13351 WHO BENEFII S FROM GOVERNMENT EXPENDITURE? A Case Study of Colombia , FILE COPY , W B R l~~~~~ Wol BakRsac Governments can improve the distribution of income and also can eradicate extreme poverty by changing the composition and direction of their public expenditure. But more important than the capacity of the fiscal budget to transfer income is its use to increase the con- sumption of such basic needs as housing, water, sewerage, and health and education services. Such use would permit the poorest groups in the population to enjoy higher levels of consumption at an earlier stage of develop- ment than would be the case if the private forces of supply and demand alone were acting during the normal course of per capita income growth. What is the present performance of developing countries in reaching the poorest income groups through public expenditure? How has this performance evolved over time? Are there constraints on the consumption of basic services other than the mere availability or supply of these services? This research study addresses these questions with the help of a case study carried out in Colombia. Another case study is described in the companion volume, Public Expenditure in Malaysia: Who Benefits and Why, by Jacob Meerman, also published by Oxford. Marcelo Selowsky is the economic adviser in the Development Economics Department of the World Bank. Who Benefits from Government Expenditure? A Case Study of Colombia A World Bank Research Publication Who Benefits from Government Expenditure? A Case Study of Colombia Marcelo Selowsky Published for the World Bank Oxford University Press Oxford University Press NEW YORK OXFORD LONDON GLASGOW TORONTO MELBOURNE WELLINGTON HONG KONG TOKYO KUALA LUMPUR SINGAPORE JAKARTA DELHI BOMBAY CALCUTTA MADRAS KARACHI NAIROBI DAR ES SALAAM CAPE TOWN © 1979 by the International Bank for Reconstruction and Development / The World Bank 1818 H Street, N.W., Washington, D.C. 20433 U.S.A. All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press. Manufactured in the United States of America. The views and interpretations in this book are the author's and should not be attributed to the World Bank, to its affiliated organizations, or to any individual acting in their behalf. Library of Congress Cataloging in Publication Data Selowsky, Marcelo. Who benefits from government expenditure? Bibliography: p. 180 Includes index. 1. Colombia-Appropriations and expenditures. 2. Income distribution-Colombia. I. International Bank for Reconstruction and Development. II. Title. HJ2083.S44 336.861 79-16384 ISBN 0-19-520098-5 ISBN 0-19-520099-3 pbk. Contents FOREWORD xiii ACKNOWLEDGMENTS xv INTRODUCTION AND SUMMARY 3 iMethodology 4 Results 5 Conclusions 7 CHAPTER 1. BACKGROUND AND MIAIN RESULTS 8 Objectives and Limitations 8 Methodology and Further Simplifications 10 Main Results 17 Appendix. Socioeconomic Stratification of Large Cities 36 CHAPTER 2. THE DISTRIBUTION OF INCOME AND OTHER POVERTY INDICATORS 38 Distribution of Income 38 Other Poverty Indicators 43 Appendix. Socioeconomic Characteristics of Households 47 CHAPTER 3. THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION 50 Enrollment by Income Group 50 Public Subsidy per Student Year 56 Distribution of Subsidies across Income Groups 63 Appendix. Comparisons with Other Studies 70 CHAPTER 4. THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH 77 Total Subsidy Received by Households 78 Consumption of Services and Distribution of Subsidies across Income Groups 86 Explanatory Variables behind the Demand for Medical Services 97 Appendix. A Framework for Analyzing the Incidence of the Financing of the Social Security System 100 v vi CONTENTS CHAPTER 5. THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES: ELECTRICITY, PIPED \\ ATER, AND SEWERAGE 105 Investment and the Availability of Services 106 Public Services and Substitutes in Urban Areas 118 Public Services and Substitutes in Rural Areas 134 Appendix. Transfers between Urban Consumers Resulting from the Tariff Structure 138 CHAPTER 6. THE DISTRIBUTION OF BENEFICIARIES OF OTHER SERVICES 143 Agricultural Loans by the Caja Agraria, 1974 143 SENA Training Courses, 1974 146 Garbage Collection Services in Urban Areas, 1974 148 STATISTICAL APPENDIX 151 REFERENCES 180 INDEX 183 Figures 1.1. Distribution of Income: Comparison of Estimates 19 1.2. Distributioni of Income and Subsidies for Ilealth and Education, 1974 24 2.1. Distributtion of Income, 1974: Rural and Urban Differences 39 3.1. Distributtion of Income and of Subsidies for Education 69 4.1. Colombian hIealth System 79 5.1. Probability of a lIousehold IIaving Electricity, as a Function of Income 113 5.2. Probability of a Household flaving Sewerage, as a Function of Income 114 5.3. Probability of a Ilousehold Ilaving Piped lWnater, as a Function of Income 115 5.4. Probability of a IIousehold Ilaving Garbage Collection, as a Function of Income 116 5.5. Probability of a lIoutsehold lIaving Street Lighting, as a Finction of Income 117 Tables 1.1. Size and Major Categories of the 1974 Sample Survey 15 1.2. Number of lIouseholds in the Sample, Classified by Quintiles in the (Expanded) Regional Distribution of Income 17 1.3. Distribution of Income, 1974 18 1.4. Per Capita Income Conversions, November 1974 20 CONTENTS vii 1.5. Percent of Households below Poverty Lines 21 1.6. Distribution of Subsidies for Education and Ilealth per Household and per Capita, Classified by Income Quintile 22 1.7. Distribution of Income and Subsidies for Edutcation and Health, Classified by Income Quintiles 24 1.8. Houtseholds with Services in 1974 and IIouseholds VVhich Were Connlected to the Network bet-ween 1970 and 1974, Classified by Income Groups 26 1.9. Electricity Utse in Urban Areas: Estimates of the Linear Probability Function 28 1.10. Piped Water Use in Urban Areas: Estimates of the Linear Probability Futnction 30 1.11. Sewerage Use in Urban Areas: Estimates of the Linear Probability Function 32 1.12. Distributioni of Subsidies and Beneficiaries of Government Services, 1974: Results from the 1974 Sample Survey 34 2.1. Percentage of Families in Each Income Quintile, Classified by Location 40 2.2. Income Distributtion Comparison 42 2.3. Conmparisont of Gini Coefficients from Four Studies 42 2.4. Distribuition of Ileads of hlousehold in Each Quintile, by Years of Schooling 44 2.5. Mean and Standard Deviations of Socioeconiomic Variables of Ilouseholds 45 2.6. Distribution of hleads of IIousehold in the Poorest Income Quintile According to Location, Sector, and Occupation 46 2.7. Sample Size in Each (Expanded) Quiintile 48 2.8. Critical V'alues of duf,n (5 Percent Significanice Level) for the Test of Means uinder Alternative Combinations of Sample Sizes 48 3.1. School Enrollment, Classified bv Income Group, 1974 51 3.2. School Enrollment: Comparison between Sample Survey Figures and Mlinistry of Education Figures, 1974 52 3.3. Adjusted School Enrollment, Classified by Location of Households, 1974 53 3.4. -lean Entrollment per Household in Different Types of Primary Schools, Classified by Quintiles 54 3.5. Alean Enrollment per hlousehold in Different Types of Secondary Schools, Classified by Quin tiles 54 viii CONTENTS 3.6. Number of Children per Ilousehold, Classified by Location, Age, and Income Quintile 56 3.7. Mfean Enrollment per IIousehold in Universities, Classified by Quintiles 56 3.8. Public Contributionts from All Sources to Public Primary Schools, 1973 59 3.9. Public Primary Education: Estimated Enrollment, Total Subsidy, and Subsidy per Student, Classified by Stratum, 1973 and 1974 60 3.10. Public Secondary Education: Estimated Subsidy per Student, Classified by Types of Schools, 1974 61 3.11. Private Primary and Secondary Schools: Enrollment, Total Subsidy, and Subsidy per Student, 1974 62 3.12. Universities: Enrollnientt, Total Subsidy, and Suibsidy per Student, 1974 63 3.13. Enrollment, Total Subsidy, and Estimated Subsidy per Student, 1974 64 3.14. Education Subsidy per IIousehold, 1974 66 3.15. Education Subsidy Figures per hIousehold and per Capita 66 3.16. Total Education Subsidy per hIousehold as a Percentage of Reported Ilousehold Annual Income 67 3.17. Primary and Secondary Education Subsidy per IIousehold as a Percentage of hIousehold A nnual Income, by Urban and Rural Regions, 1974 68 3.18. DistributionZ of the Subsidy to Education and Distribution of Personal Income, 1974 70 3.19. Distribution of Education Subsidies, 1966, Urrutia-Sandoval 72 3.20. Distribution of Education Subsidies, 1970. Jallade 73 3.21. Comparison of Enrollment per HIIousehold in Primary Education, Jallade and Sample Survey 74 3.22. ComparisonZ of Enrollnient per IIousehold in Secondary Education, Jallade and Sample Survey 75 4.1. Funding Received by Ilealth Institutions, 1974 80 4.2. Estimated Total Subsidy Received by Ihouseholds, 1974 81 4.3. National Ihealth System hlospitals: Number of Services Provided and the Distribution of the Subsidy, 1974 82 4.4. icss hIospitals: Number of Services Provided and the Distribution of the Subsidy, 1974 83 4.5. Cajas Publicas IIospitals: Number of Services Provided for 1969 and Distribution of the Subsidy in 1974 84 CONTENTS ix 4.6. Public Subsidies to Health Institutions, Classified by Location of the Institutionz, 1974 85 4.7. Number of Services Provided in hIospitals (Reported by Ilealth Institutions) and Consumed by hIouseholds (Reported in the 1974 Sample Survey) 87 4.8. Allocation of Services Provided in Urban llospitals to Rural Ilouseholds 88 4.9. hIealth Subsidies Received by hIouseholds, Classified by Location of IIousehold, 1974 89 4.10. Services Provided by National Health Service Ihospitals and Ilealth Centers, Classified by Location of llousehold 90 4.11. Services Provided by Social Security System Hospitals (icss and Cajas), Classified by Location of hlousehold 91 4.12. Households Afflliated with the Social Security System, Classified by Income Quintile 92 4.13. Health Subsidy per HIousehold, 1974 94 4.14. Health Subsidy per Household, Classified by Affiliation (AF) and Nonaffiliation (NAF) with the Social Security System, 1974 94 4.15. Health Subsidy per Household: Summary 96 4.16. Average Annual Visits to Physicians, per hlousehold and per Person, 1974 98 4.17. Regression Coefficient Results from Regressing the Number of Visits to Doctors as a Function of Socioeconomic Variables of Households in Urban Areas 100 4.18. Values of A/R (tf = 0.035; t, = 0.070) 104 5.1. Distribution of Households That Obtained Public Utility Services between 1970 and 1974 107 5.2. Investment in Electricity, 1970 to 1974 108 5.3. Investment in Piped Water and Sewerage, 1970 to 1974 108 5.4. Number of New Ilouseholds Connected to the Service and Public Investment per Household between 1971 and 1974 109 5.5. Families with Services, 1974 110 5.6. Estimated Regression Coefficients in the Linear Probability Function 112 5.7. Distribution of Urban Households That Became Connected between 1970 and 1974 128 5.8. Urban Households That Became Connected between 1970 and 1974 as a Percentage of the IIouseholds uithout the Service X CONTENTS in 1970, Classified by City Size and Urban Income Distribution 128 5.9. The Probabilities of lIaving Had a Service in 1970 (P70) and of hlaving Obtained a Service between 1970 and 1974 (-r): A Comparison of Regression Coefficients 129 5.10. Distribution of Urban Households without Services in 1974, Classified by Distance from Neighbor with Service 133 5.11. Distribution of Reasons Why Urban IIouseholds Are without the Service when There Is a Neighbor with the Service Less than One Block Away 133 5.12. Distribution of Urban Households without the Service, Classified by the Substitute Lsed 134 5.13. Sources of Services Used by Rural lIouseholds, Classified by Income Quintile in the Distribution of Rural Income 135 5.14. Regression Coefficient for the Probability of Having Electricity in Rural Areas 136 5.15. Rural Ihouseholds That Became Connected to the Electricity between 1970 and 1974 137 5.16. Rural lHouseholds without Electricity, Classified by the Distance to the Nearest Neighbor with the Service 138 5.17. Subsidies Out of the Tariff Policy as a Percentage of Household's Income, 1970 141 6.1. Estimates of the Implicit Subsidy Received by Farmers from Loans by the Caja Agraria, 1974 145 6.2. Income Distribution of Farmers, 1974 146 6.3. Attendance at SENA Courses, 1974 147 6.4. UIrban hlouseholds with Garbage Collection Service, 1974 149 SA-1. Distribuxtion of Families, by per Capita Income, According to the Official Exchange Rate, 1974 152 SA-2. Distribution of Families, by per Capita Income, According to the Kravis Parity Exchange Rate, 1974 152 SA-3. MIean N\tumber of Rooms Occupied by the hIousehold 153 SA-4. l ean NNumber of Toilets and Latrines in the Dwelling 153 SA-5. Mean Number of Persons in the hIousehold 154 SA-6. Mean Number of Income Earners in the hlousehold 154 SA-7. Mean Nufmber of Children Age 0 to 5 in the Household 155 SA-8. AMean Number of Children Age 6 to 11 in the hlousehold 155 SA-9. AMean Nzumber of Children Age 0 to 11 in the hIousehold 156 SA-10. Mfean Number of Persons Age 12 to 16 in the Household 156 CONTENTS Xi SA-ll. M1Iean Number of Pregnancies in the Household 157 SA-12. Mfean Age of 11ead of Household 157 SA-13. MIean Total Years of Schooling of the Head of Household 158 SA-14. MlIeant Age of Wife in the Household 158 SA-15. M1lean Total Years of Schooling of Wife in the hlousehold 159 SA-16. Mean Number of Children Age 6 to 11 Registered in School in the Household 159 SA-17. Mlean Number of Persons Age 12 to 16 Registered in School in the hlousehold 160 SA-18. Number of Teachers in Public Primary Schools, by Category and Stratum, 1973 160 SA-19. Monthly Wage of Teachers in Public Primary Schools, by Category and Departments, 1973 162 SA-20. Estimate of the Costs of Public Health Centers, 1974 163 SA-21. Estimate of Subsidy to Each Puesto de Salud, 1974 164 SA-22. Total Su,bsidy to P,uestos de Salud, 1974 164 SA-23. Number of Services Received in Hlospitals of the National IHealth System as Reported by Households, 1974 165 SA-24. Number of Services Received in hlospitals of the Social Security System (icss and Cajas) as Reported by Households, 1974 166 SA-25. Distribution of Pu blic Subsidies to hIospitals of the National Health System (NHS), the Social Security System (sss), and Ihealth Centers of All Types (Hc), 1974 167 SA-26. Alternate Distribution of Public Subsidies to Hospitals, 1974 168 SA-27. Percentage of Families with Electricity, by Regional Quintiles 169 SA-28. Percentage of Families with Electricity, by per Capita Household Incomze 170 SA-29. Percentage of Families with Piped Water, by Regional Quintiles 171 SA-30. Percentage of Families with Piped Water, by per Capita Household Income 172 SA-31. Percentage of Families with Sewerage, by Regional Quintiles 173 SA-32. Percentage of Families with Sewerage, by per Capita Ilousehold Income 174 SA-33. Probability of laving Electricity, by per Capita houisehold Income 175 SA-34. Probability of Having Piped Water, by per Capita Household Income 176 Xii CONTENTS SA-35. Probability of Having Sewerage, by per Capita Ilousehold Income 177 SA-36. Probability of hlaving Street Lighting, by per Capita IIousehold Income 178 SA-37. Probability of Having Garbage Collection, by per Capita Ilousehold Income 179 Foreword THIS STUDY OF COLOMBIA by MIarcelo Selowsky and the companion study of Malaysia by Jacob iMeerman were undertaken in 1974 as complemiients to other work of the World Bank on income distribu- tion. They were undertaken in recognition of the fact that the welfare of the poor is affected not only by their income, narrowly defined, but also by the services that they receive from their gov- ernmlents. By now the notion of basic needs is commonplace; attributes of systemns for meeting these needs-such as accessi- bility, reasonable prices, and cost effectiveness-are well known. Consequently, the idea of using specially designed household surveys to investigate the distribution of public expenditure and to explain it to some extent may seem less novel than it did five years ago. It is not unusual for the Bank to undertake research of an ex- periinental nature in the hope that others will be able to make use of both the results and the experience gained in attaining the results. In tlis instance both authors advocate similar investiga- tions in other countries, not only as a one-time effort in each coun- try, but also at intervals of from five to ten years to monitor progress. In essence, this is a plea that governments build on the initial efforts of these authors and institutionalize them. One thing whiclh emerges clearly from the studies, however, is the great difficulty-sometimes the impossibility-of extracting significant complementary informnation from government accounts, even good ones. On reflection this is not too surprising, since gov- ernment fiscal systems are not designed originally as information xiii xivs FOREWORD systems. Governments that command large sliares of the national products of their countries and wishl to use thlerm efficientlv for clearly specified objectives may well need to exploit more fullv their untapped sources of information. The value of these studies today rests on their intrinsic merit, but their potential value mlay be muclh greater by virtue of tlheir demonstration effect. BENJAMIN B. KING Director, Development Economics Department The World Bank June 1979 Waslhington, D.C. Acknowledgments I AMI INDEBTED to several individuals and institutions for making this researchl possible. Jaime H. Caro, Carlos Lemoine, and Fran- cisco Pereira contributed to the sample survey design and were central in planning the earlier stages of the study. COLDATOS (Corn1- pafiia Colomnbiana de Datos) carried out the field work for the survey and the initial work on data processing. DANE (Departa- mento de Estadistica) and INPES (I nstituto para Programas Especiales de la Salud) provided data on the education and healthl sector that were crucial in analyzing thie distribution of public expenditure in these sectors. Earlier discussions witlh researchers of CEDE (UTniversidad de Los Anides), as well as with Ricardo Galan, Jacol) Meermnan, Manuel Ramtrez, and Miguel UTrrutia were extremely useful in planning the research. Finally, I am indebted to Mariene I.ehwing and Zaitun \irji for valuable re- search assistance and to Aludia Oropesa for typing the final manu- script. Nancy WV. Donovan edited an early draft, and Virginia deHaven Orr edited thle final manuscript. Brian J. Sviklhart directed design and production, Florence Robinson indexed the text, and Carol Crosby Black designed the cover. MARCEI,() SELOWSKY Xi! Who Benefits from Government Expenditure? A Case Study of Colombia Introduction and Summary ONE POLICY OPTION OPEN TO GOVERNMENTS to improve the dis- tribution of income and to eradicate extreme poverty is to change the composition and direction of public expenditure. This option now has an important potential in most developing countries for two reasons. First, government expenditure has become a significant fraction of national income. The present information available on the personal distribution of income indicates that this expenditure is also quite large compared with the share of national income re- ceived by the poorest groups of the population. Hence, changes in the direction of this expenditure can have a significant effect on the real income of the lowest income groups: in a country where government expenditure and the income of the poorest 20 percent of the population account for 25 and 5 percent, respectively, of the gross national product (GNPI), reallocating 10 percent of the fiscal budget to this group would increase its income by 50 per- cent. Such a reallocation is one of the most feasible options from a political point of view. It does not represent an important trade- off with the growth rate; even if half of it were to be made at the expense of public investmenit and assuming that no assets would be created in the poorest group, the growthi rate of GNP would decline only 0.25 percentage points.' Second, and more important thani the capacity to transfer in- come, is the use of the fiscal budget to increase the consumption of specific goods and services. Not only per capita income is dis- 1. Assuming a social return to capital of 0.20, the growth rate of the GNP would diminish by (0.0125)(0.20) = 0.0025. 3 4 INTRODUCTION AND SUMMARY tributed unequally; this is also true for other welfare indicators- such as life expectancy, calorie consumption, and minimum literacy-that by now have become goals on their own. This situa- tion reflects an inequality in the consumption of food and services, such as housing, water, sewerage, education, and health. Policies to increase their consumption above the levels resulting from the regular forces of supply and demand define what has become known as a "basic needs" approach to poverty alleviation. The notion is to use the fiscal budget to direct the allocation of re- sources to reach these critical levels of consumption at earlier stages of development: that is, earlier than would have been reached in the normal growth of per capita income. The argument becomes more important in three circumstances: (a) the smaller the trickle down effect of GNP growth on the income of the poor; (b) the smaller the marginal propensity of the poor to spend on "basic needs;" and (c) the smaller the possibility of the pri- vate sector supplying these services because of complementarities and economies of scale in these sectors. What is the present performance of developing countries in reaching the poorest income groups through public expenditure? How has this performance evolved over time? Are there constraints on the consumption of these services other than the mere avail- ability or supply of these services? This research study addresses these questions with the help of a case study; the country selected is Colombia. Methodology A full evaluation of the distributive effect of government expen- diture would require a model to predict the effect of this expendi- ture on each income group, as owners of factors of production and as consumers of final goods and services. Such a model would involve specifying most factor and commodity markets and would have to be higlhly complex, such as a general equilibrium model. The objectives of this study, however, are less ambitious. It con- centrates on the publicly provided services whose consumption by individuals or houselholds can be identified and measured. This approach omits the typical "public good," such as defense and justice. It measures the consumption of public services as a final RESULTS 5 commodity, rather than tracing the distributive effect when the services are consumed as a factor of production. Specifically this study identifies the beneficiaries of publicly provided services, measures the subsidy received by households from consuming some of these services, and attempts to explain the present distribution of consumption in terms of supply and demand: that is, to what extent is the absence of consumption of particular services the result of the unavailability of supply and to what extent is it the result of demand factors governing the utiliza- tion of such supply? For this purpose a specifically designed country-wide survey of 4,019 households in Colombia was carried out in November 1974. The survey data are classified by rural and urban location, and the urban data are classified further according to city size. The survey provides household income data that are used to classify the bene- ficiaries of government services in the overall distribution of income. Results The survey was able to trace the beneficiaries of one-third of total government expenditure. The major expenditures accounting for this fraction are the public subsidies to the education and health sector and the investment in electricity, water, and sewerage. The total subsidy to education is distributed evenly across income quintiles: that is, the subsidy per household is constant across income groups. It results, however, from different subsidies to each educational level. The subsidy to primary education is progressive, whereas the subsidy to higher education is highly regressive; actually it is more unequally distributed than personal income. The health subsidy is also relatively similar across house- holds, although it varies according to the source: The National Health System has a progressive effect, whereas the Social Security System network favors the middle-income quintiles. The relative constancy of the subsidy per household across income groups does not hlold when it is expressed in per capita terms. The reason for this is the difference in family size in each per capita income group. The poorest quintile of households accounts for 25.1 percent of the population, whereas the richest accounts for 15.4 percent, a differ- 6 INTRODUCTION AND SUMMARY ence of 60 percent in family size. The per capita sulsidv to the riclhest quintile beconiies 1.6 timies larger tllan that to the poorest quintile. For electricity, water, and sewerage, data were obtained on the distribution of hlouselholds wvlo had the service in 1974 and who received the service between 1970 and 1974. A conmparison of thle figures indicates tile distributive direction of investmlent in those sectors over tilie. The distribution of beneficiaries by quintiles in 1974 is quite similar across services. Between 25 and 30 percent of the hlouse- lholds withi services belong to the bottomi 40 percent of lhouselholds, wlhereas 50 to 55 percent belong to the upper 40 percent. Almost all consumiiers are concentrated in urban areas. Thie fact that low- incomle quintiles consumle less of these services is mlore a result of those services being concentrated in urban areas-rural hlouselholds leing relatively poorer in the country distribution of incomiie- tlhan an intra-urbani discrimination against low-incomiie groups. Thle distribution of new beneficiaries is imuclh more progressive. New iinvestmiienit in these sectors hlas beeni more redistributive than in the past. Part of this clhange results froml the fact thiat invest- mlenit had a lower "urban blias:" that is, a large fractionl of the new beneficiaries hlave been rural hlouselholds, particularly for elec- tricitv and piped water. Witlhin the urban sector iiivestiniei]t has also tended increasingly to blenefit poorer 1housellolds. Wlhat factors explain the distribution of consumiiptioni of elec- tricity, water, and sewerage services? A mLultivariate analysis was carried out to identify variables associated witlh the consumiiptioni of these services by urban lhouselholds. A framlework was developed wlhere tlis association could be interpreted as a relationi of causal- ity. A situation in wlhiclh the lhousehold does not use the service because of inaccessability (thle supply network is too far) is distin- guislhed froml a situation whlere demland factors constrain the use of that supply. ThIle Colombia data suggest that hlalf of tlle urball househlolds witlhout the service fall in the latter category. Tlherefore the study made a special effort to identify the extent to wliclh variables suclh as per capita incomiie. education, and illigrant status affect the probability of a lhouselhold demlanidinig the service. The survey also identified the benieficiaries of othler services and subsidies: street lighlting and garbage collection services, educa- tional fellowshlips, adult retraininig courses at SENA (tihe Colombian Retraining Center), and the subsidy em)bodied in the loans from CONCLUSIONS 7 the Caja Agraria, the main public agency channeling subsidized credit to farmers. Although they are less important as a fraction of total government expenditure, the effect of these services and subsidies on particular groups can be important. The survey was unable to identify the beneficiaries of invest- ment in roads, a large item of public investment in Colombia. Questions regarding changes in the mode and time of travel did not yield significant results. Either investment in roads has basi- cally a benefit as an intermediate input-instead of as a consumer good-or the questionnaire metlhod used was not the proper tech- nique to measure the benefit of this type of service. Conclusions This research, initially conceived as a pilot study, slhows that specifically designed houselhold surveys can provide relatively good information on the distributive direction of a large portion of government expenditure. Ideally suchI a survey should be repeated -for example, every five years-to capture the distributive direc- tion of investment and changing government programs. Accord- ingly, the design of the survey questionnaire should be adjusted continuously to identify the beneficiaries of new kinds of programs. The advantage of a countrywide survey of this kind is the informa- tion it provides on households not consuming particular services. From a policy point of view, it is fundamental to determine wlhether this situation results from a lack of supply of the service or from demlalnd pheniomiiena governing the use of that supply. Chapter I Background and Main Results To IDENTIFY THE SPECIFIC POLICY OPTIONS that canl be used to change the composition and direction of public expenditure, thie distributive direction of the present public expenditure first must be evaluated. This study makes such an evaluation with the help of a case study. Colombia was thie country choseni for the study; it represents the typical middle-inicomiie ($600 per capita income), semi-industrialized Latin American country.' It hias a growing urban population, strong income differences between regions, and both commercial and small subsistence farming. The data base is reasonably good, and earlier work on the subject allows coiil- parison of results.2 Objectives and Limitations A full evaluation of the distributive effect of government ex- penditure-the change in the real income of individuals resulting from the total or marginal presence of the government-is beyond the scope of this research, since it would require a model of the determinants of the incidence of government intervention througlh factor and final goods markets. When this intervention provides public services wlhose benefits may not be completely internalized 1. Throughout the text, "dollar" ($) refers to U.S. dollar. 2. The most comprehensive survey of past research on income distribution is Albert Berry and Miguel Urrutia, Income Distribution in Colombia (New Haven: Yale University Press, 1967). OBJECTIVES AND LIMITATIONS 9 in market demands (for example, health), an additional difficulty arises. The benefits from the extra consumption of these services cannot be approximated by the prices individuals were paying for private substitutes before public intervention. Thus, tracing such benefits becomes a problem analogous to that in social project evaluation. This study addresses a more restricted set of questions, which can be summarized as follows: * Ilow is the consumption of publicly provided services distributed by income groups, regions, and other relevant characteristics associated with income levels? This study only covers services for which individual or household consumption can be identified and, if possible, measured. The typical pure public good, such as defense or justice, is excluded. x What is the subsidy associated with the provision of these services by the public sector? Two definitions of the subsidy are of interest. The first definition is the subsidy derived from the government pricing policy: in other words, the difference between the long-run cost of the service to the governnment and the price to consumers. It takes the pres- ence of the public sector as given and simply asks about the dis- tributive effect of the government pricing policy. The second definition is the subsidy of having a public sector with that par- ticular pricing policy as opposed to a situation in which equivalent services would have been provided by the private sector. This subsidy consists of the earlier definition (the pricing policy sub- sidy) plus the difference in cost between private and public pro- vision of the service. For particular services, this research derives estimates for the first definition of the subsidy. Estimates for the second definition would require data on the cost of equivalent services in the absence of public intervention (for example, the cost to a household of obtaining a unit of light from oil lamps)-a task beyond the scope of this study. * What factors explain the distribution of consumption addressed in the first question? Ilow much can be explained by the existing structure of supply of the service and how much by factors governing the private demand for these services? 10 BACKGROUND AND MAIN RESULTS This is a difficult area to research. It is, however, the most relevant in identifying policy instruments to increase the con- sumption of these services by low-income groups. To identify these policy instruments, it is useful to distinguish between two types of situations in which a particular household does not consume a public service: (1) the household does not consume the service because the supply network is not geograph- ically present and (2) the household is on the supply network but decides not to consume the service. In the first instance there is no consumption because of the location of the supply network: it is basically an institutional datum determined by past investment decisions in the sector. In the second, the lack of consumption is determined by demand and results from a voluntary choice. Because the factors governing this choice (variables behind the dernand) may be different from the factors determining the location of the network (variables behind the supply), it is useful to recognize explicitly this supply- demand mechanism behind consumption. This study attempts to identify the explanatory factors behind the supply and demand for particular services. - What income groups htave benefited from the changes in the supply of these services as a result of recent investment policies? Thus far, the research has addressed the distribution of con- sumption at one time: that is, the consumption stemming from the existing supply or stock of infrastructure. Of at least equal in- terest is the distribution of new consuiners of the service that results from expansions of the supply over time. For those sec- tors in whlichl investmlent has been significant as a fraction of the national budget, this study estimates the income distribution of the new househlolds consuming the service. Methodology and Further Simplifications This section outlines the metlhodology followed to address the four questions stated above. It will become evident that further simplifications are required in addition to those already mentioned. The subsidy received by a househiold from consuming a service (S) is equal to the subsidy per unit (s) multiplied by the units of METHODOLOGY AND FURTHER SIMPLIFICATIONS 11 that service being consumed by the household (Q). The subsidy per unit is equal to the difference between the long-run marginal cost to the government of providing the service (c) and the price charged to consumers (p). The quantity consumed (Q) is itself a function of the price charged and of other socioeconomic variables (x) influencing the household demand for the service.' 3. The relation between the two definitions of the subsidy discussed earlier and the extra benefits to consumers out of public provision of the service are shown in the figure below. Suppose D is the demand for the service and c and c' are the long-run marginal costs of providing the service by the public and by the private sector, respectively. Assume further that c' is greater than c (the public provision has a lower cost than the private substitute) and that the government prices the service at p, below its cost, c. The implicit subsidy of having the service provided by the public sector with that pricing policy-instead of by the private sector-is equal to s* = s + Ac; s* is there- fore equal to the pure pricing policy subsidy (the definition to be used in this study) plus the difference in cost between provision by the private and by the public sector. The extra consumer surplus resulting from public provision of the service-when it replaces more expensive private sources of the service-is s* [Qo +2 (Q2 - Qo)]. The one resulting from the pure pricing policy is equal to s [Ql + 2 (Q2 -Q)]. The concept of S used in this study is S = sQ2. Price C, CI P~~| 11 1 I li \ I I I (X) Qo Ql Q2 Qatt I ~~~~~~~~ Igntt 12 BACKGROUND AND MAIN RESULTS The problem of using the national budget to identify public services with significant subsidies and the limitations of using the budget to compute c, the long-run cost of providing these services, are discussed first. The houselhold sample survey carried out to derive data on the Qis and to identify the factors influencing con- sumption is then described. Identification of puiblic services throutgh the national buidget Public services can be thought of as a flow of services provided by a capital stock (or stock of infrastructure) and other variable resources such as labor and intermediate inputs. The long-run marginal cost of a unit of this flow includes the opportunity cost of capital required to provide this unit plus the associated variable cost. T o what extent does the government budget provide a measure for this long-run marginal cost? If the public sector hired or rented the capital stock or infrastructure from the private sector, the annual rental payments that would appear in the nationial budget could be used for this purpose. Since the public sector owns the stock, no rental paymenits are recorded in the budget. Only non- rental or running costs, that is, labor services and intermiiediate inputs, appear explicitly. The more labor intensive the production of a public service is, the more easily the long-run cost can be ap- proximated by the cost data in the national budget. The extreme, thouglh not uncommiiiioni, example is that of a capital- intensive public service with negligible labor and intermediate input cost, for example, electricity derived from hvdro sources. Ihe national budget is unable to capture the true marginal cost of providing the service during a particular year. This cost can only be estimated from micro studies of an expansion of the given system. The empirical implications of this are quite clear. By using the government's current expenditures as a substitute for the cost of the service, only the labor-intensive public services, that is, educa- tion and healtlh, can be examined. Consequently thiis study de- rives subsidy figures for these sectors only. The estimated figures will understate the true subsidy received by houselholds to the extent that the rental cost of the capital stock in these sectors is not included in the budget. METHODOLOGY AND FURTHER SIMPLIFICATIONS 13 No estimates are presented here for subsidies to capital-intensive public services.4 In the case of electricity, piped water, and sew- erage, availabilitv at one time and the supply-demand mechanism behind consumption are analyzed, and the income groups that benefited from expansions in the supply network over time are identified. Some attempts were made to estimate the benefits of invest- ment in roads, an important fraction of the national budget in Colombia. The mechanism by which this investment generates benefits, however, proved too complex to be measured by a house- hold survey. A different problem arises in dealing with the health services provided by the social security system for employees in the pri- vate sector. This system is financed largely by contributions from private employers and employees, with a minimal contribu- tion from the central government. Although the measured con- tribution from the public sector is sniall, the system induces an important transfer across income groups. Because the size of the transfer depends on the real or economic incidence induced by the legal contributions, a simple labor market model was developed in Chapter 4 to derive this incidence. The 1974 houtsehold sanmple suzrvey Two possible strategies can be followed to derive data on the consumption of public services by income groups. First, house- hold income data can be taken from the records of the institutions that provide the services. If these records do not contain income data, this information can be obtained by interviewing a sample of houselholds listed in those records. Second, a countrywide sample survey cani be mlade, including houselholds witlh and without ser- vices. Given the objectives of this study, the countrywide survey has two main advantages. First, it provides data on households without services. From a policy point of view, this information is more important than data on the distribution of houselholds withl ser- vices. Second, it generates its own incomiie distribution data against 4. Some estimates of the transfer across consumers resulting from the tariff struc- ture for electricity and water are presented in the appendix to Chapter 5. 14 BACKGROUND AND MAIN RESULTS wlhiclh to map the households to be studied; this makes it unnec- essarv to use income distribution estimates from other sources and thus avoids problems of comparability. TRADEOFFS IN THE SAMPLE DESIGN. To address simliultaneously the four questions stated above implies a tradeoff in the sample design. If the objective is to derive a statistically significant estimate for the income groups already consuming services, a minimum sample size breakdown for these groups is required. If the emplhasis is on income groups witlhout services or on ones that have only recently received themii, a minimiiumii sample size of low-incomie groups is necessary. This tradeoff becomes particularly important in testing the factors behind the availability of services in a supply-demand context. For such an analysis, a minimuml sample size of house- holds witlhout services is needed to split the group further between those who are on the supply network and those wlho are not. Since public services are available in most of the urban areas in Colombia, low-income groups must be overrepresented in the sarnple to obtain a minimum sample size of houselholds witlhout services. Thlis prob- lem will be considered again in the discussion of the results. A sample survey of 4,019 households and 22,064 individuals was undertaken especially for this study by the Compafnia Colombiana de Datos (COLDATOS) in December 1974.5 The composition of the sample according to major breakdowns appears in Table 1.1. STRATIFICATION. Urbanization, as measured by city size, is the first level of stratification used in the sample design. This stratifica- tion results from two hypotheses. The first is that urbanization is highly correlated with the consumption of governm11ent services. Thus, the first requirement in the stratification is to provide statistically significant estimates for major urban strata classified by city size. The second hypothesis is that the higher the degree of urbanization, the more the consumption of services varies across households, this variation being associated with the level of house- hold income. The second requirement, therefore, is to derive sig- 5. A full description of the sample design and stratification can be found in Compafiia Colombiana de Datos (COLDATOS) "Disefno de la Muestra del Banco Mundial" (study prepared for the World Bank, Bogota, 1976; processed). METHODOLOGY AND FURTHER SIMPLIFICATIONS 15 Table 1.t. Size and M1ajor Categortes of the 1974 Sample Suirvey Selected Percentage of households neighbor- M1ajor regions hoods or Selected In the In the (inhabitants) Strata hamlets households sample population Large cities 5S.b 184 1050 26.2 28.9 (more than 500,000)a (neighbor- hoods) (1) (51) (321) (8.0) (3.2) (2) (70) (355) (8.9) (11.3) (3) (40) (203) (5.1) (10.1) (4) (17) (94) (2.3) (2.2) (5) (6) (77) (1.9) (2.1) Intermediate cities 9 187 994 24.7 17.5 (30,000 to 500,000) (neighbor- hoods) Small towns 13 13 725 18.0 15.5 (1,500 to 30,000) (cities) Ruralareas 21 114 1,250 31.1 38.1 (less than 1,500) Total 48 498 4,019 100.0 100.0 a. "Large cities" include Bogota, Cali, Medellin, and Barranquilla. b. The five strata in the large cities are defined as follows: (1) slum neighborhoods; (2) low- income neighborhoods; (3) low-middle- andl middle-income neighborhioods; (4) middle-high- and high-income neighborhoods; and (5) municipalities attached to the large metropolitan areas. These definitions of strata are taken from the sample frame of neighborhoods used by the Bureau of Census (see the appendix to this chapter). nificant estimates of consumptioni h Imajor incomiie or socioeco- noniic groups, at least for the largest cities. The first regional breakdown of the sample is urban-rural, with the urban population classified by city size according to three inajor categories: large cities of over 500,000 inhabitants (the four largest in Colombia) ; intermediate cities of 30,000 to 500,000 inlhabitants; and small towns of 1,500 to 30,000 inhabitants. These four major categories are broken downi further into forty-eight strata: five for large cities, nine for intermediate cities, thirteen for smuall towns, and twenty-one for the rural areas. Eachi of the five strata in the large cities defines a group of neighborlloods witlh hlomogeneous socioeconomic characteristics. The geographic definitions of the neighborhoods are taken from urbani maps, and their characteristics are derived from the 1970 16 BACKGROUND AND MAIN RESULTS household survey undertaken by the Bureau of Census (Departa- mento de Estadistica; DANE).6 A total of 184 neighborhoods or clusters are selected from these strata. Intermediate cities are grouped into nine strata, with each group defined according to major geographic region and city size (be- tween 30,000 and 100,000 and between 100,000 and 500,000 in- habitants). From each stratum one city is selected. Three criteria are used to classify the small towns: a further breakdown of the size of towns, geographic region (as described above), and charac- teristics of the agricultural activities surrounding the town. One town is selected from each stratum. The twenty-one rural strata are derived from the aggregation of 786 micro-rural regions or primary sampling units (psu) that had been defined previously according to several criteria of agri- cultural homogeneity.7 One or two Psu are chosen from each of the twenty-one strata, three to eight hamlets are selected from each PsU, and ten to twelve dwellings from each hamlet. DISTRIBUTION OF HOUSEHOLDS BY INCOME QUINTILES. Table 1.2 shows the composition of the sample after the survey had been carried out and the household income data computed. The distribu- tion of households in the population is obtained by expanding the household data according to the weights derived from the stratifica- tion procedure, which corrects for any under- or overrepresentation in the sample. The expanded or weighted data can then be used to derive the distribution of households in different regions ac- cording to per capita household income. It is also possible to order households in each region according to per capita income and to compute the income figures that define quintiles (fifths) of the total population or households in the region. Table 1.2 also presents the number of sampled households in each of the income ranges defining quintiles. This information provides the basis for judging whether the initial objective of the stratification-overrepresentation of low-income groups-was 6. See the appendix to this chapter. 7. Each Psu includes one to three adjacent municipalities with the same access road and similar agricultural activities. These activities were defined as: cattle raising; coffee growing; large agro-industrialized farming (sugar, cotton); and cropping activi- ties (for example, corn and yucca) in subsistence farming. For a detailed description, see the COLDATOS report. MAIN RESULTS 17 Table 1.2. Number of Households in the Sample, Classified by Quintiles in the (Expanded) Regional Distribution of Income Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest), cities cities towns total areas total 1 285 208 137 663 278 863 2 259 227 155 624 261 886 3 209 188 145 567 239 796 4 152 198 150 499 252 786 5 145 173 138 416 220 688 Country total 1,050 994 725 2,769 1,250 4,019 a. Quintiles are calculated from the weighted number of households in the income levels in each region. Because the per capita household income defining a particular quintile differs from region to region, figures in a row do not necessarily add up to urban or to country totals. achieved. If the share of households sampled in one quintile is larger than 20 percent (the share in the population), that quintile is overrepresented in the sample. The table shows that this is indeed the case for low-income groups, particularly those in the largest cities. Main Results The main results of this study are presented in this section. Thus, it is possible to compare the distribution of beneficiaries across all the public services studied. The distributions for each service are presented and analyzed in detail in separate chapters. Distribution of income The income data from the survey can be used to estimate the distribution of income in both urban and rural areas. Such esti- mates are given in Table 1.3, which presents urban, rural, and countrywide income distribution data. Results are shown for the accumulated percentages of families, in ascending order according to household per capita income. For each accumulated percentage of families, the corresponding accumulated percentage of individuals and the accumulated share of the total income is given. For ex- ample, the poorest 40 percent of urban households make up 46.9 18 BACKGROUND AND MAIN RESULTS Table 1.3. Distribution of Income, 1974 (accumulated percentages) Percentage of house- Urban Rurala Country holds, by per capita household income Popu- Popu- Popu- (lowest to highest) lation Income lalion Income lation Income 10 12.5 1.7 13.4 3.0 13.0 1.9 20 24.7 4.6 24.9 7.6 25.1 5.2 30 36.2 8.6 36.6 13.8 36.5 9.4 40 46.9 13.3 47.7 20.9 47.2 14.3 50 56.8 18.9 59.2 29.8 57.7 20.4 60 66.1 25.5 68.6 38.6 66.9 26.9 70 75.5 34.2 76.7 47.5 75.7 34.9 80 84.4 45.7 85.5 59.4 84.4 45.6 90 92.6 62.3 93.5 73.5 92.5 60.7 95 98.5 73.9 97.0 81.9 96.4 72.8 a. Population living in conglomerates of less than 1,500 inhabitants. percent of the urban population and receive 13.3 percent of the total urban income. The data suggest that rural income is more evenly distributed, particularly at low levels of income, even if reference is made to the distribution of the population instead of to the distribution of households. This is contrary to estimates for 1964 by Berry and Urrutia, althouglh it can be partly explained by the fact that their estimates included the income of landowners living in cities in the rural income figure.8 A comparison of Gini coefficients from the 1974 sample survey and from Berry and Urrutia yields the following figures: Gini coefficients Urban Rural Country Berry and Urrutia (1964) (economically active population) 0.55 0.57 0.57 1974 Sample survey Population 0.54 0.42 0.50 Households 0.48 0.32 0.47 8. Berry and Urrutia, Income Distribution in Colombia, Chapter 2. MAIN RESULTS 19 Figure 1 A. Distribution of Income: Comparison of Estimates Percent of income 9- lC - Berry- Urrutia, 1964 (- - - 1974 sample survey 70 - / -fi 60- ___ - - - - - /0 Households / 40 - (G=0 47) Economically act ve / v /v: , ~~~~~population 0 1020 3 0 40 50 60 7 0 80 90100o Percent of householdsl Percent of population Because low-income groups have a larger familv size, the Gini coefficient for the distribution of the population is larger than for the distribution of households. Although the estimate for the urban population is almost the same as that of Berry and Urrutia, the difference between the estimates is substantial for rural areas. A comparison of Lorenz curves for the total country is presented for both estimates in Figure 1.1. The curves for the urban areas are quite similar if the distribution of the population is used for the sample survey figures. The same income data can be used to compute the fraction of 20 BACKGROUND AND MAIN RESULTS Table 1.4. Per Capita Income Conversions, November 1974 Monthly per capita income (pesos) According to the According to the Kravis parity for Annual per capita official exchange private consumption income (dollars) rate = 27.6 pesos = 13.0 pesos' 0-50 0-115 0-54 51-75 116-172 55-81 76-100 173-230 82-108 101-150 231-345 109-162 151-250 346-575 163-271 251-350 576-805 272-379 351-500 806-1,150 380-542 501-700 1,151-1,610 543-758 701-1,500 1,611-3,450 759-1,625 Over 1,500 Over 3,450 Over 1,625 a. This parity was calculated on the basis of 8.3 pesos for 1970. The parity for November 1974 becomes equal to: 8.3 El + (AP/P)co1/1 + (Ap/p)sj] = 8.3 (2.09/1.33) = 13.0 where Ap/p corresponds to the change in the price level between 1970 and November 1974 in Colombia and the United States. Irving B. Kravis, Zoltan Kenessey, Alan Heston, and Robert Summers, A System of International Comparisons of Gross Product and Purchasing Power (Balti- more: Johns Hopkins University Press for the World Bank, 1975). the population receiving a per capita income lower than a prede- termined value, such as a poverty cutoff line. To derive that in- formation for predetermined values of per capita income in dollars (to facilitate international comparisons), an exchange rate is re- quired. Two alternative values can be used for November 1974 (a month before the survey): the official exchange rate of 27.6 pesos to the dollar and the Kravis parity rate for private con- sumption of 13 pesos to the dollar. In Table 1.4 the monthly per capita income in Colombian pesos, equal to predetermined ranges of dollar annual per capita income, is shown for both exchange rates.9 9. Table 1.4 was calculated from data on the distribution of households by per capita income according to both exchange rates. These data are presented in the statistical appendix, Tables SA-1 and SA-2. MAIN RESULTS 21 Table 1.5. Percent of lIouseholds below Poverty Lines Annual Urban Rural Country per capita income Official Kravis Official Kravis Official Kravis (dollars) rate rate rate rate rate rate Below 50 5.9 1.4 10.5 1.9 7.6 1.6 Below 100 20.8 5.3 37.0 9.6 26.8 7.0 Below 150 35.9 11.3 60.0 21.0 44.8 15.0 Table 1.5 shows the percentage of households in each region be- low commonly used poverty lines: that is, witlh per capita yearly incomes of less than $50, $100, and $150. As expected, the results are extremely sensitive to the choice of exchange rates. Calcu- lated with the IKravis rate, the percent of households with a per capita income under $50 is negligible; below $100, it is a fourth of wlhat it would be if the official exchange rate were used. Education and health In 1974 the total public subsidy to education, which includes the cost of public education plus subsidies to private schools, was 11.2 percent of the total public expenditure and 2.2 percent of the GNP. The estimated public subsidy to health was 3.6 percent of the total public expenditure and 0.7 percent of the GNP, or ap- proximately one-third of the subsidy to education. As noted pre- viously, these estimates do not include the opportunity cost of the public capital stock in these sectors. Colombia's health system consists of institutions belonging to the National Health System (NHs), which theoretically benefits all individuals; institutions maintained by the Social Security of the Public Sector (Cajas Publicas), with which employees in the public sector are affiliated; and institutions run by the Colombian In- stitute of Social Security (icss), with which employees in the private sector are affiliated. In their running costs, the NHS and icss are of similar size, and together they account for approximately 90 percent of the running cost of the entire health network. Nevertheless, a different 22 BACKGROUND AND MAIN RESULTS Table 1.6. Distribution of Subsidies for Education and Health per Household and per Capita, Classified by Income Quintile (1974 pesos) Income Subsidy for education Subsidy for health quintile (poorest Univer- to richest) Primary Secondary sity Total NHS sSS Total 1 1,305 598 18 1,921 514 103 617 2 1,089 776 96 1,961 440 186 626 3 835 751 224 1,810 393 381 774 4 589 872 489 1,950 321 314 635 5 252 555 1,257 2,064 210 295 505 Country average 816 718 413 1,947 376 255 631 strategy is required to compute the subsidy received by the bene- ficiaries of the two systems. The subsidy to the NHS is relatively easy to compute because it is directly financed by the Ministry of Healtlh and by contributions from the departmental governments. The icss subsidy comes from workers and employers in the private sector, whose contributions amount to 95 percent of the total financing. The implicit subsidy received by the icss affiliates is equal to the total contributions to the system minus the real incidence borne by labor. The share borne by labor is equal to the difference between the present wage and the wage that would have prevailed without the system. This share depends not only on legal (per- centage) contributions and on the supply and demand elasticities for labor, but also on the value placed by workers on the yearly services provided by the system, which affects the new (post- system) supply price of labor and therefore the new equilibrium wage. It is estimated here that half of the total financing of the sys- tem is borne by labor. This figure is higlher than the legal incidence of 0.33 resulting from legal rates of 3.5 and 7 percent of the wage that is paid by workers and employers, respectively (0.035/0.105 = 0.33). MAIN RESULTS 23 Total Subsidy subsidies Mean Subsidy per capita as a for educa- household Annual percentage of tion and size 1974 1974 household household health (persons) pesos US$ income income 2,538 6.87 369 13.3 10,368 24.5 2,587 5.99 436 15.6 17,820 14.5 2,584 5.38 480 17.4 25,032 10.3 2,585 4.80 539 19.5 36,912 7.0 2,569 4.25 604 21.9 104,388 2.5 2,578 5.50 468 17.0 38,904 6.6 Table 1.6 presents estimates of the subsidy received by the representative household in each income quintile. These quintiles are defined by the distribution of households according to house- hold per capita income. For health, the table shows the subsidy from NHS and sss institutions, where sss refers to the overall social security network including both Cajas Publicas and icss. The total subsidy per household from education and healtlh is remarkably constant across income groups. This constancy dis- appears, however, when subsidies are broken down by type of education and by type of health service. Subsidies for primary education are much larger for low-income households, and those for higher education are much smaller. NHS subsidies are higher for low-income families, whereas sss subsidies are lower. The constancy of the subsidy as expressed per household does not hold if expressed per capita, because household size is sub- stantially larger in the poorest quintiles. The per capita subsidy for the richest quintile is 1.65 times the per capita subsidy for the poorest quintile. Expressed as a percentage of the household income, the total subsidy is substantially larger for low-inconme groups: 24.5 percent of the household income for the poorest quintile, compared with 2.5 percent for the richest quintile. 24 BACKGROUND AND MAIN RESULTS Table 1.7. Distribution of Income and Subsidies for Education and Health, Classified by Income Quintiles (percentage) Income Subsidies for education quintile (poorest Popula- Second- Univer- to richest) tion Income Primary ary sity Average 1 25.1 5.2 32.1 16.8 0.8 19.8 2 22.7 9.1 26.7 21.8 4.6 20.2 3 19.4 12.6 20.5 21.2 10.7 18.6 4 17.4 18.7 14.5 24.6 23.5 20.1 5 15.4 54.4 6.2 15.6 60.4 21.3 Figure 1 .2. Distribution of Income and of Subsidies for Health and Education, 1974 Percent of incomef Percent of subsidy 100 111 ' 90…__ _ 70~-- 80 - - - Total health - Primary education /450 Total educaion 30 - - / Personal income / 20 A@{/s 10 z S - University education 0 10 20 30 40 50 60 70 80 90 100 Percent of population MAIN RESULTS 25 Subsidies for health Average of subsidy for education NHS sss Average and health 27.3 8.1 19.5 19.7 24.2 15.1 20.5 20.8 20.5 29.4 24.2 19.7 16.9 24.5 20.0 20.0 11.1 22.9 15.8 19.8 The distribution of the subsidies and the distribution of income are compared in Table 1.7 and Figure 1.2. Figure 1.2 represents a Lorenz relation: the population is ordered on the horizontal axis from lower to higher per capita income, and the accumulated distribution of income and subsidies corresponding to that popula- tion is shown on the vertical axis. The line for primary education lies above the diagonal: that is, lower-income groups have a larger share of the subsidy. The line for higher education not only lies below the diagonal, but actually below the line for the distribution of income: the distribution of the subsidy to higher education shows a stronger inequality than the distribution of personal income. The lines for the total subsidy in education and health lie close together, between the diagonal and the line for the distribution of personal income. Thus, when mapped against the distribution of the population, the subsidy tends to favor higher-income groups, although it is more equally distributed than personal income. Electricity, piped water, and sewerage Data from the 1974 sample survey are also used to analyze the distribution of consumption and the availability of public utility services. CONSUMNPTION IN 1974 AND BETWEEN 1970 AND 1974. Table 1.8 Shows, for the whole country, both the distribution of households 26 BACKGROUND AND MAIN RESULTS Table 1.8. Ilouseholds with Services in 1974 and Households Which Were Connected to the Nietwork between 1970 and 1974, Classified by Income Groups (percent-ge) Income quintile Electricity Piped waler Sewerage Street lighting (poorest to richest) 1974 1970-74 1974 1970-74 1974 1970-74 1974 1970-74 1 13.1 26.0 13.9 31.4 10.7 24.5 12.9 25.6 2 16.1 25.1 16.3 23.6 14.9 21.3 15.9 24.4 3 19.2 25.6 19.4 26.8 18.5 29.1 18.4 22.5 4 23.0 12.0 22.5 12.6 23.7 12.0 23.1 11.9 5 28.6 11.3 27.9 5.6 32.2 13.1 29.7 15.6 Urban 90.2 50.4 88.0 41.4 96.2 81.7 96.0 100.0 that reported having services in 1974 and that of households with- out services in 1970 but obtaining themii between 1970 and 1974. The percentage of these households located in urban areas is also shown. For 1974, the distribution of consumers by quintiles is similar across services: 25 to 30 percent of houselholds with services belong to the poorest 40 percent of houselholds, wlhereas 50 to 55 percent belong to the richest 40 percent. Almost all consumers live in urban areas. Investment in electricity from all sources to generate and to transmit power fluctuated between 4.5 and 7.6 percent of the total governmenit expenditure betweeni 1970 and 1974, witlh the largest figure corresponding to 1974. For piped water and sewerage, these figures range between 1.7 and 2.7 percent. Wlhat income groups benefited from this investmelnt? To answer this question, households with the services in 1974 were asked wlhether they had had the services in 1970. With this information it is possible to compute the distribution of houselholds that were connected to the service between 1970 and 1974. The data show that this distribution tends to benefit lower-income groups: 46 to 55 percent of the newly connected houselholds belong to the poorest 40 percent of families. This is partly because in- vestment has had a lower "urban bias." In the case of electricity MAIN RESULTS 27 and piped water, more than half of the newly connected house- holds are located in rural areas (as defined in this study)."i CONSUNIPTION IN A SUPPLY-DEMIAND CONTEXT. A multivariate analysis was made to identify variables associated with the con- sumption of services by households in urban areas. It was impor- tant to develop a framework in whichl this association could be interpreted to a certain extent as a cause-effect relation. In this framework, a household comes under one of two classifi- cations: either it does not use a service because the supply network is geographically inaccessible, or it does not use the service be- cause-in spite of accessibility, that is, at the prices charged by the utility company-it has decided not to do so. The second situation is basically determined by demand. Data from the sample survey show that households without services because of lack of demand represent a substantial per- centage of the houselholds without services, particularly for elec- tricity and piped water. Thus, it is necessary to understand the reasons for this behavior to explain the different availability of services across houselholds. To do this, the factors influencing the probability of a household lhaving a particular service, P, need to be identified. By using data on the individual characteristics of households with and without services, it is possible to estimate the influence of these charac- teristics on the probability of having the service. The direct estimation of P does not fully capture the supply- demand mechanism discussed earlier. Hence, it does not allow the identification of the extent to wlhich a particular variable or char- acteristic influences P througlh the demand or supply side. A more interesting specification of the problem is to think of P as the product of two probabilities to be estimated independently: the probability of a household being on the supply network of the service, PI, that is, the probability of having access to the network at the connection cost institutionally set by public utility com- panies; and the probability of demanding the service, pd, if offered at this connection cost. Both PI and Pd are functions of a set of 10. Availability of piped water was defined as a situation in which the dwelling is connected to an aqueduct or to the primary water network. In rural areas, therefore, it includes households with any kind of access to a public aqueduct. 28 BACKGROUND AND MAIN RESULTS Table 1.9. Electricity Use in Urban Areas: Estimates of the Linear Probability Function Total urban Poorest 40 percent, Explanatory variables p pd pi p pd pe 1. Constant 0.75 0.79 0.88 0.77 0.75 1.00 (13.89) (15.12) (24.25) (5.0) (5.21) (9.80) 2. Intermediate -0.04 -0.02 -0.02 -0.07 -0.05 -0.02 cities (3.11) (2.05) (2.32) (2.74) (2.22) (1.41) 3. Small towns -0.14 -0.12 -0.03 -0.19 -0.17 -0.04 (10.86) (10.62) (3.27) (7.48) (7.03) (2.53) 4. Dirt floor -0.36 -0.28 -0.15 -0.35 -0.27 -0.15 (20.69) (16.47) (13.15) (12.53) (9.91) (7.94) 5. Rural -0.11 -0.08 -0.04 -0.17 -0.16 -0.04 migrant (4.73) (3.90) (2.64) (4.09) (3.91) (1.52) 6. Log of per 0.02 0.02 0.002 -0.01 -0.01 0.004 capita income (2.28) (2.64) (0.39) (0.22) (0.22) (0.13) 7. Log of years 0.08 0.06 0.03 0.17 0.13 0.05 of schoolingof (5.21) (4.17) (2.91) (4.70) (3.95) (2.17) head of household S. Log of age 0.09 0.07 0.06 0.11 0.14 -0.02 of head of (2.93) (2.60) (2.88) (1.7) (2.18) (0.39) household Mean 0.92 0.94 0.97 0.84 0.88 0.95 R2 0.31 0.24 0.12 0.30 0.24 0.11 Note: P = probability of a household having electricity; PI - probability of a household being on the supply network for electricity; pd = probability of a household demanding electricity. Figures in parentheses are I statistics. a. Percentage of families according to household per capita income. b. Not applicable. variables to be estimated independently. The variables influencing ps determine the utility company's behavior regarding the location of the network or supply; those influencing Pd are demand-oriented variables: that is, the cost of connection relative to the income of the household and other socioeconomic characteristics of house- holds that govern such demand. The actual estimation of P, pa, and Pd was undertaken by defining a household on the supply network as one having a neigh- bor with the service within one block. This definition reasonably reflects a situation where a household can be connected if it is will- ing to pay a typical connection cost. MAIN RESULTS 29 Poorest 20 percenta Small towns p pd ps p pd ps 0.67 0.68 0.95 0.30 0.10 0.62 (2.93) (3.11) (5.98) (2.03) (0.53) (6.38) -0.06 -0.05 -0.02 _ b _b _b (1.55) (1.34) (0.70) -0.20 -0.18 -0.04 _b _b _b (5.16) (4.89) (1.48) -0.37 -0.28 -0.18 -0.46 -0.39 -0.18 (10.08) (7.46) (6.93) (11.4) (9.44) (6.75) -0.29 -0.28 -0.07 -0.19 -0.14 -0.99 (5.28) (5.12) (1.73) (3.79) (2.77) (2.71) -0.04 -0.04 0.01 0.08 0.11 0.005 (0.49) (0.59) (0.23) (2.83) (3.55) (0.26) 0.18 0.17 0.03 0.18 0.16 0.06 (3.39) (3.29) (0.90) (3.39) (3.12) (1.85) 0.22 0.23 0.01 0.18 0.28 0.20 (2.14) (2.36) (0.20) (2.17) (2.71) (3.62) 0.81 0.86 0.94 0.76 0.81 0.93 0.35 0.28 0.13 0.31 0.26 0.14 Tables 1.9, 1.10, and 1.11 show the effects (coefficients) of different variables on P, Ps, and pd. The results can be summa- rized as follows: * The smaller the size of the city, the smaller the probability of having a service. This is particularly true for electricity and sewerage and is especially strong for the poorest quintile of urban households. * When dirt floor, a characteristic of the dwelling, is introduced, the per capita income variable tends to become insignificant.'" Dirt floor appears to be a more powerful demand variable in the case of electricity and piped water; the reverse is true for sewerage. 11. This was observed in experiments not reported here. The per capita income variable was always significant when dirt floor was not included. 30 BACKGROUND AND MAIN RESULTS Table 1.10. Piped Water Use in Urban Areas: Estimates of the Linear Probability Function Total urban Poorest 40 percent' Explanatory variables P pd p. P pd pe 1. Constant 0.84 0.92 0.91 0.73 0.91 0.76 (26.04) (35.50) (37.24) (5.50) (8.02) (7.42) 2. Intermediate -0.04 -0.02 -0.03 -0.04 -0.03 -0.01 cities (2.97) (1.48) (2.57) (1.14) (0.99) (0.52) 3. Small towns -0.06 -0.06 -0.004 -0.05 -0.07 -0.01 (3.74) (4.53) (0.37) (1.79) (2.70) (0.46) 4. Dirt floor -0.34 -0.23 -0.17 -0.33 -0.23 -0.17 (16.34) (12.71) (10.95) (9.95) (7.67) (6.57) 5. Rural -0.07 -0.06 -0.01 -0.04 -0.05 0.02 migrant (2.57) (2.74) (0.46) (0.77) (1.16) (0.44) 6. Log of per 0.04 0.02 0.02 0.07 0.03 0.06 capita income (3.29) (2.55) (2.0) (1.34) (0.65) (1.55) 7. Log of years 0.09 0.06 0.04 0.11 0.06 0.06 of schooling of (4.96) (3.83) (2.70) (2.57) (1.61) (1.79) head of household 8. Log of years -0.03 -0.03 -0.002 -0.03 -0.03 -0.003 in same (2.93) (3.48) (0.24) (1.58) (1.75) (0.17) municipality Mean 0.90 0.94 0.95 0.81 0.89 0.91 R2 0.19 0.15 0.08 0.15 0.12 0.07 Note: P = probability of a household having piped water; P' = probability of a household being on the supply network for piped water; Pd = probability of a household demanding piped water. Figures in parentheses are I statistics. a. Percentage of families according to household per capita income. b. Not applicable. * The rural-migrant characteristic tends to operate on the demand side. It has a strong effect in the case of electricity for the poorest 40 per- cent of households. * The log of years-of-schooling-of-head-of-household has a stronger effect in the poorer-income groups. A coefficient equal to 0.15 means that an increase of 50 percent in years of schooling will increase the probability by 7.5 percentage points.'2 12. When the independent variable is defined in log form, the coefficient becomes AP/A log xi. MAIN RESULTS 31 Poorest 20 percent' Small towns p pd pe p pd p. 0.85 0.94 0.83 0.81 0.87 0.93 (4.09) (5.21) (5.01) (9.81) (11.64) (15.97) 0.001 -0.01 0.01 _b _b -b (0.024) (0.25) (0.32) -0.03 -0.07 0.04 _b _b _b (0.59) (1.75) (1.12) -0.36 -0.23 -0.21 -0.33 -0.27 -0.12 (7.81) (5.37) (5.74) (8.02) (6.89) (4.29) -0.11 -0.11 0.01 -0.13 -0.12 -0.01 (1.57) (1.93) (0.16) (2.49) (2.50) (0.37) -0.01 0.003 0.02 0.05 0.05 0.01 (0.063) (0.032) (0.24) (1.79) (1.74) (0.52) 0.14 0.10 0.04 0.09 0.05 0.04 (2.09) (1.79) (0.84) (1.64) (1.06) (1.09) -0.02 -0.03 0.01 -0.08 -0.06 -0.02 (0.55) (0.96) (0.59) (3.38) (3.11) (1.09) 0.77 0.87 0.88 0.79 0.86 0.92 0.16 0.13 0.08 0.18 0.15 0.05 T The log of age-of-head-of-household was significant and was stronger on the demand side in the case of electricity. For the other services, it was not significant and was not included in the final regressions. - A negative coefficient for the log of number-of-years-in-the-same- municipality for piped water and sewerage is not easy to interpret. It can be hypothesized that the longer the household remains without the service, the lower the probability that the household will have it today. Given the nature of the sample, the data used in the analysis did not provide enough variation in the dependent variable. Most of the sampled households did have services, which reflects both the large coverage of services in urban Colombia and the fact that the design was geared more to obtaining statistically signifi- 32 BACKGROUND AND MAIN RESULTS Table 1.11. Sewerage Use in Urban Areas: Estimates of the Linear Probability Function Total urban Poorest 40 percent' Explanatory variables P pd p. P pd P, 1. Constant 0.79 0.92 0.86 0.57 0.73 0.77 (19.36) (35.43) (22.36) (3.74) (5.92) (5.27) 2. Intermediate -0.12 -0.02 -0.10 -0.13 -0.03 -0.12 cities (6.78) (2.03) (6.29) (3.67) (0.99) (3.35) 3. Small towns -0.15 -0.09 -0.09 -0.16 -0.12 -0.08 (7.81) (6.81) (4.93) (4.49) (4.25) (2.45) 4. Dirt floor -0.50 -0.21 -0.47 -0.43 -0.18 -0.43 (18.87) (18.56) (18.81) (11.36) (4.49) (11.79) 5. Rural -0.06 -0.05 -0.02 -0.004 -0.01 -0.005 migrant (1.82) (2.22) (0.75) (0.07) (0.17) (0.09) 6. Log of per 0.04 0.02 0.02 0.10 0.08 0.05 capita income (3.10) (2.46) (1.76) (1.64) (1.55) (0.85) 7. Logof years 0.16 0.05 0.13 0.22 0.12 0.17 of schooling of (6.70) (3.04) (5.97) (4.60) (2.95) (3.66) head of household 8. Log of years -0.06 -0.02 -0.05 -0.06 -0.0002 -0.07 in same (4.90) (2.19) (4.24) (2.43) (.014) (3.05) municipality Mean 0.80 0.95 0.84 0.66 0.90 0.73 R2 0.27 0.11 0.23 0.24 0.10 0.21 Note: P = probability of a household having sewerage facilities; P' = probability of a house- hold being on the supply network for sewerage facilities; pd = probability of a household demand- ing sewerage facilities. Figures in parentheses are I-statistics. a. Percentage of families according to household per capita income. b. Not applicable. cant estimates for averages of the population than to hypothlesis testing involving multivariate analysis (for the latter, an even stronger overrepresentation of low-income groups without services would have been required). The relatively large number of houselholds with services in the sample not only affects the overall variability of the dependent variable, but also influences the possibility of successfully sepa- rating supply and demand: that is, the factors behind pd from MAIN RESULTS 33 Poorest 20 percenta Small towns p pd pe p pd P, 0.85 0.73 1.08 0.58 0.67 0.80 (3.69) (3.63) (4.83) (6.17) (7.83) (8.99) -0.18 -0.06 -0.14 bb _b _b (3.13) (1.3) (2.51) -0.20 -0.16 -0.11 _ b _b _b (3.89) (3.52) (2.17) -0.44 -0.22 -0.44 -0.47 -0.33 -0.44 (8.67) (3.61) (8.91) (10.1) (5.66) (9.93) 0.01 -0.02 0.04 -0.06 -0.09 0.03 (0.19) (0.31) (0.53) (0.99) (1.84) (0.46) -0.04 0.07 -0.11 0.06 0.09 -0.01 (0.44) (0.72) (1.09) (1.73) (2.77) (0.26) 0.22 0.13 0.18 0.26 0.10 0.24 (3.06) (1.87) (2.55) (4.42) (1.85) (4.25) -0.02 0.03 -0.05 -0.09 -0.03 -0.06 (0.69) (1.07) (1.64) (3.32) (1.37) (2.51) 0.58 0.87 0.67 0.61 0.86 0.71 0.24 0.12 0.22 0.28 0.15 0.24 the factors behind P,. Households with services not only have access to the supply network, they also demand services, and this prevents the necessary variation between the data used to estimate P, and those used to estimate pd. Investment in roads and in agricitlture The sample survey attempted to measure the distribution of two large items of public expenditure in Colombia: investment in roads, wlhiclh amounts to 4.3 percent of the total public expendi- ture, and investment in agriculture, which amounts to 4.0 percent. 34 BACKGROUND AND MAIN RESULTS Table 1.12. Distribution of Subsidies and Beneficiaries of Government Services, 1974: Results from the 1974 Sample Survey (percentage) Households with service in 1974 and those who received it between 1970 and 1974 (A) Distribution of the subsidy Electricity Piped water Sewerage Income quintile (poorest to richest) Education Health 1974 A 1974 A 1974 A 1 19.8 19.5 13.1 26.0 13.9 31.4 10.7 24.5 2 20.2 20.6 16.1 25.1 16.3 23.6 14.9 21.3 3 18.6 24.2 19.2 25.6 19.4 26.8 18.5 29.1 4 20.1 19.8 23.0 12.0 22.5 12.6 23.7 12.0 5 21.3 15.9 28.6 11.3 27.9 5.6 32.2 13.1 Urban 78.5b 77.3 90.2 50.4 88.0 41.4 96.2 81.7 As of percentage of government expenditure- 11.2 3.6 7.6d 2.5d GNP 2.2 0.7 1.5 0.5 a. Not significantly different from zero. b. Refers to the subsidy received by urban households from primary and secondary schools plus the subsidy to universities. c. A figure of 62.7 billion pesos is used for government expenditure; see the text. d. Refers to investment in 1974. e. Refers to the grant component embodied in the new loans provided in 1974. f. Equal to the running cost of SENA in 1974. To capture the possible effect of public investment in roads, two variables were measured for both 1970 and 1974: the type of transportation used and the time of travel. They were derived from the following information: type of transportation of head of household to place of work; minutes it takes to walk from the dwelling to the nearest bus stop; for rural houselholds, type of transportation to the urban conglomerate of the municipality; and time it takes to get to the urban conglomerate of the municipality. No statistically significant change was observed in these vari- ables between 1970 and 1974. Either investment in roads had an effect as an intermediate input (in the transportation of goods and services) instead of as a final service to be consumed directly MAIN RESULTS 35 Houiseholds with service in 1974 and those who received it between 1970 and 1974 (A) Farm loans, Edzuca- Atten- Street Garbage Caja Agraria tional dance at lighting collection fellowships SENA (houise- (Sub- (house- (man- 1974 A 1974 A holds) sidy)' holds) months) 12.9 25.6 9.7 - 34.6 19.6 16.2 11.3 15.9 24.4 12.9 - 29.3 43.0 23.8 11.9 18.4 22.5 17.3 - 19.5 20.2 9.7 24.3 23.1 11.9 24.1 - 12.1 12.5 27.0 20.0 29.7 15.6 36.0 -a 4.5 4.7 23.3 32.5 96.0 100.0 100.0 100.0 0.5 1.0f 0.1 0.2 by households, or the technique used does not lend itself to mea- suring this type of public service. Three variables related to farming activities were measured in rural areas: changes in the sources of irrigation water between 1970 and 1974; extension services received from public agencies during 1974; and new farming loans received from public agencies in 1974. The survey was unable to measure statistically significant values for the first two variables. It did, however, measure new farming loans received by households from the Caja Agraria, a public credit institution. Other services Other services measured were garbage collection in urban areas; educational fellowships from public sources; and attendance at the Servicio Nacional de Aprendizaje (SENA), the major adult retrain- ing institution. The distribution of these services by income groups is presented in Table 1.12. 36 BACKGROUND AND MAIN RESULTS Services studied in relation to total government expenditure The results presented earlier are summarized in Table 1.12. The expenditure in the services studied accounts for 26.4 percent of total government expenditure (62.7 thousand million pesos) and 5.2 percent of Colombia's GNP in 1974 (323 thousand million pesos)."3 If the service of the debt is excluded from the definition of public expenditures, the above fraction increases to 33.2 per- cent of total public spending. The largest subsidy is to the educa- tional sector and accounts for 11.2 percent of government expendi- ture. Of this total, 4.7 corresponds to primary, 4.1 to secondary, and 2.4 to higher education. Appendix. Socioeconomic Stratification of Large Cities In 1970 the Colombian Bureau of Census (DANE) carried out a stratification of the neighborhoods of the four largest cities- Bogota, Cali, Medellin, and Barranquilla-by socioeconomic indicators. The indicators obtained from the 1970 household survey and from the agencies in charge of urbanization in each city are: average per capita income of the neighborlhood; construc- tion materials of the house; availability of public services and durable goods; index of crowding; and average educational level. 13. The figures for public expenditures are derived as follows: Consolidated Fiscal Budget for 1974 (million pesos) Current 45.500 Investment 34.014 Total 79.514 Minus: Income public companies -16.836 Total 62.678 Minus: Service public debt -12.987 Total 49.691 Source: Contraloria General de la Republica de Colombia, Informe Financiero de 1974 (processed) pp. 818, 819, 821. SOCIOECONOMIC STRATIFICATION OF CITIES 37 The relation between the six strata used by DANE and the four strata used in the 1974 sample survey is: 1974 sample survey DANE (1970) 1. Slums 1. Lowest income 2. Low income 2. Low income 3. Low middle 3. Middle 4. Middle 4. High 5. High middle 6. High Chapter 2 The Distribution of Income and Other Poverty Indicators NEW ESTIMATES OF the distribution of income in Colombia hiave been made based on the data collected in the 1974 sample survey. Other socioeconomic characteristics of households, which indicate the level of welfare, are reported also. Distribution of Income The 1974 survey recorded data on household income for the month before the interview. Because less than ten minutes of each interview was spent in deriving the income data, they may be less reliable than those usually obtained from budget and consumption surveys in which more time is given to obtaining this variable. The monthly household income figure was obtained by identifying the income earners in the family and by computing the sources of income for each. On the questionnaire five sources of income were defined: labor and wage income; income in kind received from the employer; value of the food grown and consumed on the site or plot; net profits from business or farm operations; and pension payments. The 1974 sample survey Urban, rural, and countrywide income distribution data were presented in Table 1.3, above. The data suggest that rural income is more evenly distributed than urban, particularly at the lower 38 DISTRIBUTION OF INCOME 39 Figure 2. 1. Distribution of Income, 1974: Rural and Urban Differences Percent of income 100 I I I I I I I I V 90 -…- - 80 - T - - T - - - - 7o -~~~~~~~~~~~ 70- Rr population 50 450 (G=0.42) _ 5~~~~~~ 40 - - Tolal - _ population ,_ 30 __(G=0.50) 10 ____ .- Urban population (G ~O.S54) X (C=0vI I _F 0 10 20 30 40 50 60 70 80 90 100 Percent of population levels. The poorest 20 percent of families in urban areas receive only 4.6 percent of the total urban income, whereas the poorest 20 percent of rural households receive 7.6 percent of the total rural income. These results still hold if the distribution of population rather than households is considered. This is illustrated in Figure 2.1, where the accumulated percentage of the population, ordered according to per capita income, is plotted against the accumu- lated income. The Gini coefficient is 0.54 for urban areas, 0.42 for rural areas, and 0.50 for the country as a whole. The effect of differences in rural and urban incomes is shown in Table 2.1, which classifies households in each quintile of the country income distribution according to their location. Rural households represent 53.4 percent of those in the lowest-income 40 THE DISTRIBUTION OF INCOME AND OTHER POVERTY INDICATORS Table 2.1. Percentage of Families in Each Income Quintile, Classified by Location Income quintile Inter- (poorest to Large mediate Small Urban Rural richest) cities cities towns total areas 1 13.0 10.8 22.8 46.6 53.4 2 18.2 13.5 18.6 50.3 49.7 3 25.5 16.1 14.5 56.1 43.9 4 34.1 22.5 13.1 69.7 30.3 5 54.2 24.7 8.3 87.2 12.8 Total 28.9 17.5 15.5 61.9 38.1 Note: Percentages add to 100 across rows. quintile and only 12.8 percent of those in the higlhest, wlhereas houselholds in large cities represent only 13 percent of those in the lowest-income quintile but 54.2 percent of those in the highest.1 In Chapter 1 it was shown hiow income data can also be used to compute the fraction of the population receiving a per capita income lower than a predetermined value such as a poverty line. Table 1.4 showed that the percentage of families in each region below commonly used poverty lines is hiiglhly sensitive to the ex- change rate used. Comparison with other studies Income distribution estimates are difficult to compare.2 Studies differ in the definition of income, reliability of measurement, representativeness, and type of population unit used. When the houselhold is the unit of measuremiient, the distribution of lhouse- holds or population according to houselhold per capita income is usually derived; wlhen the individual in the labor force is tlle unit of measurement, the distribution usually refers to active population or to income recipients. 1. A proper comparison obviously requires an adjustment for differences in the purchasing power of urban and rural nominal incomes. To the extent that this pur- chasing power is greater in rural areas, these figures exaggerate the percentage of the rural population in the poorest quintile of the total population. 2. See Shail Jain, Size Distribution of Income: A Compilation of Data (Baltimore: Johns Hopkins University Press for the World Bank, 1975). DISTRIBUTION OF INCOME 41 Perhaps the most comprehensive and reliable estimate of in- come distribution in Colombia is that of Berry and Urrutia (B-U).3 Their study is concerned with the economically active population and uses a variety of sources of data. The estimate of income distribution for the country combines an urban estimate based on unemployment surveys made between 1967 and 1969 by the Centro de Estudios del Desarrollo Econ6mico, Universidad de Los Andes (CEDE) and an estimate for rural areas based on the 1960 agri- cultural census and the 1964 population census.4 Table 2.2 presents a comparison of Berry and Urrutia's esti- mates and those arrived at in this study. Lorenz curves and Gini coefficients for total country estimates were shown in Figure 1.1. Coomparisons between the income distribution of households or population and that of the active population (B-U) must take into account the differences across incomiie groups in the number of earners per household as well as in the number of dependents per earner. If low-income groups have a lower number of earners per houselhold, the distribution of houselholds will show a greater inequality than the distribution of the active population. If there are a larger number of dependents per earner in low-income groups, the distribution of the total population will show a greater in- equality than the distribution of the active population. If poorer families have larger family size, a lower number of income earners per household means a larger number of dependents per earner. Consequently, if the distribution of households shows a greater inequality than the distribution of the active population, the same will be true for the distribution of the total population. The data in Table 2.2 and Figure 1.1 show different results. The distributions for households and individuals in the 1974 sam ple survey s how greater equality than the B-u estimates for the active population, particularly in the rural area. Gini coefficients The marked difference for the rural area (for B-U, the Gini coefficient is 0.57; for the sample survey, it is 0.42) is largely ex- 3. Albert Berry and Mliguel Urrutia, Income Distribution in Colombia (New Haven: Yale University Press, 1976). 4. Ibid., p. 28. 42 THE DISTRIBUTION OF INCOME AND OTHER POVERTY INDICATORS Table 2.2. Income Distribution Comparison (percentage) Urban Rural 1974 sample survey 1964 Berry-Urrutia 1974 sample survey A ctive House- Popula- popula- House- Popula- holds tion Income lion Income holds tion Income 10 12.5 1.7 12.7 0.6 10 13.4 3.0 20 24.7 4.6 25.0 2.5 20 24.9 7.6 30 36.7 8.6 30.3 4.5 30 36.6 13.8 40 46.9 13.3 41.0 9.8 40 47.7 20.9 50 56.8 18.9 51.1 15.7 50 59.2 29.8 60 66.1 25.5 60.0 21.7 60 68.6 38.6 70 75.5 34.2 70.2 30.2 70 76.7 47.5 80 84.4 45.7 79.3 40.1 80 85.5 59.4 90 92.6 62.3 89.8 56.3 90 93.5 73.5 95 98.5 73.9 96.0 72.9 95 97.0 81.9 plained by the different estimates of income for the upper 5 percent of households. In Berry and Urrutia's study, 5.1 percent of the richest active population received 40.4 percent of the rural income; in the 1974 sample survey, the richest 5 percent of houselholds received 18.1 percent of the rural income. Berry and Urrutia in- Table 2.3. Comparison of Gini Coefficients from Four Studies Year Study Urban Rural Country 1964 Berry-Urrutia 0.55 0.57 0.57 (economically active population) 1967 Rafael Prieto 0.47 - - (households in four largest cities) 1970 DANE-Polibio C6rdoba 0.55 - (income recipients) 1974 1974 sample survey Population 0.54 0.42 0.50 Households 0.48 0.32 0.47 a. Data reported only for urban category. Source: All figures, except those for the 1974 sample survey, were taken from Berry and Urrutia, Income Distribution in Colombia, Chapter 2. OTHER POVERTY INDICATORS 43 Rural Total coutntry 1964 Berry-Urrutia 1974 sample survey 1964 Berry-Urrutia A ctive House- Active population Income holds Popuilation Income population Income 11.1 1.8 10 13.0 1.9 8.8 0.54 18.7 4.1 20 25.1 5.2 25.1 4.0 30.1 8.5 30 36.5 9.4 - - 36.8 10.8 40 47.2 14.3 42.2 9.6 50.0 16.6 S0 57.7 20.4 52.5 14.5 61.5 22.7 60 66.9 26.9 60.9 19.6 73.4 30.4 70 75.7 34.9 72.1 28.6 80.5 36.5 80 84.4 45.6 80.0 36.9 90.2 49.4 90 92.5 60.7 89.1 50.6 94.9 59.6 95 96.4 72.8 95.6 66.3 clude in tlle rural distribution high-inicomiie landowners living in cities, whereas the 1974 samlple does not; in this respect, the two estimates are not directIly comparable. Table 2.3 gives Gini coefficient estimates from the B-U study as well as two others-DANE-Polibio C6rdoba (1970) and Prieto (1967)-and allows comparison witlh estimates from the 1974 sample survey for the distribution of population and houselholds, both ranked according to per capita houselhold income. The smaller coefficient for houselholds in the sample reflects the strong negative association between per capita income and family size, particu- larlv in rural areas. The coefficient for urban areas is practically the same wlhetlher the economically active population (B-U), income recipients (DANE-Polibio C6rdoba), or the total population (1974 sample survey) is used. For the country total, the 1974 survey estimate appears lower (0.50) than the B-U estimate (0.57). Other Poverty Indicators The 1974 samlple survey collected data on other variables that can be considered poverty indicators. Table 2.4 presents the dis- 44 THE DISTRIBUTION OF INCOME AND OTHER POVERTY INDICATORS Table 2.4. Distribution of Heads of Household in Each Quintile, by Years of Schooling (percentage of total in quintile) Income Years of schooling quintile (poorest 0 1 2 3 4 to richest) 1 31.4 6.7 20.6 15.0 8.6 2 27.7 6.3 18.9 16.3 9.8 3 22.7 4.1 17.3 15.4 8.6 4 19.2 3.2 9.7 10.5 8.6 5 8.6 1.5 5.4 4.3 5.4 Country average 22.0 4.4 14.4 12.2 8.2 Note: Percentages add to 100 across rows. tribution of heads of houselholds in each quintile according to years of schooling completed. The last row shows the distribution for the total country. A/lore than a fifth of the total heads of house- hold have no schooling, perhaps the best indicator of illiteracy. As expected, the amount of schooling is strongly associated with the income level of the head of houselhold: 31.4 percent in the poorest quintile have no schooling, whereas this is true of only 8.6 percent in the richest. In the two poorest quintiles, almost no heads of houselhold have completed secondary school (11 years). Only in the richest quintile have any heads of household completed the university (16 years). Data on means and standard deviations for other socioeconomic indicators, classified by location and income quintile, are presented in the statistical appendix, Tables SA-3 to SA-17. Information on sample size and some statistical formulas for identifying signifi- cance of test of the difference of two means are given in the appen- dix to this chapter. Table 2.5 presents means and standard devia- tion for some of these indicators, classified by quintiles in the distribution of income. The data tend to show that the number of dependents per earner (the ratio of column 1 over column 2) varies sharply across income groups, being substantially larger for low-income houselholds. This result, however, is higlhly sensitive to the definition of the "number of earners" variable, particularly in an environment where second earners seldom receive monetary incomes. A comparison between columns 1 and 3 suggests that OTHER POVERTY INDICATORS 45 Years of schooling 5 6-10 11 12-15 16 (primary) (secondary) (university) 12.8 4.2 0.3 0.4 0 14.3 5.9 0.2 0.6 0 18.1 11.5 1.2 1.1 0 22.7 18.5 3.8 3.3 0.5 15.2 23.6 15.5 9.2 11.3 16.6 12.8 4.2 2.8 2.4 numnber of persons per room is substantially higher for the lowest- income groups. The availability of toilets or latrines has an even stronger positive association witlh income. The occupational characteristics of the heads of household in the poorest quintile should be an important consideration for policy Table 2.5. Mean and Standard Deviations of Socioeconomic Variables of Households Mean number of I ncome Years of quintile Income Rooms Toilets or schooling Years of (poorest Persons in earners in occupied by latrines in of head of schooling to richest) household household household dwelling household of wife 1 6.87 1.24 2.51 0.59 2.29 1.67 (2.5) (0.8) (1.3) (0.6) (2.2) (1.9) 2 5.99 1.37 2.58 0.70 2.60 1.93 (2.5) (0.9) (1.3) (0.7) (2.4) (2.1) 3 5.38 1.52 2.74 0.86 3.22 2.47 (2.5) (1.0) (1.5) (0.7) (2.7) (2.8) 4 4.80 1.59 3.13 1.01 4.36 3.22 (2.4) (1.0) (1.7) (0.8) (3.5) (3.4) 5 4.25 1.72 4.40 1.81 7.84 5.34 (2.3) (1.0) (2.3) (1.2) (4.8) (5.0) Country 5.47 1.48 3.07 0.99 4.05 2.91 average (2.6) (0.9) (1.8) (0.9) (3.8) (3.5) Note: Standard deviations are given in parentheses. 46 THIE DISTRIBUTION OF INCOME AND OTHER POVERTY INDICATORS Table 2.6. Distribution of Ileads of Household in the Poorest Income Quintile According to Location, Sector, and Occupation Location and Percentage of heads occupation of houtsehold Urban areas Self-employed a. Manufacturing 5.2 16.0 b. Services 10.8 J Wage labor a. Manufacturing 4.9 b. Construction 5.5 18.6 c. Services 8.2 Rural areas a. Landowners 25.6 b. Sharecroppers 6.6 j c. Tenants and tenant farmers 3.9 18.6 d. Agricultural labor living 8.1 on farms e. Households not living on farms 14.0 Ill-defined occupations 7.2 intervention. Table 2.6 presents the percentage distribution of heads of houselhold in the poorest quintile, classified by location, sector, and occupation. The results are tentative, since the ques- tionnaire did not allow for a precise definitioni of borderline occupa- tions, wlhiclh are common in poor income groups. For example, a large fraction of those witlh ill-defined occupations (7.2 percent) were self-employed heads of houselhold workinig at home. A similar problem arose in classifying occupations in the rural area. All rural heads of houselhold classified in Table 2.6 under (a), (b), (c), and (d) live on farms and participate directly in farming activities; (d) was calculated residually, and it is not clear wlhetlher it can be interpreted as pure wage labor or farm labor. Similarly, although heads of household in (e) do not live on farms (rural areas include towns of less thani 1,500 inhabitants), they may be directly involved in farming activities. Most heads of houselhold in the poverty group live in the rural area, around 60 percent, depending on wlhether those with ill- defined occupations are included. The highest fraction-25 per- SOCIOECONONMIC CHARACTERISTICS OF HOUSEHOLDS 47 cent-is made up of landowners, that is, small farmers. Nonland- owners living on farms and directly involved in farming activities represent 18.6 percent. In rural areas, heads of household not living on farms represent 14.0 percent. In urban areas, heads of household are evenly classified between wage labor (mostly in the service sector) and self-employment. This implies that policies aimed at increasing the earning capacity of the poorest 20 percent of houselholds will require several different points of intervention. Appendix. Socioeconomic Characteristics of Households Data derived from the 1974 sample survey on means and stan- dard deviations of socioeconomic variables of households are pre- sented in the statistical appendix, Tables SA-3 to SA-17. The data are classified by quintiles in the income distribution (or house- hold per capita income) of each respective location or area: small towns, intermediate cities, large cities, urban average, rural areas, and country average. This appendix presents sample size and formulas to interpret the statistical significance of the difference between these means. Samnple size Table 2.7 shows the sample size corresponding to each (ex- panded) quintile. For a given location, the sample size differs across quintiles, because the mean expansion factor of each quintile differs, particularly for the large cities. This situation results from the fact that the low-income neighborlhoods in large cities pur- posely were overrepresented in the survey. For the same reason, the suIII of locational sample sizes for a given quintile does not necessarily add up to the aggregate sample sizes of each quintile. Statistical significance of the test of the difference of two means Table 2.8 indicates the degree of statistical significance of the differences of two means. When the variance of the two populations (a, and 0-2) are unknown and presumed unequal and when samples 48 THE DISTRIBUTION OF INCOME AND OTHER POVERTY INDICATORS Table 2.7. Sample Size in Each (Expanded) Quintile Number of households Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns total areas total 1 285 208 137 663 278 863 2 259 227 155 624 261 886 3 209 188 145 567 239 796 4 152 198 150 499 252 786 5 145 173 138 416 220 688 Total 1,050 994 725 2,769 1,250 4,019 Note: The quintiles are determined according to the expanded distribution of household per capita income in each region. The urban and country totals for a given income quintile are not necessarily arithmetic totals of the locational sample sizes because, for large cities, low-income neighborhoods were purposely overrepresented in the survey. Table 2.8. Critical Values of dMjn (5 Percent Significance Level) for the Test of Means under Alternative Combinations of Sample Sizes NJ 800 700 500 300 250 200 150 800 0.08 700 0.09 0.09 500 0.10 0.10 0.11 N2 300 0.12 0.14 250 0.15 0.15 200 0.16 0.16 0.17 150 0.17 0.18 0.19 0.20 are large, the test of means becomes: t~ X X- X2 aV/(S?IN/N) + (S22/N2) where Xi are the sample means, S2 the variance of the samples, and Ni the respective sample size. t has a "Student" distribution, but it can be referred to a table of normal probabilities for large sample sizes. SOCIOECONONIIC STRATIFICATION OF HOUSEHOLDS 49 The subscript 1 will be used for the sample witlh the larger stan- dard deviation of the two being compared: S2 = a S1, where a < 1 S22 = a2S12. Substituting into t, the result is: (X1-X2) S1 /Nl) + (a2/N2) Denoting d.1,r, = (X1 - X2)/Si as the difference of the means expressed as a fraction of the largest standard deviation (wlhich correspondingly becomes the minimiiunm or lowest of the two pos- sible ratios) generates: t (1/NV) + (a2/N2) = dmiri At a 5 percent significance level, rejection of the null hypothesis that the means come from the same population implies (for a positive Xl - X2): 1.645 ( ) ±(a2/N7) = < d,jon. Table 2.8 presents those critical values of 4ri. by assuming that a2 = 1: that is, both standard deviations are equal. If a2 is less than 1, the critical values of dMjn will represent a stronger test than the one required. Chapter 3 The Distribution of Public Subsidies for Education To ESTIMATE THE EDUCATION SUBSIDIES received by different income groups in the population, two sets of data are required: enrollment in each type of education from each income group and the subsidy per student year for each type of education. The first part of this chapter presents data derived from the 1974 sample survey on enrollment by income groups. The second presents data on public subsidies to the educational sector, pre- pared especially for this study by the Compafiia Colombiana de Datos (COLDATOS). The estimates of the preceding sections are then combined, and the distribution of subsidies across income groups are derived. Enrollment by Income Group Table 3.1 classifies students enrolled in each type of education according to quintiles in the country distribution of income. Be- cause approximately 97 percent of the total subsidy for education goes to public education, the enrollment in public schlools in each income group becomes the most important determinant of the distributive content of this subsidy. Table 3.1 shows that low- income quintiles have a larger share of enrollment in primary public schools, whereas the opposite is true in higher public edu- cation. Of the children enrolled in public primary education, 32.1 percent belong to the poorest quintile, whereas only 5.8 percent come from the richest. In higher public education 5.8 percent of 50 ENROLLMENT BY INCOME GROUP 51 Table 3.1. School Enrollment, Classified by Income Group, 1974 (percentage) Income Primary schools Secondary schools Universities quintile (poorest Pub- Pri- Aver- Pub- Pri- Aver- Pub- Pri- Aver- to richest) lic vate age lic vate age lic vate age 1 32.1 11.9 28.6 16.7 9.8 13.4 0.9 0 0.5 2 27.9 13.8 25.5 23.0 9.8 16.7 4.9 2.3 3.8 3 20.0 19.3 19.9 21.1 14.5 17.9 10.9 4.7 8.1 4 14.2 19.8 15.2 24.9 22.5 23.8 23.8 16.5 20.4 S 5.8 35.2 10.8 14.3 43.4 28.2 59.5 76.5 67.2 the students come from the poorest 40 percent of households, whereas 59.5 percent come from the richest quintile. The sharp skewness observed for primary and higher education does not hold for secondary public education. Enrollment by income group shows less variation and tends to favor the middle- income quintiles. Estimating the amount of subsidy received by each household requires information on the number of students per household and on the levels and types of schooling they attend. The 1974 sample survey provides these data for public schools, for private schools that received subsidies in 1974, and for private schools without subsidies. The data on enrollment from each household were adjusted by whatever discrepancy was found between the aggregate enrollment figures from the expanded sample survey and the figures reported by the Ministry of Education. Table 3.2 com- pares these aggregate figures and provides the corresponding adjust- ment factor; except in public higher education, this factor is not significant. The adjusted enrollment data are shown in Table 3.3. Tables 3.4 and 3.5 present the adjusted figures on enrollment per household. To allow, the discussion of the factors determining variations in enrollment, data on the number of children in each household are given in Table 3.6. Primary schools Table 3.4 slhows again the large difference across income groups in the enrollment per household in public primary schools. On 52 THE DISTRIBUTION OF PUtBLIC SUBSIDIES FOR EDUCATION Table 3.2. School Enrollment: Comparison between Sample Survey Figures and Ministry of Education Figures, 1974 (thousands of students) Primary schools Public Private Urban Rural Total Urban Rural Total Sample survey 1,698 1,195 2,893 565 31 596 Ministry of Education 1,919 1,188 3,107 528 32 560 Adjustment ratio 1.13 0.99 1.07 0.93 1.03 0.93 Source: Ministerio de Educaci6n, (processed). ICFES, "La Educaci6n en Cifras. 1970-74." The figures exclude "territorios nacionales." the average, households belonging to the poorest income quin- tile have 1.38 children enrolled in public primary schools, whereas the ones in the richest quintile have only 0.26. This is mainly because richer families have less children of school age (see Table 3.6) and because they tend to send their children to private schools, especially if they live in the urban areas. In the urban areas, fami- lies belonging to the richest quintile have 0.31 children in private primary schools, whereas only 0.26 are enrolled in public primary schools. Except for the highiest income group, enrollment per house- hold in public primary schools is larger in urban areas, even thougl these houselholds also have a large enrollment in private sclhools. The basic reason for this is the larger overall enrollmenit rate in urban areas. Secondary schools Tlhere is a great difference in the enrollment per houselhold in public secondary education between urban and rural areas, particularly in low-income groups. Rural households in the poorest 40 percent of houselholds have only 0.07 students enrolled in public schools and none in private schools. The difference appears even more significant wlhen the comparison is made for the total en- ENROLLMENT BY INCOME GROUP 53 Secondary schools Universities Public Private Public Private Total Urban Rural Total Urban Rural Total 588 133 721 629 37 666 90 74 164 0 0 641 0 0 624 73 76 149 0 0 0.89 0 0 0.94 0.81 1.02 0.91 rollment (in all schools) per household, which is four to five times larger for urban households. Enrollment per household in public secondary schools tends to be highest for the second, third, and fourth quintiles: that is, for the middle classes. The reasons for a lower enrollment in the Table 3.3 Adjusted School Enrollment, Classified by Location of Households, 1974 (thousands of students) Primary schools Secondary schools Universitiesa Private Private With Without With Without Location Public subsidies subsidies Public subsidies subsidies Public Private Large cities 746 73 219 192 90 268 - - Intermediate cities 556 41 134 181 51 111 - - Small towns 617 13 48 150 35 42 - - Urban total 1,919 127 401 523 176 421 - Rural area 1,188 5 27 118 0 27 - - Country total 3,107 132 428 641 176 448 73 76 a. Results for universities are only given as country totals since the data were not broken down by location. 54 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.4. AMean Enrollment per Ilousehold in Different Types of Primary Schools, Classified by Quintiles (number of students) Income quintile Large cities Intermediate cities Small towns (poorest to richest) T1 T2 T3 Total T1 T2 T3 Total T, T2 T3 Total 1 1.49 0.07 0.07 1.63 1.68 0.05 0.19 1.92 1.37 0 0.15 1.52 2 1.30 0.10 0.17 1.57 1.34 0.06 0.20 1.60 1.43 0 0.06 1.40 3 0.79 0.08 0.26 1.13 1.14 0.05 0.20 1.39 1.24 0.06 0.05 1.35 4 0.78 0.06 0.18 1.02 0.70 0.06 0.20 0.96 0.61 0.03 0.02 0.66 5 0.25 0.06 0.25 0.56 0.26 0.09 0.25 0.60 0.34 0.05 0.14 0.53 Country average 0.71 0.07 0.20 0.98 0.88 0.07 0.21 1.16 1.10 0.02 0.08 1.20 N'ote: Ti -public schools; T2 = private schools with subsidies; Ta = private schools without subsidies. Table 3.5. Mlean Enrollment per Ilousehold in Different Types of Secondary Schools, Classified by Quintiles (number of students) Income qointile Large cities Intermediate cities Small towns (poorest to richest) Ts T2 T3 Total T1 T2 T3 Total T1 T2 T3 Total 1 0.19 0.10 0.13 0.42 0.30 0.05 0.08 0.43 0.24 0.09 0.09 0.42 2 0.31 0.04 0.15 0.50 0.41 0.07 0.08 0.56 0.28 0.09 0.05 0.42 3 0.28 0.01 0.15 0.50 0.28 0.08 0.16 0.52 0.20 0.09 0.12 0.41 4 0.20 0.08 0.25 0.53 0.32 0.07 0.17 0.56 0.33 0.05 0.07 0.45 5 0.08 0.11 0.37 0.56 0.18 0.12 0.28 0.58 0.36 0.05 0.19 0.60 Country average 0.18 0.09 0.25 0.52 0.29 0.08 0.27 0.54 0.27 0.06 0.08 0.43 Note: Ti = public schools; T2 = private schools with subsidies; Ts = private schools without subsidies. riclbest and poorest quintiles can be understood by comparing Tables 3.5 and 3.6. In the poorest quintile, the low enrollmiient in public sclbools reflects the extremely low total scliool enrollment relative to tbe number of clhildren ages 13 to 19. Of the 1.40 clhil- dren per household between 13 and 19 years old, there are 0.24 clhildren enrolled-a rate of 17 percent. In tbe riclhest quintile, altlhouglh the total enrollnment rate is much biglber (on the order of 62 percent or 0.50 clhildren out of 0.80 betweeni 13 and 19 years ENROLLMENT BY INCOME GROUP 55 Urban average Rural areas Country average T1 T2 Ta Total Tr T2 Ts Total T, T2 T3 Total 1.48 0.04 0.16 1.68 1.26 0 0.02 1.28 1.38 0.02 0.08 1.48 1.33 0.05 0.13 1.51 0.98 0 0.01 0.99 1.15 0.03 0.08 1.26 1.00 0.06 0.19 1.25 0.72 0.01 0.02 0.75 0.88 0.04 0.11 1.03 0.72 0.05 0.15 0.92 0.41 0 0.03 0.44 0.62 0.04 0.11 0.77 0.26 0.07 0.24 0.57 0.27 0 0.03 0.30 0.26 0.06 0.21 0.53 0.86 0.06 0.18 1.10 0.86 0 0.02 0.88 0.86 0.04 0.12 1.02 Urban average Rural areas Country average T, T2 T3 Total T1 T2 Ts Total T1 T2 Ts Total 0.24 0.09 0.10 0.43 0.07 0 0.02 0.09 0.15 0.04 0.05 0.24 0.33 0.05 0.09 0.47 0.07 0 0.03 0.10 0.20 0.02 0.06 0.28 0.26 0.08 0.14 0.48 0.10 0 0.01 0.11 0.19 0.04 0.08 0.31 0.27 0.07 0.19 0.53 0.12 0 0.02 0.14 0.22 0.05 0.15 0.42 0.13 0.11 0.33 0.57 0.09 0 0.02 0.11 0.13 0.09 0.28 0.50 0.23 0.08 0.19 0.50 0.09 0 0.02 0.12 0.18 0.05 0.12 0.35 old), enrollment in public sclhools is only about a third of that in private sclhools, 0.13 and 0.37, respectively. Universities Enrollment per household in universities, whiclh is highly corre- lated with the level of household per capita income, is shown in Table 3.7. For the middle-income quintiles enrollment per house- hold in public universities is approximately twice as high as in 56 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.6. Number of Children per Household, Classified by Location, Age, and Income Quintile Income Large cities Intermediate cities quintile (poorest All All to richest) ages 6-12 13-19 ages 6-12 13-19 1 6.33 1.83 1.35 6.83 1.87 1.41 2 6.11 1.61 1.34 6.09 1.40 1.46 3 5.36 1.21 1.04 5.98 1.45 1.24 4 5.12 0.96 1.04 4.96 0.91 1.00 5 4.38 0.63 0.81 4.40 0.68 0.82 Country average 5.12 1.05 1.03 5.40 1.13 1.12 Table 3.7. Mean Enrollment per Household in Universities, Classified by Quintiles (number of students) Income quintile (poorest to richest) Public Private Total 1 0.001 0 0.001 2 0.005 0.002 0.007 3 0.011 0.005 0.016 4 0.024 0.017 0.042 5 0.061 0.081 0.142 Country average 0.020 0.021 0.041 private universities. The reverse is true for the richest quintile, where household enrollment is 0.061 in public universities and 0.081 in private universities. Public Subsidy per Student Year The concept of subsidy as it is used here includes all public contributions-whether from the central government, departnment, PUBLIC SUBSIDY PER STUDENT YEAR 57 Small towns Rural areas Country average All All All ages 6-12 13-19 ages 6-12 13-19 ages 6-12 13-19 6.93 2.02 1.44 6.99 2.14 1.42 6.87 2.05 1.40 5.93 1.52 1.21 5.95 1.47 1.24 5.99 1.50 1.28 5.60 1.27 1.20 5.12 1.05 1.06 5.38 1.19 1.10 4.53 0.57 1.00 4.41 0.68 0.81 4.80 0.81 0.96 4.32 0.59 0.90 3.39 0.35 0.60 4.25 0.59 0.80 5.76 1.37 1.20 5.63 1.37 1.14 5.50 1.23 1.11 or city-to the yearly operating cost of schools.1 For public schools the subsidv is basically equal to the labor cost (teachers' salaries). The estimate does not include the opportunity cost of the stock of physical capital invested in education: that is, depreciation and interest on public buildings. To compute the subsidy per student, the t974 sample survey classification for schools was used: public schools, where prac- tically all operating costs are financed out of public funds, and private schools, wlhiclh receive some public funding. Public primary schools To capture not only urban-rural differences but also intraurban and intrarural variations, the subsidy was broken down into the forty-eight geographic strata of the sample survey. This dictated the choice of both method and data used to derive the subsidy per student year. The operating cost of public primary schools is financed by the central government and by the department and municipality where each school is located. T he institutional meclhanisni througlh 1. Estimates given in this chapter are based on the COLDATOS report. Compaftia Colombiana de Datos (COLDATOS), "Unit Cost of Education and Health Services in Colombia in 1974" ("Costos Unitarios de los Servicios de Educaci6n y Salud en Colombia en 1974") (study prepared for the World Bank, Bogota, 1976; processed). 58 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION which these funds are allocated to specific schools is the Regional Educational Fund (FER). The Instituto Colombiano de Pedagogia (ICOLPE) has reported data for 1973 on the budgeted contributions from all sources for each of the twenty-three departments in the country.2 Actual expenditure for that year was derived on the basis of the historical evolution of budgeted and actual contribu- tions over the previous six years. The coefficient of proportionality for each department was estimated through regression analysis and was applied to the budgeted figure for 1973 (Table 3.8). The next step was to break down the contribution by depart- ment to figures for each of the forty-eight regional strata used in the sample survey. In 1972 the Bureau of Census (DANE) surveyed a sample of 17,000 schools to gather information on the number of teachers and their classification according to four wage categories.3 With these data it was possible to derive the distribution of teachers in each stratum according to wage categories. By applying this percentage distribution to the total by depart- ment,4 it was possible to arrive at the number of teachers in each wage category employed in each stratum. This information, to- gether with data on teachers' wages, provides an estimate of the total cost of teachers for each stratum.5'5 The difference between the cost of teachers for each department and the total public contribution was distributed across the different strata in propor- tion to the number of students in each stratum. Table 3.9 shows the total public contribution for each stratum as estimated for 1973 and the number of students and subsidy per student for 1973. The subsidy per student for 1974 was ob- tained by multiplying the 1973 figure by 1.20 to allow for the 20 percent average increase in teachers' salaries between these years. Public secondary schools The subsidy per student in public secondary schools was derived with the help of a sample survey of 574 schools under- 2. COLDATOS report. 3. DANE, "Investigaci6n sobre Establecimientos Educativos," 1972 (processed). 4. Ministerio de Educaci6n, "Estadisticas de la Educaci6n Primaria Oficial" (processed). 5. This wage includes the Christmas bonus, which appears in the personal services item of the department budget. 6. The data referred to are given in the statistical appendix, Tables SA-18 and SA-19. PUBLIC SUBSIDY PER STUDENT YEAR 59 Table 3.8. Public Contributions from All Sources to Public Primary Schools, 1973 (thousands of pesos) Actual Department Budgeted (estimated) Antioquia 278,157 357,754 Atlantico 78,608 78,608 Bogota, D. E. 300,108 349,838 Bolivar 65,537 68,707 Boyach 107,873 111,109 Caldas 98,895 98,895 Cauca 56,978 66,095 Cesar 33,999 33,999 Choco 25,055 25,055 C6rdoba 64,687 64,687 Cundinamarca 178,959 187,907 Huila 45,364 52,169 La Guajira 17,115 25,959 Magdalena 64,929 71,421 Meta 26,516 26,556 Narino 65,656 65,656 N. Santander 77,577 77,577 Quindio 37,196 47,983 Risaralda 40,330 44,767 Santander 110,692 132,830 Sucre 33,841 43,656 Tolima 97,897 108,665 Valle del Cauca 160,851 165,651 Country total 2,066,824 2,304,544 Source: "Budgeted" data from COLDATOS report, p. 22. taken by ICOLPE in 1972.' This survey provided data on the fund- ing of each scliool as well as enrollment figures for 1972. In deriving the 1974 figure, the subsidy per student of 1972 was multiplied by 1.25 to allow for the 25 percent increase in teaclhers' salaries between both years. Table 3.10 gives the estimates of the subsidy per student in secondary schools in urban areas, classified by city size and by 7. ICOLPE, Costos de la Educaci6n Media Oficial" 1972 (processed). 60 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.9. Public Primary Education: Estimated Enrollment, Total Subsidy, and Subsidy per Student, Classified by Stratum, 1973 and 1974 Total subsidy, Subsidy per student 1973 (thousands 1973 1974 Stratum Enrollment of pesos) (pesos) (pesos) Bogota 288,245 348,125 1,207 1,449 Cali 102,826 56,711 551 662 Medellin 139,105 100,292 720 865 Barranquilla 62,335 48,240 773 928 5 47,741 33,027 691 830 6 45,227 37,249 823 988 7 59,922 52,691 879 1,055 8 42,090 27,540 654 785 9 75,193 49,903 663 796 10 34,022 24,231 712 855 11 123,830 98,824 767 920 12 47,521 24,581 517 620 13 33,976 26,047 548 658 14 18,626 9,107 488 586 15 28,731 21,780 758 909 16 25,217 21,597 856 1,028 17 48,918 31,549 644 774 18 41,241 31,249 757 909 19 65,832 61,291 931 1,117 20 41,034 34,487 840 1,009 21 28,877 27,398 948 1,038 22 87,878 70,522 802 963 23 52,620 42,011 798 958 24 48,489 35,128 724 869 25 49,371 36,861 746 895 26 39,352 35,086 891 1,070 27 53,789 40,937 761 913 28 65,592 40,038 610 732 29 82,638 46,268 559 671 30 65,190 48,614 745 895 31 76,311 56,179 736 883 32 87,266 69,313 794 953 33 61,898 63,619 1,027 1,233 34 79,238 62,952 794 953 35 57,097 42,820 749 900 36 - - PUBLIC SUBSIDY PER STUDENT YEAR 61 Table 3.9. (Continued) Total subsidy, Subsidy per student 1973 (thousands 1973 1974 Stratum Enrollment of pesos) (pesos) (pesos) 37 27,125 23,418 863 1,036 38 49,234 40,104 814 977 39 46,561 37,221 799 959 40 60,533 49,784 822 987 41 21,400 15,977 746 895 42 87,900 64,246 730 877 43 154,002 101,115 656 788 44 3,798 1,918 505 606 45 32,889 22,356 679 815 46 54,551 33,845 620 774 47 52,626 40,977 778 934 48 36,052 17,316 480 576 Country total 2,938,907 2,304,544 784 941 a. No schools outside the cabecera. Source: COLDAIOS report, p. 23-24. Table 3.10. Public Secondary Education: Estimated Subsidy per Student, Classified by Types of Schools, 1974 (pesos) Type of school Bachil- lerato Vocacional Normalista Comercial Average Large cities 3,725 5,031 6,369 3,451 3,773 (97.5) (0.5) (1.0) (1.0) (100.0) Intermediate cities 3,744 5,067 5,020 2,631 3,760 (95.0) (0.5) (2.5) (2.0) (100.0) Small towns 3,795 7,805 4,755 2,988 3,908 (83.8) (0) (13.8) (2.4) (100.0) Note: Figures in parentheses are percentages of enrollment in each type of public secondary school. 62 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.11. Private Primary and Secondary Schools: Enrollment, Total Subsidy, and Subsidy per Student, 1974 Primary schools Secondary schools Enroll- Total Subsidy Enroll- Total Subsidy mentl subsidyb per ment' subsidyb per (thou- (millions student (thou- (millions student sands) of pesos) (pesos) sands) of pesos) (pesos) Large cities 73 4.3 58 90 56.5 628 Intermediate cities 41 4.8 117 51 35.6 698 Small towns 13 3.0 230 35 23.5 671 Rural areas 5 1.4 280 0 0 0 Country total 132 13.5 102 176 115.6 657 a. 1974 sample survey figures adjusted by the Ministry of Education totals. These figures refer only to enrollment in subsidized private schools. b. National, departmental, and municipal budgets. different types of schools. The average figure is the weighted average of the subsidy to different types of secondary schools. Private primary and secondary schools The item, "transfers," in the budget of the Ministry of Educa- tion presents data on individual transfers to private primary and secondary schools, as well as the location of each school. That item, together with the DANE directory of schools, provides the information necessary to classify private schools according to whether or not they receive subsidies. These figures from the Ministry of Education plus figures on transfers from departmental and municipal sources were used to derive the total subsidy figures in Table 3.11. The enrollment data, taken from the 1974 sample survey, refer only to enrollment in private schools that do receive subsidies. Private and public universities Subsidies to higher education were directly computed for 1974 in the COLDATOS report without a breakdown by regions. This breakdown becomes difficult when the location of the student's DISTRIBUTION OF SUBSIDIES ACROSS INCOME GROUPS 63 Table 3.12. Universities: Enrollment, Total Subsidy, and Subsidy per Student, 1974 Total subsidy Subsidy per Enrollment (millions of student University (thousands) pesos) (pesos) Public 73 1,433.8 19,641 Private 76 58.9 775 Countrytotal 149 1,492.7 10,018 Source: Data from COLDATOS report. household differs from that of his school-a typical situation in higher education. Table 3.12 presents figures for enrollment, total subsidy, and subsidy per student for public and private universities. Private universities include all nonpublic institutions, wlhether or not they receive subsidies. Distribution of Subsidies across Income Groups Table 3.13 summarizes the estimates of the subsidy per student derived earlier. The subsidy for each stratum for public primary schools was consolidated into broader categories since no logical intraregional breakdown fitted the variability of the subsidy figure. Because of the method used, the estimated figures for primary and secondary public education shown in Table 3.13 do not neces- sarily correspond to the official totals. Fortunately, the difference is not significant.8 8. The differences between the estimated and official figures are shown below (in millions of pesos). Estimated Official Public primary 2,934.3 3,104.5 Public secondary 2,444.6 2,284.7 Official data are from the Ministerio de Educaci6n, "Oficina Coordinadora de los FER y Oficina de Planeamiento de la Educaci6n," (processed); Ministerio de Educaci6n, '"Ejecuci6n Presupuestal" (processed). 64 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.13. Enrollment, Total Subsidy, and Estimated Subsidy per Student, 1974 Enrollment Total subsidy Subsidy per (thousands) (millions of pesos) student (pesos) Educa- Location tional level Public Private' Public Privatea Public Privates Large primary 746 73 819.8 4.3 1,099 58 cities secondary 192 90 716.7 56.5 3,733 628 Inter- mediate primary 556 41 480.9 4.8 865 117 cities secondary 181 51 680.6 35.6 3,760 698 Small primary 617 13 592.9 3.0 961 230 towns secondary 150 35 586.2 23.5 3,908 671 Urban primary 1,919 127 1,893.6 12.1 987 94 total secondary 523 176 1,983.5 115.6 3,793 657 Rural primary 1,188 5 1,040.7 1.4 876 280 total secondary 118 0 461.1 0 3,908b 0 Country primary 3,107 132 2,934.3 13.5 944e 102 total secondary 641 176 2,444.6 115.6 3,814 657 university 73 76 1,433.8 58.9 19,641 775 Subtotal 3,821 384 6,812.7 188.0 Grand total 4,205 7,000.7 a. Except in higher education, all private schools considered here are subsidized. b. Same figure as for small towns. c. This mean figure is different from the 941 figure arrived at in Table 3.9, given that the 1974 urban-rural weight is being used. In public education, which accounts for 97 percent of the total subsidy to education, differences in the subsidy per student are not significant across regions-at least not at the level of approxi- mation at which these figures must be interpreted. In the case of public primary education, large cities have the highest subsidy, 1,099 pesos; for urban areas the figure is 987 pesos; and for rural areas 876-a 13 percent difference. In public education, the largest variation in the subsidy is across educational levels. The subsidy per student in secondary education is approximately four times that in primary education; the subsidy in higher education is twenty times that in primary and five tinmes that in secondary. In private primary schools the subsidy varies according to the DISTRIBUTION OF SUBSIDIES ACROSS INCOME GROUPS 65 location of the school, becoming larger the smaller the size of the city. Since enrollment in these schools is low (approximately 4 percent of the total enrollment in primary schools), however, this variation is not significant in determining the total distributive effect of the subsidy. Subsidies per household Table 3.14 presents figures on the subsidy per household, ac- cording to location and income quintile. The mean annual subsidy per household in the country is 1,947 pesos or $70.50. The subsidy for primary education becomes substantially larger for the lower-income groups. For higher education the subsidy is much greater for the higher-income groups. The net result is a more or less constant subsidy per household across quintiles. In other words, the highly distributive subsidy to primary education is almost exactly compensated for by the regressive subsidy to higher education. The subsidy to secondary education is slightly higher for the middle-income quintiles. Rural households receive a substantially smaller subsidy from primary and secondary education. The difference becomes more marked in the lower quintiles; rural households in the poorest 40 percent of households receive half the subsidy of urban house- holds in the same group because of the low enrollment rate in low-income families in rural areas. Within urban areas, the mean subsidy per household is lower for large cities than for small ones because of the relatively greater enrollment in unsubsidized private schools in the larger cities. Subsidy per capita The constancy of the subsidy across quintiles, when defined per household, disappears when expressed in per capita terms: that is, the subsidy per household divided by the number of individuals in the family. This comparison is shown in Table 3.15. Since the subsidy per household is more or less constant, the variation in the subsidy per capita follows the variation in family size across income quintiles. The subsidy per capita is almost 75 percent larger for households in the richest quintile than for those in the poorest. 66 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.14. Education Subsidy per Household, 1974 (pesos) Income Large cities Intermediate cities Small towns quintile (poorest Pri- Sec- Pri- Sec- Pri- Sec- to richest) mary ondary Total mary ondary Total mary ondary Total 1 1,642 772 2,414 1,459 1,163 2,622 1,317 998 2,315 2 1,435 1,182 2,617 1,166 1,590 2,756 1,288 1,154 2,442 3 873 1,089 1,962 992 1,109 2,101 1,205 842 2,047 4 861 797 1,658 613 1,252 1,865 593 1,323 1,916 5 278 368 646 235 760 995 338 1,440 1,778 Country average 784 729 1,513 769 1,146 1,915 1,062 1,095 2,157 Note: To obtain dollar figures, divide by 27.6. Subsidy as a percentage of income Tables 3.16 and 3.17 show the subsidy per household expressed as a percentage of the annual household income reported in the 1974 sample survey. Table 3.16 shows that the typical Colombian household receives a subsidy from the educational sector equal to 5 percent of its annual income. For the poorest quintile, the sub- sidy is equivalent to 18.4 percent, whereas for the highest it is only 1.9 percent. Table 3.17 shows that even if primary and sec- ondary education only are included, urban households receive almost twice the percentage subsidy received by rural households Table 3.15. Education Subsidy Figures per Household and per Capita Subsidy per Subsidy per Income quintile household Family capita (poorest to richest) (pesos) size (pesos) 1 1,921 6.87 280 2 1,961 5.99 327 3 1,810 5.38 336 4 1,950 4.80 406 5 2,064 4.25 486 Country average 1,947 5.50 354 DISTRIBUTION OF SUBSIDIES ACROSS INCOME GROUPS 67 Urban areas Rural areas Country average Pri- Sec- Pri- Sec- Pri- Sec- Uni- mary ondary Total mary ondary Total mary ondary versity Total 1,464 969 2,433 1,103 273 1,376 1,305 598 18 1,921 1,317 1,284 2,601 858 273 1,131 1,089 776 96 1,961 993 1,039 2,032 633 391 1,024 835 751 224 1,810 715 1,070 1,785 359 469 828 589 872 489 1,950 263 565 828 237 252 589 252 555 1,257 2,064 854 925 1,779 754 352 1,106 816 718 413 1,947 in the same quintile. Although the subsidy is larger for urban households in the same quintile, the mean is larger for the rural area (4.8 percent) than for the urban area (3.7 percent). This is because more rural households are in the lower-income quintiles, where percentage subsidies are largest. The reliability of these percentage figures is proportional to the reliability of the income figures reported by households in the sample survey: that is, less reliable for rural areas and for higher- income groups. The sample survey yielded an annual household income of 38,904 pesos, equal to a dollar per capita income of Table 3.16. Total Education Subsidy per Household as a Percentage of Reported Household Annual Income Reported mean annual Subsidies (percentage) Income quintile income (poorest to richest) (pesos) Primary Secondary University Total 1 10,368 12.6 5.8 - 18.4 2 17,820 6.1 4.4 0.5 11.0 3 25,032 3.3 3.0 0.9 7.2 4 36,912 1.6 2.4 1.3 5.3 5 104,388 0.2 0.5 1.2 1.9 Country average 38,904 2.1 1.8 1.1 5.0 68 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.17. Primary and Secondary Education Subsidy per Household as a Percentage of Hlousehold Annual Income, by Urban and Rural Regions, 1974 Urban Rural Reported Number Reported Number mean of mean of Income quintile annual families Subsidy annual families Subsidy (poorest income (per- (per- income (per- (per- to richest) (pesos) centage) centage) (pesos) centage) centage) 1 10,272 15.0 23.7 10,452 27.9 13.2 2 18,024 16.8 14.4 17,604 26.9 6.4 3 26,052 17.8 7.8 23,736 22.7 4.3 4 38,712 22.5 4.6 32,808 15.8 2.5 5 108,768 27.9 0.8 74,484 6.7 0.8 Country average 48,262 100.0 3.7 23,214 100.0 4.8 $256;9 the national accounts figures show a GNP per capita for 1974 twice as large. Althouglh there are conceptual reasons why the houselhold survey would yield lower-income figures, the size of the difference suggests that the reported income figures repre- sent an underestimate. This qualification must be considered in interpreting the subsidy expressed as a percentage of houselhold incomee. Distribution of subsidies and distributtion of income Table 3.18 compares the distribution of the subsidy witlh the distribution of income. The total subsidy to education represents 11.2 percent of total governmenit expenditure and 2.2 percent of the GNP. Primary education accounts for 42.1 percent of the total subsidy-almost twice the subsidy to higlher education. Figure 3.1 presents Lorenz curves for the distribution of the subsidy compared with the distribution of income. The lhorizontal axis slhows the accumulated percentage of househiolds, ordered according to hiousehold per capita incolle; the vertical axis, the accumulated percentages of subsidy and total incoimie. The Lorenz curve for the total subsidy almost coincides with the 45 degree 9. With a mean family size of 5.5 and an exchange rate of 27.6 pesos to the dollar (1974). DISTRIBUTION OF SUBSIDIES ACROSS INCOME GROUPS 69 Figure 3.1. Distribution of Income and of Subsidies for Education Percent of income/ Percent of subsidy 100 90- 80 JPrimary education 7 _ / 60 __ _ 7 450 76°C-=- / - / caldcation - 54 = _ / Personal income 3C-- 2°_ / 7 1/te7 _7-- 2C tO C Unsety educatin 0 10 20 30 40 50 60 70 80 90 100 Percent of households diagonal; it is clearly more evenly distributed than personal income. The line above the diagonal, representing the distribution of the subsidy to primary education, shows a distribution in favor of lower-income households: lower-income families receive a higher share of the subsidy than the percentage they represent in the total number of families. The subsidy to higher education, however, not only favors higher-income households (the line is below the diagonal), but it also shows a stronger inequality than the distribu- tion of personal income: lower-income houselholds receive a smalier share of that subsidy than of the total income. When the Lorenz lines are plotted for the percentage of indi- viduals (instead of for the percentage of households), all distribu- 70 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.18. Distribution of the Subsidy to Education and Distribution of Personal Income, 1974 (percentage) Subsidy for education Income quintile House- Indi- Sec- Univer- (poorest to richest) holds viduals Primary ondary sity Average Incomea 1 20 25.1 32.1 16.8 0.8 19.8 5.2 2 20 22.7 26.7 21.8 4.6 20.2 9.1 3 20 19.4 20.5 21.2 10.7 18.6 12.6 4 20 17.4 14.5 24.6 23.5 20.1 18.7 5 20 15.4 6.2 15.6 60.4 21.3 54.4 As a percentage of: Total education subsidy 42.1 36.6 21.3 100.0 Total government expenditureb 4.7 4.1 2.4 11.2 GNP' 0.9 0.8 0.5 2.2 a. Income distribution from the 1974 sample survey. b. Corresponds to direct expenditures by the central government plus contributions to decen- tralized agencies. The total is 62.7 billion pesos. Contraloria General de la Rep6iblica de Colombia. "Informe Financiero de 1974," p. 818. c. The GNP equals 322.6 billion pesos (International Financial Statistics). tion lines show a stronger inequality. In this particular case, the relation of the lines to each other remains the same. Plotted for the population, the total subsidy line lies below the 45 degree diagonal; 25.1 percent of the poorest population receives 19.8 percent of the subsidy, whereas the richest 15.4 percent of the population receives 21.3 percent. These findings are compared witl results from earlier studies in the following appendix. Appendix. Comparisons with Other Studies Several studies on the distributive effect of government expendi- ture, particularly in education, have been made in Colombia. It is interesting to compare the earlier findings with those presented in this chapter. Urrutia and Sandoval In their pioneer work on this subject, Miguel Urrutia and Clara Elsa de Sandoval attempted to derive income distribution data COMtPARISONS WITH OTHER STUDIES 71 corrected by taxation and by the effect of government expendi- ture.'" Tlhe study was undertakeen for 1966, using 1964 income distribution data. The authors tried to allocate the public subsidy to education among income groups. In the absence of direct survev data on school enrollment, Urrutia and Sandoval had to decide on a metlhod of allocating enrollmiient according to income group. For tlis purpose, all chil- dren of primary school age (7 to 11) and secondary school age (12 to 17) were classified into three groups: those not enrolled; those enrolled in private schools; and those enrolled in public schools. The childreni not enrolled were allocated to the poorest- income deciles (tenths) ; those in private schools to the richest deciles; and those in public scllools to the remlaining deciles. The expenditure in higlher education was distributed by classify- ing students at the National University according to their parents' level of schooling" and by using the 1964 census to classify the male population 40 to 59 years old bv level of schooling. Assuming a correspondence between the distribution of individuals in that age bracket classified by schoolinig and the population classified bv per capita income (that is, if x percent of the individuals in that age bracket have 0 to 2 years of schooling, it means that they belong to hlouselholds representing the bottomii x percent of individuals classified according to per capita income), it was possible to classify the enrollmenit in universities according to income groups. Table 3.19 shows the resulting distribution of subsidies accord- ing to quintiles defined by the distributioni of the population by per capita incomiie. The figures in parentheses are figures from the present study expressed for quintiles of houselholds, not popula- tion, in the distribution of personal income. Because the poorest and richest quintiles of households account for 25.1 and 15.4 percent of the population, respectively, the figures in parentheses for these quintiles must be multiplied by 0.8 and 1.3, respectively, for comparison with those in the Urrutia-Sandoval study.'2 The 1974 survey shows a substantially more progressive sub- 10. Miguel Urrutia and Clara E. de Sandoval, "Politica Fiscal y Distribuci6n del Ingreso en Colombia," Revista Banco de la Republica, July 1974. 11. German Rama, "Origen Social de la Poblaci6n Universitaria,' Universidad Nacional, August 1969. 12. The correction factor for the middle-income quintiles approaches 1. 72 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.19. Distribution of Education Subsidies, 1966, Urrutia-Sandoval Subsidies for education Income quintiles Pri- Sec- Uni- Sec- Univer- (poorest mary ondary versity Total Primary ondary sity Average to richest) (millions of pesos, 1966) (percentage) 1 145 0 3 148 15.7 0 1.1 9.9 (32.1) (16.8) (0.8) (19.8) 2 170 0 10 180 18.5 0 3.8 12.1 (26.7) (21.8) (4.6) (20.2) 3 194 0 16 210 21.1 0 6.0 14.1 (20.5) (21.2) (10.7) (18.6) 4 267 144 16 427 29.0 47.5 6.0 28.7 (14.5) (24.6) (23.5) (20.1) 5 145 159 221 525 15.7 52.5 83.1 35.2 (6.2) (15.6) (60.4) (21.3) Country total 921 303 266 1,490 100.0 100.0 100.0 100.0 Note: Values in parentheses refer to relative comparable figures from the 1974 survey. Source: Miguel Urrutia and Clara E. de Sandoval, "Politica Fiscal y Distribuci6n del Ingreso en Colombia." sidy than the 1966 results. Two factors are behind this difference: a much higher enrollment rate in primary education for the poor- est quintiles and a much larger enrollment rate in secondary education, whiclh the 1966 study assumed to be nil for the poorest 60 percent. Part of the difference arises from the assumptions used in the 1966 study, where all nonenrolled children were allo- cated to the poorest-income groups; the implicit relation between the enrollment rate and the income level resulting from this as- sumption has an exaggerated slope and, for primary education, a discontinuity. The second source of difference is that the enroll- ment rate in the poorest-income groups may have sharply increased between 1966 and 1974. Jallade The distribution of public expenditures in education, including investment, was also estimated for 1970 by Jallade as part of a comprehensive study comparing the distribution of educational COMPARISONS WITH OTHER STUDIES 73 Table 3.20. Distribution of Education Subsidies, 1970, Jallade Percentage of subsidy Household income Households' Average (thousands of pesos) (percentage) Primary Secondary University (percentage) 0-6 19.0 11.0 1.7 - 5.9 (20.0) (32.1) (16.8) (0.8) (19.8) 6-12 20.2 17.9 1.7 0.9 9.5 (20.0) (26.7) (21.8) (4.6) (20.2) 12-24 24.9 32.5 21.8 7.6 23.7 (20.0) (20.5) (21.2) (10.7) (18.6) 24-60 22.9 31.0 50.9 40.7 38.7 (20.0) (14.5) (24.6) (23.5) (20.1) Over 60 13.0 7.6 23.9 50.8 22.2 (20.0) (6.2) (15.6) (60.4) (21.3) Note: Values in parentheses refer to quintiles of households as ordered by household per capita income from the 1974 survey. benefits with the distribution of taxation. Comparisons were made across income groups as well as regions.13 Enrollment data for urban areas were obtained by using the 1970 DANE household survey.'4 Direct data on enrollment were available for children age 12 or older. For children under 12, enroll- ment by income group was derived by comparing the number of grades passed (as reported in the survey) with the age of the child; this comparison was then used to classify children as still enrolled or as dropouts. For the rural areas, the known total of enrolled students in primary education was allocated across income groups by assuming that all households have the same number of children in the 5 to 25 year age bracket and that enrollment ratio is a func- tion of income. Table 3.20 presents the distribution of subsidies obtained in jallade's study, allocated across families according to household income. The values in parentheses are from the present study, 13. Jean-Pierre Jallade, Public Expenditures on Education and Income Distribution in Colombia, World Bank Staff Occasional Papers, no. 18 (Baltimore: Johns Hopkins University Press, 1974). 14. DANE, "Household survey." 1970. 74 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION Table 3.21. Comparison of Enrollment per Ilousehold in Primary Education, Jallade and Sample Survey Urban areas Rural areas Yearly household Percentage Enrollment Percentage Enrollment income (thou- of per of per sands of pesos) households hozusehold households household Jallade, 1970 0-6 a 45.1 0.58 14.7 0.76 (0.58) k ~~~~~(0.69) 6-12 } 31.9 0.89 (0.89) 12-24 30.3 0.91 16.5 1.27 (0.79) (1.27) 24-60 34.3 0.96 6.5 1.93 (0.74) (1.93) Over 60 20.7 0.82 0 0 (0.22) Country total 100.0 0.90 100.0 0.91 (0.64) (0.91) Urban areas Rural areas Quintile in Percentage Enrollment Percentage Enrollment country income of per of per distribution households household households household 1974 sample survey 1 15.0 1.68 27.9 1.28 (1.48) (1.26) 2 16.8 1.51 26.9 0.97 (1.33) (0.98) 3 17.8 1.25 22.7 0.75 (1.00) (0.72) 4 22.5 0.92 15.8 0.44 (0.72) (0.41) 5 27.9 0.57 6.7 0.30 (0.26) (0.27) Country total 100.0 1.10 100.0 0.88 (0.86) (0.86) Note: Figures in parentheses show enrollment in public school. COMPARISONS WITH OTHER STUDIES 75 Table 3.22. Comparison of Enrollment per Household in Secondary Education, Jallade and Sample Survey Urban areas Country average Yearly household Percentage Enrollment Percentage Enrollment income (thou- of per of per sands of pesos) households household households household Jallade, 1970 0-12 14.7 0.04 39.2 0.01 (0.04) (0.01) 12-24 30.3 0.15 24.9 0.11 (0.13) (0.09) 24-60 34.3 0.40 22.9 0.36 (0.26) (0.23) Over 60 20.7 0.73 13.0 0.71 (0.17) (0.16) Country total 100.0 0.34 100.0 0.21 (0.17) (0.10) Urban areas R ural areas Country average Quintile in country Percentage Enrollment Percentage Enrollment Percentage Enrollment income of per of per of per distribution houtseholds houtsehold households household households houisehold 1974 sample survey 1 15.0 0.43 27.9 0.09 20 0.24 (0.24) (0.07) (0.15) 2 16.8 0.47 26.9 0.10 20 0.28 (0.33) (0.07) (0.20) 3 17.8 0.48 32.7 0.11 20 0.31 (0.26) (0.10) (0.19) 4 22.5 0.53 15.8 0.14 20 0.42 (0.27) (0.12) (0.22) 5 27.9 0.57 6.7 0.11 20 0.50 (0.13) (0.09) (0.13) Country 100.0 0.50 100.0 0.12 100 0.35 total (0.23) (0.09) (0.18) Note: Figures in parentheses show the enrollment in public schools. where quintiles of households are ordered by houselhold per capita inconme. If, as appears from the 1974 study, family size is a negative functioin of houselhold per capita incomiie, then faniilies in low- incoimie quintiles will have more children than higlher-income 76 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR EDUCATION quintiles when quintiles of households are ordered by per capita income instead of household income. If this negative function is strong enough, sharp differences in the distribution of educational subsidies can be expected when families are ordered and quintiles are defined by per capita income instead of by household income. This helps to explain the differences in Table 3.20, where poorer families receive a much smaller share of subsidies when they are ordered by household income instead of by household per capita income. The source of those differences can be seen more clearly in Tables 3.21 and 3.22, which show the enrollment per household in primary and secondary education. Because most of the subsidy goes to public education, the difference in the values in parentheses across income groups is the major determinant of the distribution of that subsidy. According to the 1974 sample survey, the enrollment per household in primary education, particularly in public schools, increases sharply for lower-income groups when they are ordered by household per capita income. The reverse is found in Jallade's study when families are ordered by household income, particu- larly in rural areas. Part of this difference could result from the effect of family size on the ordering; it could also reflect genuine differences in enrollment rates for low-income families of com- parable per capita income. Chapter 4 The Distribution of Public Subsidies for Health THE COLOMBIAN HEALTH SYSTIEM las tlhree components: the Na- tional Health System (NHS), the Colombiani Institute of Social Security (Icss), and the Social Security of the Public Sector (Cajas Publicas). The National Health System is the central public health system of the country. In theory, every individual is entitled to use its services. It is run by the Ministry of Health and is financed by funds from the central government, national lotteries, and the contributions of departments and municipalities. Three types of institutions provide services: public hospitals; health centers witl at least one full-time doctor; and puestos de salud (smaller healtl centers), which usually have only a part-time doctor.' The Colombian Institute of Social Security (Icss) not only administers retirement and pension funds, but also provides healtl services to member employees in the private sector through hos- pitals and healtlh centers. Funding for the icss comes fronm tlhree sources: contributions from the Ministry of Health and com- pulsory payments made by both employers and employees. The third health system is made up of the various Social Se- curity branches that serve employees in the public sector. The most important of these are the Caja Nacional de Previsi6n 1. On the average, there are 1.88 doctors, 0.81 dentists, and 3.7 auxiliares (a type of nurse) to a health center. An average puesto de salud has 0.53 doctors and 1.06 auxiliaries. 77 78 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH (CAJANAL) and the Caja de Previsi6n de C(omunicaciones (CAPRECOM). This system is financed by contributions from tlle central government, public agencies emiploying workers belonging to the system, and the employees thenmselves. It also maintains hospitals and healtlh centers. Total Subsidy Received by Households The metlhod used to derive subsidies for lhealth was largely determined by the type of data obtained in the 1974 sample survey on hlousehold consumption of healtlh services according to location and income groups. The relation between the survey data and the structure of the lhealth system is illustrated in Figure 4.1. Sources of the suibsidies Table 4.1 gives figures for the flow of funds showni in Figure 4.1, where A, B, and C are contributions to different institutions of the National Healthl System, and D and E are contributions from the Social Security System to the lhospitals belonging to icss and Cajas Publicas.2 For the Social Security System these figures do not represent the subsidy received by the beneficiaries 2. Flow A was derived from the COLDATOS report, based on data from the Ministry of Health. (Cardex Ministerio de Salud: Presupuesto de Hospitales Oficiales y Mixtos. See COLDATOS report.) Flows B and C were estimated by using 1969 data on the running cost of health centers and an estimate of costs for puestos de salud, based on size of staff. Table SA-20 in the statistical appendix, shows the cost of each health center (with and without beds) for 1969, adjusted for the increase in the price level between 1969 and 1974. The cost of maintaining each puesto de salud (Table SA-21) was derived by com- paring the wage bill for the average puesto staff in 1969 (0.53 doctors, 1.06 auxiliaries) with the average wage bill for a health center. The 1969 figure was again adjusted by the change in the price level between 1969 and 1974 (103 percent). Table SA-22 shows the total cost of operating the puestos, by location. The cost of each puesto is assumed to be the same in all locations; thus the total subsidy becomes proportional to the number of puestos in each location. Flow D is derived from icss figures in the COLDATOS report (Informe Estadistico ICSS). Flow E includes 259.8 million pesos from CAJANAL (Item, "Prestaciones Medicas," of Presupuesto de Ingreso y Rentas) and 39 million from CAPRECOM (Item, "Servicios Medicos,- Presupuesto de Entidades Decentralizadas, Informe Financiero de 1973, Republica de Colombia). The breakdown of Flow E by location was based on the percent distribution shown in the COLDATOS report and on the 1969 INPES study on health institutions (INPES: Censo de Instituciones Hospitalarias, 1970). Figure 4. 1. Colombian Health System Services received by households Sources of financing Institutions providing as measured by the 1974 survey (excluding sale Components of the system services (by region and by of services) income group) Ministry ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~~~~ ~upain ofiHeatt Lotteries and taxes S S Public hospitals of Hospital \ _ . / ~~~~~~~~~~~~~~~~Deliveries |Departments a-nd |National Health System B ll Operations the Public Sector ~Health centers |municipalities (NHS) Inpatient days X~~~~~~~~~C I W |~~~~Ministry of H&atK34 Vst onre |Contributions Health\ D |private enterprises K Institute of Healh T ospitals a~ndI \ Social Security _ healthcetr \ Contributions from (as (ICSS) Retirement Outpatient visits pubicat employees \ em\o Deli% eries Contributions f ro Eprain public agencies Social Security of Health / osp Inpatient days |Contributions fro (Cajas Publicas) Retirement healthcetr public employe 80 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.1. Funding Received by Health Institutions, 1974 (millions of pesos) Location of institution Inter- Health system Large mediate Small Urban Rural Country and institution cities cities towns total areas total National Health System Hospitals 283.5 339.4 381.7 1,004.6 13.8 1,018.4 (A) Health centers 106.4 54.5 54.8 215.7 22.3 238.0 (B) Puestos 1.2 7.3 34.5 43.0 57.5 100.5 (C) Social Security System icss 954.2 367.3 142.1 1,436.6 5.1 1,468.7 (D) Cajas 225.1 22.9 12.8 260.8 38.0 298.8 (E) Total funding 1,570.4 791.4 625.9 2,987.7 136.7 3,124.4 Sources: See text (note 2) and Figure 4.1. of the system. The subsidy is smaller, since it must exclude the fraction of that funding financed by the laborers themselves. Excluding indebtedness, the sources of financing of the Social Security System are: ICSS Cajas Contributions (percentage) (percentage) Ministry of Health 4.8 61.9 Private enterprises (icss) 72.0 70.8 or public agencies (Cajas) 67.2 8.9 Employees 28.0 29.2 According to these figures, the contribution of workers to the icss is 29 percent of the total contribution of workers and em- ployees (0.280/0.952). It also coincides approximately with the percentage contribution fixed by law, namely, 3.5 and 7.0 percent of the wage for workers and employers, respectively. Thus, the contribution from labor is 33 percent of the total contribution. The contribution of labor to the Cajas Publicas is 29 percent, if 70.8 percent is defined as the contribution of the (public) employer. A rigorous definition of the subsidy should exclude the real TOTAL SUBSIDY RECEIVED BY HOUSEHOLDS 81 (not legal) contribution of workers: that is, the difference between the wage that would have prevailed without the system and the new net wage received by labor. This incidence depends not only on labor supply and demand, but also on the value placed by workers on the services provided by the system. This valuation will in turn affect the new (post-system) supply price of labor and tlhus the new equilibrium wage. In the appendix to this chapter, a simple framework for evaluat- ing the real incidence borne by labor is presented. It is shown that the larger the ratio of the elasticity of labor supply relative to labor demand, the smaller the incidence borne by labor. If that ratio is 2, and workers place zero value on the services they receive, the real incidence becomes equal to the legal incidence; if workers value the services in an amount equal to one quarter of the con- tribution of both labor and employers, the real incidence borne by labor is 0.5. If a reasonable range of values is used-a ratio of elasticities between 2 and 5 and a valuation of services ranging from 0.10 to 0.5 times the total contribution-the real incidence fluctuates between 0.4 and 0.6. A value of 0.5 will be used for the calculation: namely, an economic incidence larger than the legal figure of 0.29 discussed earlier. Thus the subsidy in the ICSS system becomes equal to 0.524 [0.048 + (0.5) (0.952)] times the contri- butions shown in Table 4.1. In the case of the Cajas Publicas, if the nonlabor contribution is defined as the contribution of the (public) employer, the subsidy becomes 0.5 times the contribution shown in Table 4.1. The estimated subsidies are summarized in Table 4.2. The total health subsidy for 1974 becomes 2,276.1 million pesos, one- third of the subsidy for education, (7,000.6 million pesos). It represents 3.6 percent of total government expenditure and 0.7 percent of the Colombian GNP. Subsidy classified by type of health service The next step is to classify the subsidy figures by type of service, using the same classification as the 1974 sample survey. For this an independent estimate of the relative cost of providing these services is required. The only data source for the relative unit cost of healtlh services is a 1974 study of the icss hospitals.3 The estimated costs, in terms 3. Estudio de Costos de Servicios Midicos. icss, 1974. 82 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.2. Estimated Total Subsidy Received by Ilouseholds, 1974 (millions of pesos) Location of institution Inter- Health system Large mediate Small Urban Rural Country and institution cities cities towns total areas total National H1ealth Service Hospitals 283.5 339.4 381.7 1,004.6 13.8 1,018.4 (A) Health centers 106.4 54.5 54.8 215.7 22.3 238.0 (B) Puestos 1.2 7.3 34.5 43.0 57.5 100.5 (C) Social Security System Icssa 500.0 192.5 74.5 767.0 2.7 769.7 Cajas b 112.6 11.5 6.4 130.5 19.0 149.5 Total subsidy 1,003.7 605.2 551.9 2,160.8 115.3 2,276.1 a. Figures are 52.4 percent of those in Table 4.1. b. Figures are 50 percent of those in Table 4.1. of the cost of one outpatient visit, are as follows: 1 delivery equals 1.17 outpatient visits; 1 operation equals 29.5; and 1 inpatient day equals 4. If these figures are representative for all hospitals, the number of services provided by all types of hospitals can be ex- pressed in terms of equivalent outpatient visits. The percentage distribution of each service, in equivalent outpatient visits, can thereafter be used to distribute the total subsidy. Tables 4.3, 4.4, and 4.5 shIow the numiiber of services provided by the three classes of hospitals in 1974 in terms of equivalent outpatient visits and the distribution of the subsidy according to the percentage dis- tribution of the visits.4 4. Data on the number of services provided by IcsS in 1974 were available from official statistical reports (Informe Estadistico, Icss). For the hospitals belonging to the National Health System and Cajas Publicas, 1969 data from the INPES study were used. The number of services provided in 1969 was expanded by the increase in the total number of outpatient visits, for which 1974 data were available. In other words, it was assumed that all services grew at the same rate as outpatient visits. Although this expansion is not necessary to derive the percentage distribution of the subsidy, it allows for some comparison with the absolute number of services reported by households in the sample survey. TOTAL SUBSIDY RECEIVED BY HOUSEHOLDS 83 Table 4.3. National Health System Hospitals: Number of Services Provided and the Distribution of the Subsidy, 1974 Location of institution Inter- Large mediate Small Urban Rural Country Services cities cities towns total areas total Services provided in 1974 (thousands) Number of services provided Outpatient visits 1,193 906 1,942 4,041 98 4,139 Deliveries 78 55 60 193 2 195 Operations 125 63 31 219 1 220 Inpatient days 1,657 2,189 2,075 5,921 110 6,031 Number of services in equivalent outpatient visits' Outpatient visits 1,193 906 1,942 4,041 98 4,139 Deliveries 91.3 64.4 70.2 225.9 2. 3 228.2 Operations 3,687.5 1,858.5 914. 5 6,460.5 29.5 6,490 Inpatient days 6,628 8,756 8,300 23,684 440 24,124 Total 11,599.8 11,584.9 11,226. 7 34,411.4 569.8 34,981.2 Distribution of the 1974 subsidy (millions of pesos) Outpatient days 29.2 26.5 66.0 121.7 2.4 124.1 Deliveries 2.2 1.9 2.4 6.5 0.1 6.6 Operations 90.1 54.4 31.1 175.6 0.7 176.3 Inpatient days 162.0 256.6 282.2 700.8 10.6 711.4 Total 283.5 339.4 381.7 1,004.6 13.8 1,018.4 a. The cost of 1 delivery equals the cost of 1.17 outpatient visits; of I operation equals 29.5, and of I in- patient day equals 4. Source: Data from INPES study, 1969. Table 4.6 summarizes data on the subsidy by type of service and type of health system. The values for the icss and Cajas Publicas have been combined as a total figure for the Social Se- curity System. The figures for healtlh centers were not broken down by type of service. In the case of National Health System hospitals, almost 75 percent of the subsidy goes to finance inpatient care. In the Social Security System, the subsidy is more or less evenly spread over outpatient visits and inpatient care, with approximately 40 percent of the subsidy going to each type of service. 84 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.4. icss Hospitals: Number of Services Provided and the Distribution of the Subsidy, 1974 Location of institution Inter- Large mediate Small Urban Rural Country Services cities cities towns total areas total Services provided in 1974 (thousands) Number of services provided Outpatient vis- its 4,696 2,507 884 8,087 37 8,121 Deliveries 47 20 7 74 - 74 Operations 70 22 10 102 - 102 Inpatient days 954 467 156 1,577 5 1,582 Number of services in equivalent outpatient daysa Outpatient days 4,696 2,507 884 8,087 34 8,121 Deliveries 55 23.4 8.2 86.6 - 86.6 Operations 2,065 649 295 3,009 - 3,009 fnpatient days 3,816 1,868 624 6,308 20 6,328 Total 10,632 5,047.4 11,811.2 17,490.6 54 17,544.6 Distribution of the 1974 subsidy (millions of pesos) Outpatient days 220.8 95.6 36.4 352.8 1.7 354.5 Deliveries 2.6 0.9 0.3 3.8 - 3.8 Operations 97.1 24.8 12.1 134 - 134 Inpatient days 179.5 71.2 25.7 276.4 1.0 277.4 Total 500.0 192.5 74.5 767.0 2.7 769.7 a. The cost of I delivery equals the cost of 1.17 outpatient visits; of I operation equals 29.5; and of I in- patient day equals 4. Source: Data from Informe Estadistico, icss, 1974. Consumption of Services and Distribution of Subsidies across Income Groups The 1974 sample survey provides data on the consumption of different services as reported by households. These figures can TOTAL SUBSIDY RECEIVED BY HOUSEHOLDS 85 Table 4.5. Cajas Publicas hlospitals: Number of Services Provided for 1969 and Distribution of the Subsidy in 1974 Location of institution Inter- Large mediate Small Urban Rural Country Services cities cities towns total areas total Services provided in 1974 (thousands) Number of services provided Outpatient visits 348.9 68.2 25.7 442.8 86.6 529.4 Deliveries 5.7 0.3 0.1 6.1 1.5 7.6 Operations 13.6 0.7 0.2 14.5 0.5 15.0 Inpatient days 291.5 12.6 16.2 320.3 41.0 361.3 Number of services in equivalent outpatient days, Outpatient days 348.9 68.2 25.7 442.8 86.6 529.4 Deliveries 6.7 0.4 0.1 7.2 1.8 9.0 Operations 401.2 20.6 5.9 427.7 14.8 442.5 Inpatient days 1,165.6 50.4 64.8 1,280.8 163.9 1,444.7 Total 1,922.4 139.6 96.5 2,158.5 267.1 2,425.6 Distribution of the 1974 subsidy (millions of pesos) Outpatient days 20.5 5.6 1.7 27.8 6.2 34.0 Deliveries 0.3 0.1 0.1 0.5 0.1 0.6 Operations 23.5 1.7 0.4 25.6 1.0 26.6 Inpatient days 68.3 4.1 4.2 76.6 11.7 88.3 Total 112.6 11.5 6.4 130.5 19.0 149.5 a. The cost of I delivery equals the cost of 1.17 outpatient visits; of I operation equals 29.5; and of I in- patient day equals 4. Source: Data from INPES study, 1969. now be compared with those reported by the health institutions themselves. Allocation of subsidies to zurban and rural houtseholds This comparison is necessary to allocate the subsidy between urban and rural households since an important fraction of rural 86 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.6. Public Subsidies to Health Institutions, Classified by Location of the Institution, 1974 (millions of pesos) Location of institution Inter- Health system Large mediate Small Urban Rural Country and service cities cities towns total areas total National Health System Hospitals Outpatient visits 29. 2 26.5 66.0 121.7 2.4 124.1 Deliveries 2.2 1.9 2.4 6.5 0.1 6.6 Operations 90.1 54.4 31.1 175.6 0.7 176.3 Inpatient days 162.0 256.6 282.2 700.8 10.6 711.4 Health centers 107.6 61.8 89.3 258.7 79.8 338.5 Social Security System Hospitals Outpatient visits 241.3 101.2 38.1 380.6 7.9 388.5 Deliveries 2.9 1.0 0.4 4.3 0.1 4.4 Operations 120.6 26.5 12.5 159.6 1.0 160.6 Inpatient days 247.8 75.3 29.9 353.0 12.7 365.7 households' consumption takes place in urban institutions. The comparison of the two sets of data is shown in Table 4.7. If an important part of the services provided in urban areas is consumed by rural households, an excess supply of services in urban areas (defined as an excess of services provided over services consumed by households located in that area) and an excess de- mand of services in rural areas (defined as an excess of services consumed by households located in the area over the ones pro- vided by the institutions in the area) would be expected. Although the data in Table 4.7 seem to demonstrate tlhis, the size of the excess supply in urban areas (particularly for the Social Security System) is substantially larger than the excess demand in rural areas. In other words, the country total figures reported by the healtlh institutions are larger than the country total figures re- ported by households. The largest difference-in outpatient visits to institutions of the Social Security Svstemi-mav occur because the number of visits reported by houselholds refers only to visits to CONSUMPTION OF SERVICES AND DISTRIBUTION OF SUBSIDIES 87 Table 4.7. Number of Services Provided in Hospitals (Reported by Health Institutions) and Consumed by Households (Reported in the 1974 Sample Survey) (thousands) Inter- Health system Large mediate Small Urban Rural Country and service cities cities towns total areas total National Health System Outpatient days 1,193 906 1,942 4,041 98 4,139 (891) (1,040) (939) (2,870) (1,204) (4,074) Deliveries 78 55 60 193 2 195 (34) (36) (30) (100) (76) (176) Operations 125 63 31 219 1 220 (49) (24) (11) (84) (25) (109) Inpatientdays 1,657 2,189 2,075 5,921 110 6,031 (1,611) (1,080) (660) (3,351) (2,679) (6,030) Social Security System, Outpatient visits 4,696 2,507 884 8,087 34 8,121 (1,510) (1,129) (411) (3,050) (258) (3,288) Deliveries 47 20 7 74 - 74 (22) (16) (8) (46) (9) (55) Operations 70 22 10 101 - 101 (16) (5) (1) (22) - (22) Inpatient days 954 467 156 1,577 5 1,582 (778) (230) (100) (1,158) - (1,158) Note: Figures are reported according to the locations of the institution and of the household. Figures reported by households are in parentheses. a. Figures reported by institutions include only hospitals belonging to the icss system. physicians, whereas the figures reported by the institutions refer to visits to paramedical personnel as well. For the purposes of the study, the number of services received by rural households in urban areas will be assumed to be equal to the excess demand figures for the rural areas in Table 4.7. These figures appear in the first column of Table 4.8, where they are allocated to institutions classified by city size. The rural excess demand was allocated according to the excess supply figures in each urban stratum, from small towns up to large cities. Obvi- ously, the last number allocated need not coincide with the excess 88 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.8. Allocation of Services Provided in Urban Hospitals to Rural Ho useholds (thousands) Services received Large cities by rural house- holds, minus Allocated services provided Services to rural Health system by rural provided households and service institlUtionsb (1) (2) (2) /(1) National Health System Outpatient visits 1,106 - - - Deliveries 74 78 25 0.32 Operations 24 - - Inpatient days 2,569 Social Security System Outpatient visits 204 - - - Deliveries 9 47 5 0.11 a. Figures taken from 1974 sample survey. b. Figures taken from COLDATOS report. supply of that region, since the total urban excess supply is not necessarily equal to the total rural excess demand. The figures in Table 4.8 suggest that more than half the services provided in small towns by the National Health System are consumed by rural houselholds. For inpatient care, this is also true for institu- tions located in intermediate cities. Using the information in Table 4.8, the subsidies to health institutions shown in Table 4.6 can be translated into subsidies received by houselholds, classified by location. The results are shown in Table 4.9. Consumption of services by income groups Tables 4.10 and 4.11 present the distribution of consumption according to income quintiles derived from the 1974 sample sur- vey.< Income quintiles are again defined according to the country distribution of income. 5. The absolute consumption figures are shown in the statistical appendix, Tables SA-23 and SA-24. CONSUMPTION OF SERVICES AND DISTRIBUTION OF SUBSIDIES 89 Intermediate cities Small towns Allocated Allocated Services to rural Services to rural provided households provided households (1) (2) (2)/(I) (1) (2) (2)1(1) - - - 1,942 1,106 0.57 55 19 0.34 60 30 0.50 63 4 0.06 31 20 0.64 2,189 1,154 0.53 2,075 1,415 0.68 - - - 884 204 0.23 20 4 0.20 - - - Table 4.9. Health Subsidies Received by Households, Classified by Location of Household, 1974 (millions of pesos) Location of household Inter- Large med iate Small Urban Rural Country Institution and service cities cities towns total areas total National Health System Hospitals Outpatient visits 29.2 26.5 28.4 84.1 40.0 124.1 Deliveries 1.5 1.3 1.2 4.0 2.6 6.6 Operations 90.1 51.1 11.2 152.4 23.9 176.3 Inpatient days 162.0 120.6 90.3 372.9 338.5 711.4 Health centers 107.6 61.8 89.3 258.7 79.8 338.5 Social Security System Hospitals Outpatient visits 241.3 101.2 29.3 371.8 16.7 388.5 Deliveries 2.6 0.8 0.4 3.8 0.6 4.4 Operations 120.6 26.5 12.5 159.6 1.0 160.6 Inpatient days 247.8 75.3 29.9 353.0 12.7 365.7 90 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.10. Services Provided by National Health System Hospitals and Health Centers, Classified by Location of Household (percentage distribution reported by households in the 1974 sample survey) Location of household Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns average areas average Hospitals Outpatient visits 1 16.4 12.4 33.2 20.5 26.5 22.2 2 14.6 17.6 22.8 18.4 27.6 21.1 3 18.4 17.1 26.6 20.6 22.2 21.1 4 31.5 31.0 12.4 25.0 18.8 23.2 5 19.1 21.9 5.0 15.5 4.9 12.4 Deliveries 1 26.5 22.2 36.7 28.0 31.6 29.5 2 23.5 25.0 23.3 24.0 31.6 27.3 3 17.6 19.4 20.0 19.0 21.0 19.9 4 23.5 22.3 6.7 18.0 10.5 14.8 5 8.9 11.1 13.3 11.0 5.3 8.5 Operations 1 24.5 12.5 36.4 22.6 40.0 26.6 2 6.1 16.7 18.2 10.7 32.0 15.6 3 8.2 33.3 - 14.3 12.0 13.8 4 24.5 25.0 27.3 25.0 16.0 22.9 5 36.7 12.5 18.1 27.4 - 21.1 Inpatient days 1 46.1 27.0 29.4 42.1 23.2 33.7 2 14.4 23.3 23.5 13.6 31.5 21.6 3 15.7 14.5 13.9 15.0 26.8 20.2 4 9.0 25.9 13.2 15.3 14.8 15.1 5 14.8 9.3 20.0 14.0 3.7 9.4 Health centers Visits to nurses 1 14.5 19.7 37.7 22.7 30.5 25.2 2 27.8 28.7 30.0 28.7 27.9 28.4 3 29.2 18.4 16.2 22.5 25.3 23.3 4 18.6 23.3 8.1 16.7 11.8 15.2 5 9.9 9.9 8.0 9.4 4.5 7.9 CONSUMPTION OF SERVICES AND DISTRIBUTION OF SUBSIDIES 91 Table 4.11. Services Provided by Social Security System Hospitals (icss and Cajas), Classified by Location of Household (percentage distribution reported by households in the 1974 sample survey) Location of household Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns average areas average Outpatient visits 1 2.2 7.4 6.8 4.7 18.9 5.7 2 12.1 14.4 35.5 16.1 24.8 16.7 3 24.8 18.0 21.2 21.8 29.4 22.3 4 27.7 30.5 24.8 28.4 23.1 28.0 5 33.2 29.7 11.7 29.0 3.8 27.3 Deliveries 1 - 12.4 - 4.3 33.3 9.1 2 13.7 6.3 25.0 13.0 11.2 12.7 3 40.9 6.3 37.5 28.3 22.2 27.3 4 22.7 37.5 25.0 28.3 33.3 29.1 5 22.7 37.5 12.5 26.1 - 21.8 Operations 1 12.5 20.0 - 13.6 - 13.6 2 18.9 - - 13.6 - 13.6 3 37.3 20.0 100.0 36.5 - 36.5 4 6.3 40.0 - 13.6 - 13.6 5 25.0 20.0 - 22.7 - 22.7 Inpatient days 1 7.1 4.3 38.0 9.0 - 9.0 2 16.4 5.0 29.0 14.8 - 14.8 3 40.6 8.6 20.0 31.1 - 31.1 4 18.3 60.7 13.0 28.1 - 28.1 5 17.6 21.4 - 17.0 - 17.0 As shown in Table 4.9, inpatient care accounts for most of the subsidy to the National Health Service; thus the distribution of consumption of this service determines the distributive content of the subsidy. Lower-income quintiles consume the largest share of this service, and the share declines monotonically for higher- income groups. In the case of outpatient visits, the share of services consumed is similar for each of the first four quintiles, diminishing 92 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.12. Households Affiliated with the Social Security System, Classified by Income Quintile (percentage) Location of household Income quintile Inter- (poorest Large mediate Small Rural Country to richest) cities cities towns areas average 1 5.1 6.2 12.1 19.5 8.5 (4.6) (6.6) (14.1) (27.4) (6.7) 2 13.1 13.8 27.4 26.3 17.1 (14.4) (15.9) (21.7) (21.2) (16.1) 3 17.9 20.0 23.3 28.0 20.7 (18.6) (21.9) (25.9) (26.8) (20.7) 4 25.8 26.2 21.1 18.6 24.3 (25.8) (25.9) (22.1) (17.3) (25.2) 5 38.1 33.8 16.1 7.6 29.4 (36.6) (29.7) (16.2) (6.8) (31.3) Note: The distribution of individuals is given in parentheses. sharply for the richest quintile. For surgical operations, the shares of consumption appear more erratic across income groups. In the Social Security System (icss and Cajas), 82 percent of the subsidy is accounted for by outpatient visits and inpatient care. The country total figures in Table 4.11 show that the three richest quintiles have the higlhest shares of consumption, particu- larly in the case of outpatient care. For inpatient care, the shares of consumption are concentrated in the third and fourtlh quintiles. The plausibility of the figures derived for the Social Security System can be checked by comparing them with the distribution of households or individuals (where individuals include workers plus their family) affiliated with this system. This distribution is shown in Table 4.12. The distribution of affiliated individuals resembles the distribution of outpatient visits more than it does the distribu- tion of inpatient days, particularly in the richest quintile; the resemblance still holds when the comparison is made at the level of regional breakdown. Using the figures in Tables 4.10 and 4.11 and the subsidy figures in Table 4.9, it is possible to allocate the total health subsidy CONSUMPTION OF SERVICES AND DISTRIBUTION OF SUBSIDIES 93 across income quintiles and location of households.6 The resulting subsidy per household is presented in Table 4.13. It is computed by dividing the total subsidy (allocated to each type of institu- tion) by the total number of households in the corresponding quintile and location. Thus, the subsidy per household must be interpreted as the subsidy received by the representative family in that location and income quintile. The total subsidy per house- hold is equal to the sum of subsidies received by the representative household from all types of institutions. Several conclusions can be drawn from the results in Table 4.13. (a) There is a strong difference in the subsidy for urban and rural households in the same quintile. Urban households receive a subsidy at least twice as large as that received by their rural coun- terparts; it is almost three times as large in the poorest quintile. (b) In a given quintile, the subsidy tends to be larger for house- holds living in larger cities. (c) The subsidy is largest for the first four quintiles. Within these quintiles the highest variation is found in large cities; this variance is dominated by the large National Health System sub- sidy in the first quintile and the large Social Security System sub- sidy in the third. An alternative way of presenting the subsidy per household is to compute, for each cell, a subsidy per household affiliated with the Social Security System and a subsidy per houselhold not affiliated with the system. This allows identification of the intracell differences in the subsidy that are induced by affiliation with the Social Security System. Table 4.14 presents these results. The average value for the country shows the subsidy for an affiliated household (AF) as 2.4 times the one for a nonaffiliated household (NAF). The mean is much closer to the second figure, because only 22.2 percent of heads of household are affiliated with the system.7 6. See the statistical appendix, Table SA-25. 7. Households affiliated with the Social Security System, as a percentage of the households in each quintile, are, respectively: Quintile 1 2 3 4 5 Average (poorest to richest) Percentage of households 9.4 18.4 24.7 27.0 33.0 22.2 94 THE DISTRIBUTION OF PtUBLIC SUBSIDIES FOR HEALTH Table 4.13. Health Subsidy per Household, 1974 (pesos) Location of household Income quintile Large cities Intermediate cities Small towns (poorest to richest) NHS sSS HC Total NHS SSS BC Total NHS SSS HC Total 1 1,085 404 166 1,655 546 206 156 908 246 82 205 533 2 245 683 220 1,148 412 183 176 771 216 138 193 547 3 213 1,141 173 1,527 345 263 100 708 198 242 141 581 4 189 491 82 762 326 543 89 958 198 119 76 393 5 162 399 27 588 134 294 35 462 368 58 122 548 Country average 271 587 103 961 317 323 98 738 235 129 160 524 Note: NHS = National Health Service hospital; sss = Social Security System hospital; and Hic health center. Table 4.14. Health Subsidy per Household, Classified by Affiliation (AF) and Nonaffiliation (NAF) with the Social Security System, 1974 (pesos) Location of household Income quintile Large cities Intermediate cities Small towns (poorest to richest) AF NAF Mean AF NAF Mean AF NAF Mean 1 2,062 1,556 1,655 1,221 845 908 1,089 488 533 2 1,963 714 1,148 632 828 771 671 515 547 3 3,153 605 1,527 715 704 708 1,084 436 581 4 1,325 430 762 1,514 647 958 499 359 393 5 1,118 295 588 724 283 462 205 681 548 Country average 1,697 572 961 954 627 738 701 483 524 Sensitivity analysis To evaluate the sensitivity of the results, an alternative distri- bution of the subsidy was used for both the National Health System and Social Security Systern networks.8 The total subsidy 8. The alternate distribution of public subsidies to hospitals is given in the statistical appendix, Table SA-26. CONSUMPTION OF SERVICES AND DISTRIBUTION OF SUBSIDIES 95 Location of household Urban average Rural areas Country average NHS SSS HC Total NHS SSS HC Total NHS SSS HC Total 551 201 183 935 259 17 63 339 395 103 119 617 279 347 198 824 341 23 60 424 310 186 130 626 247 657 144 1,048 331 27 65 423 284 381 109 774 235 438 83 756 285 28 43 356 250 314 71 635 173 337 38 548 159 14 39 212 172 295 38 505 275 398 116 789 295 23 58 376 282 255 94 631 Location of household Urban average Rural areas Country average AF NAF Mean AF NAF Mean AF NAF Mean 1,534 844 935 286 343 339 1,096 568 617 1,236 664 824 264 439 424 1,008 539 626 2,009 581 1,048 248 445 423 1,535 523 774 1,274 484 756 270 366 356 1,159 440 635 925 333 548 132 222 212 887 315 505 1,309 561 789 254 387 376 1,148 484 631 to the former was allocated across income quintiles according to the distribution of outpatient visits by income groups; this distri- bution is different from the one for inpatient days, which was the main determinant of the earlier estimate of the distribution of the subsidy. The total subsidy to the sss was allocated according to the income distribution of those individuals (not households) affiliated with the system, as shown in parentheses in Table 4.12. The subsidy per household derived from this distribution is shown 96 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.15. Ilealth Subsidy per Household: Summary (pesos) Location of houtsehold Inter- Income quintile Large mediate Smnall Urban Rural Country (poorest to richest) cities cities towns average area average 1 1 655 908 533 935 339 617 (959) (645) (533) (678) (364) (500) 2 1,148 771 547 824 424 626 (1,172) (844) (520) (843) (380) (613) 3 1,527 708 581 1,048 423 774 (1.090) (790) (661) (893) (381) (668) 4 762 958 393 756 356 635 (1,091) (800) (417) (871) (419) (734) 5 588 462 548 548 212 505 (745) (628) (430) (683) (278) (630) Country average 961 738 524 789 376 631 (961) (738) (524) (789) (376) (631) Note: The figures in parentheses were obtained by the alternate distribution. in parentheses in Table 4.15, together witlh the subsidy figures estimated earlier. Table 4.15 shows that at the country level the results are not particularly sensitive to this alternative method of allocating the subsidy. A change of about 20 percent is observed in the poorest and richest quintiles: the former having a lower subsidy with the alternate distribution and the latter a higher subsidy. The inajor difference is observed for households in large and intermiediate cities in the poorest-income quintile, where the subsidy under the alternate distribution is, respectively, approximlately 0.57 and 0.71 times the earlier estimate. The source of the difference is the allocation of the subsidy to thie National Health System according to outpatient visits instead of inpatient days. In the poorest quin- tile the share for inpatient days is twice the share for outpatient visits. The final conclusion is that despite the metlhod of allocation used, the subsidy for each lhousehold is largest for the middle- income quintiles-the third and fourth with the alternate metlhod -and declines to each end of the income distribution. The subsidy EXPLANATORY VARIABLES BEHIND THE DEMAND 97 to the middle-income households is approximately 25 to 50 percent larger than the subsidies to the poorest and richest quintiles. Explanatory Variables behind the Demand for Medical Services The data derived from the 1974 sample survey indicate that even in low-income households, private medicine represents a substantial share of the total consumption of health services. Visits to physicians in private practice account for more than half the total visits for both urban and rural regions, even in the poorest quintile. This suggests that an important fraction of the consumption of healtlh services, even in low-income groups, takes place outside the National Health System and Social Security System . Table 4.16 shows visits to physicians per household and per person by income group and location. The number of visits per household tends to increase with the per capita income of the family; the positive relation appears even stronger when the visits are calculated for individuals, except for rural households in the richest quintile. The data show a strong disparity between urban and rural houselholds belonging to the same quintile; the figure for urban houselholds is almost twice as large, even when visits to nurses are included in the rural figure. Definitioni of explanatory zariables To explain this disparity, it is necessary to identify possible factors influencing households' demand for medical services. The quantity demanded is defined as the number of visits to doctors by all memiibers of the household during 1974. Multiple regression experiments were undertaken for three definitions of the depen- dent variable: V1, the number of visits to doctors in Social Se- curity Systeml hospitals by households affiliated with the system; 172, the number of visits to doctors in private practice by all types of houselholds; and V3, the number of visits to all types of doctors by all types of households. V3 includes V1 and l 2 plus all the visits to doctors in institutions belonging to the National Health System. 98 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.16. Average Annual Visits to Physicians, per Household and per Person, 1974 Urban areas per household Income quintile Total (poorest Hospitals, Hospitals, Private per to richest) NHS SSS practice Total person 1 1.74 0.43 2.26 4.43 0.66 2 1.40 1.31 2.45 5.16 0.86 3 1.49 1.67 2.97 6.13 1.10 4 1.44 1.73 3.81 6.98 1.40 5 0.72 1.42 4.39 6.53 1.49 Country average 1.29 1.37 3.36 6.02 1.11 It can be hypothesized that the demand for medical services will be influenced by the purchasing power of the family, to the extent that some cost is involved in obtaining the services; by the size and age composition of the family; and by variables influ- encing the preference of households for consuming the services. Table 4.17 shows regression coefficients for explanatory vari- ables used as substitutes for these causal factors. The variables used were: PERCAP, per capita income of the household; NPER, number of persons in the household; EDUC HEAD, years of schooling of head of household; EDUC WIFE, years of schooling of wife; and SHARE CHILD, number of children in a particular age bracket, as a fraction of NPER. Regression coefficient results The figures in Table 4.17 are regression coefficient results for urban households only; no significant results were obtained for rural households. For urban households, only V2 and IZ3 yielded significant results; V1, visits to Social Security System doctors, did not appear to be associated with the variables described above. V2, visits to doctors in private practice, was the only component of the total demand for medical services responding to the hypoth- esized causal variables. Because these visits account for more than EXPLANATORY VARIABLES BEHIND THE DEMAND 99 Rural areas per household Visits to Hospitals, Hospitals, Private nurses and Total per NHS sss practice health centers Total person 0.83 0.11 1.41 0.34 2.69 0.38 0.90 0.16 1.;6 0.17 2.79 0.47 0.86 0.22 1.24 0.39 2.71 0.53 1.04 0.25 2.07 0.21 3.57 0.80 0.64 0.10 1.35 0.18 2.27 0.67 0.87 0.17 1.51 0.27 2.82 0.51 lhalf the total visits, the regressions for V3 are dominated by the belhavior of V2.I The regression results in Table 4.17 have a low R2: only 6 to 8 percent of the observed variance is accounted for. Nevertlheless, some variables appear significant, and thlis at least suggests new research. Per capita income and the education variables appear signifi- cant. The elasticity withl respect to income is larger for V1: that is, th-e income elasticity of demand for doctors in private practice is larger than it is for doctors in public institutions. Education of the lhead of houselhold has a significantly larger coefficient (almost three times as large) than education of the wife. The elasticities for NPER are substantially lower than 1, the elasticity for V2 being a tlhird of that for V3. Elasticities smaller than 1 indicate some economies of scale in the use of medical services with larger lhouselhold size, this being more important for 172. Other demograplhic variables, suchi as the shares of children below age 18, lhave negative coefficients, but not all are statisti- 9. Visits to doctors in private practice represent 51 percent of the total visits for urban households in the poorest quintile and 67 percent for the ones in the richest quintile (see Table 4.16). 100 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.17. Regression Coefficient Results from Regressing the Number of Visits to Doctors as a Function of Socioeconomic Variables of hlouseholds in Urban Areas Dependent variable Log of number of visits Independent to doctors in private Log of number of visits variable practice, log V2 to all doctors, log V3 Constant -0.278b 0.004 (2.30) (0.03) log PERCAP 0.073, 0.047, (4.37) (2.57) log NPER 0.135, 0.378, (2.58) (6.59) log.EDUC HEAD 0.171 0.182 (5.61) (5.49) log EDUC WIFE 0.058b 0.077, (2.17) (2.63) log SHARE CHILD -0.047 -0.088 (ages 0-6) (1.59) (2.72) log SHARE CHILD -0 .012 0.022 (ages 7-12) (0.36) (0.61) log SHARE CHILD -0.065b -0.010 (ages 13-18) (2.08) (0.29) 2 = 0.06 2= 0.08 F = 22.6 F= 30.6 Note: Values in parentheses are i-ratios. a. Values significant at I percent in a two-tail test. b. Values significant at 5 percent in a two-tail test. cally significant. This implies a more intensive use of medical services by adults than by children under 18. Appendix. A Framework for Analyzing the Incidence of the Financing of the Social Security System Equilibrium in the labor market without the system. Assume that the supply, LI, and demand for labor, Ld, can be written as a constant elasticity function of the wage rate: (1) Ls = aWa (2) Ld = bW-0. FINANCING OF THE SOCIAL SECURITY SYSTEM 101 In equilibrium L8 = Ld, and: (3) Wo = a where WO is the presystem equilibrium wage rate. WO is the wage paid by employers as well as the one received by workers. New equilibrium when implementing the system Assume that each worker values the services provided by the system in an amount V per year. Assume furthermore that such value is proportional to the supply price W defined earlier. De- noting as O/ (1 + 0) that factor of proportionality; (4) V- 1 W. The new supply price of labor now becomes: (5) W = W - v = I If labor contributes to the system a fraction ti of its (gross) wage, the new (or gross) supply price faced by employers becomes W1, namely: (6) W, W,-_= W1 Solving for W out of Equation (1) and substituting into Equa- tion (6) gives: +I )(L\ ") If employers also contribute a tax of t, on the wage paid, the new demand price for labor faced by workers becomes: W (8) W, += I +t Solving for W out of Equation (2) and substituting into Equa- tion (8): (9) W ( = ,) 102 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Solving for Lg and Ld from Equations (7) and (9) and making LS = Ld, the new equilibrium wage WV is defined as a function of the presystemn wage WO: (10) Jj7i = (1 (1 + t-) (kk (1-t,.),'k (1 + )/ where a + 3 = k. Of interest is the change in the net money wage received by labor as a fraction of the total contributions to the system per employed worker. Denoting that contribution as R: (11) R = (tw + t)1Wl. The change in the net wage as a fraction of R becomes: a _ (1- -t) WTV- Wo 1 F WoJ R (tw + te) WI = (t + t) L ( - tn) - - . Substituting (W°) from Equation (10) into Equation (12) produces: (1)- (1 - tw,) - (l - tW),1k (1 + 0),Ik (1 ± t)O/k (13)(t±) R (t. + t, The value of 0 Assume that the value of V can be related to the value of R, the total contributions to the systenm per employed worker. If the valuation of the services of the system V is a fraction jt of the contribution R: (14) V R = ,U(tw + tE) Wi. From Equations (4) and (5): (15) V= W'. Substituting Equation (15) into Equation (14) and solving for 4: (16) =(w +) FINANCING OF THE SOCIAL SECURITY SYSTEM 103 Substituting Equation (16) into Equation (13) yields: (17) - R 1 1 (1 - tw) - [(1 + pt~) + tw (Es - 1)]++(p/ )(1 + tE) ( /$)+ (t. + tj The fraction A/R shows the fraction of the financing of the sys- tem that is borne by labor. If the fraction equals 1, all the financing of the system is borne by labor; if it equals zero, all the financing is borne by employers. Special cases Some results for A/R under extreme values of the parameters are presented here. CASE WHEN j = 0. The expression collapses into: 1 1 A _ (1 - tw) - (1 - t) +(13/a) (1 + t,) (a /)+ ( R (tw + te) If the supply of labor is perfectly elastic, a = x, then: (I 9) A _ (1 - t.) - (1 --t~) 0 (19) R (tw + t,) and all the financing is borne by employers. If a = 0, then Equa- tion (8) can be rewritten: (20) ~A (I (- t~) - (1 ± E -1 R (tw+te) and all the financing is borne by labor. The more inelastic the sup- ply in relation to the demand for labor, the higher the incidence borne by labor. CASE WHEN t = 1. In this case Equation (17) collapses into: (21) R (1 - t t) - (I ± R (t. + t ~) 104 THE DISTRIBUTION OF PUBLIC SUBSIDIES FOR HEALTH Table 4.18. Values of A/R (tt = 0.035; t, = 0.070) a/A I X 10 5 3 2 1 1.0 -1.00 -1.00 -1.00 -1.00 -1.00 -1.00 0.5 -0.50 -0.54 -0.58 -0.62 -0.66 -0.75 0.25 -0.25 -0.31 -0.37 -0.43 -0.49 -0.62 0.10 -0.10 -0.18 -0.25 -0.32 -0.39 -0.54 0 0 -0.08 -0.16 -0.24 -0.32 -0.49 In other words, if employees value the services of the system in an amount equal to R, the incidence of the financing is borne completely by labor and independently of the values of a and ,S. POSSIBLE VALUES RELEVANT FOR COLOMBIA. Workers affiliated with icss make a legal contribution to the system of approximately 3.5 percent of their gross wage. The contribution of employers is approximately twice as much-7 percent of the wage paid to em- ployees. Table 4.18 provides values for A/R for alternative values of a/lB and A. The legal incidence, t,/(t,. + t,) = 0.33, will be larger than the real incidence for values of , below 0.25 and for ratios a/13 larger than 2. A value of u between 0.25 and 0.5 and a range of a/, be- tween 2 and 5 yield a range of A/R between -0.37 and -0.66. A value of -0.50 will be used. Chapter 5 The Distribution of Consumption of Public Utility Services: Electricity, Piped Water, and Sewerage THE SUBSIDY RECEIVED by lhouseholds from consuming electricity, water, and sewerage services is not derived here. As mentioned in Chapter 1, this would require an estimate of the long-run marginal cost of providing the services in different regions. Such an estimate can be derived only from micro-studies of the cost of expanding the system of supply of these services in different regions of the country. This chapter addresses a different set of issues concerning the distributive direction of consumption across income groups, using information from the 1974 sample survey. The first section ana- lyzes the distributive direction of investment in these sectors between 1970 and 1974; then it discusses the availability of services in 1970 and in 1974 and the association between availability, income, and region. The second section deals specifically with the availability of services in urban areas. It attempts to identify the factors determining the probability of having the service and the probability of obtaining the service in a certain time period. The third section addresses similar questions for houselholds located in the rural areas of the country. Some earlier estimates of the re- distribution of transfers across consumers resulting from the system of tariffs in these sectors are reviewed in the appendix to this chapter. 105 106 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Investment and the Availability of Services Investment in electricity (generation and transmission) be- tween 1970 and 1974 fluctuated between 4.5 and 7.6 percent of total government expenditure, the higher figure corresponding to 1974. In the same period investmenit in piped water and sewerage fluctuated between 1.7 and 2.7 percent of total government ex- penditure, the figure for 1974 amounting to 2.5 percent. For 1974 the total investment in these sectors represented 10.1 percent of all government expenditure. What was the distributive direction of this investmlent? An approximate answer can be obtained by looking at the distribution of the houselholds that received these services during this period. Distributive directioon of investment, 1970 to 1974 Table 5.1 presents data from the 1974 sample survey on1 the distribution of households that reported having received these services between 1970 and 1974. Because of the sample size, results are given for the total country, not for regions. The results show that between 45 and 55 percent of the house- holds that received the services belong to the poorest 40 percent of houselholds and that between 72 and 82 percent belong to the poorest 60 percent. Many of these houselholds are in rural areas: 58.6 percent of those receiving piped water and 49.6 percent of those receiving electricity. The large share of rural households among those whiclh received piped water seems surprising, given the nature of the service. It might be explained by the existence of small towns in rural areas,' or it may result from the way availability was defined in the ques- tionnaire. Availability of piped water was defined as a situation in whiclh the dwelling is connected to an aqueduct or to a primary network of piped water. In rural areas, therefore, it would include all househlolds with any type of access to an open aqueduct. A rough estimate of the implicit mean cost of connection can be made by comparing the absolute increase in connected houselholds 1. Rural areas include towns of fewer than 1,500 inhabitants. INVESTMENT AND THE AVAILABILITY OF SERVICES 107 Table 5.1. Distribution of Households Tlhat Obtained Public Utility Services between 1970 and 1974 (percentage) Income quintile Piped Street (poorest to richest) Electricity water Sewerage lighting 1 26.0 31.4 24.5 25.6 2 25.1 23.6 21.3 24.4 3 25.6 26.8 29.1 22.5 4 12.0 12.6 12.0 11.9 5 11.3 5.6 13.1 15.6 Percentage of households in rural areas 49.6 58.6 18.3 0.0 (estimated by expanding the survey results) with investment between 1970 and 1974. Given the lumpy nature of investment in these sectors, however, it is difficult directly to associate the aggre- gate investment figures with the increase in the number of bene- ficiaries over so brief a period. To correct partially for this in the case of electricity, only investment in subtransmission and dis- tribution is included-the assumption being that this component of investment is more closely associated to yearly increases in the number of beneficiaries than other components of investment. Tables 5.2 and 5.3 show data on investment between 1970 and 1974. The total figure for the four-year period was computed by including all of the figures for 1971, 1972, and 1973 and half of the figures for 1970 and 1974. The last column presents these figures in 1974 prices: 2,866 million pesos for electricity (subtrans- mission and distribution) and 6,022 million pesos for water and sewerage. By dividing the investment figures by the number of households connected in the period, the mean investment per household can be computed. The number of households connected is derived from the sample survey and also, in the case of water and sewerage, from independent official sources. These data are presented in Table 5.4. The investment per household in water and sewerage is com- puted using the sample data on connected households (Estimate 1) and an extreme figure from official sources (Estimate III). The official Estimate Il is almost exactly equal to the sample 108 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Table 5.2. Investment in Electricity, 1970 to 1974 (millions of current pesos) Service 1970 1971 1972 1973 1974 Generation 549 1,982 1,177 2,703 3,388 Subtransmission and distribution Urban 295 184 250 538 714 Rural 101 63 86 185 244 Total 396 247 336 723 958 Subtransmission and distribution at 1974 pricesa 797 445 545 998 958 Total for 1970-74 1/2(797) + (445) + (545) + (998) + 1/2(958) = 2,866 a. The implicit price deflator of gross fixed investment was used to derive the figures in 1974 prices. Source: Data from Departamento Nacional de Planeaci6n, Colombia. Table 5.3. Investment in Piped Water and Sewerage, 1970 to 1974 (millions of current pesos) Institutions 1970 1971 1972 1973 1974 P'ublic utility companies Bogota, Cali, Medellin, 230 356 633 841 1,141 and Barran- quilla Other compa- 18 29 97 144 121 nies Total 248 385 730 985 1,262 INSFOPAL 198 136 198 246 243 I.N.S. 41 100 121 153 100 Total at current prices 487 621 1,049 1,384 1,605 Total at 1974 prices, 981 1,118 1,701 1,910 1,605 Total for 1971- 1974 1/2(981) + (1,118) + (1,701) + (1,910) + 1/2(1,605) = 6,022 a. The implicit price deflator of gross fixed investment was used to derive the figures in 1974 prices. Source: Data from Departamento Nacional de Planeaci6n, Colombia. INVESTMIENT AND THE AVAILABILITY OF SERVICES 109 Table 5.4. Number of New Households Connected to the Service and Putblic Investment per Hoiusehold between 1971 and 1974 Estimates of the number of Investment new households (thousands) Investment per between 1970 household and 1974a Sample Official figuresb (pesos) (millions Service of pesos) I II III I III Piped water 1910 212] 2717 Pie ae 6,022 148 2152 2 243 40,689 24,782 Sewerage J 104J 91J 216 Electricity 2,866 123 23,301 a. From Tables 5.2 and 5.3. b. The two official estimates (II and III) are derived as follows: Estimate 11: According to INSFOPAL, 526,000 and 445,000 additional individuals in urban areas were served by piped water and sewerage, respectively, between 1970 and 1974. With an average family size of six, this implies 88,000 and 74,000 additional households, respectively. According to the sample survey, 41.4 percent of the additional households served by piped water are located in the urban area; this yields an implicit country total of 212,000 households. In the case of seswerage, the sample survey reports 81.7 percent of new households served as being in the urban area; this yields an absolute increment for the country of 91,000 households. Estimate HI: According to "El sector de Acueducto y Alcantarillados," Documento D.N.P., June 1976, approximately 2 and 1.6 million additional individuals received piped water and sewerage services, respectively, between 1970 and 1975. The number of households connected between 1970 and 1974 was estimated by multiplying these figures by 0.8 and then dividing them by 5.9, the weighted average family size of the households connected, according to the survey. survey figure. Because investments in water and sewerage are only available as an aggregate, the estimates on investment per house- hold are derived by using the inean figures, 148,000 and 243,000 newl houselholds, respectively, whiclh yield mean investment figures of 24,782 and 40,689 pesos. For electricity the estimated figure is 23,301 pesos. These are country averages and do not provide separate information on investment per household in urban and rural areas. Althouglh it is possible to obtain a breakdown of in- vestment figures according to location, the sample of new house- holds connected was too small to provide reliable expanded figures at the rural and urban levels. Availability of services and its association with income and region, 1970 and 1974 Table 5.5 presents data from the sample survey on the percent- age of families having services in 1974. Figures in parentheses are preliminary results from a 4 percent sample of the 1973 census of 110 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Table 5.5 Families with Services, 1974 (percentage) Piped Garbage Street Electricity water Sewerage collection lighting Large cities 98.9 95.1 90.9 82.4 97.3 (93.6) (90.0) (86.4) Intermediate cities 94.3 90.1 77.2 70.3 88.9 (86.0) (84.7) (69.0) Small towns 72.4 79.8 61.9 46.4 78.4 (69.6) (74.5) (50.7) Rural areas 15.6 19.9 5.1 -a -8 Country 62.7 63.2 51.3 - Note: Values in parentheses are preliminary figures from a 4 percent sample of the 1973 census of population and housing. a. Service is not available in this location. population and 1ousinIg, whiclh was carried out albout eiglhteen mlontlhs before the samlple survey. Even the lower bound estimate of tlle samuple survey (using a confidence interval of 95 percent) yielded a larger value than the census estimate. The difference miglht be explained by an increase in coverage between thie times the census and the survey were made. Anotlher and perhlaps more important source of difference miglht lie in the reporting of illegal connections. Probably house- hlolds withl illegal connections were less willinig to report tlhem to census interviewers thlan to survey interviewers.2 Tables SA-27 thlrouglh SA-32 in the statistical appendix slhow the proportion of lhouselholds witlh services for eachl quintile in thle regional distribution of income. The probability of hiaving a service is clearly associated witlh incomiie level and location. For a given inconlie bracket, thle proportion of hiouselholds witlh services in- creases if these lhouselholds are located in larger cities. LINEAR PROBABILITY FUNCTIONS. To demonstrate the indepen- dent effect of per capita incomiie and location, linear probability functions were estimated for 1970 and 1974. The purpose is to explain a binary dependent variable: the lhouselhold eitlher has or 2. This argument applies to electricity; it is harder to think of situations of illegal connection for the other services, especially sewerage. INVESTMENT AND THE AVAILABILITY OF SERVICES 111 does not have a particular public service. If individual (not grouped) data are used, the simplest technique is to use ordinary least squares, where a dependent variable y is 1 if the household has the service or 0 if it does not. The 1, 0 nature of the regressand enables the conditional expectation of y, given the vector of in- dependent variables X, to be interpreted as the conditional prob- ability that the event will occur given X, P(X). (1) E(y/X) = P(X) = 00 + 1lX1 + f2X2 ... The approach summarized in Equation (1), the linear prob- ability function, allows the regression coefficients to be inter- preted as the marginal contribution of the respective independent variable to the probability.' The interpretation of these coefficients must take into account that P(X) is not constrained to the unit interval; wlhat becomes relevant is to what extent the range of observations wlhose behavior is being explained lies in the interval. In this particular case a bivariate regression, PJ(x), becomes relevant, wlhere J refers to the region and x to the houselhold per capita income. After several experiments, the inverse form was adopted (based on the I-ratio of the coefficient of the income variable): (2) Pi = a + b (-). A theoretical advantage of this function is its asymptotic property if a < 1, in other words, P has an upper bound of less than 1 when bj has the correct (negative) sign. In this case the income elasticity of the probability becomes equal to: EJ=cdPJ bJ (3) Ej =x dP _ b dx x wlhere EJ is defined as the change in P wlhen per capita income clhanges by 1 percent. EJ declines with income, that is, it is in- versely proportional to x. Table 5.6 presents the estimates of aj and bj for each region and each type of service for 1970 and 1974. Although the t-ratios 3. Generalized least squares estimation should be used because of heteroskedasti- city: that is, the variance of the disturbance term depends on X. 112 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Table 5.6. Estimated Regression Coefficients in the Linear Probability Function Constant' Coefficient of (I Ix) b Inter- Inter- Large mediate Small Rural Large mediate Small Rural Service cities cities towns areas cities cities cities areas Sewerage 1970 0.93 0.84 0.65 -16.9 -34.4 -19.9 -o (5.3) (5.8) (5.1) 1974 0.95 0.86 0.70 -15.7 -32.5 -19.6 - (5.5) (5.6) (5.1) Piped water 1970 0.96 0.94 0.83 0.17 -15.6 -21.4 -15.4 -8.6 (6.0) (4.8) (4.5) (4.3) 1974 0.98 0.94 0.85 0.22 -12.3 -14.1 -13.4 -6.1 (5.8) (3.3) (4.2) (2.6) Electricity 1970 0.98 0.96 0.79 0.12 -6.2 -16.2 -15.2 -8.8 (3.3) (4.3) (4.2) (3.0) 1974 0.99 0.98 0.81 0.21 -2.3 -13.3 -10.9 -10.6 (2.2) (4.1) (3.2) (4.9) Street lighting 1970 0.98 0.89 0.75 -10.9 - 26.5 -16.8 -e (5.0) (4.9) (4.5) 1974 0.99 0.93 0.83 -6.2 -16.8 -11.3 -c (3.9) (3.8) (3.9) Garbage collection 1970 0.87 0.75 0.54 - 21.1 -36.4 - 19.9 -c (5.7) (5.5) (5.1) 1974 0.87 0.80 0.58 - 19.6 -39.4 - 25.0 (5.2) (6.2) (6.4) Note: Values in parentheses show the t-ratio of the coefficient. a. The t-ratio of the constant term is always larger than 10. b. x measures monthly per capita household income in Colombian pesos. c. Service is not available in this location. are significant (sample sizes vary from 700 to 1,200 observations), the R2 were extremely low; less than 5 percent.4 4. The interpretation of the correlation coefficient is ambiguous in the binary dependent variable case, particularly if the function is curvilinear. For a full treatment, see J. Netter and E. Scott Maynes, "On the Appropriateness of the Correlation Coefficient with a 0, 1 Dependent Variable," Journal of the American Statistical Asso- ciation (June 1970). Figure 5. 1. Probability of a Household Having Electricity, as a Function of Income Probability 1.0 1974 190 o Large cities 1970 _ _ _ _ _ _ t Intermediate 1970 cities 0.9 1974/ / 1974 0. 8 _ 1970 Small towns 0.7 / 0.6 - 0.5 _ 0.4 - 0.3 - 0.2 1974 , Rural areas 0.1 … 1970 50 100 150 200 250 300 350 400 450 500 Annual per capita income ($) 113 114 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Figure 5.2. Probability of a Household Having Sewerage, as a Function of Income Probability 0.95 - 1974 - - - - - ~Large cities 0.9 _ / Intermediate 0.8 _ 1 cities 0.7 - 1974 - - ~- - }Small towns 1l / ~ ~ ~~~~~ 1970 0.6 - / 0.5 _ 50 100 150 200 250 300 350 400 450 500 Annual per capita income ($) Tables SA-33 through SA-37 in the statistical appendix show estimates of P for alternative dollar values of annual per capita household income. In each table those values are shown for a particular service in different locations and for 1970 and 1974. They are also presented in the text in Figures 5.1 to 5.5. INVESTMENT AND THE AVAILABILITY OF SERVICES 115 Figure 5.3. Probability of a Household Having Piped Water, as a Function of Income Probability 1974 0.96 - X Large cities 0.94 - 1970 0.92 - ~~~~~~~~~~~~~~~Inter mediate 0 . 92 - - -= cities 0.84 - 1974 084 - ! ,Small towns 0).82 - X - 0.80 - / 0.78 - ().76 - 0.72 - 0. 70 - . I I I I I I I I I I, 50 100 150 200 250 300 350 400 450 500 Annual per capita income ($) INTERPRETATION OF RESULTS. The relation between the prob- ability and income appears stronger in intermediate cities and small towns. It shows that lower-income groups in these locations have relatively less access to services than do their counterparts in large cities. The relation appears to be weaker in the case of Figure 5.4. Probability of a Household Having Garbage Collection, as a Function of Income Probability 0.90 1974 Large cities 0.80 1974 / Intermediate cities 0. 70 7~~~~~~~~17 |~ 1-Small towns 0.50 -1 X_ I// 0.40 - 50 100 150 200 250 300 350 400 450 500 Annual per capita income ($) 116 INVESTMENT AND THE AVAILABILITY OF SERVICES 117 Figure 5.5 . Probability of a Household Having Street Lighting, as a Function of Income Probability 1.00 _ 1974 9 '.Large cities O0. 90 / f Intermediate / ~~~~~~~~~~1974 0.80 (Small towns / ~ ~~~~~~ - 1970- 0.70 - 1// 0.60 I I I I i I I I I I 50 100 150 200 250 300 350 400 450 500 Annual per capita income ($) electricity, particularly in large cities. As shown in Figure 5.1, the probability of having electricity in large cities is more than 95 percent, even with a per capita income of $50 a year. (The re- ported mean per capita income of the poorest 10 percent of house- holds in those cities is $59 a year.) 118 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES The weaker relation found in the case of electricity is to be ex- pected. The possibility of illegal connection, a widespread practice in Colombia, is easier for electricity than for other services; more- over, the availability of the service or the location of the supply is usually more evenly distributed among all income levels than for the other services. Investment in transmission of electricity is cheaper than in the other sectors and usually has been given pri- ority in public policymaking. Finally, on the demand side, income has less effect on the use of electricity when it is available at a normal connection cost. For other services, especially piped water and sewerage, the network can be used only after making a mini- mum investment in housing. Public Services and Substitutes in Urban Areas In undertaking a more comprehensive multivariate analysis of variables associated with the availability of services across house- holds, it is useful to develop a framework in which the association can be interpreted as a cause-effect relation. The identification of the causal variables that determine the availability of public services to households becomes important in (a) understanding the determinants of the distributive effects of those services to the extent that they provide a substantial con- sumer surplus in relation to alternative (nonpublic) sources of supply' and in (b) identifying those causal variables that can be manipulated by public policy. Availability of services in a supply-demand context In developing such a framework, it is useful to distinguish between the two situations in which a particular household does not consume a service: either because the supply network is geo- graphically inaccessible (the connection cost to the family has an infinite price) or because the household could be connected at a certain cost, but decides not to do so. 5. Obviously a comparison between the publicly supplied service and the alternative substitute ought to be carried out in efficiency units of the service. This is particularly true when strong differences in quality exist between sources: for example, light and candlelight. PUBLIC SERVICES AND SUBSTITUTES IN URBAN AREAS 119 The second situation is basically determined by demand and results from a voluntary choice by the household. Because the factors determining this choice (or the variables behind the de- mand) may be different from the variables determining the loca- tion of the network (or the variables behind the supply), this supply-demand mechanism behind the availability of services should be recognized explicitly. The empirical significance of situations of no consumption because of demand considerations becomes evident from the follow- ing figures for urban households: Piped Percentage of families without the service Electricity water Sewerage To whom the service was offered 10 16 11 With a neighbor with service less than one block away 73 54 22 Both situations are alternative concepts of being on the supply network. The first is a more strict definition, since connection is possible only if the service is offered by the public utility company. Under the second definition, connection is possible whenever the household is less than a block from a neighbor with the service.6 Households without services because of voluntary choice (lack of demand) represent a substantial fraction of the total families without services, particularly for electricity and piped water. Understanding what determines this choice becomes important in explaining the different availability of services across families. REFORMULATING THE LINEAR PROBABILITY FUNCTION. Direct estimation of P(X), the probability of having the service as a function of a vector X, does not fully capture the supply-demand mechanism discussed above; it does not identify the extent to which a particular variable or characteristic, xi, influences P through the demand or supply side. For example, in the earlier results, where the availability of the service is regressed on the per capita income of the household, 6. When poor neighborhoods are located in parts of the city below river or sea level, sewerage connection becomes difficult; when they are located higher than reser- voir altitudes, connections to the piped water system become prohibitive because water would have to be pumped. In both cases, these areas may be close to areas with services. 120 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES it is not possible to distinguish to what extent the income variable operates througlh supply or demand. On the one hand, the level of houselhold income determines the behavior of the public utility companies concerning expansion of the network in particular neighborlhoods. On the other hand, income affects the houselhold's decision to get connected, however, if connection is technically possible. The problem can be specified by considering P as the product of two independent probabilities: the probability of a houselhold being on the supply network of the service, ps, that is, the prob- ability of having access to the network at the connection cost set by public utility companies, and the probability of accepting or demanding the service, Pd, if offered at this connection cost, that is, demanding the service when the houselhold is on the supply schedule. This specification can be written as: (4) P (XI, X2) = Ps (X1) pd (X2) According to Equation (4) the probability of consuming the service, P, is equal to the probability of being on the supply sched- ule, Ps, multiplied by the probability that the houselhold will demand the service at the cost of connection, Pd. Both PI and Pd can be thouglht of as functions of vectors of variables Xi and X2, respectively, that are estimated independently. Xi includes variables determining the utility company's policy on locating the supply network. X2 measures demand-oriented variables: that is, the cost of connection relative to the income of the household and other socioeconomic characteristics of the houselhold that govern this demand. Possibly the vectors Xi and X2 will have common elements, that is, variables affecting both supply and demand. In this sense, supply and demand are not completely independent; if the location of expansions of supply is affected by the possibility of households getting connected, some of the variables entering Pd will also enter ps with the same sign. EMPIRICAL ESTIMATION. The empirical estimation of p, Pd, and PI can be illustrated as follows. Divide the total number of house- holds (the area of the square below) between the number of house- holds with the service, A, the number witlhout the service that PUBLIC SERVICES AND SUBSTITUTES IN URBAN AREAS 121 are on the supply schedule of the service, B, and the number without service that are not on the supply schedule, C. Thus, B + C is the number of households without the service. Households with the service (On the supply and demand) A Households without the service (On the supply) |(Not on the supply) B C Estimation Estimation Estimation of P of Ps of pd Sample size A + B + C A + B + C A + B Definition A=1 Ai = Ai = 1 of dependent B 0 Bi = 1 B, 0 variable Ci =0 C = 0 Mean A A + B A probability A +B +C A +B +C A+B3 The probability of having the service, P, is estimated by defining a binary dependent variable, where a family with the service gets a value 1 (Ai = 1) and a family without the service a value 0 (Bi = 0, Ci = 0). In estimating Ps, all the families who are on the supply schedule receive the value 1 (Ai = 1, B, = 1) and all that are not, a value of 0 (Ci = 0). Pd is estinmated with a subset of the households: that is, only those on the supply schedule, A + B. The families on the supply schedule who do demand the service receive the value 1 (A = 1); those who do not demand it receive a value 0 (Bi = 0). The prod- uct of the mean value for PI and Pd yields the mean value of P. In the above framework, the contribution of a variable xi to the total probability P is different if traced through P, and Pd than if it is measured by estimating P directly. Assume that the following linear probability functions for P, and pd are estimated: (5) PI(X1) = ao + a1x1 + a2x2.... (6) Pd(X2) = bo + bix1 + b2x2 .... The contribution to P of a variable xi becomes, by using Equa- 122 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES tion (4), equal to: OP OP, apd (7) -d= a Pd (X,) + Ps (X1) axi X i axi (8) aP = a,Pd (X2) + biPs (Xl). axi It is clear that aP/axi is not a constant, but is itself a function of the vectors X2 and X1. Direct specification of P as a linear probability function of all the variables included in XI and X2 can be shown to be inconsistent with Equations (5) and (6). For example, let P be specified as: (9) P(XI, X2) = Co + ClXl + C2X2 ... In this case the effect of xi on P, namely OP/Oxi = ci, is constant and independent of X1 and X2, a result contradicting Equation (8). The contradiction results from the fact that a linear specification of PI and Pd cannot yield a linear specification of P, a product of ps and pd. The correct direct specification of P is a quadratic function with interaction terms between the independent variables. From these considerations it can be concluded that it is not pos- sible directly to compare ai and bi with the parameter ci, which is obtained by direct estimation of P as a linear function. p, ps, and Pd will be estimated by defining a family on the supply schedule as one having a neighbor with the service at a distance of less than one block. This definition reasonably reflects a situation where households could effectively be connected if they were willing to pay the connection cost. The alternative definition, households to whom the service was formally offered, was considered too strict and limiting. In addition to the size of city variable defined before, the follow- ing explanatory variables are included: dirt floor in the dwelling; rural migrant (households who lived in the rural area in 1970, 4 years before the interview); logaritlhm of the monthly per capita income of the houselhold in pesos of November 1974; logarithm of the years of schooling completed by the head of houselhold; loga- rithlin of the age of the head of household; and logarithm of the num)bers of years that the houselhold has lived in the municipality. All continuous variables are included in their logarithmic form. PUBLIC SERVICES AND SUBSTITUTES IN URBAN AREAS 123 This specification was usually best in terms of the t-ratios of the coefficient. The rest of the variables-city size, dirt floor, and rural migrant-are included as dummy variables. When a continuous variable is entered in a log form in the linear probability function, its coefficient, given the 0-to-1 nature of P, can be interpreted as an elasticity. This elasticity is defined as the change in the probability when the independent variable changes by 1 percent.7 RESULTS FOR ELECTRICITY. Table 1.9 (in Chapter 1) shows the estimates for P, pd, and Ps for four population groupings: total urban population, the poorest 40 and the poorest 20 percent of families according to household per capita income, and population in small towns. At the bottom of each grouping are the mean values of P, Pd, and Pl. The product of the mean values for Pd and PI yields the mean value for P. The size of city dummies shows that small towns have a sub- stantially lower value of P than do the intermediate cities. This difference, however, is basically a demand effect as summarized by the coefficient of the small towns dummy in Pd This coefficient is four times the coefficient for small towns in P', independent of the sample group being used. This result is of interest in the sense that the lower value of P for small towns is not associated with discrimination by the public utility companies (location of the network), but with factors behind the demand for the service.8 The probability of having the service is highly sensitive to the dirt-floor characteristic of the dwelling. If P, and Pd are estimated separately, the effect is significant in both supply and demand, particularly the latter. Dirt floor is perhaps a good substitute for the permanent income of the family, a variable that would influ- ence Pd To find a significant and strong coefficient in Ps would mean that network location is negatively associated with that characteristic of households. It reflects a situation where network expansion by the electricity companies depends on the perceived probability of the household becoming connected and where dirt floor, as viewed by these companies, indicates this probability. It o9p 7. If P = ao +a ilogxj...., then a] = a log xl 8. The interpretation rests heavily on accepting the definition of being on the supply schedule. 124 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES could also reflect such institutional constraints on the company's expansion policies as the lack of legal status of the dwelling, wlhicl is often associated witlh the dirt-floor variable. Log-of-per-capita-income is significant in P and pd for the total urban population and small towns. It is not significant wlhen the population is stratified by income group, since it diminislhes the variance of the income variable. This variable is particularly strong on the demand side in small cities; the coefficient of 0.11 means that doubling the per capita lhouselhold income will increase the probability of demanding the service by 11 percentage points. It is interesting that per capita income is significant in pd, but not in PI, wlhiclh suggests that altlhouglh it influences the demand for the service, the public utility companiies do not perceive it to indicate probability of connection (to the extent that probability affects the location of supply). Thie rural-migrant clharacteristic is more statistically significant and lhas a stronger effect in pd than in Ps. The effect is particu- larly strong witlhin low-inicomiie groups, particularly in the poorest 20 percent of the urban population. Log-of-sclhooling and log-of-age-of-lhead-of-lhouselhold-variables that theoretically operate tlhrouglh demand-sometimes also appear significant on the supply side, wliclh could reflect the impossibility of separating supply and demand, given the nature of the sample. RESULTS FOR PIPED WATER. Table 1.10 shiows that if the effect of otlher variables is hield constant, the effect of city size becomes less important for piped water than for electricity. The difference between intermlediate cities and large cities appears to be basically a supply plhenomenon; the difference between large cities and small towns is basically a demand phenomenon. The effect of dirt floor is again extremely strong and significant on bothi the supply and the demand sides. It presents a pattern similar to that of electricity. The rural-migrant clharacteristic appears to affect demand, becoming insignificant on the supply side. The income variable operates basically on the demand side, but with less strengtlh than for electricity; the same is true for years-of-schooling-of-lhead-of-lhousehiold. Years-in-tlhe-same-mllunici- pality is significant on thie demand side and lhas a negative sign. One way of interpreting this result is that the longer the hlouselhold remains withlout the service, the lower the probability that the lhouselhold will demand the service today. PUBLIC SERVICES AND SUBSTITUTES IN URBAN AREAS 125 RESULTS FOR SEWERAGE. Table 1.11 shows similar city size dummies for intermediate cities and small towns in the results for P. For intermediate cities it basically operates on the supply side; for small towns it operates through both demand and supply. The dirt-floor characteristic is important, especially on the supply side; it is substantially stronger than in the case of the other two services. As expected, the income variable operates basically througlh demand. On the other hand, the schooling vari- able and the years-in-the-same-municipality variable appear more significant on the supply side; the reasons for this type of unex- pected result are discussed below. PROBLENI OF LACK OF VARIATION. The data used in the earlier analysis did not provide enough variation in the dependent vari- able. As suggested by the mean values, most observations had the value 1: that is, households witlh the service. Mean values for the total urban area are usually larger than 0.80 and were frequently larger than 0.90. This lack of variation results from the basic objective of this study-to provide statistically significant figures for the avail- ability of services by major breakdowns of the population. The sample survey was designed for this purpose and not to provide the degree of variation required for multivariate analysis. The large number of households with services sampled not only affects the overall variability of the dependent variable, but also influences the possibility of separating supply and demand: that is, the factors behind pd from the factors behind Ps. Households withl the service (A) are on the supply schedule and demand the service: that is, in estimating both Ps and Pd, they have the value 1. This common set of houselholds is large enough to prevent the necessary variation between the data used to esti- mnate pd and those used to estimate Ps. Further, a large value of A makies the results very sensitive to the size of B (households without the service, but on the supply schedule) relative to C (houselholds without the service and not on the supply schedule). The relatively small number of houseliolds withlout the service makes the estimates of P3 and pd sensitive to the distribution of these houselholds according to their position on the supply schedule. Smaller mean values of P can increase the possibility of sepa- rating demand from supply; this is shown when the results for the 126 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES total urban areas are compared witlh those for smiiall towns. In small towns the fraction of households with services is substan- tially smaller than in the total population. The variables that miglht influence demand (rural-migrant, per-capita-income, and years-of- schooling) become more significant in pd and less significant in PI when moving from the total urban estimates to the small towin estimates. OMITTING THE COST OF CONNECTION VARIABLE. Theoretically, variations in the cost of connection should help to explain varia- tions in Pd. The sample survey did not yield information on the connection cost actually paid by households with the service or the cost faced by a houselhold witlh the possibility of getting con- nected. The possibility of determining the effect of the cost of connection depends on finding enough variation in this variable across house- holds. For urban Colombia, the cost structure for each service tends to vary across cities. Generally speaking, two cost compo- nents can be distinguished: (a) a component reflecting the direct cost of installation, that is, labor and the cost of the meter, and (b) a fee, whiclh for water and sewerage is usually some function of the assessed value of the property and which for electricity is a function of meter capacity.9 Wlhat is relevant here is the extent to which the omission of the cost variable biases the coefficient of the variables that have been included: that is, the missing variable bias. This bias occurs if there is a significant association in the sample between the omitted variables and the included variables. If the coefficient of the cost variable in the demand equation is presumed to be negative, the sign of the bias will depend on the sign of the above association. If the association is positive, the estimated coefficient of the in- cluded variable will have a negative bias; if it is negative, the bias will be positive. If the cost of connection increases with property values, it will also increase with household per capita income. This suggests a negative bias in the income coefficient estimated in the linear 9. For a detailed analysis of the structure of connection cost, see Johannes F. Linn, "Public Utilities in Metropolitan Bogota: Organization, Service Levels and Financing," Urban and Regional Economics Division, Development Economics Department, (Washington, D.C.: World Bank, May 1976; processed). PUBLIC SERVICES AND StUBSTITUTES IN URBAN AREAS 127 probability functioni; the expected value of the estimate is lower thani the coefficient in the true model. In other words, the esti- mates of the income elasticities in these functionis tend to under- estimate the true coefficient. Thle m1agn1itude of tfle bias depends not only on the magnitude of the association between the omlitted and included variables, hut also on the size of the coefficient of the omitted variable in the true theoretical model: that is, the size of the cost-of-connection elasticity. If this (price) elasticity is, itself, a function of per capita inconlle, it is expected to be highler in absolute terms for lower inconmes. The bias described above will differ wheni the sample is brokeni down by income groups; it will be larger for the estimates derived for lower-income quintiles. Changes in the av(zilabilitv of sermices betwzeen 1970 and 1974 TIhe chanige over time in the availability of services to different income groups can be used to indicate the distributive direction of public investnmenit in these sectors. SHARE TO LOW-INCOMIE GROUPS. The 1974 samiiple survey gives information on1 whletlher a particular url)al liouselhold had a par- ticular service in 1970. These data allow the computations shown in Table 5.7, where househiolds wlho became connected to a par- ticular service between 1970 and 1974 are sh1owII according to their position in the urban incomiie distribution. For electricity, water, and sewerage, 60, 70, amid 55 percent of these househlolds, respectively, belonged to the two lowest-inicomiie quintiles. The distributive direction of investnment has been progressive in the sense that lower-incomiie groups have had a larger share in the expansion of the total availability of these services. This com- parison, however, ignores the fact that most households who did not hlave a service in 1970 belonged to the lower-income groups in the first place. It mighit well be that altliougil investment has, in one sense, favored lower-incomiie groups, the effect has been small relative to the potential number of beleficiaries: that is, relative to the initial number of househiolds without the service. Table 5.8 slhows the number of househlolds that received a service between 1970 and 1974 as a percentage of the lhouseholds witlhout a service in 1970. Flie results are again presented ac- cording to city size and income group in the urban income distribution. 128 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Table 5.7. Distribution of Urban Households That Became Connected between 1970 and 1974 (percentage of all households with the service) Income quintilea (poorest to Piped richest) Electricity water Sewerage 1 31.1 34.3 27.7 2 28.4 35.4 27.5 3 9.4 11.4 13.8 4 13.5 8.8 14.9 5 17.6 10.1 16.1 a. Quintiles in the urban income distribution. Table 5.8. Urban Households That Became Connected between 1970 and 1974 as a Percentage of the Households without the Service in 1970, ClassiJfied by City Size and Urban Income Distribution Inter- Income Large mediate Small Urban groapa cities cities towns average Electricity 1st quintile 61 45 12 23 2nd quintile 73 10 13 20 Upper 60 percent 47 07 14 16 Average 66 08 14 18 Piped water 1st quintile 37 24 12 25 2nd quintile 52 32 22 34 Upper 60 percent 65 19 14 25 Average 48 27 16 27 Sewerage 1st quintile 22 12 11 15 2nd quintile 34 16 07 18 Upper 60 percent 17 06 13 18 Average 28 14 13 17 a. Quintiles in the urban income distribution. PUBLIC SERVICES AND SUBSTITUTES IN URBAN AREAS 129 Table 5.9. The Probabilities of Having Had a Service in 1970 (P70) and of Having Obtained a Service between 1970 and 1974 (II): A Comparison of Regression Coefficients Electricity Piped water Sewerage Explanation variables P70 II P70 II P70 II Constant 0.95 0.69 0.90 0.45 0.84 0.32 (70.6) (5.4) (57.7) (8.7) (44.8) (9.4) Small towns -0.15 -0.49 -0.02 -0.33 -0.18 -0.14 (10.2) (6.8) (1.7) (3.7) (9.1) (3.3) Intermediate cities -0.02 -0.35 -0.06 -0.26 -0.11 -0.15 (2.01) (4.1) (4.0) (3.9) (6.1) (3.3) Dirt floor -0.34 -0.30 -0.34 -0.14 -0.47 -0.14 (16.7) (5.9) (14.5) (2.8) (16.9) (3.9) Inverse of income -6.1 -8.5 - -9.9 (3.9) (4.6) (4.5) Rural migrant - 0.28 - (1970 to 1974) (3.6) Urban migrant 0.17 (1970 to 1974) - (3.3) - - Note: Values in parentheses are t-coefficients. The last column shows that lower-income groups have benefited increasingly in relative terms from investment in electricity. This is not true in the case of sewerage or piped water. If the income bracket is held constant, the increase appears substantially larger for big cities than for intermediate cities and small towns. MARGINAL PROBABILITY OF HAVING A SERVICE. The above con- siderations can be analyzed better if the probability of obtaining the service per unit of time, in this case between 1970 and 1974, is defined as an analogue to the concept of P defined earlier. Con- sequently, H will be defined as the probability that a household living in the urban area in 1974, which did not have the service in 1970 (no matter where it was located), did obtain the service between 1970 and 1974. Table 5.9 shows the estimates of the regression coefficients for II and for P70, the probability of having had the service in 1970. Per capita income was not significant in estimating H, so it was excluded from the final equations presented in the table. 130 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES In estimating H, the concepts of urban and rural migrants are included as dummy variables. Rural migrant is defined as any houselhold that migrated to the urban area between 1970 and 1974; urban migranit is defined as any houselhold tbiat changed dwellings withliln the urban area during that period. EMPIRICAL RESULTS. For electricity and piped water, the dumii- inies for city size are substantially more negative for H than for P70, wliiclh means that being in an intermediate or small city affects negatively the probability of obtaining services over time. In other words, the negative effects hlave been stronger on the mar- ginal probability (H) than on the probability at one point in time (P70)- In interpreting these phenomiiena it is useful to resort again to the concepts of supply and demlanid behind either P70 or H. The negative effect could partly result from investmenit that benefited basically large cities, or froml relatively mlore lhouselholds movinig onto the supply network of the service. Another interpretationl (controlling for the otlher variables in the regression) is that hlouse- holds in large cities that were alreaidy on the supply network in 1970 had a stronger tendency to demlanid the service thani like houselholds in smlaller cities. Thle size of the sample. however, did not allow testing for this particular set of hypotlheses. For piped water and sewerage, the dirt-floor dummiy is less negative for H than for P70. Again, tlis result can be interpreted eitlher as a demand or supply pheniomiienoni or as a combiniationi of the two. If this variable is an indicator of permianienit incomiie (througlh demanid), it suggests that at the margin, the probability of demanding the service becomes less sensitive to incomiie. If dirt floor determines a conmpany's decisions on network location, it would mean that at the margin, dirt floor has had a lower influence on new expansions of the network in particular neighborlhoods. Perhaps the most interesting results for II are the large positive coefficients found for migrants in the case of piped water. The results mean that the probability of having obtained the service in the urban area between 1970 and 1974 is larger for houselholds that migrated during that period from1 rural areas (28 percentage points more than those that did not change their residence), as well as for those that chaniged residence withlinl the urban areas during that period (17 percentage points more thani those that did not change their residence). One explanation for this result could be that migrants are more PUBLIC SERVICES AND SUBSTITUTES IN URBAN AREAS 131 "public service achievers" than nonmigrants are-achievement being expressed in their desire for dwellings with piped water. This is particularly true if the availability of piped water is one reason for rural-to-urban or any intra-urban migration. The fact that this is true only for piped water suggests that the availability of water is a more important reason for migration than the availability of the other services. Earlier, when estimating P for water in 1974, the rural-migrant characteristic had a significant negative coefficient, -0.07 (see Table 1.10). How can this result be reconciled with the positive coefficient found in the estimation of II? The value of P for 1974 can be thought of as the weighted aver- age of P for 1974 for nonmigrants, p,M (that is, households that already were living in the urban area in 1970), and the probability that migrants from rural areas between 1970 and 1974 obtained the service on moving to the urban area, lIM. The weights are the share of the two groups in the number of households in urban areas in 1974.10 A higlher fraction of rural migrants would affect negatively the value of P for 1974 if P74' > flM. This result, however, is perfectly consistent with HI3 > II NA, that is, the probability of a migrant obtaining the service during a period, in this case 1970 to 1974, is larger than for a nonmigrant.1' 10. A distinction should be made between the probability of obtaining the service in the urban area for a migrant household that did and one that did not have the service before migrating. Given the sample size, it was not possible to test for such a difference. The implicit assumption is either that both probabilities are similar or that the fraction of migrant households that did have piped water in the rural area was negligible. 11. Denote P74 as: (1') P74 [1 - (M/T)] p74NM + (MIT) IM, where (M/T) indicates the share of rural migrants between 1970 and 1974 as a fraction of total households in urban areas in 1974. A negative effect of (MIT) on P74 can be defined as: (2') ,3P,4/O(M/T) = - p74NM + 11M < o or (3') p74NM > TM Denoting p74NM as: (4') p74NM = p70NM + (1 - P70NM) 11N = p70NM (1 - IIM) + 1NM* Substituting Equation (4') into Equation (3'), the latter can be rewritten as: (5') p70NM > (fiM -TINM) /(1 -DINM) The result, HM > HNM, found in the estimates of I presented earlier, is consistent with Equation (2') if p70NM is large enough to fulfill the condition (5'). 132 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Urban households without services For future investment policy, it is useful to identify the location of households without services in relation to the location of the supply network. The higher the percentage of households that are relatively near the network, the smaller the marginal investment required to provide them with the service. The more households who do not have a service because of demand factors, the more important are demand-oriented variables as policy instruments. Table 5.10 shows the distribution of households without various services, by city size and by location relative to the nearest point of supply. The unit of measurement used is an urban block, which in the case of Colombia is approximately 80 meters. The figures show that for electricity and piped water, more than half of these households are located less than a block from a neighbor with the service; for sewerage, it is true for one-fifth of the households. There are strong differences when households are stratified by city size. The percentage of households that are near the supply network is substantially higher in small towns than in large cities. This means that in small towns either demand factors are more important in explaining absence of consumption or public utility companies are less responsive in providing the connection even when the service is demanded. Attempts were made to derive information on the reasons for lack of consumption. The head of household was asked to give the most important reason for not having a service in 1974. Table 5.11 shows the distribution of answers for households when the nearest neighbor with the service was less than a block away: that is, for those households where distance from the network was not the obvious limiting factor. As seen by these households, the most important reason for not having a service is its cost. For electricity and water, if the second and third answers are aggregated, more than two-thirds of the households reported the cost of the service, either in itself or rela- tive to the cost of substitutes, as the main reason for not having the service. For sewerage the percentage is more than half. The percentage is significantly larger for electricity than for the other services. Because electricity has the largest coverage, this could reflect the fact that households without this service have a lower per capita income than households lacking other services. PUBLIC SERVICES AND SUBSTITUTES IN URBAN AREAS 133 Table 5.10. Distribution of Urban Households without Services in 1974, Classified by Distance from Neighbor with Service (percentage) Distance from nearest Inter- neighbor with service Large mediate Small Urban (blocks) cities cities towns average Electricity Less than 1 57.4 55.5 79.8 73.4 1 to 3 1.0 10.0 9.2 8.8 More than 3 41.6 34.5 11.0 17.8 Piped water Less than 1 45.3 37.0 67.1 54.1 1 to3 16.9 11.2 9.3 11.5 More than 3 37.8 51.8 23.6 34.4 Se-werage Less than 1 16.5 17.1 28.8 22.4 1 to 3 12.8 10.8 10.4 11.1 More than 3 70.7 72.1 60.8 66.5 Table 5.11. Distribution of Reasons Why Urban Households Are without the Service When There Is a Neighbor with the Service Less Than One Block Away (percentage) Elec- Piped Reasons for no connection tricity water Sewerage Legalstatus 5.1 5.4 7.6 Service too expensive 70.8 56.4 47.3 Cheaper substitutes 1.0 8.9 5.8 Request not answered 3.4 3.6 5.8 Network too far 0 1.0 3.4 Complexity of application 3.4 4.5 3.4 Request in process 8.6 7.9 10.4 Other reasons 7.7 12.3 16.3 Table 5.12 classifies households without services according to the substitutes used. Half of the households without direct piped water purchase water from other dwellings. When these particular households were further classified by the reason for no connection, 134 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Table 5.12. Distribution of Urban Hlouseholds without the Service, Classified by the Substitute Used (percentage) Piped water Well with pump 3.0 Well without pump 20.5 Public standpipe 20.5 Private vendors (trucks) 4.3 Purchased from other dwellings 47.4 Others 4.3 Electricity Kerosene lamps 59.0 Candle 41.0 Waste disposal Septic tank 21.8 Latrine 44.3 Without any service 33.9 53 percent of them answered that the service was too expensive. The result is of interest; although these households do pay for water, they perceive the present cost as lower than the long-run cost of obtaining the water by direct connection. The reason for this is the higlh (fixed) cost of connection relative to household consumption and because of the higlh economies of scale that can be achieved by several households using a single connection. Public Services and Substitutes in Rural Areas Table 5.13 presents data on rural families, classified by type of service being used by the household and by income quintile in the rural distribution of income. The percentage of households con- nected to aqueducts of potable water is 19.9, and the figure is larger for households of higher per capita income. Most of the rural population obtains water directly from wells or rivers without the help of a pumping system. A stronger relation between income and the availability of service is found for electricity and sewerage, although the negative PUBLIC SERVICES AND SUBSTITUTES IN RURAL AREAS 135 Table 5.13. Sources of Services Used by Rural Households, Classified by Income Quintile in the Distribution of Rural Income (percentage) Income From well or river quintile From (poorest From Withozat With other to richest) aqueduct pump pump dwellings Other Water 1 15.3 71.7 0.8 3.2 9.0 2 17.2 63.7 0.7 3.7 14.7 3 18.1 60.3 3.2 2.3 16.1 4 25.8 56.1 2.5 1.0 14.5 5 23.3 54.1 6.7 2.1 13.8 Average 19.9 61.2 2.8 2.5 13.6 Electricity (public) Kerosene Candle Light 1 8.0 61.5 30.5 2 10.8 50.8 38.4 3 16.0 43.0 41.0 4 15.0 40.2 44.8 5 20.6 41.7 37.7 Average 14.2 47.4 38.4 Septic Sewerage tank Latrine None Waste disposal 1 2.1 4.8 22.1 71.0 2 3.1 6.0 23.2 67.7 3 6.4 9.2 20.2 64.2 4 6.9 7.9 28.6 56.6 5 7.1 9.7 23.0 60.2 Average 5.1 7.5 23.4 64.0 Note: Percentages add to 100 across rows. welfare implication of a lack of sewerage in low-density rural areas is not obvious. Probabilities of having and demanding electricity The probability of having electricity as well as the probability of demanding the service was estimated. This was not done for 136 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Table 5.14. Regression Coefficient for the Probability of Having Electricity in Rural Areas Probability Probability of having of demanding the service the service Explanatory variables (P) (pd) Constant 0.27 0.91 (2.5) (3.3) Dirt floor -0.20 -0.24 (10.5) (4.5) Farmers 1 -0.09 -0.06 (4.3) (1.4) Farmers 2 -0.17 -0.40 (5.2) (4.1) Log of per capita income 0.10 0.06 (5.2) (1.5) Log of years of schooling 0.09 0.07 (2.5) (1.1) Note: Values in parentheses show the t-statistic. the other services, because the differences in the quality of services prevent comparison across households (for example, direct connec- tion to a potable water aqueduct relative to a connection through a system of canals). Moreover, in the rural areas, accessibility to the service is probably more a function of institutional constraints than a result of household behavior. Table 5.14 shows the results of estimating P and Pd for elec- tricity in the rural areas of Colombia. Dirt floor, per-capita-income, years-of-schooling, and occupational-status-of-head-of-household are significant variables. In this table farmers are defined as house- holds who live on the plots they cultivate: Farmers 1 are owners or tenants of the plot, and Farmers 2 are sharecroppers. They represent, respectively, 51 and 11 percent of the households living in the rural areas. The remaining 38 percent of households are defined as nonfarmers and are wage labor and individuals living in rural areas but not directly engaged in agricultural activities. The results for P are: (a) dirt floor has a negative coefficient, but less negative than in urban areas; (b) per-capita-income has a coefficient as large as that for small towns in the urban regressions; and (c) the coefficient for small farmers is negative, particularly PUBLIC SERVICES AND SUBSTITUTES IN RURAL AREAS 137 Table 5.15. Rural Households That Became Connected to the Electricity Network between 1970 and 1974 (percentage) As a percentage of all As a percentage of Income quintile households that households without (poorest to richest) became connected the service in 1970 1 17.7 (31.1) 3.1 (23.0) 2 11.1 (28.4) 1.9 (20.0) 3 31.1 (9.4) 5.6 4 22.2 (13.5) 4.0 (16.0) 5 17.7 (17.6) 3.9J Total/average 100.0(100.0) 3.7 (18.0) Note: Similar figures for the urban area from Tables 5.7 and 5.8 are given in parentheses. for sharecroppers. It means that farmers have a lower probability of enjoying a particular service than nonfarmers. In estimating pd, households on the supply schedule were defined as both those with electricity and those without but with a neigh- bor less than 100 meters away having it. They represented 88 per- cent of the total number of households in rural areas. The Farmers 2 variable has a significantly different coefficient in the regressions for P and pd. In estimating P, the coefficient is -0.17, but in pd it is -0.40. This reflects a large difference in the demand for services between the two types of farmers, even when the effects of other variables are held constant. Changes in the availability of electricity between 1970 and 1974 Which income groups in rural Colombia received electricity services between 1970 and 1974? The first column of Table 5.15 shows that most of the households that received the service-31.1 percent-belong to the third income quintile in the rural income distribution. Only 28.8 percent of the households that received the service belonged to the poorest 40 percent of rural households. The percentage was 59.5 for the poorest 40 percent in urban areas. The number of households connected in the period expressed as a percentage of those without services in 1970 appears for each in- come quintile in the second column of Table 5.15. Only 3.7 percent of the rural households without electricity in 1970 received the 138 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES Table 5.16. Rural Households zithout Electricity, Classified by the Distance to the Nearest Neighbor with the Service (percentage) Distance Percentage of (meters) households Less than 100 2.3 100 to 500 5.2 500 to 1,000 3.5 Over 1,000 89.0 service between 1970 and 1974; for the urban areas, the figure was 18 percent. Not onlv does the rural area have a smaller coverage of the service, but the change in that coverage relative to the initial deficit is also smaller than in urban areas. Attempts to estimate the probability of rural houselholds having received the service between 1970 and 1974, H, yielded the follow- ing result: (10) HI = 0.09 + 0.03 Log income (2.6) (2.6) - 0.08 Farmer - 0.09 Dirt floor, (4.4) (6.8) where, as before, farmers are defined as lhouseholds living on the same plots they cultivate. In this case, income becomies significant (it was not in the urban area), and the probability becomles posi- tively associated withl inconme. The probability tends to be signifi- cantly smaller for farmers than for nonfarmers. The data on the distribution of houselholds without electricity according to their distance from the network (Table 5.16) slhow that further connectioni requires substantial investment in distribu- tionI. For 89 percent of these households the nearest neighbor with the service is located mlore thani one kilomleter away. Appendix. Transfers between Urban Consumers Resulting from the Tariff Structure In Colombia, tariffs for electricity and water vary sharply between residential and nonresidential (industrial and commler- TRANSFERS BETWEEN URBAN CONSUMERS 139 cial) users, the tariffs for the latter being substantially higher. Nloreover, the level of residential tariffs differs markedly across cities. Residential electricity tariffs are characterized by prices per kilowatt-lhour that increase in blocks witlh additional consumption. Residential water tariffs consist of two components: a fixed charge per imontlh for consumption up to a certain level and a price per cubic meter of additional consumption above that level. The fixed charge increases with the assessed property value of the dwelling. Thlie price per cubic meter is, with a few exceptions, independent of the property value and increases in blocks with increased con- sumlptioll. Relationz of suibsidies and tariffs Trhe association of propertv values and consumption levels with lhouselhold income suggests that tariffs may have induced subsidies and transfers anrnong consumers classified by income groups. Thle subsidy received by a household in any income group (i) can be written as: (11) Si = (JIC - ti)Qi, wlhere MC is the long-run marginal cost of providing one unit of the service, ti is the tariff charged per unit to the ith income group, and Q, is the amounit of the service consunmed per unit of time. The subsidy (Si) can be divided into two components: the subsidy derived from charging the income group a particular tariff below the mean tariff (t) and the subsidy received by lhaving a meani tariff below the long-run cost. (12) Si = (t - t)Qi + (AIC -T)Qi The subsidy as a fraction of the household's income (Y) can be written as: (13) _i - t ) Q it (3AIC t) Qit tj y tj y (14) Si - ai + t(t -M wvhere ai is the share of income being spent at present in the con- sumption of the service. 140 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES If MC = t, the subsidy received by a consumer is financed by another consumer paying a tariff higher than the average. In this case, the tariff policy basically generates transfers across con- sumers or transfers within the system. If MC > 7, the system as a whole is being subsidized: consumers with ti < t get an addi- tional subsidy, and consumers with ti > t may also get a subsidy. Earlier estimates of the subsidy Some earlier estimates of the first component of the subsidy are reviewed here. As expected, the main empirical problem in these estimations is to locate the group of households, for which ti has been computed, in the distribution of income of the country or the particular city in question. The most comprehensive estimates are those undertaken by Martha Gutierrez de Gomez in a study done at Colomibia's Na- tional Tariff Bureau for Public Services.12 Gutierrez de Gomez computed the tariffs paid by consumers in 1974, classified by categories of assessed property values, and then used the 1970 household survey (DANE) to derive property values for different income groups from the actual or imputed rent paid by households. This allowed her to map the distribution of households according to property values on the distribution of income for 1970. In Table 5.17, the estimated subsidies, as a percentage of house- hold income, are shown for several cities. They differ markedly between cities, with the highest subsidy being received by the poorest 50 percent of households consuming electricity in Medellin. On the average, consumers belonging to the richest quintile tend to finance poorer consumers, particularly those in the middle quintiles. These distributions of consumers refer to households that were actually billed by utility companies. Households purchasing water or electricity from other dwellings (hooking up among dwellings) as well as households consuming free of charge (consumption from standpipes or illegal connection to transmission lines) are not included. These households do receive subsidies and probably belong to the poorest group among the population. 12. Martha Gutierrez de Gomez, Politica Tarifaria y Distribuci6n de Ingresos, Junta Nacional de Tarifas de Servicios Pdblicos, 1975 (processed). TRANSFERS BETWEEN URBAN CONSUMERS 141 Table 5.17. Subsidies Out of the Tariff Policy as a Percentage of Household's Income, 1970 (percentage) Deciles of the population Bogota Cali Medellin in each city (poorest Elec- Elec- Elec- to richest) Water tricity Water tricity Water tricity 1 - 0.2 - 0.2 - 1.3 2 0.4 0.8 0.2 0.5 1.6 3 - 2 0.3 0.5 002 0.9 2.1 4 0.9 0.3 0.5 0.2 0.7 2.4 5 0.7 0.3 0.3 0.2 0.4 2.0 6 0.5 0.2 0.2 0.2 0.3 1.6 7 0.4 0.3, 0-2 -0.2 0.1 - 0.2 1.3 8 0.2 - -0.2 - 0.2 1.0 9 02 _ -0.61 - -°1 -0 I 01 -04 - 03 10 -0.8 -0.1 -0.7 0 -0.6 -0.5 -0.2 a. Estimates from Lars Lundquist, "Water and Sewerage Tariffs as a Mean for Income Redis- tribution in Colombia," memorandum (Washington, D.C.: World Bank, October 23, 1973). Source: Martha Gutierrez de Gomez, Politica Tarifaria y Distribuci6n de Ingresos. The difference between officially billed consumers and actual consumers might explain the differences between Table 5.17 and Tables SA-27, SA-29, and SA-31 in the statistical appendix, which report the percentage of consumers in each quintile according to the 1974 sample survey. For water, Table 5.17 does not report officially billed consumers belonging to the poorest decile. For Bogota, none are reported in the poorest quintile. Table SA-29 shows that 87 percent of households in the poorest quintile of large cities consumed piped water. Lars Lundquist, formerly with the World Bank, undertook some estimates of the subsidy from water consumption for 1973.13 His estimates for Bogota for the second and third quintile are 2.1 and 1.2 percent, respectively-substantially higher than the esti- mates of Gutierrez de Gomez.14 The difference is not so much a 13. Lars Lundquist, "Water and Sewerage Tariffs as a Mean for Income Redistribu- tion in Colombia," memorandum (Washington, D.C.: World Bank, October 23, 1973). 14. Part of the diffeience results from using different concepts of t. Gutierrez de Gomez defines t as the average residential tariff, whereas Lundquist defines it by including commercial and industrial consumers, yielding a higher value for t. 142 THE DISTRIBUTION OF CONSUMPTION OF PUBLIC UTILITY SERVICES result of differences in the estimated absolute subsidy (Si) as of differences in the household income figures used, which in turn result from the difficulty of mapping the distribution of con- sumers according to property value categories into a distribution according to income levels. In the case of water, there is evidence that the average tariff falls short of any reasonable estimate for the long-run cost,"l Defining this cost as the sum of the operating cost, depreciation, and an 8 percent return on net fixed assets (evaluated at repro- duction cost), the shortfall ranges from 22 percent for M\Iedellin to 50 percent for Bogotd.i6 What is the additional subsidy (the second component described earlier) that a typical household in the second and third quintile would receive? With a value of ai = 0.015 and values of ti/t between one-half and two-tlhirds (implicit in the Gutierrez de Gomez estimates for these quintiles), the percentage subsidy can be computed as shown below. Fraction MC Subsidy = (t/t0) Mc- I ] shortfall (MC - t)/MC t Share spent, ae (ti/t) 1/2 (1,/1) = 2/3 0.22 1.28 0.015 0.008 0.006 0.50 2.00 0.015 0.030 0.022 For shortfalls of 0.22, the additional subsidy ranges between six-tenithls and eight-tentlbs of 1 percent of income, a figure similar to that derived for the first concept of subsidy. For shortfalls like that of Bogota (0.50), the additional subsidy becomes substan- tially larger: it can range between 2 and 3 percent of income- several times higher than the first concept of subsidy resulting from transfers across consumers. 15. See Johannes F. Linn, "The Distributive Effect of Local Government Finances in Colombia: A Review of Evidence," World Bank Staff Working Paper no. 235 (Washington, D.C.: World Bank, March 1976). 16. Ibid. Chapter 6 The Distribution of Beneficiaries of Other Services THE SAMPLE SURVEY PROVIDED statistically significant information on the consumption of other services: agricultural loans to farmers provided by the Caja Agraria, adult retraining courses offered by the Servicio Nacional de Aprendizaje (SENA), educational fellow- ships offered by public agencies, and garbage collection services provided in urban areas. This chapter presents the distribution of the beneficiaries of these services classified by income groups. Agricultural Loans by the Caja Agraria, 1974 The Caja Agraria is the major public agency channeling credit to the agricultural sector at interest rates lower than those charged by private banks. The Caja tries especially to reach small farmers, lending conditions being different for farmers with different asset values. Implicit subsidy provided bv the loans Approximately 75 percent of the value of loans of the Caja are for short-term loans, one year or less; the rest are for one to six years. In 1974, the Caja's interest rates ranged from 14 to 18 percent, depending on the value of the assets of the farmer. In- terest rates charged by the private banking system fluctuated around 25 percent. New loans extended by the Caja Agraria during 1974 amounted 143 144 TIHE DISTRIBUTION OF BENEFICIARIES OF OTHER SERVICES to 4,449 million pesos.' By assuming an average lending term of one year at 16 percent and an alternative market rate of 25 per- cent, the grant component becomes 1 - (1 + 0.16)/(1 + 0.25) or approximately 7.2 percent. If this component is multiplied by the volume of lending, the result is a subsidy of 320 million pesos.2 The figure of 7.2 percent is an average; clearly the grant com- ponent is quite sensitive to the term of a particular loan. For example, using the earlier interest rate figures, loans of six months and three years would have grant components of 4 and 20 percent, respectively. If small farmers receive shorter-term loans, the grant component tends to be smaller for them.3 The foregoing calculations assume that without loans from the Caja, all farmers could obtain credit at the market rate of 25 per- cent charged by commercial banks. In fact, poor farmers find it difficult or impossible to obtain credit from commercial banks. Given the excess demand that results from charging a low real interest (the nominal rate of 25 percent is legally fixed), credit must somehow be rationed.4 Thus, poor farmers usually resort to intermediaries charging much higher rates than commercial banks. In rural Colombia, the rates charged by lenders and other informal intermediaries have ranged between 30 and 50 percent a year. Small farmers also obtain credit by selling their crops at prices substantially below the going rate. 1. World Bank, "Economic Position and Prospects of Colombia," vol. II, report no. 1548 (a restricted-circulation document) (Washington, D.C., 1977; processed). 2. In computing the grant component of a one-year loan, the following definitions can be used: L = value of the loan; r, = concessionary or subsidized interest rate; r2 = market or alternative interest rate; RI = repayment at the end of the year at the concessionary rate, r1, RI = (1 + rI)L; V = present value of RI, discounted at the alternative or market rate r2, V = R,/(I + r2). The grant component, g, can be defined as: L -V V R,/(l + r2) -[I + rl b LL R1l(1 + r,) I + r2= If both ri and r2, the concessionary rate and the market rate, are different for farmers of different income or asset levels, the value of g will differ for different farmers. If for poorer farmers r, is smaller and r2 is larger, the grant component g will tend to be higher. 3. This will be true if poorer and smaller farmers tend to get credit for crop pro- duction, whereas larger farmers get a higher share of credit for cattle raising. 4. With inflation at 20 percent, the commercial interest rate of 25 percent is equiva- lent to a 5 percent real rate of interest. AGRICULTURAL LOANS BY THE CAJA AGRARIA, 1974 145 Table 6.1. Estimates of the Implicit Subsidy Received by Farmers from Loans by the Caja Agraria, 1974 Distribution of farmers and aggregate loan Estimates of the subsidy Income Distribu- Loan Estimated quintile tion of Distribution of to each Alternative Grant (g) subsidy to each (poorest farmers aggregate loan farmer (L) interest rate (per- farmer (g.L) to richest) (percentage) (percentage) (pesos,) (percentage) centage) (pesos) 1 34.6 19.6 4,374 40-60 17-28 743-1,224 2 29.3 43.0 11,385 30-50 11-23 1,241-2,595 3 19.5 20.2 4 12.1 12.5 7,965b 25b 7b 557b 5 4.5 4.7 a. The standard deviation, av, and the sample size. n, are: x = 4,374, ~= 2,893, ai = 43; x = 11,385, o = 13,452, n = 34; x = 7,965, crO = 9,810, it = 53. b. Because of the small sample size, the value for farmers in quintiles 3, 4, and 5 was computed as an average of the three quintiles. Distribution of the subsidy The 1974 sample survey provided data on 130 farmers who received new loans from the Caja during 1974. The distribution of these farmers in the countrv distribution of income, the mean loan received, and an estimate of the implicit subsidy appear in Table 6.1. One-third of the farmers who received loans belong to the poorest quintile in the country distribution of income. Almost two-tlhirds belong to the poorest 40 percent of households. The second column of Table 6.1 shows the distribution of new credit to farmers bv income groups in 1974. Since the mean loan in the richest tlhree quintiles is similar to the average loan, the distribution of the total credit going to these quintiles is almost the same as the distribution of farmers. This is not true for the poorest quintiles; farmers in the first quintile, although more numerous, receive half the share of the total credit received by farmers in the second quintile. The third column of the table shows figures on the mean loan to each farmer. Because of the small sample size, the value for farmers in quintiles 3, 4, and 5 was computed as an average for the three quintiles. Farmers in quintile 2 received an average 146 THE DISTRIBUTION OF BENEFICIARIES OF OTHER SERVICES Table 6.2. Income Distribution of Farmers, 1974 (percentage) Income quintile Farmers (poorest to receiving loans richest) All farmers from the Caja 1 28.3 34.6 2 26.5 29.3 3 22.0 19.5 4 17.3 12.1 5 5.9 4.5 Note: In the sample survey, farmer is defined as any rural household living on a plot under cultivation and actively involved as landlord, tenant, or sharecropper. loan of 11,285 pesos ($409), substantiallv larger than the average loan of 7,695 pesos ($279) reported by all farmers. The last three columns present a rough estimate of the implicit subsidy to each farmer from each loan. The first column shows a hypothetical range of interest rates fronm sources other than the Caja; the second, the range of the grant component for a one- year loan; and the third, the range of the implicit subsidy to each farmer. The estimates of the subsidy are for farmers who effec- tively received a loan and does not represent an average for all farmers in that quintile. In this respect these figures cannot be compared directly to the mean subsidies to each quintile esti- mnated for education and health. Finally, the income distribution of farmers whlo received loans from the Caja is compared with the income distribution of all farmers in Table 6.2. Although 28.3 percent of all farmers belong to the poorest (country) quintile, that quintile accounts for 34.6 percent of those farmers who received loans from the Caja. SENA Training Courses, 1974 Servicio Nacional de Aprendizaje (SENA) is a decentralized public agency in charge of training programs for workers already employed in the private sector. It is financed by a tax on the payroll of enterprises belonging to the urban formal, or modern, SENA TRAINING COURSES, 1974 147 Table 6.3. Attendance at SENA Courses, 1974 (percentage) Income quintile (poorest to Number of Man-months richest) students of attendance 1 12.7 11.3 2 14.7 11.9 3 24.5 24.3 4 18.9 20.0 5 29.2 32.5 sector: that is, firms wlhose workers are affiliated with the Social Security System. These enterprises select employees to be trained at SENA on a part-time or full-time basis. During 1974, SENA'S expenditures were 628 million pesos, most of wlhichl was contributed by the private sector.' In the 1974 sample survey, 377 individuals reported having participated in SENA courses during that year; information on the number of months attended per individual was also reported. Table 6.3 shows the distribution of individuals as well as the dis- tribution of total man-months of attendance, classified according to quintiles in the country distribution of income. The table shows that about 30 percent of those attending SENA courses belong to the riclhest quintile in the country income dis- tribution, and most belong to the three richest quintiles. Partici- pants from the poorest 40 percent of families account for only 23 percent of the total man-months of attendance. The smaller repre- sentation of the low-income groups can be explained by two fac- tors. First, SENA mainly trains individuals in urban areas (93 percent of the total attendance in 1974 was by urban workers), and urban households are better off than rural households in terms of the country distribution of income. Second, within the urban area, SENA favors workers employed in the modern sector, that is, the best paid labor in the urban sector. 5. Contraloria General de la Reptiblica de Colombia, Informe Financiero de 1974. 148 THE DISTRIBUTION OF BENEFICIARIES OF OTHER SERVICES Educational Fellowships from Public Agencies, 1974 In the sample survey, seventy-four houselholds reported having received educational fellowslips from public agencies during 1974. The (expanded) distribution of these houselholds according to income quintiles is: Income quintile Percentage of (poorest to richest) households 1 16.2 2 23.8 3 9.7 4 27.0 5 23.3 The distribution does not show a clear pattern across income quintiles, whlich is expected in view of the relatively snmall sample size. A more aggregate interpretation of the data suggests that approximately half of the households that received fellowships belong to the lower 40 percent and the other half to the richest 60 percent. Garbage Collection Services in Urban Areas, 1974 The 1974 sample survey collected data on the number of urban houselholds with public garbage collection services in 1974.6 Table 6.4 presents the distribution of these houselholds according to quintiles in the country income distribution. More than one-third of those with the service belong to the richest quintile, whereas 9.7 percent belong to the poorest. To show the extent to which this is a function of the fraction of urban house- holds in the higher-income quintiles, the distribution of all urban households is given in parentheses according to country quintiles. 6. Data on this service were reported earlier to estimate the probability of having the service. Here only data on households that receive the service and their distribution by income group are given. GARBAGE COLLECTION SERVICES IN URBAN AREAS, 1974 149 Table 6.4. Urban Households with Garbage Collection Service, 1974 (percentage) Income quintile (poorest to Large Intermediate Small Urban richest) cities cities towns average 1 7.2( 9.0) 8.7(12.3) 19.3(29.3) 9.7(15.1) 2 10.0(13.1) 13.6(16.0) 20.8(24.9) 12.9(16.8) 3 16.1(17.4) 17.6(18.1) 19.9(18.5) 17.3(17.8) 4 23.9(23.4) 25.2(25.6) 22.8(16.7) 24.1(22.4) 5 42.8(37.1) 34.9(28.0) 17.1(10.6) 36.0(27.9) Note: Figures in parentheses show the distribution of all urban households. Although 27.9 percent of all urban households belong to the richest quintile, their share in the distribution of the service is 36 percent; 15.1 percent of urban households belong to the poorest quintile, and their share in the service is only 9.7 percent. The distribution favors higher-income groups not only because garbage disposal is primarily an urban service, but also because there is a bias in favor of higher-income households within urban areas. Statistical Appendix 152 STATISTICAL APPENDIX Table SA-1. Distribution of Families, by per Capita Income, According to the Official Exchange Rate, 1974 (percentage) Inter- Annual per capita Small mediate Large Urban Rural Country income (dollars) towns cities cities average areas average 0-50 12.6 3.7 3.7 5.9 10.5 7.6 51-75 14.0 6.8 4.1 7.4 13.6 9.7 76-100 11.5 6.8 5.8 7.5 12.9 9.5 101-150 19.5 15.5 12.3 15.1 23.0 18.0 151-250 20.3 21.5 21.1 21.1 23.0 21.8 251-350 9.6 15.5 12.6 12.6 9.4 11.4 351-500 5.2 9.2 12.7 9.8 4.5 7.8 501-700 2.4 7.3 9.2 6.9 1.5 4.9 701-1,500 3.9 10.2 12.5 9.7 1.2 6.5 Over 1,500 1.0 3.5 6.0 4.0 0.4 2.7 Mean income 176 321 420 326 149 256 Table SA-2. Distribution of Families, by per Capita Income, According to the Kravis Parity Exchange Rate, 1974 (percentage) Inter- Annual per capita Small mediate Large Urban Rural Country income (dollars) towns cities cities average areas average 0-50 3.2 0.2 1.1 1.4 1.9 1.6 51-75 2.8 1.3 1.7 1.9 3.6 2.6 76-100 5.1 1.4 0.7 2.0 4.1 2.8 101-150 11.9 5.9 3.0 6.0 11.4 8.0 151-250 25.1 14.9 10.9 15.6 25.0 19.0 251-350 13.7 12.8 12.2 12.7 18.7 15.0 351-500 14.6 15.9 15.1 15.2 16.9 15.8 501-700 9.7 15.3 13.3 13.0 9.9 11.8 701-1,500 9.0 18.6 23.6 18.5 7.0 14.2 Over 1,500 4.9 13.7 18.4 13.7 1.5 9.2 Mean income 373 681 890 692 317 544 STATISTICAL APPENDIX 153 Table SA-3. Mean Number of Rooms Occupied by the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)' towns cities cities average areas average 1 2.63 2.83 2.94 2.78 2.15 2.51 (1.1) (1.3) (1.7) (1.4) (1.1) (1.3) 2 2.77 3.03 2.95 2.91 2.34 2.58 (1.3) (1.6) (1.7) (1.5) (1.0) (1.3) 3 2.79 3.50 3.00 3.12 2.48 2.74 (1.2) (1.8) (1.7) (1.7) (1.4) (1.5) 4 3.10 4.05 4.29 3.86 2.30 3.13 (1.6) (2.0) (2.0) (1.9) (1.1) (1.7) 5 3.77 4.27 5.24 4.71 2.80 4.40 (1.8) (2.1) (2.5) (2.3) (1.6) (2.3) Country average 3.01 3.54 3.69 3.48 2.41 3.07 (1.5) (1.9) (2.2) (1.9) (1.3) (1.8) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-4. Mean Number of Toilets and Latrines in the Dwelling Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest) a towns cities cities average areas average 1 0.74 0.99 1.05 0.89 0.31 0.59 (0.5) (0.5) (0.6) (0.6) (0.5) (0.6) 2 0.92 1.08 1.27 1.11 0.38 0.70 (0.7) (0.5) (0.6) (0.6) (0.7) (0.7) 3 0.89 1.12 1.32 1.13 0.40 0.86 (0.5) (0.7) (0.7) (0.6) (0.6) (0.7) 4 1.05 1.50 1.68 1.52 0.43 1.01 (0.6) (0.9) (0.9) (0.8) (0.6) (0.8) 5 1.29 2.07 2.44 2.11 0.48 1.81 (0.7) (1.1) (1.2) (1-2) (0.7) (1.2) Country average 0.98 1.35 1.56 1.35 0.40 0.99 (0.6) (0.9) (1.0) (0.9) (0.6) (0.9) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. 154 STATISTICAL APPENDIX Table SA-5. Mean Number of Persons in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)' towns cities cities average areas average 1 7.23 6.81 6.47 6.61 6.96 6.87 (2.5) (2.5) (2.6) (2.6) (2.4) (2.5) 2 6.45 5.90 5.25 5.97 6.41 5.99 (2.6) (2.6) (2.3) (2.5) (2.3) (2.5) 3 5.68 5.21 5.08 5.12 5.88 5.38 (2.4) (2.5) (2.4) (2.6) (2.5) (2.5) 4 5.09 4.98 4.61 4.91 4.77 4.80 (2.7) (2.4) (2.3) (2.3) (2.6) (2.4) 5 4.40 4.17 4.30 4.19 4.08 4.25 (2.2) (2.3) (2.2) (2.2) (2.3) (2.3) Country average 5.75 5.40 5.13 5.36 5.63 5.47 (2.7) (2.6) (2.4) (2.6) (2.6) (2.6) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-6. Mean Number of Income Earners in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)' towns cities cities average areas average 1 1.41 1.35 1.26 1.32 1.19 1.24 (0.9) (0.7) (0.6) (0.8) (0.8) (0.8) 2 1.46 1.58 1.46 1.45 1.19 1.37 (1.0) (1.0) (0.8) (0.8) (0.7) (0.9) 3 1.33 1.65 1.61 1.62 1.42 1.52 (0.7) (1.0) (1.0) (1.0) (1.0) (1.0) 4 1.67 1.55 1.67 1.60 1.42 1.59 (1.1) (1.0) (1.1) (1.0) (1.1) (1.0) 5 1.62 1.66 1.95 1.82 1.58 1.72 (0.9) (0.8) (1.1) (1.0) (1.7) (1.0) Country average 1.50 1.56 1.60 1.56 1.36 1.48 (0.9) (0.9) (1.0) (0.9) (1.0) (0.9) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. STATISTICAL APPENDIX 155 Table SA-7. Mean Number of Children Age 0 to 5 in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)4 towns cities cities average areas average 1 1.31 1.10 1.11 1.17 1.55 1.39 (1.2) (1.2) (1.2) (1.2) (1.3) (1.2) 2 1.09 0.90 0.77 0.92 1.36 1.05 (1.1) (1.0) (0.9) (1.0) (1.2) (1.1) 3 0.89 0.60 0.61 0.61 1.02 0.76 (1.1) (0.8) (0.8) (0.9) (1.1) (1.0) 4 0.56 0.60 0.38 0.53 0.65 0.54 (0.8) (0.8) (0.7) (0.8) (0.9) (0.8) 5 0.41 0.48 0.28 0.33 0.39 0.36 (0.7) (0.7) (0.6) (0.6) (0.8) (0.7) Country average 0.85 0.73 0.63 0.71 1.00 0.82 (1.1) (1.0) (0.9) (1.0) (1.2) (1.0) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-8. Mean Number of Children Age 6 to 11 in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)- towns cities cities average areas average 1 1.66 1.41 1.38 1.42 1.66 1.57 (1.3) (1.2) (1.3) (1.3) (1.3) (1.3) 2 1.36 0.99 0.97 1.13 1.28 1.14 (1.3) (1.2) (1.2) (1.2) (1.2) (1.2) 3 1.10 0.82 0.76 0.80 1.08 0.92 (1.2) (1.1) (1.0) (1.1) (1.1) (1.2) 4 0.74 0.74 0.58 0.70 0.73 0.64 (1.1) (1.0) (0.9) (1.0) (1.1) (0.9) 5 0.44 0.42 0.37 0.38 0.40 0.44 (0.8) (0.8) (0.7) (0.7) (0.8) (0.8) Country average 1.05 0.87 0.81 0.89 1.04 0.94 (1.2) (1.1) (1.1) (1.1) (1.2) (1.2) Note: Values in parentheses show the standard deviation. a. Quinltiles in the distribution of income in each region or location. 156 STATISTICAL APPENDIX Table SA-9. Mean Number of Children Age 0 to 11 in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest) towns cities cities average areas average 1 2.98 2.50 2.48 2.59 3.21 2.96 (1.9) (1.9) (1.7) (1.9) (2.0) (1.9) 2 2.47 1.90 1.74 2.05 2.64 2.19 (1.9) (1.6) (1.4) (1.5) (1.7) (1.6) 3 1.98 1.42 1.37 1.41 2.11 1.68 (1.6) (1.4) (1.2) (1.4) (1.7) (1.5) 4 1.30 1.34 0.96 1.23 1.38 1.18 (1.4) (1.3) (1.3) (1.3) (1.5) (1.3) 5 0.86 0.90 0.66 0.71 0.79 0.80 (1.1) (1.2) (1.0) (1.0) (1.3) (1.1) Country average 1.90 1.61 1.43 1.60 2.04 1.77 (1.8) (1.6) (1.5) (1.6) (1.9) (1.7) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-10. Mean Number of Persons Age 12 to 16 in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)' towns cities cities average areas average 1 1.05 1.19 1.00 1.04 1.06 1.05 (1.2) (1.2) (1.2) (1.2) (1.1) (1.1) 2 1.08 0.86 0.66 0.90 0.99 0.93 (1.2) (1.0) (1.0) (1.1) (1.1) (1.1) 3 0.88 0.75 0.65 0.71 0.81 0.71 (1.1) (1.0) (0.9) (1.0) (1.0) (1.0) 4 0.68 0.64 0.60 0.61 0.58 0.60 (1.1) (1.0) (0.9) (0.9) (0.9) (0.9) 5 0.55 0.50 0.41 0.46 0.48 0.49 (0.9) (0.8) (0.7) (0.8) (0.8) (0.8) Country average 0.85 0.78 0.66 0.74 0.79 0.76 (1.1) (1.0) (1.0) (1.0) (1.0) (1.0) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. STATISTICAL APPENDIX 157 Table SA-11. Mean Number of Pregnancies in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)4 towns cities cities average areas average 1 0.23 0.26 0.25 0.24 0.29 0.26 (0.5) (0.5) (0.4) (0.5) (0.5) (0.5) 2 0.23 0.23 0.18 0.21 0.21 0.20 (0.4) (0.4) (0.4) (0.4) (0.4) (0.4) 3 0.20 0.18 0.15 0.18 0.21 0.21 (0.4) (0.4) (0.4) (0.4) (0.4) (0.4) 4 0.17 0.15 0.15 0.15 0.17 0.14 (0.4) (0.4) (0.4) (0.4) (0.4) (0.4) 5 0.12 0.21 0.11 0.14 0.11 0.14 (0.3) (0.5) (0.3) (0.4) (0.3) (0.4) Country average 0.19 0.21 0.17 0.18 0.20 0.19 (0.4) (0.4) (0.4) (0.4) (0.4) (0.4) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-12. Mean Age of Head of Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest) towns cities cities average areas average 1 50.0 44.6 41.6 44.8 45.0 44.8 (13.1) (12.8) (12.7) (13.9) (13.2) (13.5) 2 46.3 44.6 42.1 43.5 43.2 44.3 (14.6) (12.8) (14.1) (12.9) (12.9) (13.2) 3 44.4 43.6 42.8 44.2 45.4 44.4 (13.8) (13.9) (11.8) (13.9) (13.6) (13.9) 4 45.9 42.2 43.5 44.0 46.0 45.3 (14.7) (13.7) (13.2) (13.6) (14.5) (14.2) 5 48.3 44.0 46.9 45.4 47.9 45.2 (15.4) (14.7) (15.0) (14.7) (15.0) (14.5) Country average 46.8 43.8 43.4 44.4 45.4 44.8 (13.6) (13.5) (13.8) (13.8) (13.9) (13.9) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. 158 STATISTICAL APPENDIX Table SA-13. Mean Total Years of Schooling of the Head of Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)' towns cities cities average areas average 1 2.03 3.42 3.96 3.03 1.70 2.29 (1.9) (2.8) (2.5) (2.5) (1.7) (2.2) 2 2.46 3.85 4.57 3.74 1.78 2.60 (2.4) (2.6) (2.8) (2.7) (1.8) (2.4) 3 2.90 4.68 5.57 4.48 2.00 3.22 (2.4) (3.2) (3.3) (2.9) (2.0) (2.7) 4 3.67 7.15 6.97 6.41 1.81 4.36 (2.7) (3.8) (4.1) (3.9) (1.9) (3.5) 5 5.78 9.77 9.50 9.04 2.52 7.84 (4.1) (4.4) (4.7) (4.6) (2.9) (4.8) Country average 3.38 5.77 6.13 5.34 1.96 4.05 (3.1) (4.2) (4.1) (4.1) (2.1) (3.8) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-14. Mean Age of Wife in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest), towns cities cities average areas average 1 36.8 37.6 35.2 36.8 37.4 37.2 (11.8) (11.2) (11.0) (11.8) (11.3) (11.6) 2 37.5 36.4 34.9 37.5 36.3 37.6 (12.0) (11.9) (10.9) (12.0) (10.9) (11.7) 3 36.9 37.1 35.9 36.9 39.3 37.4 (12.4) (13.6) (10.3) (12.4) (12.2) (12.7) 4 37.9 36.5 39.4 37.9 39.1 38.9 (12.0) (11.3) (12.5) (12.0) (13.8) (13.2) 5 38.3 35.9 39.9 38.3 43.1 38.9 (13.0) (12.5) (13.2) (13.0) (14.6) (13.0) Country average 37.5 36.7 37.0 37.5 38.8 38.0 (12.3) (12.1) (11.8) (12.2) (12.7) (12.4) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. STATISTICAL APPENDIX 159 Table SA-15. Mean Total Years of Schooling of Wife in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)' towns cities cities average areas average 1 1.31 2.11 2.73 2.05 1.34 1.67 (1.6) (2.5) (2.4) (2.2) (1.6) (1.9) 2 2.07 2.22 3.48 2.59 1.51 1.93 (2.1) (2.7) (3.1) (2.6) (1.7) (2.1) 3 2.00 3.48 3.77 3.15 1.67 2.47 (2.3) (3.7) (3.2) (3.4) (1.9) (2.8) 4 2.80 5.41 4.63 4.77 1.56 3.22 (3.1) (4.5) (4.4) (4.1) (2.0) (3.4) 5 3.96 6.11 7.00 6.10 1.89 5.34 (4.1) (5.0) (5.2) (5.1) (2.8) (5.0) Country average 2.24 3.86 4.34 3.73 1.59 2.91 (2.9) (4.1) (4.1) (3.9) (2.0) (3.5) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-16. Mean Number of Children Age 6 to 11 Registered in School in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest)' towns cities cities average areas average 1 0.92 1.06 0.96 0.88 0.85 0.85 (1.1) (1.1) (1.1) (1.1) (1.0) (1.0) 2 0.78 0.68 0.74 0.83 0.57 0.70 (1.0) (1.0) (1.0) (1.0) (0.8) (0.9) 3 0.71 0.67 0.64 0.61 0.61 0.61 (0.9) (1.0) (0.9) (1.0) (0.9) (0.9) 4 0.48 0.65 0.53 0.61 0.41 0.50 (0.8) (1.0) (0.8) (0.9) (0.8) (0.8) 5 0.37 0.39 0.35 0.34 0.24 0.38 (0.7) (0.8) (0.7) (0.7) (0.6) (0.7) Country average 0.65 0.68 0.64 0.65 0.54 0.61 (0.9) (1.0) (0.9) (0.9) (0.9) (0.9) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. 160 STATISTICAL APPENDIX Table SA-17. Mlrean Number of Persons Age 12 to 16 Registered in School in the Household Regional Inter- income quintile Small mediate Large Urban Rural Country (poorest to richest), towns cities cities average areas average 1 0.61 0.98 0.83 0.77 0.48 0.62 (0.9) (1.1) (1.1) (1.0) (0.8) (0.9) 2 0.78 0.71 0.55 0.72 0.55 0.61 (1.0) (1.0) (0.9) (1.0) (0.9) (1.0) 3 0.67 0.63 0.56 0.58 0.41 0.50 (1.0) (1.0) (0.9) (0.9) (0.8) (0.9) 4 0.53 0.55 0.55 0.54 0.30 0.47 (1.0) (0.9) (0.8) (0.9) (0.6) (0.8) 5 0.45 0.41 0.34 0.39 0.24 0.41 (0.8) (0.8) (0.7) (0.7) (0.6) (0.8) Country average 0.61 0.66 0.56 0.60 0.40 0.52 (1.0) (1.0) (0.9) (0.9) (0.8) (0.9) Note: Values in parentheses show the standard deviation. a. Quintiles in the distribution of income in each region or location. Table SA-18. Number of Teachers in Public Primary Schools, by Category and Stratum, 1973 Categories Without 4 (lowest I (highest Stratum category pay) 3 2 pay) Total Bogota 27 10 84 3,651 4,513 8,285 Cali 33 47 97 949 1,037 2,163 Medellin 10 7 32 300 2,586 2,935 Barranquilla 14 16 26 357 1,256 1,669 5 12 14 21 166 789 1,002 6 39 130 130 344 707 1,350 7 10 94 118 516 1,156 1,894 8 21 11 44 270 854 1,200 9 7 6 48 438 1,412 1,911 10 23 19 90 287 409 828 11 318 155 414 628 1,886 3,401 12 37 19 52 171 642 921 13 6 1 8 139 750 904 14 50 7 25 143 128 353 STATISTICAL APPENDIX 161 Table SA-18. (Continued) Categories Without 4 (lowest 1 (highest Stratum category pay) 3 2 pay) Total 15 50 88 72 324 272 806 16 24 117 49 278 308 776 17 63 59 53 413 583 1,171 18 34 73 60 344 596 1,107 19 30 13 99 452 1,446 2,040 20 16 20 91 384 685 1,196 21 1 5 42 232 662 942 22 156 176 238 545 1,292 2,407 23 123 86 180 317 741 1,447 24 56 67 116 327 602 1,168 25 49 46 108 240 784 1,227 26 39 29 61 432 608 1,169 27 92 47 69 436 717 1,361 28 355 245 166 544 272 1,582 29 1,130 302 190 379 189 2,190 30 543 406 215 578 277 2,019 31 1,121 231 361 513 301 2,527 32 145 65 139 986 1,143 2,478 33 43 80 157 815 1,139 2,234 34 943 151 363 646 537 2,640 35 10 38 63 594 740 1,538 36 a a 37 326 123 129 184 94 856 38 234 81 272 462 378 1,427 39 87 26 120 386 607 1,226 40 100 105 256 716 529 1,706 41 106 22 86 189 164 567 42 763 362 276 668 294 2,363 43 1,756 836 432 932 490 4,446 44 28 14 5 33 9 89 45 194 125 68 341 195 923 46 430 76 134 454 415 1,509 47 134 89 154 531 747 1,655 48 508 36 50 121 125 840 a. No schools outside the cabecera. Source: COLDATOS report, pp. 16-17. 162 STATISTICAL APPENDIX Table SA-19. Monthly Wage of Teachers in Public Primary Schools, by Category and Departments, 1973 (pesos) Categories Without 4 (lowest I (highest Department category pay) 3 2 pay) Antioquia 1,270 1,540 1,659 1,776 2,052 Atlantico 0 1,526 1,678 1,896 2,256 Bogota, D. E. 0 2,071 2,125 2,180 2,234 Bolivar 1,206 1,381 1,621 1,903 2,180 Boyaca 1,071 1,542 1,714 1,795 2,048 Caldas 1,287 1,551 1,754 1,859 2,057 Cauca 1,127 1,354 1,487 1,621 1,987 Cesar 1,298 1,628 1,749 1,892 2,090 C6rdoba 1,281 1,443 1,576 1,914 2,202 Cundinamarca 1,166 1,628 1,793 1,940 2,123 Choc6 1,265 1,321 1,561 1,740 2,040 Huila 1,038 1,343 1,487 1,665 1,998 La Guajira 1,232 1,831 1,967 2,115 2,245 Magdalena 1,038 1,622 1,804 1,908 2,046 Meta 1,265 1,845 1,610 1,848 2,145 Narino 1,021 1,376 1,510 1,654 1,987 N. Santander 1,239 1,543 1,709 1,889 2,110 Quindio 1,371 1,644 1,694 1,809 2,070 Risaralda 1,298 1,476 1,587 1,665 1,887 Santander 1,077 1,387 1,554 1,776 1,998 Sucre 1,143 1,496 1,703 1,973 2,344 Tolima 1,341 1,496 1,587 1,776 1,998 Valle del Cauca 1,624 1,820 1,853 1,929 2,134 Average 1,229 1,426 1,686 1,849 2,097 Source: COLDATOS report, p. 21. STATISTICAL APPENDIX 163 Table SA-20. Estimate of the Costs of Public hlealth Centers, 1974 Location of institution Inter- Large mediate Small Rural Country Number and costs of centers cities cities towns areas total Number of centers With beds - - 20 16 36 Without beds 148 139 262 133 682 Cost for each center, 1969, (thousands of pesos) With beds - - 131 100 Without beds 354 193 93 71 Estimated cost for each center, 1974b (thousands of pesos) With beds - - 266 203 Without beds 719 392 189 144 Cost for all centers, 1974 (millions of pesos) With beds - - 5.3 3.2 8.5 Without beds 106.4 54.5 49.5 19.1 229.5 Total 106.4 54.5 54.8 22.3 238.3 a. COLDATOS report. b. Adjusted by the change in the price level between 1969 and 1974. 164 STATISTICAL APPENDIX Table SA-21. Estimate of Subsidy to Each Puesto de Salud, 1974 (pesos) Number Institution of staff Subsidy Health center, 1969 Doctors 1.88 61,275 Auxiliares 3. 70 49,788 Total 5.58 111,063 Puesto de salud, 1969 Doctors 0.53 17,274 Auxiliares 1.06 14,263 Total 1.59 31,537 Total, 1974 pesos 64,020 Source: COLDATOS report. Table SA-22. Total Subsidy to Puestos de Salud, 1974 Location of institution Number and costs Large Intermediate Small Rural Country of puestos cities cities towns areas total Cost of each puesto 64,020 64,020 64,020 64,020 64,020 (pesos) Number of puestos 18 114 539 899 1,570 Subsidy 1.2 7.3 34.5 57.5 100.5 (millions of pesos) STATISTICAL APPENDIX 165 Table SA-23. Number of Services Received in Hospitals of the National Health System as Reported by Households, 1974 (thousands) Location of household Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns total areas total Outpatient visits 1 146 129 312 587 319 906 2 130 183 214 527 333 860 3 164 178 250 592 267 859 4 281 322 116 719 226 945 5 170 228 47 445 59 504 Total 891 1,040 939 2,870 1,204 4,074 Deliveries 1 9 8 11 28 24 52 2 8 9 7 24 24 48 3 6 7 6 19 16 35 4 8 8 2 18 8 26 5 3 4 4 11 4 15 Total 34 36 30 100 76 176 Operations 1 12 3 4 19 10 29 2 3 4 2 9 8 17 3 4 8 0 12 3 15 4 12 6 3 21 4 25 5 18 3 2 23 0 23 Total 49 24 11 84 25 109 Inpatient days 1 743 292 194 1,412 620 2,032 2 232 251 155 455 845 1,300 3 253 157 92 502 719 1,221 4 145 280 87 512 396 908 5 238 100 132 470 99 569 Total 1,611 1,080 660 3,351 2,679 6,030 166 STATISTICAL APPENDIX Table SA-24. Number of Services Received in Hospitals of the Social Security System (Icss and Cajas) as Reported by Households, 1974 (thousands) Location of household Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns total areas total Outpatient visits 1 33 83 28 144 45 189 2 183 163 146 492 59 551 3 374 203 87 664 70 734 4 418 345 102 865 55 920 5 502 335 48 885 9 894 Total 1,510 1,129 411 3,050 238 3,288 Deliveries 1 0 2 - 2 3 5 2 3 1 2 6 1 7 3 9 1 3 13 2 15 4 5 6 2 13 3 16 5 5 6 1 12 - 12 Total 22 16 8 46 9 55 Operations 1 2 1 0 3 - 3 2 3 0 0 3 - 3 3 6 1 1 8 - 8 4 1 2 0 3 - 3 5 4 1 0 5 - 5 Total 16 5 1 22 - 22 Inpatient days 1 55 12 38 105 105 2 128 14 29 171 171 3 316 24 20 360 - 360 4 142 170 13 325 - 325 5 137 60 0 197 - 197 Total 778 280 100 1,158 - 1,158 STATISTICAL APPENDIX 167 Table SA-25. Distribution of Public Subsidies to Hospitals of the National Health System (NHS), the Social Security System (sss), and Health Centers of All Types (HC), 1974 (millions of pesos) Location of household Income quintile Large cities Intermediate cities (poorest to richest) NHS sss HC Total NHS SSS HC Total 1 102.0 38.0 15.6 155.6 42.6 16.1 12.2 70.9 2 33.4 92.9 29.9 156.2 41.6 18.5 17.8 77.9 3 38.5 206.6 31.4 276.5 39.3 30.0 11.4 80.7 4 46.2 120.4 20.1 186.7 52.5 87.5 14.3 154.3 5 62.7 154.4 10.6 227.7 23.5 51.7 6.1 81.3 Total 282.8 612.3 107.6 1,002.7 199.5 203.8 61.8 465.1 Small towns Urban total NHS SSS HC Total NHS sss HC Total 1 40.4 13.4 33.7 87.5 185.0 67.5 61.5 314.0 2 30.0 19.2 26.8 76.0 105.0 130.6 74.5 310.1 3 20.4 24.9 14.5 59.8 98.2 261.5 57.3 417.0 4 18.6 11.2 7.1 36.9 117.3 219.1 41.5 377.9 5 21.7 3.4 7.2 32.3 107.9 209.5 23.9 341.3 Total 131.1 72.1 89.3 292.5 613.4 888.2 258.7 1,760.3 Rural total Country total NHS sss HC Total NHS SSS HC Total 1 99.5 6.7 24.3 130.5 284.5 74.2 85.8 444.5 2 126.1 8.5 22.2 156.8 231.1 139.1 96.7 466.9 3 103.0 8.4 20.3 131.7 201.2 269.9 77.6 548.7 4 61.8 6.1 9.4 77.3 179.1 225.2 50.9 455.2 5 14.6 1.3 3.6 19.5 122.5 210.8 27.5 360.8 Total 405.0 31.0 79.8 515.8 1,018.4 919.2 338.5 2,276.1 168 STATISTICAL APPENDIX Table SA-26. Alternate Distribution of Public Subsidies to Hospitals, 1974 (millions of pesos) Location of household Income quintile Large cities Intermediate cities (poorest to richest) NHS SSS HC Total NHS SSS HC Total 1 46.4 28.2 15.6 90.2 24.7 13.4 12.2 50.3 2 41.3 88.2 29.9 159.4 35.1 32.4 17.8 85.3 3 52.0 113.9 31.4 197.3 34.1 44.6 11.4 90.1 4 89.1 158.0 20.1 267.2 61.8 52.8 14.3 128.9 5 54.0 224.2 10.6 288.8 43.8 60.6 6.1 110.5 Total 282.8 612.3 107.6 1,002.7 199.5 203.8 61.8 465.1 Small towns Urban total NHS sss HC Total NHS SSS HC Total 1 43.6 10.2 33.7 87.5 114.7 51.8 61.5 228.0 2 29.9 15.6 26.8 72.3 106.3 136.2 74.5 317.0 3 34.9 18.7 14.5 68.1 121.0 177.2 57.3 355.5 4 16.2 15.9 7.1 39.2 167.1 226.7 41.5 435.3 5 6.5 11.7 7.2 25.4 104.3 296.4 23.9 424.6 Total 131.1 72.1 89.3 292.5 613.4 888.2 258.7 1,760.3 Rural total Country total NHS SSS RC Total NHiS SSS tiC Total 1 107.4 8.5 24.3 140.2 222.1 52.3 85.8 360.2 2 111.8 6.6 22.2 140.6 218.1 142.8 96.7 457.6 3 89.9 8.3 20.3 118.5 210.9 185.5 77.6 474.0 4 76.1 5.4 9.4 90.9 243.2 232.1 50.9 526.2 5 19.8 2.2 3.6 25.6 124.1 298.6 27.5 450.2 Total 405.0 31.0 79.8 515.8 1,018.4 919.2 338.5 2,276.1 Note: The sss subsidy is classified according to the distribution of affiliated individuals; the NHS subsidy in hospitals is classified according to the distribution of outpatient visits. STATISTICAL APPENDIX 169 Table SA-27. Percentage of Families with Electricity, by Regional Quintiles Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns average areas average 1974 1 97.5 90.7 63.2 74.2 8.0 41.4 2 98.3 89.1 70.2 79.3 10.8 49.1 3 99.5 95.4 71.1 88.9 17.7 61.7 4 99.8 98.4 86.4 90.3 15.9 73.5 5 100.0 98.7 94.5 91.3 25.7 91.3 Country average 98.9 94.3 72.4 84.9 15.6 63.2 1970 1 92.9 86.9 51.8 75.9 5.3 37.7 2 94.9 87.7 63.6 84.4 9.6 46.0 3 99.1 94.8 69.7 93.3 12.0 58.2 4 99.8 98.2 81.7 98.3 14.3 71.9 5 97.5 98.7 95.4 98.2 19.7 89.9 Country average 96.7 93.1 73.8 89.9 12.0 60.5 Note: Regional quintDles are defined in the distribution of regional income, not country income. 170 STATISTICAL APPENDIX Table SA-28. Percentage of Families with Electricity, by per Capita Household Income Annual per Inter- capita income Large mediate Small Urban Rural Country (dollars) cities cities towns average areas average 1974 0-75 95.8 89.8 59.7 77.4 7.8 41.9 76-150 98.3 88.2 70.4 85.9 14.0 50.8 151-350 99.1 96.2 88.7 95.9 16.1 67.4 351-700 99.8 97.8 100.0 99.3 20.3 87.0 Over 701 100.0 99.0 91.2 98.9 38.1 95.5 Country average 99.0 94.3 75.9 91.9 13.8 62.9 1970 0-75 90.0 86.8 54.6 72.7 5.6 38.5 76-150 94.6 85.1 68.7 83.1 11.2 47.5 151-350 97.6 95.7 84.8 94.2 15.3 64.7 351-700 99.8 98.5 100.0 99.5 19.0 86.2 Over 701 97.2 99,0 95.6 97.6 38.9 93.9 Country average 96.7 93.2 72.9 89.7 10.9 60.3 STATISTICAL APPENDIX 171 Table SA-29. Percentage of Families with Piped Water, by Regional Quintiles Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns average areas average 1974 1 87.1 81.3 62.6 77.0 15.3 44.0 2 92.3 85.5 78.7 85.8 17.2 49.5 3 97.3 90.3 79.7 90.9 18.1 62.2 4 99.4 96.1 84.2 96.8 25.8 71.8 5 100.0 96.9 92.7 98.6 23.3 89.0 Country average 95.1 90.1 79.8 89.8 19.9 63.2 1970 1 80.6 78.4 57.6 72.5 5.5 36.7 2 89.4 81.9 73.0 81.3 10.7 45.1 3 95.7 89.7 79.7 89.8 12.4 56.4 4 98.8 95.6 80.3 95.8 17.1 69.7 5 97.5 98.5 90.1 97.5 21.2 88.2 Country average 92.3 88.9 76.3 87.3 13.3 59.1 Note: Regional quintiles are defined in the distribution of regional income, not country income. 172 STATISTICAL APPENDIX Table SA-30. Percentage of Families with Piped Water, by per Capita Household Income Annual per Inter- capita income Large mediate Small Urban Rural Country (dollars) cities cities towns average areas average 1974 0-75 84.1 85.0 64.6 75.2 15.5 44.6 76-150 89.2 81.7 79.9 84.0 16.9 50.6 151-350 96.0 90.9 85.8 92.2 24.8 67.2 351-700 99.5 95.7 96.9 98.2 23.6 84.9 Over 701 100.0 97.6 97.2 99.1 22.9 93.9 Country average 95.1 89.8 79.4 89.7 19.5 62.9 1970 0-75 80.2 82.4 58.1 70.4 6.2 37.5 76-150 84.1 78.4 78.0 80.4 11.4 46.1 151-350 93.5 89.9 82.3 89.9 18.4 63.4 351-700 98.9 97.1 94.1 97.9 19.4 83.4 Over 701 97.3 98.5 97.2 97.6 31.7 93.1 Country average 92.4 88.8 75.7 87.2 13.1 58.9 STATISTICAL APPENDIX 173 Table SA-31. Percentage of Families with Sewerage, by Regional Quintiles Inter- Income quintile Large mediate Small Urban Rural Country (poorest to richest) cities cities towns average areas average 1974 1 80.0 59.7 40.0 58.4 2.1 27.6 2 88.5 73.3 54.4 73.5 3.1 36.6 3 92.1 74.6 60.8 81.7 6.4 48.2 4 96.0 87.2 69.2 90.6 6.9 61.4 5 98.7 90.7 83.9 94.3 7.1 83.3 Country average 90.9 77.2 61.9 79.7 5.1 51.3 1970 1 76.3 58.3 32.8 55.3 .8 25.1 2 84.0 69.7 53.3 69.2 2.1 35.1 3 91.7 74.2 54.7 79.6 6.7 45.0 4 94.5 86.3 63.1 88.5 6.3 61.1 5 96.0 89.2 79.9 92.0 9.0 81.7 Country average 88.4 75.6 57.1 76.9 4.9 49.4 Note: Regional quintiles are defined in the distribution of regional income, not country income. 174 STATISTICAL APPENDIX Table SA-32. Percentage of Families with Sewerage, by per Capita Household Income Annual per Inter- capita income Large mediate Small Urban Rural Country (dollars) cities cities towns average areas average 1974 0-75 77.7 64.2 43.3 58.4 1.8 29.4 76-150 83.7 68.2 56.7 70.1 5.0 37.7 151-350 90.8 77.4 74.6 83.0 6.2 54.6 351-700 96.4 87.0 83.1 92.3 10.0 77.6 Over 701 98.6 92.7 83.4 95.5 10.2 89.7 Country average 90.9 77.4 61.3 79.6 5.0 51.2 1970 0-75 73.2 58.5 36.9 52.8 1.0 26.2 76-150 80.2 67.0 51.2 67.1 4.5 35.9 151-350 88.8 75.8 68.2 80.2 7.0 53.0 351-700 95.2 85.7 79.2 90.8 9.9 76.4 Over 701 95.5 91.9 83.4 93.4 10.2 87.8 Country average 88.4 75.4 56.3 76.6 4.8 49.2 STATISTICAL APPENDIX 175 Table SA-33. Probability of Ilaving Electricity, by per Capita Household Income Probability (percentage) Per capita Urban areas annual household Inter- income, 1974 Rural Small mediate Large (dollars)a Year areas towns cities cities 50 1970 4 65 82 93 (115) 1974 12 71 86 97 75 1970 7 70 87 94 (172) 1974 15 75 90 98 100 1970 8 72 89 95 (230) 1974 16 76 92 98 150 1970 10 75 91 96 (345) 1974 18 78 94 98 250 1970 11 76 93 97 (575) 1974 19 79 96 99 350 1970 11 77 94 97 (805) 1974 20 79 96 99 500 1970 11 78 94 97 (1,150) 1974 20 80 97 99 700 1970 12 78 95 98 (1,610) 1974 20 80 97 99 1,500 1970 12 79 95 98 (3,450) 1974 21 81 98 99 Note: Predicted values from the linear probability function. a. Values in parentheses show the corresponding monthly per capita income in Colombian pesos in December 1974 (to which the coefficients of the regressions apply). The exchange rate used is 27.6 pesos for each dollar. 176 STATISTICAL APPENDIX Table SA-34. Probability of Having Piped Water, by per Capita Household Income Probability (percentage) Per capita annual Inter- household income, Small mediate Large 1974 (dollars) Year towns cities cities 50 1970 70 75 82 (115) 1974 73 82 87 75 1970 74 81 87 (172) 1974 77 86 91 100 1970 76 85 89 (230) 1974 79 88 93 150 1970 79 88 91 (345) 1974 81 90 94 250 1970 80 90 93 (575) 1974 83 91 96 350 1970 81 91 94 (805) 1974 83 92 96 500 1970 82 92 9s (1,150) 1974 84 93 97 700 1970 82 93 95 (1,610) 1974 84 93 97 1,500 1970 83 93 96 (3,450) 1974 85 93 98 Note: Predicted values from the linear probability function. a. Values in parentheses show the corresponding monthly per capita income in Colombian pesos in December 1974 (to which the coefficients of the regressions apply). The exchange rate used is 27.6 pesos for each dollar. STATISTICAL APPENDIX 177 Table SA-35. Probability of hlaving Sewerage, by per Capita Household Income Probability (percent) Per capita annual Inter- household income, Small mediate Large 1974 (dollars)' Year towns cities cities 50 1970 48 54 78 (115) 1974 53 58 81 75 1970 53 64 83 (172) 1974 59 67 86 100 1970 56 69 86 (230) 1974 61 72 88 150 1970 59 74 88 (345) 1974 64 76 90 250 1970 61 78 90 (575) 1974 66 80 92 350 1970 63 79 91 (805) 1974 67 82 93 500 1970 63 81 91 (1,150) 1974 68 83 94 700 1970 64 82 92 (1,610) 1974 69 84 92 1,500 1970 64 83 93 (3,450) 1974 69 85 94 Note: Predicted values from the linear probability function. a. Values in parentheses show the corresponding rmonthly per capita income in Colombian pesos in December 1974 (to which the coefficients of the regressions apply). The exchange rate used is 27.6 pesos for each dollar. 178 STATISTICAL APPENDIX Table SA-36. Probability of Having Street Lighting, by per Capita Household Income Probability (percent) Per capita annual Inter- household income, Small mediate Large 1974 (dollars), Year towns cities cities 50 1970 60 66 89 (115) 1974 73 78 94 75 1970 65 74 91 (172) 1974 76 83 95 1oo 1970 68 78 93 (230) 1974 78 86 96 150 1970 70 81 95 (345) 1974 80 88 97 250 1970 72 84 96 (575) 1974 81 90 98 350 1970 73 86 97 (805) 1974 82 91 98 500 1970 73 87 97 (1,150) 1974 82 92 98 700 1970 74 87 97 (1,610) 1974 82 92 99 1,500 1970 74 88 98 (3,450) 1974 83 93 99 Note: Predicted values from the linear probability function. a. Values in parentheses show the corresponding monthly per capita income in Colombian pesos in December 1974 (to which the coefficients of the regressions apply). The exchange rate used is 27.6 pesos for each dollar. STATISTICAL APPENDIX 179 Table SA-37. Probability of Having Garbage Collection, by per Capita Household Income Probability (percent) Per capita annual Inter- household income, Small mediate Large 1974 (dollars), Year towns cities cities 50 1970 37 43 69 (115) 1974 37 46 70 75 1970 42 54 75 (172) 1974 43 57 76 100 1970 45 59 78 (230) 1974 47 63 78 150 1970 48 64 81 (345) 1974 51 69 81 250 1970 50 68 83 (575) 1974 54 73 84 350 1970 51 70 84 (805) 1974 55 75 84 500 1970 52 72 85 (1,150) 1974 56 76 85 700 1970 53 73 86 (1,610) 1974 56 78 86 1,500 1970 53 74 86 (3,450) 1974 57 79 86 Note: Predicted values from the linear probability function. a. Values in parentheses show the corresponding monthly per capita income in Colombian pesos in December 1974 (to which the coefficients of the regressions apply). The exchange rate used is 27.6 pesos for each dollar. References The word processed indicates works that are reproduced by mimeograph, Xerox, or in a manner other than conventional typesetting and printing. Berry, Albert, and Miguel Urrutia. Income Distribution in Colombia. New Haven: Yale University Press, 1976. CAJANAL (Caja Nacional de Previsi6n). "Prestaciones MVl6icas," Presupuiesto de Ingreso y Rentas. CAPRECOM (Caja de Previsi6n de Comuncaciones). "Servicios Medicos," Presupuesto de Entidades Decentralizadas, Informe Financiero de 1973, Reptiblica de Colombia. Compafiia Colombiana de Datos (COLDATOS). "Design of the Sample of the World Bank Study" ("Disefio de la Muestra del BanCO Mundial'). Study prepared for the Wkorld Bank, Bogota, 1976. Processed. Compafila Colombiana de Datos (COLDATOS). "Unit Cost of Education and Health Services in Colombia in 1974' [Costos Unitarios de los Servicios de Educaci6n y Salud en Colombia en 1974]. Study prepared for the World Bank, Bogota, 1976. Processed. Contraloria General de la Repfiblica de Colombia. "Informe Financiero de 1974.' Processed. DANE (Colombia Bureau of Census). "Household Survey,' 1970. Processed. DANE (Colombia Bureau of Census). "Investigaci6n sobre Establecimientos Educativos," 1972. Processed. Departmento Nacional de Planeaci6n, Colombia. "El Sector de Acueducto y Alcantarillados." Documento D.N.P., June 1976. Gutierrez de Gomez, Martha. Politica Tarifaria y Distribuci6n de Ingresos, Junta Nacional de Tarifas de Servicios PCiblicos, 1975. Processed. ICOLPE (Instituto Colombiano de Pedagogia). "Costos de la Educaci6n Media Oficial," 1972. Processed. INPES (Instituto para Promgramas Especiales de la Salud), "Censo de Institu- ciones Hospitalarias," 1970. Processed. 180 REFERENCES 181 International Financial Statistics. Jain, Shail. Size Distribution of Income: A Compilation of Data. Baltimore: Johns Hopkins University Press, 1975. Jallade, Jean-Pierre. Public Expenditures on Education and Income Distribu- tion in Colombia. World Bank Staff Occasional Papers, no. 18. Baltimore: Johns Hopkins University Press, 1974. Kravis, Irving B., Zoltan Kenessey, Alan Heston, and Robert Summers. A System of International Comparisons of Gross Product and Purchasing Power. Baltimore: Johns Hopkins University Press, 1975. Linn, Johannes F. "The Distributive Effect of Local Government Finances in Colombia: A Review of Evidence." World Bank Staff Working Paper, no. 235. Washington, D.C.: World Bank, March 1976. Linn, Johannes F. "Public Utilities in Metropolitan Bogota: Organization, Service Levels, and Financing," Urban and Regional Economics Division, Development Economics Department. Washington, D.C.: World Bank, May 1976. Processed. Lundquist, Lars. "Water and Sewerage Tariffs as a Mean for Income Redis- tribution in Colombia." Memorandum. Washington, D.C.: World Bank, October 23, 1973. Ministerio de Educaci6n. "Ejecuci6n Presupuestal." Processed. Ministerio de Educaci6n. "Estadisticas de la Educaci6n Primaria Oficial." Processed. Ministerio de Educaci6n, IcFES. "La Educaci6n en Cifras, 1970-1974." December 1975. Processed. Ministerio de Educaci6n. "Oficina Coordinadora de los FER y Oficina de Planeamiento de la Educaci6n." Processed. Netter, J., and E. Scott Maynes, "On the Appropriateness of the Correlation Coefficient with a 0,1 Dependent Variable." Journal of the American Statistical Association (June 1970). Rama, German. "Origen Social de la Poblaci6n Universitaria." Universidad Nacional, 3, August 1969. Processed. Urrutia, Miguel, and Clara E. de Sandoval. "Politica Fiscal y Distribuci6n del Ingreso en Colombia," Revista Banco de la Repuiblica, July 1974. World Bank. "Economic Position and Prospects of Colombia," vol. II. Report no. 1548. A restricted-circulation document. Washington, D.C., 1977. Processed. Index Adult retraining, 6, 35 Consumption of public services: house- Age of head of household, 31, 124 hold versus countrywide survey of, Agriculture, public investment in, 33- 13-14; inequality in, 4; supply- 35; loans for, 143-46 demand analysis of, 5, 10, 27-33; urbanization and, 14-16 Costs of services, 12; for education, 57-58; for health, 84-88; for icss and NHS, 2 1-22 Beneficiaries of services, 5; rural-urban distribution of, 6 Berry, Albert, 8n, 18, 19, 41, 42 Budget, national, 12-13 Bureau of Census, 16, 36-37, 58 DANE. See Departamento de Estadistica DANE-Polibio C6rdoba income distri- bution study, 43 de Gomez, Martha Gutierrez, 140-42 Demand for services. See Supply-de- Caja Agraria, 7, 35; agricultural loans mand analvsis of public services by, 143-46 Departamento de Estadistica (DANE), Caja Nacional de Previsi6n (CAJANAL), 16, 36, 62, 73n 77-78 de Sandoval, Clara E., 70-72 CAJANAL. See Caja Nacional de Previ- Dirt floor, and demand for public utili- si6n ty services, 29, 122, 123, 124, 125 Caja de Previsi6n de Communicaciones Distribution of income. See Income dis- (CAPRECOM), 78 tribution Cajas Publicas. See Social Security of the Public Sector CAPRECOM. See Caja de Previsi6n de Communicaciones COLDATOS. See Compafiia Columbiana Education: cost of, 12, 57-58; and de- de Datos mand for services, 6, 30; inequality Colombian Institute of Social Security in consumption of, 4; studies on dis- (icss): costs of, 21; funding for, 77; tributive effect of expenditures on, hospitals under, 83-85, 91; subsidy 70-76; and visits to physicians, 98, to, 22-23, 92-93: workers' contribu- 99. See also Education subsidies; tions to, 80-81 Higher education; Primary educa- Compafnia Colombian de Datos (COL- tion; Secondary education; Student DATOS), 14, 50, 57n enrollment 183 184 INDEX Education subsidies, 21, 36; income 22-23, 65, 66-68; health services con- distribution and, 24-25, 68-70; per sumed by, 86-88; health subsidies to, capita, 22-23, 65; as percentage of 22-23, 81, 86- 89, 94-97; income dis- income, 66-68; per household, 22-23, tribution by, 19-21; income quintile 65, 66-68; per student, 56-65 classification of, 16-17; inequality in Electricity: dirt floor as demand varia- consumption of, 4; occupation of ble for, 29, 123; factors influencing heads of, 45-47; private medical ser- consumption of, 6; investment in, 5, vices used by, 97-100; probability of 26, 106-09; probability of having, having a public service by, 27-33; 28-29, 123-24; subsidies for, 140-42; public utility investment per, 107, tariffs for, 138-39 109; public utility services for, 107, Exchange rate, 20-21 109, 118; Social Security System affil- Expenditures for public services: distri- iation of, 92-95; socioeconomic char- butive effect of, 4-5; as percent of acteristics of, 45, 47-49; student en- total government expenditures, 36 rollment per, 51-56; subsidies per, 5-6, 10-11; years of schooling of heads of, 30, 44-45 Farmers: income distribution of, 146; subsidies from loans to, 143-46 Fellowships, educational, 6, 148 IcOI.PE. See Instituto Colombiano (le FER. See Regional Educational Fund Pedagogia icss. See Colombian Institute of Social Security Income: education subsidy as percent- age of household annual, 67-68; gov- Garbage collection services, in urban ernment expenditures as share of na- areas, 148-49 tional, 3; per capita, 3-4, 6; and Gini coefficients, 18, 41-42, 43 visits to physicians, 98, 99. See also Gross national product (GNP), 3; educa- Income distribution tion subsidies as percent of, 21 Income distribution: by family loca- tion, 40; farmers', 146; household, 19-21; as poverty indicator, 38-43; rural-urban, 17-18; and subsidies to Health centers, 77; services provided education and health, 24-25; surveys by, 90 comparing, 42-43 Health services: cost of, 12, 84-88; fi Instituto Colombianode Pedagogia, 58, nancing of, 13, inequality in con- 59 sumption of, 4; private, 97-100. See Investment. See Public investment also Health centers; Health subsi- dies; Hospitals; Physicians Health subsidies, 5; amount of, 82; to households, 22-23, 81, 86-89, 94-97; Jain, Shail, 40n income distribution and, 24-25, 95- Jallade, Jean-Pierre, 72-76 96; by location of institution, 85; per capita, 22-23 Higher education: and parents' level of schooling, 71; student enrollment in, Kravis parity rate, 20-21 55-56; subsidies to, 5, 25, 63 Hospitals, public, 77; services provided by, 82-85, 87, 90-92; subsidies to, 82-86; in urban versus rural areas, Labor: contribution to health institu- 86-88 tions by, 80-81; incidence of health Households: education subsidies per, costs borne by, 81-82, 100-04 INDEX 185 Linn, Johannes F., 142n Private schools, subsidies to, 21, 62-63 Lorenz curve, 19, 69 64-65 Lundquist, Lars, 141 Psu. See Primary sampling units Public investment, in services, 5, 6, 26- 27; in agriculture, 33-35; benefits from, 8-9; in electricity, 106-08; Mavnes, E. Scott, 112n factors influencing demand for, 6; Migrant status, and demand for public national budget and, 12-13; in roads, utility services, 6, 122, 124, 130 7, 33-34; supply-demand analysis Ministry of Education, 51, 52, 62, 63n of, 10, 27-33 Ministry of Health, 22, 77, 78n Public utility services: families with, 110; investment in, 106-09; to low- income groups, 127; marginal proba- bility of having, 129-31; new house- National Health System (NHS) 5: com- holds connected to, 109; probability ponents of, 21, 77; funding for, 77- of connection to, 110-18, 119-20; in 78; hospitals under, 82, 90; services rural areas, 106, 134-38; supply-de- to rural households by, 88; subsidies mand analysis of, 118-23. See also to, 22, 94-96; visits to physicians in Electricity; Piped water, Sewerage institutions of, 97. See also Colom- bian Institute of Social Security; So- cial Security of the Public Sector National Tariff Bureau for Public Ser- Rama, German 71n vices, 140 Regional Educational Fund (FER), 58 Netter, J., 112n Roads, public investment in, 7, 33-34 NHS. See National Health System Rural areas: consumption of services by, 16; education subsidies in, 65; health subsidies in, 93; hospitals in, 86-88; income distribution in, 17-18, Occupation, of heads of households, 21; household visits to physicians 45-47 in, 98-99; public utility services in, 106, 134-38; substitution for public utility services in, 138 Physicians, household visits to, 97- 100 Piped water: dirt floor as demand vari- able for, 29; factors influencing con- Secondary education: enrollment in, sumption of, 6; income distribution 52-55; subsidies to, 59, 61, 63-64 of households with, 26-27; inequality SENA. See Servicio Nacional de Aprend- in consumption of, 4; investment in, izaje 5, 26, 106-09; probability of house- Servicio Nacional de Aprendizaje holds having, 30-31, 124; subsidies (SENA), training programs of, 6, 35, for, 140-42; tariffs for, 138-39 146-47 Poverty: "basic needs" approach to Sewerage services: dirt floor as demand alleviating, 4; income distribution as variable for, 29; factors influencing indicator of, 38; percent of house- consumption of, 6; inequality in con- holds below, 21 sumption of, 4; investment in, 5, 26, Pricing policy, subsidy, 9 106-09; probability of having, 32-33, Prieto income distribution study, 43 125-26 Primary sampling units (Psu), 16 Social Security of the Public Sector Primary education: enrollment in, 51, (Cajas Publicas) 21, 77; hospitals 53-55, 62; financing of, 57; subsidies under, 84, 91; subsidies to, 92-93; to, 5, 57-62, 63-64 workers' contributions to, 80 186 INDEX Social Security System, 5; components Tariffs: for electricity and water, 138- of, 21; funding for, 77-78; health 39, 140, 141; subsidies induced by, subsidies to households affiliated 139-40 with, 92-95; sources of funding for, 80-81; total subsidy to, 94-95; visits to physicians in institutions of, 97. See also Colombian Institute of So- cial Security; Social Security of the Universities. See Higher education Public Sector Urban areas: and consumption of serv- Street lighting, 6 ices, 14-16; education subsidies in, Student enrollment: by income group, 65; garbage collection services to, 50-51, 71; by location, 53-55; in uni- 148-49; health subsidies to, 93; hos- versities, 55-56 pitals in, 86-88; income distribution Subsidies, for consumption of services, in, 17-18, 21; household visits to 5; from agricultural loans, 143-46; physicians in, 98-100; public utility induced by tariffs, 139-40; per house- services in, 118-34; substitutes for hold unit, 5-6, 10-11; pricing policy public utility services in, 132-34 for, 9; for public utilities, 139-42. Urrutia, Miguel: on distributive effect See also Education subsidies; Health of expenditures on education, 70-72; subsidies study on income distribution by, 8n, Supply-demand analysis of public serv- 18, 19, 41, 42 ices, 5, 10, 27-33 Survey of government expenditures: classification of data for, 5; questions addressed by, 9-10; results of, 5-7; strata for, 14-16; tradeoffs in testing, Water. See Piped water 14 World Bank, 144n The full range of World Bank publi- cations, both free and for sale, is described in the Catalog of World Bank Publications, and of the con- tinuing research program of the World Bank, in World Bank Re- search Program: Abstracts of Current Studies. The most recent edition of each is available without charge from: PUBLICATIONS UNIT THE WORLD BANK 1818 H STREET, N.W. WASHINGTON, D.C. 20433 U.S.A. Also from Oxford andi the World Bank PtJBLIC EXPENDITtTRE IN MALAYSIA: WHO BENEFITS AND WHY Jacob Meermaii Information about the mechanisms of public expenditure at work is generally meager. But this study, relying on a specially developed sample survey of pensinsular M\lalaysia, generates its ow n data for costs and household consuLimption of services to analyze the public cost of providing basic needs. I\lany new and valuable observations address the question, "Who receives public services and why?" A principal conitributioni of the study is to move the discussioni of public expenditure out of the narrow framework of national accounts and to examine the complexity of benefits and their locus, duration, and valuation. Ahbout 400 pages. Figuires, tables, bibliographv. Available in cloth ndel paper editions. Jacket design by Carol Crosby Black REDISTRIBUTION WITH GROWTH Hollis Chenery, Montek S. Ahluwalia, C. L. G. Bell, John H. I)uloy, ancl Richardl Jolly -7 "A major contribution to the literature of incomile distribution in less CD developed countries." -- Jouirna(l of D)ereloping A reo(s "''xceptionally valualble analysis of dev7elopment l)olicies . ..?edistrilitbion7 rn wilh (;rowth, a model handbook for planiners, is also an extremely usefuil , guide to the state of the discipline of development economics. ---Journal rn of IE.conomiic Litera/tre 71 324 pages. Figures, tables, bibliography. AVailabhle in cloth arnd paper editioins. -7 (GROWTH \WITH EQUIITY: THE TAIWAN CASE Q Johni C. H. Fei, Gustav Ranis, anid .Shirley W. Y. Kuo Bly providing a framew-ork for analyzing relations between growth and income distributioni in a developing economy, this volimc explains whyy rariil growtth often leads to unieven distribution of wealth. A. detailed case CD study consi(lers how TI'aiwan avoided this generally observed association. T1he methodology used can he ap)plied to the study of relevant behavioral interactions b)etwveen growth and equiity facing this "tiuzioets curve conflict. 27 .4About 228 pages. Figures, ltables, bibliography. - .47ailable int c/loth and paper editions. I NC()M EI IN EQI TALITFY AN I) POVERTY: r METUI()I)S OFIESTI'lMATI()NANI) 1P)11(OLC AIll'AI11()'ATONS Nanak IKakwam ! 'Ihis study offers new techniques, derived froni actual data, to analyze problems of size distribuItion of incomie and to evaluate altrinIat i, y fiscal policies. l3oth ethical evaluation and statistical measurement arc con- sidere(l. 1[he boosk sNystematizes existing knowledge and intr'od luces a H numnbl er of new Ificlinigs. Ahbout 320 paiges. Figures, bibliography. Avaiilable in clolh/ edition. m Oxford l niversity Press ISBN (---1(- 520098)--5 Oxford