
Community level agricultural management is very common in Asia.
Furthermore, community fisheries management is widespread in many non-
industrialised societies (Wingard, 2000). The CTQs have many potential advantages
for addressing social shortcomings of efficiency. Under a CTQ system, a large
number of people would be able to remain in the fishery at least on a culture-cycle
basis.
CTQs of CBF production
Under a CTQ system for CBF, a group of farmers would be able to get
involved in CBF activities based on a culture-cycle. Groups of farmers for CBF
could be selected among the farmers who are willing to get involved in CBF
activities. This may determine the total number of farmers in the group. Under a
CTQ system, there are two factors which may maximise the economic benefits while
minimising cost impacts:
(i) If the group of farmers is considerably large (small group favours group
stability), they can be given a community quota on the basis of the
culture cycle. The total group can be divided up into smaller groups.
Group one could be given an opportunity in the first culture cycle and
the second group could be given an opportunity in the next cycle and so
on. This system could be rotated for each consecutive culture-cycle.
169
Chapter 9: Reservoir water re-allocation and community welfare 169
(ii) Depending on the spatial MVP of rice farming, one group of farmers
with higher MVP of rice farming could cultivate rice, while others who
have a lower value of MVP could become involved in CBF, especially
during the share cropping seasons.
Selected communities (group of farmers) would provide access rotationally.
This would contribute to maintaining and improving social and economic stability
and would avoid economic dependency58
of the whole communities on one form of
production. Social capital which is a valuable asset in the context of a village
community could be further strengthened through economic independence. In
addition to communal stability, other sectors of the rural economy such as agriculture
and livestock59
would also benefit. This would also strengthen social capital
throughout the village community. Social capital exists with the form of obligation,
expectation and trust (Grafton, 2005; Teraji, 2008). Obligation and trust help farmers
to meet their goals. Information is another form of social capital which reduces the
uncertainty of CBF production. Norms and sanctions are also part of the social
capital. They allow for predictability of behaviour which reduces transaction costs
(Coleman, 1988; Grafton, 2005). Improvement of social capital may lead to
communal stability and would contribute to the long term social and economic
wellbeing of village communities.
9.5.3 CO-MANAGMENT AS A MECHANISM FOR WATER RE-ALLOCATION
Social capital plays an important role in enhancing trust and co-operation
which would reduce the misuse of the available resources among the resource users
(Grafton, 2005). As Teraji (2008) has stated, a fully protected property rights system
can achieve a higher level of trust, while unguaranteed property rights will remain at
a low level. Therefore, property rights play an important role in establishing the trust
and social capital among communities by increasing cooperation among the resource
58
FOs are highly politicised due to economic inter-dependence on the politicians and other people
(money lenders). This is done in order to obtain financial support for maintaining and repairing the
reservoirs, construction village roads, supply electricity to the village. This dependency can be
removed by increasing water user efficiency. The share of CBF production can be collected for FOs to
facilitate such common activities in the village. The CBF farmers paid 5% to 15% of the total income
of CBF production in 2009. However this is entirely a decision made by the individual FOs in the
kanna meeting. This system can be further improved and formalised under CTQs with well defined
property rights of intra-sector water allocation
59
The inefficient sector of the command area could be used for cattle grazing which will generate
positive externalities for CBF development of downstream reservoirs of the cascade.
170
170 Chapter 9: Reservoir water re-allocation and community welfare
users. Benefits of cooperation include the avoided costs of social conflict and
avoided externalities imposed by others. Wade (1987, p.98) states that the “Main
factor explaining the presence or absence of collective organisation is the net
collective benefit of the action.” More specifically, Wade (1987) focuses avoiding
external costs through cooperation. He argues that cooperation occurs in villagers
where the net benefits of cooperation are highest. Since the relative transaction and
exclusion costs will be similar for each village, the main cost is the relative benefits
of cooperation or the avoided external costs of non-cooperation. The benefits of
cooperation are highest and costs are lowest when benefits are equally distributed to
all groups gained from collective management. This is often violated in the case of
large irrigation systems where some farmers are much closer to the water source
(head-enders) while other groups are much further away (tail enders). Cooperation is
unlikely to work where the group contains both head-enders and tail-enders since
head-enders lose out as cooperation increases and their water use is limited.
Therefore, from a social capital point of view, it can be suggested that current top-
down resource management should be redirected towards a „co-management‟
approach (Grafton, 2005).
It has been shown in many parts of the world that co-management and
community-based management of natural resources could provide effective
alternatives for natural resources management (Wade, 1987; Hannesson, 1998).
Current research suggests that there are emerging characteristics which are central to
developing and sustaining institutions that support successful co-management
arrangements. Pinkerton (1989) and Ostrom (1990) have summarised and
documented some of those key conditions necessary to maintain successful co-
management institutions. From their work, co-management is likely to succeed in
resource systems where boundaries are clearly defined, membership is clearly
defined, the user group is cohesive, the user group has prior experience with the
organisation, and the benefits of management exceed costs. Additional criteria are
that there will be participation in management by those who are affected, due to the
enforcement of management rules under which these co-management approaches are
enforced. Also the user group has legal rights to organise, so that there is co-
operation and leadership at the community level. Furthermore, there is
171
Chapter 9: Reservoir water re-allocation and community welfare 171
decentralisation and delegation of authority, and there is co-ordination between the
government and the community.
In Sri Lanka, the inland fisheries development programme came to a standstill
with the decision of the government to terminate state patronage in 1990, on
religious grounds for this important sector which had been contributing 20% to the
total fish production in the country. This government policy decision has been
reversed and since 1994, development of Inland Fisheries and Aquaculture has been
given high priority because of its value as a cheap animal protein for the rural
community (Amarasinghe, 1998). It also has the potential to increase income and
employment opportunities to the people and to function as a source of foreign
exchange to the country (Sivasubramaniam & Jayasekara, 1997) .
After withdrawal of state support for inland fisheries development, annual
inland fish production declined dramatically. This decline was shown to be a result
of "growth over-fishing" (Amarasinghe & De Silva, 1999). This resulted due to the
use of small mesh gillnets in the absence of state-sponsored monitoring procedures.
This indicates that under the existing state management procedure, it is necessary to
have a Centralised Management Authority for Inland Fisheries management in Sri
Lanka (Amarasinghe & De Silva, 1999). In reservoirs with “organised” fishing, the
communities themselves have developed regulations through community based
management strategies. In such reservoirs overexploitation of fish stocks was not
evident even after state-sponsored monitoring procedures were suspended
(Amarasinghe & De Silva, 1999).
Based on these studies, an alternative approach is recommended for the
management of reservoir capture fisheries in Sri Lanka. It is recommended that
Government and resource-users have equal responsibilities in making decisions for
the management of reservoir fisheries (Amarasinghe, 1998). This acknowledges the
fact that farmers‟ involvement is equally important for the successful co-
management system as primary stakeholders.
It has been found that participation rates for collective action (FO activities) are
a positive factor for increasing TE in rice and CBF production in the case of reservoir
based irrigation in Sri Lanka. However, recent studies on major, medium and minor
irrigation systems in the Kurunagala and Anuradhapura districts of Sri Lanka have
found that the participation rate for FO activities is 38% because of lack of
172
172 Chapter 9: Reservoir water re-allocation and community welfare
accountability and transparency of FOs (Thiruchelvam, 2010). As a result,
Thiruchelvam (2010) recommended establishing strong linkages between FOs
(primary level stakeholders) and water authorities (responsible institutions) for
successful irrigation management. According to Khalkheili & Zamani (2009), the
establishment of co-operation with water authority operators will enhance farmers‟
participation in irrigation management. Furthermore, co-management practices
should promote active involvement of immediate actors to the resources for their
management rather than relying on institutional hierarchy. In the case of reservoir
water management, the effect of institutional hierarchy is shown in Figure I8.1 in
Appendix I.
Markets are another supplementary factor in the co-management of VISs. Rice
production is more popular than CBF production at the village level. However, part
of the production of rice is marketed by farmers since rice cultivation is also an
income generating activity. CBF on the other hand is mainly produced for the
market. Therefore, allocating irrigation water has to take into account the market
behaviour of these goods. The value of the water may depend on MVP. Therefore,
essentially in addition to institutions and primary level resource uses, market
motivation is another factor that should be considered in the decision making process
of reservoir water allocation (See Figure 9.2).
Source: Compiled by Author.
Figure 9.2. Co-management settings for reservoir-based agriculture in VISs
173
Chapter 9: Reservoir water re-allocation and community welfare 173
There is a possibility for all farmers in the village to be represented in FOs.
Village farmers and the village level agriculture and fisheries officers, who represent
institutions, are identified as primary level actors. The FOs represent the farmers
while ARPAs and AEO are represented by the government officials. Bidirectional
arrows in Figure 9.2 show the necessary direction of trust and cooperation. Based on
the strength of these two institutions and the power of decision-making, it will be
possible to implement a successful co-management strategy with water re-allocation.
Finally, it can be concluded that the combination of sharing responsibility of water
management, between responsible institutions and primary level stakeholders, with
the motivation of the market forces for profitable alternative water uses, is a
practicable mechanism for reservoir-based irrigation water management which can
be achieved for efficient output and higher MVP of water in VISs.
9.6 CHAPTER SUMMARY
Reservoir water productivity can be increased by five times by increasing TE
up to the frontier level of production. The only necessary requirement for water
saving is that water is used efficiently in rice farming. Increasing reservoir water
productivity should be undertaken from a practical point of view. It should ensure
water user rights of VISs for multiple users. It is important to reintroduce CTQs in
CBF production in order to select CBF farmers. Co-management of water resources
is the best institution for reservoir water management. This means sharing
responsibility between local level government institutions (ARPAs and AEO of
DAD and NAQDA) and FOs. Increasing farmers‟ economic benefits through
efficient water re-allocation in reservoir-based agriculture will remove village
dependency on external sources (subsidy, political support). The FOs can act as the
main village level institution for reservoir water management and decision-making
with the support of relevant formal institutions and the market guidance of reservoir
water demand.
175
Chapter 10: Concluding remarks 175
Chapter 10: Concluding remarks
10.1 CONCLUSIONS
The analytical core of this thesis shows that the TE of existing allocation of
water can be improved and that water can be optimally re-allocated in VISs in Sri
Lanka. The productivity of water can be increased in VISs through a combination of
sharing responsibility in water management between responsible institutions and
primary level stakeholders based on market demand. In this context, it is clear that
existing institutions (i.e., FOs) need to be re-organised, while market forces are used
to guide the efficient re-allocation decisions. In such a scenario, farmers would be
motivated to manage their water demands, not only through enforcement of rules, but
also through the development of an understanding of the importance of efficient
water use in rice farming in order to increase reservoir water productivity.
This high level of water productivity can be achieved by the development of
CBF in VISs without altering the volume of water used in rice farming. CTQs can be
introduced to the CBF production system for the selection of CBF farmers. Co-
management of water resources is the best institution for reservoir water
management with market guidance from the MVP of water. An increase of farmers‟
welfare, through reservoir-based agriculture, would remove village dependency on
external sources (subsidies and political support). The FOs could act as the main
village level institutions for reservoir water management and decision making in
collaboration with the relevant formal institutions. This PhD has demonstrated that
instead of allocating water based on farmers‟ experience (and haphazardly) they can
now base their decisions on efficient water use with a view to increasing water
productivity of VISs in order to enhance their incomes and the welfare of the
community. The next few sections of this chapter present the summary key findings
and conclusions, the limitations, recommendations for policy implementation and the
directions in which the study may be extended in future work.
10.2 SUMMARY, KEY FINDINGS AND DISCUSSIONS
Chapter 5 estimated a theoretically consistent translog stochastic production
frontier for rice farming and examined the factors influencing efficient allocation of
176
176 Chapter 10: Concluding remarks
water in reservoir-based agricultural production. The mean TE of rice farming in
VISs was 0.73, which is higher than the mean value of TE for rice farming in Asia,
generally which is 0.72 (Bravo- Ureta et al., 2007). It was shown that rice production
increased by 3.2% with a 10% increase of water in VISs. As expected, the individual
volume of water used had a positive effect on rice production, which was significant
at 1% level. The most influential factors for TE were FO membership and the
participatory rate in FO activities (collective action) while paddy field location, water
sharing and landownership issues decreased TE. These results proved that low
efficiency among rice farmers is mainly due to inefficient use of water. Similar
conclusion was reached by Bravo-Ureta et al. (2007). This chapter concluded that the
enhancement of co-operative arrangements (collective action) is effective in
increasing the TE of allocating irrigation water for rice farming. This research
showed that it is possible to improve TE in two ways. First, by formalising
transferability of land ownership and hence water user rights. Second, enhancing
institutional capacity of FOs in order to solve locational water sharing issues
In Chapter 6, the estimated consistent stochastic production frontier shows that
the mean TE of CBF in these VISs was only 33%. Furthermore, 54.2% of CBF
farmers are beyond the mean TE level. This is lower than the mean TE of 0.57 for
existing aquaculture systems in Asia. None of studies has used which water as an
input variable. In the context of CBF production, the residual volume of water used
for CBF production was highly significant (at the 1% level). With respect to the 10%
increase of residual water, CBF production increased by 4.5%. Use of group labour
in CBF production is insignificant. The large group was found to be inefficient
compared to small groups. This result contradicts previous research undertaken by
Dey et al. (2000), Kareem et al. (2009), Singh et al. (2009). Furthermore, fast
growing fish species are worse off in terms of growth than slow growing species due
to nutrient levels in the water.
The effect of random factors (i.e., weather) on TE of CBF production is
expected since water use in CBF production is entirely dependent on monsoon
rainfall in the VISs. However, the estimated results show that inefficiency is
considerable. Group stability and the number of cattle and water buffalos in the
catchment were the most significant (significant at 10% level) factors which have
positively influenced TE. On the other hand, the time spent meeting officials (i.e.,
177
Chapter 10: Concluding remarks 177
fisheries extension officers) and supplying subsidised fingerling for CBF are shown
to influence TE negatively. As such, there is a possibility of increasing CBF
production from 2,715 kg to 8827 kg per reservoir by operating at full efficiency. In
order to achieve these efficiency gains, attention has to be paid to strengthening
group stability, improving accessibility of extension services and promote a
mechanism for maintaining independent investment on CBF without depending on
subsidies and to ensure well defined water user rights.
Chapter 7 focused on the optimal allocation of water between rice farming and
CBF. The optimal value of water was estimated using a shadow value, while for rice
and CBF, the value was based on output market prices. The estimated mean capacity
of VIS is 5.421 M/ha. According to present estimations, inefficient use of water for
rice farming is approximately 32%. Efficient use of water in rice farming will enable
an increase in the residual volume of water by approximately 53%. At the given level
of TE of rice farming, the estimated optimal value of water used in rice and CBF
production is LKR 20660 per M/ha. This can be increased up to LKR 71055 per
M/ha of water which is used for CBF and rice production by removing water use
inefficiency in rice farming. Successive mechanisms, which could increase efficient
(up to the frontier level) water use in rice farming, would increase the total MVP of
reservoir water by threefold. These results are consistent with the results of Farolfi
and Perret (2002). Re-allocation of water could be made realistic with flexible water
user-rights. However, in the context of the existing water allocation mechanism in
VISs, individual water user rights do not exist for inter-sectoral water allocation.
This further supports the research findings of Kulindawa (2000). Inefficient inter-
sectoral allocation of water occurs due to a lack of active community involvement in
collective action and due to conflicting and weak institutional capacity. This finding
was also reported by Kashaigili et al. (2003).
Chapter 8 investigated intra-sectoral water allocation in the three main sectors
of the command area of VISs. Investigation of current water use patterns and the
estimation of intra-sector optimal allocation of water have enabled the researcher to
make allocation decisions, which will eventually increase sectoral water productivity.
The calculated marginal value of water used for rice farming at the existing level of
TE is LKR 78350 (approximately AU$ 783) per M/ha. At the frontier level of
production, the marginal value of reservoir water can be increased up to LKR
178
178 Chapter 10: Concluding remarks
159350 (approximately AU$ 1593) per M/ha while the volume of water use in HEFs
and TEFs can be reduced by 10% and 23% respectively. This can be re-allocated for
MFs where lower levels of water usage exist. By re-allocating the water, the overall
sectoral water productivity can be increased twofold. The results from this study
show that the most technically inefficient and lower productive sectors are the MFs.
Charavorthy and Roumasset (1991) presented a theoretical overview of the head-tail
dilemma of water allocation. The empirical estimation of inter-sectoral water
allocation by Ekanayake and Jayasooriya (1987); Charavortty et al. (1995) and
Wichelns (2002) confirmed that HEFs receive more water, while TEFs receive less
irrigated water in the main irrigation systems. However, these findings and the
theoretical explanation in the literature are contradicted by the results of this PhD
study. Furthermore, Daleus et al. (1988) suggested that there is a perfect negative
relationship between yield and distance to the reservoir dam from individual paddy
fields. Nevertheless, this study found a U shape relationship between head-end,
middle and TEFs. The reason for this relationship was discussed in Chapter 8.
As a whole, the value of reservoir water can be increased by re-allocating water
through minimising the inefficient volume of water use in the different sectors. The
main problem of the intra-sector water allocation at present is that farmers have no
proper motivation to engage effectively in field-level water management. However,
results from this study show that intra-sectoral efficient use of water increases the
inter-sectoral reservoir water allocation, which ultimately increases the total reservoir
water productivity, thereby helping the farmers.
Chapter 9 discussed the economic gains of re-allocating water which was
necessary to estimate farmers‟ welfare in comparison to that of competing water
users. The allocation of water in VISs is assumed to be at sub-optimum levels when
water usage is inefficient. The total potential expansion of CBF with optimal
efficiency in water allocation in rice farming, which is estimated by measuring the
net farmers‟ welfare, is LKR 21553 in rice and CBF production per M/ha of reservoir
water. These results are consistent with results of Zhou et al. (2009) which found
that water re-allocation impacts on crop production and farmers‟ incomes. The total
net benefits of CBF production are lower than the total net benefit of rice due to the
low levels of given TE. However, it can be concluded that the total net benefit of
reservoir water re-allocation can be increased which is mainly attributed to the
179
Chapter 10: Concluding remarks 179
marginal value of water productivity of CBF production. Therefore, the main
constraints to increasing reservoir water productivity by incorporating CBF
production are sharing economic benefits of CBF production and the selection of
CBF farmers. These two constraints, which arise with water re-allocation, are due to
the absence of well-defined water user rights for CBF production and the non-
existence of transferability of water user rights in rice farming.
10.3 POLICY IMPLICATIONS
Policies are considered as alternative institutions (Griffin, 2006). Water re-
allocation aims to allocate water for enhancement of the total reservoir water
productivity. The preceding analysis of MVP of water shows that the optimal
allocation of water between rice and CBF production enables increases in reservoir
water productivity.
After political independence in 1948, agricultural policies in Sri Lanka were
mainly focused on food security, self-sufficiency by rice and import-substitution
practices. Water supplies for rice production were mainly based on reservoir-based
irrigation systems. Nevertheless, it was found that water productivity in VISs were
very low (0.07 $/m) compared to other major and medium irrigation systems
(Thiruchelvam, 2010). Therefore, the investigation of TE and factors influencing
technical inefficiency were important for policy-making on optimal allocation of
water in VISs.
There are two policy objectives in the ten year development policy framework
of the inland fisheries and aquatic resources sector in Sri Lanka: (i) to improve the
nutritional and food security of the people by increasing the national fish production,
and (ii) to increase employment opportunities in fisheries and related industries and
the socio-economic status of the fisher community (Anon, 2007). Furthermore, the
government expects to increase annual CBF production to 74,450 tonnes in 2016
from 33,180 tonnes in 2004 by increasing fish production in VISs.
The development plan assumes inadequate stocking, low level of social
acceptance, religious and cultural prejudices, environmental concerns and the
instability of government policies as constraints to the development of the CBF
sector. In addition to these constraints, lack of proper water allocation between
sectors that is based on well-defined water user rights among multiple users and
180
180 Chapter 10: Concluding remarks
inappropriate institutional responsibility and coordination between the Ministry of
Fisheries and DAD have considerable impact on the development of the sector.
Therefore, the policy can be revised to strengthen water user rights, enabling the
transfer of water rights within the existing institutional framework of the Ministry of
Fisheries and Aquatic Resources and the Department of Agrarian Services.
Promoting CBF is an incentive to use water efficiently in rice farming if all rice
farmers are represented in the CBF group. In this study, it has been revealed that
head-end and MFs are technically less efficient than TEFs. Historically, the
sharecropping system has been instrumental for allocating land in HEFs or MFs. This
is entirely due to water constraints. However, based on the results of this study, this
traditional method of allocation cannot be recommended
The empirical estimates of TE in rice farming from VISs have proven to be
useful. Especially with respect to the water resource allocation, it is important for
policy makers to know by how much agricultural production can be increased by
increasing its TE without altering available water, given the technology involved. It
has been estimated in this research that for the same quantity of input, it is possible to
increase output by up to 28% in rice farming in VISs. It also has been found that
enhancing the institutional capacity of FOs will further improve TE. Furthermore, it
has been shown that if it is possible to put in place a system to transfer land
ownership and hence water user rights to solve locational sharing issues, this will
improve the institutional capacity of the FOs and will thereby help to reduce
technical inefficiency. Overall findings of this research show that the total benefits of
the reservoir water can be increased by improving water use efficiency in rice
farming and improving the TE of CBF production. Therefore, it is logical to argue
that the anticipated policy should focus on increasing the reservoir water
productivity, which can be achieved through water use efficiency in rice farming and
TE of CBF production. Furthermore, this PhD study identified six important areas
which need to be addressed in order to achieve a higher level of water productivity of
VISs in Sri Lanka. The six areas are listed below.
1. Efficient use of irrigation water increases the residual volume of reservoir water,
which can be used for multiple purposes.
2. At present, group labour used in CBF is over utilised. Therefore, there is a need to
identify mechanisms for the efficient use of labour in CBF production.
181
Chapter 10: Concluding remarks 181
3. Fast growing fish species have no positive impact on CBF production due to
inadequate nutrition in reservoir water. Therefore, the possibility of introducing a
cost effective integrated farming system or promoting artificial fish feeds should
be explored.
4. The inter-sectoral water allocation mechanism is made effective by introducing an
acceptable transferable water user rights system.
5. The MFs of the command areas are less efficient and less productive. Water
management at the field level should be enhanced by increasing farmers‟
motivation to improve their water management practices in the command area.
6. Total benefits of reservoir water can be increased by solving two constraints:
establishing water user rights for CBF production and by ensuring transferable
water user rights are established in rice farming.
Most of the issues relating to the enhancement of water productivity must be
dealt with using the existing WUAs (i.e., FOs) on an apolitical basis. Co-
management of the water resources is the most appropriate mechanism that can be
recommended where a combination of both farmers and formal institutions would
share the management responsibilities in the market environment.
Based on the findings of the study, it is clear that these 6 options are ideal for
dealing with the water re-allocation issues. Various biological productivity-related
problems, such as a lack of an effective means of selecting suitable reservoirs and a
lack of guaranteed supply of fingerlings for stocking (De Silva, 2003) have
constrained CBF development in Sri Lanka since its beginning in the 1980s.
Furthermore, weak institutional linkages, lack of legislation and poorly planned
social mobilisation procedures were also responsible for the unsustainability of CBF
activities. Some of these constraints, especially at the grassroots levels, have been
overcome through concerted efforts of active biological research to some extent and
the barriers at the institutional levels can be solved. Therefore, CTQs are proposed as
a possible policy instrument within the framework of the co-management strategy
which can be implemented through DAD and NAQDA.
182
182 Chapter 10: Concluding remarks
Co-management strategy
Co-management of water resources is the most appropriate mechanism to be
implemented in combination with market forces. Established water user rights and
transferable water user rights must be initiated at the existing village level
institutions (i.e., FOs). Ministries of Agrarian Services and Fisheries and Aquatic
Resources should formulate relevant policies for further strengthening the relevant
institutions at the national level. The responsible legal body for solving water
allocation issues with FOs is the network of agrarian services. The NAQDA should
facilitate the technical aspects of CBF production. Legislation of FOs should make
provisions to enable expansion of their membership beyond the current holdings.
Table 10.1
Decision-making of kanna meetings in the framework of co-management strategy
Decisions Agricultural activities Activities of CBF production
1. Cleaning canals,
bunds and sluices
Cleaning and construction of small
canals, bunds and sluices by the
relevant farmers
Decide on fish culture and repair
reservoir, remove logs, fill pits for
brick making and fill the wells dug
during the low rainfall season
2. Duration of water
supply
Selecting the method of cultivation,
type of paddy, place of buying,
price, transport & the quantity
Select the group of fish culture,
species, place of buying, price,
transport, and the quantity
3. First date of water
supply
Use rainwater for ploughing fields
in order to save reservoir water
Stock fingerlings based on the
level of water in the reservoir
4. Broadcasting of
paddy and
protection rice
crop from birds,
etc.
Release less water for agriculture to
save reservoir water, make fences,
remove domestic animals (e.g.,
cattle), prevent disease, pesticide
and fertiliser contamination
Prevent escape of the fingerlings
from sluice, outlet, feeder streams,
and birds
Allow cattle and other animals to
graze in the catchment area
5. Last date of water
issue
Decide to close the reservoir sluice
Decisions can change based on
special requests with the approval
of FOs and Divisional Office.
Decisions should be flexible
183
Chapter 10: Concluding remarks 183
This is aimed at strengthening group participation. Collaboration of these two
institutions with FOs would considerably improve collective action of the farmers
and will further advance the co-management strategy of production as shown in
Table 10.1.
CTQs as a policy instrument
The idea of implementing CTQs is not a new phenomenon in reservoir-based
agriculture. However, the CTQs need to be reinstated and re-established as formal
institutions under the umbrella of a FO system, in order to increase total productivity
of reservoir-based agriculture.
The development plan is based on the assumption of absence of inadequate
stocking, low levels of social acceptance due to religious and cultural prejudices,
environmental concerns and instability and uncertainty of government policy. These
factors are all constraints in varying degrees. In addition to those constraints, lack of
proper water allocation between sectors, which is based on well-defined water user
rights and inappropriate institutional responsibility and coordination between
fisheries and agrarian authorities, has a considerable impact on the further
development of the sector. Therefore, the policies needed most at present are those
that strengthen water user rights, which will thus enable the transfer of water use
rights between sectors within the existing institutional framework established by the
Ministry of Fisheries and Aquatic Resources and DAD.
Policies should be formulated with the proper understanding of the factors
influencing technical inefficiency of intra-sectoral water allocation. The common
reason for the intra-sector production inefficiency is due to the water sharing issue
between the sectors. The water sharing issue in the different sectors can be solved by
taking into consideration the different positive factors, which influence TE. The
water is allocated based on collective agreement, and all farmers must have FO
membership to contribute to the collective action organised by the FOs. On the other
hand, as it has been revealed, understanding the soil fertility and the environmental
services of VISs (especially services provided by kattakaduwa) is necessary and has
to be communicated to farmers through formal or informal farmer education. For the
MFs, collective management and collective actions organised by the FOs are
important to increase water use efficiency. These issues should be addressed by FOs
to avoid any negative influence on the level of TE in enhancing reservoir water
184
184 Chapter 10: Concluding remarks
productivity. As a technical measure, in order to improve the efficiency of the head-
end re-establishment of the kattakaduwa is recommended in VISs. Accordingly;
policies can be directed to promote multiple agricultural activities (i.e., animal
husbandry) in areas where less efficient sectors such as MFs exist. The other
technical answer for the sectoral water shortage is the establishment of original
settings of the command area, such as the re-construction of multi canal systems for
the three sectors of the command area instead of a mono-canal operation. There is a
possibility of encouraging integration of a crop-animal system in the watershed areas
and the catchment areas within a framework of an integrated sustainable agricultural
organic CBF system. This will also solve the issue of nutrient deficiency problems of
stocking fast growing fish species. This reduces the need to feed fish artificially.
Selection of farmers for CBF production can be streamlined by re-introducing
CTQs. This is already being practiced by rice farmers in Sri Lanka and is known as
the thattumarau system. This system ensures transferable water user rights for all
farmers between the different consecutive cropping years and culture cycles. This
solves the issue of over utilisation of the group labour in CBF production. In
addition, lack of institutional ability, the higher level of fish poaching and water use
conflicts between rice and CBF farmers could be solved by giving an opportunity for
all farmers to be involved in CBF production through CTQs. At present most fish
poaching occurs due to villagers having no opportunity to participate in CBF
production (See Figure 1.1). CTQ systems facilitate maximum involvement in both
rice and CBF production. Furthermore, farmers are likely to be motivated by more
efficient intra-sectoral water management due to the increased benefits received from
CBF production. This may also solve the problem of inefficient use of water in the
MFs of the command areas. Therefore, establishing both water user rights for CBF
production and ensuring the existence of a transferable water user rights structure for
rice farming can be achieved by establishing a CTQ system in reservoir-based
agriculture in Sri Lanka.
Implementation
Established water user rights and transferable water user rights must be
initiated at the existing village level institutions (FOs). The Ministry of Agrarian
Services and the Ministry of Fisheries and Aquatic Resources should formulate
relevant policies for further strengthening relevant institutions. The responsible legal
185
Chapter 10: Concluding remarks 185
body for solving water allocation issues with FOs is the DAD network. NAQDA
should facilitate the technical aspects of CBF production. Collaboration of these two
institutions with FOs would considerably improve collective action of farmers and
would advance the co-management strategy further. Selection of farmers for CBF
production in particular VISs can cope with the re-introduction of CTQs and as
mentioned earlier are already being practised in rice farming60
. A Thattumaru system
can be successfully used for the selection of CBF farmers without introducing new
selection criteria as it is inherently practised by village farmers.
In addition, there is a possibility to encourage livestock farming in the
watershed areas within a framework of integrated agriculture (Prein, 2002) for
sustainable organic CBF. As such, a revival and re-establishment of such integration
of a crop-animal system as formal institutions under the umbrella of a FO system
which is already in existence is useful in order to increase the total productivity of
reservoir-based agriculture.
10.4 LIMITATIONS AND FUTURE DIRECTION OF RESEARCH
There are two main limitations of the study. First, the results of the allocation
estimates are limited only to VISs, which have unique economic and hydrologic
characteristics. This analysis of inter-sectoral water allocation models is therefore not
applicable to other irrigation systems because of their differences in scale and other
administrative characteristics.
Second, the MVP analysis assumes that water is used only for two uses.
However at present reservoir water is being used for many purposes (domestic uses,
animal husbandry and cottage industry). Any steps which are taken to increase the
residual volume of water will be beneficial for all the other alternative uses.
60
Thattumaru is the rotational cultivation of one plot of land by several children within one
household. One of the children cultivates the entire plot for one season, the next season
another son/daughter will cultivate the entire plot, etc. Thattumaru prevents the division of
land into smaller and smaller plots. In each village, Thattumaru is applied on average by 4 or
5 families with small landholdings. Thattumaru is practiced to prevent conflicts among
children. Thattumaru is subject to creative arrangements, such as selling one‟s share to one‟s
brother or sister, or in combination with sharecropping or a private lease. Thattumaru is most
likely to be practised when further fragmentation of lands within a family is no longer
feasible.
186
186 Chapter 10: Concluding remarks
Although this PhD thesis reliably demonstrates the principle of optimal allocation of
water, it does not take into account longitudinal changes in TE.
In addition, the following specific limitations are recognised in this study.
The analysis of TE estimation of VISs does not estimate the impact of
government fertiliser subsidy programmes on water user efficiency in village
irrigation systems.
The functions and legal strengthening of FOs (which is the main institution in
VISs management) should further be investigated for better water management
of VISs.
CBF production is organised as a group activity. Therefore, group TE was
estimated for CBF production. Individual characteristics of farmers are not
represented in TE estimation.
Effects of some environmental factors such as trophic characteristics (i.e., water
quality; alkalinity, conductivity), water quality and catchment characteristics,
which could also have positive effects on CBF yields and TE, were not included
in the estimated model.
A sharecropping system, which is known as Bethma, is practised for allocating
land either from HEFs or MFs, entirely due to water constraints. However, from
the results of this study, it is not possible to recommend this traditional way of
allocating land for cropping practices alone. This study has not undertaken the
estimation of costs and benefits of water allocation in the tail end with
conveyance losses.
All analyses were based on the two extremes: given level of TE and the frontier level
of production. The changes in water productivity were not estimated taking into
account dynamic TE scenarios. Therefore, a prospective case study should
concentrate on and investigate longitudinal variations in TE, water allocation and
benefit sharing of multiple uses of water in VISs.
Bibliography 187
Bibliography
Abeyaratne, S., & Perera, J. (1984). Changes in irrigation management in small
communities: A tank and irrigation systems in Monaragala District, Sri Lanka.
Journal of Agrarian Studies, 5(1), 71-84.
Agarawal, A. (2001). Common property institution and sustainable governance of
resources. World Development, 29(10), 1649-1672.
ADB (2001). Water for all. The water policy of the Asian Development Bank: Asian
Development Bank. The official policy paper approved by the Asian
Development Bank Board of Directors on 16 October 2001.
Aheeyar, M. M. M. (2001). Socio-economic and institutional aspects of small tank
system in relation to food security. In H. P. M. Gunasena (Eds.), Food security
and small tank system in Sri Lanka (pp.64-78). Colombo, Sri Lanka: National
Science Foundation.
Aheeyar, M. M. M., Henegedara, G. M., & Rupasena, L. P. (2005). The cost of
production of rice in Kegalle and Kurunagala Districts of Sri Lanka. Research
study 115. Hector Kobbekaduwa Agrarian Research and Training Institute,
Colombo, Sri Lanka:
Ahmed, M., Viswanathan, K. K., & Valmonte-santos, R. A. (2004). Collective
action, and property rights in fisheries management. Focus 11, Brief 7 of 16.
CGIAR System-wide Programme on Collective Action and Property Rights
(CAPRI), International Food Policy Research Institute, Washington DC.
Aigner, D. J., Lovell, C. A. K., & Schmidt, P. (1977). Formulation and estimation of
stochastic frontier production models. Journal of Econometrics, 6(1), 21-37.
Al-hassan, S. (2008). Technical efficiency of rice farmers in Northern Ghana. AERC
Research Paper, African Economic Research Consortium, Nairobi, 1-41.
Alam, F. M., & Murshed-e-Jahan, K. (2008). Resource allocation efficiency of the
prawn–carp farmers of Bangladesh. Aquaculture Economics & Management,
12(3), 188-206.
Amarasinghe, U. S. (1998). Reservoir fisheries management in Sri Lanka:
Achievement, mistakes and lessons for future. International Review of
Hydrobiology, 83, 523-530.
Amarasinghe, U. S., & De Silva, S. S. (1999). Sri Lankan reservoir fishery: a case for
introduction of co-management strategy. Fisheries management and ecology,
6(5), 387-399.
Amarasinghe, U. S. (2008). Culture-based fisheries in village reservoirs of Sri Lanka,
SIL News, 52, 11-12. www.limnology.org/news/silnews52.pdf
188
188 Bibliography
Amarasinghe, U. S., & Nguyen, T. T. T. (2009). Culture-based fisheries
development: Successfull secondary use of a water resources for enhancing rural
farmer income through fish production with special refeence to Sri Lanka. In D.
S.S. De Silva & F Brian (Eds.), Success stories in Asian aquaculture (pp. 100-
126). Network of Aquaculture Centres in Asia Pacific, Bangkok, Thailand.
Springer.
Amarasinghe, U., Samad, M., & Anputhas, M. (2005). Spatial clustering of rural
poverty and food security in Sri Lanka. Food Policy, 30, 493-509.
Anon. (1964). Administrative report of the director of fisheries for 1962-1963. 84,
Colombo: Government Publication Bureau.
Anon. (2007). Ten year development policy framework of the fisheries and aquatic
resources sector, 2007-2016. Colombo, Sri Lanka: Ministry of Fisheries and
Aquatic Resources.
Anon. (2009). Annual report: Central Bank of Sri Lanka, Colombo.
Anon. (2009 a). Fisheries statistics. Department of Fisheries, Colombo: Ministry of
Fisheries and Aquatic Resources.
Appasamy, P. P. (2004). Managing conflict in the Bhavani Basin. In A.Vaidyanathan
& H. M. Oudshoorn (Eds.), Managing water scarcity: experiences and projects,
(pp.1-52). New Delhi, Manohar Publishers.
Arnason, R. (2005). Property right in fisheries: Iceland‟s experience with ITQs.
Review in Fish Biology and Fisheries, 15, 243-264.
Arnason, R. (2008). On the economics of releasing cultured fish into the aquatic
environment. Review in Fisheries Science, 16, 135-145.
Ayer, H. W. (1997). Grass roots collective action: Agricultural oppotunities. Journal
of Agricultural and Resource Economics, 22(1), 1-11.
Bailey, S. J. (1995). The economic rationale for government public sector economics
theory, policy and practice. Macmillan Press, London.
Bakker, M., & Matsuno, Y. (2001). A framework for valuing ecological services of
irrigation water: A case of an irrigation-wetland system in Sri Lanka. Irrigation
and Drainage Systems, 15(2), 99-115.
Bandara, C. M. M. (1985). Catchments ecosystems and village tank cascades in the
dry zone: Riedel Publishing Company, Dordecht, The Netherlands.
Bandara, C. M. M. (Eds.). (1999). Catchment eco-system and traditional village tank
cascades in the dry zone of Sri Lanka. 99-113: GeoJournal Library, Dordecht.
Bandara, C. M. M. (2009). Village tank cascade system of Sri Lanka: A traditional
technology of drought and water management. Paper presented at the Third
Annual Workshop on Disaster Reduction Hyper base Asian Application (DRH-
Asia) JST Hall, Science Plaza, Tokyo, Japan, 8-9 January 2009.
Bandara, J. M. R. S. (2007). Nature farming integration of traditional knowledge
system with modern farming in rice. Leusden, The Netherlands: ETC/COMPAS.
189
Bibliography 189
Bardhan, P. (2000). Irrigation and cooperation: An empirical analysis of 48 irrigation
communities in South India. Economic Development and Cultural Chanage,
48(4), 847-863.
Barker, R., & Tuong, L. Y. (2001, March). Water saving irrigation for rice.
Proceedigs of an international workshop. Paper presented at the International
workshop on water saving irrigation for rice, 23-25 March 2001, Wuhan, China.
Barland, J. M., & Platteau, J. P. (1996). Halting degradation of natural resources: Is
there a role for rural communities? Oxford: Clarendon Press.
Battese, G. E., & Corra, G. S. (1977). Estimation of production frontier model: With
application to the pastoral zone of eastern Australia. Australian Journal of
Agricultural Economics, 21(3), 169-179.
Battese, G. E., & Coelli, T. J. (1992). Frontier production functions, Technical
efficiency and pannel data; With application to paddy farmers in India. Journal
of Productivity Analysis, 3, 153-169.
Battese, G. E., & Coelli, T. J. (1995). A model for technical inefficiency effects in a
stochastic frontier production function for panel data. Empirical Economics, 20,
325-332.
Battese, G. E., & Broca, S. S. (1997). Functional form of stochastic frontier
production function and models for technical inefficiency effects: A
comparative study for wheat farmers in Pakistan. Journal of Productivity
Analysis, 8, 395-414.
Bell, K. P., & Irwin, E. G. (2002). Spatially explicit micro-level modeling of land use
changes at the rural-urban interface. Agricultural Economics, 27, 217-232.
Belloumi, M., & Matoussi, M. S. (2006). A stochastic frontier approch for measuring
technical efficiency of date farms in Southern Tunisia. Agricultural and
Resources Economics Review, 35(2), 285-298.
Berkes, F. E. (1989). Common property resource: Ecology and community-based
sustainable development. London: Belhaven Press.
Berkes, F. E., & Farver, T. (Eds.). (1989). The evolution of appropriate resources
management system, common property resources. London: Belhaven press.
Binam, J. N., Tonye, J., Wandiji, N., Nyambi, G., & Akoa, M. (2004). Factors
affecting the technical efficiency among smallholder farmers in the slash and
burn agriculture zone of Cameroon. Food Policy, 29(5), 531-545.
Bokusheva, R. A., & Hockmann, H. (2006). Production risk and technical
inefficiency in Russian Agriculture. European Review of Agricultural
Economics, 22(1), 93-118.
Bostock, J., McAndrew, B., Randolph, R., Jauncey, K., Telfer, T., Lorenzen, K., &
Corner, R. (2010). Aquaculture: global status and trends. Philosophical
Transactions of the Royal Society B, 365, 2897-2912.
190
190 Bibliography
Bravo, D., Sauvant, D., Bogaert, C., & Meschy, F. (2003). Quantitative aspects of
phosphorus excretion in ruminants. Reproductive Nutrition Development, 43,
(3), 285-300.
Bravo-Ureta, B., Solis, D., Victor, H., Lopez, M., Maripani, J. F., Thaim, A., &
Rivas, T. (2007). Technical efficiency in farming: a meta-regression analysis.
Journal of Production Analysis, 27, 57-72.
Brennan, D. (2008). Missing markets for storage and the potential economic cost of
expanding the spatial scope of water trade. Agricultural and Resource
Economics, 52(4), 471-485.
Brohier, R. L. (1934). Ancient irrigation works in Ceylon (Vol. 1, 37): Colombo
Ceylon Government press.
Bromley, D. W. (1991). Environment and economy: Property rights and public
policy: Oxford, UK, Basil Blackwell,.
Brooks, R., & Harris, E. (2008). Efficiency gains from water markets: Empirical
analysis of watermove in Australia. Agricultural Water Management, 95(4),
391-399.
Brumbelow, K., & Georgakakos, A. (2007). Optimization and assessment of
agricultural water-sharing scenarios under multiple socioeconomic objectives.
Journal of Water Resources Planning and Management, 133(3), 264-274.
Chakrabarty, R. D., & Samaranayake, R. A. D. B. (1983). Fish culture in seasonal
tanks in Sri Lanka. Journal of Inland Fisheries, Sri Lanka, 2, 125-140.
Chakravorty, U., Hochman, E., & Zilberman, D. (1995). A spatial model of optimal
water conveyance. Journal of Environmental Economics and Management,
29(1), 25-41.
Chakravorty, U., & Roumasset, J. (1991). Efficient spatial allocation of irrigation
water. American Journal of Agricultural Economics, 73(1), 165-173.
Chambers, R. (1988). Managing canal irrigation: Practical analysis from South
Asia: Cambridge University Press.
Chandrasoma, J., & Kumarasiri, W. S. A. A. L. (1986). Observations on polyculture
of fish in seasonal tanks in Sri Lanka. Journal of Inland Fisheries, Sri Lanka, 3,
49-55.
Chou, Y. K. (2002). Modelling social capital and growth: Department of Economics,
Research paper 865, Department of Economics, University of Melbourne,.
Christensen, L. R., Jorgenson, D. W., & Lawrence, J. L. (1973). Transcendental
logarithmic production frontiers. The Review of Economics and Statistics, 55(1),
28-45.
Coase, R. (1960). The problem of social cost. Journal of Law and Economics, 3, 1-
44.
Cochran, W. G. (1960). Sampling Techniques (3rd
ed.). New York: Wiley Publishers.
191
Bibliography 191
Coelli, T., & Battese, G. E. (1996). Identification of factors which influence the
technical inefficiency of Indian farmers. Australian Journal of Agricultural
Economics, 40(2), 103-128.
Coelli, T., Rao, D. S.P., O‟Donnell, C.J. & Battese, G. (2005). An Introduction to
efficiency and productivity analysis (2nd
ed.), Springer, New York.
Coelli, T., & Henningsen, A. (2009). Frontier: Stochastic frontier analysis. R
package version 0.991: http://CRAN.R- project.org
Coleman, J. S. (1988). Social Capital in the Creation of Human Capital. The
American Journal of Sociology, 94(S95-S120).
Costa, H. H., & De Silva, P. K. (1995). Limnological research and training in Sri
Lanka: State of the art and future needs. In B. G. R. G. Wetzel (Ed.), Limnology
in Developing Countries (Vol. 1, pp. 63-103). New Delhi, India: International
Association for Limnology Publication.
Costanza, R., Cumberland, J., Daly, H., Goodland, R., & Norgaard, R. (1997). An
introduction to ecological economics., Boca Raton, Florida,USA, St.Lucie
Press.
DAD (2000). Data book for village irrigation schemes of Sri Lanka. Colombo:
Ministry of Agriculture and Lands, Department of Agrarian Development,
Water Management Division.
Daleus, E., Palm, O., Sandell, K., Jayawardena, S. N., & Siripala, G. D. (1988).
Management and environmental constraints to rice yield within a village
irrigation system - A case study from Sri Lanka. GeoJournal, 17, 401-412.
Daleus, E., Palm, O., & Lundqvist, J. (1989). Water allocation, land tenure and yield
in a purana village, Sri Lanka. International Journal of Water Resources
Development, 5(1), 25-37.
Dasgupta, P. (1998). The economics of poverty in poor countries. Scandinavian
Journal.of Economics, 100(1), 41-68.
Debreu, G. (1959). Theory of value. Yale University Press, New Haven.
Demsetz, H. (1967). Towards a theory of property rights. American Economic
Review, 57(2), 347-359.
De Silva, S. S. (1988). Reservoirs of Sri Lanka and their fisheries. FAO Fisheries
Technical Paper 298, Food and Agriculture Organisation, Rome.
De Silva, S. S. (2003). Culture-based fisheries: an under utilized opportunity in
aquaculture development. Aquaculture Economics & Management, 221(1-4),
221-243.
De Silva, C. S., Weatherhead, E. K., Knox, J. W., & Rodriguez-Diaz, J. A. (2007).
Predicting the impact of climate change - A case study of paddy irrigation water
requirements in Sri Lanka. Agricultural Water Management, 93, 19-29.
192
192 Bibliography
Dennis, V. C., & Arriens, W. L. (2005). Understanding water rights and water
allocation. Paper presented at the 1st NARBO Thematic Workshop on Water
Rights and Water Allocation, Hanoi, Viet Nam 5-9 December 2005.
Dey, M., Rab, M., Paraguas, F., Piumsumbun, S., Bhatta, R., Ferdouse, A., &
Ahmed, M. (2005). Fish consumption and food security: A disaggregate analysis
by types of fish and classes of consumers in selected Asian countries.
Aquaculture Economics & Management, 9(1),(89-111).
Dey, M. M., Ferdinand, J., Paraguas, Bimbavo, G. B., & Regaspi, P. B. (2000).
Technical efficiency of tilapia growout pond operations in the Phillppines.
Aquaculture Economics & Management, 4(1), 33-47.
Dey, M. M., & Garcia, Y. T. (2008). Demand for fish in Asia: A cross-country
analysis. The Australian Journal of Agricultural and Resources Economics,
52(3), 321-338.
Dharmasena, P.B. (1994). Conservation farming practices for small reservoirs
watershed: a case study from Sri Lanka. Agroforestry System, 28, 203-212.
Diewert, W. E., & Wales, T. J. (1987). Flexible functional forms and global
curvature conditions. Econometrica, 55(1), 43-68.
Dinar, A. M. W., Rosegrant, R., & Meinzen-Dick, R. (1997). Water Allocation
Mechanisms: Principles and examples. The World Bank Policy Research
Working Paper 1776. World Bank, Agriculture and Natural Resources
Department.
Dixit, A. (Ed.) (1997). Indo-Nepal water resources development: Cursing the past or
moving forward: Department of History, University of Calcutta.
Doll, J. D., & Orazem, F. (1984). Production economics, theory with application,
(2nd
Ed.),NewYork, Wiley Publications.
Dudu, H., & Chumi, S. (2008). Economics of irrigation water management: A
literature survey with focus on partial and general equilibrium models. The
World Bank Policy Research Working Paper, 4556.
Dugan, P., Dey, M. M., & Sugunan, V. V. (2006). Fisheries and water productivity in
tropical river basins: Enhancing food security and livelihoods by managing
water for fish. Agricultural Water Management, 80, 262-275.
Ekanayake, S. A. B., & Jayasuriya, S. K. (1987). Measurement of firm-specific
technical efficiency: a comparison of methods. Journal of Agricultural
Economics, 38, 115-122.
Essafi, B. (1997, April). A simple method for the optimal spatial and temporal
allocation of water shortages. Proceeding of the Sustainabitity of Water
Resources under Increasing Uncertainty conference. (pp.145-152).The Rabat
Symposium S1, IAHS.
Estache, A., & Kouassi, E. (2002). Sector organization, governance, and the
inefficiency of African water utilities. World Bank Policy research working
paper, 2890, 1-21.
193
Bibliography 193
FAO (1994). Aquaculture production 1986-1992. FAO Fisherey Circuler, 815
(Revision 6), 214 pp. Food and Agriculture Organisation, Rome.
Farmer, B. H. (Ed.) (1977). Technology and change in rice growing areas. London:
The Macmillan Press Ltd.
Farolfi, S., & Perret, S. (2002). Inter-sectoral competition for water allocation in
rural South Africa: Analysing a case study through a standard environmental
economics approach. Department of Agricultural Economics, Extension and
Rural Development,University of Pretoria, South Africa, Working paper 2002-
23, 1-14.
Farrell, M. J. (1957). The measurement of productive efficiency. Journal of the
Royal Statistical Society, 120, 253-290.
Ferguson, C. A. (1992). Water allocation, inefficiency and inequity in a government
irrigation system. Journal of Development Economics, 38, 165-182.
Fernando, C. H. (1993, March). Impact of Sri Lankan reservoirs, their fisheries,
management and conservation. Proceeding of The Ecology and Landscape
Management in Sri Lanka: The International and Interdisciplinary Symposium.
(pp.351-374). 1990, Colombo, Sri Lanka.
Fernando, C. H., & Ellepola, W. B. (1969). A preliminary study of two village tanks
(reservoirs) in the Polonnaruwa area with biological notes on these reservoirs in
Ceylon. Bulletin of Fisheries Research Station Ceylon, 20, 3-13.
Florax, R. J. M. F., Voortman, R. L., & Brouwer, J. (2002). Spatial dimension of
precision agriculture: a spatial econometric analysis of millet on Sahelian
coversands. Agricultural Economics, 27, 425-443.
Freebairn, J. (2003). Policy Forum: Water pricing and availability: Principle for the
allocation of scarce water. The Australian Economic Review, 36(2), 203-212.
Furubotn, E. G., & Pejovich, S. (1972). Property rights and economic theory: A
survey of recent literature. Journal of Economic Literature, 10, 1137-1162.
Furubotn, E. G., & Richter, R. (2005). Institutions and eonomic theory. The
contribution of the new institutional economics, The university of Michigan
Press.
Gardner, R., Ostrom, E., & Walker, J. (1994). Rules, games and common pool
resources: The University of Michigan.
Gopalakrishnan, C. (1967). Economics principles of resource allocation to water
pricing. Journal of the American Water Resources Association, 3(2), 6-9.
Gough, H. G. (1957). Manual for the California psychological inventory: Consulting
Psychologists Press.
Grafton, R. Q., Squires,D., & Fox, K.J. (2000). Private property and economics
efficiency: A study of a common-pool resource. Journal of Law and Economics,
43(2), 679-698.
194
194 Bibliography
Grafton, R. Q., Knowles, S., & Owen, P.D. (2004). Total factor productivity, per
capita income and social divergence. Economic Record, 80(250), 302-313.
Grafton, R. Q. (2005). Social capital and fisheries governance. Ocean & Coastal
Management, 48, 753-766.
Griffin, R. C., Montgomery, J. M., & Rister, M. E. (1987). Selecting functional form
in production function analysis. Western Journal of Agricultural Economics, 12
(2), 216-227.
Griffin, R. C. (2006). Water resources economics: The analysis of scarcity, policies,
and projects. Cambridge, London. The Massachusetts Institute of
Technology(MIT) Press.
Gulati, H. S., & Murty, V. V. N. (1979). A model for optimal allocation of canal
water based on crop production functions. Agricultural Water Management,
2(1), 79-91.
Hahn, R. W. (1984). Market power and transferable property rights. The Quarterly
Journal of Economics, 99, 735-765.
Hanley, N., Shogren, F. F., & White, B. (1997). Environmental Economics. In theory
and practice. Market failures, United Kindom, Macmillan Text in Economics.
Hanna, S., & Munasinghe, M. (Eds.) (1995). Property rights in a social and
ecological context: Case studies and design application. The Beijer
International Institute of Ecological Economics and the World Bank.
Washington, D.C.
Hannesson, R. (1988). Studies on the role of fishermen's organisation in fishery
management: Theoretical consideration and experiences from industrial
countries: FAO Fisheries technical paper 300, 1-27. Food and Agriculture
Organisation, Rome.
Hardin, G. (1968). Tragedy of the commons. Science, 62, 1243-1248.
Heaney, A., Beare, S., & Bell, R. (2001). Evaluating improvements in irrigation
efficiency as a salinity mitigation option in the South Australian Riverland. The
Australian Journal of Agricultural and Resource Economics, 45(3), 477-493.
Hearne, R. R., & Easter, K. W. (1995). Water allocation and water market: An
analysis gains-from-trade in Chile. World Bank Technical Paper, 315. The
World Bank, Washington, D.C.
Heltberg, R. (2000). Property rights and natural resources management in developing
countries. Journal of Economic Surveys, 16, 190-214.
Henningsen, A., & Henning, C. H. C. A. (2009). Imposing regional monotonicity on
translog stochastic production frontiers with simple three-step procedure.
Journal of Productivity Analysis, 32, 217-229.
Howell, A. T. (2001). Enhancing water use efficiency in irrigated agriculture.
Agronomy Journal, 93, 281-289.
195
Bibliography 195
Huang, C. J., & Liu, J. T. (1994). Estimation of a non-neutral stochastic frontier
production function. Journal of Productivity Analysis, 5, 171-180.
Huang, C. J., Tang, A. M., & Bagi, F. S. (1986). Two view of efficiency in Indian
agriculture. Canadian Journal of Agricultural Economics, 34(2), 209-206.
Hunt, R. C. (1989). Appropriate social organization? Water users in Bureaucratic
canal irrigation systems. Human Organization, 48(1), 79-90.
Hussain, I., & Hanjra, M. A. (2004). Irrigation and poverty alleviation: Review of the
empirical evidence. Irrgation and Drainage, 53(1), 1-15.
Indrasena, H. H. A. (1965). The development of fresh water fisheries in Ceylon.
Bulletin of Fisheries Research Station, Ceylon, 17, 287-289.
Jayasinghe, A., & Amarasinghe, U. S. (2007). Buffaloes in favor of culture-based
fisheries in Sri Lanka. Aquaculture Asia, 12(3), 3-6.
Jayasinghe, A., Amarasinghe, U. S., & De Silva, S. S. (2005). Limnology and
culture-based fisheries in non-perennial reservoirs in Sri Lanka. Lake &
Reservoirs: Research and Management, 10(3), 157-166.
Jayasinghe, U. A. D., Amarasinghe, U. S., & De Silva, S. S. (2005 (a)). Trophic
classification of non-perennial reservoirs utilized for the development of culture-
based fisherioes, Sri Lanka. International Reviews of Hydrobiology, 90(2), 209-
222.
Jennings, E., Mills, P., Jordan, P., Jenson, J., Sonndergaard, M., & Barr, A. (2003).
Eutrofication from agricultural sources: Seasonal pattern and effect of
phosphoxrus, (p. 61). 2000-LS-2.1.7.-Ms, Final report. Wexford.
Jinfen, W., Jilei, W., Zhiyong, W., Changming, L., & Jingjie, Y. (2004). An
optimized spatial-temporal-sectoral allocation model for water resources.
Geojournal, 59, 227-236.
Johansson, R. C., Tsur, Y., Roe, T. L., Doukkali, R., & Dinar, A. M. W. (2002).
Pricing irrigation water: A review of theory and practice. Water Policy, 4, 173-
179.
Jondrow, J., Lovell, C. A. K., Materov, I. S., & Schmidt, P. (1982). On the
estimation of technical efficiency in the stochastic frontier production function
model. Journal of Econometrics, 19, 233-238.
Junna, J. S., & Strzepek, K. M. (2006, August). Inter-sectoral water use in South
Africa: efficiency versus equity. Paper presented at the 26th International
Association of Agricultural Economists Conference. (pp.1-16) Goald Coast,
Queensland, Australia.
Kadigi, R. M. J., Kashaigili, J. J., & Mdoe, N. S. (2004). The economics of irrigated
paddy in Usangu Basin in Tanzania: Water utilization, productivity, income and
livelihood implications. Physics and Chemistry of the Earth, 29, 1091-1100.
Kalirajan, K. P. (1981). An econometric analysis of yield variability in paddy
production. Canadian Journal of Agricultural Economics, 29, 283-294.
196
196 Bibliography
Kalirajan, K. P., & Flinn, J. C. (1983). The measurement of farm-specific technical
efficiency. Pakistan Journal of Applied Economics, 2, 167-180.
Kalirajan, K. P., & Shand, R. T. (1986). Estimating location-specific and firm-
specific technical efficiency: An analysis of Malaysian agriculture. Journal of
Economic Development, 11, 147-160.
Kalirajan, K. P., & Shand, R. T. (1999). Frontier production functions and technical
efficiency measures. Journal of Economic Survey, 13(2), 150-172.
Karagiannis, G., Tzouvelekas, V., & Xepapadeas, A. (2003). Measuring irrigation
efficiency with a stochastic production frontier. Environmental and Resource
Economics, 26, 57-72.
Kareem, R. O., Aromolaran, A. B., & Dipeolu, A. O. (2009). Economic efficiency of
fish farming in Ogun state Nigeria. Aquaculture Economics & Management,
13(1), 39-52.
Karunagoda, K. (2004). Changes in labour market and domestic agriculture. Sri
Lankan Journal of Agricultural Economics, 6(1), 82-97.
Katz, E. G. (2000). Social capital and natural capital: a comparative analysis of land
tenure and natural resource management in Guatemala. Land Economics, 76(1), 114-132.
Khalkheili, T. A., & Zamani, G. H. (2009). Famer participation in irrigation
managment: The case of Doroodzan dam irrigation network, Iran. Agricultural
Water Management, 96, 859-865.
Khan, A., Huda, F. A., & Alam, A. (2010). Farm household technical efficiency: A
study on rice producers in selected areas of Jamalpur District in Bangladesh.
European Journal of Social Sciences, 14(2), 262-271.
Khan, S., Traig, R., Yuanlai, C., & Blackwell, J. (2006). Can irrigation be
sustainable? Agricultural Water Management, 80(1-3), 87-99.
Kularatne, M.G., Amarasinghe, U.S., De Silva, S.S. (2008). Influence of
socioeconomic heterogeneity on culture-based fisheries in non-perennial
reservoirs of Sri Lanka. In M.J.S. Wijeyaratne & U.S. Amarasinghe (Eds.),
Participatory Approaches to Reservoir Fisheries Management: Issues,
Challenges and Policies (pp.135-150). Sri Lanka: Colombo: Sri Lanka
Association for Fisheries and Aquatic Resources,
Kularatne, M. G., Amarasinghe, U. S., Wattage, P., & De Silva, S. S. (2009).
Evaluation of community participation for the development of culture-based
fisheries in village reservoirs of Sri Lanka. Aquaculture Economics &
Management, 13(1), 22-38.
Kulindwa, K. (2000). Economics of water resources: Key issues and challenges.
Tanzanian Journal of Population Studies and Development, 7(1-2), 17-35.
Kumbhakar, S. C. (1987). The specification of technical and allocative inefficiency
in stochastic production and profit frontiers. Journal of Econometrics, 34, 335-
348.
197
Bibliography 197
Kumbhakar, S. C. (1994). Efficiency estimation in a profit maximizing model using
flexible production function. Agricultural Economics, 10(2), 143-152.
Kumbhakar, S. C., & Lovell, C. A. K. (2000). Stochastic frontier analysis. New
York, NY: Cambridge University Press.
Lallana, C. W., Krinner, T., Nixon, E. S., Leonard, J., & Berland, J. M. (2001).
Water scarcity managment in the context of WFD. Sustainable water use in
Europe, Demand managment, European Environment Agency, 2, 107-108.
Lau, L. J. (1986). Functional forms in econometric model building. In Z. Griliches &
M. D. Intriligator (Eds.), Handbook of Econometrics (Vol. III, pp. 1516-1566):
Elseier Science Publishers BV.
Leach, E. R. (1961). Pul Eliya, A village in Ceylon: a study of land tenure and
kinship. New York, NY: Cambridge University Press.
Le Moigne, G., Dinar, A., & Giltner, S. (1997). Principles and examples for the
allocation of scarce water resources among economic sectors. Optíons
Méditerranéennes, 31, 87-101.
Lewis, W. A. (1954). Economic development with unlimited supplies of labour. Manchester School of Economic and Social Studies, 22, 139-191.
Li, S. K., & Ng, Y. C. (1995). Measuring the productive efficiency of a group of
firms. International Advance in Economic Research, 1(4), 377-390.
Lindara, L. M. J. K., Johnsen, F. H., & Gunathilaka, H. M. (2006). Technical
efficiency in the spice based agroforestry sector in Matale District, Sri Lanka.
Agroforestry Systems, 68, 221-230.
Linuma, M., Sharma, K. R., & Leung, P. S. (1999). Technical efficiency of carp
pond culture in peninsula Malaysia: an application of stochastic production
frontier and technical efficiency model. Aquaculture, 175, 199-213.
Long, R. B. (1991). Short run marginal returns from irrigation water. Water
Resources Development, 7(1), 39-44.
Lorenzen, K. (2008). Understanding and managing enhancement fisheries systems.
Review in Fisheries Sciences, 16(1-3), 10-23.
Mahendrarajah, S., & Warr, P. G. (1991). Water management and technological
change: Village dams in Sri Lanka. Journal of Agricultural Economics, 42, 309-
324.
Marikar, F., Wilkin-Wells, J., Smolnik, S., & Sampath, R. K. (1992). Irrigation
system performance and its impact on crop productivity in Sri Lanka. Water
Resources Development, 8, 226-234.
Meeusen, W., & Van den Broeck, J. (1977). Efficiency estimation from Cobb-
Douglas production functions with composed error. International Economic
Review, 18, 435-444.
198
198 Bibliography
Meinzen-Dick, R., & Jackson, L. A. (1996). Multiple uses, multiple users of water
resources. Paper presented at the International Association for the Study of
Common Property Berkeley, CA June 5-9 1996, IFPRI, Washington, D.C.
Meinzen-Dick, R., & Bakker, M. (2001). Water rights and multiple water uses,
framework and application to Kirindioya Irrigation system Sri Lanka. Irrigation
and Drainage Systems, 15, 129-148.
Meinzen-Dick, R., & Ringler, C. (2006). Water Reallocation: Challenges, Threats,
and Solutions for the Poor. Human Development Report 2006 (pp. 1-13): United
Nations Development Programme (UNDP).
Mendis, A. S. (1965). A preliminary survey of 21 Ceylon lakes. Limnology and fish
production potential. Bulletin of Fisheries Research Station, Ceylon, 16, 7-16.
Mendis, A. S. (1977). The role of man-made lakes in the development of fisheries in
Sri Lanka, Proceedings of Indo-Pacific Fisheries Council, 17(3), 247-254.
Mendola, M. (2007). Farm households production theories:A review of institutional
and behavioral responses. Asian Development Review, 24(1), 49-68.
Molden, D., Oweis, T., Steduto, P., Bindraban, P., Hanjra, M. A., & Kijne, J. (2010).
Improving agricultural water productivity: Between optimism and caution.
Agricutural Water Management, 97, 528-535.
Molle, F., & Berkoff, J. (2009). Cities vs. agriculture: A review of intersectoral water
re-allocation. Natural Resources Forum, 33, 6-18.
NAQDA. (2008). Fisheries Statistics Year Book 2008: National Aquaculture
Development Authority, Colombo, Sri Lanka.
Nash, D., & Halliwell, D. (1999). Fertilizers and phosphorus loss from productive
grazing systems. Australian Journal of Soil Research, 37, 403-429.
Nguyen Khoa, S., Lorenzen, K., Garway, C., Chamsing, B., Siebert, D., & Randon,
M. (2005). Impact of irrigation on fisheries in rain-fed rice-farming landscapes.
Journal of Applied Ecology, 42, 892-900.
Nelson, C. N. (2002). Introduction to the special issue on spatial analysis for
agricultural economists. Agricultural Economics, 27(3), 197-200.
NSF (2000). Natural Resources of Sri Lanka 2000: N6ational Science Foundation of
Sri Lanka, Colombo, Sri Lanka.
O'Donnell, C. J., & Coelli, T. J. (2005). A Bayesian approach to imposing curvature
on distance functions. Journal of Econometrics, 126(2), 493-523.
Oglesby, R. T. (1981). A synthesis of the reservoir fisheries in Sri Lanka. FI:
TCP/SRL/8804 Field Document 2 , Food and Agriculture Organisation, Rome.
Oliver, P. E., & Marwell, G. (1988). The paradox of group size in collective action: a
theory of the critical mass II. American Sociological Review, 53(1), 1-8.
Ostrom, E. (1990). Governing the commons: the evaluation of institutions for
collective action: New York, NY: Cambridge University Press.
199
Bibliography 199
Otsuki, T. (2002). The implication of property rights for joint agriculture-timber
productivity in Brazilian Amazon. Environmental and Development Economics,
7, 299-323.
Pain, A. (1986). Agricultural research in Sri Lanka: an historical account, Modern
Asian Studies, 20(4), 755-778.
Panabokke, C. R. (1958). A pedologic study of dry zone soils. Tropical Agriculture
(Sri Lanka), 114, 151-174.
Panabokke, C. R. (2001). The nature and properties of small tank systems of the dry
zone and their sustainable production thresholds. Paper presented at the Food
security and small tank systems in Sri Lanka; agricultural science and forestry,
National Science Foundation, Colombo.
Pascoe, S., & Coglan, L. (2002). The contribution of unmeasurable inputs to fisheries
production: An analysis of technical efficiency of fishing vessels in the English
channel. American Journal of Agricultural Economics, 84(3), 585-597.
Pazvakawambwa, G. T., & Van der Zaag, P. (2001). The value of irrigation water in
Nyanyadzi smallholder irrigation scheme, Zimbabwe Value of Water Research
Report Series (Vol. 4): IHE-Delft, The Netherlands.
Peris, K., Narayana, M., & Wijesinghe, S. (2008). Ecosystem based indigenous water
management (Vol. 3). Colombo: National Science Foundation.
Phengphaengsy , F., & Okudaira, H. (2008). Assessment of irrigation efficiencies
and water productivity in paddy fields in the lower Mekong river basin. Paddy
Water Environment, 6(1), 105-114.
Pinkerton, E. W. (Ed.). (1989). Introduction: attaining better fisheries management
through co-management prospects, problems, and propositions. Vancouver:
University of British Columbia Press.
Pitt, M. M., & Lee, L. F. (1981). Measurement and sources of technical inefficiency
in the Indonesian weaving industry. Journal of Development Economics, 9, 43-
64.
Prein, M. (2002). Integration of aquaculture into crop-animal systems in Asia.
Agricultural Systems, 71, 127-146.
Rabbani, M. G., Hossain, M. I., Islam, M. S., Hossain, T. M. B., & Mannan, M. A.
(2004). Factors affecting alternate rice-fish production of Mymensingh District
in Bangladesh. Pakistan Journal of Biological Sciences, 7(5), 667-669.
Renwick, M. E. (2001). Valuing water in a multiple-uses system, irrigated
agriculture and reservoir fisheries. Irrigation and Drainage Systems, 15(2), 149-
171.
Rosegrant, M. W., Ringler, C., & Gerpacio, R. V. (1997). Water and land resources
and global food supply. Paper presented at the 23rd International Conference of
Agricultural Economics on Food Security, Diversification, and Resources
Management: Refocusing the role of agriculture, Sacramento, California.
200
200 Bibliography
Rosenthal, H. (1979). Preliminary report and recommendations on reservoir and tank
stocking activities in inland fisheries in Sri Lanka. Interim Report to FAO for
the Project TCP/SRL/8804, Development of fisheries in the man-made lakes and
reservoirs. Food and Agriculture Organisation, Rome.
Ryther, J. H. (1981). Mariculture, ocean ranching, and other culture-based fisheries.
BioScience, 31(3), 223-230.
Saikia, S. K., & Das, D. N. (2008). Rice-fish culture and its potential in rural
development: A lesson from Apatani farmers, Arunachal Pradesh, India. Journal
of Agriculture & Rural Development, 6(1&2), 125-131.
Salman, A. Z., Al-Karablieh, E. K., & Fisher, F. M. (2001). An inter-seasonal
agricultural water allocation system (SAWAS). Agricultural Systems, 68, 233-
252.
Samad, M. (2005). Water institutional reforms in Sri Lanka. Water Policy, 7, 125-
140.
Sampath, R. K. (1992). Issues in irrigation pricing in developing countries. World
Development, 20(7), 967-977.
Sauer, J., Frohberg, K., & Hockmann, E. (2006). Stochastic efficiency measurement:
the curse of theoretical consistency. Journal of Applied Economics,9(1), 139-
165.
Sauer, J., & Hockmann, H. (2005). The need for theoretically consistent efficiency
frontiers. Paper presented at the XIth European Association of Agricultural
Economists (EAAE) congress, Copenhagen, Denmark.
Schoengold, K., & Zilberman, D. (2005). The Economics of Water, Irrigation, and
Development, Handbook of Agricultural Economics, (Vol.3): Elsevier
Publishing Company.
Scott, C. A., & Ochoa, P. S. (2001). Collective action for water harvesting irrigation
in the Lerma-Chapla Basun, Mexico. Water Policy, 3, 555-572.
Senaratne, A., & Karunanayake, K. (2006). Transaction cost and institutional
innovation: Sustainability of tank aquaculture in Sri Lanka. South Asian
Network for Development and Environmental Economics (SANDEE) working
paper 18-6.
Sengupta, N., Sheladia, S., & Ostrom, E. (2001). Sustainability, equity, and
efficiency of irrigation infrastructures. In B. S. L. R. Costanza, E. Ostrom, and J.
Wilson (Eds.), Institutions, Ecosystems, and sustainability (pp. 77-118): Florida.
USA. CRC Press.
Seneviratne, G., Kulasooriya, S. A., & Rosswall, T. (1994). Sustainment of soil
fertility in the traditional rice farming, dry zone, Sri Lanka. Soil Biology &
Biochemistry, 26(6), 681-688.
Sharif, N. R., & Dar, A. A. (1996). An empirical study of the patterns and sources of
technical efficiency in traditional and HYV rice cultivation in Bangladesh. The
Journal of Development Studies, 32(4), 612-616.
201
Bibliography 201
Shankar, B., Halls, A. S., & Barr, J. (2005). The effects of surface water abstraction
for rice irrigation on floodplain fish production in Bangladesh. International
Journal of Water Resources Development, 3(1), 61-68.
Sharma, K. R., & Leung, P. S. (1998). Technical efficincy of carp production in
Nepal: An application of stochastic frontier production function approach.
Aquaculture Economics & Management, 2(3), 129-140.
Sharma, K. R., & Leung, P. S. (2000). Technical efficiency of carp production in
India: a stiochastic frontier production function analysis. Aquaculture Research,
31, 937-947.
Sharma, K. R., Praddhan, N. C., & Leung, P. S. (2001). Stochastic frontier approach
to measuring irrigation performance: An application to rice production under the
system in the Tarai of Nepal. Water Resources Research, 37(7), 2009-2018.
Singh, K., Dey, M. M., Rabbani, A. G., Sudhakaran, P. O., & Thapa, G. (2009).
Technical efficiency of freshwater aquaculture and its determinants in Tripura,
India. Agricultural Economics Research Review, 22, 185-195.
Siriweera, W. I. (1994). A study of the economic history of pre-modern Sri Lanka:
New Delhi, Vikas Publishing House Pvt Ltd.
Sivasubramaniam, K., & Jayasekara, B. M. (1997). Aquatic resources sector
contribution to the GDP and the economy. Economic Review, 23(6), 2-4.
Speelman, S., Farolfi, S., Frija, A., Dhaese, M., & Dhaese, L. (2010). The impact of
the water rights system on smallholder irrigators' willingness to pay for water in
Limpopo province, South Africa. Environmental and Development Economics,
15, 465-483.
Squires, D., & Tabor, S. (1991). Technical efficiency and future production gain in
Indonesian agriculture. The Developing Economics, 29, 258-270.
Stevenson, G. G. (1991). Common Property Economics, a general theory and land
use applications .New York, NY: Cambridge University Press.
Stiglitz, J. E. (2002). Information and the change in the paradigm in economics. The
American Economic Review, 92, 460-501.
Suppiah, R. (1985). Four types of relationship between rainfall and paddy production
in Sri Lanka. GeoJournal, 10(1), 109-118.
Swanson, T. (2003). Introduction to property rights and biodiversity conservation:
Convergence of conflicts. Land Economics, 79(4), 457-459.
Tadesse, B., & Krishnamoorthy, S. (1997). Technical efficiency in paddy farms of
Tamil Nadu: An analysis based on farm size and ecological zone. Agricultural
Economics, 16, 185-192.
Tennakoon, M. U. A. (1986). Drought hazard and rural development: A study in
perception of and adjustment to drought. Central Bank of Sri Lanka, Colombo.
Teraji, S. (2008). Property rights, trust, and economic performance. The Journal of
Socio-Economics, 37, 1584-1596.
202
202 Bibliography
Thayaparan, K. (1982). The role of seasonal tanks in the development of freshwater
fisheries in Sri Lanka. Journal of Inland Fisheries, Sri Lanka, 1, 133-152.
Thilakaratne, L., Yanagita, Y., & Imai, K. (1997). Land and labour use pattern of
paddy farming practices in Sri Lanka peasant farm sector. Research Bulletin of
the Faculty of Agriculture, Gifu University, 62, 33-43.
Thiruchelvam, S. (2002). Agricultural production efficiency of Bethma cultivation in
Mahaweli System H. Sri Lankan Journal of Agricultural Economics, 7, 1-20.
Thiruchelvam, S. (2003). The study on the cost of production of rice in Sri Lanka;
Anuradhapura and Polonnnaruwa districts Final Report. Colombo: National
Science Foundation of Sri Lanka.
Thiruchelvam, S. (2010). Enhancement of capacity of farmer organizations for
sustainable irrigation systems in Anuradhapura and Kurunegala Districts.
Paper presented at the The national conference on water, food security and
climate change in Sri Lanka: Policies, institutions and data needs for water
managment, BMICH, Colombo, Sri Lanka, 9-11 June 2009.
Tilmant, A., Pinte, D., & Goor, Q. (2008). Assessing marginal water values in
multipurpose multireservoir systems via stochastic programming. Water
Resources Research, 44(W12431), 1-17.
Tuong, T. P., & Bouman, B. (2002). Rice production in water-scarce environment.
Paper presented at the Water productivity workshop, 12-14 November 2001,
International Water Management Institute, Colombo, Sri Lanka.
Turner, K., Georgiou, S., Clark, R., & Brouwer, R. (2004). Economic valuation of
water resources in agriculture. From the sectoral to a functional perspective of
natural resources managment (Vol. 27). Food and Agriculture Organisation,
Rome.
Udawattage, U. D. S. (1985). The development of micro-catchments in Sri Lanka.
Journal of Hydrology, 80, 351-359.
Ulluwishewa, R. (1991). Soil fertility management of paddy fields by traditional
farmers in the dry zone of Sri Lanka. Journal of Sustainable Agriculture, 1(3),
95-106.
Ulluwishewa, R. (1995). Traditional practices of inland fishery resources
management in the dry zone of Sri Lanka: implications for sustainability.
Environmental Conservation, 22, 127-133.
Uphoff, N. (1985). Experience with people‟s participation in water management: Gal
Oya, Sri Lanka. In J.C. Gracia-Zamor (Eds.), Public Participation in
development planning and management: cases from Africa and Asia (pp. 131-
178). Boulder, Colorado, Westview Press.
Van der Molden, P. (2001). Rains, droughts and dreams of prosperity: Resourceful
strategies in irrigation management and beyond. The Sri Lankan case
(Unpubished doctoral dissertation). University of Twente, Enschede, the
Netherlands.
203
Bibliography 203
Van der Zaag, P. (2007). Asymmetry and equity in water resources management;
critical institutional issues for southern Africa. Water Resources Management,
21, 1993-2004.
Varian, H. R. (1992). Microeconomic analysis (3rd
ed.). London, W.W. Nortan &
Company.
Verhoef, E. T. (1999). Externalities, In J. C. J. M.van den Bergh (Eds) Handbook of
environmental economics (pp 197-214). Cheltenham, Edward Elgar Publishers.
Villano, R. A., & Fleming, E. M. (2006). Technical inefficiency and production risk
in rice farming: Evidence from Central Luzon Philippines. Asian Economic
Journal, 20(1), 29-46.
Wade, R. (1982). The system of administrative and political corruption: Canal
irrigation in south India. Journal of Development Studies, 18, 287-327.
Wade, R. (1987). The management of common property resources: Collective action
as an alternative to privatization or state regulation. Cambridge Journal of
Economics, 11(1), 95-106.
Wadud, A., & White, B. (2000). Farm household efficiency in Bangladesh: a
comparison of stochastic frontier and DEA methods. Applied Economics,
32(13), 1665-1673.
Ward, F. A., & Michelsen, A. (2002). The economic value of water in agriculture:
concepts and policy application. Water Policy, 4,(423-446).
Wijenayake, W. M. H. K., Jayasinghe, U. A. D., Amarasinghe, U. S., Athula, J. A.,
Pushpalatha, K. B. C., & De Silva, S. S. (2005). Culture-based fisheries in non-
perennial reservoirs in Sri Lanka: production and relative performance of
stocked species. Fisheries Managemnet and Ecology, 12, 249-258.
Wingard, J. D. (2000). Community transferable quotas: internalizing externalities
and minimizing social impacts of fisheries management. Human Organization,
59(1), 48-57.
Yao, S., & Liu, Z. (1998). Determinants of grain production and technical efficiency
in China. Journal of Agricultural Economics, 49(2), 171-184.
Young, R. (1996). Measuring economic benefits for water investment and policies.
World Bank Technical Paper, 338. The World Bank, Washington, D.C.
Zhou, Y., Zhang, Y., Abbaspour, K. C., Mosler, H., & Yang, H. (2009). Economic
impacts on farm households due to water reallocation in China's Chaobai
watershed. Agricultural Water Management, 96, 883-891.
Zhu, X., & Lansink, A. O. (2010). Impact of CAP subsidies on technical efficiency
of crop farms in Germany, the Netherlands and Sweden. Journal of Agricultural
Economics, 61(3), 545-564.
Zubair, L. (2002). El-nino-southern oscillation influences on rice production in Sri
Lanka. International Journal of Climatology, 22, 249-260.
204
204 Bibliography
205
Appendices 205
Appendices
APPENDIX A
Figure A1. The distribution of VISs. From “Evaluation of community participation
for the development of culture-based fisheries in village reservoirs of Sri
Lanka” by M.G. Kularatne at el., 2009 Aquaculture Economics &
Management, 13(1) p.25.
206
206 Appendices
Table A1.1.
Distribution of small reservoirs in administrative districts
Districts Working
reservoirs
Abandoned
reservoirs
Total
Ampara 181 87 268
Anuradapura 2,333 665 2,998
Badulla 259 128 387
Batticaloa 132 110 242
Colombo 03 02 05
Galle 00 00 00
Gampaha 24 33 57
Hambantota 446 23 469
Kalutara 06 01 07
Kandy 47 11 58
Kegalle 07 03 10
Kurunegala 4,192 77 4,269
Mannar 61 51 112
Matara 24 03 27
Matale 278 33 311
Monaragala 285 151 436
Nuwara Eliya 54 17 71
Polonnaruwa 79 36 115
Puttalam 743 175 918
Ratnapura 59 08 67
Trincomalee 428 196 624
Vavuniya 453 101 554
Total 10,094 1,911 12,005
Source: Department of Agrarian Services, Sri Lanka (2000).
207
Appendices 207
Figure A2. The village irrigation system. Adapted from “Soil fertility
management of paddy fields by traditional farmers in the dry zone of
Sri Lanka” by R. Ulluwishewa, 1995, Journal of Sustainable
Agriculture, 1(3) p.97.
208
208 Appendices
Table A2.
Variation of rice farming in Sri Lanka
a. Extent of of sown and harvested land area (hectares) in Sri Lanka, 2008
Yala season (2008) Maha season (2008)
Gross extent sown
Net extent harvest
471395
417167
581600
508338
b. Rice yield variations (kg/ha) by main seasons, main rainfall zones and main
irrigation schemes in 2008
Main seasons Rainfall zones Main irrigation schemes
Major Minor
Maha season
Yala season
High rainfall zone
Low rainfall zone
High rainfall zone
Low rainfall zone
4105
4507
3978
4751
3486
3699
3427
3854
Source: Sri Lanka Department of Census and Statistics 2008.
Appendices 209
APPENDIX B
Figure B1. The density of village reservoirs as a cascade system (Medawachchiya and Anuradhapura sheet: not to scale). From “Rains, droughts
and dreams of prosperity,” by P. Van der Molden, 2001, p.376.
Appendices 210
Source: Compiled by Author.
Figure B2. Water management hierarchy.
Source: Adapted from Daleus et al. (1988).
Figure B3. Paddy fields distribution in the command area.
Appendices 211
APPENDIX C
Source: compiled by Author.
Figure C1. Study area and selected sample.
Appendices 212
Table C1.
Reservoirs used for CBF from 2006
No Kurunagala District Anuradhapura District
AD Division Number of
reservoirs
AD Division Number of
reservoirs
Excluded
reservoirs
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
TOTAL
Abanpola
Aehetuwewa
Bingiriya
Galgamuwa
Ganewaththa
Giribawa
Ibbagamuwa
Katupotha
Kobeigane
Kotawehera
Kuliyapitiya
Kurunegala
Mahawa
Nikaweratiya
Paduwasnuwara
Polpithigama
Udubaddawa
Wariyapola
Wellawa
07
10
13
44
14
03
04
01
16
05
02
01
07
04
04
16
03
08
03
165
Andiyagala
CNP
Elayapathtuwa
ENP
Galenbindunuwewe
Galkiriyagama
Galnewe
Gambirigaswewe
Horowpothana
Ipalogama
Kahatagasdigiliya
Kebathigollawa
Kekirawa
Medawachchiya
Mahailluppallama
Mahapaladikulama
Mihintale
Muriyakadawala
Nochchiyagama
Palagala
Palugaswewe
Rajanganaya
Rambewa
Saliyapura
Seeppukulama
Talawa
Tantirimale
Thirappane
Vlachchiya
02
04
02
04
12
01
04
01
07
01
15
02
04
11
01
01
07
05
10
02
09
02
03
01
02
03
04
10
39
169
-
-
-
01
-
-
01
-
-
-
03
01
-
-
-
-
-
-
-
-
-
01
-
-
-
-
-
-
01
09
Sources: DAD, District office, Kurunagala and Anuradhapuraya 2009.
213
Appendices 213
APPENDIX D
DO FILE FOR THE SIMPLE THREE-STEP PROCEEDURE FOR TESTING
THEORETICAL CONSISTANCY OF TRANSLOG FUNCTION.
Source file and results of stochastic production frontier for rice farming
# load R packages "micEcon","frontier", "quadprog"#
library( "car" )
library( "micEcon" )
library( "frontier" )
library( "quadprog" )
# Load data set on rice production in Sri Lanka#
# data ( riceFinalD )#
riceFinalI <- read.table( "C:/R/WORK/riceFinalI.csv", header = TRUE, sep = "," )
# add information on panel structure#
riceFinalI <- data.frame( riceFinalI, c( "YEAR", "SAMPLE" ) )
# ***********************************************************
# UNRESTRICTED STOCHASTIC FRONTIER ESTIMATION (STEP 1)
# ***********************************************************
# estimate the unrestricted stochastic frontier model #
sfaStep1Result <- frontierQuad( yName = "PROD",
xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,
zNames = c( "AGE", "EDU", "PRATE", "FOM", "LHE", "LMID", "LOISSU", "LOWN",
"DPEST", "DWEED", "WMGT" ) , data = riceFinalI )
# Efficiency estimate from the unrestricted model #
riceFinalI$uEfficiency <- efficiencies( sfaStep1Result, asInData = TRUE )
# Beta coefficients of the unrestricted model#
uCoef <- coef( sfaStep1Result ) [ 1: 21 ]
# Inverse of the covariance matrix of the unrestricted beta coefficients #
uCovInv <- solve( vcov( sfaStep1Result ) [ 1:21, 1:21 ] )
# ******************************************
# MINIMUM DISTANCE ESTIMATION (STEP 2)
# ******************************************
# Matrix to impose monotony #
monoRestr <- translogMonoRestr( xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
data = riceFinalI, dataLogged = TRUE )
# Minimisation of the difference by quadratic programming#
minDistResult <- solve.QP( Dmat = uCovInv, dvec = rep( 0, length( uCoef ) ) ,
Amat = t( monoRestr ), bvec = - monoRestr %*% uCoef )
# Beta coefficients of the restricted model #
cCoef <- minDistResult$solution + uCoef
# **************************************************
# FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
# **************************************************
# fitted frontier output of the restricted model (assuming efficiency =1) #
riceFinalI$lcFitted <- translogCalc( xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
214
214 Appendices
data = riceFinalI, coef = cCoef , dataLogged = TRUE )
# estimate a stochastic frontier model with the constrained frontier #
sfaStep3Result <- frontier( yName = "PROD" , xNames = c( "lcFitted" ),
zNames = c( "AGE", "EDU", "PRATE", "FOM", "LHE", "LMID", "LOISSU", "LOWN",
"DPEST", "DWEED", "WMGT" ), data = riceFinalI )
# Efficiency estimate from the restricted model#
riceFinalI$cEfficiency <- efficiencies( sfaStep3Result, asInData = TRUE )
# adjusted beta coefficients of the restricted production frontier#
caCoef <- cCoef * coef ( sfaStep3Result ) [ 2 ]
caCoef [ 1 ] <- caCoef [ 1 ] + coef ( sfaStep3Result ) [ 1 ]
# ****************************************
# TESTING MONOTONICITY RESTRICTIONS
# ****************************************
# Binding restriction (with zeros for the deltas, sigma, and gama)#
monoRestrBind <- cbind( monoRestr [ minDistResult$iact, ],
matrix( 0, nrow = length( minDistResult$iact ) , ncol = 4 ) )
# wald test#
waldTest <- linear.hypothesis(sfaStep1Result, monoRestrBind)
# Likelihood ratio test#
lrTest <- -2 * ( logLik( sfaStep3Result ) - logLik( sfaStep1Result ) )
lrTestDf <- nrow( monoRsetrbind)
lrTestProb <- pchisq( lrTest, lrTestDf, lower.tail = FALSE )
# ************************************
# PARTIAL PRODUCTION ELASTICITIES
# ************************************
# Unrestricted model#
uEla <- elas( sfaStep1Result )
# restricted (adjusted) model#
cEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE )
# restricted adjusted model #
caEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE )
# ***********************************************************************
# COMMAND FOR DISPLAY RESULTS;1. Unrestricted stochastic frontier estimation
(step1) # -estimated parameters#
#************************************************************************
coef(summary( sfaStep1Result ) )
# Check for monotonicity#
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalI, coef = uCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST"
) ,
data = riceFinalI, coef = uCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
# 2. MINIMUM DISTANCE ESTIMATION (STEP 2)#
215
Appendices 215
# Parameter estimation #
cCoef
# Check monotonicity #
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST"
) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
#3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
# Parameter estimation#
coef(summary( sfaStep3Result ) )
# Adjusted (restricted) co efficiencies#
caCoef
# Check monotonicity#
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalI, coef = caCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST"
) ,
data = riceFinalI, coef = caCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
# TESTING MONOTONICITY RESTRICTIONS
waldTest
# Likelihood ratio test ( test statistics, degree of freedom, P-value) #
lrTest
lrTestDf
lrTestProb
# PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS #
# Partial production elasticities of the unrestricted model #
# Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME #
colMeans( uEla )
# Partial production elasticities of the restricted model#
colMeans( caEla )
# EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS #
# Mean efficiencies of the unrestricted model #
colMeans ( riceFinalI [ , c( "uEfficiency", "cEfficiency" ) ] )
## estimation final likelihood estimate ##
summary( sfaStep3Result, effic = FALSE,
logDepVar = TRUE )
## estimation of individual technical effiiciency scores##
summary( sfaStep3Result, effic = TRUE,
logDepVar = TRUE )
216
216 Appendices
Tables D1
Results of the simple three steps procedure for imposing theoretical consistency of
stochastic translog function for rice farming.
1.1. Unrestricted stochastic frontier estimation (step1)
estimated parameters
Estimate Std. Error z value Pr(>|z|)
a_0 0.275388724 0.061523587 4.47614869 7.600157e-06
a_1 0.310820893 0.034986140 8.88411516 6.443152e-19
a_2 0.161971172 0.029788198 5.43742763 5.405529e-08
a_3 0.163121269 0.036401574 4.48115984 7.423848e-06
a_4 0.052059005 0.026992965 1.92861381 5.377882e-02
a_5 0.110043094 0.030015914 3.66615832 2.462215e-04
b_1_1 0.150261405 0.056347895 2.66667292 7.660618e-03
b_1_2 -0.044486729 0.040592074 -1.09594619 2.731023e-01
b_1_3 -0.057810165 0.034681783 -1.66687408 9.553945e-02
b_1_4 0.007332395 0.031208626 0.23494769 8.142493e-01
b_1_5 -0.007223513 0.037899536 -0.19059633 8.488419e-01
b_2_2 0.041845157 0.059283142 0.70585256 4.802798e-01
b_2_3 0.080814846 0.048887582 1.65307512 9.831558e-02
b_2_4 -0.009932764 0.035158663 -0.28251255 7.775505e-01
b_2_5 -0.012141975 0.036711281 -0.33074233 7.408391e-01
b_3_3 0.144531803 0.046807549 3.08778833 2.016520e-03
b_3_4 0.034517647 0.035371466 0.97586134 3.291332e-01
b_3_5 -0.079809940 0.039788015 -2.00587892 4.486917e-02
b_4_4 -0.075415138 0.046441219 -1.62388368 1.044006e-01
b_4_5 0.006872198 0.033379822 0.20587880 8.368856e-01
b_5_5 0.123656682 0.052302099 2.36427764 1.806527e-02
Z_AGE 0.002912864 0.006629012 0.43941149 6.603634e-01
Z_EDU 0.001833783 0.031024257 0.05910805 9.528661e-01
Z_PRATE -0.013829543 0.007738583 -1.78708992 7.392296e-02
Z_FOM -0.631576797 0.275013329 -2.29653159 2.164551e-02
Z_LHE 0.319375844 0.238240224 1.34056222 1.800626e-01
Z_LMID 0.652846089 0.336792470 1.93842246 5.257170e-02
Z_LOISSU 0.994043532 0.390767860 2.54382111 1.096472e-02
Z_LOWN 0.452919639 0.324831847 1.39432030 1.632209e-01
Z_DPEST 1.235348877 0.497884413 2.48119613 1.309423e-02
Z_DWEED -1.049469174 0.590638689 -1.77683784 7.559492e-02
Z_WMGT -0.011546941 0.005609727 -2.05837839 3.955382e-02
sigmaSq 0.785954382 0.275572401 2.85207945 4.343424e-03
gamma 0.852418528 0.051626502 16.51125879 3.044657e-61
Check for monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 272 out of 460 observations (59.1%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 460 out of 460 observations (100%)
- 'LABOR' is fulfilled at 451 out of 460 observations (98%)
- 'POWER' is fulfilled at 408 out of 460 observations (88.7%)
- 'ITIME' is fulfilled at 374 out of 460 observations (81.3%)
- 'PEST' is fulfilled at 396 out of 460 observations (86.1%)
Check for quasiconcavity
This translog function is quasiconcave at 2 out of 460
observations (0.4%)
217
Appendices 217
1.2. MINIMUM DISTANCE ESTIMATION (STEP 2)
Parameter estimation
a_0 a_1 a_2 a_3 a_4
0.2917592184 0.3227480305 0.1622377999 0.1338698838
0.0586163762
a_5 b_1_1 b_1_2 b_1_3 b_1_4
0.1147049454 0.1658511874 -0.0160568435 -0.0332303210
0.0066078916
b_1_5 b_2_2 b_2_3 b_2_4 b_2_5
-0.0112890677 0.0508935459 0.0136110091 -0.0117018252 -
0.0063667408
b_3_3 b_3_4 b_3_5 b_4_4 b_4_5
0.0226133233 0.0162788763 -0.0050120728 -0.0195950429
0.0003493144
b_5_5
0.0473820092
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 460 out of 460 observations (100%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 460 out of 460 observations (100%)
- 'LABOR' is fulfilled at 460 out of 460 observations (100%)
- 'POWER' is fulfilled at 460 out of 460 observations (100%)
- 'ITIME' is fulfilled at 460 out of 460 observations (100%)
- 'PEST' is fulfilled at 460 out of 460 observations (100%)
Check for quasiconcavity
This translog function is quasiconcave at 389 out of 460
observations (84.6%)
1.3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
coef(summary( sfaStep3Result )
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.005518296 0.062236856 -0.08866605 9.293473e-01
lcFitted 1.001222093 0.045532573 21.98913932 3.658610e-107
Z_AGE 0.004663744 0.006131042 0.76067728 4.468498e-01
Z_EDU -0.006029988 0.024286426 -0.24828634 8.039129e-01
Z_PRATE -0.012052283 0.006231026 -1.93423730 5.308396e-02
Z_FOM -0.592926306 0.265200579 -2.23576550 2.536714e-02
Z_LHE 0.340990912 0.250308138 1.36228456 1.731081e-01
Z_LMID 0.597606793 0.288144349 2.07398408 3.808079e-02
Z_LOISSU 0.914882285 0.337012326 2.71468494 6.633885e-03
Z_LOWN 0.459380846 0.291394331 1.57649205 1.149124e-01
Z_DPEST 1.050003283 0.436559015 2.40518062 1.616447e-02
Z_DWEED -0.845842528 0.476920633 -1.77354987 7.613764e-02
Z_WMGT -0.009587195 0.004474243 -2.14275222 3.213300e-02
sigmaSq 0.644528309 0.213184770 3.02333187 2.500079e-03
gamma 0.794731676 0.073658040 10.78947622 3.859831e-27
218
218 Appendices
Adjusted (restricted) coefficiencies
a_0 a_1 a_2 a_3 a_4
0.2865974795 0.3231424588 0.1624360696 0.1340334853
0.0586880109
a_5 b_1_1 b_1_2 b_1_3 b_1_4
0.1148451256 0.1660538730 -0.0160764664 -0.0332709315
0.0066159670
b_1_5 b_2_2 b_2_3 b_2_4 b_2_5
-0.0113028640 0.0509557426 0.0136276431 -0.0117161259 -
0.0063745215
b_3_3 b_3_4 b_3_5 b_4_4 b_4_5
0.0226409589 0.0162987706 -0.0050181980 -0.0196189899
0.0003497413
b_5_5
0.0474399144
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 460 out of 460 observations (100%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 460 out of 460 observations (100%)
- 'LABOR' is fulfilled at 460 out of 460 observations (100%)
- 'POWER' is fulfilled at 460 out of 460 observations (100%)
- 'ITIME' is fulfilled at 460 out of 460 observations (100%)
- 'PEST' is fulfilled at 460 out of 460 observations (100%)
Check for quasiconcavity
This translog function is quasiconcave at 389 out of 460
observations (84.6%)
TESTING MONOTONICITY RESTRICTIONS
Likelihood ratio test ( test statistics, degre of freedom, P-
value) #
lrTest
[1] 16.58081
attr(,"nobs")
[1] 460
attr(,"df")
[1] 15
attr(,"class")
[1] "logLik"
lrTestDf
[1] 8
lrTestProb
[1] 0.03478239
attr(,"nobs")
[1] 460
attr(,"df")
[1] 15
attr(,"class")
[1] "logLik"
PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS
219
Appendices 219
Partial production elasticities of the unrestricted model
Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME #
colMeans( uEla )
WATER LABOR POWER ITIME PEST
0.3108209 0.1619712 0.1631213 0.0520590 0.1100431
Partial production elasticities of the restricted model#
colMeans( caEla )
WATER LABOR POWER ITIME PEST
0.32274801 0.16223780 0.13386989 0.05861638 0.11470495
EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS
Mean efficiencies of the unrestricted model #
uEfficiency cEfficiency
0.7278041 0.7347953
estimation of individual technical effiiciency scores##
summary( sfaStep3Result, effic = TRUE,
+ logDepVar = TRUE )
Efficiency Effects Frontier (see Battese & Coelli 1995)
Inefficiency decreases the endogenous variable (as in a
production function)
The dependent variable is logged
Iterative ML estimation terminated after 31 iterations:
log likelihood values and parameters of two successive iterations
are within the tolerance limit
3.Final maximum likelihood estimates (Step 3)
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.0055183 0.0622369 -0.0887 0.929347
lcFitted 1.0012221 0.0455326 21.9891 < 2.2e-16 ***
Z_AGE 0.0046637 0.0061310 0.7607 0.446850
Z_EDU -0.0060300 0.0242864 -0.2483 0.803913
Z_PRATE -0.0120523 0.0062310 -1.9342 0.053084 .
Z_FOM -0.5929263 0.2652006 -2.2358 0.025367 *
Z_LHE 0.3409909 0.2503081 1.3623 0.173108
Z_LMID 0.5976068 0.2881443 2.0740 0.038081 *
Z_LOISSU 0.9148823 0.3370123 2.7147 0.006634 **
Z_LOWN 0.4593808 0.2913943 1.5765 0.114912
Z_DPEST 1.0500033 0.4365590 2.4052 0.016164 *
Z_DWEED -0.8458425 0.4769206 -1.7735 0.076138 .
Z_WMGT -0.0095872 0.0044742 -2.1428 0.032133 *
sigmaSq 0.6445283 0.2131848 3.0233 0.002500 **
gamma 0.7947317 0.0736580 10.7895 < 2.2e-16 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
log likelihood value: -298.3766
cross-sectional data
total number of observations = 460
efficiency estimates
efficiency
1 0.81801421
220
220 Appendices
2 0.66017048
3 0.88218982
4 ...
5 ...
6 ...
458 0.91011679
459 0.72037026
460 0.84425273
mean efficiency: 0.7347953
221
Appendices 221
APPENDIX E
Source file:The simple three step procedure for testing theoretical consistency of
translog production function estimation for CBF
# load R packages "micEcon","frontier", "quadprog"#
library( "car" )
library( "micEcon" )
library( "frontier" )
library( "quadprog" )
# Load data set on rice production in Sri Lanka#
# data ( riceFinalD )#
riceFinalI <- read.table( "C:/R/WORK/riceFinalI.csv", header = TRUE, sep = "," )
# add information on panel structure#
riceFinalI <- data.frame( riceFinalI, c( "YEAR", "SAMPLE" ) )
# ***********************************************************
# UNRESTRICTED STOCHASTIC FRONTIER ESTIMATION (STEP 1)
# ***********************************************************
# estimate the unrestricted stochastic frontier model #
sfaStep1Result <- frontierQuad( yName = "PROD",
xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,
zNames = c( "AGE", "EDU", "PRATE", "FOM", "LHE", "LMID", "LOISSU", "LOWN",
"DPEST", "DWEED", "WMGT" ) , data = riceFinalI )
# Efficiency estimate from the unrestricted model #
riceFinalI$uEfficiency <- efficiencies( sfaStep1Result, asInData = TRUE )
# Beta coefficients of the unrestricted model#
uCoef <- coef( sfaStep1Result ) [ 1: 21 ]
# Inverse of the covariance matrix of the unrestricted beta coefficients #
uCovInv <- solve( vcov( sfaStep1Result ) [ 1:21, 1:21 ] )
# ******************************************
# MINIMUM DISTANCE ESTIMATION (STEP 2)
# ******************************************
# Matrix to impose monotony #
monoRestr <- translogMonoRestr( xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
data = riceFinalI, dataLogged = TRUE )
# Minimisation of the difference by quadratic programming#
minDistResult <- solve.QP( Dmat = uCovInv, dvec = rep( 0, length( uCoef ) ) ,
Amat = t( monoRestr ), bvec = - monoRestr %*% uCoef )
# Beta coefficients of the restricted model #
cCoef <- minDistResult$solution + uCoef
# **************************************************
# FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
# **************************************************
# fitted frontier output of the restricted model (assuming efficiency =1) #
riceFinalI$lcFitted <- translogCalc( xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
222
222 Appendices
data = riceFinalI, coef = cCoef , dataLogged = TRUE )
# estimate a stochastic frontier model with the constrained frontier #
sfaStep3Result <- frontier( yName = "PROD" , xNames = c( "lcFitted" ),
zNames = c( "AGE", "EDU", "PRATE", "FOM", "LHE", "LMID", "LOISSU", "LOWN",
"DPEST", "DWEED", "WMGT" ), data = riceFinalI )
# Efficiency estimate from the restricted model#
riceFinalI$cEfficiency <- efficiencies( sfaStep3Result, asInData = TRUE )
# adjusted beta coefficients of the restricted production frontier#
caCoef <- cCoef * coef ( sfaStep3Result ) [ 2 ]
caCoef [ 1 ] <- caCoef [ 1 ] + coef ( sfaStep3Result ) [ 1 ]
# ****************************************
# TESTING MONOTONICITY RESTRICTIONS
# ****************************************
# Binding restriction (with zeros for the deltas, sigma, and gama)#
monoRestrBind <- cbind( monoRestr [ minDistResult$iact, ],
matrix( 0, nrow = length( minDistResult$iact ) , ncol = 4 ) )
# wald test#
waldTest <- linear.hypothesis(sfaStep1Result, monoRestrBind)
# Likelihood ratio test#
lrTest <- -2 * ( logLik( sfaStep3Result ) - logLik( sfaStep1Result ) )
lrTestDf <- nrow( monoRsetrbind)
lrTestProb <- pchisq( lrTest, lrTestDf, lower.tail = FALSE )
# ********************************************************
# PARTIAL PRODUCTION ELASTICITIES# Unrestricted model#
# ********************************************************
uEla <- elas( sfaStep1Result )
# restricted (adjusted) model#
cEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE )
# restricted adjusted model #
caEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE )
# ***********************************************************************
# COMMAND FOR DISPLAY RESULTS;1. Unrestricted stochastic frontier estimation
(step1) # -estimated parameters
#************************************************************************
coef(summary( sfaStep1Result ) )
# Check for monotonicity#
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalI, coef = uCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST"
) ,
data = riceFinalI, coef = uCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
# 2. MINIMUM DISTANCE ESTIMATION (STEP 2)#
# Parameter estimation #
223
Appendices 223
cCoef
# Check monotonicity #
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST"
) ,
data = riceFinalI, coef = cCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
#3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
# Parameter estimation#
coef(summary( sfaStep3Result ) )
# Adjusted (restricted) co efficiencies#
caCoef
# Check monotonicity#
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalI, coef = caCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST"
) ,
data = riceFinalI, coef = caCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
# TESTING MONOTONICITY RESTRICTIONS
waldTest
# Likelihood ratio test ( test statistics, degre of freedom, P-value) #
lrTest
lrTestDf
lrTestProb
# PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS #
# Partial production elasticities of the unrestricted model #
# Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME #
colMeans( uEla )
# Partial production elasticities of the restricted model#
colMeans( caEla )
# EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS #
# Mean efficiencies of the unrestricted model #
colMeans ( riceFinalI [ , c( "uEfficiency", "cEfficiency" ) ] )
## estimation final likelihood estimate ##
summary( sfaStep3Result, effic = FALSE,
logDepVar = TRUE )
## estimation of individual technical effiiciency scores##
summary( sfaStep3Result, effic = TRUE,
logDepVar = TRUE )
224
224 Appendices
Tables E1 Results of the simple three steps procedure for imposing theoretical consistency of
stochastic translog function for CBF production
1. Unrestricted stochastic frontier estimation (step1)
Estimate Std. Error z value Pr(>|z|)
a_0 1.247380594 0.2416894288 5.1610888 2.455177e-07
a_1 0.451353904 0.0721485640 6.2558959 3.952404e-10
a_2 -0.058634076 0.0858220336 -0.6832054 4.944771e-01
a_3 0.285020365 0.0972974115 2.9293725 3.396471e-03
b_1_1 0.398438098 0.1259519899 3.1634125 1.559312e-03
b_1_2 0.040080546 0.0677087480 0.5919552 5.538806e-01
b_1_3 -0.196934695 0.0987459403 -1.9943574 4.611301e-02
b_2_2 0.081197503 0.1398692695 0.5805243 5.615611e-01
b_2_3 0.005030219 0.1000365815 0.0502838 9.598962e-01
b_3_3 0.118590538 0.1657341372 0.7155468 4.742712e-01
Z_GROUPS -0.368413829 0.3320777544 -1.1094204 2.672489e-01
Z_TIME 0.017098512 0.0069243945 2.4693151 1.353720e-02
Z_WRISK 0.353024754 0.3145319236 1.1223813 2.617003e-01
Z_SUBSI 0.798219727 0.3295247974 2.4223358 1.542109e-02
Z_CATBUF -0.001145305 0.0008300587 -1.3797881 1.676519e-01
Z_DSGS -0.156123342 0.3332726477 -0.4684553 6.394590e-01
Z_DFGS 0.404372834 0.4738651635 0.8533500 3.934652e-01
Z_MUW -0.036601882 0.0426110334 -0.8589766 3.903534e-01
sigmaSq 2.690489672 0.5199010859 5.1750030 2.279073e-07
gamma 0.797576690 0.0761043069 10.4800467 1.066910e-25
1.2. Check for monotonicity
This translog function is monotonically increasing in WATER,
LABOR, TOTALF at 72 out of 325 observations (22.2%)
The monotonicity condition for the exogenous variable
'WATER' is fulfilled at 274 out of 325 observations (84.3%)
'LABOR' is fulfilled at 84 out of 325 observations (25.8%)
'TOTALF' is fulfilled at306 out of325 observations (94.2%)
1.3. Check for quasiconcavity
This translog function is quasiconcave at 7 out of 325
observations (2.2%)
2. MINIMUM DISTANCE ESTIMATION (STEP 2)
2.1 Parameter estimation
a_0 a_1 a_2 a_3
b_1_1
1.489423e+00 4.469748e-01 4.585024e-03 2.656198e-01
1.647927e-01
b_1_2 b_1_3 b_2_2 b_2_3 b_3_3
1.587508e-03 -9.092850e-02 -8.029165e-05 3.835936e-04
7.033910e-02
225
Appendices 225
2.2. Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, TOTALF at 325 out of 325 observations (100%)
The monotonicity condition for the exogenous variable
'WATER' is fulfilled at 325 out of 325 observations (100%)
'LABOR' is fulfilled at 325 out of 325 observations (100%)
'TOTALF' is fulfilled at 325 out of 325 observations (100%)
2.3. Check for quasiconcavity
This translog function is quasiconcave at 302 out of 325
observations (92.9%)
3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
3.1. Final stochastic frontier estimation
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.014257731 0.270862924 0.05263818 9.580202e-01
lcFitted 0.999210757 0.120815060 8.27058112 1.333076e-16
Z_GROUPS -0.386222554 0.324890628 -1.18877715 2.345274e-01
Z_TIME 0.016591550 0.006580037 2.52149782 1.168564e-02
Z_WRISK 0.318785244 0.294719548 1.08165626 2.794053e-01
Z_SUBSI 0.890853592 0.313954879 2.83752110 4.546534e-03
Z_CATBUF -0.001167067 0.000727689 -1.60379929 1.087583e-01
Z_DSGS -0.165057140 0.302053595 -0.54644984 5.847568e-01
Z_DFGS 0.550641545 0.440642310 1.24963385 2.114333e-01
Z_MUW -0.040792762 0.040885158 -0.99774012 3.184054e-01
sigmaSq 2.717152688 0.482156016 5.63542214 1.746300e-08
gamma 0.815028104 0.064226614 12.68988117 6.728780e-37
3.2. Adjusted (restricted) coefficiencies
a_0 a_1 a_2 a_3 b_1_1
1.502505e+00 4.466220e-01 4.581406e-03 2.654101e-01
1.646626e-01
b_1_2 b_1_3 b_2_2 b_2_3 b_3_3
1.586255e-03 -9.085674e-02 -8.022828e-05 3.832908e-04
7.028358e-02
3.3.Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, TOTALF at 325 out of 325 observations (100%)
The monotonicity condition for the exogenous variable
'WATER' is fulfilled at 325 out of 325 observations (100%)
'LABOR' is fulfilled at 325 out of 325 observations (100%)
'TOTALF' is fulfilled at 325 out of 325 observations (100%)
3.4 Check for quasiconcavity
This translog function is quasiconcave at 302 out of 325
observations (92.9%)
226
226 Appendices
4. TESTING MONOTONICITY RESTRICTIONS
Likelihood ratio test (test statistics, degre of freedom, P-
value)
lrTest
[1] 6.584605
attr(,"nobs")
[1] 325
attr(,"df")
[1] 12
attr(,"class")
[1] "logLik"
lrTestDf
[1] 5
lrTestProb
[1] 0.2534112
attr(,"nobs")
[1] 325
attr(,"df")
[1] 12
attr(,"class")
[1] "logLik"
5.PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS
Partial production elasticities of the unrestricted model
Mean values:
WATER LABOR TOTALF
0.45135390 -0.05863409 0.28502036
Partial production elasticities of the restricted model
WATER LABOR TOTALF
0.446621998 0.004581406 0.265410133
EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS
Mean efficiencies of the unrestricted model
uEfficiency cEfficiency
0.3423334 0.3250660
6. Estimation final likelihood estimate
Efficiency Effects Frontier (see Battese & Coelli 1995)
Inefficiency decreases the endogenous variable (as in a
production function)
The dependent variable is logged
Iterative ML estimation terminated after 22 iterations:
log likelihood values and parameters of two successive iterations
are within the tolerance limit
227
Appendices 227
final maximum likelihood estimates
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.01425773 0.27086292 0.0526 0.958020
lcFitted 0.99921076 0.12081506 8.2706 < 2.2e-16 ***
Z_GROUPS -0.38622255 0.32489063 -1.1888 0.234527
Z_TIME 0.01659155 0.00658004 2.5215 0.011686 *
Z_WRISK 0.31878524 0.29471955 1.0817 0.279405
Z_SUBSI 0.89085359 0.31395488 2.8375 0.004547 **
Z_CATBUF -0.00116707 0.00072769 -1.6038 0.108758
Z_DSGS -0.16505714 0.30205359 -0.5464 0.584757
Z_DFGS 0.55064155 0.44064231 1.2496 0.211433
Z_MUW -0.04079276 0.04088516 -0.9977 0.318405
sigmaSq 2.71715269 0.48215602 5.6354 1.746e-08 ***
gamma 0.81502810 0.06422661 12.6899 < 2.2e-16 ***
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
log likelihood value: -533.4631
Cross-sectional data ; total number of observations = 325
7. estimation of individual technical effiiciency scores
sample efficiency
1 0.73932291
2 0.34771041
3 0.06322439
...
...
322 0.23179416
323 0.35913201
324 0.27130335
325 0.58909283
Mean efficiency: 0.3250660
228
228 Appendices
229
Appendices 229
APPENDIX F
Estimation of inter-sector optimal allocation of water
The water allocation of VISs are administrated under the user-based water
allocation mechanism in Sri Lanka (Renwick, 2002). Particularly for this study,
estimated average capacity of a VIS is 5.421M/ha in two sample districts
(Anuradhapura and Kurunagala). This represents 339 reservoirs in total. The water
user-association (FOs) allocate 62.5% (3.3881Mha) from the reservoir capacity (at
full supply level) for rice farming and 37.5% (2.0329 M/ha) maintains as a storage
(residual volume) for other multiple uses. Appendix F details the steps followed in
estimating inter-sector optimal allocation of water.
The production relationship is assumed as follows:
Lny lni i i i iw v u (1)
2
R 0 1 Ri 2 RilnY = β + β lnw + β lnw (2)
2
RlnY = 0.2866 + 0.3231lnw + 0.1661lnw (3)
The log value of rice production of ith
farmer (Lnyi) is a function of log value
of water use by ith
farmer (lnwi). vi and ui are defined as Equation (3.17). the empirical
model at frontier level (lnYR) for rice is given in Equation (2). Output elasticity of
water represented by 1β and 2β shows the output elasticity of the square root of water.
Equation (2) is the estimated production function, which coefficients have abstracted
from table 4.4. Similarly the estimated production function at frontier level (lnYF) for
CBF is
2
F 0 1 Fi 2 Fi
2
F
lnY = β + β lnw + β lnw (4)
lnY = 1.5025 + 0.4466lnw + 0.16471lnw (5)
The marginal product can be derived from the production function utilising the
relationship between the production elasticity and marginal product (i.e., elasticity is
equal to the marginal product divided by the average product). This can be shown as:
230
230 Appendices
ln
ln
y y w
w w y and hence,
ln
ln
y y y
w w w (6)
Therefore,
ln * * . .
ln
Y Y YMVP P MVP p
w w w (7)
where, Y denotes the frontier level of production and W the volume of water used.
Therefore, the relationship between the TE of the current level of MVP of individual
producer and the imvp can be stated as:
-umvp = e MVPi
, since
__-u
i
2
i 0 1 2
i 0
y = e y
y ln ln ln ln(mean efficiency)
y ln(mean efficiency)
or
Y w w (8)
where e-u
denotes TE and MVP at frontier level denotes as “MVP”. The output at the
given efficiency denotes as yi
Hence,
MVPR= MVP for rice production at frontier level
MVPF= MVP for fish production at frontier level
The EE (optimal allocation) condition with rival use of water hold (Griffin, 2006)
when,
MVP MVP
R F (9)
1. Estimation of frontier level optimal water allocation between rice farming
and CBF production
MaxT P Y P YR R F F
(10)
. . S T W W WR F
(11)
231
Appendices 231
where,
( ) (12)
( ) (13)
Y f WR R
Y f WF F
The Lagrangian under joint maximisation is:
( - - ) (14)T P Y P Y W W W
R R F F R F
The Kuhn-Tucker (necessary first –order) conditions are,
- 0YT RP
RW WR R
(15)
- 0
YT FPFW W
F F (16)
- - 0T
W W WR F (17)
Solve for maximum use of WR and WF,
(15) - 0,Y Y
R RP P MVPR R RW W
R R
(18)
(16) - 0Y Y
F FP P MVPF F FW W
F F
(19)
Therefore,
,Y Y
R FP PR FW W
R F
λ=shadow price of water (20)
Therefore,
MVP MVPR F
, (21)
(17) - - 0W W W W W WR F F R
(22)
(7) .Y
RMVP PR R R w
R (23)
where ,
232
232 Appendices
PR = average market price for paddy/ kg
εR = input elasticity of water use for rice farming = ln
2 ln1 2ln
yR w
RiwR
(2ln
2w
Ri is assumed on average level)
YR = Sample mean of paddy production
WR = Mean volume of water allocated for rice farming
Similarly,
(7) .Y
FMVP PF F F w
F (24)
PF = average market price for fish / kg
εF = input elasticity of water use for CBF = ln
ln
yF
wF
= 2ln
1 2w
Fi
(2ln
2w
Fi is assumed on average level)
YF = Sample mean of CBF production
WF = Mean volume of water allocates for CBF production
Then,
. .P YR R RMVP
R wR (25)
and,
. .
P YF F FMVP
F wF (26)
Then the optimal level of water (W ) allocation can be estimated as follows,
(13) MVP MVPR F (27)
and from equations,
(25) and (26) . . . .P Y P Y
R R R F F F
w wR F (28)
233
Appendices 233
Estimated total capacity of the VISs at the full supply per cropping season (W) =
5.421 (M/ha)
Average market price (PR) for kg of rice (in 2009) = 30 LKR
Average market price (PF)61
for kg of fish (in 2009) = 125 LKR
Frontier level average paddy yield/season per farmer (YRi/MTE)
= 1183/0.73= 1620kg
Frontier level average fish production per culture cycle (YFi/MTE)
= 2715 /0.33 = 8227.2kg
ε = 0.2866R
ε = 1.5025F
Then estimate demand functions for water use for rice production and CBF
production as follows:
P .ε .YR R R(28) λ = (29)
wR
P .ε .YR R RWater demand for rice farming W = (30)
R λ
P .ε .YF F F(28) λ= (31)
wF
P .ε .YF F FWater demand for CBF w = (32)
F λ
Then the aggregate demand function for water use for rice farming and CBF
production is:
F R
P .ε .Y P .ε .YR R R F F F(22) W=W +W (33)
w wR F
W
Then the shadow value at the aggregate demand for water:
61
This price is considered as farm gate price (the price of fish producers received from the fish
vendors at the reservoir). However, the market price which consumers receive from the fish sellers
varies from 250 to 300 LKR (approximately AU$ 2.50 to 3.00) per Kg.
234
234 Appendices
. . . . . . . .
(34)
P Y P Y P Y P YR R R F F F R R R F F F
w w w wR F R FW
W
Then by substituting the shadow value of water in Equation 28, the optimal water
allocate for rice and CBF can be calculated as:
. . . .(35)
. . . .(36)
P Y P YR R R R Rw
RwR
P Y P YF F F F F Fw
FwF
235
Appendices 235
APPENDIX G
For estimation of the sectoral production function of head-end, middle, and TEFs, the
same source file has been used: i. riceFinalIDHE.csv, ii. riceFinalIDM.csv and iii.
riceFinalIDTE.csv three different data files.
# load R packages "micEcon","frontier", "quadprog"#
library( "car" )
library( "micEcon" )
library( "frontier" )
library( "quadprog" )
# Load data set on rice production in Sri Lanka
# data ( riceFinalIDHE )
riceFinalIDHE <- read.table( "C:/R/WORK/riceFinalIDHE.csv", header = TRUE,
sep = "," )
# add information on panel structure
riceFinalIDHE <- data.frame( riceFinalIDHE, c( "YEAR", "SAMPLE" ) )
# ***********************************************************
# UNRESTRICTED STOCHASTIC FRONTIER ESTIMATION (STEP 1)
# ***********************************************************
# estimate the unrestricted stochastic frontier model #
sfaStep1Result <- frontierQuad( yName = "PROD",
xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,
zNames = c( "AGE", "EDU", "PRATE", "FOM", "LOISSU", "LOWN",
"DPEST", "DWEED", "WMGT" ) , data = riceFinalIDHE )
# efficiency estimate from the unrestricted model #
riceFinalIDHE$uEfficiency <- efficiencies( sfaStep1Result, asInData = TRUE )
# Beta coefficients of the unrestricted model#
uCoef <- coef( sfaStep1Result ) [ 1: 21 ]
# Inverse of the covariance matrix of the unrestricted beta coefficients #
uCovInv <- solve( vcov( sfaStep1Result ) [ 1:21, 1:21 ] )
# *******************************************************************
# MINIMUM DISTANCE ESTIMATION (STEP 2) # matrix to impose monotony#
********************************************************************
monoRestr <- translogMonoRestr( xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
data = riceFinalIDHE, dataLogged = TRUE )
# Minimisation of the difference by quadratic programming#
minDistResult <- solve.QP( Dmat = uCovInv, dvec = rep( 0, length( uCoef ) ) ,
Amat = t( monoRestr ), bvec = - monoRestr %*% uCoef )
# beta coeffcients of the restricted model #
cCoef <- minDistResult$solution + uCoef
236
236 Appendices
# **************************************************
# FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
# **************************************************
# fitted frontier output of the restricted model (assuming efficiency =1) #
riceFinalIDHE$lcFitted <- translogCalc( xNames = c( "WATER", "LABOR",
"POWER", "ITIME", "PEST" ) ,
data = riceFinalIDHE, coef = cCoef , dataLogged = TRUE )
# estimate a stochastic frontier model with the constrained frontier #
sfaStep3Result <- frontier( yName = "PROD" , xNames = c( "lcFitted" ),
zNames = c( "AGE", "EDU", "PRATE", "FOM", "LOISSU", "LOWN", "DPEST",
"DWEED", "WMGT" ), data = riceFinalIDHE )
# efficiency estimate from the restricted model#
riceFinalIDHE$cEfficiency <- efficiencies( sfaStep3Result, asInData = TRUE )
# adjusted beta coefficients of the restricted production frontier#
caCoef <- cCoef * coef ( sfaStep3Result ) [ 2 ]
caCoef [ 1 ] <- caCoef [ 1 ] + coef ( sfaStep3Result ) [ 1 ]
# ****************************************
# TESTING MONOTONICITY RESTRICTIONS
# ****************************************
# Binding restriction ( with zeros for the deltas, sigma, and gama)
monoRestrBind <- cbind( monoRestr [ minDistResult$iact, ],
matrix( 0, nrow = length( minDistResult$iact ) , ncol = 4 ) )
# wald test#
waldTest <- linear.hypothesis(sfaStep1Result, monoRestrBind)
# Likelihood ratio test#
lrTest <- -2 * ( logLik( sfaStep3Result ) - logLik( sfaStep1Result ) )
lrTestDf <- nrow( monoRsetrbind)
lrTestProb <- pchisq( lrTest, lrTestDf, lower.tail = FALSE )
# ******************************************************
# PARTIAL PRODUCTION ELASTICITIES # Unrestricted model#
# ******************************************************
uEla <- elas( sfaStep1Result )
# restricted (adjusted) model#
cEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalIDHE, coef = cCoef, dataLogged = TRUE )
# restricted adjusted model #
caEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalIDHE, coef = cCoef, dataLogged = TRUE )
237
Appendices 237
******************************************************************
# COMMAND FOR DISPLAY RESULTS;1. Unrestricted stochastic frontier
estimation (step1) # -estimated parameters#
******************************************************************
coef(summary( sfaStep1Result ) )
# Check for monotonicity#
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
data = riceFinalIDHE, coef = uCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalIDHE, coef = uCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
# 2. MINIMUM DISTANCE ESTIMATION (STEP 2)#
# Parameter estimation #
cCoef
# Check monotonicity #
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
data = riceFinalIDHE, coef = cCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalIDHE, coef = cCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
#3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
# Parameter estimation
coef(summary( sfaStep3Result ) )
# Adjusted (restricted) coefficiencies
caCoef
# Check monotonicity#
summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER",
"ITIME", "PEST" ) ,
data = riceFinalIDHE, coef = caCoef, dataLogged = TRUE ) )
# Check for quasiconcavity#
translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME",
"PEST" ) ,
data = riceFinalIDHE, coef = caCoef, dataLogged = TRUE,
convexity = FALSE, quasi = TRUE )
# TESTING MONOTONICITY RESTRICTIONS
waldTest
# Likelihood ratio test (test statistics, degre of freedom, P-value) #
lrTest
238
238 Appendices
lrTestDf
lrTestProb
# PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS #
# Partial production elasticities of the unrestricted model #
# Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME #
colMeans( uEla )
# Partial production elasticities of the restricted model#
colMeans( caEla )
# EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED
MODELS #
# Mean efficiencies of the unrestricted model #
colMeans ( riceFinalIDHE [ , c( "uEfficiency", "cEfficiency" ) ] )
## Estimation of individual technical efficiency scores##
summary( sfaStep3Result, effic = TRUE,
logDepVar = TRUE )
239
Appendices 239
Tables G1
Results of the simple three steps procedure for imposing theoretical consistency of
stochastic translog function for rice farming at HEFs
2.1. Unrestricted stochastic frontier estimation (step1)
Estimate Std. Error z value Pr(>|z|)
a_0 0.238954847 0.077419698 3.0864864 2.025372e-03
a_1 0.297919386 0.050939369 5.8485095 4.959972e-09
a_2 0.157323566 0.046330259 3.3956980 6.845385e-04
a_3 0.138878487 0.063016646 2.2038381 2.753573e-02
a_4 -0.016437596 0.042965045 -0.3825807 7.020307e-01
a_5 0.201150414 0.052670636 3.8190238 1.339808e-04
b_1_1 0.121236909 0.092557835 1.3098503 1.902465e-01
b_1_2 -0.124950590 0.065574246 -1.9054827 5.671736e-02
b_1_3 -0.013920332 0.088169868 -0.1578808 8.745507e-01
b_1_4 -0.004946552 0.058879179 -0.0840119 9.330470e-01
b_1_5 0.163673435 0.090513310 1.8082803 7.056289e-02
b_2_2 0.308501356 0.099643036 3.0960654 1.961071e-03
b_2_3 0.177972625 0.082809188 2.1491893 3.161939e-02
b_2_4 0.014868155 0.048600182 0.3059280 7.596595e-01
b_2_5 -0.049966245 0.064027697 -0.7803849 4.351644e-01
b_3_3 0.188830178 0.083617498 2.2582615 2.392936e-02
b_3_4 -0.067596436 0.067919502 -0.9952434 3.196179e-01
b_3_5 -0.213335726 0.082295669 -2.5923081 9.533436e-03
b_4_4 -0.112746303 0.060759942 -1.8556026 6.351023e-02
b_4_5 -0.023386940 0.053139177 -0.4401073 6.598594e-01
b_5_5 0.023150724 0.099600997 0.2324347 8.162004e-01
Z_AGE -0.027615551 0.034012674 -0.8119194 4.168379e-01
Z_EDU -0.184162943 0.180743756 -1.0189173 3.082422e-01
Z_PRATE 0.004098227 0.009512151 0.4308413 6.665838e-01
Z_FOM -2.177027384 1.655172738 -1.3152871 1.884134e-01
Z_LOISSU 1.212863245 0.980510927 1.2369707 2.160980e-01
Z_LOWN 1.044965733 0.945654367 1.1050187 2.691515e-01
Z_DPEST 0.951299122 0.947932827 1.0035512 3.155950e-01
Z_DWEED -2.395846506 2.260681628 -1.0597894 2.892404e-01
Z_WMGT 0.022929427 0.019953638 1.1491352 2.505002e-01
sigmaSq 1.242868849 1.113412084 1.1162703 2.643064e-01
gamma 0.951661329 0.044344422 21.4606773 3.629779e-102
Check for monotonicity#
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 27 out of 160 observations (16.9%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 155 out of 160 observations (96.9%)
- 'LABOR' is fulfilled at 104 out of 160 observations (65%)
- 'POWER' is fulfilled at 110 out of 160 observations (68.8%)
- 'ITIME' is fulfilled at 76 out of 160 observations (47.5%)
- 'PEST' is fulfilled at 150 out of 160 observations (93.8%)
Check for quasiconcavity
This translog function is quasiconcave at 0 out of 160
observations (0%)
240
240 Appendices
2.2. MINIMUM DISTANCE ESTIMATION (STEP 2)
a_0 a_1 a_2 a_3 a_4
a_5
0.246722957 0.311663779 0.134049485 0.100063095 0.040017409
0.182049146
b_1_1 b_1_2 b_1_3 b_1_4 b_1_5
b_2_2
0.162565498 -0.003956613 0.023388362 0.003247033 0.026419175
0.064392019
b_2_3 b_2_4 b_2_5 b_3_3 b_3_4
0.013723666 0.001989614 -0.012011331 0.042073296 -0.031034114
b_3_5 b_4_4 b_4_5 b_5_5
-0.075776230 -0.004033117 0.008953390 0.016746173
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 160 out of 160 observations (100%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 160 out of 160 observations (100%)
- 'LABOR' is fulfilled at 160 out of 160 observations (100%)
- 'POWER' is fulfilled at 160 out of 160 observations (100%)
- 'ITIME' is fulfilled at 160 out of 160 observations (100%)
- 'PEST' is fulfilled at 160 out of 160 observations (100%)
Check for quasiconcavity#
This translog function is quasiconcave at 36 out of 160
observations (22.5%)
2.3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.0006761980 0.107665362 -0.006280553 9.949889e-01
lcFitted 1.0204430480 0.072224878 14.128691852 2.527966e-45
Z_AGE -0.0077233039 0.019024178 -0.405973077 6.847624e-01
Z_EDU -0.1067586372 0.120015767 -0.889538435 3.737138e-01
Z_PRATE -0.0007008132 0.009238783 -0.075855577 9.395340e-01
Z_FOM -1.1146856687 0.982362614 -1.134698790 2.565015e-01
Z_LOISSU 0.7698592871 0.713429605 1.079096357 2.805448e-01
Z_LOWN 0.6265502622 0.817474862 0.766445908 4.434110e-01
Z_DPEST 0.5893035781 0.430362695 1.369318449 1.708998e-01
Z_DWEED -0.5731854383 0.796169228 -0.719929153 4.715686e-01
Z_WMGT 0.0062957935 0.008229191 0.765056163 4.442381e-01
sigmaSq 0.6445157697 0.507849337 1.269108223 2.044025e-01
gamma 0.8475945176 0.114137331 7.426093709 1.118518e-13
241
Appendices 241
Adjusted (restricted) coefficiencies
a_0 a_1 a_2 a_3 a_4
a_5
0.251090529 0.318035137 0.136789865 0.102108689 0.040835487
0.185770785
b_1_1 b_1_2 b_1_3 b_1_4 b_1_5
b_2_2
0.165888832 -0.004037498 0.023866492 0.003313412 0.026959264
0.065708388
b_2_3 b_2_4 b_2_5 b_3_3 b_3_4
b_3_5
0.014004220 0.002030288 -0.012256879 0.042933403 -0.031668546
-0.077325327
b_4_4 b_4_5 b_5_5
-0.004115566 0.009136425 0.017088516
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 160 out of 160 observations (100%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 160 out of 160 observations (100%)
- 'LABOR' is fulfilled at 160 out of 160 observations (100%)
- 'POWER' is fulfilled at 160 out of 160 observations (100%)
- 'ITIME' is fulfilled at 160 out of 160 observations (100%)
- 'PEST' is fulfilled at 160 out of 160 observations (100%)
Check for quasiconcavity
This translog function is quasiconcave at 36 out of 160
observations (22.5%)
TESTING MONOTONICITY RESTRICTIONS
Likelihood ratio test (test statistics, degre of freedom, P-
value)
lrTest
[1] 26.17691
attr(,"nobs")
[1] 160
attr(,"df")
[1] 13
attr(,"class")
[1] "logLik"
lrTestDf
[1] 8
lrTestProb
[1] 0.000979526
attr(,"nobs")
[1] 160
attr(,"df")
[1] 13
attr(,"class")
[1] "logLik"
242
242 Appendices
PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS
Partial production elasticities of the unrestricted model
Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME
WATER LABOR POWER ITIME PEST
0.316456200 0.092873221 0.105063376 0.001715152 0.228705397
Partial production elasticities of the restricted model
WATER LABOR POWER ITIME PEST
0.30772270 0.12256740 0.09865822 0.04281043 0.18900987
EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS
Mean efficiencies of the unrestricted model
uEfficiency cEfficiency
0.7319102 0.7405257
estimation final likelihood estimate
estimation of individual technical effiiciency scores
Efficiency Effects Frontier (see Battese & Coelli 1995)
Inefficiency decreases the endogenous variable (as in a
production function)
The dependent variable is logged
Iterative ML estimation terminated after 31 iterations:
log likelihood values and parameters of two successive iterations
are within the tolerance limit
final maximum likelihood estimates
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.00067620 0.10766536 -0.0063 0.9950
lcFitted 1.02044305 0.07222488 14.1287 < 2.2e-16 ***
Z_AGE -0.00772330 0.01902418 -0.4060 0.6848
Z_EDU -0.10675864 0.12001577 -0.8895 0.3737
Z_PRATE -0.00070081 0.00923878 -0.0759 0.9395
Z_FOM -1.11468567 0.98236261 -1.1347 0.2565
Z_LOISSU 0.76985929 0.71342960 1.0791 0.2805
Z_LOWN 0.62655026 0.81747486 0.7664 0.4434
Z_DPEST 0.58930358 0.43036269 1.3693 0.1709
Z_DWEED -0.57318544 0.79616923 -0.7199 0.4716
Z_WMGT 0.00629579 0.00822919 0.7651 0.4442
sigmaSq 0.64451577 0.50784934 1.2691 0.2044
gamma 0.84759452 0.11413733 7.4261 1.119e-13 ***
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
log likelihood value: -87.98355
cross-sectional data
total number of observations = 160
efficiency estimates
efficiency
1 0.8142269
2 0.9049821
3 0.8542603
243
Appendices 243
4 ...
5 ...
6 ...
158 0.7439952
159 0.8059457
160 0.8589782
mean efficiency: 0.7405257
Tables G2
Results of the simple three steps procedure for imposing theoretical consistency of
stochastic translog function for rice farming at MFs
3.1.Unrestricted stochastic frontier estimation (step1)
Estimate Std. Error z value Pr(>|z|)
a_0 0.7092697945 0.0336637150 2.106927e+01 1.522707e-98
a_1 0.3427803825 0.0408968306 8.381588e+00 5.221836e-17
a_2 0.1775658005 0.0323769084 5.484335e+00 4.150285e-08
a_3 0.0006063881 0.0402401718 1.506922e-02 9.879770e-01
a_4 0.1549260366 0.0441126838 3.512052e+00 4.446611e-04
a_5 0.0048687076 0.0291611244 1.669588e-01 8.674024e-01
b_1_1 -0.2438909357 0.1389876657 -1.754767e+00 7.929921e-02
b_1_2 -0.0366691472 0.0674365340 -5.437579e-01 5.866081e-01
b_1_3 -0.0217264539 0.0719640427 -3.019071e-01 7.627229e-01
b_1_4 0.2098688473 0.0763423911 2.749047e+00 5.976875e-03
b_1_5 -0.1172718006 0.1201564638 -9.759924e-01 3.290682e-01
b_2_2 -0.0920581093 0.0562811054 -1.635684e+00 1.019057e-01
b_2_3 0.0807655777 0.0742810721 1.087297e+00 2.769056e-01
b_2_4 -0.0233269051 0.0368919698 -6.323031e-01 5.271889e-01
b_2_5 0.0594436493 0.0456706821 1.301571e+00 1.930630e-01
b_3_3 0.0871066927 0.1115468891 7.808976e-01 4.348628e-01
b_3_4 0.0741988645 0.0752718795 9.857448e-01 3.242584e-01
b_3_5 0.0017051896 0.0944687638 1.805030e-02 9.855987e-01
b_4_4 -0.1242596082 0.0380023546 -3.269787e+00 1.076285e-03
b_4_5 -0.2047505551 0.0955369707 -2.143155e+00 3.210064e-02
b_5_5 0.2507682886 0.0636539016 3.939559e+00 8.163159e-05
Z_AGE 0.0162874854 0.0088959343 1.830891e+00 6.711682e-02
Z_EDU 0.0237859674 0.0373703541 6.364930e-01 5.244552e-01
Z_PRATE -0.0061679499 0.0060544706 -1.018743e+00 3.083250e-01
Z_FOM -0.3286164251 0.3878981394 -8.471720e-01 3.968993e-01
Z_LOISSU 0.7548476316 0.2396271240 3.150093e+00 1.632187e-03
Z_LOWN 0.0099187518 0.3193221496 3.106190e-02 9.752202e-01
Z_DPEST 0.6057106859 0.3035137435 1.995661e+00 4.597078e-02
Z_DWEED 0.2752346904 0.7183350517 3.831564e-01 7.016038e-01
Z_WMGT -0.0152174732 0.0057266748 -2.657297e+00 7.877011e-03
sigmaSq 0.6334691352 0.1144096255 5.536852e+00 3.079570e-08
gamma 0.9999997843 0.0000031251 3.199897e+05 0.000000e+00
Check for monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 14 out of 152 observations (9.2%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 136 out of 152 observations (89.5%)
- 'LABOR' is fulfilled at 142 out of 152 observations (93.4%)
- 'POWER' is fulfilled at 73 out of 152 observations (48%)
244
244 Appendices
- 'ITIME' is fulfilled at 111 out of 152 observations (73%)
- 'PEST' is fulfilled at 74 out of 152 observations (48.7%)
Check for quasiconcavity
This translog function is quasiconcave at 0 out of 152
observations (0%)
3.2. MINIMUM DISTANCE ESTIMATION (STEP 2)
a_0 a_1 a_2 a_3 a_4
6.568479e-01 2.738447e-01 1.760853e-01 -1.477768e-16
1.853527e-01
a_5 b_1_1 b_1_2 b_1_3
b_1_4
7.276319e-02 8.654806e-03 -8.347479e-02 -6.591949e-17 -
7.855355e-02
b_1_5 b_2_2 b_2_3 b_2_4
b_2_5
5.631210e-02 7.280742e-03 -1.942890e-16 3.919532e-03 -
3.944137e-03
b_3_3 b_3_4 b_3_5 b_4_4
b_4_5
1.249001e-16 -6.938894e-17 -1.530893e-16 -6.356412e-03 -
2.040639e-02
b_5_5
1.134924e-02
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 147 out of 152 observations (96.7%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 152 out of 152 observations (100%)
- 'LABOR' is fulfilled at 152 out of 152 observations (100%)
- 'POWER' is fulfilled at 147 out of 152 observations (96.7%)
- 'ITIME' is fulfilled at 152 out of 152 observations (100%)
- 'PEST' is fulfilled at 152 out of 152 observations (100%)
Check for quasiconcavity
This translog function is quasiconcave at 109 out of 152
observations (71.7%)
3.3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.004517614 0.149484535 0.03022128 9.758906e-01
lcFitted 1.028401258 0.089347403 11.51014152 1.172858e-30
Z_AGE 0.020622755 0.007251766 2.84382517 4.457550e-03
Z_EDU 0.024075043 0.032928277 0.73113583 4.646962e-01
Z_PRATE -0.004525326 0.005488278 -0.82454389 4.096306e-01
Z_FOM -0.387156296 0.314400963 -1.23140938 2.181698e-01
Z_LOISSU 0.392417698 0.207915153 1.88739345 5.910743e-02
Z_LOWN -0.014233350 0.226324884 -0.06288902 9.498549e-01
Z_DPEST 0.664820236 0.358069078 1.85668151 6.335648e-02
Z_DWEED 0.082605892 0.501354291 0.16476550 8.691286e-01
245
Appendices 245
Z_WMGT -0.011732763 0.004866011 -2.41116663 1.590158e-02
sigmaSq 0.485345427 0.135471708 3.58263311 3.401482e-04
gamma 0.899578851 0.071956250 12.50174730 7.302829e-36
Adjusted (restricted) coefficiencies
a_0 a_1 a_2 a_3 a_4
6.800208e-01 2.816222e-01 1.810864e-01 -1.519738e-16
1.906170e-01
a_5 b_1_1 b_1_2 b_1_3 b_1_4
7.482975e-02 8.900613e-03 -8.584557e-02 -6.779169e-17 -
8.078457e-02
b_1_5 b_2_2 b_2_3 b_2_4 b_2_5
5.791144e-02 7.487524e-03 -1.998071e-16 4.030851e-03 -
4.056155e-03
b_3_3 b_3_4 b_3_5 b_4_4 b_4_5
1.284474e-16 -7.135967e-17 -1.574373e-16 -6.536942e-03 –
2.098596e-02 b_5_5
1.167158e-02
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 146 out of 152 observations (96.1%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 152 out of 152 observations (100%)
- 'LABOR' is fulfilled at 152 out of 152 observations (100%)
- 'POWER' is fulfilled at 146 out of 152 observations (96.1%)
- 'ITIME' is fulfilled at 152 out of 152 observations (100%)
- 'PEST' is fulfilled at 152 out of 152 observations (100%)
Check for quasiconcavity
This translog function is quasiconcave at 102 out of 152
observations (67.1%)
TESTING MONOTONICITY RESTRICTIONS
Likelihood ratio test ( test statistics, degre of freedom, P-
value) #
lrTest
[1] 51.46044
attr(,"nobs")
[1] 152
attr(,"df")
[1] 13
attr(,"class")
[1] "logLik"
lrTestDf
[1] 8
lrTestProb
[1] 2.139086e-08
attr(,"nobs")
[1] 152
attr(,"df")
[1] 13
246
246 Appendices
attr(,"class")
[1] "logLik"
PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS
Partial production elasticities of the unrestricted model
Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME
colMeans( uEla )
WATER LABOR POWER ITIME PEST
0.350872298 0.169291063 -0.004173004 0.147438633 0.028320491
Partial production elasticities of the restricted model
colMeans( caEla )
WATER LABOR POWER ITIME
PEST
2.721284e-01 1.843117e-01 -1.508169e-16 1.944268e-01
6.851226e-02
EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS
Mean efficiencies of the unrestricted model
colMeans ( riceFinalIDM [ , c( "uEfficiency", "cEfficiency" ) ]
)
uEfficiency cEfficiency
0.5590765 0.5549092
estimation final likelihood estimate
estimation of individual technical effiiciency scores
Efficiency Effects Frontier (see Battese & Coelli 1995)
Inefficiency decreases the endogenous variable (as in a
production function)The dependent variable is logged
Iterative ML estimation terminated after 25 iterations:
log likelihood values and parameters of two successive iterations
are within the tolerance limit
final maximum likelihood estimates
Estimate Std. Error z value Pr(>|z|)
(Intercept) 0.0045176 0.1494845 0.0302 0.9758906
lcFitted 1.0284013 0.0893474 11.5101 < 2.2e-16 ***
Z_AGE 0.0206228 0.0072518 2.8438 0.0044575 **
Z_EDU 0.0240750 0.0329283 0.7311 0.4646962
Z_PRATE -0.0045253 0.0054883 -0.8245 0.4096306
Z_FOM -0.3871563 0.3144010 -1.2314 0.2181698
Z_LOISSU 0.3924177 0.2079152 1.8874 0.0591074 .
Z_LOWN -0.0142334 0.2263249 -0.0629 0.9498549
Z_DPEST 0.6648202 0.3580691 1.8567 0.0633565 .
Z_DWEED 0.0826059 0.5013543 0.1648 0.8691286
Z_WMGT -0.0117328 0.0048660 -2.4112 0.0159016 *
sigmaSq 0.4853454 0.1354717 3.5826 0.0003401 ***
gamma 0.8995789 0.0719562 12.5017 < 2.2e-16 ***
---
247
Appendices 247
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
log likelihood value: -105.1074
cross-sectional data
total number of observations = 152
efficiency estimates
efficiency
1 0.26788346
2 0.80914416
3 0.88186694
4 ...
5 ...
6 ...
150 0.26715793
151 0.25281330
152 0.12471164
mean efficiency: 0.5549092
248
248 Appendices
Tables G1
Results of the simple three steps procedure for imposing theoretical consistency of
stochastic translog function for rice farming at TEFs
4.1.Unrestricted stochastic frontier estimation (step1)
Estimate Std. Error z value Pr(>|z|)
a_0 0.177340552 0.194131715 0.91350634 3.609763e-01
a_1 0.365654116 0.066376190 5.50881447 3.612584e-08
a_2 0.228881736 0.062987171 3.63378338 2.792954e-04
a_3 0.241112522 0.064858805 3.71749871 2.012050e-04
a_4 0.107571370 0.057397156 1.87415853 6.090858e-02
a_5 0.089021023 0.053912328 1.65121830 9.869401e-02
b_1_1 0.105136485 0.103324263 1.01753917 3.088970e-01
b_1_2 0.125568085 0.069093358 1.81736840 6.916072e-02
b_1_3 -0.125950728 0.072861304 -1.72863674 8.387413e-02
b_1_4 -0.019654567 0.066595479 -0.29513365 7.678918e-01
b_1_5 0.053438477 0.058930246 0.90680898 3.645078e-01
b_2_2 -0.147025546 0.113938903 -1.29038934 1.969155e-01
b_2_3 0.031813426 0.076719415 0.41467243 6.783817e-01
b_2_4 -0.044382764 0.069481410 -0.63877179 5.229714e-01
b_2_5 -0.010446579 0.068189034 -0.15320028 8.782403e-01
b_3_3 0.120548666 0.110173102 1.09417512 2.738782e-01
b_3_4 -0.019151845 0.079228308 -0.24172983 8.089895e-01
b_3_5 0.055644547 0.060187065 0.92452668 3.552122e-01
b_4_4 0.008374517 0.095643868 0.08755937 9.302269e-01
b_4_5 -0.038358363 0.061558092 -0.62312463 5.332026e-01
b_5_5 0.143752246 0.091262613 1.57514936 1.152220e-01
Z_AGE 0.002998859 0.011745738 0.25531467 7.984801e-01
Z_EDU 0.110300956 0.080483911 1.37047212 1.705396e-01
Z_PRATE -0.014597936 0.014084939 -1.03642167 3.000055e-01
Z_FOM -0.673352110 0.552658640 -1.21838701 2.230769e-01
Z_LOISSU 1.021100007 0.614340534 1.66210750 9.649122e-02
Z_LOWN -0.229547552 0.416146118 -0.55160325 5.812202e-01
Z_DPEST 4.350220936 11.799912676 0.36866552 7.123771e-01
Z_DWEED -4.640242907 11.851108046 -0.39154507 6.953944e-01
Z_WMGT -0.007069576 0.008310144 -0.85071638 3.949269e-01
sigmaSq 0.520268745 0.391781244 1.32795725 1.841922e-01
gamma 0.752949772 0.249167036 3.02186752 2.512205e-03
Check for monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 101 out of 148 observations (68.2%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 147 out of 148 observations (99.3%)
- 'LABOR' is fulfilled at 138 out of 148 observations (93.2%)
- 'POWER' is fulfilled at 143 out of 148 observations (96.6%)
- 'ITIME' is fulfilled at 144 out of 148 observations (97.3%)
- 'PEST' is fulfilled at 116 out of 148 observations (78.4%)
Check for quasiconcavity
This translog function is quasiconcave at 27 out of 148
observations (18.2%)
249
Appendices 249
4.2. MINIMUM DISTANCE ESTIMATION (STEP 2)
Parameter estimation
a_0 a_1 a_2 a_3 a_4
a_5
0.243490923 0.399204615 0.204318776 0.222618283 0.102701355
0.094682638
b_1_1 b_1_2 b_1_3 b_1_4 b_1_5
b_2_2
0.136942023 0.048765149 -0.083205633 0.024176125 0.006792186 -
0.048360597
b_2_3 b_2_4 b_2_5 b_3_3 b_3_4
b_3_5
-0.012367142 -0.050758312 0.009215443 0.018440774 -0.006963036
-0.013624602
b_4_4 b_4_5 b_5_5
0.021188805 -0.016151107 0.047843112
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 147 out of 148 observations (99.3%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 148 out of 148 observations (100%)
- 'LABOR' is fulfilled at 148 out of 148 observations (100%)
- 'POWER' is fulfilled at 147 out of 148 observations (99.3%)
- 'ITIME' is fulfilled at 148 out of 148 observations (100%)
- 'PEST' is fulfilled at 148 out of 148 observations (100%)
Check for quasiconcavity#
This translog function is quasiconcave at 116 out of 148
observations (78.4%)
4.3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.015411240 0.133436976 -0.1154945 9.080532e-01
lcFitted 1.001642128 0.068998008 14.5169716 9.459815e-48
Z_AGE 0.003626539 0.014773018 0.2454840 8.060817e-01
Z_EDU 0.057220519 0.051990045 1.1006053 2.710685e-01
Z_PRATE -0.011450968 0.013743301 -0.8332036 4.047299e-01
Z_FOM -0.719524248 0.518834738 -1.3868082 1.655003e-01
Z_LOISSU 1.423246735 0.959524208 1.4832838 1.379990e-01
Z_LOWN -0.119353770 0.530064842 -0.2251682 8.218484e-01
Z_DPEST 2.894019829 6.185125762 0.4678999 6.398562e-01
Z_DWEED -3.410340070 6.468788131 -0.5271992 5.980552e-01
Z_WMGT -0.007174847 0.008957422 -0.8009946 4.231347e-01
sigmaSq 0.593414550 0.364275954 1.6290248 1.033078e-01
gamma 0.768542379 0.167610779 4.5852802 4.533781e-06
250
250 Appendices
Adjusted (restricted) coefficiencies
a_0 a_1 a_2 a_3 a_4
a_5
0.228479526 0.399860159 0.204654294 0.222983850 0.102870004
0.094838119
b_1_1 b_1_2 b_1_3 b_1_4 b_1_5
b_2_2
0.137166899 0.048845227 -0.083342267 0.024215825 0.006803339 -
0.048440011
b_2_3 b_2_4 b_2_5 b_3_3 b_3_4
b_3_5
-0.012387450 -0.050841664 0.009230576 0.018471056 -0.006974470
-0.013646976
b_4_4 b_4_5 b_5_5
0.021223600 -0.016177629 0.047921677
Check monotonicity
This translog function is monotonically increasing in WATER,
LABOR, POWER, ITIME,
PEST at 147 out of 148 observations (99.3%)
The monotonicity condition for the exogenous variable
- 'WATER' is fulfilled at 148 out of 148 observations (100%)
- 'LABOR' is fulfilled at 148 out of 148 observations (100%)
- 'POWER' is fulfilled at 147 out of 148 observations (99.3%)
- 'ITIME' is fulfilled at 148 out of 148 observations (100%)
- 'PEST' is fulfilled at 148 out of 148 observations (100%)
Check for quasiconcavity
This translog function is quasiconcave at 116 out of 148
observations (78.4%)
TESTING MONOTONICITY RESTRICTIONS
waldTest
Error: object 'waldTest' not found
Likelihood ratio test (test statistics, degre of freedom, P-
value)
lrTest
[1] 12.19845
attr(,"nobs")
[1] 148
attr(,"df")
[1] 13
attr(,"class")
[1] "logLik"
lrTestDf
[1] 8
lrTestProb
[1] 0.1425666
attr(,"nobs")
[1] 148
attr(,"df")
[1] 13
251
Appendices 251
attr(,"class")
[1] "logLik"
PARTIAL PRODUCTION EFFICIENCIES OF THE UNRESTRICTED AND
RESTRICTED MODELS
Partial production elasticities of the unrestricted model
Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME
WATER LABOR POWER ITIME PEST
0.3677595 0.2222485 0.2485890 0.0977435 0.1014768
Partial production elasticities of the restricted model#
WATER LABOR POWER ITIME PEST
0.41151297 0.19167464 0.21295311 0.10305827 0.09292656
EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS
Mean efficiencies of the unrestricted model
uEfficiency cEfficiency
0.8195621 0.8049358
estimation of individual technical effiiciency scores
Efficiency Effects Frontier (see Battese & Coelli 1995)
Inefficiency decreases the endogenous variable (as in a
production function)The dependent variable is logged
Iterative ML estimation terminated after 39 iterations:
cannot find a parameter vector that results in a log-likelihood
value
larger than the log-likelihood value obtained in the previous
step
final maximum likelihood estimates
Estimate Std. Error z value Pr(>|z|)
(Intercept) -0.0154112 0.1334370 -0.1155 0.9081
lcFitted 1.0016421 0.0689980 14.5170 < 2.2e-16 ***
Z_AGE 0.0036265 0.0147730 0.2455 0.8061
Z_EDU 0.0572205 0.0519900 1.1006 0.2711
Z_PRATE -0.0114510 0.0137433 -0.8332 0.4047
Z_FOM -0.7195242 0.5188347 -1.3868 0.1655
Z_LOISSU 1.4232467 0.9595242 1.4833 0.1380
Z_LOWN -0.1193538 0.5300648 -0.2252 0.8218
Z_DPEST 2.8940198 6.1851258 0.4679 0.6399
Z_DWEED -3.4103401 6.4687881 -0.5272 0.5981
Z_WMGT -0.0071748 0.0089574 -0.8010 0.4231
sigmaSq 0.5934146 0.3642760 1.6290 0.1033
gamma 0.7685424 0.1676108 4.5853 4.534e-06 ***
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
log likelihood value: -84.00497
cross-sectional data
total number of observations = 148
252
252 Appendices
efficiency estimates
efficiency
1 0.6454554
2 0.8958543
3 0.8879304
4 ...
5 ...
6 ...
146 0.8139508
147 0.9202070
148 0.8648573
Mean efficiency: 0.8049358
253
Appendices 253
APPENDIX H
Estimation of intra-sector optimal allocation of water
The analytical background of intra-sectoral allocation is provided in Section
3.3.4 of Chapter 3 and more details of the estimation are discussed in the
introduction of Chapter 8.
As assumed in inter-sectoral analyses the production relationship is considered as
follows:
ln lni i i i iy w v u (1)
We can derive the marginal product from the production function utilising the
relationship between the production elasticity and marginal product (i.e., elasticity is
equal to the marginal product divided by the average product). This can be shown as:
ln ln and hence, (2)
ln ln
y y w y y y
w w y w w w
Therefore, when the frontier marginal value product (MVP) is equal to the
individual marginal value product i(mvp ) , then:
lnY Y =P* * .
lnw w w
YMVP MVP (3)
where, Y denotes the frontier level of production. Hence, the relationship between
the TE at the current level of MVP of a individual producer and the imvp can be
stated as:
-u
imvp = e MVP , since
__-u
iy = e y (4)
2
i 0 1 2
i 0
y ln ln ln ln(mean efficiency)
y ln(mean efficiency)
or
Y w w
where e-u
denotes TE and MVP at frontier level denotes as “MVP”. The output at the
given efficiency denotes as yi where e-u
denotes TE.
The production function is assumed as i i i i iLny =β lnw +v - u and empirical models
and estimated production functions are shown as in Table 1.
254
254 Appendices
Table H 1.
Empirical and estimated sectoral models for rice farming
Locations Production function Estimated function
H 2
0 1 2lnY = + lnw + lnwH
Ri i i
2lnY = 0.0.2511 + 0.3180lnw + 0.1659lnwH
R
M 2
0 1 2lnY = + lnw + lnwM
Ri i i
2lnY = 0.6800 + 0.2816lnw+ 0.0089lnwMR
T 2
0 1 2lnY = + lnw + lnwT
Ri i i
2lnY = 0.2285 + 0.3999lnw + 0.1372lnwTR
The log value of rice production of the ith
farmer (Lnyi) is a function of log
value of water used by the ith
farmer (lnwi). vi and ui are defined as in Equation (3.17).
The three empirical models at the frontier level (lnYR) for H, M and T are shown in
Table 1. The output elasticity of water represented by 1β and 2β shows the output
elasticity of the square root of water use. The coefficients have abstracted from the
estimated sectoral production functions in Chapter 8.4.
The total volume of water allocated for rice farming at the existing level is
estimated to be 2.4665 M/ha in Chapter 6. The total water demand for rice farming
for the three sectors is:
R H M TW = W + W + W (5)
where,
RW = total volume of water used for rice,
HW = total volume of water use for HEFs,
MW = total volume of water use for MFs and
TW = total volume of water use for TEFs.
Then the total benefit function for water allocated for rice farming is
(6)
. . (7)
MaxT P Y P Y P YR H R M R T
S T W W W WH M T R
where,
255
Appendices 255
(8)
(9)
(10)
W W WH H R
W W W WM M H R
W W W W WT T M H R
The Lagrangian under joint maximisation is:
( ) (11)T P Y P Y P Y W W W WR H R M R T R H M T
The Kuhn-Tucker (necessary first order) conditions are,
- 0 (12)Y YT H HP P
R RW W WH H H
- 0 (13)Y YT M MP P
R RW W WM M M
- 0 (14)Y YT T TP P
R RW W WT T T
- - 0 (15)T
W W W W W W W WR H M T R H M T
The sectoral MVP or shadow price for water can be estimated as described in Section
3.3.4 and then solved for maximum use of WH, WM and WT.
From Equations (12) to (14):
(16)Y
HP MVPR WHW
H
(17)YMP MVP
R WMWM
256
256 Appendices
(18)YTP MVP
R WTWT
Therefore, the shadow value of water equals the MVP of each sector. This is also
called as efficient (optimal) allocation of rival use of water (Griffin, 2006).
(19)MVP MVP MVPWH WM WT
The MVP at the frontier level is denoted as MVP.
Therefore:
MVPH = MVP at frontier level of rice production in HEFs
MVPM = MVP at frontier level of rice production in MFs
MVPT = MVP at frontier level of rice production in TEFs
The rest of the analyses follow the same steps of estimation of inter-sectoral water
allocation as shown in Appendix F.
257
Appendices 257
APPENDIX I
Figure I 8.1. Institutional hierarchy of reservoir water management
258
258 Appendices
Figure I8.2. Farmers‟ welfare benefits of reservoir water at the existing level of TE.
* *
* * * * * ( ) - ( ) ( ) - ( )
0 0
4.2338 1.187287469 24528 - (20660 4.2338) - (20660 1.1872
0 0
4.2-187469 - 87470
0
w WR F
TMVP MVP w dw w MVP w dw wR R R F F F
dw dww w
R F
w dwR
338 1.1872-124528 - 24527
0
4.2332 1.187287469ln( ) - 87470 24528ln( ) - 24527
0 0
4.2332 1.187287469ln( ) - 87470 24528ln( ) - 24527
4.2332 1.18720 0
[(87
w dwF
w wR F
w wR F
469 1.443) - 87470] [(24528 1.1715) - 24527]
(126226 - 87470) (4208 - 24527)
38756 - 20319
* 18438TMVP
259
Appendices 259
Figure I 8.3. Farmers welfare benefits of reservoir water at the frontier level of
production.
**
* * * * ( ) - ( ) ( ) - ( )
0 0
2.31 3.111164138 221051 - (71055 2.31) - (71055 3.111)
0 0
2-1164138 - (71055 2.31)
0
Ww frTMVP MVP w dw w MVP w dw w
r r r f f f
dw dww w
r f
wr
.31 3.111-1221051 - (71055 3.111
0
3.1112.31164138 ln( ) - 164137 221051ln( ) - 221052.1
0 0
3.1112.31164138 ln( ) - 164137 221051ln( ) - 221052
2.31 3.1110 0
wf
w wr f
w wr f
[(164138 0.9658) - 164137] [(221051 1.1349) - 221052]
(137424 - 164137) (250880 - 221052)
- 26713 29828
3115TMVP
260
260 Appendices
261
Appendices 261
APPENDIX J
262
262 Appendices
263
Appendices 263
264
264 Appendices
265
Appendices 265
266
266 Appendices
267
Appendices 267
268
268 Appendices
269
Appendices 269
270
270 Appendices
271
Appendices 271
272
272 Appendices
273
Appendices 273
274
274 Appendices