Infrastructure and Poverty in Viet Nam · 2016. 7. 17. · Viet Nam is poor both in terms of...

LSIfS R. Is/1Living StandardsMeasurement StudyWorking Paper No. 121

Infrastructure and Poverty in Viet Nam

Dominique van de Walle

.j

I'

a

____________ ___________ ____________ __________.

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

Pub

lic D

iscl

osur

e A

utho

rized

LSMS Working Papers

No. 49 Scott and Amenuvegbe, Sample Designsfor the Living Standards Surveys in Ghana andMauritania/Plans de sondage pour les enquetes sur le niveau de vie au Ghana et en Mauritanie

No. 50 Laraki, Food Subsidies: A Case Study of Price Reform in Morocco (also in French, 50F)

No. 51 Strauss and Mehra, Child Anthropometry in Cote d'lvoire: Estimatesfrom Two Surveys, 1985and 1986

No. 52 van der Gaag, Stelcner, and Vijverberg, Public-Private Sector Wage Comparisons andMoonlighting in Developing Countries: Evidencefrom Cote d'Ivoire and Peru

No. 53 Ainsworth, Socioeconomic Determinants of Fertility in C6te d'ivoire

No. 54 Gertler and Glewwe, The Willingness to Payfor Education in Developing Countries: EvidenceflomRural Penr

No. 55 Levy and Newman, Rigidite des salaires: Donnees microeconomiques et macroeconomiques surl'ajustement du marche du travail dans le secteur moderne (in French only)

No. 56 Glewwe and de Tray, The Poor in Latin America during Adjustment: A Case Study of Peru

No. 57 Alderman and Gertler, The Substitutability of Public and Private Health Carefor theTreatment of Children in Pakistan

No. 58 Rosenhouse, Identifying the Poor: Is "Headship" a Usefiul Concept?

No. 59 Vijverberg, Labor Market Performance as a Determinant of Migration

No. 60 Jimenez and Cox, The Relative Effectiveness of Private and Public Schools: Evidencefrom TwoDeveloping Countries

No. 61 Kakwani, Large Sample Distribution of Several Inequality Measures: With Application to C6ted'lvoire

No. 62 Kakwani, Testingfor Significance of Poverty Differences: With Application to Cote d'Ivoire

No. 63 Kakwani, Poverty and Economic Growth: With Application to C6te d'lvoire

No. 64 Moock, Musgrove, and Stelcner, Education and Earnings in Peru's Informal Nonfarm FamilyEnterprises

No. 65 Alderman and Kozel, Formal and Informal Sector Wage Determination in Urban Low-IncomeNeighborhoods in Pakistani

No. 66 Vijverberg and van der Gaag, Testingfor Labor Market Duality: The Private Wage Sector inC6te d'Ivoire

No. 67 King, Does Education Pay in the Labor Market? The Labor Force Participation, Occupation,and Earnings of Peruvian Women

No. 68 Kozel, The Composition and Distribution of Income in Cote d'lvoire

No. 69 Deaton, Price Elasticitiesfrom Survey Data: Extensions and Indonesian Results

No. 70 Glewwe, Efficient Allocation of Transfers to the Poor: The Problem of Unobserved Household lucome

No. 71 Glewwe, Investigating the Determinants of Househiold Welfare in Cote d'Ivoire

No. 72 Pitt and Rosenzweig, The Selectivity of Fertility and the Determinatnts of Human CapitalInvestments: Paranmetric and Semiparametric Estimates

No. 73 Jacoby, Shadow Wages and Peasant Family Labor Supply: An Econometric Application to thePeruvian Sierra

No. 74 Behrman, The Action of Human Resources and Poverty on One Anothier: What We HaveYet to Learn

No. 75 Glewwe and Twum-Baah, The Distribution of Welfare in Ghana, 1987-88

No. 76 Glewwe, Schooling, Skills, and the Returns to Governmenit Investmetnt in Education: AnExploration Using Data from Ghanla

No. 77 Newman, Jorgensen, and Pradhan, Workers' Benefitsfrom Bolivia's Emergency Social Fund

No. 78 Vijverberg, Dual Selection Criteria with Multiple Alternatives: Migration, Work Status, and Wages

No. 79 Thomas, Gender Differences in Househiold Resource Allocations

No. 80 Grosh, The Household Survey as a Tool for Policy Chanlge: Lessons from the Jamaican Surveyof Living Conditions

No. 81 Deaton and Paxson, Patterns of Aging in Thailand and Cote d'lvoire

No. 82 Ravallion, Does Undernutrition Respond to Incomes and Prices? Dominance Testsfor Indonesia

No. 83 Ravallion and Datt, Growth and Redistribution Components of Changes in Poverty Measure: A

(List continues on the inside back cover)

The Living Standards Measurement Study

The Living Standards Measurement Study (LSMS) was established by the WorldBank in 1980 to explore ways of improving the type and quality of household datacollected by statistical offices in developing countries. Its goal is to foster increaseduse of household data as a basis for policy decisionrmaking. Specifically, the LSMSis working to develop new methods to monitor progress in raising levels of living,to identify the consequences for households of past and proposed governmentpolicies, and to improve communications between survey statisticians, analysts,and policymakers.

The LSMS Working Paper series was started to disseminate intermediate prod-ucts from the LSMS. Publications in the series include critical surveys covering dif-ferent aspects of the LSMS data collection program and reports on improvedmethodologies for using Living Standards Survey (Lss) data. More recent publica-tions recommend specific survey, questionnaire, and data processing designs anddemonstrate the breadth of policy analysis that can be carried out using LSS data.

LSMS Working PaperNumber 121


Dominique van de Walle

The World BankWashington, D.C.

Copyright C 1996The International Bank for Reconstructionand Development/THE WORLD BANK1818 H Street, N.W.Washington, D.C. 20433, U.S.A.

All rights reservedManufactured in the United States of AmericaFirst printing February 1996

To present the results of the Living Standards Measurement Study with the least possible delay, thetypescript of this paper has not been prepared in accordance with the procedures appropriate to formalprinted texts, and the World Bank accepts no responsibility for errors. Some sources cited in this paper maybe iInformal documents that are not readily available.

The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) andshould not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of itsBoard of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy ofthe data included in this publication and accepts no responsibility whatsoever for any consequence of their use.

The boundaries, colors, denominations, and other information shown on any map in this volume do notimply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsementor acceptance of such boundaries.

The material in this publication is copyrighted. Requests for permission to reproduce portions of it shouldbe sent to the Office of the Publisher at the address shown in the copyright notice above. The World Bankencourages dissemination of its work and will normaliy give permission promptly and, when the repro-duction is for noncommercial purposes, without asking a fee. Permission to copy portions for classroomuse is granted through the Copyright Clearance Center, Inc., Suite 910, 222 Rosewood Drive, Danvers,Massachusetts 01923, U.S.A.

The complete backlist of publications from the World Bank is shown in the annual Index of Publications,which contains an alphabetical title list (with full ordering information) and indexes of subjects, authors, andcountries and regions. The latest edition is available free of charge from the Distribution Unit, Office of thePublisher, The World Bank, 1818 H Street, N.W., Washington, D.C. 20433, U.S.A., or from Publications,The World Bank, 66, avenue d'Iena, 75116 Paris, France.

ISSN: 0253-4517

Dominique van de Walle is an economist in the World Bank's Policy Research Department.

Library of Congress Cataloging-in-Publication Data

Van de Walle, Dominique.Infrastructure and poverty in Viet Nam / Dominique van de Walle.

p. cm. - (LSMS working paper: 121)Includes bibliographical references.ISBN 0-8213-3544-81. Infrastructure (Economics)-in Vietnam. 2. Poverty-Vietnam.

3. Vietnam-Economic conditions-Regional disparities. 4. Householdsurveys-Vietnam. I. World Bank. II. Title. III. Series.HC444.Z9C38 1996363'.09597-dc2O 95-52159

CIP

Contents

Foreword .............................................. viiAbstract ............................................. ixAcknowledgments ....................................... xl

1. Introduction ........................................... 1

2. Poverty and Infrastructure in Viet Nam, 1992-93. . . 22.1 Availability of Physical Infrastructure in Rural Viet Nam. 32.2 Drinking Water .52.3 Sewerage and Sanitation .102.4 Access to Irrigation .132.5 Sources of Energy .152.6 Roads .182.7 Summary and Implications .18

3. Explaining Crop Income . . . 203.1 Determinants of Crop Income ....................... 203.2 The Benefits from Irrigation: Policy Simulations .... ........ 293.3 The Cost of Household Labor ....................... 383.4 The Cost of Irrigation Expansion ..................... 39

4. Conclusions ........................................... 43

References . ............................................... 45

Tables

Table 1: Rural Infrastructure and Poverty in Viet Nam .................... 3Table 2: Rural Infrastructure in North and South Viet Nam ................. 4Table 3: Source of Drinking Water in Rural and Urban Areas of

North and South Viet Nam (%) ........................... 7Table 4: Source of Drinking Water by Region (%) ....................... 9Table 5: Toilet Facilities in Rural and Urban Areas of North and South Viet Nam (%) . 12Table 6: Toilet Facilities by Region (%) ............................. 13Table 7: Average per Capita Square Meters of Irrigated, Non-irrigated,

Other and Total Land ................................. 14Table 8: Average per Capita Square Meters of Irrigated, Non-irrigated,

Other and Total Land by Region .......................... 15Table 9: Lighting Source in Rural and Urban Areas of North and South Viet Nam (%) 16Table 10: Cooling Fuel in Rural and Urban Areas of North and South Viet Nam (%) 16Table 11: Variable Definitions and Summary Data ....................... 22Table 12: Regression Results: Crop Incomes ........................... 26

v

Table 13: Marginal Effect on Net Crop Income Allowing for Interaction Effects ..... 28Table 14: National Distribution of Impacts of Irrigation Under Different Scenarios .... 31Table 15: Regional Distribution of Per Capita Impacts of Simulation 1 .... ....... 32Table 16: Regional Distribution of Per Capita Impacts of Simulation 2 .... ....... 33Table 17: Regional Distribution of Per Capita Impacts of Simulation 3 .... ....... 34Table 18: Regional Distribution of Per Capita Impacts of Simulation 4 .... ....... 35Table 19: Regression Results: Family Labor Costs ....................... 40Table 20: Marginal Effect on Family Labor Costs Allowing for Interation Effects .... 42

Figures

Figure 1: Safe Water Sources in Rural Viet Nam ........................ 8Figure 2: Sources of Safe Water in Viet Nam .......................... 8Figure 3: Sanitation Facilities in Rurl Viet Nam ........................ 11Figure 4: Sanitation Facilities in Urban Viet Nam ........................ 12Figure 5: Total and Irrigated Annual Land Distribution in Viet Nam, 1992-93 ... ... 17

vi

Foreword

Viet Nam is poor both in terms of household living standards and physical infrastructure.How important are future infrastructural investments likely to be in promoting pro-pooreconomic growth in Viet Nam? This is an important question for the government and donorsalike, as Viet Nam moves through its transition to a market economy.

This study uses the Viet Nam Living Standards Survey of 1992-93 to examine theassociation between household living standards and the level of access to variaus infrastructuralservices. It also explores in depth the distributional impact of an expansion in irrigationinfrastructure. The paper is part of a larger effort in the Policy Research Department tounderstand how public spending policies affect household welfare.

Lyn Squire, DirectorPolicy Research Department

vii

Abstract

Viet Nam has poor physical infrastructure and high levels of income poverty. What rolemight better infrastructure play in reducing poverty in Viet Nam? This paper explores the linkbetween poverty and lack of infrastructure using the 1992-93 Viet Nam Living Standards Survey.The household data indicate that although there are some regional and urban-rural imbalances,in general access to infrastructure is not very different between poor and non-poor-infrastructure tends to be bad for everyone. Simulations of the potential benefits from anexpansion of irrigation infrastructure and under certain assumptions about how it would bedistributed, suggest that the policy would be redistributive, representing proportionately greatergains to the poor. The most pro-poor impacts would occur in Viet Nam's poorest regions andunder a policy which targeted irrigation expansion to small per capita landholding households.The average annual economic rate of return of the irrigation investments considered would beat least 20 percent. The paper also finds evidence that various constraints over and above thatpresented by lack of irrigation appear to diminish the benefits of irrigation to poor and non-pooralike.

ix

Acknowledgments

Financial assistance from the World Bank Research Committee (RPO BB67883) isgratefully acknowledged. I would like to thank Shanta Devarajan, Paul Glewwe, FrannieHumplick, Nauman Ilias, Jeanie Litvack, Amit Mohindra, Martin Ravallion, Hedy Sladovich,and Tom Wiens for their help and useful comments.

xi

1. Introduction

Various arguments can be made as to why basic infrastructure investments in a countrysuch as Viet Nam would reduce poverty.' One is that the poor have least access toinfrastructure and so will benefit most from new investments. If the non-poor have captured allthe benefits of past infrastructure projects and are now satiated then new projects must benefitthe poor. Another argument is that the poor are concentrated in sectors of the economy whererates of return to infrastructure investments are high. For example, the poor in Viet Namdepend heavily on agriculture, and rural infrastructure investments could have high agriculturalreturns.

This paper attempts to throw light on these arguments by asking: How large, and howpro-poor are the gains from infrastructure investments-specifically irrigation-likely to be? TheViet Nam Living Standards Survey (VNLSS) household data-collected during 1992-93-contains much information which is suggestive in examining this question. The paper onlyaddresses this question in any depth for irrigation. The data have partly influenced thischoice-the attraction of modelling irrigation is that it is household specific, and so there isample scope for identifying interaction effects with other variables and assigning benefits at thehousehold level.

The paper begins in section 2 by linking household living standards as revealed in theVNLSS with the level of various infrastructural services. Using standard descriptive techniques,an overall picture of the state of infrastructure, and how access varies by standards of living,is provided. Section 3 then attempts to explore in much greater depth one aspect ofinfrastructure-irrigation-and its association with living standards. Here marginal, as opposedto average, effects of irrigation expansion are estimated and the distributional implicationsassessed. Farm household crop incomes are modelled as functions of household characteristics,community characteristics and land, irrigated and non-irrigated. The impact of irrigation onfamily labor inputs is also explored. The final section draws some conclusions.

1. The paper defines economic infrastructure to consist of services from public utilities such as sanitation, powerand water supply and public works such as road and transport networks and irrigation systems (World Bank 1994a).Such services are distinguished by the fact that they share technical features (such as economies of scale) andeconomic features (such as spillover from users to non-users)' (World Bank 1994a, p. 2). For these reasons,govermment provision is often seen to be necessary. Linkages between poverty and infrastructure are discussed inWorld Bank (1994a), Lipton and Ravallion (1995) and Jimenez (1995). For sector specific discussions see Howeand Richards (1984), Binswanger et al. (1993) and Goldstein (1993).

1

2. Poverty and Infrastructure in Viet Nam, 1992-93

Except where noted, the analysis is based on the nationally representative 1992-93 VietNam Living Standards (VNLSS) survey. The survey covers 4800 households (23,790 persons)of which 3840 (19,094 persons) are rural, and includes a separate questionnaire on thecommunes in which sampled rural households are found.2 Collected information covers a widespectrum of aspects of living standards. The household survey touches upon access to and usageof infrastructure facilities in the context of its focus on household members' activities, incomesources, health, education, housing and so on. The community survey provides detailedinformation on the availability of infrastructure services in each rural household's commune ofresidence. It does not cover urban areas. For certain types of infrastructure, the communitysurvey is the sole source of information in the VNLSS. For others, details are also providedat the household level. However, the latter are often conditional on the household's usage andso tend to provide a skewed view of "access". For example, for households who do not reportan illness or whose member's illness was not externally treated, the survey reveals nothing aboutthe household's accessibility to health facilities. The commune level data must also be treatedwith caution. Because communes vary in size and distances differ, the figures do not reveal aUthat we would ideally like to know about household access to infrastructure services. These datawere supplemented by a number of field trips to rural areas of the North, Center and South ofViet Nam during 1993 and 1994.

Throughout, the paper uses household consumption expenditure per person as the welfareindicator. Since prices vary spatially, each household's expenditure is deflated by the regionspecific poverty line relative to the national poverty line.3 This provides a measure ofhousehold per capita expenditures at what can be termed "all Viet Nam prices". Al monetaryunits are also converted into real values in this way. The analysis is thus based on realexpenditure values representing purchasing power parity throughout the country. For thedistributional analysis in section 3 and the figures, individuals are ranked by the convertedhousehold per capita consumption expenditures and placed into 14 class intervals defined on percapita expenditures.

In the following sections, the paper first briefly looks at the general availability ofphysical infrastructure in the rural communes to which households belong. It then turns toaccess to drinking water, sewerage and sanitation, irrigation, energy sources and roads for eachhousehold in both urban and rural areas.

2. Expansion factors are not needed as the survey is self-weighted. The community questionnaire relies oninterviews of village leaders, health care workers, teachers and local govermment officials.

3. Regional poverty lines are estimated based on the "cost of basic needs methodology (Ravallion 1994), anddetailed in World Bank, 1994c. Deflating by region specific poverty lines is an alternative to using a regional priceindex. Because the weighing diagram used in deriving poverty lines tends to be more appropriate to the poor thanthat typically used in spatial price indices, their use is often preferred for investigations concerning poverty.

2

2.1 Availability of Physical Infrastructure in Rural Viet Nam

Tables 1 and 2 combine information on household living standards from the VNLSShousehold level survey with information on infrastructure facilities in each household's communeof residence from the community schedule. As mentioned, this is possible only for ruralhouseholds. Table 1 shows availability across various household groups for all of rural VietNam while Table 2 desegregates this information across North and South household groups. Forexample, (from first row, Table 1) 70.2% of the population as a whole are estimated to live incommunes which have a passable road, while this is true of 74.7% of people living in 'non-poor" households and of 67.3% of those living in "poor" households (with consumption perperson above and below the poverty line, respectively). Using a lower poverty line (arbitrarilyset at close to two thirds of the national poverty line) 72.8 and 62.5 % of those living in non-poor and poor households respectively live in communes with a passable road.

7'able 1: Rural Infrastructure and Poverty in Viet NamPercent of rural ,,pulation living in communes Headcount

with this infrastructure Index of Poverty (% poor

High Poverty Line Low Poverty Line among those with thisinfrastructure)

High Povwrty LowiNFRASTRUCTURE Total Non-Poor Poor Non-Poor Poor Line Poverty Line

Passable road 70.2 74.7 67.3 72.8 62.5 58.6 22.3

Passenger transport 52.3 56.2 49.8 54.0 47.3 58.2 22.7

Electricity 43.1 47.2 40.6 45.8 35.3 57.6 20.6

Pipe-borne water 5.2 7.2 3.9 5.5 4.2 45.8 20.3

Post office 34.4 36.2 33.2 34.9 32.9 59.0 24.0

Lower secondary school 87.9 87.7 88.0 88.7 85.4 61.2 24.4

Upper secondary school 9.7 10.7 9.1 10.2 8.3 57.3 21.5

Clinic 93.3 93.6 93.1 93.7 92.1 61.0 24.8

Total 61.1 25.1

Note: The table combines data from the household and community questionnaires. Poor defined by higherpoverty line are those with yearly per capita expenditures deflated by regional poverty line which are less thanthe national poverty line of 1,209,300 Dongs. Poor defined by lower line are those with per capitaexpenditures deflated by regional poverty line which are less than (0.65)*national poverty line. Electricity isdefined as most households in commune have it; pipe-borne water is defined as at least some households haveit.

Source: 1993 VNLSS.

3

Table 2: Rural Infrastructure in Noith and South VWet Nam

Percent of rural population living in communes Headcount Index of Povertywith this infrastructure (% poor among those with this

High Poverty Line Low Poverty Line infrastructure)

High LowINFRASTRUCTURE Total Non-Poor Poor Non-Poor Poor Poverty Line Poverty Line

RURAL NORTH

Passable road 76.8 89.6 70.4 82.5 62.5 61.1 23.1


Electricity 55.9 68.1 49.8 61.1 42.6 59.4 21.6

Pipe-borne water 3.5 6.2 2.2 4.1 2.1 41.9 17.0

Post office 27.7 29.5 26.7 28.4 25.9 64.3 26.6



Clinic 93.9 97.1 92.3 95.3 90.4 65.6 27.3

Total 66.7 28.4

RURAL SOUTH

Passable road 58.3 56.5 60.0 57.3 62.4 52.4 20.3


Electricity 20.2 21.7 18.8 21.3 15.5 47.4 14.6

Pipe-borne water 8.1 8.4 7.8 7.8 9.7 49.0 22.8

Post office 46.5 44.4 48.6 45.2 52.1 53.2 21.3



Clinic 92.2 89.3 95.0 91.2 96.7 52.4 19.9

Total 50.9 19.0

Note: The table combines data from the household and community questionnaires. Poor defined by higherpoverty line are those with per capita expenditures deflated by regional poverty line which are less thannational poverty line of 1,209,300 Dongs. Poor defined by lower line are those with per capita expendituresdeflated by regional poverty line which are less than (0.65)*national poverty line. Electricity is defined as mosthouseholds in commune have it; pipe-borne water is defined as at least some households have it.

Source: 1993 VNLSS.

Infrastructure for social services-schools and clinics-is much more widely accessiblethan other physical infrastructure such as electricity and water (Tables I and 2). There areclinics in communes accounting for 93% of the total rural population, lower secondary schoolsin communes covering 88% and primary schools (not reported) exist in every sampled ruralcommune. Facilities tend to be somewhat more prevalent in the North. Differences between

4

poor and non-poor are not large. Thus, according to the VNLSS, communes tend to be quitewell-provisioned in at least basic social services. However, the data also remind us that thequality of social services may leave a lot to be desired. For example, although all surveyedrural communes report having a primary school, 20% of children not attending school say thisis because the school is too far; and 64% of communes complain of poor material conditions asthe number one problem facing their commune's primary school.

Forty-three percent of the rural population live in communes in which "most" householdshave electricity.4 The variation from North (56%) to South (only 20%) is striking. Pipe-bornewater is even less frequently present in communes. Only 5% of the rural population reside incommunes where at least some households have piped water. This percentage is somewhathigher in the South. The availability of electricity and piped water is also related to livingstandards, with the poor less likely to make their home in communes where these are obtainable.Particularly for water in the North, headcount indices for households in communes with thisinfrastructure are considerably lower than for the population at large.

Some of these data must be interpreted carefully. For example (as noted), the surveyindicates that 70% of the rural population are found in communes serviced by a road which ispassable year round. Two caveats should be mentioned. First, in the South, coastal areas andparts of the North, canals and waterways are widely used to transport goods and passengers, sothat roads may not be the relevant entity. Second, the survey gives little indication of the qualityof the roads or how it defines "passable". Based on casual observation during my field workin rural Viet Nam, it seems likely that being passable by a motorcycle or bicycle may have beensufficient to qualify as "passable". For these reasons, the availability of passenger transport maybe a more informative indicator of accessibility. Tables 1 and 2 thus include this variable as aproxy for the presence of a serviceable road or waterway. Around half the population are incommunes in which some kind of passenger transport is available. Transport is more frequentlyfound in the South probably reflecting widespread use of boats as well as road vehicles there.There is also a more pronounced difference between poor and non-poor in terms of access toa passable road and transport in the North, indicating the remoteness of some of the pooresthouseholds in the North.

In the rest of section 2 the household questionnaire is used to further examine access tospecific infrastructure services at the household level in both urban and rural areas.

2.2 Drinking Water

Over half the population of Viet Nam (52 %) secure their drinking water primarily fromwells which are not equipped with a purnp. A further 20% obtain it mostly from rivers, lakesand other water bodies, while another 11 % rely on rainwater. There are sharp regional and

4. The questionnaire asks whether 'most' or 'just a few' have electricity.

5

urban-rural differences, as well as some variation related to living standards. Table 3 presentspercentages of various population subgroups according to their drinking water source. In ruralareas the pattern closely follows the national one: wells without pumps are most common,followed by water bodies and rainwater. However, wells are much more prevalent in the Northwhere they are by far the most common source of drinking water (71 %). The population in therural South rely somewhat more on surface water (41 %) than wells (33%). Public standpipesand private taps, whether inside or outside the residence, are almost non-existent in rural areas.Comparable data for Pakistan, Ghana, Tanzania and Peru indicate that 14 percent or more ofthese countries' rural populations have access to piped water compared to none in Viet Nam.'

Differences between rural poor and non-poor are not large, though the poor almostalways have less access to the more desirable sources of drinking water. Figure 1 which plotshow use of a water drinking source varies across expenditure per capita groups reinforces thisconclusion. Though the slopes show a tendency to slightly incline or decline as living standardsrise, on the whole the variation across expenditure groups is not dramatic.

Differentials between rich and poor are more pronounced in urban than in rural areas.This is illustrated by Figure 2 which adds the urban pattern to the rural to give the nationalequivalent of Figure 1. Access to water taps exhibits a steady rise beyond the seventh classinterval while use of wells steadily drops from about the sixth. Table 3 shows that the betteroff are considerably more likely to have access to an inside tap, outside tap or public standpipethan the poor. Still, although piped water systems are limited to urban areas, less than half thetotal urban population have access to tap water (public or private). Again, this compares poorlywith the urban areas of our 4 previous comparator countries, where the lowest access is foundin Pakistan at 57% of urban households.6 The data clearly show how inadequate these networkscontinue to be. There are also distinct North-South differences between urban areas. Privateindoor taps are more prevalent in the South (44 versus 24%) while publicly provided standpipesare more standard in the North (16 versus 4%). Wells with pumps are also more frequent inthe South (10 versus 2%). As in rural areas, wells without pumps are the most common sourceof drinking water for the North's urban populations (43%). It is striking to note that as muchas 14% of the South's urban non-poor population relies on assorted water bodies, while 30%of the poor do so.

5. These countries are chosen as comparators because they have Living Standard Measurement Surveys whichfollow the same methodologies and ask similar questions. The data on access to piped water in rural areas is thefollowing: Ghana: 13.5 percent of households (1991/92 Ghana Living Standards Survey); Tanzania: more than 21.5%of households (1993/4 Human Resources Development Survey) though this number excludes. obtaining water froma neighbor's piped water supply; Pakistan: 15 % of households (Pakistan Integrated Household Survey 1991/2); andPeru: 43 % of the rural Sierra population (1991 Living Standard Measurement Survey).

6. Results on access to piped water in urban areas are the following: Ghana:73.5 percent of households (1991/92Ghana Living Standards Survey); Tanzania: more than 56 % of households (1993/4 Human Resources DevelopmentSurvey) though this number excludes obtaining water from a neighbor's piped water supply; Pakistan: 57% ofhouseholds (Pakistan Integrated Household Survey 1991/2); and Peru: 91 % of the rural Sierra population (1991Living Standard Measurement Survey).

6

Table 3: Source of Drinking Water in Rural and Urban Areas of North and South Viet Nam (%)Rural North Rural South

Non-poor Poor Total Non-poor Poor Total

Private Tap 2.2 0.1 0.8 0.4 0.0 0.2

Public Standpipe 1.3 0.1 0.5 0.1 0.5 0.3

Well w/ pump 2.0 1.4 1.6 10.6 7.6 9.0

Well no pump 68.2 72.2 70.9 31.8 34.1 33.0

River/lake/pond 7.1 15.2 12.5 38.6 43.1 40.9

Rainwater 17.7 9.3 12.1 15.5 11.7 13.6

Other 1.4 1.6 1.5 3.0 3.0 3.0

Urban North Urban South


Inside tap 33.2 9.0 24.2 52.9 13.0 44.2

Outside tap 7.9 3.9 6.4 4.1 1.2 3.5

Public standpipe 18.6 11.8 16.1 3.9 2.6 3.6

Wel w/pump 2.4 1.1 1.9 11.3 5.4 10.0

Well no pump 31.1 61.9 42.6 10.3 39.5 16.6

River/lake/pond 1.2 4.0 2.3 13.6 30.3 17.3

Rainwater 4.8 5.3 5.0 3.9 8.0 4.8

Total 100 100 100 100 100 100

Note: The figures are % of persons in each subgroup according to their household's primary source of drinkingwater. Totals may not add up to 100-remainder is attributable to 'other". Private inside and outside taps areaggregated for rural areas. Bottled water is one of the options though it is rare.

Source: 1993 VNLSS.

7

Figure 1: Sqfe Water Sources in Rural Viet Nam

Population %70

60

so

40

30

20

10

0I 2 3 4 5 6 7 a 9 lo I 1 12 13 14

Per Capita Expenditure Group1+Well w/Pip Well w/o.Pump o.Surieace 4Raisn

Figure 2: Sources of Safe Water in Viet Nam

Population %80

70

60

50

40

30

20

10

0

I 2 3 4 5 6 7 a 9 10 I 1 12 13 14

Per Capita Expenditure Group

|4Pvt. Tap 4-Well v:Surface 44Rain

8

Table 4 presents some of the same information disaggregated across urban and rural areasof !he seven geographical regions. Among rural areas, the deltas-particularly theMekong-stand out. In contrast to all other rural regions where populations overwhelminglyrely on wells without pumps, in the Mekong use of water from rivers and lakes, rain and wellswith pumps is higher than elsewhere. In both deltas, over 20% depend on rainwater. A largemajority of the urban populations of the Northern Uplands and the North Coast also procuretheir drinking water from wells (no pumps). It is noteworthy that piped water is found primarilyin the Red River, Central Coast and Mekong Delta regions reflecting the better provisionedurban centers of Hanoi, Danang, and Ho Chi Minh City respectively.

Table 4: Source of Drinking Water by Region (%)Northern Red North Central Central South Mekong

RURAL Uplands River Coast Coast Highlands East Delta Total

Well w/pump 0.44 3.5 0.5 0.8 0.0 7.0 11.4 4.3Well no pump 72.8 54.7 85.5 83.1 79.4 74.2 9.2 57.4River/lake/pond 18.3 11.5 6.7 13.5 19.2 5.2 58.3 22.6

Rainwater 5.5 25.7 7.0 0.0 0.0 1.0 20.7 12.6

Other 3.0 1.3 0.2 1.6 0.5 12.0 0.0 2.1

URBAN

Inside tap 0.0 35.9 0.0 36.4 -- 64.8 15.5 33.8

Outside tap 0.0 12.1 0.0 7.3 -- 2.3 5.2 5.0

Public standpipe 0.0 38.3 0.0 11.1 -- 5.0 1.8 10.1

Well w/pump 1.3 0.8 0.0 3.9 -- 4.4 17.9 5.8

Well no pump 80.0 0.8 100.0 39.0 -- 23.4 7.2 30.1

River/lake/pond 6.5 1.4 0.0 1.0 -- 0.0 41.3 9.5Rainwater 8.3 9.7 0.0 0.1 -- 0.2 11.1 4.9

Total 100 100 100 100 100 100 100 100

Note: The table gives the percent of persons in each subgroup according to their household's primary source ofdrinking water.

Source: 1993 VNLSS.

The VNLSS data also suggest that up to 80% of the population lives within 100 metersof their source of non-piped drinking water and almost all (98%) within 1 kilometer. Close to79% of all households obtain bath and laundry water from the same source. Almost all the restalso live within a kilometer of their household's source for non-drinking water needs. Thoughthere may be seasonal variations not captured by the VNLSS-the extended dry season mayrequire collection from surface water at greater distances (World Bank 1990)-these data do not

9

lend support to the claim that water collection presents a severe burden on women and children(UNICEF 1994, NEDECO 1993).

Finally, it is important to point out that the data tells us nothing about the quality orsafety of the obtained drinking water. Piped water is reputed to often go untreated. Wells arefor the most part shallow and the water prone to contamination. Indeed, one study found thatup to 80% of wells were of unacceptable standard and harbored harmful bacteria thoughhouseholds still preferred them to alternative sources (UNICEF 1994). Given sewerage andsanitation conditions (see below), the presumption must be that surface water is rarely safe fordrinking. Evidence on the country's health profile and high incidence of water-related diseasesimplies severe water and sanitation problems. Viet Nam's National Programme of Action forChildren (NPA) reckons that 21% of the rural population have a hygienic and ample watersupply, though UNICEF cautions that this is on the high side of most estimates (UNICEF 1994).

2.3 Sewerage and Sanitation

Untreated industrial and residential waste waters tend to be dumped straight into existingsewerage systems and waterways. Waste from flush toilets enters sewers directly or via septictanks and is eventually discharged into rivers and other water bodies. Waste water systemsgenerally manage both flood and waste waters. Other than that provided through septic tanks,there is little treatment (World Bank 1990). For the most part, sewer systems were built pre-1954 in the North and pre-1975 in the South. Coverage is limited and performance of existingnetworks is poor.

A fraction of the population is serviced by conventional sewerage systems. The VNLSSindicates the style of toilet used by each household. The survey lists flush, double vaultcomposting latrines (DVCL), pit latrines, other, and no toilet as choices. The first three areconsidered hygienic and desirable relative to other methods by the Ministry of Health. "Other"which includes bucket and fishpond latrines, toilets hanging over water bodies, and animal andhuman waste manure tanks, are of lower standard and not officially sanctioned.

Human wastes are used extensively in agriculture and aquaculture and are a majorconcern for rural sanitation. A 1989 national study revealed extremely high rates ofparasitism-as high as 90-95%-in villages where DVCL are common and excreta used asfertilizer, and a much lower rate-40%-in similar villages where fishpond latrines are prevalent(World Bank 1990). Although DVCLs are hygienic when operated correctly-including allowingsufficient composting time in a sufficiently large vault-in practice, proper use appears to berare. The NPA's figure for the percent of the rural population with access to adequate sanitationfacilities is 13% (UNICEF 1994).

Tables 5 and 6 show population subgroups according to the type of toilet used by thehousehold. Nationally, 26% of the population reports having no toilet. This indicates a worsesituation than that found in Peru (17%), Tanzania (5.3% of households) and Ghana (25% of

10

households).7 A further 33% of the population use pit latrines, 22% use "other" and the restflush and double vault composting toilets. A tiny percentage have access to flush toilets in ruralareas. While not commonly in use anywhere in the South, DVCLs are employed by 13% of therural North's population and by 21% of its rural non-poor. Pit latrines are by far the mostcommon toilet type in the rural North while other and no toilet are most prevalent in the South.The urban areas of the Central Coast and the South East account for most flush toilets.However, the former also betrays one of the worst waste management situations in the country.A household without access to a flush toilet in the urban Central Coast is most likely to haveaccess to no toilet at all. In its rural hinterland, the percentage of the population without toiletfacilities of any kind reaches a national high of 55 %. How household waste management variesacross expenditure groups can be seen in Figure 3 for rural areas and in Figure 4 for urbanareas. Figure 3 indicates a clear and steady decline in the proportion of the population withoutrecourse to any kind of toilet facility as living standards rise. The use of DVCLs starts at a verylow rate and rises with welfare levels, as does the use of flush toilets. Interestingly, pit latrinesexhibit more of a flat inverted u shape. These are used primarily in the mountainous areas ofthe North. The graph reflects the fact that the poorest in these regions probably have worseaccess than the less poor, while less of the higher expenditure population is found in these areas.In urban Viet Nam, patterns are less clear except in the case of flush toilets which exhibit asteady increase as per capita expenditures rise.

Figure 3: Sanation FaciMiies in Rural Viet Nam0Population %

60

50

40

30

20

10

01 2 3 4 5 6 7 8 9 10 11 12 13 14

Per Capita Expenditure Group|0flush *DVCL *Pit -Nonel

7. Sources as detailed in footnote 5.

11

Figure 4: Sanition Facilities in Urban Viet NamPopulation %

100

suo

60

40

20

01 2 3 4 5 6 7 8 9 1 0 1 1 1 2 13 14

Per Capita EXPeDditure Group UfiFluh -#DVCL -&Pit -None

Table 5: Toilet Facilities in Rural and Urban Areas of North and South Viet Nam (%)

Rural North Rural South

Toilet Type Non-poor Poor Total Non-poor Poor Total

Flush 4.3 0.5 1.8 3.8 0.8 2.2

DVCL 21.4 8.2 12.6 0.7 0.2 0.4

Pit latrine 47.2 44.8 45.6 21.3 18.4 19.8

Other 7.9 16.8 13.8 51.3 39.7 45.4

No toilet 19.2 29.7 26.2 22.9 41.0 32A1


Toilet type Non-poor Poor Total Non-poor Poor Total

Flush 48.1 21.6 38.2 67.7 20.0 57.3

DVCL 11.1 12.4 11.6 1.9 2.0 1.9

Pit latrine 23.6 30.8 26.3 6.3 32.3 12.0

Other 0.8 3.3 1.7 13.3 29.9 16.9

No toilet 16.3 31.9 22.1 10.7 15.8 11.8

Total 100 100 100 100 100 100

Note: The table gives the percent of persons in each subgroup according to the toilet facility used by theirhousebold. DVCL is a double vault compost latrine.

Source: 1993 VNLSS.

12

Table 6: Toilet Facilities by Region (%)Northern Red North Central Central South Mekong

RURAL Uplands River Coast Coast Highlands East Delta Total

Flush 0.1 1.0 0.8 7.3 2.1 5.9 0.9 1.9

DVCL 11.4 13.3 12.0 13.7 2.2 0.2 0.2 8.3

Pit latrine 58.8 54.6 34.3 20.2 53.2 40.2 6.2 36.4

Other 6.5 16.2 25.4 4.2 2.0 18.3 63.3 25.1

No toilet 23.2 14.8 27.4 54.6 40.4 35.4 29.4 28.3

URBAN

Flush 2.0 50.4 3.8 60.7 -- 82.8 21.9 47.4

DVCL 10.5 8.1 42.4 6.0 -- 0.00 4.6 6.9

Pit latrine 64.8 19.7 31.1 6.4 -- 14.4 8.7 19.4

Other 0.9 0.6 11.0 0.5 -- 0.00 40.5 9.1

No toilet 21.8 21.2 11.7 26.4 -- 2.9 24.3 17.2

Total 100 100 100 100 100 100 100 100

Note: The table gives % of persons in each subgroup according to the toilet facility used by their household.DVCL is a double vault compost latrine.

Source: 1993 VNLSS.

2.4 Access to Irrigation

Around half of Viet Nam's annual crop cultivation area is currently irrigated. Irrigationneeds are largely supplied by surface water (Vu and Taillard 1993). Irrigation includes bothlarge-scale (networked investments) and small-scale (wells, bore-holes) systems. The two majorriver deltas are characterized by complex hydraulic systems dating back hundreds of years whichincorporate navigation, flood control, drainage, saline intrusion control and irrigation functions.The latter rely on a system of canals with pumping stations and on-farm water controlarrangements. These networks are in a state of severe disrepair. It is estimated that byrehabilitation of the existing infrastructure alone, there is the potential to expand irrigation tosome 700,000 hectares assuming there is sufficient water (SPC et al. 1989). Outside the deltas,irrigation-like other physical infrastructure-is less well developed, though there are expansionpossibilities. In some areas, this requires storage dam construction and gravity irrigationsystems. In others the development of small scale irrigation systems such as based on smallelectric pumps drawing water from reservoirs and natural water bodies is more feasible.

13

Some features of the distribution of irrigation infrastructure based on information fromthe VNLSS can be seen in Tables 7 and 8, and in Figure 5.s The tables present mean percapita square meters of total land and its various components, including irrigated and non-irrigated annual crop land and perennial land, across poor and non-poor groups (Table 7) andacross Viet Nam's seven geographical regions (Table 8). Differences between the rural Northand South are the most striking feature of Table 7. First, overall per person land amounts arequite a bit higher in the South as are all components with the exception of "other" land. But,the most notable difference is in the variation between poor and non-poor land endowments.These are relatively equitable in the North, though the poor have less irrigated land. However,these differences are dwarfed by the disparities in the South. On average, the Southern poorhave access to less than half the amount of land that the non-poor control. Table 8 reveals someof the geographical variation-attributable in large part to environmental terrain and variablepopulation densities. Average total land cultivated per person varies from 702 m2 in the RedRiver Delta to 1977 m2 in the Mekong Delta, while irrigated land ranges from 17 in the CentralHighlands to 713 M2 in the Mekong River Delta. Figure 5 combines the geographical anddistributional aspects of land allocation. Total land and total irrigated annual land are graphedseparately (both in square meters per capita) by per capita expenditure groups for all of ruralViet Nam and individually for the 7 regions. Past land reform has ensured relatively low intra-regional inequality of access to land (Vu and Taillard 1993). Significant variation in farm sizeand land quality can be found between regions. Landholdings are more correlated with livingstandards in some regions. In particular, all three regions of the South reveal a morepronounced positive association of land size with per capita expenditure levels. In general thedistribution of irrigation appears to be more equitable or about the same as that of total land.Based on the tables and Figure 5, there appears to be considerable room for expandingirrigation.

Table 7: Average per Capita Square Meters of Irrigated, Non-Irrigated, Other and Total LandRural North Rural South National

Non-poor Poor Total Non-poor Poor Total Non-poor Poor Total

Irrigated annual land 414.8 333.4 360.1 825.9 346.0 584.2 590.3 336.8 434.0

Non-irrigated annual 288.9 378.2 348.9 1149.8 660.9 903.5 656.6 454.4 531.9land

Perennial land 55.0 50.8 52.2 373.9 212.8 292.7 191.2 94.4 131.5

Otherland 173.8 126.1 141.8 156.0 29.7 92.4 166.2 100.1 125.5

Total land 932.4 888.6 902.9 2505.7 1249.4 1872.8 1604.3 985.8 1222.9

Note: Per capita m2 of land are calculated over the rural farm population. Other land includes forest, watersurface, and 'other' as defined in footnote 10.

Source: 1993 VNLSS.

8. According to the interviewer's instruction manual, irrigated land in the VNLSS includes all land which isirrigated either through a system of canals or by means of electric or petrol pumps which prevent flood and drought.

14

Table 8: Average per Capita Square Meters of Irrigated, Non-Irrigated, Other and Total Land byRegion

Northern Red North Central Central South MekongRural Uplands River Coast Coast Highlands East Delta Total

Irrigated 229.3 521.2 307.9 325.9 17.1 484.5 713.1 434.0annual land

Non-irrigated 697.2 78.7 349.3 321.2 1015.5 823.1 905.9 531.9annual land

Perennial land 76.7 31.8 58.9 42.8 398.3 354.9 256.6 131.5

Other land 311.0 69.8 112.8 50.3 27.2 101.3 101.5 125.5

Total Land 1314.2 701.5 828.8 740.2 1458.0 1763.7 1977.2 1222.9

Note: Per capita m2 of land are calculated over the rural farm population. Other land includes forest, watersurface, and "other" as defined in footnote 10.

Source: 1993 VNLSS.

2.5 Sources of Energy

The North is generally better endowed in energy, with coal reserves and hydroelectricpower plants, while the South is well known for its electricity shortages. Throughout Viet Nam,electricity is also available through diesel operated generators generally run by local authorities.Population percentages by lighting source are given in Table 9. Two methods dominate:electricity is used by 49% of the national population while 50% utilize gas, oil and kerosenelamps. There are pronounced differences between the North and South and across expendituregroups. Electricity networks are better developed and reputed to be more reliable in North VietNam and this is reflected in household usage. Practically half of the North's rural populationrely on electricity compared to less than a quarter (22%) in the rural South. For the non-poor,this rises to 64% in the North and 27% in the South. The rest make use of gas, oil andkerosene accounting for 61 and 80% of the poor and 34 and 68% of non-poor, in the North andSouth respectively. In urban Viet Nam a large majority of the population depends on electriclighting though the proportions are smaller in the South than in the North and among the lesswell off than the better-off. Again, comparison with countries for which similar data exists canhelp place the Viet Nam numbers in context. On the whole, household access to electricity inViet Nam compares favorably to that in both Ghana (69% of urban and 9% rural households)and Tanzania (35% and 1 %) but less well to Peru (95% of the total population).

Electricity is rarely used for cooking. Table 10 indicates that wood and leavespredominate in the rural areas of the North and wood dominates in the South's rural sector,while coal and kerosene also take on importance in the urban household sector. The differencesbetween poor and non-poor rural groups are small. In the urban areas of both parts of thecountry however, the poor are more likely to use wood and the better off to use coal orkerosene.

15

Table 9: Lighting Source in Rural and Urban Areas of North and South Viet Nam (%)Rural North Rural South


Electricity 63.9 38.0 46.6 27.4 16.4 21.8Battery 1.0 0.4 0.6 4.9 0.6 2.7Gas/oil/kerosene 34.2 60.9 52.0 67.7 80.3 74.1

Other 1.0 0.7 0.8 0.0 2.7 1.4



Electricity 97.0 80.9 91.0 91.0 64.9 85.3Gas/oil/kerosene 3.0 19.1 9.0 9.0 35.1 14.7

Total 100 100 100 100 100 100

Note: The table gives % of persons in each subgroup according to their household's source of lighting.Source: 1993 VNLSS.

Table 10: Cooking Fuel in Rural and Urban Areas of North and South Viet Nam (%)

Rural North Rural South


Wood 39.3 46.0 43.8 86.7 88.3 87.5Leaves 51.1 52.7 52.1 9.3 11.3 10.3

Coal 8.9 1.3 3.8 3.7 0.4 2.0



Wood 34.7 50.7 40.7 57.9 94.6 65.9Leaves 5.0 16.1 9.1 2.2 3.2 2.5

Coal 48.7 31.7 42.3 23.8 0.8 18.7

Electricity 3.9 0.4 2.6 1.9 0.0 1.5

Kerosene 7.7 1.1 5.2 13.7 1.4 11.0

Total 100 100 100 100 100 100

Note: The table gives % of persons in each subgroup according to their household's cooking fuel. Totals maynot add up to 100. The remainder is attributable to 'other' and kerosene and electricity in rural areas, and to otherand bottled gas in urban areas.

Source: 1993 VNLSS.

16

Figure 5: Total and Irrigated Annual Land Distribution in Viet Nam, 1992-93 (m2/person)

Northern Uplands Red River Delta South East

m~~~~~~~~~~~3

ff ' t L.2W44S.-:~~~~~~~~~~~~~~~~~~~~~~~~~~~~WPer C pi* a Expendhure CroupNrhCentral CentraldCoastoNational

co~ ~ ~ ~~~~~~~~~~~~~~~~~~~~M

0 0 11 a ;3~~~~~~~~~~~~~~~~~~~i

Per apit Exendiure rou

2.6 Roads

Other than the information (from the community schedule) of whether a road passesthrough the commune, there is little information in the VNLSS to illustrate the poor state of thecountry's transportation sector. The public road network consists of around 105,100 km ofroads: 11,400 of national roads, 14,200 of provincial level roads, 25,300 of district level roads,46,200 of village roads and 2,600 urban roads and 5,400 special roads (World Bank 1994b).In 1992, around 12 percent of Viet Nam's existing road network was paved compared to 30%of India's in 1985 and 48% of Indonesia's in 1990 (World Bank 1994b). Average road densityis quite low at 0.32 km per sq km of land area and 1.6 km per 1000 inhabitants. Notunexpectedly, densities are highest in the two deltas and lowest in the more mountainousregions. Together with the rest of the infrastructure stock in Viet Nam, the roadnetwork-dating largely from before the 1970s in the South and pre-1954 in the North-is oldand in severe disrepair. This is also true of other transport infrastructure including inlandwaterways (a 40,000 km network), ports, and the railway system (World Bank 1994b).

2.7 Summary and Implications

The current state of physical infrastructure in Viet Nam is clearly poor by moststandards. Nearly a third of the rural population live in communes without a passable road.Nearly half do not have access to passenger transport. More than half do not have electricity.Barely half of annual crop land is irrigated. All but 5% of the rural population live incommunes where no one has access to piped water. There are also marked differences in accessto infrastructure between urban and rural areas with urban areas generally favored, as well asconsiderable regional imbalances.

The poor tend to have worse access to infrastructure than the non-poor. However, formany types of infrastructure the poor in rural Viet Nam do not have appreciably worse accessthan do the non-poor: many types of basic infrastructure are equally bad for both. Basicinfrastructure ventures will not automatically be redistributive. It cannot be argued that the non-poor already have adequate basic infrastructure and the poor have none such that newinvestments will necessarily benefit the poor.

What do these data imply for the distributional impacts of future investments ininfrastructure? The answer will depend on the marginal benefits from infrastructure investments.For example, take irrigation. If a household's land is fully irrigated then clearly the marginalbenefits to that household from expanding total irrigated area are zero. Those who benefit froma general expansion of irrigation, say, will be those who have non-irrigated land. If it were truethat the rich have fully irrigated land while the poor don't, then the benefits would go to thepoor. However, if anything it is the non-poor who have more of both non-irrigated and irrigatedland (Table 7 and Figure 5). Looking at Figure 5 one would not think that undifferentiatedirrigation infrastructure would be an important redistributive instrument as such; the poor would

18

benefit but probably less so than the non-poor.9 There are, however, a number of other factorswhich are correlated with the marginal benefits from irrigation. It is often argued for examplethat smaller farms are more productive so that marginal benefits may be higher for the poor.Or it might be argued that marginal benefits will tend to be higher for those with more humancapital, likely to be the not-so-poor. On balance, it is not clear what the outcome would be.Inferring the potential gains from irrigation using cross-sectional data thus requires controls forthese other factors. To properly address such questions we need to go beyond the simplydescriptive analysis and investigate marginal impacts.

9. It may be a different question if one could adequately target irrigation to the poor, but targeting can often bedifficult and costly (van de Walle 1995a).

19

3. Explaining Crop Income

This section attempts a detailed assessment of the likely distributional impacts of anexpansion in the irrigated land area. The data set contains detailed information on householdland assets and farm incomes. The vast majority of Viet Nam's population derives itslivelihoods from farming (van de Walle 1995b). The extent to which annual crop land isirrigated and water adequately managed are widely recognized as key factors in the productivityand success of agriculture at the local level. Although the aggregates hide considerable regionalvariation, around one half of agricultural annual crop land is currently under irrigation.10 Forthese reasons, it is possible to look closely at the impact of irrigation infrastructure onlivelihoods in rural Viet Nam. How large would the gains in incomes from further irrigationinvestments be? Would those gains be pro-poor? Such questions are of considerable interestgiven the multiple policy choices faced by policy makers in a country setting where budgets areheavily constrained and simultaneously challenged by a plethora of real investment andconsumption needs.

3.1 Determinants of Crop Income

In attempting to throw some light on these questions the paper looks at the determinantsof net farm crop income and the role played by irrigation. The size of the difference inmarginal returns between irrigated and non-irrigated land determines the income gains fromirrigating a unit of land. To quantify the gains from irrigation the paper posits the profit

function 4(p, L N L', z) giving the farm-household's maximum profit conditional on a vector

of prevailing output and input prices (p), amounts of non-irrigated (L N) and irrigated (L')annual crop land, and a vector of other fixed factors (z) including other types of agriculturalproduction land, human capital, location specific agro-ecological variables and other constraintsarising from market imperfections (such as the underdeveloped state of labor markets and

supervision costs of labor). In specifying z a wider range of variables are allowed than onewould normally posit in a profit function, recognizing that this is a transition economy in whichmarkets are still underdeveloped. For example, in many parts of Viet Nam householddemographic factors can be an important constraint on production arising out of labor-marketimperfections and institutionalized non-market modes of factor allocation.

10. In addition to annual crop land, households derive agricultural incomes from perennial land (used for perennialtree crops), forest land (natural forest or reforested areas used for inseminating of young plants and growing offorest tree crops), water surface land (for raising water products) and what I will call other land. The latter includesvarious other land categories listed in the survey: vacant lots and bald hills (land managed by household but notcultivated for at least 12 months); virgin land (burnt and fallow land); and other (area of road and dike sides, riverbanks, etc).

20

Output and variable input prices are assumed to vary between but not within communes,so these can be represented by a vector of commune dummy variables. For the jt household,the profit function is assumed to be linearized as follows:

(1) @i (p, Li Li, z ) = a + [37LjN + PiLi' + y

where the marginal returns to non-irrigated and irrigated land are given by

(2) 137 N (2)~ ~ ~ ~ ~~P =PO+ PlXdj + P2 z}+ P3L

and

(3) ~ ~ ~ ~ ~ ~ P <=PO+ Pldj+ P2Zj+ P3Lj'

respectively, and where d is a vector of regional dummy variables. The error term in (1) isassumed to be independently and identically normally distributed. The regression includes a fullset of commune dummy variables which are meant to pick up prices as well as any spatialvariations in other omitted or fixed factors, and should also capture the influences of communitylevel characteristics, including geographical and infrastructural variations. These spatial effectsare compressed into the 7 regional entities in their effects on marginal returns, though a full setof commune dummies is allowed in the intercept of the profit function. Profit is measured bynet farmrl crop income, net of variable costs." Table 11 lists the right hand-side variables andprovides a description and summary statistics.

11. Total revenue from agricultural land production includes all crops evaluated at harvest prices (missing valuesare replaced by average community prices); the value of crop byproducts consumed or sold (such as thatch, straw,trunks of cassava, maize, jute, etc.); land incomes (rents from other households and govermment assistance); andincome from leasing out farm production equipment. The total costs of production are then subtracted. Theseinclude hired labor expenses; costs of seeds and young plants; fertilizer, manure and insecticide costs; animal rental,transport, packaging and storage, equipment rental, repair and maintenance fees, fuel oil and electricity costs; anaccounting depreciation charge for owned farming equipment (5 %); taxes and fees to cooperative (such as forirrigation, crop protection, plowing); land payments such as rent for land leased in and land taxes paid togovernment or cooperative.

Transformation of home grown crops (such as producing cured tobacco, peanut oil, rice noodles) is notincluded, but treated as family off-farm enterprises using farm inputs. Note also that the opportunity costs ofhousehold farm labor are not included. Section 3.3 provides a test of whether the results are sensitive to thisassumption. Also note that profits from livestock raising are not part of net crop income.

21

Table 11: Variable Definitions and Summary Data

Variable Definitions

cropinc Net household crop income, 1993 Dongssick Dummy for household member being sick in last yearsexhhh Gender of household headhhsvgs Initial stock of household savings, 1993 Dongshhsize Size of the householdpropO6 Proportion of household members who are 6 years and youngerprop716 Proportion of household members who are 7 to 16 years, inclusivepfadlt Proportion of household members who are female adults (17 +)pmadlt Proportion of household members that are male adults (17 +)hedl Years of primary education of household headhed2 Years of post-primary education of household headoedl Years of primary education of other adult household members (17 +)oed2 Years of post-primary education of other adult household members (17 +)irrigated Irrigated annual crop land area (m2)nonirrigated Non-irrigated annual crop land area (m2)

perennial Perennial land area (m2)forest Forest land area (m2)waterland Water surface land area (m2)

otherland Other land area (m2)proplt Proportion of annual land which is long-termpropauct Proportion of annual land which is auctionedproppriv Proportion of annual land which is privatepropshare Proportion of annual land which is sharecropped/rentedpropall Proportion of annual land which is allocatedurban Dummy variable for urban residencenu Dummy variable for the Northern Uplands regionrr Dummy variable for the Red River Delta regionnc Dummy variable for the North Coast regioncc Dummy variable for the Central Coast regionch Dummy variable for the Central Highlands regionmk Dummy variable for the Mekong River Delta region

22

Table 11 (continued) Summary DataVariable Obs Mean Std. Dev. Min Max

cropinc 3049 2282069 2391173 -6061975 2.63e+07

sick 3049 0.933 0.250 0 1

sexhhh 3049 0.809 0.393 0 1

hhsvgs 3049 939223 8990056 0 4.66e+08

hhsize 3049 5.033 1.992 1 15

propO6 3049 0.156 0.176 0 .75

prop716 3049 0.213 0.204 0 .80

pfadlt 3049 0.327 0.169 0 1

pmadlt 3049 0.282 0.160 0 1

hedl 3049 4.379 1.114 0 5

hed2 3049 2.513 2.842 0 16

oedl 3049 6.872 5.372 0 42

oed2 3049 4.111 5.287 0 51

irrigated 3049 2267.58 3997.50 0 80000

nonirrigated 3049 2605.92 5632.34 0 97150

perennial 3049 678.43 2169.50 0 50000

forest 3049 279.22 1970.98 0 50000

waterland 3049 122.89 1203.53 0 50000

otherland 3049 217.50 2106.11 0 70000

proplt 3049 0.20 0.380 0 1

propauct 3049 0.023 0.092 0 1

proppriv 3049 0.227 0.341 0 1

propsbare 3049 0.043 0.165 0 1

propall 3049 0.507 0.431 0 1

urban 3049 0.057 0.231 0 1

nu 3049 0.183 0.387 0 1

rr 3049 0.275 0.447 0 1

nc 3049 0.178 0.383 0 1

cc 3049 0.090 0.286 0 1

ch 3049 0.020 0.139 0 1

mk 3049 0.20 0.40 0 1

23

Explanatory variables aiming to capture the influence of household characteristics includehousehold size and composition; gender of the household head; years of primary schooleducation (0 to 5) and of any additional education of the household head; ditto for all other adulthousehold members (aged over 17)12; access to various kinds of land"3; proportions of annualcrop land in various forms of ownership"4; the stock of household savings, and dummyvariables for urban residence and whether a household member was sick in the last year.15

For the most part, land is not allocated to households through the market mechanism inViet Nam. Thus the usual concerns about regressing outputs on inputs chosen by the household(and hence endogenous) do not arise in this setting with respect to land. And although theremay be endogeneity of placement regionally for irrigation, the existence of irrigation in an areacan be treated as exogenous at the household level given that there has been little to no mobilityin the country. However that does not mean that there are not other, more subtle, forms ofendogeneity. Over a long period, land has traditionally been allocated by administrative fiat.There may be some omitted variable in the model which also influences the amount of non-irrigated and irrigated land allocated to the household. Then land will be correlated with theerror term in the regression giving a bias. There is nothing that can be done about this moresubtle potential form of endogeneity problem in a cross-section data set such as the VNLSS.

A number of functional form specifications were tested including linear, semi-log, anddouble log forms, with and without quadratics in land and education. Explanatory variableswere also tested logged and unlogged. OLS is used on a regression sample consisting of allfarm households (including some urban households who farm) for which the data are complete(3049 households). Results are reported in Table 12. Two regressions are presented: one,referred to as the unrestricted model, contains all variables, while the restricted model is theoutcome of pruning variables with t-ratios below 1 in the unrestricted model and following

12. The education of school age children is omitted to avoid possible endogeneity problems. The latter could resultif, for example, households with unobserved factors contributing to higher farm profits are more likely to pull theirchildren out of school.

13. See footnote 10.

14. It may be important to distinguish between land ownership rights. During survey implementation and beforethe new land law of late 1993, land was classified into 5 types: i) Allocated: (applicable in the North) land fromthe cooperative's land fund which was distributed to households according to number of workers. ii) Auctioned:(applicable in the North) around 5 to 10 % of cooperative's land which was reserved for bidding by those whowanted more land. This land was more expensive and had a 3 to 5 year tenure depending on the region. iii) Longterm use: the South's equivalent of the North's allocated land. iv) Private: land used by the household as a gardenarea. Often of lower quality, this land required no payment. v) Sharecropped or rented.

15. In addition, cash income received from the Government Social Fund, dummy variables for the household head'sethnicity, age, religion, language, and whether born in present residence were tried and found to be insignificantand to have no effect on the other regression coefficients.

24

iterations. Note however, that in an effort to make Table 12 more wieldy, variables with t-ratiosbelow 1 and the 119 commune dummy variables are not reported."6

Because of the many interaction effects, the impact of individual variables is difficult toassess directly from the regression coefficients. Table 13 presents the calculated total marginaleffects and t-statistics of such variables evaluated at the mean points. Turning to this table first,a few things are worth noting. Annual land, both irrigated and non-irrigated, and perennialland all have high significant positive overall effects on crop income. The highest is fromirrigated land, an impact which is more than twice as large as that of non-irrigated land. Thereare also high returns to perennial land (higher than to non-irrigated annual land). Returns toforest, water surface, and other land are much lower, though only water surface land in therestricted model is statistically significant.

All education variables are found to have pronounced and significant positive impacts oncrop income. In particular, one extra year of primary education for the head of householdincreases crop income by an amount equal to about 8 percent of mean crop income. There aredecreasing returns to the education of the household head (the coefficient on hedl is muchgreater than that on hed2), but not to that of the rest of the household (coefficients on oedl andoed2 are roughly similar). Finally, larger households have higher crop incomes. This impliesthat family labor endowments matter in agricultural production probably because labor marketsare underdeveloped. The marginal effects of demographic composition variables are notstatistically significant.

The regression shows strong, though diminishing, impacts of annual crop land-bothirrigated and non-irrigated-on crop income (Table 12). A higher share of allocated orauctioned annual land significantly increases annual crop profits. Household size matters,though composition effects do not appear to be of consequence independently of the interactionwith land. However, many of the interaction effects are significant and of interest. Forexample, household size appears important in its interactions with nearly all land variables (allare significant in the restricted model), as does the share of females in household adults. Theseeffects are positive in most cases and demonstrate the importance of family labor inputs, andparticularly female ones."7 They suggest a dependency on own household labor, probablyindicating the presence of labor market imperfections and the inability of many households tohire labor in or out. In contrast, interacting household size and the female adult share withirrigated land, and the female adult share with non-irrigated land, results in a significant negativeimpact. A careful investigation finds that when the sample is partitioned across geographicalregions, this effect holds only in the South, and particularly in the Mekong delta, though it isclearly strong enough to influence estimation results for the national model.

16. Full regression details are available from the author.

17. Women play a major role in agriculture in much of Viet Nam. The VNLSS indicates that women averagedthe equivalent of 182.5 8 hour days work on the family farm and men 159.4 days.

25

Table 12: Regression Results: Crop IncomesUnrestriced Model Restricted Model

cropinc Coefficient t-ratio Coefficient t-ratio

urban 1093640 1.08 837371 1.25sick -318465.4 2.59 -317534.3 2.61hhsize 81451.7 2.12 67405.6 2.61propO6 -586514.9 1.18 -429757.1 2.08hedl -468646.3 2.69 -527229.5 3.12hedl*hedl 67862.6 2.66 76644.0 3.13oedl*oedl -1932.7 2.48 -1566.7 3.22oed2 21023.8 1.24 27026.3 2.85irrigated 352.40 4.27 362.59 4.78

irr*irr -0.0030 4.42 -0.0030 4.82nonirrigated 238.40 3.81 206.72 8.20nonirrig*nonirrig -0.0036 9.60 -0.0034 10.63perennial -277.04 1.73 -238.17 2.11perennial*perennial -0.0097 6.43 -0.0099 7.10forest -372.88 1.35 -80.33 1.02forest*forest -0.0026 1.38 -0.0022 2.0waterland*waterland -0.0401 3.80 -0.0042 5.12otherland -611.27 1.22 -426.07 2.16otherland*otherland -0.0024 1.32 -0.0016 1.19propauct 1116555 2.54 1048419 2.62proppriv 325505.5 1.49 215568 1.50

propall 470198.4 2.10 337486.1 2.07hedl*irrigated 47.87 6.06 49.80 6.93

hedl *otherland -113.39 2.58 -111.97 3.40hed2*irrigated -6.46 1.53 -5.10 1.46

hed2*perennial 21.90 2.53 25.85 4.09hed2*forest 23.01 1.62 26.73 3.23hed2*waterland 72.61 1.58 30.70 2.12hed2*otherland 33.45 1.51 22.67 1.14oedl*irrigated 20.74 8.03 20.74 8.58oedl*nonirrigated 7.27 3.38 5.66 3.51oedl*perennial 5.42 1.21 5.10 1.20oedl*forest -21.37 1.72 -12.80 3.20oedl*otherland -49.20 4.66 -39.95 4.48oed2*irrigated -4.179 2.18 -4.57 2.52oed2*nonirrigated 1.741 1.04 1.990 1.34oed2*perennial -10.914 2.65 -10.694 2.76oed2*otherland 33.814 4.0 25.259 3.60hhsize*irrigated -35.991 6.94 -35.865 7.51hhsize*nonirrigated 4.639 1.13 7.243 2.22

26

Table 12 (continued)

Unrestricted Model Restricted Model

cropinc Coefficient t-ratio Coefficient t-ratio

hhsize*perennial 52.933 4.62 51.712 4.79hhsize*forest 37.473 1.68 28.892 2.38hhsize*otherland 79.081 2.62 64.675 2.28pfadlt*irrigated -176.63 2.16 -189.17 2.55pfadlt*nonirrigated -137.02 2.07 -115.10 2.30pfadlt*perennial 610.10 3.33 628.69 4.12pfadlt*otherland 1941.44 4.18 1751.47 4.48pmadlt*irrigated -162.40 1.71 -142.17 1.70pmadlt*perennial 289.39 1.92 290.70 2.63prop716*irrigated 155.85 2.03 132.86 1.94rr*irrigated 271.75 4.06 260.97 4.17ff*forest 135.35 1.03 74.85 1.19mk*irrigated -67.71 1.94 -83.64 2.65mk*perennial -158.92 2.94 -147.73 2.95nu*irrigated 255.81 3.47 241.34 3.43nu*perennial -199.58 2.27 -215.25 2.77nu*otherland 434.86 1.20 361.41 5.02nc*perennial -218.53 3.02 -205.53 2.96nc*otherland 528.01 1.39 480.14 3.09cc*iffigated -203.38 3.63 -211.93 3.97cc*nonirrigated -152.25 2.62 -147.69 3.03cc*perennial -228.68 1.26 -226.31 1.26ch*irrigated -973.79 1.66 -1051.07 1.81ch*nonirrigated -134.37 2.65 -130.90 3.29ch*perennial 310.57 4.37 326.09 4.96ch*waterland 5195.78 1.31 4078.46 2.70

Number of obs = 3049 Number of obs = 3049F(233, 2815) = 19.06 F(183, 2865) = 24.04Prob > F = 0.0000 Prob > F = 0.0000R-square = 0.6120 R-square = 0.6057Adj R-square = 0.5799 Adj R-square = 0.5805Root MSE = 1.5e+06 Root MSE = 1.5e+06

Note: The restricted model results from the pruning of all variables with t-ratios less than I in the unrestrictedmodel. The unrestricted model also contained the following variables: demographic composition variables,pnum716, pfadlt, pmadlt and interactions with land variables; education variables: hed2, hed22, oedl, oed22 andinteractions with land; land: waterland and interactions between types of land and regions; propit, propshare.

27

Table 13: Marginal Effect on Net Crop Income Allowing for Interaction Effects

Unrestricted model Restricted Model

Marginal effecton net crop Marginal effect on

Variable income t-ratio net crop income t-rauio

irrigated annual land Dongs/lOOm2 48,571.5 16.1 48,226.3 17.9

non-irrigated annual Dongs/100m2 19,994.0 8.1 21,876.2 16.3land

perennial land Dongs/100m2 21,269.1 4.1 23,385.0 6.7

forest land Dongs/100m2 8,722.1 1.9 6,325.0 1.8

water surface land Dongs/lOOm2 -86,491.2 0.1 15,729.2 3.3

other land Dongs/100m2 10,422.2 1.2 2,346.8 0.4

household size Dongs/person 59,065.9 2.0 62,154.9 2.8

prop female adults Dongs/% -2,366.5 0.1 78,456.6 0.4point

prop male adults Dongs/% -1,165.1 0.2 -125,228.9 0.6point

prop aged 7-16 Dongs/% 1,041.6 0.2 301,322.2 1.9point

primary ed (head) Dongs/year 191,875.8 3.0 232,762.7 4.1

ed > primary (head) Dongs/year 38,584.9 2.0 22,132.4 2.3

primary ed (other Dongs/year 35,094.1 2.5 31,466.8 4.7adults)

ed > primary (other Dongs/year 22,195.7 1.8 20,094.3 2.6adults)

mean crop income 2,282,069 2,282,069

Note: Marginal effects are evaluated at mean points.

The latter results can be interpreted to indicate that the market labor constraint does notbite as much for irrigated land in the South. For households with larger amounts of irrigatedland, family labor becomes less of a constraint. It is the way in which household laborinfluences crop income which is important here. If the household could buy or sell as muchlabor time as it required then one would not expect household demographics to be significantin the crop income equation. The fact that they are significant can then be taken as animplication of labor market failure. Family labor becomes an input to production but the extentto which this matters depends on how much market conditions apply to each household. Theresults indicate that family labor is generally a constraining factor in farm production in the

28

North, but less so in the South and particularly less so for households with lots of irrigated landin the Mekong delta." A test of the linear restriction that the overall influence of householdsize is zero when evaluated at mean sample values is not rejected for the South (though thenumber is positive), but it is found to be positive and significant in the North. (The same isfound for the share of female adults.) Thus, the importance of the labor market constraint variesfrom household to household and from region to region.

Education is found to be of considerable importance to agricultural productivity. Theprimary schooling of the household head is important on its own and is found to be convex inits impact on crop incomes, implying increasing returns to schooling. Interaction effectsbetween education variables and land are generally positive. Notable exceptions include anegative effect of both primary education variables interacted with other land and post-primaryyears of education of adults other than the household head interacted with both irrigated annualand perennial land. Interestingly, the results imply that primary education interacts strongly withirrigated land to increase crop income while post-primary education does not.

Finally, almost half of the 119 commune dummies are significant at the 5 % significancelevel. There are also non-negligible spatial differences in the effects of both irrigated and non-irrigated land, and other land types. These effects are all relative to land impacts on cropincomes in the South East (left out of the regression) and show expected signs and magnitudes.

3.2 The Benefits from Irrigation: Policy Simulations

Irrigation is in part a private good. Individual households can invest in a bore-hole orpump but require capital to do so and face capital market constraints. It is also a collective goodwhere substantial resources are needed to set things up and benefits are distributed across manypeople. The combination of these two factors helps explain why there could be underinvestmentin irrigation in Viet Nam. Some amount has to be publicly provided while another amount canbe provided privately, provided credit is available for that purpose.

What are the potential benefits from irrigating a unit of non-irrigated land, holding totalcultivable land area constant? How would those benefits be distributed across expendituregroups? And how might they vary with other factors? Here, an attempt is made to quantifythose benefits using the above regression model.

18. A number of things lend support to this interpretation. Commune level collected wage data provide someevidence that labor markets are better developed in the South. Wage rates were unavailable, and hence missing inthe survey, for a larger proportion of surveyed households in the North than in the South for both agricultural andunskilled non-agricultural work. Salinger (1993) provides further corroboration for the underdeveloped state oflabor markets in North relative to South Viet Nam. In addition, more so than elsewhere, the Mekong delta ischaracterized by large areas which are either irrigated or not irrigated. As the paper shows in section 3.3, irrigationappears to increase the labor input requirement. It may be surmised that labor markets are likely to have adaptedand more fully developed in areas of the Mekong which have large irrigated farms.

29

It is assumed that non-irrigated annual land can feasibly be irrigated (though costs mayvary widely across regions), and that perennial and other land types cannot. Conversion ofannual land to irrigation may be through rehabilitation and expansion of existing irrigationnetworks or through new construction. The paper considers the distributional impact whichwould arise from irrigating around 10 percent of the annual crop land currently not underirrigation. The impacts are examined under four possible policy scenarios for how the irrigationexpansion is distributed across farms. In each case the same total amount of land is broughtunder irrigation.

A first simulation simply extends irrigation to all farm households who have non-irrigatedland. Because some farm households have little or no non-irrigated land, in practice a policyof bringing 10% of non-irrigated annual land under irrigation allows up to a maximum 500 m2

of newly irrigated land to farm households to reach a mean of around 260 m2 per household.

The current distribution of the share of annual land under irrigation tends to be bimodal.There is a tendency for farm households to have all their annual land irrigated, or none at all.It may be argued that a policy of converting 10 percent of the country's non-irrigated land toirrigation would more realistically be implemented in areas where farm households presentlyhave little irrigation. Simulation 2 limits the expansion of irrigation to farms currently lackingaccess to irrigated annual land. The increment is zero if the farm has any irrigated land or hasno non-irrigated land.

Finally, simulations 3 and 4 target smallholders. Section 2.4 showed that poor farmhouseholds tend to have less annual, as well as less irrigated land than non-poor households.It is therefore of interest to examine how the distributional effects of bringing 10 percent of thecountry's non-irrigated annual land under irrigation would differ if those improvements weretargeted to households with low total annual crop land holdings. Simulation 3 distributes theirrigation on the basis of low total household annual landholdings, while simulation 4 targets onthe basis of low per capita annual landholdings. Once again, the simulations hold total annualland constant. Given the existing distribution of irrigated and non-irrigated land acrosshouseholds, simulation 3 results in irrigating all the non-irrigated land of households who haveless than 3250 m2 total annual land and simulation 4, the non-irrigated land of all with less than620 mi per person.

The expected marginal benefit from irrigation-the change in household crop income

from irrigating one unit of non-irrigated land-can be estimated by 0 - pO, where 0 and ON

are estimated at each data point, using the parameter estimates for the relevant interaction effectsin Table 12 applied to the household-specific values of the relevant variables. To simulatepolicy impacts one can then multiply the marginal benefit by the household specific increment.

However, the marginal benefit function, - ,Bis only a first-order approximation and strictlyvalid for small changes only. For estimating the gains from discrete changes, a more accuratemethodology is to recalculate the value of the function after substituting constrained land changesinto the profit function as follows:

30

(4) A nj = C(p, LjN - ALj, Lj + AL,, zj) - % (pj, L1 , Li', z1 )

where, for example in the case of simulation 1 (and as appropriate for the others)

ALj = 0 if LjN = 0

= Li if L, S 5O

so that the exact amount of land shifted into irrigation is appropriate to each household'scircumstances-zero for those who have no unirrigated land and up to 500 m2 for householdswho do. The results are termed the "simulated total impacts." The distribution of the impactin per capita Dongs and as a percent of per capita household expenditures across all farmhouseholds classified into expenditure groups, is shown in Table 14 for the four simulations.Tables 15 to 18 show the results disaggregated across regions (with the exception of the CentralHighlands where there are too few observations for the breakdown).

Table 14: National Distribution of Impacts of Irrigation Under Different Scenarios

Total impact as % of householdSimulated total impacts (per capita) expenditure

Expenditure group % cffarm('000 Dongs/personlyr) population 1 2 3 4 1 2 3 4

1 0-500 4.1 9099.3 6700.7 11878.1 19310.3 2.13 1.57 2.78 4.52

2 501-600 4.1 12645.5 15513.5 13267.7 15314.7 2.29 2.81 2.40 2.77

3 601-700 6.9 16059.4 13118.9 14217.7 14604.0 2.47 2.02 2.18 2.24

4 701-800 9.6 14132.9 12312.0 15185.0 16657.2 1.88 1.64 2.02 2.22

5 801-900 9.2 13750.1 11443.8 15974.4 10287.4 1.62 1.35 1.88 1.21

6 901-1000 9.2 14775.1 10704.2 14482.1 9108.2 1.56 1.13 1.53 0.96

7 1001-1100 9.1 12924.6 9183.1 11281.8 11328.6 1.23 0.88 1.08 1.08

8 1101-1250 11.5 11396.7 10261.8 11310.6 7977.2 0.97 0.87 0.96 0.68

9 1251-1400 8.0 15035.7 12469.1 14279.8 14383.4 1.14 0.94 1.08 1.09

10 1401-1550 7.3 14185.6 13287.7 13194.1 9476.3 0.96 0.90 0.90 0.64

11 1551-1800 7.2 10240.5 10226.4 6653.9 2836.9 0.62 0.62 0.40 0.17

12 1801-2200 6.9 9947.0 9328.1 10000.7 3602.2 0.50 0.47 0.51 0.18

13 2201-3000 5.0 10142.9 11421.5 5143.5 3433.6 0.41 0.46 0.21 0.14

14 3001-4500 2.1 10935.7 13185.7 5215.5 3900.1 0.28 0.34 0.13 0.10

Total 100 12844.6 11226.5 12221.3 10293.6 1.05 0.92 1.00 0.84

Note: Results are based on the unrestricted model. A conversion of 10% of non-irrigated annual land to irrigationis common to all simulations. Under simulation (1): irrigation is distributed to all households subject to feasibility;(2) irrigation is distributed only to households without irrigated land; (3) irrigation is targeted to households withlow total annual landholdings; and (4) irrigation is distributed to households with low per capita annual landholdings.

31

Table 15: Regional Distribution of Per Capita Impacts of Simulation INorthern Uplands Red River Delta North Coast Central Coast South East Mekong Delta

Total Simulated Total Simulated Total Simulated Total Total TotalExp. Simulated impact % total impact % total impact % total impact % Simulated impact % Simulated impact %group total impact of hh exp impact of hh exp impact of hh e-xp impact of hh exp total impact of hh exp total impact of hh exp

1 27985.5 6.53 8335.3 1.88 14788.3 3.35 1855.0 0.47 1216.7 0.30 741.1 0.172 31584.0 5.80 5188.6 0.93 18062.1 3.28 2989.3 0.54 12305.0 2.25 -699.6 -0.133 35212.4 5.37 9028.8 1.39 15553.4 2.40 2645.0 0.41 13792.3 2.08 6001.8 0.914 34651.3 4.62 5162.9 0.68 16729.4 2.23 2704.7 0.36 10336.9 1.38 4027.0 0.545 35820.9 4.22 6623.6 0.78 18529.7 2.19 5195.5 0.61 14442.1 1.69 3895.0 0.466 35807.7 3.81 10776.5 1.14 20855.1 2.19 5480.6 0.57 6382.4 0.67 4785.5 0.507 30698.2 2.93 5547.2 0.53 19847.3 1.89 6768.6 0.64 14822.4 1.41 3409.2 0.33

8 33514.2 2.83 11012.6 0.93 18230.6 1.54 5822.9 0.50 4831.1 0.41 4704.2 0.409 33915.7 2.58 12650.2 0.95 22188.9 1.68 7073.4 0.54 14254.1 1.06 4175.5 0.3110 30031.7 2.05 9656.7 0.67 25121.2 1.70 4587.0 0.31 14492.9 0.96 8923.5 0.6011 29949.4 1.82 8392.4 0.50 22120.2 1.35 3201.8 0.20 18261.2 1.10 7429.5 0.4512 36663.7 1.86 10786.7 0.55 26043.7 1.32 6518.4 0.33 16186.4 0.83 7982.8 0.4013 29741.7 1.23 11884.4 0.47 22954.4 0.93 9928.3 0.41 21402.4 0.86 7954.7 0.3114 28907.6 0.84 8488.1 0.22 2900.7 0.08 10387.3 0.31 9820.4 0.20 11955.2 0.31

33211.4 3.00 8798.9 0.72 18767.2 1.96 5151.6 0.42 12805.3 0.86 5839.3 0.40

Note: Results are based on the unrestricted model. Under simulation I the conversion of 10% of non-irrigated annual land to irrigation is distributed to allhouseholds subject to feasibility.

Table 16: Regional Distribution of Per Capita Impacts of Simulation 2Northern Uplands Red River Delta North Coast Central Coast South East Mekong Delta

Total Total Total Total Simulated Total Simulated TotalExp. Simulated impact % Simulated impact % Simulated impact % Simulated impact % total impact % total impact %group total impact of hh exp total impact of hh exp total impact of hh exp total impact of hh exp impact of hh exp impact of hh exp

1 26082.1 6.08 14515.3 3.28 14568.9 3.30 521.3 0.13 1302.0 0.33 488.8 0.112 41024.3 7.53 8358.8 1.49 18248.0 3.31 3641.7 0.66 15214.6 2.78 2187.2 0.393 32702.6 4.99 8242.8 1.27 8442.5 1.30 -1617.5 -0.25 23844.4 3.60 9507.0 1.454 32412.3 4.32 3453.1 0.46 12476.5 1.67 2698.4 0.36 11907.8 1.59 6280.3 0.845 34634.2 4.08 2869.5 0.34 14847.7 1.75 5327.0 0.63 13564.9 1.59 6345.0 0.756 27266.9 2.90 4734.2 0.50 16719.4 1.75 6941.2 0.73 7792.4 0.82 8003.3 0.847 18956.6 1.81 4392.2 0.42 14756.4 1.41 3512.5 0.33 25629.1 2.43 5471.4 0.528 34830.1 2.94 6677.5 0.57 18470.4 1.56 4240.4 0.36 7473.9 0.63 7758.6 0.669 28508.3 2.17 5480.9 0.41 22731.8 1.73 3667.2 0.28 16530.1 1.23 6941.2 0.5210 33139.2 2.26 1548.1 0.11 24518.8 1.66 1473.0 0.10 21100.6 1.40 13724.1 0.9311 36228.6 2.20 0.00 0.00 20050.8 1.22 1858.6 0.11 28771.1 1.74 12848.4 0.7712 28678.7 1.46 1006.7 0.05 27010.3 1.37 4097.1 0.21 24500.3 1.26 13416.3 0.6813 24897.8 1.03 4849.1 0.19 16223.0 0.66 13118.5 0.54 29843.5 1.20 13759.6 0.5414 36872.1 1.07 4315.9 0.11 0.00 0.00 17885.7 0.53 7788.0 0.16 18823.8 0.49

Total 30499.2 2.76 4323.2 0.35 15755.8 1.65 4153.3 0.34 17769.4 1.19 9677.5 0.67

Note: Results are based on the unrestricted model. Under simulation 2 the conversion of 10% of non-irrigated annual land to irrigation is distributed onlyto households without irrigated land.

Table 17: Regional Distribution of Per Capita Impacts of Simulation 3Northern Uplands Red River Delta North Coast Central Coast South East Mekong Delta

Total Total Total Total Simulated TotalExp. Simulated impact % Simulated impact % Simulated impact % Simulated impact % Simulated Total impact total impact %

group total impact of hh exp total impact of hh exp total impact of hh exp total impact of hh exp total impact % of hh exp impact of hh exp

1 40040.0 9.34 0.00 0.00 16744.0 3.79 201.9 0.05 1287.6 0.32 1670.3 0.38

2 28688.4 5.26 8789.7 1.57 18282.8 3.32 1763.3 0.32 13122.9 2.40 1577.5 0.28

3 41771.8 6.37 2980.7 0.46 16166.5 2.49 -4232.3 -0.66 0.00 0.00 6011.3 0.92

4 32566.1 4.34 3814.9 0.50 22552.6 3.01 3308.1 0.44 23583.0 3.15 1821.5 0.24

5 48170.9 5.68 2630.6 0.31 30108.0 3.56 9489.9 1.12 15552.6 1.83 2026.1 0.24

6 30457.6 3.24 10523.6 1.12 18341.9 1.92 6740.2 0.71 2634.5 0.28 4037.7 0.43

7 22029.0 2.10 3662.2 0.35 21273.1 2.03 7292.6 0.69 16304.3 1.55 652.7 0.06

4A. 8 29176.5 2.46 9711.9 0.82 23350.0 1.97 7016.3 0.60 9025.0 0.76 -473.9 -0.04

9 31302.4 2.38 11247.7 0.85 17343.6 1.32 3902.9 0.30 27273.7 2.03 1390.3 0.10

10 36372.9 2.48 8862.0 0.60 31239.5 2.11 4978.5 0.33 4394.7 0.29 3578.1 0.24

11 21393.9 1.30 6116.1 0.37 24493.6 1.50 4465.9 0.27 10752.1 0.65 1530.2 0.09

12 30343.8 1.54 9582.8 0.49 41661.6 2.11 6564.8 0.33 11809.2 0.61 4639.1 0.23

13 44089.7 1.82 14057.4 0.56 22001.1 0.89 23895.0 0.98 7336.1 0.29 152.4 0.01

14 522.3 0.02 6134.8 0.16 0.00 0.00 9968.8 0.29 5342.3 0.11 5280.6 0.14

Total 32870.9 2.97 7061.2 0.58 21878.8 2.29 6166.9 0.98 12051.9 0.81 2144.9 0.14

Note: Results are based on the unrestricted model. Under simulation 3 the conversion of 10% of non-irrigated annual land to irrigation is targeted tohouseholds with low total annual landholdings.

Table 18: Regional Distribution of Per Capita Impacts of Simulation 4

Northern Uplands Red River Delta North Coast Central Coast South East Mekong Delta

Total Total Total Total Total Total

Exp. Simulated impact % Simulated impact % of Simulated impact % of Simulated impact % of Simulated impact % of Simulated impact %

group total impact of hh ctp total impact hh exp total impact hh exp total impact hh exp total impact hh exp total impact of hh exp

1 66093.4 15.42 0.00 0.00 23270.5 5.27 -12.8 -0.003 1287.6 0.32 156.1 0.04

2 51335.4 9.42 385.7 0.07 15460.6 2.81 12444.7 2.26 13122.9 2.40 1097.9 0.20

3 31943.0 4.87 2162.0 0.33 25377.8 3.92 -2818.3 -0.44 0.00 0.00 5435.2 0.83

4 43557.8 5.80 4634.6 0.61 21889.9 2.92 572.5 0.08 22153.2 2.96 -2719.8 -0.37

5 35629.3 4.20 2384.4 0.28 13680.9 1.62 3166.1 0.37 15552.6 1.83 649.7 0.08

6 26856.1 2.86 2441.6 0.26 21172.2 2.22 10350.7 1.08 2634.5 0.28 1183.4 0.12

, 7 27165.8 2.59 2837.3 0.27 21864.7 2.08 2006.5 0.19 16304.3 1.55 -2227.8 -0.21

8 26032.9 2.20 8312.0 0.71 21085.6 1.78 4904.6 0.42 10021.0 0.84 -473.9 -0.04

9 37681.2 2.87 11383.2 0.86 8878.4 0.67 5234.4 0.40 16070.5 1.20 2678.8 0.20

10 25623.7 1.75 3072.4 0.21 18073.4 1.22 6787.5 0.46 4394.7 0.29 1855.4 0.13

11 1761.8 0.11 4892.5 0.29 4787.3 0.29 4066.6 0.25 1635.3 0.10 1530.2 0.09

12 8570.6 0.43 5237.3 0.27 14804.9 0.75 6902.3 0.35 7262.8 0.37 2064.8 0.10

13 29762.1 1.23 2424.2 0.10 10597.2 0.43 25819.3 1.06 7336.1 0.29 152.4 0.01

14 522.3 0.02 6134.8 0.16 0.00 0.00 0.00 0.00 0.00 0.00 5280.6 0.14

Total 31234.6 2.83 4660.9 0.38 19199.2 2.01 5785.7 0.47 9284.6 0.62 906.9 0.06

Note: Resuls are based on the unrestried model. Under simulation 4 the conversion of 10% of non-irriated annual land to irriation is distrbuted to households with low

per capit annua landholdings.

Converting 10 percent of non-irrigated land to irrigation produces an increase in cropincomes equal on average to around 1 percent of mean household expenditures. This impliesan elasticity of 0.1. The elasticities vary only slightly across the simulations. However, thelevel and distribution of per capita impacts differs across national expenditure groups accordingto how the irrigation is distributed (Table 14). This reflects the method of allocating theirrigation expansion combined with the existing household distribution of irrigated and non-irrigated annual crop land and the influence of other household and community specific factorsentering the marginal benefit of irrigation function such as education, household size and region.Under equal distribution to all households subject only to land constraints (simulation 1), impactsare smaller at the lower and upper ends of the distribution but otherwise relatively steady acrossexpenditure groups. Simulation 2 tends to be more generous towards the upper end of thedistribution and less so at the bottom end though it is not altogether that different from impactsunder simulation 1. Targeting the irrigation expansion to smallholders results in larger absoluteimpacts at the lower end of the distribution which fall much more sharply when targeting is doneon the basis of per capita than household annual landholdings. Impacts under all 4 simulationsare certainly progressive-declining as a proportion of household expenditures as living standardsrise-and so inequality reducing. Progressivity is most pronounced for simulation 4 whichconfers large benefits to the poorest groups (worth 4.5 percent of household expenditures for thepoorest group and only 0.1 percent for the wealthiest expenditure group).

Gzins are very concentrated regionally (Tables 15 to 18). The potential benefits of.rrigation appear to be strongest in the Northern Uplands where simulated total impacts are

turgest for all simulations (mean impacts of up to 3 percent of mean household expenditures).IiTigation expansion is inequality reducing there, exceptionally so when irrigation expansion islut rgeted to low per capita landholding farm households. However, the net absolute gains tendto be relatively steady across the distribution of per capita expenditures in all except simulation4. The next most substantial impacts are found in the North Coast and South East regions. Inthe North Coast, impacts are generally inequality reducing on the whole, though the gradient ismuc h lower than for the Northern Uplands. Absolute benefit levels tend to increase withexpe. nditure class except under simulation 4 which, here too, is found to result in the mostprogi essive distribution of benefits. In the South East, total impacts tend to be larger for thebetter off (with the exception of simulation 4) and flat or only somewhat progressive whenexpre. ;sed as a proportion of household expenditures. The smallest total impacts are evidencedfor tha, Central Coast and the Mekong River Delta. In both regions the benefits are also farfrom pi ogressive. Indeed, the simulated per capita total impacts, though small, tend to increasefor high er expenditure groups. One interesting finding from the above is that concentration ofbenefits and progressivity appear to go hand in hand. Benefits tend to be higher where theirdistribuLti on is also more pro-poor. The results hint towards targeting irrigation expansion to theNorthern Uplands and North Coast regions, where absolute benefits are not only higher but alsowell distributed. These are also Viet Nam's poorest regions (World Bank 1994c; Dollar andGlewwe 1995).

Thet overall regional picture is quite robust across simulations. Nationally, there is notmuch of aii obvious tradeoff between the ways of distributing the irrigation across regions.

36

Interestingly, however, there is a distinct regional pattern to which simulation has the greatestimpact on regional absolute benefit levels. This no doubt reflects characteristics of how annualland is distributed regionally. Simulation 1 produces the highest absolute gains for both theNorthern Uplands and the Red River; simulation 3 for the North and Central Coasts; andsimulation 2 for the South East and Mekong. In each case these are contiguous regions.Simulation 4 is distinguished not by producing the largest benefits in any region but by tendingto favor the poor with larger absolute impacts and by producing the most progressive distributionof benefits almost universally across regions.

At first sight, the simulation outcomes appear surprising for the Mekong River Delta.They also appear robust. This is the country's primary producer of rice with, as yet, only halfits total cultivated area under irrigation. It is sometimes said that extending irrigation will enabledouble and triple cropping and boost production and incomes formidably in this ideal setting forpaddy cultivation. However, the Mekong delta situation is complex. Various characteristics ofthe region's ecosystem and economy appear to provide credible explanations for the simulationresults. Irrigation systems in the Mekong Delta have been plagued by problems of sea waterintrusion and acid-sulphate soils. In recent years, as supplementary areas are brought underirrigation in upstrean areas, the level and flow of the Mekong river has dwindled, resulting insalt water intrusion in previously productive irrigated fields downstream (NEDECO 1991).'9This has meant that only one crop can be grown annually or, under the worse case, thatcontinued rice cultivation is rendered impossible. In the latter case, the areas are oftentransferred to aquaculture activities such as the farming of brackish shrimp which can be veryprofitable but would be reflected in lower crop incomes. Furthermore, fully irrigated areas mayalso suffer from extensive flooding and water logging for long periods of the year. In such areasof the Mekong, what is needed is better water management and drainage control rather thanirrigation as such.

The issue seems to be that, because of the heterogeneity of irrigated land in the MekongDelta, it is hard to generalize about the impact of irrigation infrastructure there. If the dataallowed a separation between irrigated areas which suffer from salinity and acidity problems andother irrigated land, we would probably get strong impacts of additional irrigation investmentsin the Mekong River Delta. Irrigation can be very positive depending on whether it is upstreamor not. The results indicate low marginal benefits on average, where they are being averagedover a considerable amount of heterogeneity. The profits from irrigation vary by region butthere is also variation within region.

19. NEDECO 1991 quotes farmers in the center of the delta complaining about this.

37

3.3 The Cost of Household Labor

The costs of household labor inputs on the family farm were ignored in defining net cropincomes.30 This is entirely defensible if one is concerned solely with the impact of irrigationon family consumption (since the implicit payment for own-labor inputs is exactly matched bythe receipts leaving consumption unchanged). However, to assess the gains to farm profits thefamily labor cost should be debited. And, like other costs of production, family labor inputsmay well be related to whether land is irrigated or non-irrigated. It is not obvious that irrigationwould save on family labor. However, one then faces the long-standing issue of what wage rateone should use for valuing family labor inputs; with surplus labor in rural areas and supervisioncosts, the opportunity cost of family labor may be well below the market wage rates for similarwork.

Here I try to assess the possible bias in the above results due to the omission of familylabor input costs. The results may either over- or under-estimate the net returns to irrigation,depending on family labor requirements on irrigated versus non-irrigated annual land. Theshadow wage of family labor is somewhere between zero and the agricultural wage rate. Iassume that the shadow wage is a constant proportion (+) of the prevailing market wage. Thecost of family inputs to own-farm production by household j is then given by

(5) Fj, - tOkWkFjk

where wk is the wage rate for the k'th demographic-group (adult men and women, and children),

and Fjk is the labor time devoted to farm work by demographic group for household j.Information on wage rates are available in the community survey separately for men, womenand children for a series of tasks (preparation, planting/transplanting, weeding, andharvesting).2" However, the household survey does not include the time allocation for eachmember by those tasks. Furthermore, in practice the wage data are very incomplete reflectingthe lack of labor markets in many of the communes. Thus demographic-specific commune meanagricultural wages are formed over all tasks for which wage rates are recorded and these areused to value the time on all farm tasks by each household member. Missing data at thecommune level are then replaced by the regional means for males, females or children asappropriate.22

20. Recall that non-family labor costs are included.

21. Note that since the wage rates can only be obtained from the community schedule, they are not householdspecific.

22. Data are missing for 398, 430, and 2406 households on wage rates for men, women, and children respectively.

38

The parameter + is unknown. To get an upper bound estimate, family labor input costsare evaluated at agricultural market wage rates (so + = 1), and the family labor cost is regressedagainst the same right-hand-side variables used to explain crop incomes. The net marginalimpact of irrigation over non-irrigation on the cost of the family labor input can then becompared to the previously calculated net marginal effect of irrigated over non-irrigated land oncrop incomes. If there is no significant difference between irrigated and non-irrigated land, thenwe need not worry; for any value of + my results for crop income carry over to profits net offamily labor. If there is a difference then we can ask if there is an admissible value of < whichreverses the earlier conclusions.

The regression results (given in Table 19) indicate that irrigation tends to increase workon the family farm. It is also notable that, other things being equal, bigger households and oneswith a larger proportion of adults and teenagers tend to use more family labor. Table 20presents the total marginal effects of the main variables on labor costs allowing for all interactioneffects. The effect on the market value of family labor time of irrigation over non-irrigated landis estimated to be 7,279 Dongs per 100 m2. Subtracting this full amount (i.e., at + = 1) fromthe average net impact on crop incomes of converting 100 m2 of non-irrigated land to irrigationreduces the latter to 21,299 Dongs, a 25 percent decline. Of course this is an upper boundestimate which may considerably overestimate the costs. If the opportunity cost of family laboris half of the market wage (+ = 0.5), then the gain in farm profit from irrigating 100 m' of non-irrigated land is 24,939 Dongs, a 13 percent decline.

In conclusion, the earlier results overestimated the marginal effect of irrigation on farmprofits, though the difference is not prohibitively large, representing a maximum of 25% of theprevious estimates of net income gains.

3.4 The Cost of Irrigation Expansion

Information on the costs of irrigation expansion is hard to come by, and generalizationsacross regions and types of irrigation investments are risky. Nonetheless, even a rough senseof the cost-benefit appraisal can be useful. Irrigation project costs have been estimated by anumber of agencies for various regions. Estimated average costs-including for a World Bankirrigation rehabilitation project in the Central Coast and for a large number of water resourcedevelopment projects in the Mekong River Delta drawn up as part of the Mekong Delta MasterPlan-fluctuate around 85,000 Dongs per 100 mi2 ; these are averages over appraisals for multiple

39

Table 19: Regression Results: Family Labor CostsUnrestricted Model Restricted Model

laborcost Coefficient t-ratio Coefficient t-ratio

soxhh 158669.5 1.45 160862.6 1.52

hhsvgs -0.0231 3.58 -0.0239 3.78hhmi 246193.1 5.31 269214.9 6.93

prop716 1471817.0 2.29 1185507 4.24

pfadlt 1507441.0 2.20 1435968 4.09

pmadlt 1754666.0 2.50 1711478 5.2

oedloedl 2276.06 2.42 3135.26 5.03

oed2 -24043.93 1.17 -19321.13 1.19

oed2*oed2 -1079.62 1.53 -1059.71 1.68

irrigated 322.208 3.24 306.387 6.16

irrigated*irrigated -0.0021 2.52 -0.00258 4.66

nonirrigated -126.674 1.68 -58.595 1.20

noniffig*nonirrig -0.00252 5.65 -0.00290 7.23

perennial -197.220 1.02 -283.297 1.88

perennial*perennial -0.00858 4.72 -0.00887 6.51

waterland 12564.61 2.82 13290.42 3.07

propauct 842682.1 1.60 939847.9 1.95

propall 488905.0 1.81 509353.9 2.80hedl*perennial 46.6154 2.17 34.251 2.30

hed2*irrigated -8.9632 1.77 -12.528 3.30

hed2*nonirrigated -8.4584 1.97 -8.368 2.61

hed2*wateriand 119.046 2.15 57.577 1.67

oedal irrigated 9.620 3.09 9.627 3.96

oedlinonirrigated -5.151 1.99 -2.592 1.38

oodl*foremt 25.395 1.70 12.208 2.06

oed2*irrigated -2.499 1.08 -2.863 1.32

oed2*perennial -7.088 1.43 -7.862 1.90

o*d2ottherland 15.0682 1.48 6.371 1.91

hhsize*iffigated 13.962 2.24 13.641 2.70

hhsize*nonirrigated 17.257 3.48 15.766 3.55

hhsize*perennial 22.0446 1.60 27.969 3.14

pfadlt*irrigated -179.628 1.82 -178.334 2.60

pfadlt*nonimigated 182.699 2.29 131.447 2.00

pfadlt*perennial 273.755 1.23 327.912 1.90

pmadlt*nonirrigated 268.217 3.54 193.512 3.15

pmadlt*perennial 373.679 2.06 458.372 3.33

pmadlt*forest -684.857 1.77 -399.353 2.17

prop716*irrigated -318.899 3.45 -326.617 4.48prop7l6*perennial 203.644 1.14 300.683 1.91

rrfirrigated 85.881 1.065 99.855 1.34

rr*waterland -13158.55 2.99 -13419.56 3.10

mk"*irrigated -125.50 2.99 -115.506 3.25

40

Table 19 (contnued)

Unrestricted Model Restricted Modellaborcost Coefficient t-ratio Coefficient t-ratio

mk*nonirrigated 51.107 1.25 66.613 2.39mk*waterland -12923.81 2.93 -13184.76 3.05nw*irrigated -144.161 1.63 -133.200 1.60nw*nonirrigated 48.626 1.03 65.681 1.87nw*waterland -13165.85 2.99 -13408.45 3.10nc*irrigated -123.831 1.36 -118.265 1.37nc*nonirrigated 88.077 1.58 106.054 2.35nc*perennial -129.440 1.49 -104.656 1.36nc*waterland -12856.66 2.91 -13153.33 3.03cc*perennial -284.91 1.30 -266.521 1.27cc*waterland -106753.7 1.42 -106546.7 1.42ch*irrigated -798.508 1.13 -877.190 1.27ch*perennial -260.649 3.05 -245.386 3.20ch*otherland 789.428 1.58 782.188 3.02ch*waterland -13286.87 2.79 -13973.44 2.98

Number of obs = 3025 Number of obs = 3025F(232, 2792) = 16.32 F(232, 2792) = 16.32Prob > F = 0.0000 Prob > F = 0.0000R-square = 0.5756 R-square = 0.5756Adj R-square = 0.5404 Adj R-square = 0.5404Root MSE = 1.9e+06 Root MSE = 1.9e+06

Note: The restricted model results from the pruning of all variables with t-ratios less than 1 in the unrestricted model. Theunrestricted model contained exactly the same variables as the crop income regression reported in table 12.

projects, though the variance is low.23 For these costs, the estimated model indicates an annualgain in net crop income of around 28,600 Dongs per 100 m2, falling to a gain of 21,300 Dongsin farm profit at the maximum shadow wage for family labor. This represents a rate of returnof at least 25 to 35 % per year, assuming the project delivers such benefits indefinitely. But evenunder conservative assumptions of a project life of only 10 years and with the maximum shadowwage for family labor, the rate of return is about 20%.24

23. The World Bank project average costs are estimated at about 83,150 Dongs per square meter when excludingconsultant costs as well as physical and price contingencies. The Mekong Master Plan projects average 85,830Dongs per m2 (table 8.2), 87,650 Dongs (table 8.3), and 87,120 Dongs (table A2.1) all in NEDECO 1993.

24. These are internal rates of return which equate the present value of the stream of benefits over the chosen timeperiod with initial costs.

41

Table 20: Marginal Effect on Family Labor Costs Allowing for Interadion Effects

Unrestricted model Restricted Model

Marginal effect on Marginal effect onVariable family labor cost t-ratio family labor cost t-ratio

irrigated annual land Dongs/100m2 19,249.7 5.3 20,032.5 5.7

non-irrigated annual land Dongs/100m2 11,970.8 4.0 10,864.3 6.3

perennial land Dongs/100m2 23,231.6 3.7 21,530.7 5.3

forest land Dongs/100m2 -3,777.0 0.7 -2,795.3 1.4

water surface land Dongs/100m2 -764,911.9 1.1 -755,668.6 1.1

other land Dongs/100m2 12,901.6 1.2 2,612.2 1.9

household size Dongs/person 342,305 9.5 360,211.8 11.2

prop female adults Dongs/% point 17,197.3 2.7 1596511 5.3

prop male adults Dongs/% point 25,086.1 3.9 2415231 7.9

prop aged 7-16 Dongs/% point 9,143.5 1.5 648723.4 2.8

primary ed (head) Dongs/year -31,351.0 0.4 23,235.7 2.3

ed > primary (head) Dongs/year -49,473.3 2.2 -43,136.9 3.6

primary ed (other adults) Dongs/year 78,568.7 4.7 61,565.8 7.9

ed > primary (other Dongs/year -41,642.1 2.8 -38,451.3 3.1adults)

mean family labor costs 3,034,006 3,034,006

Note: Marginal effects are evaluated at mean points.

42

4. Conclusions

Viet Nam has poor infrastructure and high poverty. These two facts are intimatelyconnected. However, the nature of those connections and their implications for the role ofinfrastructure investments in fighting poverty are complex to disentangle. This paper hasfocused on some aspects of the link between poverty and lack of infrastructure using theVNLSS.

Access to infrastructure services tends to be poor for the majority of Vietnamese. Urbanareas are better provisioned and some regions certainly fare worse than others. In particular,there are some distinct differences between the North and South of the country. Imbalances arealso evidenced among infrastructure services. For example, the provision of social servicefacilities is generally superior than that of other physical infrastructure. Piped water provisionand electricity reveal considerable disparities between poor and non-poor groups. But, by andlarge, the data indicate that basic infrastructure services are generally inadequate for all groups,though generally worst for the poor. As a result, it cannot be surmised that an expansion ininvestment in basic infrastructure will be well-targeted to the poor. Indeed, there is ample scopefor the non-poor to capture the lion's share of the direct gains from infrastructure investment inViet Nam.

To assess the impacts on poverty it is necessary to examine the distribution of themarginal benefits of specific infrastructure investments. The paper focuses on irrigationinvestments to explore this issue in more depth. The cross-sectional variation is used to estimatethe marginal impacts of converting non-irrigated annual crop land over to irrigation. Inparticular, a policy of irrigating 10 percent of currently non-irrigated annual land is simulatedbased on a regression model for crop income which includes irrigated and non-irrigated land asexplanatory variables. The simulations allow for four different ways of distributing the irrigationexpansion across households: in simulation (1): irrigation is distributed to all households subjectto feasibility; in (2) it goes only to households currently without irrigated land; in (3) it istargeted to households with low total annual landholdings and in simulation (4) it is targeted tohouseholds with low per capita annual landholdings. In general, at the national level theabsolute income gains across expenditure groups imply that an undifferentiated expansion ofirrigation would be redistributive-having higher proportionate gains to poorer households.Targeting the irrigation expansion to households with small per capita landholdings produces themost progressive incidence of gains as well as the largest absolute benefits to the poor. Theresults under all simulations show the highest total impacts on net crop incomes would occur forViet Nam's two poorest regions-the Northern Uplands and the North Coast, where the impactsalso show the most pro-poor distribution.

These substantial potential gains from irrigation from an equity point of view are likelyto be accompanied by sizable average rates of return. Even under quite conservativeassumptions-namely a project life of only 10 years and valuing family labor inputs at themarket wage for similar work-the average annual rate of return implied by my estimates of thegains to farm profits, and recent estimates of the investment cost of irrigation, is about 20%.

43

An even larger impact may be possible with a more differentiated expansion ofirrigation-emphasizing key regions such as the Northern Uplands and addressing the need forrehabilitation of existing irrigation infrastructure, to realize its full potential. Conversely, therate of return will undoubtedly be lower in some areas where irrigation expansion is particularlycostly.

Lack of irrigation infrastructure is clearly not the only constraint to reducing ruralpoverty in Viet Nam. The quantity (in particular household size) and quality (education) of thefamily's human resources also matter greatly. And not only do other important constraints exist,but these are inextricably bound to the benefits which can ultimately be derived from irrigationinfrastructure. The analysis uncovers important complementarities between education,particularly primary education, and the gains from irrigation. Demographics are also found tobe key. Finally, one can conjecture that the current lack of other infrastructure such as roads,electricity, communications and so forth, must also conspire to reduce the impacts which canbe garnered from irrigation alone.

44

References

Binswanger, Hans P., Shahidur R. Khandker and Mark R. Rosenzweig. 1993. "HowInfrastructure and Financial Institutions Affect Agricultural Output and Investment inIndia. " Journal of Development Economics, 41(2): 337-366.

Dollar, David and Paul Glewwe. 1995. "Poverty and Inequality in Viet Nam: The CurrentSituation." Mimeo, EAlCO and PRDPH, World Bank.

Goldstein, Ellen. 1993. "The Impact of Rural Infrastructure on Rural Poverty." Mimeo, SouthAsia Region, World Bank.

Howe, J. and P. Richards. 1984. Rural Roads and Poverty Alleviation. ILO, IntermediateTechnology Publications Ltd., London, U.K.

Jimenez, Emmanuel. 1995. "Human and Physical Infrastructure: Public Investment and PricingPolicies in Developing Countries." In Jere Behrman and T.N. Srinivasan, eds.,Handbook of Development Economics, Volume 3, Amsterdam: North-Holland.

Lipton, Michael and Martin Ravallion. 1995. "Poverty and Policy." In Jere Behrman and T.N.Srinivasan, eds., Handbook of Development Economics, Volume 3, Amsterdam: North-Holland.

NEDECO. 1991. "Mekong Delta Master Plan: Inception Report." May 13, The Netherlands.

. 1993. "Draft Master Plan for the Mekong Delta in Viet Nam: A perspective forSustainable Development of Land and Water Resources." June, The Netherlands.

Ravallion, Martin. 1994. Poverty Comparisons. Chur, Switzerland: Harwood Academic Press,Fundamentals in Pure and Applied Economics, Volume 56.

Salinger, Lynn B. 1993. "Viet Nam's Agricultural Comparative Advantage and ExportPotential." Associates for International Resources and Development, Cambridge, Mass.

State Planning Committee, UNDP, FAO and World Bank. 1989. "Viet Nam Agricultural andFood Production Sector Review." Mission report.

UNICEF. 1994. Situation Analysis of Women and Children in Viet Nam. Hanoi.

van de Walle, Dominique. 1995a. "Targeting and Incidence: An Overview of Implications forResearch and Policy." In D. van de Walle and K. Nead Public Spending and the Poor:Theory and Evidence. London and Baltimore: The Johns Hopkins University Press.

45

. 1995b. "Rural Poverty in an Emerging Market Economy: Is Diversification intoNon Farm Activities in Rural Viet Nam the Solution?" Mimeo, PRDPE, World Bank.

Vu, Tu Lap and Christian Taillard. 1993. Atlas du let Nam. Montpellier et Paris: Reclus, LaDocumentation Francaise.

World Bank. 1990. 'Viet Nam: Water Supply and Sanitation Sector Study." April.

_ 1994a. World Development Report 1994. Infrastructure for Development. NewYork: Oxford University Press.

. 1994b. "Viet Nam Transport Sector: Serving an Economy in Transition." ReportNo. 12778-VN, August.

. 1994c. "Viet Nam: Poverty Assessment and Strategy." Report No. 13442VN,September.

46

LSMS Working Papers (continued)

Decomposition with Applications to Brazil and India in the 1980s

No. 84 Vijverberg, Measuring Incomefrom Family Enterprises with Household Surveys

No. 85 Deaton and Grimard, Demand Analysis and Tax Reform in Pakistan

No. 86 Glewwe and Hall, Poverty and Inequality during Unorthodox Adjustment: The Case of Peru,1985-90

No. 87 Newman and Gertler, Family Productivity, Labor Supply, and Welfare in a Low-lncome Country

No. 88 Ravallion, Poverty Comparisons: A Guide to Concepts and Methods

No. 89 Thomas, Lavy, and Strauss, Public Policy and Anthropometric Outcomes in Cdte d'lvoire

No. 90 Ainsworth and others, Measuring the Impact of Fatal Adult Illness in Sub-Saharan Africa: AnAnniotated Household Questionnaire

No. 91 Glewwe and Jacoby, Estimating the Determinants of Cognitive Achievement in Low-IncomeCountries: The Case of Ghania

No. 92 Ainsworth, Economic Aspects of Child Fostering in C6te d'lvoire

No. 93 Lavy, Investment in Hurman Capital: Schooling Supply Constraints in Rural Glhana

No. 94 Lavy and Quigley, Willingness to Payfor the Quality and Intensity of Medical Care:Low-Income Households in Ghana

No. 95 Schultz and Tansel, Measurement of Retlurs to Adult Health: Morbidity Effects on Wage Ratesin C6te d'lvoire and Ghana

No. 96 Louat, Grosh, and van der Gaag, Welfare Implications of Female Headship in Jamaican Households

No. 97 Coulombe and Demery, Household Size in Chte d'lvoire: Sampling Bias in the CILSS

No. 98 Glewwe and Jacoby, Delayed Primary School Enrollment and ChildhIood Malnultrition inGhana: An Economic Analysis

No. 99 Baker and Grosh, Poverty Reduction throuigIh Geographic Targeting: How Well Does It Work?

No. 100 Datt and Ravallion, Income Gainsfor the Poorfrom Public Works Employment: Evidence fromTwo Indian Villages

No. 101 Kostermans, Assessing the Quality of Anthropometric Data: Background and IllulstratedGuidelines for Survey Managers

No. 102 van de Walle, Ravallion, and Gautam, How Well Does the Social Safety Net Work? TheIncidence of Cash Benefits in Huingary, 1987-89

No. 103 Benefo and Schultz, Determinants of Fertility and Child Mortality in C6te d'lvoire and Ghana

No. 104 Behrman and Lavy, Children's Health and Achievement in Schlool

No. 105 Lavy and Germain, Quality and Cost in Health Care Choice in Developing Countries

No. 106 Lavy, Strauss, Thomas, and De Vreyer, The Impact of the Quiality of Health Care 1on Children's Nutritionand Survival in Ghana

No. 107 Hanushek and Lavy, School Quality, Aclhievement Bias, and Dropout Behavior in Egypt

No. 108 Feyistan and Ainsworth, Contraceptive llse and the Quality, Price, and Availability of Family Planning

No. 109 Thomas and Maluccio, Contraceptive Choice, Fertility, and Public Policy in Zimbabwe

No. 110 Ainsworth, Beegle, and Nyamete, The Impact of Female SChOOling On Fertility and Contraceptive Use:A Study of Fourteen Siib-Saharan Countries

No. 111 Oliver, Contraceptive Use in Ghana: The Role of Service Availability, Quality, and Price

No. 112 Montgomery, Kouame, and Oliver, The Tradeoff between Numtber of Children and Chlild Schooling:Evidence from Cdte d'Ivoire and Ghana

No. 113 Pradhan, Sector Participation Decisions in Labor Suipply Models

No. 114 Beegle, The Quality and Availability of Family Planntinig Services and Contraceptive Use in Tanzania

No. 115 Lavy, Spratt, and Leboucher, Changing Patterns of Illiteracy in Morocco: Assessment Methods Compared

No. 116 Lavy, Palumbo, and Stern, Health Care in Jamaica: Quality, Ouitcomes, and Labor Supply

No. 117 Glewwe and Hall, Who Is Most Vulnerable to Macroeconomic Shocks? Hypotheses Tests Using PanelData from Peru

No. 118 Grosh and Baker, Proxy Means Tests-for Targeting Social Programs: Simiulations and Speculation

No. 119 Pitt, Women's Schooling, the Selectivity of Fertility, and Child Mortality in Sub-Saharan Africa

No. 120 Grosh and Glewwe, A Guide to Living Standards Measurement Study Surveys and thzeir Data Sets

THE WORLD BANK

Head uarters Euro ean Office Tok oOffice

9 780821 335444

ISBN 0-8213-3544-8

Infrastructure and Poverty in Viet Nam · 2016. 7. 17. · Viet Nam is poor both in terms of...

Documents

Transcript of Infrastructure and Poverty in Viet Nam · 2016. 7. 17. · Viet Nam is poor both in terms of...