Contents - Previous - Next

This is the old United Nations University website. Visit the new site at


1. Batista Filho M, Sigulem DM, Nobriga FJ. Nomenclatura e classificação da desnutrição: la parse desnutrição proteico-calórica (DPC). J Pediatr 1976;41(7-8):45-51.

2. Aylward F, Jul M. Protein and nutrition policy in low-income countries. London and Tonbridge: Charles Knight, 1975.

3. Campino AC, Farina EMMQ. Questão alimentar no curto prazo: um desafio pare a política económica. São Paulo: Fundação Instituto de Pesquisas Econômicas, 1986 (mimeo).

4. Reutlinger S. Food security and poverty in less developed countries. Finance Dev 1985 ;22(4) :7- 11.

5. World Bank. Feeding Latin America's children. Internal Discussion Paper no. 49. Washington, DC: World Bank, 1989.

6. Lustig N. Food subsidy programmes in Mexico. IFPRI Working Papers on Food Subsidies, no. 3. Washington, DC: International Food Policy Research Institute, 1986.

7. Rogers BL, Overholt CA, Kennedy E, et al. Consumer food price subsidies. In: Austin JE, Zeitlin MF, eds. Nutrition intervention in developing countries: an overview. Cambridge, Mass, USA: Oelgeschlager, Gunn end Hain, 1981:99-110).

8. Basiotis PP, Johnson SR, Morgan KJ. Chen JSA. Food stamps, food costs, nutrient availability and nutrient intake. J Policy Model 1987;9(3):725-29.

9. Johnson SR, Burt JA, Morgan KJ. The food stamp programme: participation, food cost and diet quality for low income households. Food Technol 1981 ;35: 60-70.

10. Nelson Jr PE, Perrin J. Technical supplement to analysis of the economic impacts of the U.S. food stamp and school lunch programmes. Washington, DC: Economic Research Service, USDA, 1976.

11. Hollenbeck D, Ohls JC, Posner B. The effects of cashing out food stamps on food expenditures. Am J Agric Econ 1985;67(3):609-13.

12. US Department of Agriculture, Economic Research Service. Impact of the food stamp programme on three local economies: an input output analysis. Washington, DC: USDA/ERS, 1972.

13. Marques RM, Berquo E, Yunes J, Marcondes E. Crecimento de niños brasileños: peso y altura in relación con la edad y el sexo y la influencia de factores socio-económicos. Publicación Cientifica no. 309. Washington, DC: OPS, 1975.

14. IMPEP/IPE. O estado nutricional de crianças de 6 a 60 meses no município de São Paulo. São Paulo: IMPEP/IPE, 1975.

15. INAN, FIBGE, IPEA. Pesquisa nacional sobre saúde e nutrição: resultados preliminares. Brasilia: INAN, 1990 (mimeo).

16. Campino AC. Nutrition interventions in Brazil: the possible impact of food price subsidies. Report submitted to the Kellogg International Fellowship Program in Food Systems, 1989.

17. Gray CW. Food consumption parameters for Brazil and their application to food policy. IFPRI Research Report no. 32. Washington, DC: International Food Policy Research Institute, 1982.


PROAB a food-price subsidy programme in low-income urban areas of Recife, Brazil: Its impact on food consumption, nutrition status, and low birth weight

Yony Sampaio

The experience of PROAB

The status of Brazilians' health and nutrition is rather inadequate in comparison with the country's economic position. Despite a per capita income of over US$2,000, the infant mortality is around 70 per 1,000, exceeding 100 per 1,000 in the north-eastern region. The proportion of children with moderate and severe undernutrition is about 7%, and although this represents a sizeable decrease from 17% for urban and 22% for rural children in 1974-1975, there are severe regional imbalances. Therefore, several food and nutrition programmes have been implemented in the last 15 years, most of them operational in the northeastern region.*

One of these programmes, PROAB, is a food-price subsidy programme targeting low-income urban areas and operating through the private retail system. It is a low-cost programme, whose only cost to government is the subsidy given and whose administration is very simple. It could turn out to be effective if low-income families actually receive most of the benefit.

Berg [1], reviewing the World Bank's experience with nutritional interventions, concluded that PROAB worked smoothly and that, owing to its low cost, expansion on a national scale would consume less than 2% of the national budget. He further concluded that the geographical targeting adopted "proved to be more effective in identifying those in need and less of an administrative burden than eligibility requirements based on household income" [1, p. 96] but also cautiously remarked that "results look promising but more needs to be known" about its impact [p. 100].

This study evaluated the effects of PROAB on a target population in the Recife area, specifically its impact on food expenditures and consumption and on the nutrition status of children younger than 60 months with low birth weights.

PROAB began in 1979 with the following objectives: (1) to improve the health conditions of low-income families by increasing the consumption of basic foods through subsidies, (2) to decrease prices paid by consumers through subsidies supporting small retailers and through direct intervention in the market, and (3) to develop a low-cost alternative to the supplementary food programmes.

The Brazilian Food Company (COBAL) is in charge of buying and distributing food to selected retailers. It also selects the participating retailers located in low-income districts under the supervision of the National Food and Nutrition Institute (INAN). The prices charged by retailers for basic food products are determined by adding a margin of 7% to the price COBAL pays producers, minus a government subsidy of 20%, plus a margin of 11% for retailers (cost + 7% - 20% + 11% = retail price). Eleven basic products were selected for subsidization: rice, sugar, dry beans, manioc flour, cornmeal, dry milk, eggs, dry meat, macaroni, fish, and edible oil.

A basic hypothesis of the programme is that by limiting the distribution of food to poor districts only, it will be of most benefit to undernourished families. Thus, it is intended to be a targeted programme with very low administrative costs. It is recognized that some leakage of benefits to higher-income groups occurs, but its magnitude is not known.

The programme was started in the Recife area and later expanded to other provincial capitals in the north-east. Although it has been growing, the amount of food distributed per capita has decreased, which raises doubts about potential coverage and about whether the family calorie and protein targets have been met (table 1). In this regard, it was reported that the supply of food by COBAL has shown irregularities (FUNDAJ/INAN unpublished report, 1985). In particular, 1986 was a year of food shortages due to increased demand, which was a result of the Cruzado Plan. Some commodities were not regularly distributed for several months.

An exploratory survey carried out by FUNDAJ (Joaquim Nabuco Foundation) in Recife found that COBAL prices without a subsidy were quite similar to (and sometimes even higher than) the prices at supermarkets (FUNDAJ/INAN unpublished report, 1985). But a different conclusion was reached by Berg on the basis of unpublished papers prepared for the United Nations and the World Bank [1]. He reported that consumers served by PROAB-licensed stores paid 20% to 25% less than normal market prices because COBAL purchases in bulk at lower prices. In fact, the success of COBAL led to its expansion, and "the government of Colombia, . . . asked for and received [World] Bank assistance in setting up a similar program" [1, p. 22]. A more detailed analysis concluded that COBAL prices without subsidy are higher than prices in local supermarkets [5]. Still, because of the subsidy, it is advantageous to buy in the PROAB-licensed stores.

The subsidy itself represents nearly the only cost to government. Administrative costs are low since no additional employees were hired either by INAN or COBAL. COBAL charges 7% to cover its costs and is supposed to have no operational deficit.

TABLE 1. PROAB-number of participant retailers, amount of food distributed, and number of beneficiaries in Recife and Pernambuco Province

  Participant retailers Food distributed (tons) Beneficiaries (thousands) Food per capita (kg)
Recife Province Recife Province Recife Province Recife Province
1980 121 121 9,241 9,241 100 100 92 92
1981 112 716 8,687 16,883 160 460 54 36
1982 141 1,034 7,563 30,466 160 460 47 66
1983 231 2,631 15,507 77,354 300 1,420 52 54
1984 331 3,081 27,470 88,562 516 2,122 53 42
1985 619 3,345 22,832 71,785 672 2,228 34 32
1986 725 4,092 32,772 107,905 696 2,794 47 39
1987 762 4,246 39,080 121,725 770 3,425 51 36
1988 804 4,263 12,944a 36,576a 770 3.433 - -

Sources: Refs. 2-4. C. da Silva, unpublished report, 1987.
The figures for tthe province included thpse for Recif, which is its capital.
a. January-June only.


The area covered by PROAB in Recife consists of seven large suburbs: Ibura, Brasilia Teimosa, Guararapes, Jequiá, Nova Descoberta, Beberibe (Peixinhos), and Maranguape. Ibura, included in 1987, is quite new to the programme. Brasilia Teimosa experienced an influx of low-income squatters in the late 1950s, but housing and sanitary conditions have improved considerably in the last 20 years; average income is much higher than in the other areas. Guararapes and Jequiá are slum sub-areas with very insanitary conditions and a large number of underemployed families. Conditions in Nova Descoberta, Beberibe (Peixinhos), and Maranguape are between these two extremes.

A three-stage sampling procedure was adopted for the study: first to sample the area, second to sample sub-areas, and third to sample the households. Jequiá was sampled, and within it two sub-areas, Mustardinha and Mangueira, were surveyed. A third sub-area, Mini-Central, was added because it is a very poor slum area. The control group consisted of sub-areas in a neighbouring area with similar income and housing characteristics. The survey was carried out from 30 June to 10 August 1987. Data on projected and realized samples are presented in table 2.

A comparison of target and control groups showed that when family composition was similar, significant differences were seen in average income and income distribution. Mangueira and Mustardinha have a greater number of high-income households, with average incomes of Cz$1,252 (0.63 minimum wages) and Cz$1,518 (0.77 minimum wages) respectively; households in Mini-Central have a lower average income of Cz$1,018 (0.52 minimum wages) and income distribution slightly worse than that of the target group.

Whereas 70% of the families in the target group participate in the programme, only 15% of the families in the control group do. As expected, participation decreases with income. Because of this, it was decided that analysis by participation would bias the results when comparing low-income families with higher-income families. Therefore, although the degree of participation was taken into consideration, the target and control groups were defined by geographical criteria.

Comparative and multivariate analyses were used to study the impact of PROAB. The effects on total food expenditures, expenditures and consumption of the 11 subsidized commodities, retail prices, calorie and protein consumption, the percentage of children with low birth weights, and children's nutrition status are analysed by comparisons between the target and control groups.

A more accurate analysis of dissimilar characteristics can only be made through multivariate analysis. The inclusion of variables representing dissimilar characteristics allows a more explicit analysis to be made of the programme's effect on household calorie consumption, nutrition status, and birth weight. The estimate of PROAB's effect on household calorie consumption is obtained from the regression equation

CAEU = f(YAEU, S, TF, E, Dm), (1)

where CAEU is calorie consumption per adult equivalent unit, Y is real household income, S is a dummy to account for the PROAB subsidy, TF is family size, E is the educational level of the wife, and Dm is a dummy variable to account for sub-area differences.

The impact of the PROAB subsidy is measured through the increase in household incomes and through a direct effect (S) hypothesized by Garcia and Pinstrup-Andersen "to come about primarily because households [can] treat the real income increase from food subsidies differently from other income increases-that is, the marginal propensity to spend on food [may differ] between subsidy income and other real income" [6, p. 29]. Positive coefficients are expected for YAEU, S, and E and negative coefficients for TF.

PROAB effects on the growth of preschoolers are estimated by an equation of the following form:

W = g(I, I², E, NP, PC, A, SX, Y, S, IS, D), (2)


W = weight as a percentage of the standard weight for a particular age (Gomez index),
I = the age of the mother (and 12 is the square of the age variable),
E = the mother's educational level,
NP = household size,
PC = birth order of the child,
A = the child's age,
SX = the child's sex,
Y = total household income,
S = a term to account for PROAB subsidy,
IS = an index of sanitary conditions (potable water or disposable waste facility),
D = a dummy to account for diarrhoea in the child (1 if the child had diarrhoea in the past two weeks).

Positive coefficients are expected for Y, IS, I, and E and negative coefficients for I², D, A, PC, and NP.

PROAB effects on the birth weight of preschoolers are estimated on the basis of an equation of the following form:

PN = h(I, I², E, NP, SX, Y, IS, H, F, NC, AB, P), (3)


PN = birth weight,
H = a dummy to account for prenatal health care (1 if the mother had undergone prenatal examinations),
F = a dummy to account for maternal smoking (1 if the mother did not smoke during pregnancy),
NC = number of parturitions,
AB = a dummy to account for abortions (1 if the mother had a prior abortion),
P = a dummy for the type of parturition (1 if it was a caesarean),

and the other variables are as previously defined.

TABLE 2. Projected and realized number of households sampled, Recife, 1987

Sub-area Projected Realized Income group
sample sample 1 2 3 4
  Mangueira 100 102 13 21 34 34
  Mustardinha 100 100 8 21 36 35
  Mini-Central 45 45 14 19 4 8
Control 100 100 24 31 27 18

Because of the small sample size (only children under I year of age were considered) not all variables are included. The model is estimated for participant and non-participant households. Positive coefficients are expected for Y, IS, H, F, I, and E and negative ones for I² and NG.

Finally, a crude analysis of leakages is made on four sources of leakages. The first is the sale of subsidized products outside the target area. To estimate the importance of this source, the sample was expanded and the projected total consumption of each subsidized food was compared to total sales in the month of July 1987.

The second source [of leakage] is households participating that are not deficient in calories or that do not have preschoolers. The third source is substitution between food and nonfood items and among foods resulting in a net increase in household food expenditures and caloric consumption that is smaller than the equivalent real income transfer. The fourth is when some of the net increase in household caloric consumption is consumed by household members other than preschoolers [6, p. 34]


Comparative analysis of the effects of PROAB

The effects of PROAB are estimated by comparing the target group and the control group in terms of total food expenditures, consumption of the subsidized food, prices paid, nutrition status, and birth weight.


Total food expenditure

Average total food expenditures per household and per capita were less for the subsidized PROAB target group than for the control group (table 3). The possible conclusion is that the subsidy allowed the target households to acquire the same bundle of goods at a lower price, spending the extra income on nonfood items. Since the subsidy has been in operation for several years in this area, this behaviour represents a stable adjustment, perhaps changes that occurred after PROAB began.

Two comparisons were made in relation to this result: a rough calculation of the income effect (quantities consumed times price differentials), and expenditures on subsidized food items. The income effect is negligible, representing 0.81% of total expenditures for income group 1 and 0.62% for the other groups. Given the fact that products subsidized by PROAB represent about 22% of total expenditures, one would expect that the subsidy would result in a modest 4.4% increase in income if the consumption bundle remained constant. But, as will be shown in the next section, the price differentials are lower, and quantities bought by PROAB households are only about 70% of the total consumption of these commodities by the control group (table 4).


Consumption of subsidized commodities

A comparison of the target and control groups reveals that, in terms of the quantities of subsidized commodities, consumption patterns of both groups are very similar, with the exception of dried fish. During the period of data collection, however, PROAB was not selling dried or salted fish. Slightly more dry meat was consumed by the control group. For the other commodities the differences are not significant. The prices paid by the test group are, in general, lower, but in a few cases they are above the prices paid by the control group. The differences are greater for sugar, rice, eggs, and dry milk for all income groups, for cornmeal for income group 4, and for macaroni for income groups 2 and 4.

TABLE 3. Comparison of household food expenditures by PROAB target group and control group

Income level   Food expenditures Difference (%)
Total   Per capita   Total Per capita
Target Control Target Control
1 3,363 3,649 818 867 -7.8 -5.6
2 4,138 4,182 836 869 - 1.0 -3.8
3 5,145 6,011 875 1,249 -14.4 -30.0
4 7.213 7.542 1.225 1.218 -4.4 0.5

The case of dry milk is of considerable interest. Consumption of dry milk by the test group is two to three times that of the control group. However, the total quantity of milk consumed-fluid milk plus dry milk converted into fluid (130 g of dry milk is considered equivalent to 1 litre of fluid milk)-is smaller for the target group. This is surprising, since the net effect is a decrease in milk consumption. In addition, fluid milk is cheaper than dry milk, which means that households from the target group actually moved to the more expensive commodity. Two possible explanations can be advanced. First, the conversion rate of dry to fluid milk used is one that maintains similar nutrient content; however, it is known that, particularly in low-income families, people frequently add much more water, creating the illusion that more milk is consumed. Second, dry milk is more convenient because it does not require cold storage and thus does not have to be purchased daily. This decreases both the cost of storage and the time required for daily purchase.

A final comparison was made to ascertain whether the percentage of families consuming the subsidized commodities varied between control and target groups. Practically all households consume sugar, rice, dry beans, eggs, and macaroni, and most consume dry meat and vegetable oil. But as income increases, the percentage of those in the test group consuming manioc flour and cornmeal decreases; manioc flour income elasticities are negative and substitution can be expected. Income elasticities for cornmeal, although low, are positive. Dry or salted fish is seldom consumed by the two groups. Most households in the control group consume fluid milk, whereas target households consume dry milk. Thus, although PROAB may have had an effect on consumption of the subsidized commodities, this effect has been small and appears to have been negative.

TABLE 4. Monthly per capita consumption of subsidized commodities: target versus control group and price differentials by income level

Commodity Target group Control group Price differentials (%)
1 2 3 4 1 2 3 4 1 2 3 4
Dry meat (kg) 0.37 0.35 0.48 0.65 0.48 0.44 0.41 0.72 -1.7 -5.2 -4.5 -11.1
Salted fish (kg) 0.00 0.04 0.01 0.01 0.02 0.00 0.01 0.00 - - -  
Sugar (kg) 2.14 2.00 2.25 2.48 2.48 2.14 2.01 2.39 - 13.7 - 15.9 -16.0 - 13.3
Oil (500 ml can) 0.55 0.42 0.50 0.55 0.43 0.50 0.40 0.57 +4.0 -2.0 -2.3 -4.7
Rice (kg) 1.30 1.32 1.44 1.56 1.47 1.35 1.34 1.50 - 16.2 -22.7 - 16.1 -24.8
Dry beans (kg) 1.17 1.32 1.41 1.55 1.49 1.31 1.37 1.33 - 1.2 +3.3 -5.8 - 1.8
Manioc flour (kg) 1.24 1.47 1.53 1.27 1.49 1.30 1.13 1.13 +9.1 -0.8 - 10.5 -4.6
Cornmeal (500 g) 1.22 1.53 1.48 1.22 1.38 1.17 1.44 1.17 +4.2 +1.8 -4.8 -25.5
Macaroni (500 g) 1.10 1.30 11.8 1.30 1.06 1.21 1.37 1.24 -4.3 -31.1 -12.8 -25.7
Dry milk (200 g) 1.22 0.93 0.94 0.79 0.29 0.64 0.77 0.23 -32.5 -11.9 -25.1 -31.2
Fluid milk (litre) 3.55 3.84 3.26 4.15 6.98 4.57 4.31 3.74 - - -  
Fggs(units) 11.55 12.64 12.64 15.65 13.30 10.42 14.71 15.57 -19.0 -20.4 -13.6 -18.5

Price differentials

A basic hypothesis of PROAB is that COBAL is effective in reducing marketing costs because it deals with large quantities of products. Thus. by choosing to operate with small retailers the programme would benefit the participant retailers and the poor consumers whom they supply.

The evidence collected through prior evaluation of PROAB and reviewed by Musgrove showed that PROAB prices without the subsidy were equivalent to or even higher than supermarket prices, and that price differentials among retailers were not significant [6].

The data presented in table 4 show that, with the exception of dry milk, eggs, rice, cornmeal, and macaroni (for some income groups), prices without the subsidy would be higher for the target group. Even with the subsidy, the differentials are minimal for oil, dry meat, dry beans, cornmeal, and manioc flour. Thus, this analysis reinforces earlier studies that shed doubt on the efficiency of government in reducing market prices through direct intervention.


Calorie and protein consumption and adequacy

Daily per capita calorie consumption is higher for the target group in income groups 2 and 3 and lower for the target group in income group 4 (table 5). Daily per capita protein consumption for the target group is lower in income group 1 and slightly higher in the other income groups. Although food expenditures decreased and the consumption of the 11 subsidized commodities varied little, PROAB's effect on calorie and protein consumption was slightly positive.

TABLE 5. Comparison of daily per capita calorie and protein consumption

Income level Target Control Difference (%)
Calories (kcal)
1 2,279 2,261 0.8
2 2,457 2,223 10.5
3 2,325 2,149 8.2
4 2,476 2,693 -8.0
Protein (g)
1 77 81 -4.9
2 77 76 1.3
3 84 82 2.4
4 108 107 0.9

This positive effect can also be seen in the decrease in the percentage of families with nutrition inadequacies, in particular for those whose diets are less than 60% adequate (table 6). Although the proportion of families experiencing protein inadequacy is small, the proportion of the sample population with diets less than 60% adequate in protein is smaller for the target group.


Nutrition status

Children's nutrition status was compared between the target and control groups, and a few variables were tested statistically to verify differences between them.

Among the several measures of nutrition status, two were chosen: the percentile distribution of children according to the NCHS standard presently recommended by WHO and widely used by nutritionists; and the Gomez index, because of its simplicity and the ease of comparison with available results.

TABLE 6. A comparison of calorie and protein adequacy

Adequacy (%) Calories Protein
Target Control Target Control
<60 8.5 15.0 2.0 5.0
<70 18.2 18.0 2.8 6.0
<80 24.7 25.0 4.0 7.0
<90 32.4 38.0 4.8 7.0
<100 46.6 43.0 6.0 10.0
>100 53.4 57.0 94.0 90.0

Values are cumulative percentages for those with diets < 100% adequate.

Only indices of weight for age are presented here.

Data in table 7 show a worse situation for the target group than for the control group. A high proportion of the latter children are overweight, and a lower percentage are in the first and second deciles. The conclusions are similar when standard deviations of median values or indices of height for age and weight for height are used.

The results are rather surprising, because the conclusions are consistent for all indices: the control group households earn lower average incomes, but the children have better nutrition status. Thus, a major conclusion is that PROAB apparently was not effective in improving nutrition status. Table 8 shows the nutritional status of children broken down by sub-area. This gives a picture more consistent with income figures. Despite a single case of DIII at Mustardinha, nutrition status is worst in the Mini-Central sub-area. Still, the proportion of overweight children in the control groups is incredibly high.

Some preliminary statistical tests (chi-square) were done to verify simple associations between nutritional status and socio-economic variables. The housing conditions (e.g., type of wall construction) were not significant. However, sanitary conditions (sewage and water systems) and the health condition of children (respiratory diseases) were significant.

TABLE 7. Nutrition status of children: percentile distribution using the NCHS standard weight for age

Percentile Target group Control group
Sex   Sex Total
M F Total M F
<3 11.1 20.0 15.4 8.7 7.5 7.9
3-10 13.6 12.0 12.8 17.4 2.5 7.9
10-90 70.4 60.0 65.4 60.0 77.5 71.4
90-97 3.7 2.7 3.2 8.7 2.5 4.9
>97 1.2 5.3 3.2 4.3 10.0 7.9
Total 100 100 100 100   100

TABLE 8. Nutrition status of children: percentages categorized by the Gomez status index

Nutrition status Target group Control group (N = 63)
= 64)
Mini-Central (N = 29) Mustardinha (N = 63) Total
= 156)
Overweight 10.9 - 7.9 7.7 20.6
Normal 47.0 48.3 47.6 47.4 55.6
DI 35.9 41.4 38.1 37.8 22.2
DII 6.2 10.3 4.8 6.4 1.6
DIII - - 1.6 0.7 -
Total 100 100 100 100 100

TABLE 9. Birth weight: percentages of total births by weight categories

Weight (g) Target group Control group (N = 11)
= 20)
Mini-Central (N = 8) Mustardinha (N = 5) Total
= 33)
<2,500 10 50 20 21 18
2,500-2,999 10 0 20 9 45
3,000-3,499 45 38 0 36 18
3,500-3,999 20 0 20 15 18
>4,000 15 13 40 28 0
Total 100 100 100 100 100

Birth weight below 3,000 g is considered insufficient.

Birth weight

It has been shown under similar conditions that an improved natural diet is associated with increased birth weight [7]. Thus, if PROAB is effective in improving nutrition, the target group would be expected to have a smaller percentage of low-birth-weight children. In accordance with this expectation, 64% of the children in the control group were of low or insufficient weight, compared with only 30% in the target group. Disaggregated data show the situation in the target group to be worse in the Mini-Central sub-area, an area served by PROAB, but which has worse sanitary conditions and lower incomes (table 9). An analysis of possible causes for this difference between groups suggests that variables such as maternal smoking, previous abortions, and number of parturitions were not significant. However, the amount of prenatal care was significantly related to birth weight.

Contents - Previous - Next