Embed Size (px)
Citation preview
RESEARCH REPORT
THE PILOT FOOD PRICE SUBSIDYSCHEME IN THE PHILIPPINES:ITS IMPACT ON INCOME, FOODCONSUMPTION, ANDNUTRITIONAL STATUS
Marito GarciaPer Pinstrup-Andersen
August 1987
INTERM4TION/1F O O DPOLICYRESEARCHINSTITUTE
The International Food Policy ResearchInstitute was established in 1975 to identifyand analyze alternative national and inter-national strategies and policies for meetingfood needs in the world, with particular em-phasis on low-income countries and on thepoorer groups in those countries. While theresearch effort is geared to the precise ob-jective of contributing to the reduction ofhunger and malnutrition, the factors involvedare many and wide-ranging, requiring analy-sis of underlying processes and extendingbeyond a narrowly defined food sector. TheInstitute's research program reflects world-wide interaction with policymakers, adminis-trators, and others concerned with increasingfood production and with improving theequity of its distribution. Research resultsare published and distributed to officials andothers concerned with national and inter-national food and agricultural policy.
The Institute receives support as a consti-tuent of the Consultative Group on Interna-tional Agricultural Research from a numberof donors including the United States, theWorld Bank, Japan, Canada, the UnitedKingdom, the Netherlands, Australia, theFord Foundation, Italy, Norway, the FederalRepublic of Germany, India, the Philippines,Switzerland, Belgium, and the People's Re-public of China. In addition, a number ofother governments and institutions contrib-ute funding to special research projects.
Board of Trustees
Dick de ZeeuwChairman, Netherlands
Ralph Kirby DavidsonVice Chairman, U.S.A.
Eliseu Roberto de Andrade AlvesBrazil
Yahia BakourSyria
Ivan L. HeadCanada
Dharma KumarIndia
Anne de LattreFrance
James R. McWilliamAustralia
Philip NdegwaKenya
Saburo OkitaJapan
Sukadji RanuwihardjoIndonesia
Theodore W. SchultzU.S.A.
Leopoldo SolisMexico
M. SyeduzzamanBangladesh
Charles Valy TuhoCote d'lvoire
John W. Mellor, DirectorEx Officio, U.S.A.
THE PILOT FOOD PRICE SUBSIDYSCHEME IN THE PHILIPPINES:ITS IMPACT ON INCOME, FOODCONSUMPTION, ANDNUTRITIONAL STATUS
Marito GarciaPer Pinstrup-Andersen
Research Report 61International Food Policy Research InstituteAugust 1987
Copyright 1987 International Food PolicyResearch Institute.
All rights reserved. Sections of this report maybe reproduced without the express permissionof but with acknowledgment to the InternationalFood Policy Research Institute.
Library of Congress Catalogingin Publication Data
Garcia, Marito, 1951-The pilot food price subsidy scheme in the
Philippines.
(Research report / International Food PolicyResearch Institute ; 61)
Bibliography: p. 90.1. Food relief—Philippines—Case studies.
2. Food consumption—Case studies. I. Pinstrup-Andersen, Per. II. International Food PolicyResearch Institute. III. Title. IV. Series: Researchreport (International Food Policy Research Insti-tute) ; 61.
HV696.F6G36 1987 363.8'82'09599 87-17237ISBN 0-89629-063-8
CONTENTS
Foreword
1. Summary 9
2. The Pilot Food Price SubsidyScheme ' 2
3. The Study Areas 17
4. Assessment of Implementation 23
5. Analytical Methodology 27
6. Descriptive Analysis of Income,Consumption, and Nutritional Pat-terns 38
7. A Comparative Analysis of the Ef-fects of the Food Price SubsidyScheme 52
8. A Multivariate Analysis of the Ef-fects of the Food Subsidy Scheme 62
9. Cost-Effectiveness Analysis 70
10. Conclusions 78
Appendix: Supplementary Tables 80
Bibliography 90
TABLES
1. Price discounts on rice and cook-ing oil and their share of the retailprice, 1983/84 16
2. Study areas of the Philippine pilotfood price subsidy scheme 19
3. Characteristics of sample house-holds compared with the rest ofthe Philippines 21
4. Experimental design of the pilotfood price subsidy study 29
5. Overview of the cost-effectivenessmeasures used in this study 35
6. Average incomes and incomesources by province, survey round,occupation, and income quartile 39
7. Percent of total household ex-penditures on major commoditygroups, by province, survey round,occupation, and income class,1983-84 40
8. Proportion of the food budget al-located to various foods, by prov-ince, survey round, occupation,and income group, 1983-84 42
9. Diet composition by food groups,survey round, occupation, and in-come group, 1983-84 44
10. Calorie acquisition by food groups,survey round, occupation, and in-come group, 1983-84 46
11. Calorie intakes of groups of indi-vidual household members 48
12. Nutritional status of preschoolchildren, 1983-84 51
13. Weekly food expenditures foreach subsample and survey round,1983-84 53
14. Change in household food expen-ditures due to price subsidy ornutrition education 53
15. Daily calorie acquisition for eachsubsample and survey round 54
16. Change in daily calorie acquisitiondue to price subsidy or nutritioneducation 54
17. Daily household calorie consump-tion for each subsample and sur-vey round 55
18. Change in daily household calorieconsumption due to price subsidyor nutrition education 56
19. Individual calorie adequacy by ageand gender for each subsampleand survey round 57
20. Change in calorie adequacy for se-lected groups of individuals dueto price subsidy or nutrition edu-cation
21. Mean weight-for-age of all pre-schoolers for each subsample andsurvey round
22. Change in weight-for-age of pre-schoolers due to price subsidy ornutrition education
58
60
60
23. Mean height-forage of all pre-schoolers for each subsample andsurvey round 61
24. Change in height-for-age of pre-schoolers due to price subsidy ornutrition education
29. Key coefficients for nutritionalstatus regressions
61
25. Key coefficients for householdfood expenditure and food acqui-sition regressions 63
26. Parameters for estimating the im-pact of the pilot subsidy schemeon household food expendituresand calorie acquisition 64
27. Key coefficients and parametersfor estimating the impact of thepilot subsidy scheme on the acqui-sition of rice, maize, oil, and fish 67
28. Regressions of calorie intake ofchildren aged 13 to 83 months 68
69
30. Fiscal costs of the pilot food pricesubsidy scheme and their distribu-tion 71
31. Estimated net impact of the pilotsubsidy scheme on food acquisi-tion and nutrition 72
32. Cost-effectiveness indicators 73
33. Cost-effectiveness comparisonwith other programs 76
34. Regression results for householdfood expenditures and food acqui-sition 80
35. Regression results for acquisitionof rice, maize, oil, and fish 81
36. Regression results for calorie in-take of children aged 13 to 83months 82
37. Regression results for weight ofchildren aged 13 to 83 months 83
38. Regression results for weight as apercent of standard weight-for-age,children aged 13 to 83 months 84
39. Regression results for height as apercent of standard height-for-age,children aged 13 to 83 months 86
40. Regression results for z-scores forweight-for-age, children aged 13to 83 months 87
41. Regression results for z-scores forweight-for-height, children aged13 to 83 months 88
ILLUSTRATIONS
1. Pilot food discount project deliverysystem 13
2. Map of the Philippines indicatingstudy areas 18
3. Simplified version of the conceptualmodel underlying the analyses 27
4. Conceptual model underlying themultivariate analyses 30
FOREWORD
Food costs represent more than 80 per-cent of total expenditure for the poorestfamilies in developing countries and are asignificant budget item even for upper- andmiddle-income people. Hence, the provi-sion of food subsidies as a means of incomedistribution in developing countries has hada strong attraction for policymakers. Subsidycosts compose as much as 15 to 20 percentof national budgets, providing importanttrade-offs among inflation rates, exchangerates, food subsidies, other types of incomedistribution measures, and development ex-penditures, which raise incomes in thelonger run. In view of this, the InternationalFood Policy Research Institute has, rightfrom its inception, included studies of foodsubsidy schemes as a major part of its pro-gram.
If government expenditure on food sub-sidies is to be reduced, it is immensely im-portant to know who the beneficiaries offood subsidies are, particularly regardingrural-urban distribution and the distributionamong various income groups. There is con-siderable interest in the impact of varioustargeting schemes designed to ensure thata high proportion of the benefits go to lower-income people.
The research for this report by MaritoGarcia and Per Pinstrup-Andersen was con-ducted in the Philippines in collabora-tion with the National Nutrition Councilof the Philippines. It was financed by theUnited Nations Development Programme,the Government of the Philippines, andthe U.S. Agency for International Develop-ment.
The particular focus of the study is thetargeting of food subsidies to the poor bydirecting subsidies to regions with particu-larly high proportions of low-income people.Because administrative costs represent alarge share of the cost of targeting, devicessuch as targeting by location help reducethese costs.
In keeping with the tradition of suchstudies at IFPRI, a carefully drawn sampleof households is analyzed in detail to deter-mine the impact of food subsidies. The studyis designed to permit an analysis of the costof the delivery system, and comparisons aremade between this and other means oftargeting. The results should be useful notonly in the Philippines but in other coun-tries that are contemplating means of reduc-ing the cost and increasing the effectivenessof measures to increase the food consump-tion of low-income people. The study alsoanalyzes the interaction of nutrition educa-tion with food subsidies, finding that nutri-tion education is much more valuable inconjunction with increased food consump-tion. This information should be useful ina developmental as well as a distributionalcontext.
Although this study does not analyze thereturns to expenditure on food subsidies ifthose returns have been spent in some otherway, it provides a basis for such an examina-tion by presenting comparative cost figures.
John W. Mellor
August 1987Washington, D.C.
ACKNOWLEDGMENTS
This consumer food subsidy experimentwould not have been possible without thesupport and cooperation of the PhilippineNational Nutrition Council (NNC) and its di-rector, Delfina B. Aguillon, to whom we aregrateful for continuous support throughoutthe study. The massive day-to-day logisticalproblems of running this pilot program fellsquarely on the shoulders of the field person-nel of NNC and the Ministry of Agriculture,who did an outstanding job. In particular,the home management technicians in theprovinces of Abra, Antique, and South Cot-abato were instrumental in the monitoringof the sample households and making surethat designated shops were open to sell riceand cooking oil at subsidized prices. They,too, were competent interviewers for thefour consumption and socioeconomic sur-veys that were conducted as part of the study.
We gratefully acknowledge the assis-tance of the Nutrition Foundation of thePhilippines and the Food and Nutrition
Research Institute in training our surveyteams. Melba Aligaen provided excellentorganization of the supervisory and fieldsurvey teams along with Benilda Labayog,who was seconded to the project from theBureau of Agricultural Extension. MindaCaedo was a great help in the initial de-sign of the logistics of the subsidy experi-ment.
This report benefited from the valuablecomments and suggestions from many col-leagues including John W. Mellor, JereBehrman, Harold Alderman, Joachim vonBraun, Eileen Kennedy, and Gunvant Desai.We are also indebted to Elizabeth Jacintoand Fe Lisondra, who provided excellentcomputing assistance.
Finally, we wish to thank the UnitedNations Development Programme, the Gov-ernment of the Philippines, and the U.S.Agency for International Development Mis-sion in Manila, all of whom provided finan-cial support for this research.
1SUMMARY
This report presents the results of a pilottargeted food price subsidy scheme im-plemented in three provinces of the Philip-pines for 12 months beginning in mid-1983.It assesses the economic and nutritional ef-fects of the scheme, analyzes its technicaland administrative feasibility, and considerspossible alternatives.
The scheme consisted of price discountson rice and cooking oil and a nutrition edu-cation component. It was made available tohalf of 14 villages selected for their highincidence of malnutrition and poverty. Theother half acted as a control population. Be-cause targeting was geographical, all house-holds in villages selected to receive thediscount were eligible. Each household wasissued a ration card indicating its monthlyquota of rice and oil, based on family size.The rice ration subject to a price discountwas only about half the amount usually con-sumed by most of the households, but theoil ration exceeded the amount usually pur-chased prior to the subsidy. Thus, consumerrice prices were not reduced at the margin,but oil prices were.
To evaluate its effects, one-third of thehouseholds from the selected villages weresurveyed. Both comparative and multivariateanalyses were made to estimate the impactof the scheme and its components on house-hold food expenditures, acquisition, andconsumption. Effects on consumption bypreschoolers and their nutritional statuswere also monitored. Data on a variety ofsocioeconomic, environmental, and biolog-ical variables relating to food consumptionand nutrition were collected before, during,and after the scheme. Two alternative meth-ods were used to collect data on householdfood consumption. In the food acquisitionmethod, called flexible period recall, allfoods obtained by family members for aweek were recorded. Under the food con-sumption method, all food consumed by
household members was weighed for a 24-hour period. Food consumption by individ-ual preschoolers was estimated on the basisof 24-hour food weighing for a subsampleof the survey households.
The average monthly income of the sam-ple households was 910 pesos, which issubstantially lower than the Philippine aver-age. Total expenditures of the sample house-holds were 84 percent of total incomes, onaverage. The poorest showed a dissaving,while those who were better off were ableto save significant amounts.
Although the total sample was drawnfrom agricultural and fishing villages in ruralareas, only about one-third of the incomesoriginated directly from farming and fishing;almost one-half came from salaries andwages. Food accounted for 79 percent oftotal expenditures of the poorest quartileand 71 percent, on average, for the sampleas a whole.
About 40 percent of the average foodbudget was spent on rice. Food expenditurepatterns and diet composition differed con-siderably among occupational and incomegroups, with the poorest obtaining a largershare of calories from rice and maize. Aver-age daily calorie consumption per adultequivalent unit (AEU) was 1,700 calorieswhen based on 24-hour food weighing and1,837 calories when based on flexible pe-riod recall. The 7 percent difference isexplained primarily by the exclusion ofweekends from food weighing.
Individual food consumption data col-lected from a subsample of 140 householdsindicate that the distribution of food withinthe households was biased in favor of adults.The calorie adequacy rate is estimated tobe about 0.80 for adults, 0.60 for male pre-schoolers, and 0.55 for female preschoolers.School-age children and pregnant and lactat-ing women also had low calorie adequacyrates.
The average weight of the sample pre-schoolers was about 83 percent of the stan-dard weight-for-age. About one-fourth of thepreschoolers were malnourished (less than75 percent of the standard). Although mostof the sample households were poor, theextent of malnutrition varied considerablyamong occupational groups. Forty percentof the preschoolers of hired fishermen weremalnourished compared with 15 percent forprofessionals and salary earners. As incomeincreased, malnutrition decreased, rangingfrom 30 percent among the poorest quartileto 16 percent among the richest.
According to the comparative analyses,the subsidy component of the scheme causedan increase in household food expendituresand calories acquired and consumed, as wellas in calories consumed by most individualhousehold members. Although adults ob-tained the largest share, the average weightof preschoolers also increased.
The second component of the scheme—nutrition education—had a small positiveeffect in households where it was accompan-ied by the subsidy. When education wasprovided without additional purchasingpower, however, no effect could be de-tected. But the subsidy without the educa-tion component was still effective.
Multivariate analysis confirms and fur-ther refines these findings. The impact of theprice subsidies on household food expendi-tures and acquisition was highly significant.Increased purchasing power, lower oil prices,and the subsidy itself all contributed.
The income elasticities of total food ex-penditures and total calorie acquisition indi-cate that a 10 percent increase in householdincomes could result in a 7 percent increasein food expenditures and a 3 percent in-crease in calorie acquisition. Each additionalpeso of income from sources other than thesubsidies is estimated to expand daily cal-orie acquisition by about 150 calories perAEU. However, an additional peso of pur-chasing power of food subsidies is estimatedto result in an increase of about 230 calories.These findings indicate that consumers aremore likely to increase their food consump-tion if foods are subsidized than if incomesare raised directly.
When multivariate analysis is employed,the nutrition education component of thescheme shows no significant impact on house-hold food expenditures and acquisition, butfood consumption and the nutritional statusof preschoolers are strongly affected, indi-cating that the nutrition messages increasedthe focus on children.
Estimation of the effects of the schemeon the nutritional status of preschoolers isbased on five indicators: weight; weight asa percent of standard weight-for-age; thez-score (a method of standardizing distri-bution) of weight-for-age; the z-score ofweight-for-height; and height as a percentof standard height-for-age. The nutritionalstatus of preschoolers was most affectedby household incomes and calorie con-sumption—variables influenced by the sub-sidies.
The evidence indicates that the subsidyhad positive effects on both households andpreschoolers. The scheme resulted in netincreases in household calorie acquisitionof 136-138 calories per AEU per day, whichis about 7 percent of current calorie con-sumption; in calorie consumption by pre-schoolers of 31 -55 calories per child per day(4-6 percent of current consumption); andin the weight of preschoolers of 0.12-0.14kilograms.
Eighty-four percent of the cost of thescheme was the subsidy itself. Administra-tive costs accounted for about 9 percent andthe incentive payment to retailers to assureefficient distribution of subsidized food wasabout 7 percent. The fiscal cost of each$1.00 transferred to participating house-holds was $1.19 in U.S. dollars. However,if only transfers to households with mal-nourished preschoolers are considered abenefit, the cost increases to $3.61. Simi-larly, the annual cost of a net increase incalorie consumption of 100 calories perAEU per day among all households is esti-mated to be $6.75 per AEU. The amountincreases to $ 13.66 if only the food receivedby households with malnourished pre-schoolers is considered.
The annual cost of eliminating caloriedeficiencies in the study population is esti-mated to be $25 per AEU. Adding 1 kilo-
10
gram to the weight of each preschoolerwould cost $24 per year. If only weightgains among the malnourished are countedas benefits, the cost would increase to $56.
In comparison with other food and nu-trition programs, the scheme's cost-effec-tiveness is favorable. Costs were low be-cause, first, geographical targeting based ongrowth monitoring costs less than targetingbased on household income levels; second,the use of existing private-sector retail out-lets for the distribution of subsidized foodscosts less than a separate distribution net-work; and third, the use and expansion ofexisting local bureaucratic structures costless than the creation of a new and indepen-dent structure.
If the sole goal of such a scheme is toexpand food consumption by households withmalnourished preschoolers and to improvethe nutritional status of these preschoolers,its cost-effectiveness could be significantly im-proved by additional targeting based ongrowth monitoring. However, the calorieadequacy rates of school-age children werealmost as low, which casts doubt on the wis-dom of targeting to one group of childrenalone.
Finally, the study finds a strong relation-ship between malnutrition and poverty.Groups such as landless farm workers andtenant farmers are most likely to be mal-nourished because of their limited purchasingpower.
11
THE PILOT FOOD PRICE SUBSIDY SCHEME
IntroductionIn 1980 the Philippine Ministry of Agri-
culture together with the National NutritionCouncil and the National Economic Devel-opment Authority formulated the PhilippineFood and Nutrition Plan (FNP) to providethe framework for action and research innutrition, food consumption, and food pro-duction for the 1980s. The main objectivesof the Plan were to increase and diversifythe production of food and other agriculturalcommodities; to improve the quality of thediet of all Filipinos; and to assure a basicminimum diet for the undernourished.These objectives were to be achievedthrough the following broad strategies:stimulating the growth of the food economy;increasing export earnings and producingimport substitutes; maintaining consumerprices at reasonable levels; and undertakingimmediate programs to prevent malnutri-tion and to correct the most serious nutri-tional deficiencies, particularly calorie defi-ciency.
A food price discount or subsidy programwas one of the immediate and short-termprograms identified under the FNP to reducecalorie deficiencies among low-incomehouseholds.1 The scheme was conceived asa stop-gap measure aimed at bringing aboutimmediate improvements in the energy in-takes of malnourished individuals. Althoughviewed as a temporary measure, its phase-out was inextricably linked to the successof other elements of the FNP that seeklonger-term solutions, such as increasingproductivity and incomes of the beneficiar-ies, which would reduce the need for foodsubsidies.
The food subsidy scheme evolved out ofthe recognition that undernutrition cannot
be solved effectively without expanding theability of the poor to acquire food. Althoughthe food production program of the govern-ment, particularly for rice, was a success,there were clear indications that theseachievements, though important, were notsufficient to bring food consumption andnutritional status of the majority of the poorup to an acceptable level.
Past programs in nutrition have pointedtoward direct intervention, such as supple-mentary feeding, health protection, nutri-tional rehabilitation, nutrition education,and family planning. Less attention has beenfocused on programs to improve the acces-sibility of food, particularly to the more de-prived groups. Since 1981, a marketingscheme called KADIWA has been promotedto lower the costs of food in the main urbanareas. The level of coverage, however, islow, and the scheme has reached only asmall portion of the disadvantaged groups.Although the KADIWA program focuses onfood, it has no expressed nutritional objec-tive.
To assess the feasibility of FNP's pro-posed food subsidy program and to help as-sure that its design would be cost-effective,a one-year pilot scheme was implementedduring 1983-84 in three provinces of thePhilippines. The pilot program was specifi-cally designed to permit its evaluation fromviewpoints of incomes, nutrition, and tech-nical and administrative feasibility. In additionto periodic surveys of selected householdsin the areas where the pilot scheme wasimplemented, data for the economic andnutritional assessments were obtained fromextensive monitoring of all parts of thescheme. This report presents findings fromthe surveys and monitoring.
1 Philippines, Ministry of Agriculture, National Nutrition Council, and National Economic and DevelopmentAuthority, The Philippine Food and Nutrition Plan for the 80s (Manila: Ministry of Agriculture, 1980).
12
Description of the PilotScheme
The principal element of the pilot sub-sidy scheme was a reduction in the priceof rice and edible oil offered for sale inselected areas identified as having high ratesof malnutrition.2 Nutrition education formedanother element. Figure 1 is a simplifiedflowchart of the operation of the pilot foodsubsidy scheme. Each household in the proj-ect area was issued a ration (discount) card,which guaranteed a monthly quota of riceand cooking oil at a subsidized price. Par-ticipating families could use the ration cardonly in the accredited stores and only for
the discounted purchase of rice and edibleoil. The coded, nontransferable card, whichshowed the monthly quota of the householdbased on household size, provided space torecord the purchases and signature of thestore owner. For each sales transaction, theparticipating households were required tosign a sales logbook and the retailer wasrequired to sign in a space provided on theration card. The distribution and certifica-tion of the ration cards were handled by theMinistry of Agriculture through its localofficers—the home management technicians(HMTs). Cards were valid for a period ofone month, and they were issued at thebeginning of each month.
Figure 1—Pilot food discount project delivery system
CommercialRice/Cooking Oil
Wholesalers
Reimbursement ofDiscount Value
Food Discount Project
Special Savings Deposit
Home Management_ _ Technicians
Barangay LicensedGrain Retailer
(Private)
Target Beneficiaries
Note: NFA is the National Food Administration, NNC is the National Nutrition Council, and MA is the Ministryof Agriculture.
'• These areas are described in detail in the next chapter.
13
The retail distribution outlets chosen forthe pilot scheme were the neighborhoodvariety stores (called sari-sari stores) usuallylocated within each village. The advantageof using local neighborhood stores in thescheme was their accessibility to the targetbeneficiaries. Locating the outlets in theneighborhood is compatible with the food-buying practices of the poor, who often buyfood on a day-to-day basis and in smallamounts. A study of such practices indicatesthat, among the very poor, small but fre-quent purchases are resorted to deliberatelyto ensure that food will not be consumedat once—a survival technique that allowsfamilies to stretch their budgets.3
Sari-sari stores are typically small, family-run enterprises that carry food, beverages,household items, tobacco and cigarettes,and other items that are normally sold inpublic markets. They perform a unique func-tion in the life of the poor because theyserve immediate needs of households. With-out such outlets, households would have tobuy from the nearest market town, whichin many cases is far from the village.
Sari-sari stores normally operate withsmall revolving capital; not all are able tocarry large quantities of rice. Under Philip-pine law, a license from the National FoodAuthority is necessary for any enterprise tosell grains. Under the pilot scheme, the localoffice of the Ministry of Agriculture followeda set procedure to select the sari-sari storesthat would be accredited under the scheme.Among the criteria for selection were acces-sibility to participating households, size ofrevolving capital, license to retail grains, andacceptability of the retailer to local commu-nity leadership.
The ultimate distribution mechanismunder the pilot scheme was left entirely tothe private sector. Hence, the sari-sari storeowner was responsible for the procurementof food commodities, as well as handlingand final sales to the participating house-holds. The function of the store owner wasto serve all card-carrying participants within
his designated area of jurisdiction. The gov-ernment's role was to monitor the retailprices of subsidized food in the stores, audittheir accounts for proper redemption of thesubsidies, and reimburse retailers for thedifference between the market price of thesubsidized food commodity and the desig-nated subsidized program price. The gov-ernment depended on the services of theHMTs from the local extension office of theMinistry of Agriculture to carry out thesetasks.
The sari-sari store owners procured sub-sidized rice and edible oils mostly from com-mercial sources, but in some instances theywere procured from the government-ownedNational Food Authority. In general, storeowners preferred to buy from the commer-cial sources where they were normallygiven credit.
The retailers were reimbursed for thesubsidy only after the sales transactionswere made. In return, the scheme providedan incentive to the retailers of 7 percent ofthe gross sales of the subsidized com-modities. Subsidy accounting, which wasdone by the extension officer, was com-puted from the discount cards. These cardswere redeemed every month along with theretailers' sales books. The program usedlocal banks to reimburse the participatingretailers. A special savings deposit account,which was opened in each area, was jointlyheld by the program office and the retailer.An accounting form called a discount reim-bursement voucher was required for thetwice-a-month withdrawal from the banksubsidy account. These vouchers helped theprogram office keep track of the accounts.
Selection of Food Commoditiesfor the Scheme
The 1982 Second Nationwide NutritionSurvey carried out by the Food and NutritionResearch Institute (FNRI) indicated that themost important nutritional problem in thecountry was inadequate calorie intake. The
3 Simeon G. Silverio, The Neighborhood Sari-Sari Store, Institute of Philippine Culture Poverty Series 2 (QuezonCity, Philippines: Ateneo de Manila University, 1975).
14
average Filipino consumed 200-250 calo-ries4 below the recommended dietary allow-ances, and the problem was more severeamong low-income groups.5
In the light of these findings, two calorie-rich foods, rice and edible oil, were selectedfor the program. These foods contributenearly two-thirds of the calorie consumptionof an average Filipino. Rice was selected forthe scheme because it is universally avail-able and the preferred staple food in thecountry, composing 56 percent of the calorieconsumption of the average Filipino.
Apart from these nutritional considera-tions, the use of rice was justified on eco-nomic grounds, rice having high income andprice elasticities for low-income house-holds.6 In addition, the marketing systemfor rice in the Philippines is efficient andreaches even the most widely dispersedpopulations.
Edible oil from coconuts, on the otherhand, is nutritionally important because ofits caloric density. This is particularly criticalin the case of infants and small children,whose digestive systems may be unable toabsorb the necessary energy from diets thatare based on a high-bulk food such as rice.FNRI considered the average annual con-sumption of 2.9 kilograms of vegetable oilper capita inadequate.7
The Size of the SubsidiesThe amount of rice that target house-
holds could obtain under the pilot schemewas smaller than the amount consumed bymost but not all households prior to thescheme: that is, it was inframarginal formost households. Under the scheme, eachhousehold member irrespective of age wasentitled to 5 kilograms of subsidized riceper month. The average rice consumptionof the targeted households before the startof the project was about 10 kilograms percapita per month.
The amount of edible oil, 400 grams permonth, that each household member couldobtain at the discounted price was higherthan the average consumed prior to im-plementation of the scheme. The higher dis-counted quantity of edible oil reflected agovernment policy to expand its domesticconsumption. The National Nutrition Coun-cil saw increasing consumption of edible oilin the diet of Filipinos as a strategy to narrowthe calorie gap. FNRI concluded that thecurrent share of oils and fats in the diet wastoo low, while the potential for its use wasconsiderable because the Philippines isamong the world's leading exporters ofcoconut oil.
The initial price subsidy was 32 percentfor rice and 50 percent for cooking oil.These discounts were initially projected totransfer approximately 200-250 calories perperson per day—a level that was computeddirectly from the food consumption elas-ticities available at the start of the pilot ex-periment. The initial plan was to maintainthese percentage discount rates throughoutthe study period. In the latter half of theexperiment, the last quarter of 1983, do-mestic consumer rice and oil prices in-creased abnormally—the rice price rosefrom 2.90 pesos (P) per kilogram to P 4.25per kilogram, while the price of cooking oildoubled. Thus it was necessary to reducethe percentage of the retail price to be dis-counted in order to keep the cost of thescheme within its original budget. Furtherprice increases were followed by a reductionin the percentage of the discount. The pricediscounts are given in Table 1.
Nutrition EducationThe nutrition education component of
the scheme was included as a complemen-tary intervention to the price subsidy. Theform of nutrition education adopted closelyfollowed that used under the Philippine Nu-
4 All calories referred to in this report are kilocalories.5 Philippines, Food and Nutrition Research Institute, National Science and Technology Authority, Second Nation-wide Nutrition Survey, Philippines 1982 (Manila: NSTA, 1984).6 Howarth Bouis, "Rice Policy in the Philippines" (Ph.D. dissertation, Stanford University, 1982).7 Ibid.
15
Table 1—Price discounts on rice and cooking oil and their share of the retailprice, 1983/84
Period
July-August 1983September 1983 - April 1984May-June 1984July 1984
Price Discount as a Percentof Retail Price
CookingRice Oil
32 5029 4521 2319 22
Price Discount
RiceCooking
Oil
(pesos/person/month)
4.95 2.254.90 2.255.10 2.605.15 2.60
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Note: The exchange rate in 1983/84 was U.S. $1.00 to PI9.
trition Program, which had been in exis-tence for 10 years at the time of the study.The method consisted of face-to-face exten-sion education supplemented by handoutmaterials for information dissemination.
The nutrition education in this schemewas aimed at changing behavior relating tofood consumption, first, to encourage opti-mal use of the additional oil from thescheme; second, to ensure that food con-sumption and nutritional benefits would berealized from the rice and cooking oil sub-sidy; and third, to improve child feedingpractices. The first two items were specificto the project scheme, while the third isstandard in all of the nutrition educationmessages in the Philippine Nutrition Pro-gram. In this respect, the scheme's nutritioneducation messages and the desired be-havioral changes differed from regular nutri-tion education in that nutrition educationwas used as a complement to, rather thanas a substitute for, the subsidies.
Mothers, the primary audience of thenutrition education scheme, regularly at-tended mothers' classes. Outreach washigh: only 15 percent of mothers in thestudy villages did not regularly participate.In households where cases of second- andthird-degree malnutrition were identified,extension workers made monthly homevisits.
The field workers used as educators inthe scheme consisted of the HMTs of theMinistry of Agriculture supported by localparaprofessional volunteers, known asbarangay nutrition scholars (BNS). All fieldworkers had prior experience in nutritioneducation methods. Retraining of theseworkers for the pilot scheme emphasizedthe nutrition education messages. The cen-tral staff of the project monitored thescheme to attain uniformity in methods andmessages across all of the treatment villages,to ensure that classes would be consistent,and to maximize attendance by mothers.
16
THE STUDY AREAS
Criteria for Site SelectionThe pilot areas were selected from eco-
nomically depressed provinces. Although noattempt was made to select a sample of areasrepresentative of the country as a whole,the three provinces chosen were drawnfrom each of the three main geographicalgroupings: Luzon, Visayas, and Mindanao.The three provinces were Abra, located inthe northern Philippines; Antique in thecentral part; and Cotabato in the south (seeFigure 2). Four villages were selected fromeach of the three provinces. The rice andoil subsidies were implemented in two vil-lages, and the other two were used as control(comparison) areas. In each of the provinces,nutrition education was introduced in oneof the two subsidy villages and in one ofthe two control villages. In one of the prov-inces (Antique), two additional villageswere selected to test the impact of a schemeusing edible oil as the only subsidized com-modity, with and without education. Thenumber of the study villages was, therefore,14. This sample size was largely dictated byresearch costs and the ability to maintainan acceptable level of control over experi-mental conditions, and it represents an in-complete experimental design in that notall combinations of treatment and controlswere included. In selecting the villages, rep-resentation of the dominant socioeconomicand ecological environments in the countrywas considered, as well as the nutritionalsituation of the village's children, based onthe anthropometric reports of the NationalNutrition Council. The treatment and com-parison villages within each province werechosen for maximum comparability in eco-nomic base, topography, demography, eth-nic origin, infrastructure services, and othersocioeconomic characteristics.
It should be noted that because an aimof the pilot scheme was also to test its ad-
ministrative feasibility, the pilot areas werechosen to include a variety of infrastructuraland extension service conditions. Hence,areas with both adequate and inadequateextension services and distribution outletswere represented.
As can be seen from the summary ofcharacteristics of the pilot areas in Table 2,the various types of socioeconomic andecological environments covered includeupland, lowland, and coastal areas, andareas whose economic base varies, such asmaize, rice, and coconut farming, fishing,handicrafts, and other indigenous indus-tries.
Characteristics of Study AreasThe first group of study villages is located
in the province of Abra in northern Luzon,situated approximately 450 kilometers northof Manila. The area is primarily in the up-lands, generally mountainous with smallpatches of flatlands. The dominant sourceof livelihood is tobacco, maize, and vege-table farming, which accounted for 44 per-cent of total employment at the time of thestudy. Nearly half of all farmers owned land,but the average farm size of 2 hectares wasless than half the national average. Becauseof the generally poor soil and hilly terrain,farm productivity was considerably lowerthan the national average. Some 15 percentof the population were engaged in livestockproduction, with freshwater (river) fishingas a secondary source of livelihood. About21 percent of the household breadwinnerswere employed outside the province (7 per-cent overseas as contract workers or asseamen), mainly because of the limited eco-nomic opportunities in their home villages.The nonfarm labor force in Abra was ab-sorbed by small home industries, by the in-formal and service sectors as wage earners
17
Figure 2—Map of the Philippines indicating study areas
O
Abra
(30 percent), or as professional and salariedworkers in government and private sectors(10 percent).
The low household income per capitain Abra reflects the poor economic condi-tions in the study villages. Nearly three-fourths of household expenditures went tofood, compared to 57 percent for the rest
of the Philippines. The average family sizein Abra was 6.30 people, and the averageyears of schooling were 7-06 years forfathers and 7.04 for mothers (Table 3). Athird of all children were below 7 years ofage, indicating a young population and ahigh dependency ratio. One of every fivepreschool children was either moderately
18
'S.
0>
5
c3OEcTO
UD 2
3O
'ac
IIO..S
c3O
cro
"§§ II I(J
oU
£
O
i II3 1
U 60SS! E >
IIIl
a .2 «
§ 1 1
if) Jj g ro
O
II•a
S.S
IICO ro
,S2 S <J .12£ E •£ HL, • l l l l l
iHi
•g
> .b -3 5 .ia
i l l l io
if | If
I1a. ^E -0
o «>
Si 8I «5-a
§ 1 00 S S
a.E
13 13
c'>oa.E
u"> cap
o 0
— IS
1!^ e
IQ.
CO
o
fi
.•a
•a
|•a
I
ac
^ 3 Jrr op —
2 =
(Ham
O3T3C
Etd
CO
c'3O
d
n(A
ni
>
Ani
ni-
o^J
19
•o
ISo
S3
c"
>S
og-'£
isc'S.a.
ig |.2»
u o o2 D.S
Iu00I-s
oi -a P P P.2c §S j |
03 •*-!
2 8.a 32.S
e
1 S E
_o
oCO
£i £J
I
o
IIo
u " o- ^
I I # 1QJ r-> H
co5
S u S — ja~ O O 3
CO U ( =
8"C <u
o
20
Table 3—Characteristics of sample households compared with the rest ofthe Philippines
Characteristic
Population of study areas, 1983
Number of households in sample survey"
Household characteristicsHousehold sizeAdult equivalent units (AEU)Years of schooling, fatherYears of schooling, motherPer capita income (P/year)Percent of income spent on foodPercent of expenditures spent on foodCalorie acquisition in AEUsCalorie intake in AEUsCalorie deficit in AEUs
Occupation of main income earnerFarmers with landTenant farmersLandless farm laborersWage earnersFishermenProfessionals/salariedEmployed overseasOthers
Characteristics of preschool children6
Percent of all children below 6 yearsMean weight as percent of standard for agePercent of children second and third
degree malnourishedMean Z-scores for weight for age
Abra
3,628240
6.304,967.067.04
1,68071.1373.901,8181,755
327
23.6412.687.38
30.664.43
10.227.143.81
33.3082.84
20.72-1.60
Antique
5,682360
6.525.077.067.53
1,76860.3769.501,9151,720
383
6.801.903.47
29.4514.6224.5211.077.74
27.3682.71
26.46-1.72
SouthCotabato
5,478240
6.415.096.256.74
1,50571.6071.001,7361,613
410
(percent)27.3017.6921.4120.894.871.005.761.79
32.5883.09
23.19-1.68
Averagefor the
Philippines
6.14b
4.80b
n.a.n.a.n.a.
50.20b
57.00b
l,808c
224C
14.20b
n.a.19.70"29.30b
8.60b
17.00b
1.10b
10.10"
26.09"84.0d
17.2d
n.a.
Sources: Data for the three provinces were collected by the International Food Policy Research Institute, PhilippinesNational Nutrition Council, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey,1983/84," Philippines. Averages are taken from the sources indicated below.
Notes: The z-score is a method used in standardizing the distribution of actual weight of the child relative tothe standard weight for a child of that sex and age. The standards devised by the U.S. National Centerfor Health Statistics (NCHS) were used in the study, n.a. means not available.
* This is the number of households randomly selected to be surveyed at the start of the experiment, 1983.b Philippines, National Economic and Development Authority, Philippine Statistical Yearbook (Manila, NEDA,1984).c Philippines, Food and Nutrition Research Institute, Nationwide Food Consumption Survey, 1985.d Strictly speaking, these results taken from Philippines, Food and Nutrition Research Institute (FNRI), are notcomparable to the data for the three provinces included in the survey because FNRI uses Philippine child growthstandards, whereas the study uses international (NCHS) standards.e These data are from the first survey round.
or severely underweight—a higher figurethan in most parts of the country.
The second group of pilot villages is lo-cated in Antique, a province on the islandof Panay in the central part of the Philip-pines. The villages are situated along a nar-row coast bounded by mountains to the east
and the sea to the west. The area is one ofthe least arable in the country.
Because of the limited land resources,most of the area's population is engaged innonfarm employment. At the time of thesurvey, about 30 percent were wage work-ers in local industries and crafts, 24 percent
21
were salaried and professional workers, and15 percent were fishermen. Only 12 per-cent of the heads of households in the studyareas in Antique were engaged in farming,mainly in marginal maize, rice, and coconutproduction.
Per capita incomes in Antique werehigher than in the two other pilot areas,but still below the national average. The 60percent proportion of income spent on foodwas nearer the national mean than that ofthe two other areas. However, the incidenceof second- and third-degree malnutrition, at26 percent, was highest among the threepilot areas and particularly prevalent amongsmall fishermen.
The third group of pilot villages is lo-cated in the province of South Cotabato, onthe island of Mindanao. Located in a richriver basin, the area is one of the mainmaize-producing provinces in the country.In 1983, two-thirds of its population wereengaged in farming, mostly growing maizeused as animal feed. South Cotabato had ahigh tenancy rate; less than half of the farm-ers owned land, and 21 percent were land-less farm laborers. About 20 percent of theheads of households were wage earners insecondary and service industries.
In the pilot areas of South Cotabato, thepopulation is composed largely of secondgeneration migrants from Luzon and Vis-
ayas. About 15 percent are natives (tribalminorities) who have been assimilated intothe Christian population of immigrants. Percapita incomes in South Cotabato were thelowest among the pilot areas. The share oftotal income spent on food—72 percent—was high compared to the national average.Per capita calorie deficits exceeded 400 perday, and the native minority groups werethe most affected by calorie inadequacy.Malnutrition among preschool children, es-timated at 23 percent, was moderately high.Levels of educational attainment were alsolower in South Cotabato than in Abra orAntique (Table 3).
Population Coverage of thePilot Scheme
The total population of the study villageswas 14,788 at the start of the experimentin May 1983. Out of these, some 1,407households (8,611 people) received the sub-sidy, while the rest were treated as a controlgroup. A total of 840 households with 5,376members, or about one-third of the totalpopulation of the villages, were randomly se-lected for the pilot study. A total of 88 house-holds out of the 840 households droppedout in the second round of interviews andhence were excluded from the study.
22
ASSESSMENT OF IMPLEMENTATION
The purpose of this chapter is to assessthe effectiveness of the design of the schemebased on experience and monitoring. For acontinuous period of 12 months, the pilotsubsidy scheme was observed under actualoperating conditions. This involved themonitoring of key participants of the pro-gram: the beneficiaries, retailers, wholesal-ers, participating banks, extension officers,and local political functionaries. The feasi-bility of the program instruments—rationcards, reimbursement vouchers, and saleslogbooks—were tested under local condi-tions. A full-time study team from the cen-tral office of the National Nutrition Council,Ministry of Agriculture, and the Interna-tional Food Policy Research Institute super-vised the scheme for the entire period ofoperation. In addition, local extension offic-ers were designated to provide periodic on-site monitoring.
The findings exposed a number of imple-mentation problems that were not foreseenin the design of the scheme. The implica-tions of these constraints are discussed herebecause they are critical to a possible larger-scale future program.
The Geographical TargetingMethod
Eligibility for participation in a schemethat targets beneficiaries by geographic areais determined by residency. From an admin-istrative point of view, the requirements forbeneficiary certification in such a schemeare simple. For the pilot scheme, problemsin verification of residence using the villagecensus were few because families were usu-ally known to the extension workers cover-ing those areas.
Because the subsidized ration was basedon household size, some families lied aboutthe size of their households in order to drawlarger quotas. To settle conflicting claims ofhousehold membership, cases were nor-
mally decided by the barangay captain, thelocal political leader. The barangay captainalso decided whether families were bonafide residents of the barangay or precinct,entitled to participate in the program. Atleast 20 percent of the participating house-holds padded reports of household size.There were a few migrant households andtemporary settlers who were not in the cen-sus but were nevertheless included in theinitial program listing.
Although area targeting is easier to ad-minister than other forms of screening, suchas income eligibility or sex and age criteria,several unanticipated effects were observed.A few households invited relatives fromneighboring villages excluded from the sub-sidy into the targeted villages to share thesubsidy benefits with them. This was ob-served only among the very poor householdsand only in the first half of the experiment.It is unlikely that large-scale populationmovements would occur as a result of anarea-targeted subsidy such as this. However,movement would likely be proportional tothe size of the subsidy envisioned. In thepilot study, the level of subsidization com-pared to incomes was quite low; hence thein-migration was limited.
The geographical area targeting schemeassured coverage for all the needy house-holds in the depressed village because thesubsidy was available to everyone withinthe area. The major disadvantage of such amethod, however, is that poor householdswho live outside the target area are excluded,while well-off households and well-off in-dividuals in deficit households in the areaare included.
Possible ProblemsLeakages
The goal of the subsidy scheme was toincrease and sustain the ability of the food-
23
deficit poor households to purchase enoughfood to meet nutritional requirements. Bydefinition, therefore, program benefits thatwere lost in the process of implementationor that accrued to households that did notneed them are considered "leakages." Alter-natively, leakage can be defined as the dif-ference between the value of food or cashtransfer, and the value of the net incrementof the food intake of the food-deficit or needyhouseholds.8
Several overt forms of leakage were ob-served in the pilot scheme. First, becausethe program did not discriminate againstany household within a project village, house-holds without a deficit (mostly higher-incomehouseholds) in the villages were includedin the program. Consequently, leakageswere low if the majority of the householdsin the villages selected were calorie-deficithouseholds. The efficiency of the area methodof targeting is therefore a function of thenational agency's accuracy in identifyingpoverty areas and the degree of geographicalconcentration of malnutrition. The food con-sumption survey conducted just prior to theenforcement of the scheme found that 73percent of the households were deficient,which shows that targeting was relativelyefficient.
In two villages in the south, a few eco-nomically well-off households shared theirrations with poorer neighbors, thus reduc-ing leakages. Some better-off households al-lowed their poor neighbors the use of theirfood discount cards, although discountcards, strictly speaking, were not transfera-ble. These transactions, which were donewith the knowledge of the store owners,involved no compensation.
During the 12-month monitoring periodsome cases of reselling of subsidized foodwere observed, although this was not ram-pant. All reselling cases involved cookingoil, which was attractive for the very poorfamilies because the product commanded agood price in the open market. For instance,a family of six with a monthly allocation of2.4 kilograms of cooking oil stood to gain
about P 15 if it sold all its ration, an amountthat was almost equal to half a day's wages.There was almost no resale of rice observed,mainly because rice is a staple food, and theration for each family was only about halfof what a typical household consumes.
In two villages in the north several en-terprising households used all of their cook-ing oil ration in their small native delicacybusiness, not for home consumption. In anumber of instances some retailers preemp-ted the rations of households that did notconsume all of their quota for the month.These practices were particularly difficultto check because the retailers could forgesignatures in the salesbook records, andmonitoring by the extension officers couldonly be done once a week.
Use of Food Discount Cards
Some problems were encountered inthe use of the food discount ration cards inthe retail stores. Two signatures wereneeded for every sales transaction in thestores: the retail store owner had to signthe ration card and the card holder had tosign the store salesbook. This procedure de-layed the transaction, especially in thestores that served larger populations. Illiter-ate beneficiaries who could not sign thesalesbook created another administrativeproblem. For this group—about 8 percentof the target households—a thumb markwas used to carry out the transaction.Another problem was generated by partici-pants who in the first two months of theexperiment deliberately changed the rationamount stated on the discount cards by tam-pering with the figures on the cards. Thesewere easily detected when compared withthe records of the stores.
Accountability and ControlsThe use of local banks in the handling
of subsidy reimbursements to the retail store
8 The measurement of leakage is explicitly treated in Chapters 6 and 10.
24
owners was quite efficient and would prob-ably be just as feasible in a larger program.The special savings deposit provided a con-venient procedure for the redemption of thesubsidies by the store owners. Because thesize of the subsidy was predetermined fromthe potential number of beneficiaries, theamount transferred every month by the cen-tral program office into the deposit accountswas accurate to within 94 percent of theactual subsidies.
The weekly auditing of the books in thestore presented some difficulties at the startof the project, but once the extension officerand the retailer became familiar with theprocedures, the accounting tasks progressedwithout exceptional problems. Part of thedifficulty in the auditing procedure couldbe traced to the small but frequent pur-chases, which resulted in a large numberof bookkeeping entries.
Administration of Food OutletsThe procurement and selling perfor-
mances of the accredited neighborhood sari-sari stores varied by area. The differencesin efficiency were dependent on a numberof factors, including the size of the popula-tion served, frequency of purchase, amountof revolving capital and supply credit, theamount of credit extended to customers,the location of the store, and the characterof the retailers and their community accep-tance.
The outlets in Antique, which served alarger clientele, experienced more difficultythan those in the other two provinces. Thequeuing and crowding in the stores, espe-cially during the heavy buying days, some-times resulted in bookkeeping errors inthese larger villages. The owners of the re-tail outlets indicated that, given the frequentpurchasing habits of households, the op-timum number that each storekeeper couldefficiently handle was not more than 120households, assuming that the storewas attended to by the store owner and anassistant, who was usually an immediatemember of the family.
In most areas poor households made fre-quent purchases but in small quantities.
Many households, for example, boughtcooking oil three times a day, and the pur-chases were as small as one-half cup (80grams). Store records compiled for the proj-ect indicated that 28 percent of all house-holds made daily purchases. This practiceincreased the transaction time in the storesper unit purchased. Most stores reportedheavier selling late in the afternoons whenthe breadwinners arrived from work orwhen the day's fish catch was sold.
Stores that had more than P 10,000(U.S. $900) of revolving capital were effec-tive in maintaining uninterrupted selling op-erations. On the other hand, smaller storesexperienced intermittent shortages in thesupply of the commodities. Althoughweekly credit was available from commer-cial rice wholesalers, the amount alloweddepended on the capitalization of the store.Hence, smaller stores were not assured ofsufficient credit to meet the demand. Onthe basis of the 12-month store records, itis estimated that in order to support a clientbase of 120 families a retail outlet shouldput up an initial revolving capital of at leastP 10,000 in 1983 pesos.
It was a common practice in neighbor-hood stores to give weekly credit to regularclients. Larger stores offered credit on thesubsidized food, but the smaller stores didnot because of limited revolving capital.Families located on the fringes of the sub-sidized villages experienced difficulty inobtaining subsidized food because of theirdistance from the stores. These familiesusually made one-time bulk purchases, buteven so transport costs increased the totalfood acquisition cost of these beneficiaries.The geographical distribution of target house-holds should therefore play a crucial role inthe choice of outlets for a program.
In some instances neighborhood politicshampered the scheme's delivery mechanism.In Abra, for example, one retailer refusedto sell subsidized rice to constituents notbelonging to her political party. Anotherstore owner refused to sell to people whohad a long-standing feud with her family.Some cultural minority groups were notproperly attended to in the stores in SouthCotabato and were effectively denied their
25
subsidized rations. However, more than halfof the retailers were observed to be quitesensitive to community needs, even open-ing their stores late at night in order to servetheir clients. Based on this experience, it isquite apparent that the choice of the retaileris crucial in program delivery. One of theimportant considerations in selecting an out-let is community acceptance of the retailers.
Subsidy Take-out TrendsRecords compiled at the stores for the
initial month of the experiment indicatethat only 85 percent of the rations madeavailable were claimed by the villagers, butthis take-out rate increased to 95 percentafter the third month. Households that didnot get all of their subsidized food rationswere those that had insufficient cash avail-able during the month, were growing rice,or were too far from the stores to be ableto take advantage of the lower price.
The rate of ration utilization varied bycommodity. Almost all rice rations weretaken but only 80-85 percent of cooking oilrations, probably because the cooking oilration was larger than the amount con-sumed before the scheme, while the riceration was not. Under these conditions,households can be expected to consume allof their rice allocations, but they may decideto buy only as much as or slightly morecooking oil than they used before, if resell-ing is difficult.
Monitoring
The monitoring clearly demonstratedthat the efficiency of scheme implementation
depended largely on adequate understand-ing by program participants of the program'sobjectives, mechanics, and benefits, and onthe retail store owners' understanding ofprogram procedures. Social preparation ac-tivities were conducted by local extensionofficers of the Ministry of Agriculture, as-sisted by the BNS. Monitoring of both theretail stores and the beneficiaries was doneby extension officers as part of their workon the project areas. Prior to implementa-tion, village assemblies were held to explainthe scheme's mechanics and objectives.These assemblies were undertaken with thecooperation of the local political councils inthe project areas.
There were no major difficulties in deliv-ering nutrition education because all of theextension workers had prior experience inthis type of intervention. Eighty-five percentof all mothers in the project areas attendednutrition education classes regularly. Forthe most part, the mothers who did notattend were unable to combine attendancewith wage labor. It is difficult, however, toassess the degree of effectiveness across thevillages because there were obvious differ-ences in the abilities of the extension offi-cers in delivering the nutrition educationmessages. The effects of nutrition educationon the outcome variables (calorie intake andchild nutrition) are discussed in Chapters 8and 9.
In summary, except for a few details,the scheme's design as initially conceivedwas generally feasible from an administra-tive standpoint. Several important lessonswere learned from the monitoring on waysto enhance its viability if the scheme wereapplied on a larger program level.
26
ANALYTICAL METHODOLOGY
The analysis of the impact of the pilotscheme on household food expenditures,acquisition, and consumption, and on foodconsumption and the nutritional status ofpreschoolers is based on a conceptual frame-work shown in a simplified version in Figure 3.
The price discount element of the schemeimplies an increase in the real purchasingpower or income of the recipient households.This increase can in turn be expected toincrease household food expenditures, ac-quisition, and consumption. It may also re-sult in increases in food consumption bythe household members most likely to becalorie-deficient—preschoolers—and thusimprove their nutritional status (Figure 3).
If the amount of subsidized rice or oil ex-ceeds the amount that would be purchasedby the households without the subsidies,the subsidies will reduce the price house-holds pay for the last unit (at the margin).This could be expected to increase the con-sumption of the subsidized foods. Theseprice-induced increases would not be ex-pected, however, if the amount subsidizedwere less than the amount the householdwould purchase without the subsidies be-cause the price at the margin would not beaffected. The price subsidies may also affectfood expenditures, consumption, and nutri-tional status directly through changes inhousehold behavior.
Figure 3—Simplified version of the conceptual model underlying the analyses
Pilot Scheme
HouseholdIncomes
> Household food expenditures
Nutrition Education
Household food consumption •<-
Food consumption by preschoolersand pregnant and lactating women
-> Nutritional status of preschoolers ••-
27
The nutrition education component ofthe scheme may also affect food expendi-tures, consumption, and nutritional status bychanging household behavior regarding foodpurchases, dietary patterns, intrahouseholdfood distribution, and other aspects. The ef-fects of each of the two components—pricesubsidies and nutrition education—on eachof the five indicators shown in Figure 3—household food expenditures, acquisition,and consumption, food consumption by in-dividual preschoolers, and their nutritionalstatus—are estimated by means of compara-tive static and multivariate analyses.
Comparative AnalysisThe experimental design used in this
study is illustrated in Table 4. Data werecollected from four rounds of surveys in the14 villages over a 17-month period. Abouttwo months after the first survey round wasinitiated, subsidized rice and oil were madeavailable to survey villages in each of thethree provinces. In addition, subsidized oilalone was made available to two villages inone of the provinces. The two other villagesfrom each province that were included inthe study were not given access to subsi-dized food; they served as controls. In eachprovince, a nutrition education programwas introduced in one of the two villagesthat received subsidies and one of the twothat did not.
The second survey round was executedabout 2 months after the introduction ofthe subsidy scheme, the third 12 monthsafter the first, and the fourth and last roundwas undertaken about 2 months after thescheme was terminated, that is, 12 monthsafter the second survey. The spacing of 12months between surveys was done to avoidconfounding effects of seasonal variations.
The control villages were selected to beas similar as possible to the "treatment"villages—those receiving subsidies. Thus,effects of exogenous variables that could notbe controlled or accounted for in direct com-parison are assumed to have the same effecton control and treatment villages and cantherefore be separated from the effects ofthe scheme.
The effects of the subsidies and nutritioneducation are estimated simply by compar-ing the appropriate treatment and controlvillages after adjusting for differences priorto the introduction of the scheme (the treat-ment). Analysis of variance is used to testfor significant differences. The percentchange in each of the five indicators of inter-est due to subsidies and/or nutrition educa-tion is estimated:
(1)
where
Aj = percent change in the indicatorfor month i relative to the base-line value of the indicator dueto the treatment, for example,the percent change in house-hold food consumption due tonutrition education;
Tj and T{ = average value of the indicatorforround 1 (baseline) and month ifor treatment households; and
C, andCj = average value of the indicatorin round 1 (baseline) and monthi for control households.
The results of the comparative staticanalyses are presented in Chapter 7.
Multivariate AnalysisResults from comparative static analyses
are suspect because the effects of other fac-tors on the groups are not fully accountedfor. Multivariate analyses suffer considerablyless from this problem because the effectsof many factors may be isolated from theeffects of the pilot scheme. Therefore, inaddition to comparative static analyses, mul-tivariate analyses are used to estimate theimpact of the scheme. These analyses arebased on the conceptual model presentedin Figure 4, which shows the variables hy-pothesized to influence the indicators andthus included in the multivariate analyses.
28
Table 4—Experimental design of the pilot food price subsidy study
CommoditiesSubsidized
Rice and cooking oil
No subsidy
Oil only6
NutritionEducation
Yes
No
Yes
No
YesNo
Province
AbraAntiqueSouth Cotabato
AbraAntiqueSouth Cotabato
AbraAntiqueSouth Cotabato
AbraAntiqueSouth Cotabato
AntiqueAntique
l a
XXX
X
X
XXXX
XXX
X
X
Survey Rounds2b
X
X
XXXX
XXX
XXX
X
X
3C
XXX
XXX
XXX
XXX
X
X
4d
XXX
XXX
XXX
XXX
X
X
a This was the baseline survey, 2 months prior to introduction of the scheme.b This survey took place 2 months after introduction of the scheme.c This survey took place 10 months after introduction of the scheme.d This survey took place 2 months after the scheme was terminated.e In two villages of one province, only oil was subsidized at the request of the Minister of Agriculture.
Household Calorie ConsumptionFunctions
Four of the five indicators—householdfood expenditures and acquisition, and foodconsumption and nutritional status of pre-schoolers—are regressed on the variablesshown in Figure 4 to estimate the effect ofthe price subsidies and nutrition educationon each of the indicators and to isolate theeffects of other variables hypothesized toinfluence these indicators. The impact ofthe price subsidies is measured partlythrough the increase in household incomesand the decrease in rice and oil prices andpartly through a direct effect hypothesizedto come about primarily because householdstreat the real income increase from foodsubsidies differently from other income in-creases—that is, the marginal propensity to
spend on food differs between subsidy in-come and other real income.9
The general estimating model used toestimate the impact of the scheme on house-hold food expenditures and acquisition is
= f(Y,S, PR,P0,NE,Z), (2)
where C is total food expenditure (or calorieacquisition), Y is real household income, Sis the subsidy term, PR and Po are the realprice of rice and cooking oil, respectively,NE is nutrition education, and Z is a set ofother variables hypothesized to influencethe dependent variable. Here the dependentvariable is a composite of expenditures onall foods or an aggregation of total caloriesfrom all of the foods acquired by the house-hold.
9 Evidence of such differential treatment of subsidy income has been found for the U.S. food stamp program (BenSenauer and Nathan Young, "The Impact of Food Stamps on Food Expenditures: Rejection of the TraditionalModel," American Journal ofAgricultural Economics 68 [February 1986]: 37-43). None has been found, however,for the Sri Lankan food stamp program (Neville Edirisinghe, The Food Stamp Scheme in Sri Lanka: Costs, Benefits,and Options for Modification, Research Report 58 [Washington, D.C.: International Food Policy Research Institute,1987]).
29
Figure 4—Conceptual model underlying the multivariate analyses
Pilot Scheme
Household size
Education of father
Education of mother
Nutrition education
Food assistance
Farm ownership
Garden ownership
Women's incomes
Province
Household incomes
Food prices
Price Subsidy Nutrition Education
HouseholdIncomes
FoodPrices
1 IHousehold food expenditures
IHousehold food acquisition
IHousehold food consumption
Food consumption by preschoolers .
INutritional status of preschoolers
Age of child
Sex of child
Previously breastfed
Birth order
Mother's time spent on
child care
Diarrhea
Fever
Drinking water supply
Household composition
The household-level estimating modelis defined:
C/AEU = b0 b,Ln(Y/AEU)
b5EducHj
+ b4EducSj
b6NutredCj
+ b7NutrelVj + bgFoodasst,
+ bgOwnfarnij + bjgOwngardj
b1IPwomenYi + b12L2i b,3L3i
b14Price, b15Poil
where
C/AEU = food consumption of house-hold i defined either as totalfood expenditure per adultequivalent unit (AEU), or as
total calorie acquisition perAEU;
Y/AEU = total income per AEU of house-hold i;
Sj = subsidy term defined either asa percent of total incomes, oras a dummy, 1 if householdsreceived the subsidy and 0otherwise;
HHsize, = household size;
. u . ; (3] EducSj = educational level of the wife;
EducHj = educational level of the hus-band;
Nutredc, = a dummy defined as 1 if thehousehold received nutritioneducation from governmentprograms, 0 otherwise;
30
NutrehA = a dummy defined as 1 if thehousehold participated in thenutrition education classes un-der the pilot subsidy scheme,0 otherwise;
Foodasstj = a dummy defined as 1 if thehousehold participated in foodassistance programs, 0 other-wise;
Ownfarnij = a dummy defined as 1 if thehousehold owned a farm, 0otherwise;
Owngardj = a dummy defined as 1 if thehousehold owned a garden, 0otherwise;
PwomenYj = proportion of total incomescontributed by women;
L2i = location dummy for Antique;
L3i = location dummy for SouthCotabato;
Pricej = unit price paid for rice by thehousehold; and
Poilj = unit price paid for oil by thehousehold.
The two alternative specifications of thesubsidy are used to test the robustness ofthe estimates of the scheme's impact. Eachof these implies different assumptions andlevels of constraint imposed on the con-sumption function. The first alternative usesa proportion term (the subsidy as a percentof total incomes from all sources includingthe subsidy itself). If the coefficient of thissubsidy term is not significantly differentfrom zero, the marginal propensity to con-sume (MPC)—to acquire food—does notchange with the changing proportion of totalincomes coming from subsidies. A positiveand significant coefficient would indicate
that income transfers embodied in thescheme have a higher MPC than other house-hold income, and vice versa.
The specification of a dummy for subsidyrecipients is interpreted as a parallel shiftin the relationship between consumptionand income. Therefore, a positive and sig-nificant coefficient of the dummy subsidyvariable would indicate an effect above theincome effect. However, this subsidy vari-able predicts an impact that does not varyacross income groups. This model is henceless intuitive, although empirically predic-tive of the subsidy effects at the mean levelof income.
The use of an AEU scale in the foodconsumption and income variables explicitlyaccounts for the differences in householdsize and composition. The AEU approach isa refinement to the consumption per capitaand income per capita approach, as it ex-plicitly accounts for the differences of agecomposition of households.10 For this study,the AEU scale used to translate family mem-bers of different ages into equivalent personunits is derived from the Philippine recom-mended dietary allowances (RDAs) for cal-ories for each sex-age group.1' Thus, formu-lating incomes and consumption in AEUsallows flexibility in economies of scale andsize and composition of households.
The income variable is specified in nat-ural logarithm to allow for decreasing MPCfor food with increasing incomes. Becauseof inherent problems in correctly measuringincome, particularly in semisubsistencehouseholds, total expenditures were usedas income proxy throughout the regressions.Because, as discussed in Chapter 8, house-hold incomes are not strictly exogenous, thevalidity of the results is tested using wagerates instead of incomes.
The impact of the other component ofthe pilot scheme, the nutrition educationintervention, is tested with a dummy vari-able equal to 1 if the household received
10 S. J. Prais and H. S. Houthakker, The Analysis of Family Budgets (New York: Cambridge University Press,1971); and Angus Deaton and John Muellbauer, Economics and Consumer Behavior {Hew York: CambridgeUniversity Press, 1980).11 Food and Nutrition Research Institute, and National Nutrition Council, Recommended Dietary Allowances forFilipinos (Manila: FNRI, 1976).
31
the nutrition education and 0 if not. Varioussocioeconomic and locational variables areincluded as explanatory variables, specifi-cally to isolate the effects of incomes andsubsidies from other socioeconomic, demo-graphic, and locational factors.
The prices of rice and cooking oil usedas independent variables in the regressionsare data reported from the survey. In usingthe unit price for rice as an explanatoryvariable, the market price paid is used forhouseholds where the rice subsidy is infra-marginal; in the extramarginal householdsthe price used is the price that the house-holds paid in the subsidy stores. Price,expenditure, and income variables are de-flated by price indexes for the particularmonth when the survey was carried out,and for the particular region where thesamples were located. The use of real valuesin the regressions is designed to separatethe effect of the increase in real incomethrough the subsidy scheme from an increasein expenditures through inflation. The firstsurvey is used as the baseline (index = 100)in deflating the nominal values.12 The re-sults from the regressions are shown in theAppendix, Table 34, and the main findingsare presented and discussed in Chapter 8.
Household Consumption Functionsfor Rice, Oil, Fish, and Maize
In order to better understand how thescheme influenced the consumption of theprincipal food commodities, the calorie con-sumption function is complemented withconsumption functions for each of the fourprincipal food commodities—rice, oil, fish,and maize. The commodity functions arespecified using the calorie consumptionfunctions above, except that the prices ofmaize and fish are added. The coefficientsestimated by these functions are shown inthe Appendix, Table 35, and discussed inChapter 8.
Calorie Consumption Functionsfor Preschoolers
To estimate the impact of the schemeon the consumption of calories by individualpreschoolers, this model is used:
= b0 b3CBreastf(
b4PBreastfj + b5Birth0j + b6EducHj
+ b7EducSi + b
+ b10Perchd6i j ^
+ b^Water, + b,3Nutrelv,
+ bpHHsize,
b14NutredC; + b15Foodasst; + b16L2i
b17L3i
+ b,9Ln(Y/AEU) + b20Sj + b^Price
+ b22Poili + b-^Fever,. + (4)
where
IC, = daily calorie consumption bypreschooler j ,
= log of the age of preschooler j ,
= sex of preschooler j (1 if male),
CBreastfj = zero-one dummy (1 if the pre-schooler is currently breast-fed),
PBreastfj = zero-one dummy (1 if the childwas breast-fed in the past),
BirthOj = birth order of preschooler j ,
ChTime- = time spent in child care by themother of preschooler j ,
Perchd6j = proportion of children with ageless than six years in household i,
Diarrhea, = zero-one dummy (1 if the pre-schooler had diarrhea in the pastweek),
Wate^ = zero-one dummy (1 if the house-hold where preschooler belongs
12 The price indexes were taken from Philippines, National Economic and Development Authority, PhilippineStatistical Yearbook (Manila: NEDA, 1985).
32
has sanitary potable water facil-ity), and
j = zero-one dummy (1 if the pre-schooler had fever during thepast week).
Other terms have already been defined.The dependent variable in the equation
is the calorie content of foods consumed byindividual children between the ages of13 and 83 months. Children less than 13months of age are excluded from the analy-sis because breast-feeding of that age groupis widespread and the calorie content ofbreast milk consumed was not estimated.
The pilot scheme's net effect on individ-ual children is hypothesized to come fromincreases in overall incomes, nutrition edu-cation, and from the effects of the subsidyscheme itself. A semilog functional form ischosen to express the relationship betweenincome and individual child consumption.A variant of this model is also tested usinghousehold calorie acquisition in place ofhousehold incomes. The inclusion of indi-vidual child variables such as age, sex, mor-bidity, birth order, and breast-feeding serveto control for child-specific variations andbiological factors, whereas other householdvariables, such as the parent's education,household size and composition, women'sincomes, child care time, rice and oil prices,and geographical location, are included inorder to account for household differencesthat could influence individual calorie con-sumption of children. The demand equationsare estimated by ordinary least squares forthe 589 sample preschool children aged 13to 83 months. Results of the regressions aregiven in the Appendix, Table 36, and thekey findings are presented and discussed inChapter 8.
Anthropometric Functions forPreschoolers
The effects of the scheme on the growthof preschoolers are estimated on the basisof estimating equations of the following gen-eral form:
W] = f(Y1,S1,P r l0e,Pol l,C1,NE, lZ1J)> (5)
where
Wj = an anthropometric indicator of thenutritional status of preschooler j ,
Y, = total income per capita of the house-hold to which the preschooler be-longs (i),
S, = an indicator for the food subsidy,
C: = calorie consumption by the house-hold to which the preschooler be-longs (used only as an alternative to
NEj = an indicator for whether the house-hold received nutrition educationfrom the scheme, and
Zj • = a set of other variables hypothesizedto influence W(. All Z-variables areidentical to those specified in the cal-orie consumption function for pre-schoolers.
The dependent variable is specifiedin five alternative ways: (1) weight of thepreschooler, (2) weight as a percent of thestandard weight for a particular age, (3) astandardized Z-score for weight-for-age, (4)a standardized Z-score of weight-for-height,and (5) the height as a percent of the stan-dard height for a particular age. Z-scores area method used to standardize a distribution.International standards are used in estimat-ing the Z-scores because they facilitate com-parison with other studies. The Z-scores forweight variables are defined as:
ZWA = (W - W*)/s.d. (6)
where
W = actual weight of the child,
W* = standard weight for a child of a par-ticular sex and age, that is, the me-dian of the reference population, and
s.d. = standard deviation from the standardfor a particular sex and age.
Weight-related measures are expectedto be more appropriate than height-related
33
measures for analyses of effects on childgrowth in the short run. Thus, for the periodof time involved in this study, the schemeis less likely to have an effect on height.
Cost-Effectiveness AnalysisFor estimating the cost-effectiveness of
the scheme, three sources of cost are identi-fied: the cost of administration, the cost ofthe subsidy or price discount (the price dis-count per unit of rice and cooking oil timesthe quantity for which the discount wasgiven), and the cost of the subsidized foodpaid by the participating households. Thefirst two are borne by the public sector andreferred to as "fiscal costs," whereas thelast refers to the amount consumers pay forthe subsidized commodities. The last twoadd up to the value of the subsidized com-modities at nonsubsidized prices. The costof the subsidy equals the subsidy benefitsto the consumer because all costs of thescheme other than the price subsidy timesthe quantity subsidized are included undercost of administration.13 The transactioncosts incurred by participants are assumedto be identical to those they would incurwith the purchase of nonsubsidized rice andoil. The transaction costs associated withparticipation in the nutrition education pro-gram, including the value of the time spentby the participants, are ignored in the costestimations.
Effectiveness is measured, first, by theamounts of real income and calories trans-ferred to the participating households in-cluding those households with either severecalorie deficits or children weighing lessthan 75 percent of standard, and, second,by the change in the nutritional status ofchildren.
Table 5 provides an overview of thevarious measures used to estimate cost-effectiveness. These measures represent dif-ferent indicators of benefits, such as changesin household incomes and food acquisition,food consumption by preschoolers, and the
weight of preschoolers. They also representfive different target groups: all householdsin the target villages, only households withcalorie consumption below 80 percent ofrequirements, only households with mal-nourished preschoolers, only preschoolers,and only malnourished preschoolers.
Thus, if the goal of the scheme is toreduce malnutrition among preschoolers,the measures reflect different degrees ofleakage. Three sources of leakages are con-sidered. The first source is households parti-cipating in the scheme that are not deficientin calories or that do not have preschoolers.The second is substitution between food andnonfood items and among foods resultingin a net increase in household food expen-ditures and calorie consumption that issmaller than the equivalent real incometransfer. The third is when some of the netincrease in household calorie consumptionis consumed by household members otherthan preschoolers.
The fiscal cost of transferring one U.S.dollar of real income to all participatinghouseholds is a crude indicator of cost-effec-tiveness from a nutrition point of view be-cause all three sources of leakages are pre-sent. However, it may be an appropriateindicator if the goal is to transfer purchasingpower at a low fiscal cost. One source ofleakage is removed if only transfers tocalorie-deficient households or householdswith malnourished preschoolers are consid-ered (Table 5).
The second source of leakage—reductionin food acquisition from sources other thanthe subsidy—may be deleted from the cost-effectiveness measure by considering thecost of a net increase in household calorieconsumption. The third source of leakage isdeleted in the indicators that only deal withthe effects on malnourished preschoolers.
In specifying the estimation proceduresfor each of the indicators, an estimate ismade of the fiscal cost of transferring U.S.$ 1.00 of real income, first, to all participat-ing households (H,),
13 Shlomo Reutlinger and Judith KatonaApte, The Nutritional Impact of Food Aid: Criteria for the Selection ofCost-Effective Foods, Discussion Paper WB-ARU Report No. 12 (Washington, D.C.: World Bank, 1984).
34
Table 5—Overview of the cost-effectiveness measures used in this study
Fiscal Cost
Fiscal cost of transfer-ring U.S. $1.00
Fiscal cost of a net in-crease in food acqui-sition of 100 caloriesper AEU per day
Fiscal cost of eliminat-ing the calorie gap
Fiscal cost of a netincrease in food con-sumption of 100 cal-ories per preschoolerper day
Fiscal cost of increasingthe average weightof preschoolers by1 kilogram
All Par-ticipating
Households(H,)
X
X
X
HouseholdsConsuming
Less than80 Percent
ofRDA(H2)
X
X
X
HouseholdsWith
MalnourishedPreschoolers3
(H3)
X
X
X
AllPreschoolChildren
(Cr)
X
X
PreschoolChildren
ConsumingLess Than80 Percent
ofRDA(C2)
X
X
Mal-nourished
Preschoolers'1
(C3)
X
X
' Malnourished preschoolers are those with weight-forage below 75 percent of standard.
(CA + CS) /CS (7)
where CA is the cost of administration ofthe scheme, and C s is the cost of the subsidy(price discount x total quantity of subsidizedcommodities). This is equal to the value ofthe subsidy received by all participatinghouseholds.
Second, the cost of transferring U.S.$1.00 to households with a calorie deficitin excess of 20 percent of requirements (H2)is estimated:
C s ) /CS 8 0 , (8)
where CS80 is the value of the subsidy re-ceived by households with a calorie deficitin excess of 20 percent of requirements.
Finally, the cost of transferring U.S.$1.00 to households with at least one pre-schooler whose weight-for-age is below 75percent of the standard (H3) is estimated:
(C CS)/C SMC (9)
where CSMC is the value of the subsidy re-ceived by households with at least one pre-schooler whose weight-for-age is below 75percent of standard.
The next step is to estimate the fiscalcost of increasing net household calorieacquisition by 100 calories per AEU per dayamong households in groups H,, H 2 , andH 3 . The cost is given by CA + C s . The im-pact on calorie acquisition is estimated fromthe calorie acquisition function (3):
[a(C/AEU)/ln(Y/AEU)]VS
+ [d(C/AEU)/S]S + b 7
+ [S(C/AEU)/Poi,]APoil, (10)
where
VS = value of the subsidy, and
APoil = change in PoU due to subsidy, andb7 = coefficient for nutrition education
from the pilot scheme.
35
Thus, the fiscal cost of increasing the nethousehold calorie acquisition by 100 caloriesper AEU per day is given by (CA + Cs) 100,divided by equation (10) above.
Now, the fiscal cost of increasing thenet calorie consumption of preschoolers by100 calories per AEU per day among groupsof C, (all preschoolers), C2 (preschoolersconsuming less than 80 percent RDA), andC3 (malnourished preschoolers) is estimated.The cost is as shown above. The impact isestimated from the calorie consumptionfunction for preschoolers, equation (4), as
[d(IC)/ln(Y/AEU)]VS,3
+ [3(IC)/Poil]APoil. (11)
The equation is evaluated at the meanfor each of the three groups.
Finally, the fiscal cost of increasing theweight of preschoolers is shown above. Theimpact is estimated from equation (5),
[a(w)/ln(Y/AEU)JVS + [3(w)/SlS
+ [d(w)/3(NE)jNE + [3(w)/Poil]APoll. (12)
Data Collected and SurveyCharacteristics
The analytical framework discussed inthis chapter was the basis for choosing thedata that were collected in the study.
A stratified random survey of house-holds in the pilot areas was conducted as acomponent of the project. A random sampleof households was drawn from each of theeight subsidized and six nonsubsidized vil-lages for a total sample of 840 households.These households were visited four timesduring the period between May 1983 andNovember 1984. Between the first andfourth survey round, 88 households droppedout of the survey. As stated earlier, the firstsurvey was conducted two months prior tothe introduction of the food subsidy. Thissurvey provided the baseline informationfor the scheme. Within two months fromthe start of the scheme, the second surveywas conducted using the same questionnaire
as in the first survey. Exactly 12 monthsafter the first survey, a third survey roundwas conducted. Thus, the first and thirdsurveys were conducted in similar monthsof the first and second year in order to elimi-nate possible seasonal effects on income,production, consumption, and nutrition.
The subsidy scheme was in effect for aperiod of 12 months—July 1983 to June1984. Two months after the withdrawal ofthe subsidies, the fourth survey was con-ducted. One of the purposes of the fourthsurvey was to understand how householdconsumption behavior was affected by thewithdrawal of the subsidy.
Data were collected for all variablesshown in Figure 4. Several modules wereincluded. The first was a socioeconomic sur-vey, including income of each householdmember; household expenditures on foodand nonfood items; size, age, sex, occupa-tion, and education of all household mem-bers; time allocated separately by the wifeand husband to various activities; a seriesof questions on household decisionmaking,such as who within the household decideson expenditure patterns and purchases ofspecific items; environmental sanitation,including toilet and water facilities; andhousehold participation in local governmentprograms.
The second module was a consumptionand food acquisition survey including house-hold food consumption; household food ac-quired from purchases, own production, gifts,and wages; expenditures and prices of food;and only for a subsample of households, foodintake by each household member.
Two dietary survey methods were usedin the same questionnaire: the 24-hour foodweighing method (here called food con-sumption) and the flexible period recallmethod (here referred to as food acquisi-tion). The 24-hour food-weighing methodwas used in both the household and individ-ual food-weighing surveys. Under thismethod, all items used in food preparationwere weighed and so were leftovers. Snacksor foods eaten between meals were alsotaken into consideration. For the subsamplewhere individual food intake was weighed,interviewers were present at the meal table.
36
Care was taken not to disrupt eating be-havior during meals. Data collection wascarried out by trained nutritionists andhome economists, all females with at leasta bachelor's degree, whose average workexperience was six years. The interviewerscame from the same provinces where thesurveys were conducted in order to over-come language barriers. All interviewersunderwent a week of extensive theoreticaland field training on the procedures in orderto minimize disruptions at mealtimes. Forthe other households where individual in-takes were not taken, foods were weighedprior to cooking, just before the morningmeals, and after dinner, when leftoverswere weighed.
In the flexible period food recall method,the respondent mothers were asked to recallfood expenditures (quantity, unit value, andmonetary values) for the immediate past—aday, a week, or a month—which includedall foods purchased from the market, foodsfrom their own production either from homegardens or farms, and foods received as giftsor wages. The recall period was made onthe basis of the normal frequency of pur-chase stated by respondent households.
The third module was the anthropomet-ric data collected monthly throughout thestudy period to determine the health statusof preschool children, including weight,height, age, and morbidity of each child lessthan seven years of age.
37
DESCRIPTIVE ANALYSIS OF INCOME,CONSUMPTION, AND NUTRITIONAL PATTERNS
In this chapter the results of the surveysfor the five indicators are disaggregated byprovince, survey round, occupational group,and income quartile and analyzed to identifydifferences among population groups thatmay be useful for targeting of future subsidyprograms or other interventions. A secondarybut important objective is to gain knowledgeabout the survey households and individualsover and above what can be learned fromthe analyses based on the total sample.
Household Income SourcesThe overall average monthly income of
the sample households during the surveyperiod was P 140 per capita (equivalent toa household income of P 910 per month)as compared to total expenditures of P 117(Table 6). Some households in the poorestquartile spent slightly more than theyearned. This occurred primarily amonglandless farm workers and tenant farmersother than those producing rice and maize,the two groups that were clearly the poorestof the sample. For the sample as a whole,incomes were considerably higher than ex-penditures during 1983, but these apparentsavings did not continue in 1984, whichprobably reflects the rapid increase in pricesduring late 1983 and early 1984. About one-half of all incomes came from salaries andwages and about one-third originated inagriculture and fishing.
Food Acquisition andConsumption Patterns
Household consumption expendituresare used in this study as a proxy for incomebecause consumption expenditure is hypoth-
esized to be a better indicator of permanentincome and hence a more important deter-minant of consumption behavior.14 Con-sumption expenditures are also easier tomeasure, and nonsampling errors for thisvariable are likely to be much lower thanthose for incomes, which have been chron-ically understated in many household surveys.
The average share spent on major con-sumption items is shown in Table 7. Onthe average, 71 percent of total householdexpenditures went to food. This figure wasmuch higher than the national average of57 percent, which indicates that conditionsin the study areas were relatively poorer.
The share spent on food fell as incomesrose in concurrence with Engelian relation-ships. While the households in the lowestincome quartile allocated 79 percent of theirtotal expenditures to food, those in the high-est quartile allocated 63 percent. The shareof income spent on food increased overtime, which may be explained by the higherrate of inflation for rice than for other majorbudget items between 1983 and 1984. Inearly 1984, the ceiling price of rice wasallowed to increase by about 45 percentover the previous year. Rice accounted forabout 39 percent of the total food budgetand nearly 26 percent of total householdexpenditures. Thus, rice prices were impor-tant in overall purchasing power.
Table 7 summarizes the budget sharesof major expenditure items by province, sea-son (survey round), occupation, and incomegroup. The shares spent on nonfood itemsincreased as incomes rose, particularly cloth-ing and footwear, education, medical ser-vices, and housing. Shares spent on foodvaried widely between farm and nonfarmhouseholds. These variations may reflect
14 Milton Friedman, A Theory of the Consumption Function (Princeton: Princeton University Press, 1957).
38
Table 6—Average incomes and income sources by province, survey round,occupation, and income quartile
Category
Entire sampleProvince
AbraAntiqueSouth Cotabato
Survey roundMay 1983November 1983May 1984November 1984
Occupational categoriesRice farmer landownerMaize farmer landownerFarmer, other cropsTenant rice farmerTenant maize farmerTenant, other cropsLandless farm laborerWage earnerProfessional/salariedFishermen, boat ownersHired fishermenEmployed overseasOccupation unclassified
Income groupFirst quartile (lowest)Second quartileThird quartileFourth quartile
Number ofHousehold
Visits"
3,167
9081,382
877
840799776752
200222
7787
13650
275817116384
39256508
791791791791
IncomePer CapitaPer Month"
TotalExpenditurePer CapitaPer Month
(pesos)
140
140149125
122129151159
1481351361121219098
133300110142191152
78100130252
117
131106120
9675
149153
13112812711311810310410721810594
132123
8296
118173
Salaries/Wages
Income SourceFarm/Live-
stock/Fishing
Busi-ness
Rent/Interest/Pensions/
Gifts
(percent of total income)
51.3
43.855.750.9
50.447.054.553.5
21.812.623.019.921.017.169.470.483.613.280.883.533.7
41.140.349.959.8
34.4
37.930.339.4
36.439.030.331.9
66.474.865.370.770.469.018.219.411.477.614.010.428.1
41.646.037.925.6
6.5
4.08.64.6
6.36.16.86.6
4.14.61.30.90.51.83.14.61.74.21.73.1
20.6
4.86.66.27.1
7.8
14.35.45.1
6.97.98.48.0
7.78.0
10.48.58.1
12.19.35.63.35.03.53.0
17.6
12.57.16.07.5
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
a "Household visits" is defined as the number of sample households times the number of interviews performedin each household.b Income includes the value of own production consumed.
differences in income as well as the semisub-sistence nature of farming. The professionalsand salaried workers spent 60 percent onfood compared to the poorer wage earners'81 percent. Surprisingly, food producers,such as rice farmers and fishermen, tendedto spend more on food than those who wereemployed. It is important to note, however,that the present definition of food expen-ditures covers not only cash food purchasesfrom the market but also the imputed valuesof food consumed from the households' ownfood production and food received as giftsand wages.
Food Consumptionby Source of Food
For the entire sample, nearly 80 percentof total food expenditures was accountedfor by food purchases (including food con-sumed away from home), 15 percent byown production, 5 percent by gifts received,and a small proportion by foods received aswages. The province of South Cotabatospent more on food purchases and less onfood from own production. Nearly one-fifthof Abra's food consumption came from pro-duction of their own rice and maize.
39
isG.
sO"oa-Hr-oNooo co o r-od c\i —co d —- d «
—• co d m <N 06 06 in i/i i t cd
—; cs| c> «J IN. ^ i uii —; cs[ tn tv, •T}| f-..in co r- n ro O; »*
Ic
Ia.a.
1
•B§
•o m o co — oo t^ co oo co -*t c~ co in ^- cs coO N « t m n N N O O O O O d ) N
40
Between 1983 and 1984 the absolutepercentage of subsistence consumption—food grown for the household's own con-sumption—increased, partly indicating theeffects of the increase in food prices overthe period.
The food acquisition pattern by sourcesof food varied by occupation. The shares offood consumption from own production, asexpected, were highest among the food pro-ducers, such as the rice farmers and fisher-men. However, even among rice farmerswho owned land, about two-thirds of thetotal food consumed was accounted for bypurchases from the market, while 19 percentcame from own production and 4 percentwas received as gifts. Similar consumptionpatterns among rural groups were observedby Trairatvorakul in Thailand.15
Among the farming groups, the surveyindicates a larger proportion of subsistenceconsumption among rice farmers than maizefarmers. This is because most of the peoplein the sample were rice eaters, and maizeis used more as livestock feed than for humanconsumption in most parts of the Philippines.The landless farm laborers depended onmarket purchases for 61 percent of theirhousehold food expenditures.
It is estimated that 21 percent of thefood consumed by the families of boat-owningfishermen came from their own production.Hired fishermen depended on the marketfor a higher proportion of their householdfood supply than did boat-owning fishermen.Food received as wages composed 8 percentof the total food consumed by hired fisher-men, the second highest percentage of alloccupational categories. This reflects thepractice of paying hired workers from theday's catch of fish.
Although the samples were all drawnfrom rural areas, about 48 percent of thehouseholds had nonfarming, nonfishing oc-cupations. The main sources of livelihoodfor the major breadwinners in many house-holds were urban-based or urban-related,despite their rural residences. For thesegroups, 85 percent of all food consumed
came from purchases. About 10 percent oftheir food supply came from their own pro-duction, mostly from backyard fruit and veg-etable gardens.
Allocation of the Food Budgetby Food Group
Table 8 gives the breakdown of the av-erage household food budget allocated toeach food group. There were 124 foodgroups coded in the survey, and these wereaggregated into the 12 main food groupsreported in Table 8. The last one is theresidual from the first 11 and is designatedas "other foods."
For the entire sample, rice accountedfor nearly 40 percent of total food expendi-tures, fish for 18.6 percent, and vegetablesand fruits for 11.3 percent.
A comparison of the patterns of foodexpenditure across provinces indicates thatminor variations occurred in rice expendi-tures, but major differences were found innonstaple foods. Although South Cotabatois a maize-producing province, its popula-tion is basically rice-eating. This explainsits lower food expenditures for maize com-pared to Abra, where part of the populationeats maize during certain periods of theyear. Abra spent about 2.8 percent of itsfood expenses on maize, the highest amongthe three provinces. As expected, the sharespent for fish was highest in the coastalprovince of Antique—23 percent comparedto only 13 percent in Abra. Abra's fish sup-ply came mostly from freshwater sources,mainly out of the Abra River. Abra spentmore on poultry, fruits, and vegetables thanthe other two provinces. The other foodgroups did not vary much by province.
The major differences in the pattern offood expenditures observed were acrossoccupational categories. Certain food itemswere heavily consumed by the producersthemselves. Rice, maize, fruits, and vege-tables, for example, were most heavily con-sumed by farmers, tenants, and farm laborers,
15 Prasarn Trairatvorakul, 77ie Effects on Income Distribution and Nutrition of Alternative Rice Price Policies inThailand, Research Report 46 (Washington, D.C.: International Food Policy Research Institute, 1984).
41
60
is
-- ro"o<>; o O ^ •— O 0 0 -
—' f«j a d o' N o- o " <N *-* o o
C0C0O;O; i/1 O in N f l O - J 1
inmcocsi r-'coin-—> O;^ ' - " t^! r*s od 06
^ ,_ ,_, -o O CO CO <> O; CO ^" ^ O; CO CO '-J O in t>. C>C 6 < N O 6 "^ co* co in in co * ~- cd co co " m* ~-* d co <N ro co in •
(N co Tt in 0;OC0O| 0;'-|<jeoo)0;0;(Sr<lO;riinM O;cq<NinfSl ~ fsq oq ^ c^ csj ^, ^ CN) r-j " <N — ^-<N CO'^ IN (N (N ^ ^
in coo^co 00 o •—' 0;oqinocO'—incocoincocoo
^-^- , — cq ro cod d d d d i
-—• co o - r>«
~ ~ do
in i** in in ^ 1 ° ^ ^ ^ cqcooO'-;cO'o* IN." t*s" co* r>i ts'-d^ic o - - • • - - •
co O O r O cOcsj-Ot-s.CN. C O N N - - - -
•c
ail_. **3 rt 21 -t\ E r^. ^H - L "
o
3" S
•a
(A
I i
1 Im
S 5
OCO " T-
3fS&
42
and fish and fish products by the boat-owningfishermen and the hired fishermen. Thediets of professionals and overseas workersvaried more, as indicated by the smallerproportions spent on staples and the in-creasing percentage allocated to foods suchas meat, milk, eggs, bread, and cooking oil.Within the farming groups, the landlessfarm laborers had the least varied diet, spend-ing most of their food budget on rice andmaize. In general, the composition of thefood budget of land-owning farmers wasmore varied than that of the tenant farmers.
Population groups forming the lowestquartile spent larger proportions of theirbudgets on staples, such as rice, maize,sweet potatoes, and cassava, and smallerproportions on milk, eggs, meat, and cook-ing oil. From a high of 46 percent in thelowest income quartile, rice's share of thefood budget declined to 33 percent in thehighest quartile. Expenditure shares ofbread, pork, beef, and cooking oil werelargest in the highest income quartile.Shares spent on fish did not appear to varyby income group. This may be because thehigher income quartiles in the sample werestill at the lower end of the income distribu-tion of the Philippines as a whole.
Food ConsumptionIn the food-weighing procedure used to
estimate food consumption, all food servedto household members was weighed beforemeals and the leftovers after for a given24-hour period, and each of these foods wasconverted into calorie equivalents.
Daily calorie consumption, which isbased on the 24-hour weighing method, isestimated at 1,701 calories per AEU for theentire sample. This is about 374 caloriesless than the standard recommended byFNRI.16 Thus, it appears that the sample
area as a whole was deficient in calories.Average measures, however, tend to maskthe deficiency levels for particular popula-tion groups. For purposes of this analysis,the calorie consumption and adequacy ofthe sample population was disaggregated bylocation (province), survey round, occupa-tional grouping, and income group.
Statistics summarized in Table 9, usinginformation from the 24-hour food weighingsurvey, suggest that the people in SouthCotabato were worse off than those in theother two provinces. It was the only prov-ince where calorie deficits averaged morethan 400 calories per day. As shown inTable 6, South Cotabato had the lowest percapita incomes, indicating a likely relation-ship between undernutrition and economicdeprivation.
Caloric intakes vary significantly evenwithin subsectors of socioeconomic groupsbased on occupation. For instance, withinthe farming sector, it is clear that rice farm-ers who are landowners have higher caloricintakes than farmers growing maize or othercrops. It is, however, incorrect to concludethat all rice producers are relatively well-fed.It is evident from the results that the smallerrice producers, like tenant rice farmers,have large calorie deficiencies, and the prob-lem is even more pronounced among farmlaborers working in the rice sector: theircalorie deficits are the highest of all occupa-tional groups. The significant differences incaloric intakes among the various occupa-tional subcategories within rice farmingsupport the theory that the rice-farming sec-tor is not homogeneous; hence, analysis ofpolicies affecting the sector should explorethe distributional patterns within.
The disaggregation of the fishing sectorinto fishermen who own boats and hiredfishermen indicates varying degrees of de-ficiency between these groups. The signifi-cant difference in their means amplifies the
16 The Philippine standards for calories were based on the assumption that energy expenditures for Filipinosdiffer from those of other people because of differences in body size and type, as well as duration of physicalactivities, rather than because of intrinsic physiological differences. FNRPs energy expenditure and intake studiesbased on the Philippine population formed the basis for the calorie requirement standards used in this study. Asummary of the standards for all sex, age, and physiological status categories can be found in Philippines, Foodand Nutrition Research Institute, Publication No. 75 (Manila: FNRI, May 1977).
43
oppi
8
1
co" rn ri f*i in^d>-<q
' - » d ^ «d" in oi oi
co roror^j c*J c*i csj ro cqc^eocNJCMC^cvicsieoc^^coc^
in >q oq od d r^ to
ro i/> ^H O irt O O
~ -"^ fN) <S ^ —' ^-*('-lONomTf o — o o
ONfO ^^ ) i—< ^ t l / l CO <•
-;v>o^tsNO;roainqfOoq
in oa co o;d — oi
fN, © ^M pT) Q f\) H ^ If) ^ 1/^ f^ ^H r—I CO ^™ ^~" ^^ IO *O l / l fSI CO ^ ^
^ ^(sj^ ^^{sj^ ^ o o o ^ o o ^ (N ^ r»j *^ d d n p j
00 ^^ t^ 00 <> L^ ^ ^ (^ ^ < ^^ LO ^3 t" fO ^ ^ ^3 ^ ^ rO ^ 00 ^3 ^^ ^^* C? ^O
•— inoco ino i n o i - -Hrv ists>o N
CO_>«—'fN.csi oooooo '— r ^1'— S a o ' -IP^PQilOr?P>oaNO>fOO> r j o a s
O Is*" ^O t*1""
c
2
'a..2-
44
need to look at the patterns within populationgroups in order to understand the impactof particular policies. Among the nonfarm-ing occupational groups, wage earners wereparticularly deficient in calories, while, asexpected, the professionals and salariedworkers had the lowest average deficits.
Diet CompositionA disaggregation of the sources of calo-
ries for the sample households illustratesthe overwhelming dominance of rice in thediet, accounting for more than three-fourthsof total calories. The dependence on riceholds true in all areas, in practically all sea-sons, for all occupational categories, and forall income groups. The overall dependenceon one staple commodity also implies that,in general, there is little diversity in the diet.
The composition of the average diet var-ied slightly among the three provinces. Abrashowed the least dependence on rice, butits consumption of maize was much higherthan that of the other two provinces becausemaize is eaten in Abra during the leanmonths. The coastal province of Antiquederived the highest proportion of its calories(and protein) from fish.
Meal patterns of a typical household inthe sample consisted of staples such as rice(326 grams per capita daily), dried or freshfish, bagoongor fish sauce, leafy vegetablessuch as kangkong, malunggay, talong, andmung beans, and fruit. As mentioned be-fore, in areas where maize is grown, maizesupplements rice as a staple, which is eatenas a rice-maize mix. Bread is eaten in theform of a roll made of wheat flour calledpan-de-sal, which is usually eaten for break-fast. Pork, poultry, and beef are part of themeal only on occasion.
Household Calorie AcquisitionAs shown in Table 10, average calorie
acquisition for the sample as a whole, basedon the flexible recall period (one week orone month), exceeded calorie consumptionby 136 calories per AEU per day, or 7 per-
cent. As mentioned earlier, there are severalreasons for this slight divergence. First,because no food weighing was done onSaturdays and Sundays, food consumptionestimates refer to an average daily consump-tion for the week based on actual consump-tion for only five days, Monday to Friday,whereas food acquisition covered all sevendays of the week. Moreover, Philippinehouseholds usually consume more foodduring the weekend. Second, part of thedifference may also be explained by the fail-ure to take into account wastage in the esti-mates based on food acquisition.
Although it is not possible to determinethe bias of each of the two estimates, andthus the "true" value, it may be hypothesizedthat the two estimates provide a lower andupper bound. It is also likely that the esti-mate based on acquisition is closer to thetrue value because it reflects consumptionduring all days of the week and because themethod itself is unlikely to influence actualconsumption.
There was also a slight seasonal vari-ation in the composition of the diets in thesample. The relative importance of rice de-clined slightly during the lean month ofNovember, compared to May, due to theincrease in consumption of maize in Abra.A large proportion of Abra's population eatsrice mixed with maize grits on a 60-40 or70-30 ratio during the maize harvest monthsfrom August to November.
Tenant farmers of cash crops, landlessfarm laborers, and fishermen had the leastdiverse diets, while the most diverse were,as expected, those of the professionals andsalaried workers, the households of workersemployed overseas, and the landowningrice farmers. Professionals depended less oncereals for their caloric needs and obtaineda relatively higher proportion from suchfoods as bread, cooking oil, milk, meat, andeggs. The opposite was true for the low-in-come occupational groups, such as the land-less farm laborers, who depended on ricefor 82 percent of total caloric intakes.
Table 9 indicates the increasing diver-sification in the average diet as income in-creases. This is marked by the decreasingbut still significant role of rice and maize in
45
?eo
4)
1
N S S fv C* f> t-s
O O t*s m O N ' - H
oo ^" co LT> -o in in
~ d d ^f oi co co <N CO CO CO -0 co ro co co co co
co co^-co ^tcoforo cocococococococnuntNco^t^ cococo^t
tso>q <>oqcocoi ' i
i i i * i o0 0 0 ui o o vn o in o o m o o o o 0 0 0 0
rOtvcOOCOCOOfN^-O^OOCv
2
•a
I
2
1
9-3
|
It' i f•aS
I22 sS S
sS£ tota •aIIo o
§ 1
I
46
overall food consumption and the increasingrelative importance of such calorie sourcesas bread, cooking oil, sugar, beef, pork, poul-try, milk, and eggs. Although the poorestincome group, as expected, derived thehighest proportion of total calories from riceand maize, their diet was surprisingly highin fish, fruits, and vegetables. One possibleexplanation is that the lowest income groupreceives a relatively higher proportion oftheir nonstaple foods as wages (such as fish)or as gifts from neighbors (backyard produceof fruits and vegetables).
The share of calories from cooking oilfor the highest income quartile was almosttwice that of the lowest income group. Thesame trend can be observed for milk, eggs,pork, beef, and bread.
It appears that the 24-hour food weigh-ing method overestimated the consumptionof rice and underestimated the consumptionof virtually all other commodities. This pat-tern was consistent for all population groups.As a consequence, the relative importanceof rice in the diet is much lower if foodacquisition data are used.
The difference between food acquisitionand food consumption data is large for sugar,fruits and vegetables, and other foods.17 Itis conceivable that these commodities playeda more important role in weekend mealsand therefore were underrepresented in the24-hour weighing. Or, sugar used in foodpreparation and in coffee may not have beenfully accounted for in the food weighing,which focused on the main meals. Finally,fruits and vegetables consumed betweenmeals may not have been fully reflected inthe food weighing.
Maize acquired under flexible period re-call is reported to be more than twice theamount of maize consumed under 24-hourweighing. This difference can be traced al-most totally to maize in Abra, which wasestimated to be 125 calories per AEU perday using the weighing method and 318
calories per AEU per day using recall. Thereis no obvious explanation for this difference.
Individual Calorie Intakeand Adequacy
For a subsample composed of 140 house-holds, individual food consumption datawere collected, using a combination of tech-niques. Food weighing and observationtechniques were used in determining intra-familial distribution of food, while the recallmethod was used in determining food con-sumed outside the home. Data were collectedfor every member of the household over a24-hour period, and the survey was repeatedfor the same households and members fourtimes during the study period.
To assess whether a person consumedan amount more, less, or equal to the aver-age calorie requirement, the study com-pares the calories consumed with the Philip-pine RDA for calories.18 Table 11 providesdata on calorie intake and calorie adequacyratios (intake divided by require ments) bysex and age group for each of the four pop-ulation quartiles. The mean calorie intakeof preschool children was far below require-ments, particularly for females, who ob-tained only 50 to 55 percent of their require-ments. Male preschool children, althoughbetter off than their female counterparts,still consumed much less than the norm.The difference between adequacy rates formale and female preschoolers is statisticallysignificant at the 5 percent level.
The calorie adequacy of adults above18 years of age was higher than that forchildren and adolescents. There was no sig-nificant difference between the husbands'and wives' calorie adequacy, except whenthe wives were pregnant or lactating. Wo-men who were pregnant or lactating at thetime of the survey had adequacy ratios of0.69 compared to 0.78 for all wives and
17 Such differences were also found in Egypt. See Harold Alderman and Joachim von Braun, The Effects of theEgyptian Food Ration and Subsidy System on Income Distribution and Consumption, Research Report 45(Washington, D.C.: International Food Policy Research Institute, 1984).18 Philippines, National Nutrition Council, Nutrition Research Committee, Philippine Recommended DietaryAllowances, Part I, Nutrients (Manila: National Nutrition Council, 1976).
47
i
173
I(A
S1
o(A
I
&
O COro TT
ooooo"
— — — (N <N
^ CO f l Tt i—'Tt i—'0 0 * 0 1/) COOONNON-- §2
SMOCOOO r oo
VIN'OOO-— fN — O —
CSJ CO CO f O CO O C^ CO coS3
o o d
M - in «-O "fNNNO00-CDN O O- O O- ootoro
^M*O «J N
<>7 I I I SI I ncow
^- t~N r- — ^ - -O
oo o-«j- £— M o p
to o. o in
1 g1
•C 3 .S
l> 2 «sS
<S
Hi
• S 1
a
eg w i3
o3l
9U ,2"S
il°3
fO O — — <N
8.
co to r o o r
LO t^ <N 00 — —<N O to 00 Tf —
O- IT* CO fOIA[S
d odd do
ocsooc-O r o O O N
•S5 I I
48
r oo o oo o
O tx CM fM 00 t~»00 (N NO
p-Tc-T w
r— IT)
? r iP *~* o ro" • "—' CO CO
IO
f
I
CM ~ CO O ^
7 n i S
ts — ^
| 8T§
I ISJ^S g
111! ii&S
« CO
• a T3 «< <o
III!
49
0.80 for husbands. Again, the calorie ade-quacy rate tended to rise as household in-come increased for most groups. The pat-terns observed in this study essentially con-form to those observed in other regions inthe Philippines.19
These findings suggest that in order toreach nutritionally vulnerable householdmembers, targeting households with mal-nourished preschoolers may be inadequate.School-age children and adolescents, partic-ularly females, were as deficient in caloriesas preschoolers.
Nutritional Status ofPreschool Children
Anthropometric data for all preschoolchildren in the sample households providethe basis for determining their nutritionalwell-being relative to a particular growthstandard. Four indicators of nutritionalstatus are reported in Table 12: meanweight as a percent of standard, proportionof children below 75 percent of standardweight-for-age, proportion below 60 per-cent, and Z-scores of weight-for-age. Al-though a local Philippine weight standardis available, this portion of the study usesinternational standards developed by theNational Center for Health Statistics of theUnited States in order to facilitate interna-tional comparisons of results.
Results reveal that one of every four chil-dren in the study areas is malnourished toeither the second or third degree (less thanthe 75 percent cutoff suggested in the
Gomez classification system). Malnutritionappears to have been higher in the provinceof Antique than in the two other sampleprovinces. The data suggest that malnutri-tion declined dramatically from 34 to 18percent during the 17 months between thefirst and last survey round. The improve-ment in child weight relative to the standardis also reflected in the mean weight-for-age,which increased from 80.7 to 85.0 percent.The Z-scores for weight-for-age are also con-sistent with nutritional improvement dur-ing the period.
The incidence of child malnutritionvaried according to the occupational groupof the parents. The highest percentage ofmalnourished children belonged to familiesof hired fishermen (40 percent), boat-own-ing fishermen (30 percent), tenant farmersof nonfood crops (28 percent), maize farm-ers (24 percent), wage earners (23 percent),and landless farm laborers (22 percent). Thegroups that were relatively well off werethe children of professionals and salariedworkers and landowning rice farmers—theoccupational groups with the highest in-comes (see Table 6). These results indicatethat income and sources of income are fac-tors that account for differences in nutri-tional status across occupational groups.The analysis of nutritional level by incomegroup also bears out this observation. Itshows that 15.9 percent of children in thehighest income quartile were malnourished,compared with 29.8 percent of childrenbelonging to the lowest income quartile.Z-scores of weight-for-age and mean weight-for-age show similar improvements as incomeincreases.
19 Robert E. Evenson, Barry M. Popkin, and Elizabeth K. Quizon, "Nutrition, Work and Demographic Behaviorin Rural Philippine Households," in Rural Households in Asia, ed. Hans B. Binswanger et al. (Singapore: SingaporeUniversity Press, 1980); Melba B. Aligaen and Cecilia A. Florencio, "Intra-Household Nutrient Distribution andAdequacy of Food and Nutrient Intake of Filipino Urban Households," Philippine Journal of Nutrition (January-March 1980): 11-19; and Rosario Valenzuela, "A Study on Nutrient Distribution within the Family and FactorsAffecting Nutrient Intake," University of the Philippines, Diliman, 1977 (mimeographed).
50
Table 12—Nutritional status of preschool children, 1983-84
Category
Entire sample0
Province0
AbraAntiqueSouth Cotabato
Survey roundMay 1983November 1983May 1984November 1984
Occupational categories0
Rice farmer landownerMaize farmer landownerFarmer, other cropsTenant rice farmerTenant maize farmerTenant, other cropsLandless farm laborerWage earnerProfessional/salariedFishermen, boat ownersHired fishermenEmployed overseasOccupation unclassified
Income groupc
First quartile (lowest)Second quartileThird quartileFourth quartile
NumberofObser-vations
3,053
8031,336
914
768766770749
194228
8394
13853
258801100410
35233426
763763763764
Mean Weightas a Percentof Standard
Weight-for-Age
(percent)
83.12
83.8582.7183.09
80.7483.1883.9085.03
86.2882.9784.9482.3784.2280.9783.4783.0986.5381.2380.0383.8682.75
81.3582.0683.4086.07
MeanZ-Scores
ofWeight-for-Age
-1.64
-1.58-1.71-1.60
-1.87-1.66-1.57-1.47
-1.33-1.70-1.48-1.70-1.43-1.82-1.63-1.64-1.31-1.87-2.03-1.63-1.64
-1.84-1.73-1.64-1.34
Percentage ofMalnourished ChildrenSecondDegree"
ThirdDegree1"
(percent)
23.16
20.3326.2420.85
29.8824.4020.9916.82
20.2722.9217.1422.5618.0127.0621.5422:'.O514.8730.6341.4220.9424.78
29.2026.3521.9814.86
1.84
1.001.812.65
4.421.360.530.90
1.723.650.950.001.901.181.031.781.032.382.860.492.83
2.921.801.491.19
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Note: The z-score is a method used in standardizing the distribution of actual weight of the child relative tothe standard weight for a child of that sex and age. The standards devised by the U.S. National Centerfor Health Statistics (NCHS) were used in the study.
a Second-degree malnutrition includes those children between 60 and 75 percent of the standard weight-forage.b Third-degree malnutrition includes children weighing less than 60 percent of the standard weight-forage.c These data are for all survey rounds combined.
51
A COMPARATIVE ANALYSIS OF THE EFFECTSOF THE FOOD PRICE SUBSIDY SCHEME
In this chapter the effects of each of thetwo components of the scheme—price sub-sidy and nutrition education—are estimatedbased on direct comparisons between house-holds that received these components andhouseholds that did not. As explained ear-lier, the first survey round was done beforethe scheme was implemented and the fourthround after the scheme was completed. Toavoid the effects of seasonal variations, thethird round was done exactly one year afterthe first and the fourth round one year afterthe second. Thus, a direct comparison be-tween rounds one and three should yieldan estimate of the impact of the schemeprovided no other changes took place duringthe year that influenced the outcome vari-able. But such changes may have occurred.Therefore a control subsample—one thatwas not a part of the scheme—is needed.Assuming that any external changes werethe same for the control subsample and thesubsample participating in the scheme, theseexternal effects can be isolated from theeffects of the scheme.
This is the approach used here. Meansfor control and treatment groups for eachsurvey round and means for survey roundsone and three and two and four were testedstatistically for significant differences. Thenthe changes in the treatment groups fromthe price subsidy and the nutrition educa-tion program were estimated using the meth-ods presented in Chapter 5. Estimates areprovided for the impact on household foodexpenditures, calorie acquisition, and calo-rie consumption, as well as calorie adequacyof groups of individuals, and mean weightand height for age of preschoolers.
Household Food ExpendituresHousehold food expenditures increased
greatly for all household groups because food
prices rose steeply during the study period(Table 13). The increase was larger for house-holds receiving the subsidy, that is, about9 percent—approximately the size of thesubsidy—after the scheme had been in ef-fect for 10 months (Table 14). These house-holds continued to spend more on food afterthe scheme was discontinued. Householdsreceiving both the subsidy and nutritioneducation showed the largest increase, butnutrition education without the subsidydoes not seem to have influenced food ex-penditures. Clearly, enhancement of pur-chasing power of the poor and improvementof their nutrition-related knowledge arestrongly complementary. The effect of oneis greatly influenced by the presence of theother.
Household Food AcquisitionHousehold calorie acquisition increased
with the introduction of the subsidy and fellback almost to presubsidy levels when thesubsidy was discontinued (Table 15). House-holds receiving the subsidy increased theirdaily calorie acquisition by about 250 cal-ories per AEU, while the calories acquiredby households in the control group actuallydeclined by almost 30 calories per AEU perday. On the assumption that this fall wascaused by factors (primarily price increases)that also influenced the subsidized house-holds, the impact of the subsidy was judgedto be about 280 calories per AEU per day,including the decrease avoided. Thus, house-holds receiving the subsidy acquired about19 percent more calories than they wouldhave without the subsidy (Table 16). Al-though food acquisition fell when the schemeended, it was still about 11 percent aboveprescheme levels after adjusting for effectsof external factors. Thus, it appears that the
52
Table 13—Weekly food expenditures for each subsample and survey round
Subsample/Levels of Significance
Subsample receiving subsidyWith nutrition educationWithout nutrition education
Subsample not receiving subsidyWith nutrition educationWithout nutrition education
Subsample receiving nutrition educationSubsample receiving no nutrition educationLevels at which means are significantly different8
Subsidy versus no subsidyTotal sampleSubsample receiving nutrition educationSubsample receiving no nutrition education
Nutrition education versus nutrition education
1
20.020.419.520.421.419.420.919.5
n.s.n.s.n.s.
0.01
Survey Round2
24.225.023.321.922.521.323.822.3
0.0010.0050.05
n.s.
3
36.536.636.334.235.832.336.234.5
0.05n.s.
0.05n.s.
4
34.034.233.732.934.631.334.432.5
n.s.n.s.
0.050.05
Levels at WhichMeans are Signifi-cantly Different3
BetweenRoundsland 3
0.0010.0010.0010.0010.0010.0010.0010.001
0.0010.0010.0010.001
BetweenRounds2 and 4
0.0010.0010.0010.0010.0010.0010.0010.001
0.0010.0010.0010.001
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
a Significance is based on t-test; n.s. means not significant at the 0.10 level.
Table 14—Change in household food expenditures due to price subsidy ornutrition education
Scheme
Subsidy versus no subsidyTotal sampleSubsample receivingnutrition education
Subsample receiving nonutrition education
Nutrition education versusno nutrition education
Two MonthsAfter
InitiationofScheme
12.7
16.6
8.8
-0 .4
Time of Measurement
Ten MonthsAfter
InitiationofScheme
(percent)
8.9
7.2
11.8
-2.1
Two MonthsAfter
TerminationofScheme
5.4
3.7
7.1
-1.2
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Notes: Percentages of change in food expenditures are in comparison to the baseline and corrected for exogenouschanges reflected in the control sample. Changes are estimated on the basis of direct comparisons betweencontrol and treatment populations. The value of the subsidy was 9.25 percent of food expenditures ofhouseholds receiving subsidies in survey round 2 and 8.5 percent in round 3.
53
Table 15—Daily calorie acquisition for each subsample and survey round
Subsample/ Levels of Significance
Subsample receiving subsidyWith nutrition educationWithout nutrition education
Subsample not receiving subsidyWith nutrition educationWithout nutrition education
Subsample receiving nutrition educationSubsample receiving no nutrition educationLevels at which means are significantly different
Subsidy versus no subsidyTotal sample (Subsample receiving nutrition educationSubsample receiving no nutrition education
Nutrition education versus no nutritioneducation
1
,742,798,686,836,889,784,844,735
).050n.s.n.s.
n.s.
Survey Round2
2,0382,0632,0131,7931,8441,7421,9541,878
0.0010.0010.001
n.s.
3
2,0092,0291,9891,8021,8731,7311,9511,860
0.0010.0050.001
0.005
4
1,7881,7461,8311,7101,7781,6431,7621,737
0.050n.s.
0.001
n.s.
Levels at WhichMeans are Signifi-cantly Different
BetweenRoundsland 3
0.0010.0010.001
n.s.n.s.n.s.
0.050n.s.
BetweenRounds2 and 4
0.0010.0010.0010.0500.050
n.s.0.0010.005
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Note: Levels of significance are based on t-test; n.s. means not significant at the 0.10 level.
Table 16—Change in daily calorie acquisition due to price subsidy or nutritioneducation
Scheme
Subsidy versus no subsidyTotal sampleSubsample receivingnutrition education
Subsample receiving nonutrition education
Nutrition education versusno nutrition education
Two MonthsAfter
Initiationof Scheme
20.9
17.8
21.6
-2.1
Time of Measurement
Ten MonthsAfter
Initiationof Scheme
(percent)
18.6
14.0
21.0
-1.3
Two MonthsAfter
Terminationof Scheme
11.2
3.4
17.3
-4 .6
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Notes: Percentages of change are in comparison to the baseline and corrected for exogenous changes reflectedin the control sample. Changes are estimated on the basis of direct comparisons between control andtreatment populations.
54
scheme had a long-term effect on food acqui-sition over and above the effect of the in-come transfer itself. This effect does notseem to originate with nutritional education,which, if anything, had a negative effect oncalorie acquisition.
Household CalorieConsumption
According to data from 24-hour foodweighing, household calorie consumptionby subsidized households increased slightly,although these changes were not statisti-cally significant. During the same period,calories consumed by nonsubsidized house-holds decreased significantly (Table 17).Thus it appears that the subsidy effectivelycountered a decrease stemming from otherfactors and resulted in calorie consumptionabout 10 percent above what it would havebeen without the subsidy (Table 18). Thenegative effects of external factors are fullyreflected in the calorie consumption of thesubsidized households after the scheme wasdiscontinued. Whereas consumption levels
of the two groups were significantly differ-ent during the latter part of the scheme,this difference disappeared as soon as thescheme was terminated, leaving both groupsconsuming less than before. Nutrition edu-cation appears to have had little or no effecton calorie consumption as measured by 24-hour food weighing.
Calorie Consumption byIndividuals
The calorie adequacy rate increased dur-ing the period of the subsidy and fell afterthe subsidy was terminated for virtually allmembers of subsidized households (Tables19 and 20). But adult males and femalesappear to have gained considerably morefrom the scheme than children. Some of theadults' relative gain seems to have beenmaintained after the scheme was discon-tinued. However, school-aged children,who did gain some from the scheme, wereconsiderably worse off after the scheme wasterminated than before it was initiated.There is no obvious explanation for thisphenomenon.
Table 17—Daily household calorie consumption for each subsample andsurvey round
Subsample/Levels of Significance
Subsample receiving subsidyWith nutrition educationWithout nutrition education
Subsample not receiving subsidyWith nutrition educationWithout nutrition education
Subsample receiving nutrition educationSubsample receiving no nutrition educationLevels at which means are significantly different
Subsidy versus no subsidyTotal sampleSubsample receiving nutrition educationSubsample receiving no nutrition education
Nutrition education versus no nutritioneducation
1
1,7411,7241,7571,7511,7911,7101,7381,734
n.s. in.s. (n.s.
n.s.
Survey Kound2
,676,669,684,763,814,710,742,697
).O5O).OO5
n.s.
n.s.
3
1,7881,8041,7721,6391,6361,6411,7201,707
0.0010.001 (0.001
n.s.
4
,651,613,688,689,717,661,665,675
n.s.).050
n.s.
n.s.
Levels at WhichMeans are Signifi-cantly Different
BetweenRoundsland 3
n.s.0.050
n.s.0.0010.001
n.s.n.s.n.s.
BetweenRounds2 and 4
n.s.n.s.n.s.
0.050n.s.n.s.
0.050n.s.
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Notes: Levels of significance are based on ttest; n.s. means not significant at the 0.10 level.
55
Table 18—Change in daily household calorie consumption due to pricesubsidy or nutrition education
Scheme
Subsidy versus no subsidyTotal sampleSubsample receivingnutrition education
Subsample receiving nonutrition education
Nutrition education versusno nutrition education
Two MonthsAfter
Initiationof Scheme
-4 .4
-4 .4
-4 .2
2.4
Time of Measurement
Ten MonthsAfter
InitiationofScheme
(percent)
9.7
14.5
5.9
0.5
Two MonthsAfter
TerminationofScheme
-1.7
-2.4
-1.1
-0 .8
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Notes: Percentages of change are in comparison to the baseline and corrected for exogenous changes reflectedin the control sample. Changes are estimated on the basis of direct comparisons between control andtreatment populations.
According to this measurement, nutri-tion education had a significant and positiveeffect on the calorie adequacy of most groups,especially pregnant and lactating women,when provided along with food subsidies.No impact is detected when nutrition edu-cation was provided alone, confirming onceagain the importance of providing nutritioneducation together with expanded purchas-ing power.
Nutritional StatusThe weight-for-age of the preschoolers
in the sample relative to standard weight-for-age increased during the study period(Table 21). The increase was largest for pre-schoolers in the households that partici-pated in the subsidy scheme (Table 22).
Based on these comparisons, it may be con-cluded that the scheme improved the aver-age weight of the preschoolers relative tothe standards. The effects continued for atleast two months beyond the terminationof the scheme. Nutrition education does notappear to have contributed to these im-provements.
Table 23 shows changes in the height-for-age of the preschoolers studied. A gen-eral increasing trend was found in all groups.The increase was slightly larger among pre-schoolers from subsidized households. Thus,it appears that the scheme caused an in-crease in height of 2.7 percent during thefirst 10 months of its existence. The effectwas slightly higher in the group receivingboth subsidy and nutrition education thanin those receiving only one of those compo-nents (Table 24).
56
Table 19—Individual calorie adequacy by age and gender for each subsampleand survey round
Survey Round/Sample
Survey round 1SubsidizedWith educationWithout education
NonsubsidizedWith educationWithout education
Survey round 2SubsidizedWith educationWithout education
NonsubsidizedWith educationWithout education
Survey round 3SubsidizedWith educationWithout education
NonsubsidizedWith educationWithout education
Survey round 4SubsidizedWith educationWithout education
NonsubsidizedWith educationWithout education
Hus-bands
0.730.70
0.960.85
0.880.78
0.850.68
0.880.82
0.770.80
0.720.70
0.850.88
Wives
0.780.74
0.880.71
0.860.87
0.780.72
0.840.82
0.760.76
0.800.74
0.800.76
Adults
Pregnant/LactatingWomen
0.650.70
0.710.56
0.760.77
0.670.60
0.790.77
0.730.62
0.710.72
0.710.71
All 14Years
Male
0.750.73
0.900.74
0.850.79
0.790.62
0.820.81
0.750.69
0.680.72
0.810.82
Fe-male
Children1-7 Years
Male
(percent)
0.790.78
0.810.69
0.860.82
0.760.69
0.850.85
0.770.71
0.800.80
0.770.76
0.550.60
0.550.67
0.630.62
0.560.60
0.660.60
0.550.62
0.650.71
0.600.68
Fe-male
0.560.57
0.570.46
0.620.61
0.530.38
0.610.62
0.540.54
0.520.67
0.620.45
7-14 Years
Male
0.660.72
0.640.64
0.650.65
0.690.41
0.720.61
0.660.65
0.610.68
0.740.72
Fe-male
0.610.77
0.590.51
0.680.65
0.550.47
0.690.56
0.550.58
0.520.58
0.640.71
All
Male
0.670.67
0.730.69
0.730.68
0.680.58
0.740.68
0.660.65
0.660.70
0.710.72
Fe-male
0.660.69
0.710.59
0.730.71
0.630.55
0.730.72
0.660.63
0.640.70
0.690.64
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
57
S$
I
I!i
I
•3; in O; to^ o o oI I I
— CM O;O-' CO i n
CM O —
in O —• CO
CM CM O —-d —• <> <>CM CO — I
Iin
or
28.0
O ;CO
15.0
20.7
29.1
12.8
24.7
24.4
29.2
-6.6
mto
15.9
3.0
-6.2
-10.
9-9
.9
l
I
CO " ^ • q - CM
in r~ co od
co in qo ^t <5in CM |
•q; CM t o<6 -d c>CO CO CO
D.
M a i lS C o !> C
« S £ S i g -B
S E a S a o i " c
IIII11115 z
s.
Tf -O ol —
~ TT o tv— CM I
in r^ co— I |
q — o coto o c> inCM — CM
CO CO -O t v— co <i -G
co q ~ TtCM o od d
CM in -<t in•q1 CM' in -^
(SjCO
qod
ro
a.o roc M
O 3 S
c S3•? B ? B C .2
g 6 o . § a g « c
C/3
O 3
qo-
2 S
— COo --
Oro <M
IV O— CM
I I
CO CO
in in
coI
58
in ro
I I
I I
Cv. IT)
O <>
11
O
c
g gS3T O
Mr; w
iilp! 3 ra
IIII*a o «» c6 ;S e .2
Z
53 tj •
tifS.1
59
Table 21 —Mean weight-for-age of all preschoolers for each subsample andsurvey round
Subsample/Levels of Significance
Subsample receiving subsidyWith nutrition educationWithout nutrition education
Subsample not receiving subsidyWith nutrition educationWithout nutrition education
Subsample receiving nutrition educationSubsample receiving no nutrition educationLevels at which means are significantly different
Subsidy versus no subsidyTotal sampleSubsample receiving nutrition educationSubsample receiving no nutrition education
Nutrition education versus no nutritioneducation
1
77.379.974.484.382.885.780.481.4
0.050.05n.s.
n.s.
Survey Round2 3
(percent)
81.281.580.984.281.686.481.083.5
n.s.n.s.n.s.
n.s.
84.483.985.182.281.083.482.784.2
0.005n.s.
0.005
n.s.
4
84.985.384.683.381.684.983.784.6
0.0050.0500.050
n.s.
Levels at WhichMeans are Signifi-cantly Different
BetweenRoundsland 3
0.0010.0010.001
n.s.n.s.
0.0500.0010.001
BetweenRounds2 and 4
0.0010.0050.005
n.s.n.s.
0.0500.0100.001
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Notes: Levels of significance are based on t-test; n.s. means not significant at 0.10 level.
Table 22—Change in weight-for-age of preschoolers due to price subsidy ornutrition education
Scheme
Subsidy versus no subsidyTotal sampleSubsample receivingnutrition education
Subsample receiving nonutrition education
Nutrition education versusno nutrition education
Two MonthsAfter
Initiationof Scheme
5.3
3.5
8.0
-1.7
Time of MeasurementTen Months
AfterInitiationofScheme
(percent)
12.1
7.3
17.4
0.7
Two MonthsAfter
TerminationofScheme
11.3
8.3
14.9
-0.3
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Notes: Percentages of change are in comparison to the baseline and corrected for exogenous changes reflectedin the control sample. Changes are estimated on the basis of direct comparisons between control andtreatment populations.
60
Table 23—Mean height-for-age of all preschoolers for each subsample andsurvey round
Subsample/Levels of Significance
Subsample receiving subsidyWith nutrition educationWithout nutrition education
Subsample not receiving subsidyWith nutrition educationWithout nutrition education
Subsample receiving nutrition educationSubsample receiving no nutrition educationLevels at which means are significantly different
Subsidy versus no subsidyTotal sampleSubsample receiving nutrition educationSubsample receiving no nutrition education
Nutrition education versus no nutritioneducation
1
89.889.290.791.892.890.990.790.8
0.0050.001
n.s.
n.s.
Survey Round2 3
(percent)
91.390.791.992.093.490.891.891.4
n.s.0.001
n.s.
n.s.
93.493.793.293.094.791.494.192.4
n.s.n.s.
0.050
0.010
4
93.293.093.392.494.490.793.692.2
n.s.n.s.
0.001
0.050
Levels at WhichMeans are Signifi-cantly Different
BetweenRoundsland 3
0.0010.0010.0100.0500.050
n.s.0.0010.010
BetweenRounds2 and 4
0.0010.005
n.s.n.s.n.s.n.s.
0.010n.s.
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Notes: Levels of significance are based on t-test; n.s. means not significant at the 0.10 level.
Table 24—Change in height-for-age of preschoolers due to price subsidy ornutrition education
Scheme
Subsidy versus no subsidyTotal sampleSubsample receivingnutrition education
Subsample receiving nonutrition education
Nutrition education versusno nutrition education
Two MonthsAfter
InitiationofScheme
1.4
1.0
1.4
0.5
Time of Measurement
Ten MonthsAfter
InitiationofScheme
(percent)
2.7
2.9
2.2
2.0
Two MonthsAfter
TerminationofScheme
3,1
2.5
3.1
1.6
Notes: Percentages of change are in comparison to the baseline and corrected for exogenous changes reflectedin the control sample. Changes are estimated on the basis of direct comparisons between control andtreatment populations.
61
8A MULTIVARIATE ANALYSIS OF THE EFFECTSOF THE FOOD SUBSIDY SCHEME
Direct static comparisons may not yieldreliable results because the effects of thescheme are not effectively separated from theeffects of other factors.20 Therefore, in thischapter comparative analysis is supplementedwith empirical results of multivariate analy-sis, employing the analytical methodologypresented in Chapter 5.
A comparison of the food consumptionpatterns between the first (baseline) surveyand the subsequent survey rounds indicatesthat the rationed subsidies for rice wereinframarginal for about 90 percent of therecipient households. On average, rice ac-quisition per capita was estimated at 118kilograms annually before the subsidy wasintroduced, whereas the per capita rationwas only 60 kilograms per year. For cookingoil, however, the rationed quantity was ex-tramarginal. On average, per capita intakeof cooking oil was about 2.20 kilograms peryear, while the subsidized quantity was es-timated at 4.70 kilograms per year.
Effects on HouseholdFood Expenditures andCalorie Acquisition
The parameters of equation (3) in Chap-ter 5 are estimated by ordinary least squaresfrom the pooled sample from the four surveyrounds. The regression results are reported
in the Appendix, Table 34, and the parame-ter estimates for the key variables are sum-marized in Table 25. The estimated priceand income elasticities and the MPCs arereported in Table 26. As shown in Table25, changes in household income signifi-cantly affected both food expenditures andcalorie acquisition. Furthermore, the sub-sidy seems to have had an impact on thesehousehold variables over and above the ef-fect expected from the real income em-bodied in the subsidy. Changes in the priceof rice did not have a significant effect oneither food expenditures or calorie acquisi-tion. Changes in the price of cooking oilaffected food expenditures as well as calorieacquisition significantly. No significant ef-fect of the nutrition education componentof the scheme was detected.
The estimated income elasticity for totalfood expenditures, which was about 0.68,is in the plausible range and is similar tothose obtained in studies of Sri Lanka (0.72),Thailand (0.65), Egypt (0.69), and Bangladesh(0.67).21 The calorie acquisition elasticityof 0.33 is about half of the food expenditureelasticity, indicating that even poor ruralhouseholds have a tendency to shift tohigher-priced sources of calories as incomesincrease.22 Thus, the average increase inhousehold incomes of 8.0 percent broughtabout by the rice and oil subsidy is estimatedto have caused an increase of 5.6 percent
20 Per Pinstrup-Andersen, "An Analytical Framework for Assessing Nutrition Effects of Policies and Programs,"Food Policy, ed. Charles K. Mann and Barbara Huddleston (Bloomington: Indiana University Press, 1985), pp.55-66.21 David E. Sahn, "Malnutrition and Food Consumption in Sri Lanka: An Analysis of Changes from 1969 to1982," International Food Policy Research Institute, Washington, D.C., 1986 (mimeographed); Trairatvorakul,Effects on Income Distribution and Nutrition; Alderman and von Braun, Effects of the Egyptian Food Ration andSubsidy System; and Odin K. Knudsen and Pasquale L. Scandizzo, Nutrition and Food Needs in DevelopingCountries, Staff Working Paper 328 (Washington, D.C.: World Bank, 1979).22 This value is about the same as the average estimated by Alderman for a number of countries (Harold Alderman,The Effect of Food Price and Income Changes on the Acquisition of Food by Low-Income Households (Washington,D.C.: International Food Policy Research Institute, 1986]).
62
Table 25—Key coefficients for household food expenditure and foodacquisition regressions
IndependentVariable
Log income
Subsidy
Price of rice
Price of cooking oil
Nutrition education
Regression R2
N
Dependent VariableHousehold Food Expenditures
Model A-1 Model A-2
(pesos/AEU/day)
2.20(57.82)a
0.03(7.30)a
0.006(0.37)
0.012(4.11)a
-0.02(-0.73)
0.73
2,509
2.16(56.72)a
0.16(4.96)a
0.006(0.34)
0.004(1.49)
-0.02(-0.86)
0.72
2,509
Household Calorie AcquisitionModel B-l Model B-2
(calories/AEU/day)
633.29(27.37)a
8.94(3.32)a
-12.11(-1.12)
-5.98(3.5 l)b
22.67(1.28)
0.38
2,509
623.07(27.00)a
68.23(3.34)a
-12.22(1.12)
-7.45(5.14)a
21.86(1.24)
0.38
2,509
Notes: The adult equivalent unit (AEU) is a method used to convert the consumption of persons of different agesand sex to standard consumption units. The complete set of coefficients is given in the Appendix, Table34. The figures in parentheses are t-ratios.
a Significant at the 5 percent level.b Significant at the 10 percent level.
in total food expenditures and 2.6 percentin calorie acquisition if the income transferfrom the subsidy is treated the same as otherincome.
The estimated price elasticities for riceof-0.006 to -0.020 and -0.020 to -0.050for cooking oil seem low, but these shouldnot be confused with own-price elasticities.The price elasticities shown in Table 26 areprice elasticities for total food expendituresand total calories, and they show the neteffect after substitutions among com-modities have taken place. Therefore, theywould be expected to be lower than directprice elasticities for individual foods.
One of the issues addressed in thisstudy—whether the subsidy scheme influ-ences household food consumption overand above the effect of the income transferas measured by the income elasticityabove—is tested and evaluated using twodifferent model specifications (see Chapter5). As reported in Table 25, the subsidycomponent of the pilot scheme positivelyaffects both household food expenditures
and calorie acquisition: it is highly signifi-cant. The effect detected by the subsidyterm is over and above the price and incomeeffects. The effect of the nutrition educationcomponent of the pilot scheme is weak butpositive. Although significant at only a 0.20level, it appears important as a complemen-tary intervention in the scheme. To furtherexplore the effects of nutrition educationon households, interaction between nutri-tion education and the subsidy—that is,both schemes operating simultaneously—are tested in regressions. Such tests indicateno differential effects of nutrition education-subsidy interaction on calorie and food ac-quisition.
The estimates of absolute effects can begleaned from the MPCs and the price andincome elasticities. The MPC results showthat an additional P 1.00 of subsidy incomewould add P 0.69-0.98 to food expendi-tures. The 0.50 MPC for expenditures onfood from all income sources other than sub-sidies is within the range usually found inpoor rural areas in the developing world.
63
Table 26—Parameters for estimating the impact of the pilot subsidy schemeon household food expenditures and calorie acquisition
IndependentVariable
Marginal propensity to consumeAll income net of subsidySubsidy income
Consumption parametersIncome elasticity(net of subsidy)
Rice price elasticityOil price elasticity
Nutrition education effects
Dependent VariableHousehold Food ExpendituresModel A-1
0.50a
0.98a
0.66-0.01-0.04
0.02
Model A-2
0.53a
0.69a
0.70-0.01-0.04
0.02
Household Calorie AcquisitionModel B-l
142b
363"
0.32-0.02-0.0422.67
Model B-2
153b
222b
0.34-0.02-0.0421.86
Notes: See the Appendix, Table 34, for a complete set of regression coefficients. The adult equivalent unit (AEU)is a method used to convert the consumption of persons of different ages and sex into standard consumptionunits.
a Pesos per AEU per day.b Calories per AEU per day.
The difference in the MPC betweencash and subsidy incomes is also large forcalories. Thus, the marginal propensity toincrease calorie consumption from subsidyincome is about twice that of the MPC forall other sources of income. In absoluteterms, every additional peso of subsidy in-creases calorie consumption by 222-363calories per AEU. The impact of an addi-tional peso of other income is about 150calories per AEU.
Different MPCs for different sources ofhousehold income have also been found inother studies of consumption behavior inboth developed and developing countries.In Kerala, India, there is evidence that par-ticipation in rice subsidy programs in-creased the MPCs of some households.23
Several studies on the consumption effectof food stamps on U.S. households indicatethat the MPC for subsidy transfer is twiceto three times that of the MPC for moneyincome.24
The behavioral change that caused ahigher MPC for subsidy income cannot befully explained. Perhaps the scheme gener-
ated an increased awareness among par-ticipants about food and nutrition needsover and above the explicit nutrition educa-tion intervention that accompanied theprice subsidy. It is conceivable that mothers'awareness of the nutritional objectives ofthe scheme helped realign family budgetpriorities. In the pilot scheme, there was astrong persuasive element introduced by ef-forts to draw mothers into nutrition educa-tion classes administered by the extensionworkers.
Another plausible explanation is thatthe use of rice and oil as the food com-modities in the scheme altered food budgetsin such a way that the caloric content perunit of the household's food bundle wasincreased. Per capita daily calories acquiredfrom rice increased from 326 grams priorto the subsidy to 423 grams when the sub-sidy was in force. Daily calories acquiredfrom cooking oil increased by 30 percent.This raised the MPC for calories for someof the families. The calorie density per kilo-gram of rice and oil is about twice that offoods in the average food bundle of the study
23 Shubh Kumar, The Impact of Subsidized Rice on Food Consumption and Nutrition in Kerala, Research Report5 (Washington, D.C.: International Food Policy Research Institute, 1979).24 Senauer and Young, "Impact of Food Stamps on Food Expenditures."
64
households. Thus, in designing a project itis important to consider not only the eco-nomic value of the subsidized good but alsoits nutritive value if the program is to benutritionally effective.
Evaluated at their mean values, the totalenergy increase from the scheme is equiva-lent to 136-138 calories per AEU per day.This calorie effect combines net income ef-fects, price effects, and effects caused byhigher MPCs for subsidy. The scheme's totalimpact on calories is equivalent to 7 percentof the calories consumed per person in thestudy households. In terms of food expendi-tures, the scheme effectively increased totalfood budgets by P 0.27-0.34 per AEU perday, basically the same as the average trans-fer value of the subsidy, which was P 0.31per AEU per day.
Household income can be influenced bydecisions regarding the allocation of timeof each household member to various activ-ities, including income earning. Food sub-sidies may cause a change in time allocationin general and in the allocation of time toincome-earning activities in particular.Thus, it may be hypothesized that the sub-sidies reduced the time allocated to income-earning activities. This implies substitutionbetween the real income embodied in thesubsidies and household income from othersources.
Such substitution is ignored in the modelsused here and, if it is significant, the modelsoverestimate the effect of the subsidies. Toexplore whether such a substitution oc-curred, the household income variable wasreplaced by predicted wage rates for thehusband and wife of each household, andthe models were rerun. Neither the size ofthe coefficient of the subsidy variable norits level of significance changed appreciably.Furthermore, when the number of hoursworked in income-earning activities andhousehold incomes excluding the subsidyvalue were regressed on the value of thesubsidy and a set of other variables, no sig-nificant effect of the subsidy on hours workedor household incomes from sources other
than the subsidies was detected. Thus, itappears that the use of household incomesas an explanatory variable is acceptable forestimating the effect of the scheme.
Effects of Other FactorsThe regression of food expenditures on
household size gives negative and signifi-cant results, indicating that the presence ofmore family members reduces food consump-tion per AEU. Household size is generallymore closely related to food expendituresthan to calorie acquisition.
The education of the wife is stronglycorrelated with food expenditures, after con-trolling for incomes and nutrition education:the more educated the wife, the more sheis likely to spend on food for the household.Perhaps she purchases more expensive cal-ories, such as processed foods, because theopportunity cost of time spent on food prep-aration increases with education. Or perhapshigher education leads to a better under-standing of the importance of adequatenutrition. Evidence from Nicaragua and else-where indicates that women's schoolingplays a substantial role in family nutrition.25
Higher education of the husband does notappear to affect food expenditures or calorieacquisition significantly.
Participation in food assistance programssuch as those administered by Catholic Re-lief Services, CARE, and the World FoodProgramme significantly affected food ex-penditures and calorie acquisition amongthe study households (Appendix Table 34).The quantities received by households fromsuch programs are not included in the esti-mates of total food expenditures or calories,and the significant and negative sign demon-strates that food from these programs wassubstituted for food from other sources.
The regressions show that a large shareof total household income earned by womenis associated positively with food expendi-tures and negatively with calorie acquisition(Appendix Table 34). This is consistent with
25 Jere Behrman and Barbara Wolfe, "More Evidence on Nutrition Demand: Income Seems Overrated and Women'sSchooling Underemphasized," Journal of Development Economics 14 (January-February 1984): 105-128.
65
the hypothesis that households where theopportunity cost of women's time is hightend to purchase more expensive calories—such as processed foods—to reduce foodpreparation time.
The degree to which households supplytheir own food is partly captured by theOWNFARM and OWNGARDEN variables.These variables significantly and positivelyaffect calorie acquisition. The positive signon total calorie effects indicates that house-holds with a higher ratio of own-produce,all things being equal, tend to acquire morecalories. Under the definition the calorieacquisition term covers calories from bothpurchased and own-produced foods.
Effects on Consumption ofRice, Oil, Fish, and Maize
The subsidy component of the schemehad a positive and significant effect on theconsumption of rice, oil, and fish and a nega-tive effect on the consumption of maize(Table 27). The MPC for rice from totalhousehold income, excluding the subsidy,is estimated to be 88 calories or slightlymore than one-half of the MPC for totalcalories. The MPC from subsidy income is210 calories compared to 220 for totalcalories. Thus, the increase in total calorieconsumption brought about by the subsidyis almost totally accounted for by increasingrice consumption. The MPC for oil is 20calories from the subsidy and 8 calories fromother incomes.
The income elasticity is estimated to beabout 0.2 for rice, 0.4 for oil, and 0.5 forfish, while it is negative but nonsignificantfor maize. Own-price elasticities for rice andmaize are negative but not significantly dif-ferent from zero. Own-price elasticities foroil and fish are estimated to be about -0 .6and -0 .8 , respectively. Thus, price sub-sidies for extramarginal quantities of oilwould be expected to increase oil consump-tion by 6 percent for each 10 percent reduc-tion in its price. The nutrition educationcomponent of the scheme appears to havehad a positive and significant impact on riceconsumption.
Effects on CalorieConsumption by PreschoolChildren
As reported in Table 28 and in the Ap-pendix, Table 36, calorie demand equationsfor individual preschoolers yielded statisti-cally significant coefficients for the incomeand subsidy variables.
The scheme resulted in an additionaleffect on calories consumed by preschoolersover and above the effect operating throughincreases in the households' incomes andcalorie consumption. This may be partly ex-plained by the families' increased awarenessof the nutritional problems of children. Theregular monitoring of children's weights,conducted as part of the experiment, mayhave increased the mothers' awareness,especially since the nutrition educationcampaign stressed the importance of correctchild feeding practices.
According to the demand models, theimpact of the pilot scheme on daily calorieconsumption of preschoolers between ages13 and 83 months was 31-55 calories perchild—about 3 to 6 percent of the averagecalorie intake of preschoolers in the sample.These figures are below the average esti-mated for the household both in absoluteterms and relative to individual calorie re-quirements. For the household, the averageeffects per AEU were in the neighborhoodof 136-138 calories per day.
Effects on the NutritionalStatus of Children
Whether the calorie increases for pre-school children resulting from the subsidyscheme translated into child growth is acritical question from a policy standpoint.An increase in calorie consumption may beviewed not as a measure of the ultimateoutcome of policy, but rather as a measureof an intermediate outcome. This distinc-tion is often critical in measuring the impactof a program. Indexes of child health andnutritional well-being are considered bymany to be important indicators of currenteconomic welfare of developing countries.It is one set of measures of the quality of
66
Table 27—Key coefficients and parameters for estimating the impact of thepilot subsidy scheme on the acquisition of rice, maize, oil, and fish
Independent Variable
Log income per AEU
(including subsidy)
Dummy for subsidy
Price of rice
Price of maize
Price of oil
Price offish
Nutrition education
R2
Number of observationsMarginal propensity to consume0
All incomes (net of subsidy)Subsidy income
Consumption parametersIncome elasticity(net of subsidy)
Price elasticityRiceMaizeOilFish
Nutrition education effectsImpact of subsidy on acquisition
(calories per AEU per day)
Dependent Variable: Calorie Acquisition Per AEU Per Day
Rice
333.13(9.29)a
122.02(3.89)a
-5.81(-0.35)
7.89(0.65)
-10.66(-4.76)b
-3.90(-1.23)
78.14(2.28)a
0.152,131
88210
0.23
-0.013.0 x 10~4
-0.08-0.0278.14a
227
Maize
-32.67(-0.96)
-104.79(-3.53)a
33.87(2.15)a
-11.12(-0.97)
7.97(3.77)a
-5.51(-1.82)
14.92(0.58)
0.162,131
- 9-113
-0.18
0.55-0.004
0.49-0.2414.92
-108
Oil
28.55(6.50)a
12.22(3.18)a
-1.62(-0.79)
0.54(0.36)
-4.22(-15.41)a
-0.43(-1.09)
3.01(0.91)
0.232,131
820
0.38
-0.064.0 x 10~4
-0.62-0.05
3.01
15
Fish
39.64(6.14)a
52.40(9.27)a
4.10(1.37)
0.22(0.10)
-0.77(-1.90)"
-7.57H3.20)3
-4.27(-0.88)
0.342,131
1063
0.52
0.161.7 x 10~4
-0.11-0.78-4.27
56
Notes: The numbers in parentheses are t-values. The adult equivalent unit (AEU) is a method used to convertthe consumption of persons of different ages and sex into standard consumption units.
a Significant at the 5 percent level.b Significant at the 10 percent level.c Calories per AEU per peso.
life, which is itself a policy objective. A child'sgood health and nutritional status signifi-cantly influence his or her intelligence,health, and nutritional status as an adultand have a direct impact on adult produc-tivity and earnings.26
The current state of knowledge on howto measure the effects of an intervention
program on child nutritional status is quitepoor, partly because of the difficulty in mod-eling the interactions among biological,behavioral, cultural, environmental, andsocioeconomic factors that influence thenutritional status and health of children. Itis quite possible, for example, that an in-crease in calories may increase energy
26 Barbara Wolfe and Jere Behrman, "Determinants of Child Mortality, Health, and Nutrition in a DevelopingCountry," Journal of Development Economics 11 (No. 1, 1982): 163-193.
67
Table 28—Regressions of calorie intake of children aged 13 to 83 months
Independent Variable
Log income per AEU per day
Household calorie acquisitionper AEU per day
Subsidy
Price of rice
Price of cooking oil
Nutrition education
R2
Model A-1
140.39(3.64)'
11.34(3.64)'
68.60(1.91)b
-0.34(-0.54)
-55.97(-1.90)b
0.43
Dependent Variable:
Model A-2
129.05(3.35)'
100.27(3.31)'
45.87(1.37)
-0.25(-0.40)
-55.48(-1.88)"
0.43
Calorie Intake Per Day
Model B-l
0.10(3.34)'
9.87(2.87)'
63.36(1.76)b
-0.37(-0.58)
-62.14(2.10)'
0.43
Model B-2
0.09(3.02)'
88.35(2.87)'
43.71(1.31)
-0.28(-0.44)
-61.21(-2.07)b
0.43
Note: See the Appendix, Table 36, for the complete set of regressions. The numbers in parentheses are t-values.a Significant at the 5 percent level.b Significant at the 10 percent level.
expenditures and thereby improve the well-being of the child without changing anthro-pometry. There are also many measurementproblems associated with anthropometricvariables. In most cases, regression analysesbased on cross-sectional data give very lowR2, indicating an insubstantial knowledgeof factors that influence nutritional statusor an inability to quantify and incorporatethese factors into regression analysis.
The anthropometric data in this study—weight, height, and age of preschool chil-dren—were collected monthly from a panelof children from study households prior toand during the time the scheme was inforce.
As opposed to the various food-relatedmeasures of outcome used in calorie regres-sions, anthropometric indicators reflectboth food- and health-related factors. Thus,the statistical estimate of effect must be ableto control for factors such as age, sex, birthorder, breast-feeding, morbidity, and childcare, which, along with household variablessuch as family size, education of parents,and accessibility to clean water are expected
to influence the nutritional status of pre-schoolers.
A complete list of the explanatory vari-ables used in the regressions and the esti-mated coefficients is given in the Appendix,Tables 37 to 41. Table 29 shows the keycoefficients for estimating the impact of thescheme, using the five indicators of nutri-tional status as dependent variables.
Increases in household incomes appearto have a positive effect on all five an-thropometric indicators. This effect is highlysignificant for the short-term indicators,weight and weight-for-age. The subsidy doesnot seem to influence nutritional status overand above the effect of the income from thesubsidy. The nutrition education compo-nent of the scheme appears to have had apositive impact on both short- and long-termindicators.
As shown in the Appendix, Tables 37-41, a number of other variables appear toinfluence the nutritional status of preschool-ers. These include diarrhea in the recentpast, birth order, access to clean drinkingwater, and household size.
68
Table 29—Key coefficients for nutritional status regressions
Dependent Variable
WeightPercent of total income
Dummy
Weight as a percent ofstandard weight-foragePercent of total income
Dummy
Z-scores of weight-for-agePercent of total income
Dummy
Z-scores of weight-for-heightPercent of total income
Dummy
Height as a percent ofstandard height-foragePercent of total income
Dummy
Log Income PerAEUPerDay
0.68(3.33)a
0.68(3.28)a
3.78(2.77)a
3.70(2.70)a
0.42(2.43)a
0.41(2.35)a
0.24(1.40)
0.21(1.25)
1.11(1.38)
1.16(1.45)
Independent Variables
Subsidy
0.03(1.44)
-0.12(-0.76)
0.21(1.71)"
-0.70(-0.65)
0.03(1.97)a
-0.11(-0.79)
0.04(2.48)b
0.10(0.75)
-0.03(-0.37)
-0.82(-1.31)
Price ofRice
-0.04(-0.23)
-0.21(-1.17)
0.30(0.23)
-0.92(-0.77)
0.01(0.06)
-0.17(-1.12)
0.43(2.68)b
0.28(1.87)b
-2.03(-2.71 f
-2.16(-3.1 l)a
Price ofCooking Oil
46xlO^4
(1.37)
59xlO- 4
(1.77)b
0.02(1.08)
0.03(1.52)
87x 10~5
(0.31)
23x l0~ 4
(0.81)
11 x10~5
(-0.04)
91 x10^5
(0.33)
-0.02(-1.20)
-0.02(-1.33)
NutritionEducation
0.25(1.59)
0.21(1.30)
2.27(2.17)b
1.94(1.86)b
0.28(2.08)a
0.23(1.71)b
-0.17(-1.29)
-0.20(-1.51)
2.01(3.29)a
1.94(3.17)a
RegressionR2
0.72
0.72
0.14
0.14
0.12
0.12
0.11
0.10
0.16
0.17
Notes: See the Appendix, Table 36, for the complete set of regressions. The numbers in parentheses are t-values.a Significant at the 5 percent level.b Significant at the 10 percent level.
69
COST-EFFECTIVENESS ANALYSIS
Food price subsidy schemes are but oneof many instruments that governments useto alleviate calorie-protein deficiencies andto improve nutrition among poor households.In the Philippines, existing interventionsfocus on growth monitoring, extension ofnutrition education to mothers, and directsupplemental feeding of malnourished chil-dren. Several other schemes that are health-related (such as immunization, mother andchild health care, oral rehydration) andcommunity-based schemes, such as homeand community gardens, have been imple-mented with varying degrees of cost andeffectiveness.27 Although these interven-tions have reached a significant proportionof the high-risk population, their impact andrelative cost-effectiveness are not clear.
In this chapter the cost-effectiveness ofthe pilot subsidy scheme will be measuredin absolute terms and relative to othernutrition-related intervention programs orpolicies. Cost comparisons often suffer fromlack of adequate data, particularly regardingthe net effects of programs. Varying degreesof methodological rigor also make meaning-ful comparisons among the various types ofinterventions and programs difficult. For thesereasons, the following analysis deals mostlywith determining the cost-effectiveness ofthe scheme itself. Comparative analyseswith other nutrition-related interventionswill be included to the extent permitted byavailability of estimates of the effectivenessof other programs.
Program CostThe fiscal cost of operating the food price
subsidy scheme is given in Table 30. The
items reflected are, strictly speaking, gov-ernment financial or fiscal costs, as theyrefer to explicit budgetary costs. Economicor implicit costs of the subsidies and thetransaction costs incurred by beneficiariesare ignored.
The fiscal cost of the subsidy schemeconsists of three main components: the costof the price subsidy, the cost of the incentivepaid to retailers, and the cost of adminis-tration and management.
The cost of the price subsidy is com-puted as the price discount allowed for riceand cooking oil multiplied by the subsidizedquantity. The price discount is the differencebetween the market price and the subsidizedprice paid by the participating households.As shown in Table 30, this component ac-counted for about 84 percent of the totalfiscal costs. To compensate retailers foradded costs, the scheme offered an incen-tive of 7 percent of the gross sales of theprogram commodities to the accredited re-tailers. The reliance of the subsidy schemeon the private trade network for its procure-ment and distribution effectively saved thegovernment the costs of the provision ofwarehousing, distribution outlets, and ve-hicles. The administrative overhead coveredsalaries and benefits for the extension work-ers and paraprofessional workers, the costof travel of monitoring staff, salaries of cen-tral office project administration, officeoverhead, vehicles, and costs of printing thediscount cards. It is difficult to estimate thecost of the nutrition education componentseparately from the administrative overheadfor the subsidy operations because exten-sion staff members performed both as sub-sidy monitoring officers and as nutritioneducators. If person-hours were the mea-
27 Stewart Blumenfeld et al., PL 480 Title II: A Study of the Impact of Food Assistance Programs in the Philippines(Washington, D.C.: U.S. Agency for International Development, 1982).
70
Table 30—Fiscal costs of the pilot food price subsidy scheme and theirdistribution
Province/Subsidy
AbraRice and oil
AntiqueRice and oil
AntiqueOil
South CotabatoRice and oil
Average all areas
SubsidyCost
16,595(83.5)
18,958(83.9)
4,941(72.7)
13,861(82.8)
49,414(83.4)
Cost Per Month
Retailers'Incentive
Adminis-trativeCost
(pesos)
1,430(7.2)
1,633(7.2)
221(3.3)
1,195(7.1)
4,258(7.2)
1,856(9.3)
2,001(8.9)
1,634(24.0)
1,686(10.1)
5,543(9.4)
TotalProgram
Cost
19,881(100.0)
22,592(100.0)
6,796(100.0)
16,742(100.0)
59,215(100.0)
PilotPopulationCoverage
2,829
2,615
1,797
1,912
6,816
Fiscal Cost Per CapitaMonthly Annually
(pesos)
7.03
8.64
3.78
8.76
8.69
84.36
103.68
45.36
105.12
104.28
Annually
(U.S. $)
7.67
9.43
4.12
9.56
9.48
Source: Based on data collected by the International Food Policy Research Institute, Philippines National NutritionCouncil, and the Philippines Ministry of Agriculture, "Pilot Food Subsidy Survey, 1983/84," Philippines.
Note: The numbers in parentheses are percentages of the total cost.
sure, however, the extension staff spent 60percent of project time on subsidy monitor-ing and 40 percent for nutrition education.
Administrative costs accounted for 9.4percent of the total cost of the pilot scheme.It is likely that in a national program, severallayers of supervisory infrastructure, fromthe central offices down to the region, prov-ince, and village levels, would be necessary.This heavy supervisory infrastructure wouldlikely put pressure on the implementing andcoordinating agencies (in this case, theMinistry of Agriculture and the National Nu-trition Council). Such burdens are associatedwith larger fiscal outlays for salaries andoffice overhead. It is estimated that theseoutlays would add 25 percent to the admin-istrative costs shown in Table 30.
the different levels of targeting, and programcosts. Estimates of the net effects of thescheme given in Chapter 8 are summarizedin Table 31. It must be noted that theseestimates are computed at the mean for allparticipating households and children withinthese households. The leakage issues havebeen discussed in Chapter 5 and will notbe repeated here.
Cost-effectiveness indicators are devel-oped in this study for alternative programgoals. (Estimation procedures are describedin Chapter 5.) Therefore, a program designwith a goal to improve the nutrition of mal-nourished preschool children can be com-pared with a design where the goal is toimprove the nutrition of all household mem-bers.
Cost-Effectiveness MeasuresIn estimating cost-effectiveness, three
main factors are considered: the size of thenet program benefits to intended benefici-aries, the extent of leakage associated with
Fiscal Cost of TransferringU.S. $1.00
If the goal of the program is to efficientlytransfer purchasing power, the fiscal cost oftransferring U.S. $ 1.00 may be an appropri-
71
Table 31—Estimated net impact of the pilot subsidy scheme on foodacquisition and nutrition
Variable Model Estimated Net Impact"
HouseholdFood expenditure
Calorie acquisition
Individuals (preschoolers)Calorie intake
Weight
Percent of total incomeDummy
Percent of total incomeDummy
Percent of total incomeDummyPercent of total incomeDummy
P 0.27 per AEU per dayP 0.34 per AEU per day136 calories per AEU per day138 calories per AEU per day
31 calories per day55 calories per day
0.12 kilograms0.14 kilograms
' The net impact is computed at the mean.
ate indicator of cost-effectiveness. As dis-cussed in Chapter 5, this indicator is at besta crude measure of cost-effectiveness froma nutritional standpoint because severalsources of leakage are present.
The cost of transferring U.S. $1.00 toall households participating in the subsidyscheme is calculated at U.S. $1.19 in 1984prices (Table 32). This estimate assumesthat all participating households are the tar-get of the program. However, some reci-pient households are not nutritionally defi-cient. If the definition of the targets includesonly those households that consume lessthan 80 percent of the recommendedcalories, the fiscal cost of transferring U.S.$ 1.00 rises to U.S. $ 1.63 under the assump-tion that program benefits accruing tohouseholds above the 80 percent cutoff con-stitutes a leakage. It is estimated from theconsumption surveys that about 27 percentof the participating households were abovethe 80 percent cutoff line.
If the goal is to reach only those house-holds with malnourished preschool chil-dren, the cost of transferring U.S. $1.00increases by a factor of three because theextent of leakage from an untargeted pro-gram would be high. About one-third of thestudy households had at least one mal-nourished child. The rest of the householdsare considered nontarget under this as-sumption.
Fiscal Cost of IncreasingFood Acquisition by100 Calories per AEU per Day
If all households in the targeted villagesare defined as program beneficiaries, itwould cost U.S. $6.75 annually to increasethe daily caloric intake by 100 calories perAEU (Table 32). This estimate is based onnet impact, as it already accounts for sub-stitution effects (see the discussion in Chap-ter 8). In order to bring the average calorieacquisition of households in the study areasto recommended levels, it would cost thegovernment about U.S. $25 per AEU peryear.
If the program is intended only forhouseholds with calorie acquisition of lessthan 80 percent of the recommended levels,then the subsidy received by householdsabove this cutoff may be considered a leak-age in a scheme targeted by area. Given thesize of the leakage, it is estimated that itwould cost U.S. $7.40 to increase the calorieintake of households below the 80 percentcutoff by 100 calories per AEU per day forone year. If the program goal is to reachonly those households with malnourishedpreschool children, the annual fiscal costwould increase to U.S. $13.66 per AEU,inasmuch as two-thirds of the householdsdo not have malnourished preschool chil-
72
jI
1 •§
oq-6IN
• o i n
o
s";
73
dren; hence, benefits to such householdswould be considered leakage.
to well-nourished children and other house-hold members) is considerably higher.
Fiscal Cost of IncreasingPreschoolers' Daily CalorieConsumption by 100 Calories
The analysis has shown that preschoolchildren are at great nutritional risk becausetheir share of the family food basket is smallrelative to their RDA. Hence, it is importantto examine the range of cost required toraise their calorie intakes to a particular con-sumption level.
Based on the impact on calorie con-sumption for the preschool child derivedfrom the individual demand equations inChapter 5, it is estimated that it would costthe government U.S. $26 annually to in-crease the average daily intake of preschool-ers by 100 calories per day per preschooler.This figure is roughly four times the cost ofachieving the same level (100 calories) ofbenefit on average for all household mem-bers. The higher cost reflects the large shareof additional household food consumptionthat will be captured by members other thanpreschoolers. The average calorie incre-ment at the household level will also exceedthe increment accruing to preschoolerswithin the same household because the dis-tribution is uneven, and that will also bereflected in the cost.
Not all preschool children within thetargeted villages are equally at risk of mal-nutrition. Some are more predisposed to theproblem than others because of the varyingamounts of calories consumed. If thescheme's objective is to reach only thosechildren consuming less than 80 percent ofRDA, then the benefits accruing to childrenabove this cutoff are considered leakage.This translates to a higher cost: it is esti-mated that it would cost U.S. $45 per childto increase the daily calorie intake of pre-schoolers currently below the 80 percentcutoff by 100 calories for a one-year period.
If the program target is defined to in-clude only the malnourished children, thenthe cost escalates to U.S. $74 per 100 calo-ries because the leakage (benefits accruing
Fiscal Cost of Increasingthe Weight of Preschoolers
The expected effects of the scheme onchild weight are summarized in Table 31.A net increment of 0.12 kilograms was thecontribution of the scheme to the growthof children between 13-83 months. Suchan impact represents 1.0 percent of the av-erage body weight of the sample preschool-ers. These estimates are based on actualex-post data of a continuous and sustainedlevel of participation in the subsidy schemeamong the study children.
The incremental weight gain of childrenreported in Table 31 represents the averagenet impact produced by the scheme. Inorder to increase a child's weight by 1 kilo-gram, it would cost the government U.S.$76 annually. This assumes that all of thechildren in the study area are program tar-gets. If the goal is to improve the weight ofonly malnourished children, the fiscal costof increasing child weight by 1 kilogramincreases to U.S. $101, since benefits de-rived by children who are not malnourishedcan be considered leakage.
The average weight of the preschool chil-dren in the sample was 13.3 kilograms,which is about 2.0 kilograms below the stan-dard weight-for-age. Thus, the above increaseof 1 kilogram would reduce the average gapby about one-half.
Improving Cost-EffectivenessThe preceding sections have clearly de-
monstrated that cost-effectiveness hingeson the method of targeting. The geographic-area targeting scheme that was used in thepilot subsidy scheme provides benefits toall households in the target areas irrespec-tive of nutritional status and food deficiency.This implies that the cost-effectiveness ofthe scheme could be increased by targetingthose households expected to be high-risk.Once the area is identified, a second level
74
of screening of participants could targetthose households that need the subsidy.According to the cost-effectiveness indi-cators in Table 32, the cost of increasingthe acquisition of a given amount of foodby households with at least one mal-nourished preschooler could be reduced byhalf if the program were targeted to thosehouseholds only.
The use of child weight as a targetingcriterion may be feasible in the Philippinesetting because of the extensive child weigh-ing program carried out nationwide since1975, called Operation Timbang. The useof growth monitoring for targeting wouldentail the additional cost of periodic weigh-ing. However, this cost is very small andcan be disregarded in cost-effectiveness es-timates.
It is important to emphasize that target-ing a price subsidy scheme is practical onlyat the household level. The presence of amalnourished child may be used to identifytarget households, but efforts to target thebenefits of the scheme exclusively to theindividual child are not likely to be success-ful.
Cost-Effectiveness Relativeto Other Programs
It may be misleading to compare relativecost-effectiveness of alternative programsunless the context in which these programsare evaluated is based on a comparable setof objectives, targeting, coverage, timing ofinputs, and duration of program and projectorganization. Given the state of the art inthe evaluation of nutrition and health pro-grams, including the shortage of data, anycomparison, at best, can only detect generalindications of relative cost-effectiveness.
Relative cost-effectiveness should beevaluated in terms of particular programgoals. Hence, if the program goal is to de-liver a certain number of calories, the mea-sure of relative cost- effectiveness could givethe cost of transferring, say, 1,000 calories.
Alternatively, the measure could be givenin terms of the cost to deliver U.S. $ 1.00worth of subsidy if the concern is incometransfer efficiency. However, these are crudemeasures from the point of view of nutritionbecause they do not account for leakages(see Chapter 5).
A more refined measure is the fiscal costof a certain net increase in daily consump-tion by malnourished individuals. It should,however, be emphasized that in order toget meaningful results, calculations shouldbe based on the net calorie impact, that is,net of substitution effects. However, fewavailable program evaluations estimate sub-stitution.
Cost-effectiveness indicators for severalprograms are reported in Table 33. Theannual fiscal cost of a net increase in cal-orie consumption by 100 calories per dayis available as a net figure only for the Philip-pine pilot subsidy scheme and for the SriLankan food stamp program.28 The cost-effectiveness measures for other programsreported in Table 33 are estimated on thebasis of gross calorie estimates of impact, andthus they overestimate the actual increasein consumption. The costs reported are notstrictly comparable, although they givesome guidance on relative cost-effectiveness.Except for the Philippine pilot subsidy andSri Lankan food stamp program, the cost-effectiveness figures for the programs re-ported do not estimate leakage throughsharing and substitution.
The results indicate that the Philippinepilot subsidy scheme is relatively more cost-effective than any of the programs listed inTable 33. The food stamp program in SriLanka is almost as cost-effective, but all theother programs are considerably less cost-effective. The cost of a gross transfer of 100calories per person per day by any of theother programs is more than twice that ofthe Philippine program. However, theseresults must be interpreted in a proper con-text. Each of the programs cited has par-ticular goals and methods of targeting; eachuses different food commodities, and each
1 Edirisinghe, The Food Stamp Scheme in Sri Lanka.
75
1o
ou3S
I S !
HI
CO
I
00 00ro
A (ll (B nl
c d a d
CO
r>oCO
ro s o o ooOV1 M O
—< - o oq — ~•d 06 ro —• « r > oin ro ro M««O
" ^ § •-•£
.g
1•§•0
•3 <u e
1 2 1| | J | |
ffffs-i, O O f C 3
^ J N— ^ (i) O
a " s . ^ § oo
•2
76
is implemented on a different scale. Whileonly pilot programs have been implementedfor the Philippine, Colombian, and Brazilianschemes, the Sri Lankan scheme has beenin operation for a longer period on a nationalscale. The Tamil Nadu, India, program isnarrowly targeted to malnourished chil-dren, while some of the other programs arenot. Thus, the cost of transferring a certainamount of calories may not be an appropri-ate measure of the impact on nutrition.
As expected, the fiscal cost of transfer-ring $1.00 worth of subsidy is determinedlargely by the degree of targeting. The Philip-pine scheme, which uses geographical areatargeting, shows the lowest cost. The pre-school feeding programs in Brazil, Indonesia,and India had the highest cost because theseare the most tightly targeted among thosereported in Table 33.
Among the factors that contribute to the
relatively lower cost of delivery of the Philip-pine pilot subsidy scheme, the commoditymix is quite important. Besides being locallyproduced, rice and oil are the cheapestsource of calories in the country, and theyare available in almost all parts of the country.These foods are bought in raw form, unlikethe expensive processed weaning foods usedin the Colombian food subsidy program.
Another crucial factor is the substan-tially lower percentage of administrativecosts—about 9.4 percent of the total cost.These costs are low because targeting bygeographical area eliminates the burden-some and costly screening of beneficiariesfound in some of the other programs. Inaddition, the administrative costs in thePhilippine program are an add-on type ofcost because the delivery system is builtupon an already existing infrastructure ofextension officers.
77
10CONCLUSIONS
It is evident that the pilot subsidyscheme was successful in increasing foodconsumption among participating house-holds. Although distribution of the addi-tional food within the households favoredadults, preschool children consumed morecalories and showed improvement in theirnutritional status. Increases in food con-sumption came mostly from the increasesin purchasing power resulting from theprice subsidies on rice and cooking oil. Fur-thermore, the marginal propensity to spendon food out of the real income embodied inthe price subsidies appears to be considera-bly higher than the marginal propensity tospend on food out of other income. Thelower oil prices resulted in substitution ofoil for other commodities. Finally, the ef-fects of the nutrition education componentwere positive and strong for children butweak on the household level.
The scheme resulted in net increases incalories acquired by households of 136-138calories per AEU per day, which is about 7percent of current calorie consumption;calorie consumption by preschoolers of 31 -55 calories per child per day; and the weightof preschoolers of 0.12-0.14 kilograms.Eighty-four percent of the cost of the schemewas the subsidy itself. Administrative costsaccounted for about 9 percent and the in-centive payment to retailers to assure effi-cient distribution of subsidized food wasabout 7 percent.
The results of the assessment of cost-effectiveness can be summarized as follows.The fiscal cost of each U.S. $1.00 trans-ferred to participating households is esti-mated to be U.S. $1.19. However, if onlythe transfers received by households withmalnourished preschoolers are considereda benefit, whereas transfers received byother households are considered leakage,the cost increases to U.S. $3.61. Similarly,the annual cost of a net increase in calorie
consumption of 100 calories per AEU perday among all households is estimated tobe U.S. $6.75 per AEU, increasing to U.S.$13.66 if only food received by householdswith malnourished preschoolers is consid-ered of interest to the scheme.
The annual cost of eliminating caloriedeficiencies in the study population is esti-mated to be U.S. $25 per AEU. Adding 1kilogram to the weight of each preschooleris estimated to cost U.S. $76 per year. Ifonly weight gains among the malnourishedare counted as benefits, the cost increasesto U.S. $101 per year.
Comparisons with other programs showthat the cost-effectiveness of the Philippinepilot subsidy scheme was extremely favor-able, primarily because costs were kept lowthrough geographical targeting based ongrowth monitoring, the use of existing pri-vate sector retail outlets for the distributionof subsidized foods, and the use and expan-sion of existing local bureaucratic struc-tures.
If the sole goal of the scheme were toexpand food consumption by households withmalnourished preschoolers and to improvethe nutritional status of these preschoolers,the cost-effectiveness of the scheme wouldbe significantly improved by a two-step tar-geting procedure based on growth monitor-ing. The first step would be to identify targetvillages with a high concentration of under-weight preschoolers. This was the approachused in this scheme. The second step wouldbe to target individual households in theselected villages and to remove from thescheme those households that do not includepreschoolers at high nutritional risk. Accord-ing to the estimates here, a second targetingstep would reduce the cost of providingbenefits to high-risk preschoolers to less thanone-half the cost of the scheme when un-targeted within the village.
Low calorie adequacy rates among pre-
78
schoolers, both in absolute terms and relativeto the adequacy rates of adults, clearly justifyfocusing on this age group. However, adequ-acy rates of school-age children and adoles-cents were almost as low. Other studieshave shown similar results.29 If this findingis of general validity throughout the country,it raises serious questions about the useful-ness of targeting nutrition programs to pre-schoolers alone. It might be more appropri-ate to use a two-step targeting procedurebased on growth monitoring of all children,not just preschoolers.
A strong relationship between malnutri-tion and incomes is found in this study.Certain low-income occupational groups—landless farm workers, some wage earners,and tenant farmers, for example—are muchmore likely to be malnourished than others.This is primarily because their purchasingpower is limited. Efforts to expand food con-sumption and improve the nutritional statusof preschoolers in those groups through
nutrition education are unlikely to be suc-cessful unless accompanied by expandedpurchasing power.
Although nutrition education clearlyplays a role in expanding food consumptionby preschoolers, such expansion is morelikely to come about as part of an overallincrease in household food consumption.Efforts to reallocate an inadequate amount offood among household members so that pre-schoolers receive a larger share are unlikelyto be successful in households operatingunder severe income constraints. Further-more, such reallocation may endanger thesurvival of the household as a whole by fur-ther eroding the income-earning capabilitiesof economically active adult members whomust have enough energy to work produc-tively. However, as this study shows, nutri-tion education may be effective in assuringthat a larger share of additional income isspent on food for the household membersmost at risk of malnutrition.
29 Valenzuela, "Nutrient Distribution Within the Family"; and Aligaen and Florencio, "Intra-Household NutrientDistribution."
79
APPENDIX: SUPPLEMENTARY TABLES
Table 34—Regression results for household food expenditures and foodacquisition
Household Food Expenditures Household Calorie AcquisitionPerAEUPerDay PerAEUPerDay
Independent Variable
Intercept
Log income per AEU(including subsidy)3
Subsidy in percent ofincome
Subsidy dummy
Log of household size
Education of husband
Education of wife
Nutrition education dummy
PNPb
Pilotc
Food assistance dummy
Percent women's income
Price of rice
Price of oil
Dummy for Antique Province
Dummy for South CotabatoProvince
Own-farm dummy
Own-garden dummy
F
R2
Model A-1
0.151(1.39)
2.202(57.82)
0.032(7.30)
-0.051(-7.83)
0.002(0.49)
0.017(3.63)
0.023(0.63)
-0.021(-0.73)
0.056(1.86)
0.001(1.81)
0.006(0.37)
0.012(4.11)
-0.059(-1.47)
-0.135(-3.75)
0.013(0.40)
-0.035(-1.06)
378.65
0.73
Model A-2
0.255(2.37)
2.167(56.72)
0.168(4.96)
-0.054(-7.70)
0.0004(0.08)
0.016(3.24)
0.062(1.68)
-0.025(-0.86)
0.052(1.73)
0.001(1.58)
0.006(0.34)
0.004(1.49)
-0.069(-1.71)
-0.110(-3.05)
0.009(0.28)
-0.009(-0.28)
371.77
0.72
Model B-l
984.29(14.89)
633.95(27.38)
8.94(3.32)
-5.92(-1.40)
-5.52)(-1.97)
-1.92(-0.65)
-13.68(-0.61)
22.67(1.29)
39.45(2.16)
-1.03(-2.50)
-12.11(-1.12)
-5 .98(-3.51)
248.40(10.18)
26.96(1.23)
98.46(5.02)
27.50(1.36)
87.19
0.38
Model B-2
998.38(15.35)
623.07(27.01)
68.23(3.35)
-5 .98(-1.41)
-6.07(-2.17)
-2.46(-0.83)
-3.76(-0.17)
21.86(1.24)
36.26(1.98)
-1.06(-2.56)
-12.23(-1.13)
-7.45(-5.14)
245.04(10.04)
33.06(1.52)
98.79(5.03)
33.17(1.66)
87.21
0.38
Notes: Figures in parentheses are tratios. The number of observations for each model was 2,509. Models A-land B-l specify the subsidy variable in terms of the proportion of subsidy to total income; Models A-2and B-2 specify the subsidy variable in terms of a zero-one dummy for the presence of subsidy.
a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy study.
80
Table 35—Regression results for acquisition of rice, maize, oil, and fish
Independent Variable
Intercept
Log income per AEU(including subsidy)11
Subsidy dummy
Log of household size
Education of husband
Education of wife
Nutrition education dummy
PNPb
Pilotc
Food assistance dummy
Percent women's income
Price of rice
Price of maize
Price of oil
Price of fish
Dummy for Antique Province
Dummy for South CotabatoProvince
Own-farm dummy
Own-garden dummy
F
R2
Rice
785.36(6.28)
333.13(9.29)
122.02(3.89)87.08(2.03)
-8.94(-2.07)-6.50(-1.43)
78.14(2.88)17.50(0.51)
105.44(3.73)
-0.59(-0.92)-5.81(-0.35)
7.89(0.65)
-10.66(-4.76)-3.90(-1.23)331.52
(8.04)
62.12(1.83)73.92(2.45)
13.15(0.43)21.56
0.15
Maize Oil
(calories/AEU/day)
319.63(2.70)
-32.67(-0.96)
-104.79(-3.53)-6.23(-0.15)-1.71
(-0.42)
2.02(0.47)
14.92(0.58)
100.36(3.10)
-49.04(-1.83)-1.22
(-2.04)33.87(2.15)
-11.12(-0.97)
7.97(3.77)
-5.51(-1.83)
-446.72(-11.46)
-485.15(-15.14)
80.98(2.83)61.95(2.11)
23.41
0.16
112.72(7.37)
28.55(6.50)12.22(3.18)
-24.97(-4.76)
0.51(0.97)0.95
(1.71)
3.01(0.91)7.33
(1.75)-2.63(-0.76)-0.01(-0.16)-1.62
(-0.79)
0.54(0.36)
-4.22(-15.41)
-0.43(-1.09)-0.11(-0.02)
10.99(2.65)
-1.65(-0.45)-3.16(-0.84)36.95
0.23
Fish
82.13(3.65)
39.64(6.14)52.40(9.27)
-40.08(-5.19)-2.51(-3.22)
1.11(1.36)
-4.27(-0.88)
8.88(1.44)8.41
(1.65)-0.22(-1.90)
4.10(1.37)0.22(0.10)0.77(1.90)
-7.57(-13.20)101.19(13.63)
9.22(1.51)
-7.07(-1.30)
-4.89(-0.88)
65.01
0.34
Note: Figures in parentheses are t-ratios. The number of observations was 2,132.a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy study.
81
Table 36—Regression results for calorie intake of children aged 13 to 83months
Independent Variable
Intercept
Log income per AEU(including subsidy)*1
Subsidy in percent of income
Subsidy dummy
Education of husband
Education of wife
Sex of child
Log age in months
Currently breastfed dummy
Past breastfed dummy
Nutrition education dummy
PNPb
Pilotc
Child care time in hoursofwife
Diarrhea dummy
Fever dummy
Food assistance dummy
Percent children below6 years
Dummy for Antique Province
Dummy for South CotabatoProvince
Birth order
Water quality dummy
Household size
Model A-1
-948.70(-4.78)
140.39(3.64)11.34(3.31)
-9.74(-2.39)-4.52(-0.97)122.02
(4.50)
395.59(12.04)70.54(1.72)
-75.53(-1.91)
-42.74(-1.19)
-55.97(-1.90)
-0.36(-0.30)
-38.44(-0.86)-79.91(-2.50)20.69(0.72)
1.47(1.41)
211.60(4.73)
-37.98(-1.06)
-19.34(-1.15)55.24(1.81)
-1.95(0.11)
Model A-2
-843.35(-4.40)
129.05(3.35)
100.27(3.31)
-10.54(-2.58)-6.50(-1.39)124.95
(4.61)391.57(11.88)65.58(1.60)
-71.10(-1.81)
-45.75(-1.27)
-55.48(-1.88)
-0.21(-0.17)
-29.93(-0.67)
-68.32(-2.15)20.14(0.70)
1.00(0.97)
207.82(4.65)
-29.04(-0.81)
-23.82(-1.41)43.06(1.40)
-4.05(-0.23)
Model B-l
849.39(-4.41)
9.87(2.87)
. . .
-8.23(-2.04)-4.16(-0.89)122.49
(4.51)395.70(12.01)64.81(1.58)
-72.90(-1.84)
-36.16(-1.01)
-62.14(-2.10)
0.16(0.13)
-41.55(-0.92)
-67.55(-2.12)
4.97(0.17)
1.31(1.25)
168.44(3.79)
-52.03(-1.45)
-16.43(-0.97)60.94(2.00)
-10.19(-0.59)
Model B-2
-761.75(-4.07)
88.35(2.87)
-9.03(-2.23)-5.88(-1.27)125.06
(4.61)
392.18(11.87)60.78(1.48)
-69.33(-1.76)
-39.20(-1.09)
-61.21(-2.07)
0.25(0.21)
-33.75(-0.75)
-58.18(-1.84)
5.71(0.20)
0.91(0.88)
168.41(3.79)
-43.13(-1.21)
-20.61H.22)49.91(1.63)
-7.50(-0.43)
(continued)
82
Table 36—Continued
Independent Variable
Percent women's income
Price of rice
Price of cooking oil
Household calorie acquisition
FR2
Model A-1
0.02(0.02)
-68.60(-1.91)
-0.34(-0.54)
15.84
0.43
Model A-2
-0.13(-0.14)
-45.87(-1.37)
-0.25(-0.40)
15.84
0.43
Model B-l
0.10(0.11)
-63.36(-1.76)-0.36(-0.58)
0.10(3.34)
15.69
0.43
Model B-2
-0.04(-0.04)
-43.71(-1.31)-0.28(-0.44)
0.09(3.02)
15.68
0.43
Notes: Figures in parentheses are t-ratios. The number of observations for each model was 2,509. Models A-land B-l specify the subsidy variable in terms of the proportion of subsidy to total income; Models A-2and B-2 specify the subsidy variable in terms of a zero-one dummy for the presence of subsidy.
a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy study.
Table 37—Regression results for weight of children aged 13 to 83 months
Independent VariableProportional
ModelZero-One
Dummy Model
Intercept
Log income per AEU (including subsidy)3
Subsidy in percent of income
Subsidy dummy
Education of husband
Education of wife
Sex of child
Log age in months
Currently breastfed dummy
Past breastfed dummy
Nutrition education dummyPNPb
Pilotc
-5.80(-5.49)
0.68(3.33)
0.03(1.44)
0.0024(0.11)
0.03(1.04)
1.01(7.00)
4.41(25.11)
-0.19(-0.86)-0.51(-2.42)
0.02(0.09)
0.25(1.59)
-5.27(-5.15)
0.68(3.28)
-0.13(-0.76)
-0.0041(0.19)
0.03(1.04)
1.02(7.05)
4.44(25.16)
-0.21(-0.93)-0.43(-2.04)
0.05(0.25)
0.21(1.30)
(continued)
83
Table 37—Continued
Independent VariableProportional
Model
0.004(0.06)
-0.47(-1.96)
-0.16(-0.91)
0.0029(0.02)
-0.0037(-0.66)
-0.36(-1.52)
-0.07(-0.39)
-0.35(-3.89)
-0.48(-2.96)
-0.44(-4.73)
-0.04(-0.23)
-0.0046(-1.37)
56.51
0.72
Zero-OneDummy Model
-0.00005(-0.08)
-0.48(-2.01)
-0.12(-0.73)
0.04(0.25)
-0.0051(-0.92)
-0.37(-1.57)
-0.06(-0.30)
-0.35(-3.88)
-0.47(-2.83)
-0.44(-4.62)
-0.21(-1.17)-0.0059(-1.77)
56.27
0.72
Child care time in hours of wife
Diarrhea dummy
Fever dummy
Food assistance dummy
Percent children below 6 years
Dummy for Antique Province
Dummy for South Cotabato Province
Birth order
Water quality dummy
Household size
Price of rice
Price of cooking oil
F
R2
Notes: Figures in parentheses are t-ratios. The number of observations was 509. The proportional model specifiesthe subsidy variable as a proportion of the subsidy to total income, whereas the dummy model specifiesone for the presence of the subsidy, zero otherwise.
a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy scheme.
Table 38—Regression results for weight as a percent of standard weight-for-age, children aged 13 to 83 months
Independent Variable
Intercept
Log income per AEU (including subsidyf
Subsidy in percent of income
Subsidy dummy
84
ProportionalModel
87.35(12.45)
3.78(2.77)
0.21(1.71)
Zero-OneDummy Model
91.34(13.42)
3.70(2.70)
-0.70(-0.65)
(continued)
Table 38—Continued
Independent VariableProportional
ModelZero-One
Dummy Model
Education of husband
Education of wife
Sex of child
Log age in months
Currently breastfed dummy
Past breastfed dummy
Nutrition education dummyPNPb
Pilotc
Child care time in hours of wife
Diarrhea dummy
Fever dummy
Food assistance dummy
Percent children below 6 years
Dummy for Antique Province
Dummy for South Cotabato Province
Birth order
Water quality dummy
Household size
Price of rice
Price of cooking oil
F
R2
0.04(0.25)
0.22(1.34)
1.77(1.84)
-5.47(-4.69)
0.40(0.27)
-4.38(-3.13)
0.20(0.16)
2.26(2.17)
0.0076(0.18)
-3.35(-2.10)
-0.54(-0.47)
0.04(0.04)
-0.002(-0.05)
-2.39(-1.53)
-0.02(-0.01)
-1.91(-3.19)
-2.94(-2.71)
-2.60(-4.17)
-0.30(-0.23)
-0.02(-1.08)
3.66
0.14
0.05(0.33)
0.22(1.31)
1.83(1.90)
-5.27(-4.49)
0.27(0.18)
-3.80(-2.72)
0.41(0.32)
1.94(1.85)
0.0012(0.03)
-3.44(-2.15)
-0.29(-0.26)
0.29(0.29)
-0.01(-0.35)
-2.49(-1.59)
0.12(0.10)
-1.91(-3.19)
-2.84(-2.59)
-2.54(-4.06)
-0.92(-0.77)
-0.03(-1.52)
3.52
0.14
Notes: Figures in parentheses are tratios. The number of observations was 509. The proportional model specifiesthe subsidy variable as a proportion of the subsidy to total Income, whereas the dummy model specifiesone for the presence of subsidy, zero otherwise.
a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy scheme.
85
Table 39—Regression results for height as a percent of standard height-for-age, children aged 13 to 83 months
Independent Variable
Intercept
Log income per AEU (including subsidy)8
Subsidy in percent of income
Subsidy dummy
Education of husband
Education of wife
Sex of child
Log age in months
Currently breastfed dummy
Past breastfed dummy
Nutrition education dummyPNPb
Pilot0
Child care time in hours of wife
Diarrhea dummy
Fever dummy
Food assistance dummy
Percent children below 6 years
Dummy for Antique Province
Dummy for South Cotabato Province
Birth order
Water quality dummy
Household size
Price of rice
ProportionalModel
100.56(24.42)
1.11(1.38)
-0.03(-0.37)
0.13(1.50)
0.09(0.97)
0.47(0.83)
-2.29(-3.35)
-0.53(-0.63)
-1.51(-1.84)
0.03(0.04)
2.01(3.29)
0.01(0.44)
-1.49(-1.60)
0.19(0.28)
-1.91(-3.21)
0.04(1.66)
0.54(0.59)
1.38(1.87)
-0.50(-1.43)
-3.07(-4.84)
-0.78(-2.12)
-2.03(-2.71)
Zero-OneDummy Model
100.81(25.35)
1.16(1.45)
-0.82(-1.31)
0.13(1.59)
0.11(1.09)
0.46(0.82)
-2.22(-3.23)-0.53(-0.63)
-1.40(-1.72)
0.10(0.13)
1.94(3.17)
0.008(0.33)
-1.57(-1.68)
0.17(0.25)
-1.85(-3.10)
0.04(1.71)
0.53(0.58)
1.34(1.83)
-0.47(-1.35)
-2.97(-4.65)-0.75(-2.05)
-2.16(-3.11)
(continued)
86
Table 39—Continued
Independent VariableProportional
ModelZero-One
Dummy Model
Price of cooking oil
F
R2
- 0 . 0 2-1 .20)
4.32
0.16
-0.02(-1.33)
4.41
0.17
Notes: Figures in parentheses are t-ratios. The number of observations was 509. The proportional model specifiesthe subsidy variable as a proportion of the subsidy to total income, whereas the dummy model specifiesone for the presence of the subsidy, zero otherwise.
a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy scheme.
Table 40—Regression results for z-scores for weight-for-age, children aged13 to 83 months
Independent VariableProportional
ModelZero-One
Dummy Model
Intercept
Log income per AEU (including subsidy)8
Subsidy in percent of income
Subsidy dummy
Education of husband
Education of wife
Sex of child
Log age in months
Currently breastfed dummy
Past breastfed dummy
Nutrition education dummy
PNPb
Pilotc
Child care time in hours of wife
Diarrhea dummy
-1.03(-1.15)
0.42(2.42)
0.03(1.97)
0.0099(0.54)
0.01(0.02)
0.30(2.46)
-0.60(-4.05)
-0.04(-0.24)
-0.48(-2.68)
-0.02(-0.12)
0.28(2.08)
0.0069(1.26)
-0.34(-1.66)
-0.44(-0.51)
0.41(2.35)
-0.11(-0.79)
0.01(0.62)
0.01(0.64)
0.31(2.52)
-0.57(-3.82)
-0.06(-0.34)
-0.39(-2.20)
0.01(0.08)0.23
(1.71)0.0059(1.07)
-0.35(-1.71)
(continued)
87
Table 40—Continued
Independent VariableProportional
Model
-0.10(-0.69)
0.11(0.89)
-0.004(-0.77)
-0.20(-1.00)
0.19(1.17)
-0.21(-2.89)
-0.40(-2.89)
-0.29(-3.60)
-0.010(-0.006)
-0.0009(-0.31)
3.15
0.12
Zero-OneDummy Model
-0.06(-0.44)
0.15(1.18)
-0.005(-1.11)
-0.21(-1.07)
0.21(1.30)
-0.22(-2.80)
-0.39(-2.75)
-0.28(-3.47)-0.17(-1.12)
-0.0023(-0.81)
2.98
0.12
Fever dummy
Food assistance dummy
Percent children below 6 years
Dummy for Antique Province
Dummy for South Cotabato Province
Birth order
Water quality dummy
Household size
Price of rice
Price of cooking oil
F
R2
Notes: Figures in parentheses are t-ratios. The number of observations was 509. The proportional model specifiesthe subsidy variable as a proportion of the subsidy to total income, whereas the dummy model specifiesone for the presence of the subsidy, zero otherwise.
a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy scheme.
Table 41—Regression results for z-scores for weight-for-height, children aged13 to 83 months
Independent Variable
Intercept
Log income per AEU (including subsidy)8
Subsidy in percent of income
Subsidy dummy
Education of husband
Education of wife
ProportionalModel
-2.00(-2.27)
0.24(1.40)
0.04(2.48)
-0.02(-0.95)
0.0076(0.37)
Zero-OneDummy Model
-1.46(-1.71)
0.21(1.25)
0.10(0.75)
-0.02(-0.97)
0.004(0.20)
(continued)
88
Table 41—Continued
Independent VariableProportional
ModelZero-One
Dummy Model
Sex of child
Log age in months
Currently breastfed dummy
Past breastfed dummy
Nutrition education dummyPNPb
Pilot0
Child care time in hours of wife
Diarrhea dummy
Fever dummy
Food assistance dummy
Percent children below 6 years
Dummy for Antique Province
Dummy for South Cotabato Province
Birth order
Water quality dummy
Household size
Price of rice
Price of cooking oil
F
R2
0.14(1.21)
-0.09(-0.63)
0.10(0.53)
-0.18(-1.04)
0.05(0.29)
-0.17(-1.29)
-0.0029(-0.54)
-0.13(-0.63)
-0.13(-0.91)
0.39(3.03)
-0.0079(-1.70)
-0.52(-2.64)
-0.29(-1.87)
-0.16(-2.13)
0.26(1.92)
-0.18(-2.36)
-0.43(-2.68)
-0.0001(-0.04)
2.87
0.11
0.16(1.29)
-0.08(-0.55)
0.08(0.42)
-0.13(-0.70)
0.06(0.38)
-0.20(-1.51)
-0.003(-0.59)
-0.12(-0.60)
-0.09(-0.61)
0.41(3.18)
-0.01(-2.10)
-0.53(-2.70)
-0.27(-1.68)
-0.17(-2.21)
0.25(1.82)
-0.18(-2.31)
-0.28(-1.87)
-0.0009(-0.33)
2.58
0.10
Notes: Figures in parentheses are tratios. The number of observations was 509. The proportional model specifiesthe subsidy variable as a proportion of the subsidy to total income, whereas the dummy model specifiesone for the presence of the subsidy, zero otherwise.
a Income is approximated by total expenditures in all regressions.b This dummy is for the Philippine Nutrition Program (PNP), a government program begun in 1974.c This dummy is for the nutrition education program held in conjunction with the pilot food price subsidy scheme.
89
BIBLIOGRAPHY
Alderman, Harold. The Effect of Food Price and Income Changes on the Acquisition ofFood by Low-Income Households. Washington, D.C.: International Food Policy Re-search Institute, 1986.
Alderman, Harold, and von Braun, Joachim. The Effects of the Egyptian Food Ration andSubsidy System on Income Distribution and Consumption. Research Report 45.Washington, D.C.: International Food Policy Research Institute, 1984.
Aligaen, Melba B., and Florencio, Cecilia B. "Intra-Household Nutrient Distribution andAdequacy of Food and Nutrition Intake of Filipino Urban Households." PhilippineJournal of Nutrition (January-March 1980): 11-19.
Behrman, Jere, and Wolfe, Barbara. "More Evidence on Nutrition Demand: Income SeemsOverrated and Women's Schooling Underemphasized." Journal of DevelopmentEconomics 14 (January-February 1984): 105-128.
Benus, J; Kmenta, J.; and Shapiro, H. "The Dynamics of Household Budget Allocation toFood Expenditures." Review of Economics and Statistics 58 (No. 2,1976): 129-138.
Blumenfeld, Stewart; Flores, Maria B.; Mack, M.; Norton; and Pooley. PL 480 Title II: AStudy of the Impact of Food Assistance Programs in the Philippines. Washington,D.C.: U.S. Agency for International Development, 1982.
Bouis, Howarth. "Rice Policy in the Philippines." Ph.D. dissertation, Stanford University,1982.
Carloni, Alice S. "Sex Disparities in the Distribution of Food Within Rural Households."Food and Nutrition 7 (1981): 3-12.
Deaton, Angus, and Muellbauer, John. Economics and Consumer Behavior. New York:Cambridge University Press, 1980.
Edirisinghe, Neville. The Food Stamp Scheme in Sri Lanka: Costs, Benefits, and Optionsfor Modification. Research Report 58. Washington, D.C.: International Food PolicyResearch Institute, 1987.
Evenson, Robert E.; Popkin, Barry M.; and Quizon, Elizabeth K. "Nutrition, Work andDemographic Behavior in Rural Philippine Households." In Rural Household Studiesin Asia. Edited by Hans P. Binswanger, Robert Evenson, Cecilia A. Florencio, andBenjamin White. Singapore: Singapore University Press, 1980.
Fraker, Thomas; Devaney, Barbara; and Cavin, Edward. "An Evaluation of the Effect ofCashing Out Food Stamps on Food Expenditures." American Economic Review 76(May 1986): 230-234.
Friedman, Milton. A Theory of the Consumption Function. Princeton: Princeton UniversityPress, 1957.
Hassan, Nazmul, and Ahmad, Kamaluddin. "Intra-Familial Distribution of Food in RuralBangladesh." Food and Nutrition Bulletin 6 (No. 4, 1984): 34-42.
Hymans, S., and Shapiro, H. "The Allocation of Household Income to Food Consumption."Five Thousand American Families—Patterns of Economic Progress, Vol. II. Editedbyj . Morgan. Ann Arbor: University of Michigan Press, 1975.
90
Knudsen, Odin K., and Scandizzo, Pasquale L. Nutrition and Food Needs in DevelopingCountries. Staff Working Paper 328. Washington, D.C.: World Bank, 1979.
Kumar, Shubh. The Impact of Subsidized Rice on Food Consumption and Nutrition inKerala. Research Report 5. Washington, D.C.: International Food Policy ResearchInstitute, 1979.
Mellor, John W. "Food Aid and Nutrition." American Journal of Agricultural Economics62 (December 1980): 979-983.
Philippines, Food and Nutrition Research Institute. Publication No. 75. Manila: FNRI, 1977.
Philippines, Food and Nutrition Research Institute, National Science and Technology Author-ity. Second Nationwide Nutrition Survey, Philippines, 1982. Manila: NSTA, 1984.
Philippines, Food and Nutrition Research Institute and National Nutrition Council. Recom-mended Dietary Allowances for Filipinos. Manila: FNRI, 1976.
Philippines, Ministry of Agriculture, National Nutrition Council and the National Economicand Development Authority. The Philippine Food and Nutrition Plan for the 80s.Manila: Ministry of Agriculture, 1980.
Philippines, National Economic and Development Authority. Philippine Statistical Yearbook.Manila: NEDA, 1985.
Philippines, National Nutrition Council, Nutrition Research Committee. Philippine Recom-mended Dietary Allowances, Part I, Nutrients. Manila: National Nutrition Council,1976.
Pimoz, E., and Viteri, F. "Understanding Health and Nutrition Behavior Through the Studyof Household Decision-Making and Intra-Household Resource Distribution."Washington, D.C.: Pan-American Health Organization, 1983.
Pinstrup-Andersen, Per. "An Analytical Framework for Assessing Nutritional Effects ofPolicies and Programs." In Food Policy. Edited by Charles K. Mann and BarbaraHuddleston. Bloomington: Indiana University Press, 1985.
Pinstrup-Andersen, Per, ed. Consumer-Oriented Food Subsidies: Costs, Benefits, andPolicyOptions for Developing Countries. Baltimore: Johns Hopkins University Press forInternational Food Policy Research Institute, forthcoming.
Prais, S. J., and Houthakker, H. S. The Analysis of Family Budgets. New York: CambridgeUniversity Press, 1971.
Reutlinger, Shlomo, and Katona-Apte, Judith. The Nutritional Impact of Food Aid: Criteriafor the Selection of Cost-Effective Foods. Washington, D.C.: World Bank, 1984.
Sahn, David E. "Malnutrition and Food Consumption in Sri Lanka: An Analysis of Changesfrom 1969 to 1982." International Food Policy Research Institute, Washington,D.C., 1986 (mimeographed).
Senauer, Ben; Garcia, Marito; and Jacinto, Elizabeth. "Determinants of IntrahouseholdAllocation of Food in Rural Philippines." American Journal of Agricultural Economics(forthcoming, February 1988).
Senauer, Ben, and Young, Nathan. "The Impact of Food Stamps on Expenditures: Rejectionof the Traditional Model." American Journal ofAgricultural Economics 68 (February1986): 37-43.
91
Silverio, Simeon G. The Neighborhood Sari-Sari Store. Institute of Philippine Culture PovertySeries 2. Quezon City, Philippines: Ateneo de Manila University, 1975.
Trairatvorakul, Prasarn. The Effects on Income Distribution and Nutrition of AlternativeRice Price Policies in Thailand. Research Report 46. Washington, D.C.: InternationalFood Policy Research Institute, 1984.
Valenzuela, Rosario. "A Study on Nutrient Distribution within the Family and FactorsAffecting Nutrient Intake." University of the Philippines, Diliman, 1977 (mimeog-raphed).
Valenzuela, Rosario; Florencio, Cecilia; and Guthrie, H. "Distribution of Nutrients Withinthe Family." Nutrition Report International 19 (No. 4, 1979).
West, Donald A., and Price, David W. "The Effects of Income, Assets, Food Programs, andHousehold Size on Food Consumption." American Journal of Agricultural Economics58 (1976): 725-730.
Wolfe, Barbara, and Behrman, Jere. "Determinants of Child Mortality, Health, and Nutritionin a Developing Country." Journal of Development Economics \ 1 (1982): 163-193.
92
RECENT IFPRI RESEARCH REPORTS (continued)
41 RURAL GROWTH LINKAGES: HOUSEHOLD EXPENDITURE PATTERNS IN MALAYSIA ANDNIGERIA, September 1983, by Peter B. R. Hazell and Ailsa Roell
40 FOOD SUBSIDIES IN EGYPT: THEIR IMPACT ON FOREIGN EXCHANGE AND TRADE, August1983, by Grant M. Scobie
39 THE WORLD RICE MARKET: STRUCTURE, CONDUCT, AND PERFORMANCE, June 1983, byAmmar Siamwalla and Stephen Haykin
38 POLICY MODELING OF A DUAL GRAIN MARKET: THE CASE OF WHEAT IN INDIA, May1983, by Raj Krishna and Ajay Chhibber
37 SERVICE PROVISION AND RURAL DEVELOPMENT IN INDIA: A STUDY OF MIRYALGUDATALUKA, February 1983, by Sudhir Wanmali
36 AGRICULTURE AND ECONOMIC GROWTH IN AN OPEN ECONOMY: THE CASE OF ARGEN-TINA, December 1982, by Domingo Cavallo and Yair Mundlak
3 5 POLICY OPTIONS FOR THE GRAIN ECONOMY OF THE EUROPEAN COMMUNITY: IMPLICA-TIONS FOR DEVELOPING COUNTRIES, November 1982, by Ulrich Koester
34 EGYPT'S FOOD SUBSIDY AND RATIONING SYSTEM: A DESCRIPTION, October 1982, byHarold Alderman, Joachim von Braun, and Sakr Ahmed Sakr
33 AGRICULTURAL GROWTH AND INDUSTRIAL PERFORMANCE IN INDIA, October 1982, byC. Rangarajan
32 FOOD CONSUMPTION PARAMETERS FOR BRAZIL AND THEIR APPLICATION TO FOODPOLICY, September 1982, by Cheryl Williamson Gray
31 SUSTAINING RAPID GROWTH IN INDIA'S FERTILIZER CONSUMPTION: A PERSPECTIVEBASED ON COMPOSITION OF USE, August 1982, by Gunvant M. Desai
30 INSTABILITY IN INDIAN FOODGRAIN PRODUCTION, May 1982, by Peter B. R. Hazell
29 GOVERNMENT POLICY AND FOOD IMPORTS: THE CASE OF WHEAT IN EGYPT, December1981, by Grant M. Scobie
28 GROWTH AND EQUITY: POLICIES AND IMPLEMENTATION IN INDIAN AGRICULTURE,November 1981, by J. S. Sarma
27 AGRICULTURAL PRICE POLICIES UNDER COMPLEX SOCIOECONOMIC AND NATURALCONSTRAINTS: THE CASE OF BANGLADESH, October 1981, by Raisuddin Ahmed
26 FOOD SECURITY IN THE SAHEL: VARIABLE IMPORT LEVY, GRAIN RESERVES, AND FOREIGNEXCHANGE ASSISTANCE, September 1981, by John Mclntire
25 INSTABILITY IN INDIAN AGRICULTURE IN THE CONTEXTOFTHE NEWTECHNOLOGY, July1981, by Shakuntla Mehra
24 THE EFFECTS OF EXCHANGE RATES AND COMMERCIAL POLICY ON AGRICULTURAL IN-CENTIVES IN COLOMBIA: 1953-1978, June 1981, by Jorge Garcia Garcia
23 GOVERNMENT EXPENDITURES ON AGRICULTURE IN LATIN AMERICA, May 1981, by VictorJ. Elias
Marito Garcia has been a research fellow at the International Food PolicyResearch Institute since 1982. Per Pinstrup-Andersen was a researchfellow at IFPRI from 1980 to 1986. He now directs the Cornell UniversityNutritional Surveillance Program.
RECENT IFPRI RESEARCH REPORTS
60 POPULATION POLICY AND INDIVIDUAL CHOICE: A THEORETICAL INVESTIGATION, June1987, by Marc Nerlove, Assaf Razin, and Efraim Sadka
59 PRODUCTION INCENTIVES IN PHILIPPINE AGRICULTURE: EFFECTS OF TRADE AND EX-CHANGE RATE POLICIES, May 1987, by Romeo M. Bautista
58 THE FOOD STAMP SCHEME IN SRI LANKA: COSTS, BENEFITS, AND OPTIONS FOR MODIFI-CATION, March 1987, by Neville Edirisinghe
57 CEREAL FEED USE IN THE THIRD WORLD: PAST TRENDS AND PROJECTIONS TO 2000,December 1986, byj. S. Sarma
56 THE EFFECTS OF TRADE AND EXCHANGE RATE POLICIES ON AGRICULTURE IN ZAIRE,November 1986, by Tshikala B. Tshibaka
55 THE EFFECTS OF TRADE AND EXCHANGE RATE POLICIES ON AGRICULTURE IN NIGERIA,October 1986, by T. Ademola Oyejide
54 WEATHER AND GRAIN YIELDS IN THE SOVIET UNION, September 1986, by Padma Desai
53 REGIONAL COOPERATION TO IMPROVE FOOD SECURITY IN SOUTHERN AND EASTERNAFRICAN COUNTRIES, July 1986, by Ulrich Koester
52 FOOD IN THE THIRD WORLD: PAST TRENDS AND PROJECTIONS TO 2000, June 1986, byLeonardo A. Paulino
51 DETERMINANTS OF AGRICULTURAL POLICIES IN THE UNITED STATES AND THE EUROPEANCOMMUNITY, November 1985, by Michel Petit
50 GOVERNMENT EXPENDITURES ON AGRICULTURE AND AGRICULTURAL GROWTH IN LATINAMERICA, October 1985, by Victor J. Elias
49 LIVESTOCK PRODUCTS IN THE THIRD WORLD: PAST TRENDS AND PROJECTIONS TO 1990AND 2000, April 1985, by J. S. Sarma and Patrick Yeung
48 RURAL HOUSEHOLD USE OF SERVICES: A STUDY OF MIRYALGUDA TALUKA, INDIA, March1985, by Sudhir Wanmali
47 EVOLVING FOOD GAPS IN THE MIDDLE EAST/NORTH AFRICA: PROSPECTS AND POLICYIMPLICATIONS, December 1984, by Nabil Khaldi
46 THE EFFECTS ON INCOME DISTRIBUTION AND NUTRITION OF ALTERNATIVE RICE PRICEPOLICIES IN THAILAND, November 1984, by Prasarn Trairatvorakul
45 THE EFFECTS OF THE EGYPTIAN FOOD RATION AND SUBSIDY SYSTEM ON INCOME DISTRI-BUTION AND CONSUMPTION, July 1984, by Harold Alderman and Joachim von Braun
44 CONSTRAINTS ONKENYA'S FOOD AND BEVERAGE EXPORTS, April 1984, by Michael Schluter
43 CLOSING THE CEREALS GAP WITH TRADE AND FOOD AID, January 1984, by Barbara Huddleston
42 THE EFFECTS OF FOOD PRICE AND SUBSIDY POLICIES ON EGYPTIAN AGRICULTURE,November 1983, by Joachim von Braun and Hartwig de Haen
(continued on inside back cover)
International Food Policy Research Institute1776 Massachusetts Avenue, N.W.Washington, D.C. 20036 USA
