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


8.4 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


9.4 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


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


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


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


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


(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


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


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.


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


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


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


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


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


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


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


rice farming

CHAPTER 6


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


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”.


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


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.


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


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).


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


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


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


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.


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).


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.


Figure 2.3. Graphical presentation of land and water relationship.

28


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.


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


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.


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


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).


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


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).


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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).


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


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


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


( )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


__ __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


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


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


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


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 ) .


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


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


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


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


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


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


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


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


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


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


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.

http://www.fao.org/fishery/culturedspecies/search/en

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


(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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


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


(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


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


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


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


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.


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


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


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


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


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


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


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


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***


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


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


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


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


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)



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


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


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


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


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


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


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


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


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


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


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


(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


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


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


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


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


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


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


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


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


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


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


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.


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


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


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


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


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


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


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


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


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


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


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


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


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


(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


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


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.


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


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


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


* 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

.


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


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


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


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


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


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


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


(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


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


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


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).


Figure 9.2. Co-management settings for reservoir-based agriculture in VISs

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


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


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


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


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


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


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


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


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


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


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


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" ),


"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" ) ,


# ***********************************************************************

# 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 #


"PEST" ) ,

data = riceFinalI, coef = cCoef, dataLogged = TRUE ) )



) ,

data = riceFinalI, coef = cCoef, dataLogged = TRUE,


#3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)

# Parameter estimation#


# Adjusted (restricted) co efficiencies#

caCoef

# Check monotonicity#


"PEST" ) ,

data = riceFinalI, coef = caCoef, dataLogged = TRUE ) )



) ,

data = riceFinalI, coef = caCoef, dataLogged = TRUE,



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



PEST 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%)



observations (84.6%)

1.3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)

coef(summary( sfaStep3Result )


(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













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 )


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##


+ 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)


(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


library( "car" )




# 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" ) )

# ***********************************************************


# ***********************************************************





"DPEST", "DWEED", "WMGT" ) , data = riceFinalI )

# Efficiency estimate from the unrestricted model #

riceFinalI$uEfficiency <- efficiencies( sfaStep1Result, asInData = TRUE )





# ******************************************

# MINIMUM DISTANCE ESTIMATION (STEP 2)

# ******************************************

# Matrix to impose monotony #


"ITIME", "PEST" ) ,

data = riceFinalI, dataLogged = TRUE )




# Beta coefficients of the restricted model #


# **************************************************


# **************************************************


riceFinalI$lcFitted <- translogCalc( xNames = c( "WATER", "LABOR", "POWER",

"ITIME", "PEST" ) ,

222

222 Appendices

data = riceFinalI, coef = cCoef , dataLogged = TRUE )




"DPEST", "DWEED", "WMGT" ), data = riceFinalI )

# Efficiency estimate from the restricted model#

riceFinalI$cEfficiency <- efficiencies( sfaStep3Result, asInData = TRUE )




# ****************************************


# ****************************************

# Binding restriction (with zeros for the deltas, sigma, and gama)#



# wald test#






# ********************************************************

# PARTIAL PRODUCTION ELASTICITIES# Unrestricted model#

# ********************************************************



cEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,



caEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME", "PEST" ) ,


# ***********************************************************************

# COMMAND FOR DISPLAY RESULTS;1. Unrestricted stochastic frontier estimation

(step1) # -estimated parameters

#************************************************************************




"PEST" ) ,

data = riceFinalI, coef = uCoef, dataLogged = TRUE ) )



) ,

data = riceFinalI, coef = uCoef, dataLogged = TRUE,




223

Appendices 223

cCoef



"PEST" ) ,

data = riceFinalI, coef = cCoef, dataLogged = TRUE ) )



) ,

data = riceFinalI, coef = cCoef, dataLogged = TRUE,



# Parameter estimation#


# Adjusted (restricted) co efficiencies#

caCoef



"PEST" ) ,

data = riceFinalI, coef = caCoef, dataLogged = TRUE ) )



) ,

data = riceFinalI, coef = caCoef, dataLogged = TRUE,



waldTest

# Likelihood ratio test ( test statistics, degre of freedom, P-value) #

lrTest

lrTestDf

lrTestProb


RESTRICTED MODELS #



colMeans( uEla )


colMeans( caEla )

# EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED MODELS #


colMeans ( riceFinalI [ , c( "uEfficiency", "cEfficiency" ) ] )

## estimation final likelihood estimate ##

summary( sfaStep3Result, effic = FALSE,

logDepVar = TRUE )

## estimation of individual technical effiiciency scores##


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)


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


LABOR, TOTALF at 72 out of 325 observations (22.2%)


'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


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


LABOR, TOTALF at 325 out of 325 observations (100%)


'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



3. FINAL STOCHASTIC FRONTIER ESTIMATION (STEP 3)

3.1. Final stochastic frontier estimation


(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


LABOR, TOTALF at 325 out of 325 observations (100%)


'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



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


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


Mean efficiencies of the unrestricted model


0.3423334 0.3250660

6. Estimation final likelihood estimate








227

Appendices 227

final maximum likelihood estimates


(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


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



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.


library( "car" )




# 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" ) )

# ***********************************************************


# ***********************************************************




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 )





# *******************************************************************

# MINIMUM DISTANCE ESTIMATION (STEP 2) # matrix to impose monotony#

********************************************************************


"ITIME", "PEST" ) ,

data = riceFinalIDHE, dataLogged = TRUE )




# beta coeffcients of the restricted model #


236

236 Appendices

# **************************************************


# **************************************************


riceFinalIDHE$lcFitted <- translogCalc( xNames = c( "WATER", "LABOR",

"POWER", "ITIME", "PEST" ) ,

data = riceFinalIDHE, coef = cCoef , dataLogged = TRUE )



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 )




# ****************************************


# ****************************************

# Binding restriction ( with zeros for the deltas, sigma, and gama)



# wald test#






# ******************************************************

# PARTIAL PRODUCTION ELASTICITIES # Unrestricted model#

# ******************************************************



cEla <- translogEla( xNames = c( "WATER", "LABOR", "POWER", "ITIME",

"PEST" ) ,

data = riceFinalIDHE, coef = cCoef, dataLogged = TRUE )


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#

******************************************************************



summary(translogCheckMono(xNames = c( "WATER", "LABOR", "POWER",

"ITIME", "PEST" ) ,

data = riceFinalIDHE, coef = uCoef, dataLogged = TRUE ) )


translogCheckCurvature( xNames = c( "WATER", "LABOR", "POWER", "ITIME",

"PEST" ) ,

data = riceFinalIDHE, coef = uCoef, dataLogged = TRUE,




cCoef



"ITIME", "PEST" ) ,

data = riceFinalIDHE, coef = cCoef, dataLogged = TRUE ) )



"PEST" ) ,

data = riceFinalIDHE, coef = cCoef, dataLogged = TRUE,



# Parameter estimation


# Adjusted (restricted) coefficiencies

caCoef



"ITIME", "PEST" ) ,

data = riceFinalIDHE, coef = caCoef, dataLogged = TRUE ) )



"PEST" ) ,

data = riceFinalIDHE, coef = caCoef, dataLogged = TRUE,



waldTest

# Likelihood ratio test (test statistics, degre of freedom, P-value) #

lrTest

238

238 Appendices

lrTestDf

lrTestProb


RESTRICTED MODELS #



colMeans( uEla )


colMeans( caEla )

# EFFICIENCY ESTIMATE OF THE UNRESTRICTED AND RESTRICTED

MODELS #


colMeans ( riceFinalIDHE [ , c( "uEfficiency", "cEfficiency" ) ] )

## Estimation of individual technical efficiency scores##


logDepVar = TRUE )

239

Appendices 239

Tables G1


stochastic translog function for rice farming at HEFs

2.1. Unrestricted stochastic frontier estimation (step1)


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#





- 'WATER' is fulfilled at 155 out of 160 observations (96.9%)







observations (0%)

240

240 Appendices


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










Check for quasiconcavity#





(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


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















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


RESTRICTED MODELS


Mean values:lWATER, lLABOR, lPEST, lWEED, lPOWER, lITIME


0.316456200 0.092873221 0.105063376 0.001715152 0.228705397



0.30772270 0.12256740 0.09865822 0.04281043 0.18900987




0.7319102 0.7405257

estimation final likelihood estimate

estimation of individual technical effiiciency scores










(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





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


Tables G2


stochastic translog function for rice farming at MFs

3.1.Unrestricted stochastic frontier estimation (step1)


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







- 'LABOR' is fulfilled at 142 out of 152 observations (93.4%)


244

244 Appendices





observations (0%)


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















(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


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














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"


RESTRICTED MODELS



colMeans( uEla )


0.350872298 0.169291063 -0.004173004 0.147438633 0.028320491


colMeans( caEla )

WATER LABOR POWER ITIME

PEST

2.721284e-01 1.843117e-01 -1.508169e-16 1.944268e-01

6.851226e-02



colMeans ( riceFinalIDM [ , c( "uEfficiency", "cEfficiency" ) ]

)


0.5590765 0.5549092

estimation final likelihood estimate




production function)The dependent variable is logged






(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





efficiency

1 0.26788346

2 0.80914416

3 0.88186694

4 ...

5 ...

6 ...

150 0.26715793

151 0.25281330

152 0.12471164


248

248 Appendices

Tables G1


stochastic translog function for rice farming at TEFs

4.1.Unrestricted stochastic frontier estimation (step1)


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







- 'LABOR' is fulfilled at 138 out of 148 observations (93.2%)







249

Appendices 249


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










Check for quasiconcavity#





(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


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














waldTest

Error: object 'waldTest' not found


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"


RESTRICTED MODELS




0.3677595 0.2222485 0.2485890 0.0977435 0.1014768

Partial production elasticities of the restricted model#


0.41151297 0.19167464 0.21295311 0.10305827 0.09292656




0.8195621 0.8049358




production function)The dependent variable is logged


cannot find a parameter vector that results in a log-likelihood

value

larger than the log-likelihood value obtained in the previous

step



(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




252

252 Appendices


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



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

OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS...

Documents

Transcript of OPTIMAL ALLOCATION OF WATER IN VILLAGE IRRIGATION SYSTEMS...