Contents - Next

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

Hunger, health, and society

The nutritional impact of food aid: criteria for the selection of cost-effective foods
Proposed guidelines for designing evaluation for nutrition and health programmes
Determinants of dietary intake and dietary adequacy for pre-school children in Bangladesh
Intra-familial distribution of food in rural Bangladesh
Maternal and infant nutrition in developing countries, with special reference to possible intervention programmes in the context of health


The nutritional impact of food aid: criteria for the selection of cost-effective foods

Shlomo Reutlinger
Senior Economist, World Bank, Washington, D.C., USA

Judit Katona-Apte
Nutrition Adviser. World Food Programme


Some food aid programmes, both in the United States and in other countries, provide people in need with aid in kind rather than in cash. Food for Peace, operating in many developing countries under US Public Law 480, Title II, typically delivers food commodities—grain, cooking oil, non-fat dry milk, etc.—to selected recipients for supplementation of their diets. In some domestic US programmes, such as the supplementary feeding programme for women, infants, and children, the recipients are given vouchers that they can exchange for selected foods, such as fortified cereals, dairy products, and infant formulae.

While one important motivation for delivering food rather than cash is often that governments acquire food as a result of programmes designed to support farm incomes, it is also often alleged that aid in kind, particularly food, is preferable to cash aid because the former is nutritionally more effective. This latter allegation requires close scrutiny. Is food aid nutritionally more effective because it leads recipients to allocate a larger share of their total expenditures to food than they would when provided an equivalent amount of cash aid? Or is it presumed that the recipients would consume the particular foods delivered in the programme in addition to their otherwise existing diets?

Precise answers to these questions have profound implications for the way food aid should be administered. Acting on the basis of how aid givers would like recipients to use the aid rather than on how recipients actually use it can be very costly in terms of the intended outcome. The nutrient content of the food-aid commodity must not be confused with its nutritional effect. And most importantly, correct perceptions about the effects of food aid lead to criteria for selecting implementation procedures that can increase potential benefits. Our purpose here is to sharpen these perceptions and to suggest appropriate criteria for the selection of commodities to be used in some food aid programmes.


As already suggested, recipients of food aid have essentially two extreme options for using the food provided. They might use it as a net addition to the food they are consuming in the absence of any such aid. This kind of utilization should be expected only in the case of emergency feeding or disaster relief when the recipients would have very little food from any other source. The nutritional effect is then predictable on the basis of pathway A described in figure 1; i.e., the nutritional effect stands in direct proportion to the nutrient content of the foods provided through aid. Similarly, the nutritional cost effectiveness of different foods is directly related to the nutritional value per unit cost of the different foods. Because a nutritionally effective diet requires an appropriate mix of energy and specific nutrients, the most cost-effective mix of foods is the one that supplies all the nutritional needs of the recipients at the least possible cost.

The other option is for the recipients to substitute the food supplied in the aid programme all or in part for the same or similar foods they would be purchasing otherwise. The impact of the food is then essentially determined by the size of the potential savings realizable as a consequence of the aid programme's distributed food and what the recipients choose to purchase with the added purchasing power. This option is described by pathway B in figure 1. The basic feature of this pathway is that the amount and composition of foods consumed as a result of the food aid may or may not bear any resemblance to the food aid package provided. As we will show below, the net addition to energy or protein intake may be more or less than that contained in the provided foods. Moreover, some of the additional purchasing power is likely to be used for other than added food consumption. This may be disappointing to some donors of aid in kind who believe that food aid is only effective as long as it is used exclusively to increase the consumption of food and even of that particular donated food. However, it is not at all clear whether some of the recipients' non-food-related expenditures (e.g. for housing, clothing, education), indirectly made possible by the added purchasing power released as a consequence of the food, are any less important than added food consumption to the improvement of the recipients' nutritional status.

This second pathway of nutritional effect suggested here might appear new and innovative. However, numerous studies in the past have suggested that the nutritional effect of food aid is not necessarily caused by the nutritional content of the food aid package, but by its income transfer potential. A CARE report for example, claims, "Mothers stated that they were able to spend less on foods for their family since being enrolled in the CARE feeding program " ( 1, p. 87). In a study of 302 families with pre-school children and pregnant and lactating women in Peru who were supplemented with two protein levels (2), it was found that the supplemental food was really a substitute food, since it displaced foods of lesser nutritional value. In fact, total energy intake remained the same, but foods of higher biological value were replacing foods normally consumed. There was found to be no advantage to the higher level of protein in the supplement in terms of mortality and growth.

Findings from Colombia are similar. Herrera et al. concluded that food supplementation cannot be considered "as purely nutritional since the intervention served to free family resources for satisfying other material needs" 13, P. 176). More et al. offer a non-nutritional explanation for the results by stating that "supplemented families utilized a significant portion of the supplements to substitute for food previously purchased, thus achieving an economic advantage over the unsupplemented subjects" (4, pp. 173-174).

A comprehensive, recent survey of supplementary feeding programmes for young children in developing countries by Beaton and Ghassemi (5) provides the most persuasive evidence in support of the existence of a second pathway. They report that, in both the 8 supervised feeding and the 13 take-home food programmes for which they had data, "the net increase in intake by the target recipient was 45-70% of the food distributed with one program showing a net effect of only 10-15%." They speculate that the "leakage" must be due to sharing of the food with other members of the households (in the case of take-home food programmes) and also to displacement of foods in the existing diets of the target recipients.

In a more recent write-up of the Colombia data (6), it was claimed that education programmes by themselves do not improve diets unless an increase in household resources accompanies them to enable the recipients to acquire the nutritionally preferable foods. They stated that supplemented families achieved better nutrition because "the food supplementation program increases family income because it increases total resources available for family consumption.... the number of food items increased and these other foods contributed an increasing share to both calories and protein adequacy" (p. 1160). In a supplemental feeding programme in Guatemala food supplementation was also "found to increase variability in nutrient intake" (7, p. 304).

FIG. 1. Pathways of Food Aid into Nutritional Status

Given that the indirect pathway is quite commonplace, two basic questions arise: Is food aid at all nutritionally effective? How can it be made more effective through the application of appropriate criteria to the selection of foods in food aid programmes? The remainder of this paper will address these questions.


If the foods provided in supplementary take-home food programmes are all or in part substituted for foods in the current diet, can food aid be expected to make any significant contribution at all to nutrition improvement? In other words, if the food aid package is "just" a vehicle for transferring purchasing power, can the recipients be entrusted with using that purchasing power to improve their nutrition?

The answer undoubtedly depends in the first place on what is meant by improvement of nutrition. A nutrition planner who has a great deal of confidence in his ability to determine precisely a particular nutritional deficiency or the insufficiency of intake even of a particular food in a population might well be disappointed when the recipients do not commensurately increase the consumption of the food supplied by the programme. More realistically, however, unsatisfactory nutritional status is a very complex phenomenon attributable to inadequate intakes of energy and/or of specific nutrients and morbidity factors that interact with nutritional ones. Moreover, the relative deficiencies will vary a great deal among individuals even in a seemingly homogeneous population. Viewed in this context, the ability of individuals to make use of food aid in a flexible way, perhaps suited to their particular circumstances, may indeed be a great blessing rather than a shortcoming.

Empirical evidence from many sample surveys of food consumption and food expenditures by households with different levels of income confirm the supposition that poor people use additional purchasing power to increase their energy and nutrient intake.

Table 1 summarizes food expenditures based on cross sectional survey data collected in several countries The range of income groups considered coincides with groups having reported per capita energy intakes in the range of approximately 1,500 to 2,000 calories, with the lowest and higher income groups excluded. The analysis concentrates on the low (but not the lowest) income groups, for which data are more reliable and on which supplementary feeding programmes are likely to focus. This group comprises between 30 and 50 per cent of the total population of the countries.

TABLE 1. Average Food Expenditures, Elasticities, and Marginal Propensities to Spend on Food of Low-Income Populations in Selected Countries

Country (Currency) Monthly Income Monthly Food Expenditure
(and Proportion of Income)
Daily Energy Intake
Low High Low Income High Income Low Income High Income
Bangladesh (take) 53.3 74.8 40.1 56.4 1,543 1,927
  (.75) (.75)      
India (rupees) 26.0 38.4 21.7 31.4 1,528 1,904
  (.83) (.83)      
Indonesia (rupiahs) 2,503 4,458 2,065 3,479 1,712 2,057
  (.83) (.78)      
Morocco (dirhams) 348 504 262 357 1,652 2,033
  (.75) (.71)      
Pakistan (rupees) 26.4 46.4 18.4 25.9 1,739 1,993
  (.70) (.56)      

Food expenditures in this group generally take up between 70 and 80 per cent of total income. The only notable exception appears to be Pakistan. The explanation could be that in that country, food is heavily subsidized through the ration shops and thus is generally available at relatively lower prices.

Income elasticities and marginal propensities calculated from the grouped data presented in table 1 are presented in table 2. Income elasticities are the percentages of change in food expenditures, energy intake, and the cost per unit of energy for a 1 per cent change in income: the marginal propensities indicate the change in food expenditures and energy intake for each dollar of additional income.

On average, the food expenditure elasticity is 0.85, and the energy intake elasticity is about 0.50. The discrepancy between the food expenditure elasticity and the energy intake elasticity is the elasticity of the cost per unit of energy. As income rises, people purchase more expensive foods per unit of energy. Noteworthy are the observed lower energy elasticities in Indonesia and Pakistan. In Indonesia, a particularly large discrepancy is observed between food expenditure and food energy elasticity. Apparently, as income rises there is a high propensity to replace low-cost energy from tubers with high-cost energy from rice. In the case of Pakistan, the low food energy intake elasticity, like the low food expenditure elasticity, might be related to the extensive food subsidy schemes operated by the government.

Many more household food expenditure and consumption surveys tell a similar story. At low levels of income, households with higher income spend more on food and also increase their energy intake. Between 60 and 70 cents out of every additional dollar is spent for food. The additional dollar is spent for food. The additional energy intake is usually much less than what could be bought at the per unit cost of energy in the existing diet. Clearly, people use additional purchasing power to purchase more expensive foods.

From the surveys cited so far, we cannot tell whether people spend more per unit of energy because they buy more nutritious, more tasty, or more convenient foods. The results of the surveys reported in tables 3 and 4 suggest that the observed tendency to purchase higher cost foods is indicative of both a propensity to consume tastier and/or more convenient foods and to increase the quantity and quality of protein in the diet (9-14). In general, however, the percentage increase in protein intake is at least equal to, and in most cases higher than, the percentage increase in energy as income rises.

To return to the question posed earlier: Will donated foods lead to improved nutrition if they substitute for foods in the existing diet and thereby release purchasing power? The evidence cited provides a qualified positive answer:
- If food-aid-recipient households with low incomes behave like other households currently observed to have higher incomes, they will spend some of the conveyed purchasing power on food. Similarly, they will also marginally improve the quality of their diets.

TABLE 2. Income Elasticities and Marginal Propensities to Spend on Food and Increase Energy Intake

  Income Elasticities of Marginal Propensity per Dollar
Food expenditures Energy intake Cost per unit of energy Food expenditures (dollars) Energy intake (calories)
Bangladesh 1.00 0.67 0.33 0.75 4,450
India 0.95 0.56 0.39 0.78 7,520
Indonesia 0.91 0.32 0.59 0.73 2,230
Morocco 0.84 0.55 0.29 0.61 4,020
Pakistan 0.62 0.24 0.38 0.38 4,230
Average 0.86 0.47 0.40 0.65 4,490

Source: ref. 8.

The respective exchange rates used per US dollar were as follows: Bangladesh (19731974),8.1 take; India (1974),8.0 rupees; Indonesia (1976),415 rupiahs; Morocco (1971), 4.6 dirhams; Pakistan (1971-1972), 11.0 rupees.

TABLE 3. Income Elasticities of Food Energy and Protein

Survey Elasticity
Energy Protein
Brazil (Ceara)a    
rural (1973) 0.53 0 74
rural (1975) 0.52 0.68
urban (1975) 0.33 0.51
urban 0.69 0.90
rural 0.52 0.47
0.52 0.55
rural 0.24 0.15 0.63
rural low-income 0.41 0.24 2.41

Sources: (a) ref.9; (b) ref. 10; (c) ref. 11; (d) ref. 12.

TABLE 4. Percentage of Increase of Food Energy, Total Protein, and Animal Protein between Low and Middle-Income Group Diets

Survey Percentage of Increase
Energy (calories) Total protein (grams) Protein from milk, meat, fish, eggs (grams)
Kerala 19 45 39
Andhra 15 16 79
Karnataka 9 14 54
Tamil Nadu 5 12 28
rural 21 19 31
urban 35 39 59

Sources: (a) ref. 13; (b) ref. 14.

Results from statistical analyses of food consumption and income in many countries over time provide some evidence that the observed elasticities are quite robust.

- Nevertheless, a good case could be made for suggesting that the marginal propensity to spend on food from a food aid programme could be either lower or higher than the differences in food consumption observed among households having different incomes. There may be factors that jointly determine income and food consumption. For instance, people who need more food for physiological or psychological reasons may work harder and earn higher incomes, or people who have better access to remunerative employment may therefore also require more food to sustain higher activity levels. The latter could explain, however, only a small portion of the observed differences in energy intakes.
- It has also been suggested, but without much empirical evidence, that the marginal propensity to spend on food is higher when additional purchasing power is conveyed through food aid than from ordinary income. One explanation could be that women can influence the allocation of food aid more than the allocation of additional income acquired in other ways (15).
- There is also some evidence that the propensity to spend on food can be influenced by nutrition education, particularly if households are provided at the same time with additional purchasing power.
- In any case, how much of the food aid will be utilized for increasing food consumption and how much will be diverted into reducing other deprivations will vary a great deal from one population to another and among households in the same population. If we accept that people do not behave arbitrarily but in their best interest, this might itself be a good reason for providing food aid in a form that affords the recipients opportunities to make use of it in ways suited to their particular circumstances.


Once it is accepted that under certain conditions the changes in quantities and kinds of foods consumed by the recipients of food aid bear little resemblance to the contents of the food aid package, criteria for selecting foods that are based exclusively on nutrient contents of the food aid commodities are obviously erroneous. An entirely new approach is needed for achieving the best possible nutritional cost-effectiveness in food aid programmes.

The new approach discussed below is applicable wherever food aid provides households with a relatively small, rationed amount of food that is taken home to supplement the otherwise existing diet. Under these circumstances-in contrast to food aid programmes that provide for disaster relief-foods are available in the local market, and if households are not consuming enough it is for lack of sufficient purchasing power. The household can then be expected to substitute the food aid commodity, all or in part, for foods in the existing diet and to use the savings for the purchase of other food and non-food items. To know the precise nutritional cost-effectiveness we always need to know the household's marginal propensities for spending the saved purchasing power. Fortunately, however, these propensities need not be known for determining the relative nutritional cost-effectiveness of different food aid commodities in many circumstances. It is then quite easy to select the commodities that maximize the nutritional cost-effectiveness of the programme.

First, let us consider the case of a commodity that is identical to a food in the existing diet of the recipient household. In this case, the cost-effectiveness of the commodity i with respect to nutritional unit j (calories, protein etc.), eij, is simply



Vi= monetary value per unit of food i to recipient.
Ci = cost per unit of food i (including cost of delivery) to programme.
mj = household's marginal propensity to spend on nutritional unit j.

The appropriate monetary value of the commodity to the recipient depends on the food likely to be all or in part replaced in the existing diet. It could be the same or a food similar to the one supplied by the programme. Its value would be the retail price if normally purchased in the open market. If the food aid commodity potentially replaces a food obtained at a concessionary price in a food ration shop, the concessionary price would be the appropriate value, or, if the food aid commodity is supplied to a farming household and is potentially used to replace a food that is in marketable surplus, the appropriate value is the farm-gate price of the food.

Since the marginal propensity to spend on nutritional unit j (mj) is independent of the food aid commodity, it is easy to see that the relative nutritional cost-effectiveness of a food aid commodity that replaces the same food (or another food that is the same with regard to the nutritional unit) in the existing diet is simply


An a value of 1.00 means that a dollar spent by the food aid programme on the commodity is worth a dollar to the recipient. Similarly, an a value of 0.50 means that a dollar spent on that commodity is worth only 50 cents to the recipient. If the a value of a commodity is twice the a value of another commodity, it is twice as nutritionally costeffective with respect to any nutritional unit, provided that the marginal propensity of the recipient to spend additional income on the nutritional unit is at all positive.

The second important case involves the substitution of a food aid commodity for another food in the existing diet. This would occur when the commodity supplied by food aid is not part of the existing diet. To the extent that the nutritional units in the two foods differ, the nutritional cost-effectiveness would then be



Nij = nutritional unit j in a unit of the delivered commodity i.
Nsj = nutritional unit j in a unit of food replaced by commodity i.
Vs = value of food replaced by commodity i.

The most common case of a food substituted for another food in the diet is a fortified commodity. Both the fortified and the replaced commodity contain the same amount of energy and, therefore, the first term in equation 3 is zero. Consequently, a comparison of a fortified and a non-fortified commodity with respect to energy cost effectiveness can still be made without knowing the marginal propensity to spend on energy. A fortified (and therefore more costly) food will always be less energy cost-effective than an equivalent non-fortified food because of its higher cost per unit. However, for comparing the protein cost-effectiveness of a fortified food with other commodities, it is necessary to have knowledge about the marginal propensity of spending on protein and the difference between the protein content of the supplied and the replaced food in addition to knowing the values and costs of the different commodities.

This, in a nutshell, is the theory. The next section will illustrate that using the proper cost-effectiveness criteria can be very important to the success of aid-in-kind programmes.


The reasons that a values and therefore cost-effectiveness can vary greatly among commodities and food aid programmes are many.

First, the recipients of food aid may live under a regime of relative food prices that bear little resemblance to the relative prices paid by the food aid donor. This is particularly relevant to international food aid programmes. The governments of countries in which the food aid recipients live often will use trade restrictions and selective taxes or subsidies to hold the prices of some foods below and the prices of other foods above international prices. Moreover, domestic food prices may differ from international prices depending on whether the country or a particular region in the country is deficient or self-sufficient in a particular food or has a surplus.

TABLE 5. Imputed Retail Prices, PL 480 Title II Costs, and a Values for Selected Commodities, 1981 (prices and costs in US dollars per metric ton)

  Wheat Corn Rice Oil Milk (fluid)
Retail pricea
Egypt 94 (0.25) 105(0.32) 370(0.58) 763 (0.76) 624(5.20)
India 187 (0.50) 213(0.65) 328(0.51) 1,889 (1.89) 403(3.36)
Pakistan 178 (0.48) 185(0.56) 451(0.70) 873 (0.87) 332(2.76)
average 153 (0.41) 159(0.51) 383(0.60) 1,175 (1.17) 453(3.78)
Food aid costb 370 330 640 1,000 120

Source: ref. 16.
Figures in parentheses are a values
a. Retail prices are imputed from data reported from earlier years.
b. The food aid cost for the commodities are the acquisition and average shipping costs to the PL 480
Title II programme. The cost of milk is based on the concessionary price charges for non-fat dry milk (NFDM) (60 cents per kilogram including shipping charges) and a conversion ratio of 5 litres fluid milk per kilogram of NFDM, because it is presumed that NFDM has a lower value to the recipients than fluid milk.

TABLE 6. Indices of Cost-Effectiveness (taking the values for wheat as 100)

Index of a values Wheat Corn Rice Oil Milk(fluid)
Egypt 100 128 232 304 2,080
India 100 130 102 378 672
Pakistan 100 117 146 181 575
average 100 125 146 285 922
Index of energy per dollar in food* 100 114 59 95 63
Index of protein per dollar in food* 100 83 32 0 190

*Derived from energy and protein content per kilogram, respectively, and the costs reported in table 5.

Second, as mentioned earlier, some commodities may be considered "surplus" by the government providing food aid. Therefore, some commodities may be charged to the food aid programme at far below market prices while others are charged at their full market price.

Third, the relative cost of delivering commodities for food aid may be quite unlike the relative cost of handling commodities in commercial channels. For instance, the cost of shipping wheat in bags for food aid to Asia is about US$150 per ton while the cost of shipping wheat in bulk for commercial transactions is US$50 per ton. However, the mode and cost of shipping vegetable oil is more or less the same in commercial and food aid transactions.

Finally, some foods may be available to food aid recipients at their normal retail value and some at a concessionary price (if they are obtained, for instance, through a ration shop) and others should be valued at their farm-gate selling price (if the food delivered by the aid programme to a farmer results in more sales rather than reduced purchases).

As a consequence of a combination of these circumstances, the a values and hence the relative nutritional cost" effectiveness of commodities used in different food aid programmes can be expected to vary a great deal. Table 5 shows retail prices of selected foods for 1981, imputed from data reported for 1975 in Egypt, 1980 in India, and 1979 in Pakistan and adjusted for inflation (16). The cost of the same commodities to the PL 480 Title II programme shown refer to acquisition costs plug average ocean shipping charges to the three countries. The a values shown in brackets are the ratios of the retail prices to the costs of the commodities in the PL 480 Title II programme.

Table 6 shows the relative monetary and nutritional cost-effectiveness of the different commodities based on their values and compares them with the relative energy and protein cost-effectiveness of the commodities when imputed from their energy and protein contents. The cost of commodities in all cases are as shown in the last row of table 5.

By the a -value criterion, a dollar spent on milk (NFDM) is twenty times and oil is three times as cost-effective as wheat in a supplementary food aid package provided by PL 480 Title II to consumers in Egypt. In a similar programme in India, milk would be six times and oil four times as costeffective as wheat. Such dramatic differences are not common, but they do illustrate the point. Milk has a high ? value because it is a genuine surplus commodity for which the PL 480 Title II programme is charged far less than its market value. Moreover, in Egypt basic staples are highly subsidized and in India the oil/wheat price ratio is far above the international oil/wheat price ratio.

It is particularly important to note how the ranking of commodities would be different if the (incorrect) nutritional-unit-per-dollar-in-food criterion rather than the a-value criterion were used. Even with the highly concessionary cost of milk, wheat rather than milk would be supplied if the programme's primary objective were to enhance the energy supply in the diets. Obviously, oil would be the least desirable commodity if the programme's primary objective were to improve the protein content of the diet.

The empirical evidence provided here serves to illustrate that the a values of commodities can range from far below to far above unity. They are likely to be different among countries and even among different regions within countries, and they will also change over time as relative prices change. The cost, i.e. the denominator in the a ratio, also depends on the objective of the programme. By imputing only the acquisition and shipping costs of commodities as we have done in the above calculations, we have assumed that the objective is to maximize the nutritional effect from a PL 480 Title II programme dollar. The PL 480 Title II programme does not pay for distribution costs within the country. These costs are borne by voluntary agencies, such as Catholic Relief Services, and host country governments. Nor have we considered the costs of handling the commodities by the recipient households. To the extent that these additional costs are included, all a values will be lower. Moreover, the relative values of the commodities may be different to the extent that the costs left out are not in the same proportion to the included costs for all commodities.


Contents - Next