Contents - Previous - Next

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

A framework for tracing policy effects on intra-household food distribution

Shubh K. Kumar
International Food Policy Research Institute, Washington, D.C., USA


In order to develop an understanding of policy impact on intrahousehold food distribution, it is imperative first to understand the dynamics in the functioning of households. Unfortunately, not much attention has been given to this by two major disciplines that have to date been involved in nutrition policies. Neither nutritional science nor economics has addressed the question seriously. At the very basis is the need for an explicit recognition of the diversity in structure, composition, and function of households under various socio-cultural and economic environments. Interesting insights have been provided by anthropological investigation into the various dimensions of household units. The paper by Messer in this issue shows some of the variations in household-provisioning mechanisms adopted by household units under diverse sociocultural conditions. A recent review by Dwyer (1) shows, in addition, the need for a gender-differentiated approach to household behaviour.

The framework postulated here has as its core the nature of the provisioning mechanisms adopted by household units It is suggested that these provisioning mechanisms are conditioned or even derived (the direction of causality here is not essential for the framework) from the social and cultural milieu, including the theological-legal environment. In addition, change in economic factors can influence these provisioning mechanisms directly as well as indirectly by inducing changes in the socio-cultural scene.

There are two possible ways in which intra-household food distribution is influenced by the household provisioning mechanism. First, by determining intrahousehold distribution of capital (both physical and human), division of labour, and resource/income generation it affects intra-household control of resource allocation (including time). Second, it determines preference functions for intra-household investment in nutrition and health. This in turn provides the basis for an intergenerational transfer and exchange of resources and the start of the process all over again.


Access to Physical Capital

Ownership of or rights to land or other assets are a foremost determinant of how a household establishes its entitlements, since food and cash are primarily derived from agriculture for the majority of developing country populations. Socio-cultural and theo-legal factors are perhaps most important in traditional societies. In the process of change, land legislation and development programmes can introduce new ways of securing access to land. Ownership and rights of individuals in a household to land or other assets influence how households derive entitlements, how they use their labour in the process, and eventually what real income is available and how its allocation is controlled. A great deal of variation exists in conditions that have been documented across Asia, Africa, and Latin America. Those who have access to physical capital are likely to be the ones who derive entitlements from it for other household members (with their labour input as well) and are likely to be primarily engaged in its allocation.

Access to Human Capital

Human capital plays an important role in household provisioning. Both education and nutrition/health levels of individuals in a household are relevant. The levels desired and obtained are likely to be determined by employment opportunities and socio-cultural factors that influence who can or cannot perform certain activities and by returns expected for the household by alternative types and levels of human capital. Health, nutrition, and educational programmes can also influence levels desired by reducing the cost to households of making these investments. Children embody potential human capital for a household unit, and consequently household resources will be spent in raising them, which, under conditions of scarce resources, will depend on their expected benefit to the household.

When level of physical assets is lower, then the importance of human capital in household provisioning increases as a means of securing income. Depending on the type of employment opportunities and their relative rates of return, different types of strategies may be adopted by households.

Income Generation

Households use their combination of physical and human capital in the production of income. In agriculture, decisions are made on the choice of crop mix and area, amount of household labour used, division of labour, choice of technology and inputs, and disposal of products. Policies in agriculture can influence each of these processes that lead to the production of real income in agriculture. Similarly, the production of non-farm income by self employment or wages can be influenced by policies. Finally, time for converting the goods or cash represented by income into items of direct consumption or use by households is an important component of real income. Food consumption of children is especially dependent on mother's or substitute's time, and when it is short could influence intra-household food consumption.

The division of household labour among production/ income activities and for its processing or consumption related activities is an important factor in the outcome of household provisioning. Not only does it determine the size and mix of components of income available to the household, but perhaps as importantly, it may provide a basis for control in allocation of income. It may then be argued that control over allocation of household income is influenced by (a) ownership of assets, physical or human, required in provisioning, and (b) time allocation in (i ) the production and (ii) the processing of income for household consumption.

It is possible that there is a hierarchy of rights based on ownership and time spent among various household members that manifests itself in the allocation of income and intra-household distribution of consumption.


Little is known about how households choose to allocate income among various goods and household individuals. The study of nutrition in today's Third World countries is a field of investigation barely 50 years old. It is characterized by paradigms of household behaviour that were at best proposed by the early Western scholars from their own experience. In this paradigm, the household is nuclear, with a male breadwinner and a female "homemaker" who makes decisions on how to allocate income for food consumption. Any deficiencies could be explained by ignorance of the finely tuned requirements for nutrients by the body. Consequently, those household members with proportionately greater nutritional needs, such as infants, children, and pregnant and lactating mothers were expected and found to receive a lower share of their food consumption requirements. Singular emphasis on nutrition education programmes aimed at women were a natural outcome of these early theories in the 1950s and 1960s.

In the case of economic analysis of consumption there has been little scope for studying intra-household questions. Neither micro- nor macro-economic theories have looked within households. Micro-economic theories are geared to individual behaviour of a firm or household whether in making production or consumption decisions. Intrahousehold distribution of food can only be a manifestation of a single combined household utility or preference function. No attention has been paid to how this is derived, and little has been given to "the profound problem in, on the one hand, internalizing all the family members' satisfactions in one utility function and, at the same time, using this same utility function to determine the number and 'quality' of the family members themselves" (2). While accumulating anthropological evidence suggests a more complex phenomenon in the allocation of household income that cannot be examined with only a single household utility, there may be some scope within a single utility function for examining the fertility-nutrition trade off in aggregate child investments by a household (3).

It seems clear, however, that a better understanding of the intra-household dimension of household preference functions is necessary for understanding intra-household food distribution and how policies can influence it.


In figure 1 a simplified framework is proposed as a basis for examining policy implications on household provisioning mechanisms (intra-household factors). Among the components in the process of household provisioning we can identify a socio-cultural dimension and an economic dimension. Intra-household access to physical and human capital and the sexual division of labour can be seen as being largely a part of the socio-cultural dimension (this will hold in respect of the sexual division of labour for larger population groupings, even though subgroups may respond to economic pressures by modifying their sexual division of labour), while the actual activities undertaken in the income generation process are part of the economic dimension in that they are primarily influenced by such policies. Development programmes and policies can influence any of the components of household provisioning.

In addition to influencing the generation of household incomes, policies can also influence preference functions involved in the allocation of household resources. Education, advertising, etc., are the most obvious examples of how preferences can be modified. There is an obvious

intra-household dimension in preference functions since it is linked to decision-making. Further understanding of intrahousehold decision-making can provide a valuable basis for examining preference functions. Recent work in Nepal has shown that decision-making in allocation of household income is closely related to the individual's role in the generation of the income (4).


How an individual policy measure interacts with household provisioning strategy will determine the outcome on intrahousehold distribution. The same policy measure can have very different repercussions under varying circumstances or characteristics of how households provide their entitlements. Consequently, policy analysis needs to be made within a stated context. In order to do this, more attention must be given to understanding better the underlying socio-cultural and economic dimensions of household provisioning set out in figure 1. Ultimately, the objective of the policy analysis should be:

Figure 1


1. Dwyer, "Women and Income in the Third World: Implications for Policy," The Population Council, International Programs, Working Paper no. 18 (New York, 1983).

2. M. Nerlove, "Toward a New Theory of Population and Economic Growth," in T. Shultz, ea., Economics of the Family (University of Chicago Press, Chicago, 1974), pp. 527-545.

3. G. Becker and H. Lewis, "Interaction between Quantity and Quality of Children," in T. Schultz, ea., Economics of the Family (University of Chicago Press, Chicago, 1974), pp. 81-90.

4. M. Acharya and L. Bennett, "Women and the Subsistence Sector: Economic Participation and Household Decision Making in Nepal," World Bank Staff Working Paper no. 526 (Washington, D.C., 1983).


Estimating the nutritional impact of food policies: A note on the analytical approach

Per Pinstrup-Andersen
International Food Policy Research Institute, Washington, D.C., USA

Food policies are rarely assessed for their effect on human nutrition. Yet, such assessment, if properly done, would facilitate the incorporation of nutrition goals into the choice and design of these policies with the likely result of improving nutritional impact. The degree to which expected nutritional effects should influence the choice and design of policies with multiple objectives would ideally be determined on the basis of: the importance of the nutritional problem relative to other problems towards which a given policy is aimed and the cost of achieving a certain nutritional improvement through the particular policy relative to the cost of achieving such improvement by means of the least-cost alternative policy or programmes. Clearly, certain nutrition problems are most efficiently dealt with through direct nutrition intervention and/or health programmes, while others should be approached through broader food policies. However, in order to select the most appropriate approach it is important to understand not only how the various programmes and policies affect nutritional status, but also how the impact is transmitted.

Many, although by no means all, past evaluations of nutrition programmes have attempted to assess the impact on selected indicators of nutritional status while partially or totally ignoring the intermediate steps or relationships that brought about the impact as well as other factors that might have exercised impact simultaneously with the programme being evaluated.

If the sole purpose of a given study is to evaluate, ex post facto, the impact of a particular programme on the nutritional status of a particular group of people during a particular time period and within certain environmental influences, such an approach may suffice. A caricature of an evaluation of this nature is shown in figure 1. No attempts are made to analyse the mechanism by which the impact is transmitted, i.e., it is unknown what happened inside the "black box." The study merely compares a situation where the programme is present to a situation where it is absent, either over time for the same population group or at a given point in time across population groups.

But if analyses of past and current programmes are to be truly useful for the choice and design of new programmes and policies and modifications or termination of current ones, it is necessary to know not only by how much but also how nutritional 'statue is influenced by the various programmes. We must understand why some programmes and policies are more or less effective than others. This requires understanding of the mechanisms by which nutritional status is affected and how this mechanism and its key components link immediate programme effects to nutritional impact. The process components must be identified and their interaction understood. The job of evaluating a given programme then becomes one of tracing programme impact through the relevant processes while estimating the impact on each of the relevant components.

FIG. 1. The "Black Box" Approach to Programme and Policy Evaluation. Often all the multiple factors and processes through which programmes and policies influence nutritional status are ignored and only nutritional impact is measured.

The key elements in most or all of the policies and programmes are relatively few, and many of them are common to different kinds of policies and programmes. On the other hand, there is an almost unlimited number of possible policies, programmes, and programme combinations that may be designed. If we understand how a given policy and programme affects the key elements and in turn how these key elements affect nutritional status within clearly specified or identified environments, we can design effective policies and programmes by selecting and combining the elements that are most appropriate for that environment. The number of key elements in any of the programmes is obviously smaller than the number of possible combinations of these elements in programmes. Thus, it is more effective to study the elements than the programmes. But to assure immediate programme relevance, a study of the elements should take place within a policy or programme framework.

FIG. 2. Factors and Processes Influencing Nutritional Status. This simplified analytical framework for determining the processes by which programmes and policies affect nutritional status replaces the "black box" of figure 1. The items marked with an asterisk (*) are sometimes used as indicators of nutritional impact.

Furthermore, the effects of factors other than those directly affected by a given policy or programme, including those falling into what is here called the environment, must be understood and quantified. In the case of a food price policy or a supplementary feeding programme, for example, it is important to know not only how nutrient intakes by the malnourished are affected, but also how the utilization of the additional food is affected by sanitary conditions and health factors, and how changes in sanitary conditions, health, and educational levels of parents may contribute to a better utilization of the food. While such factors may be assumed constant for the purpose of the evaluation of a single programme, such an assumption will not be valid in the case of using the project findings for the design of new policies and programmes under different circumstances.


The processes that determine the nutrition effects of programmes and policies are complex. This complexity, together with insufficient data and lack of appreciation for the utility of generalizing study findings beyond a particular programme, population group, time period, and environment is the most common reason why evaluation studies often avoid an analysis of the processes, thus leaving the black box partially or totally untouched. To change this situation, an appropriate analytical framework must simplify the complexity by identifying the most important factors, relationships, and data needs and demonstrate that these factors and relationships are not programme-specific and that empirical findings may be generalized across programmes and policies. In the context of previous jargon, the black box must be opened, but rather than emptying everything out-and thus overwhelming even the most ambitious researcher or evaluation officer-we must select the most important parts for study and clearly identify what is left inside.

A set of factors and relationships that might be used to make up such a simplified analytical framework is proposed in figure 2. The intent of the framework is not that all programme and policy assessments should estimate empirically every causal relationship shown in the figure. Rather, the purpose of the framework is to help identify the key factors and relationships for a particular programme and policy and the data needs for tracing the programme/policy effects through these relationships to the final impact on the specified nutrition indicator.

The framework shown in figure 2 contains three main factors through which food and nutrition programmes and policies may eventually influence nutrition: (a) the ability of households with malnourished members to acquire food, (b) household food acquisition behaviours, and (c) intrahousehold distribution of food. These three factors may be influenced by programmes and policies through changes in a number of other factors (fig.2) and may, in turn, influence nutrition through changes in the acquisition of food by households and individuals and the utilization of food by these individuals.

Ideally, analyses would focus on the estimation of the coefficients that explain the magnitude and strength of each of the causal relationships found in the processes of interest, thus linking quantitatively the various steps through which a given policy or programme affects nutritional status.

The most appropriate specification of the analytical model depends on a number of issues, including (a) the type of policy or programme and how it is expected to influence nutritional status, i.e., the first step in the process, (b) the population groups whose behaviour is most important in determining nutritional effects, and (c) which of many nutrition indicators will be used. Each of these issues is briefly discussed below.

Policy and Programme Types

Food and nutrition policies and programmes may influence nutritional status through their impact on any of the factors shown in the top row and the left column of figure 2. Food supply and rural development programmes and policies influence human nutrition primarily through changes in the ability of malnourished households to acquire food. This ability is influenced through changes in food availability on the farm, food prices, and rural incomes as well as the fluctuations in these factors.

While household food acquisition behaviour determines the extent to which ability results in actual household food procurement, these policies and programmes may also have a direct impact on behaviour through changes in income composition, intra-household income and budget control, and women's time allocation. The nutrition impact of consumer oriented food price policies (e.g., food price subsidies), food-linked income transfers (e.g., food stamp programmes), and food transfer programmes also occurs primarily through changes in household ability to acquire food.

The impact of a particular programme depends on its nature, including its limitation to certain commodities, rations, and/or target groups. If programme rations are inframarginal with no purchase requirements, the effect is expected to be determined by the real income embodied in the transfers. In other cases some substitution between programme commodities and other goods is expected to take place. If the intra-household control of incomes from these programmes is distributed differently from that of other incomes, and if the marginal propensity to spend on the particular foods varies among household members, a direct impact on household behaviour may occur. Food transfer programmes are frequently targeted to certain household members, e.g., malnourished children or pregnant women. Except for certain programme types, such targeting is likely to fail because of adjustments in food consumption by all members and/or adjustments in food acquisition from other sources. Such failure is usually referred to in the nutrition literature as "leakage." Whether, in fact, leakage of this nature should be considered a serious failure of the programme is, of course, open to debate. In spite of large leakages, most food supplementation programmes still attempt targeting on particular individuals by imposing various types of restrictions on the use of transferred food.

Nutrition education and awareness programmes influence human nutrition through household acquisition and allocation behaviour related to food, health, and sanitation. Perceived needs for food, nutrients, and health and sanitary services may change, as may child care and breast-feeding practices. Food fortification programmes may influence nutrition through the ability of the household to acquire deficient nutrients. Two opposing effects may occur: first, fortification may result in a higher content of particular nutrients in a given quantity of food, and second, it may result in a higher price per unit of food. Similar effects would be expected in programmes using formulated foods.

Population Groups and Their Behaviour

The nutrition impact of a particular programme or policy depends not only on the programme or policy design, but also on the behaviour of the various groups or individuals acting within the processes illustrated in figure 2. These "actors" may enhance or reduce the intended nutrition impact. In some cases they may purposely oppose programme objectives. At the time of programme design, ignoring the possibilities for conflict between programme goals and the goals of the various actors is likely to lead to disappointing programme results. Furthermore, as substantiated below, programme evaluation that assumes that programme goals are shared by all relevant actors is likely to add little to our understanding of why programmes perform as they do. The principal actors are:

The importance of each of these actors varies among programme and policy types. The behaviour of households, programme implementation bodies, and the local power structure are of particular importance for all programmes and policies and are further discussed below.

Household Behaviour

In-depth understanding of household behaviour as it relates to acquisition and intra-household distribution of food is essential to successful programme and policy design. Household food acquisition and allocation behaviour determines the extent to which changes in household ability are reflected in food intakes by the malnourished.

Demand parameters such as commodity-specific price and income elasticities go a long way in explaining or predicting the relationships between changes in household food acquisition ability and the resulting change in household food consumption. Since the concern is for households with malnourished members, the parameters must be relevant for these households. In societies with a very skewed income distribution and considerable malnutrition, average estimates are not likely to represent the behaviour of households with malnourished people. Thus, the relevant parameters must be estimated by income group. Reliable estimates of such parameters are of recent origin and their use in food policy design has been very limited indeed. During the last few years, however, there has been a considerable increase in research efforts to estimate demand parameters by income stratum.

Data scarcity is the principal barrier to direct estimation of such disaggregated parameters. Cross-sectional data sets may provide an acceptable basis for the estimation of income parameters and, thus, the income effects of price changes. However, they may not serve for reliable estimation of price parameters or the substitution effect of price changes unless they refer to various points in time or various geographical locations and therefore provide for sufficient and relevant price variation. Periodic and directly comparable household surveys over a number of years would alleviate the data constraints.

In addition to reliable estimates of demand parameters related to household incomes and food prices, the ability to predict with a high degree of precision household food acquisition and allocation behaviour and the household's reaction to food policies and programmes depends on a better understanding of other behavioural factors. Changes in the demand for women's time, intra-household budget control, income composition, the range of foods and services competing for the household budget, the degree to which incomes are considered transitory or permanent, and other changes brought about by public policies and programmes all influence behaviour. Also, structural changes such as rural to urban migration or transformation of subsistence farming areas into an exchange economy may have a direct impact on household food acquisition and allocation behaviour and thus either make existing income and price parameters invalid or incapable of explaining household food acquisition behaviour.

Past research on the impact of these factors on food acquisition and allocation is relatively limited. It appears plausible, on the basis of available evidence, that a number of unexplained behavioural issues reflect the influence of factors such as those mentioned above. However, additional empirical research is needed to provide information in this area that will be useful for policy and programme design.


The process by which national government programmes and policies are implemented or translated into action at the local level is a key ingredient in their success or failure. Yet little is known about the determinants of a "successful" implementation process. To date, relatively few efforts have been made to analyse how the black box at the local level influences the outcome of specific national programmes. Until recently it was erroneously assumed that the black box at the local level was largely passive in nature, i.e., content to replicate policy decisions made at higher levels. However, recent studies have suggested that in many cases what the national government orders or commands is not necessarily what the local level actually does. It may indeed be true that local-level forces lack the power and resources to determine national policy. Yet, since national decisions at the local level must be implemented by or through local forces, these forces possess an important power to constrain or deflect the character of national programmes. Local-level forces act as a critical filter or prism capable of screening, altering, or even impeding the implementation of national health and nutrition programmes.

One of the central problems associated with policy implementation at the local level is that of leakage. One type of leakage occurs when nutritional programmes designed for lower income groups fail to reach their targeted groups. Leakage is produced by a variety of factors, such as weaknesses or inadequacies in the delivery mechanism at the local level. Yet the net effect of this type of leakage is always the same, namely a significant difference between the promise of national government programmes and the reality of local-level delivery.

In many rural areas the problem of leakage is directly related to the dynamics of local power structure. The skewed distribution of land and economic resources in these areas means that many poor peasants are economically and socially dependent upon the patronage services provided by rich peasants. These services include the provision of agricultural employment, emergency loans, and inter cessionary services with officials. On the one hand, such patronage services are quite important because they ensure the daily survival of large numbers of poor peasants. Yet the importance of these patron-client ties serves to complicate greatly the process of local-level project implementation. By virtue of their control over human and material resources at the local level, rich peasants expect to dominate all local delivery mechanisms established by the national government officials whose power is typically not grounded in the economic structure of the local community.

In most cases, local government administrators simply lack the resources and will-power needed to assure that programme benefits reach the target groups. At the same time, poor peasants, fearful of alienating the rich peasants on whom they depend, are most reluctant to press for access to national health and nutrition programmes. From the standpoint of the typical poor peasant, national programmes and policies that come and go are not to be trusted or pursued at the expense of antagonizing the local elite. As a consequence, the benefits associated with national programmes and policies are often captured by rich peasants and their favoured clients.


The choice of indicator of nutrition impact varies among studies and is a function of (a) the particular programme or policy being assessed, (b) data availability, (c) cost and time considerations, (d) the disciplinary orientation of the researcher or evaluation officer, (e) implied or assumed relationships among process components, and probably a number of other factors.

The choice of indicator is reflected in the degree of penetration of a particular study into the process, as illustrated in figure 2. In general, data requirements and magnitude of the study increase with greater penetration. The least penetration is illustrated by the use of programme and policy impact on total food availability as an indicator. Although grossly ineffective and often misleading, this indicator is frequently used in food production programmes and policies. A slightly greater degree of penetration is provided by the impact on the ability of households with malnourished members to acquire food. While a greater improvement over total food availability, this frequently used indicator is still unlikely to be closely associated with the ultimate criteria because it ignores factors and relationships downstream in the framework, e.g., the effects of household food acquisition behaviour (and possible programme input on this behaviour), food distribution among well and malnourished members, and health and sanitation issues.

Use of actual household food acquisition as a nutrition indicator is a further improvement because it takes into account household behaviour. This indicator is widely used in assessments of food policies and as an indicator of existing malnutrition and its distribution in a given population. Use of estimates of intakes by malnourished individuals provides yet another improvement over household food acquisition data. However, although sometimes used to evaluate food supplementation programmes, particularly those targeted to particular household members, the use of such estimates is not nearly as frequent as estimates of total household food acquisition because of difficulty in obtaining reliable data.

Anthropometric measurement of growth and development is a very commonly used indicator of the impact of nutrition intervention programmes on the nutritional status of children. This is a relatively convenient approach that, if currently applied, yields reliable estimates of the extent to which the physical development of a particular child deviates from the norm. However, except for severely malnourished children, it may be difficult to separate the effect of nutritional improvements from other effects such as genetic variation. As opposed to the various food-related indicators mentioned above, anthropometric indicators reflect both food- and health-related factors.

The activity level of an individual is another possible indicator of nutritional effect. This indicator is based on the premise that individuals suffering from insufficient energy intakes tend to reduce energy usage by lowering the activity level. Such lower levels may affect the development of children and reduce labour supply and productivity. Except for a few studies of the impact of food supplementation on labour productivity, the use of this indicator has been rare. The limited usage is due, at least in part, to the severe difficulties of measuring activity levels with sufficient accuracy.

The rate of mortality and morbidity have also been used as indicators of nutritional impact. They are probably good indicators of the impact on severe malnutrition provided that (a) programme impact can be separated from that of other factors, (b) the sample from which data are drawn is sufficiently large, and/or (c) these rates were relatively higher before the programme began. Finally, clinical and biochemical methods are sometimes used as indicators. While the former are used mainly for severely malnourished individuals, the latter are sometimes used in relation to extensive household surveys.

Contents - Previous - Next