Contents - Previous - Next

Methodology of field studies related to socioeconomic effects of chronic energy deficiency


* Institute of Nutrition of Central America and Panama, P.O. Box 1188, Guatemala City, Guatemala.

1. Introduction
2. Conceptual framework
3. Study designs
4. Human capital studies
5. Economic behavioral studies
6. Participation of the community in research
7. Concluding remarks

1. Introduction

Chronic energy deficiency (CED) is a poverty phenomenon. Essentially, it reflects an inadequate access to food at the household level, and - at the individual level - may also reflect intra-household food distribution practices, as well as the effect of different disease states negatively affecting the biological utilization of ingested food. Poverty is a multidimensional problem, and thus the solutions to be found must also be multisectoral in scope, and may require structural changes. Research should provide results which can be fed directly into decision-making and action-taking processes, which in turn lead to reduced poverty. It is within this framework that we must undertake research related to CED and its consequences.

The use that can be expected to be made of the research results should determine the research questions asked and the methodology applied in obtaining answers to those questions. In fact, those who actually utilize and internalize the research results (such as policy-makers, program-managers, community leaders) should fully participate in the formulating and prioritizing of the research questions, to avoid the formulation and testing of hypotheses which are scientifically interesting, but have little practical applicability.

CED may have negative social and economic consequences. This implies social costs which will accumulate over the time-period that CED persists. Thus, urgency is involved inasmuch as results from research contribute effectively to actions which improve the social and economic conditions which produce CED. Research results produced for advocacy purposes may not be directly relevant in the formulation of policies and programs. This requires the full participation of decision-makers and resource-managers at different levels (central, local, community) to ensure that research results are directly transformable into specific actions, thereby reducing over time the social costs of CED (and increasing the real returns from CED research). The basic challenge lies in putting this idea into practice.

In many developing countries the financial and technical capacity of the public sector to provide integrated and multisectoral solutions to poverty problems is increasingly being reduced. Important reasons include inflation, high levels of foreign debt, and rapidly rising public sector budget deficits. Thus, making research results available only to public sector policy-makers and program-managers may yield little in the way of effective solutions to poverty problems, both in the short and long term. Action programs must be undertaken using resources available through non-governmental, popular and community organizations. Research programs should produce results, therefore, useful to those non-public sector organizations that manage resources and deal directly with poverty problems.

Field studies by their very nature involve communities. They offer opportunities for bringing together researcher, decision-maker and resource-manager to find answers to questions jointly. The classical models of obtaining data in the field and conducting intervention studies have always raised ethical questions, since the study communities rarely ever benefit directly from their participation in the research effort. Full participation by the communities researched may be an effective way of having research results from field studies transformed rapidly into economically feasible and culturally relevant actions leading in turn to reduced poverty, at the same time ethically justifying the research undertaking.

In Latin America newly emerging social processes are bringing into existence more and more popular and community organizations which directly assume responsibility for reducing the poverty problems of the population groups which they represent. Thus, opportunities for participatory research by communities exist, and participatory research methods should form part of field study methodologies.

In the present paper we focus on methodological issues of socioeconomic studies related to CED, identifying basically two kinds of studies: a) human capital studies, and b) economic behavioral studies. Both have a common conceptual framework, but ask different questions and have different analytical designs. Each is subject to similar limitations which lead to imprecise estimations of the economic consequences of CED. A number of recommendations for future studies are made, including studies which use participatory research methodologies.

2. Conceptual framework

Figure 1 presents a conceptual model that attempts to incorporate some other models which have previously been presented elsewhere. For example, at the heart of the diagram we have included the model presented by SPURR (1984). Levels of physical activity and nutritional status interact with each other, and each is a determinant of physical work capacity. In the Spurr model, physical work capacity is the determinant of worker productivity, an economic outcome of the biological process.

As VITERI et al. (1981) have pointed out, however, the biological process interacts with social, economic, cultural and motivational factors to produce specific socioeconomic outcomes. So, for example, the labor demand that a rural worker faces will influence his daily activity pattern and, thereby, his physical work capacity. At the same time, a change in the labor demand which is reflected in a change in wage-rate may provide a (dis-)incentive for actual worker productivity, independently of its effects on the biological process.

It is also well to remember that an individual is a member of a household unit, and that there is likely to exist interdependence within the household in individual food intake levels and activity patterns. This means that the economic effect of a change in the food intake level of one household member may not be confined to that one household member. Thus, to capture the socioeconomic effects more fully, a more valid unit of analysis is the household in any case.

3. Study designs

A brief discussion of designs of field studies is necessary to understand the measurement limitations of the socioeconomic effects of the CED studies presented in the following sections.

Basically, two types of study designs have been employed in relevant field studies: cross-sectional, and longitudinal or time-series. In cross-sectional designs the investigators attempt to establish a statistical association between indicators of energy deficiency status and economic outcomes, such as worker productivity. Examples of these types of studies are: (a) a study of the relationship between body size (body weight, height and lean body weight) and work output among factory workers in India (SATANARAYANA et al., 1977); (b) studies of the relationship of body composition parameters, physical working capacity, and work output of sugar cane cutters in Colombia (SPURR, BARAC-NIETO and MAKSUD, 1977), Tanzania (DAVIES, 1973) and Jamaica (HEYWOOD, 1974). The study by VITERI (1971) and co-workers of agricultural laborers represents a variant of the cross-sectional design; it is what COOK and CAMPBELL (1979) call the posttest-only design with non-equivalent groups, in which a cross-sectional comparison is made between two groups, one of which has experienced a premeasurement change in the variable of interest, in this case: daily energy intake levels.

Figure 1. Chronic energy deficiency: Conceptual framework.

Cross-sectional studies cannot establish that chronic energy deficiency causes certain socioeconomic outcomes, even when holding constant all other relevant factors. As we shall see below, human capital studies have employed exclusively the cross-sectional design, since longitudinal studies are not feasible for this type of study.

Longitudinal studies are usually more definitive: depending on the (quasi-)experimental design that they employ, stronger inferences about cause and effect can be made than in cross-sectional studies. Energy supplementation studies have employed longitudinal designs; some of the relevant studies involved (a) rice farmers in India (BELAVADY, 1966); (b) road construction workers in Kenya (WOLGEMUTH et al., 1982); and (c) sugar cane cutters in Guatemala (IMMINK et al., 1986; IMMINK and VITERI, 1981). The first two studies involved relatively short supplementation periods; in all three studies comparison groups were present, and their quasi-experimental designs can be identified as: pretest-posttest control group design in (a) and (b), or multiple time-series design in (c) (COOK and CAMPBELL, 1979).

There have been few longitudinal studies which have used naturally (seasonally) occurring variations in food availability and/or in energy needs as the experimental variable to measure socioeconomic outcomes. Seasonality has been considered in relation to energy balances, that is, energy produced in relation to energy consumed in agricultural production during different seasons (BAYLISS-SMITH, 1981; HASWELL, 1981). In a short review article, LONGHURST and PAYNE (1981) have documented marked seasonal variations in food availability occurring in rural areas of Africa, particularly where there is dependence on a single harvest. Other studies have demonstrated that labor requirements may change in agricultural production (CLAY, 1981) or with shifts from subsistence farming to cash crop production (GROSS and UNDERWOOD, 1971). In one study in rural Philippines, seasonal effects may have confounded the measured effects of energy supplementation on energy expenditure levels, as the subjects gained weight with supplementation and the energy cost of activities decreased (DE GUZMAN et al., 1985). Changes in labor requirements during a harvest season were shown to produce similar productivity responses among groups of Guatemalan sugar cane cutters with different body composition without producing marked changes in body composition in any of the groups (IMMINK et al., 1987).

Figure 2 Agricultural cycle of traditional and diversified farmers in the Highlands of Guatemala.

Figure 2 is shown to illustrate how seasonal variations in food availability and labor demand may be introduced as experimental variables in longitudinal studies. It is based upon a small survey conducted among farmers in the Guatemalan Highlands. For traditional farmers a critical period of high labor requirements and low food availability can clearly be identified, as well as periods of relative food abundance with changing labor requirements. This allows for a design which tests for between- and within-period effects on economic outcomes of short-term changes in energy availability and needs. Incorporating for comparison a group of diversified farmers who face a relatively stable level of food availability and labor requirements during the whole agricultural cycle, strengthens the design, allowing the confounding effects of external factors to be separated from the experimental effects. This design can be expanded to incorporate energy supplementation among traditional farmers during, e.g., the "critical" period.

In the following sections we distinguish between human capital and economic behavioral studies, and indicate how different study designs may affect the estimates of the economic consequences of CED in developing countries.

4. Human capital studies

4.1. Methodological aspects
4.2. Analytical limitations

Human capital studies view the individual's stock of human capital as associated with his/her long-term nutritional intake level. Improvements in nutritional intake augment the stock of human capital, i.e., represent human capital formation, and food is seen as an investment good and not, as is more usual, as a consumption good.

Human capital theory was developed during the 1960's and 1970's by economists such as SCHULTZ (1971), BECKER (1975) and others, as it became clear that what explained economic growth was not just factors of production such as land, labor services, physical capital and technological change, but also the quality of labor services, i.e., the stock of human capital. Human capital theory was later incorporated into a more generalized theory of the allocation of time and goods over the life cycle (GHEZ and BECKER, 1975).

Investments in schooling, on-the-job training, health and nutritional status all increase an individual's human capital, which has both a physical and a mental dimension. An individual's stock of human capital is subject to deterioration over time due to factors associated with age, and to obsolescence of skills and knowledge. Investments in human capital may be complementary, and may produce a synergistic effect, as for instance when nutritional improvements take place in children attending school, improving the return on educational investments.

Studies which employ the human capital approach basically produce results for policy and program advocacy purposes, in that they attempt to measure what the rate-of-return is on different levels of investment in human capital. The measurement approach involves the assumption that earnings (or some other indicator of an individual's productivity) are proportional to an individual's stock of human capital. We can then operationally construct profiles of earnings by age for individuals with different stocks of human capital over their productive life cycle. By comparing, at a point in the life cycle, the present value of the earnings differential associated with different levels of investment in human capital, with the present value of the economic cost associated with higher levels of investment, we obtain the internal rate-of-return of different investment levels. Relevant CED studies have applied this analytical approach in the case of adult workers (IMMINK, VITERI and HELMS, 1982), or in relation to higher levels of intake in children and associated earnings profiles during the productive life cycle (BELLI, 1971; SELOWSKY, 1978; SELOWSKY and TAYLOR, 1973). The results from these studies are limited, since data on the economic costs associated with long-term higher intake levels were not available and thus internal rate-of-return analysis was not possible.

Table 1. Methodological aspects of human capital studies

Methodological aspects:

Cross-sectional design (within occupation)

Longitudinal studies (with or without supplementation)

1. Experimental variable

Variation in nutritional status (body composition, energy intake) encountered in study subjects

Change induced in nutritional status (body com position, energy intake)

2. Study subjects

a. Economically active adult population (within- generation analysis)
b. Cohort of children and adult group from same socioeconomic stratum(intergeneration analysis)

a. Economically active adult population (within generation analysis)

3. Key economic indicators

a. Total income from productive activities/u.t.
b. Earnings/u.t.
c. Number of work units/ u.t., valued at market prices

Same (a, b or c) Economic costs of supplementation (internal rate of return analysis)

4.1. Methodological aspects

Certain methodological aspects of human capital studies have been summarized in Table 1. These studies in the past have mainly employed a cross-sectional design, using as the experimental variable the variation in nutritional status encountered in the study subjects. Body composition indicators or daily energy intakes have been used to represent levels of chronic energy deficiency.

The few relevant studies which are available have used economically active adults as subjects within a homogeneous occupational group. In one case, a parameter of childhood nutrition (adult stature) was associated with adult productivity in order to quantify human capital formation via nutritional investment during childhood (IMMINK et al., 1984). In another study, it was demonstrated that in a cohort of children protein-energy intake levels were related to their mental development. In a cohort of adult workers from the same social stratum as the cohort of children, mental development was related to productivity (SELOWSKY and TAYLOR, 1973). The general conclusion of the study was that undernutrition in children represents disinvestment in human capital.

Alternative key economic indicators which can be used for the analysis are (a) total income from all productive activities, including the imputed market value of home production activities; (b) total earnings, in wage employment, or (c) number of work units performed and valued at market wage rates. The first indicator is preferable though costly in terms of data requirements. Direct questions regarding income earned usually result in a systematic under-reporting1.

1 This may not be serious it the reporting bias is a constant proportion of true income, but this is usually not the case.

Alternatively, as a proxy of earned income, total expenditures (plus net changes in savings if relevant) are often used.

Longitudinal studies which involve an induced change in nutritional status, either by means of supplementation or naturally occurring change in energy availability, offer an opportunity to undertake internal rate-of-return analysis, assuming that the measurement period is sufficiently long to measure the full impact on productivity. Relevant supplementation studies (see section 3) did not demonstrate any significant productivity effect during the measurement periods. To the author's knowledge no human capital analysis has been applied in longitudinal CED studies.

4.2. Analytical limitations

We can distinguish two groups of analytical limitations in relation to human capital studies which can be expected to produce underestimates of the human capital formation effects of CED (Table 2). In several instances, and depending on the study design, the effect on the human capital formation estimate cannot be known a priori. Intertemporal behavioral assumptions are likely to underestimate the effect on human capital formation in a cross-sectional design, and to a lesser extent in a longitudinal design (with or without supplementation).

In the cross-sectional design, increased social mobility as a result of higher levels of human capital formation is assumed not to take place, primarily because the analysis is in most case done within homogeneous occupational groups. Increased social mobility is likely to have an income-effect which, if ignored, leads to an underestimation of human capital formation. Assuming that the time-period over which measurements are made is sufficiently long, increased social mobility can be measured.

The final effect of human capital formation is conditioned by interactions with exogenous socioeconomic, cultural and other factors. In cross-sectional designs in which these interactions are ignored, the effect on the estimate of human capital formation is a priori unknown, while in longitudinal designs the actual estimate of human capital formation should reflect at least the measurable interactions with exogenous factors.

Table 2. Analytical limitations of human capital studies and their effect on estimates of human capital formation

Analytical limitations


Cross-sectional design

Longitudinal design with or without supplementation

A. Intertemporal behavioral assumptions:

1. Social mobility



2. External socioeconomic factors



3. Income effect on human capital formation



4. Individual preferences for income and leisure at different levels of nutritional intake



5. Interrelationship among individual labor supply and consumption functions of household members



B. Quantification problems:

1. Psychic income from leisure



2. Wealth management



3. Variation in nutritional status encountered in homogenous group



4. Internal rate of return measured

Possible only if food intake is


An initial investment in the human capital of an individual may have an income-effect which in turn may have secondary human capital formation and income-effects. This multiplier effect is ignored in cross-sectional designs, but can be included, conceptually at least, in longitudinal designs.

In the cross-sectional design of human capital studies, it is assumed that individual relative preferences for income and leisure remain constant with human capital formation. Labor supply theory would predict an increase in supply of labor time as the marginal productivity of the worker increases. This may lead to an underestimation of human capital formation in cross-sectional studies, but not necessarily in longitudinal studies, in which the labor supply response is directly measurable.

Interrelationships among individual labor supply and consumption functions of individual household members may have as a result that improvements in the energy status of one household member induce human capital formation in other household members. There is, however, no solid empirical evidence of this. In the Philippines' supplementation study (DE GUZMAN et al., 1985) the household as a whole, rather than an individual household member, was provided with a food ration because of the postulated interrelationship of consumption functions among individual household members.

Human capital studies suffer from certain quantification problems, independently of whether a cross-sectional or a longitudinal study design is employed. More intensive use of leisure time is likely to render additional returns in what economists call "psychic income" which is difficult to measure. Better wealth management (not very relevant to low-income populations) may result from human capital formation, and these returns are usually not measured. Cross-sectional designs depend on the variation in energy status which they encounter in the study sample. Given that these samples are normally quite homogeneous, this variation may be expected to be small. Furthermore, field studies of this kind involve another measurement bias: they are undertaken on workers who are actually employed and who are not necessarily representative of the total adult population in a developing country. Among those members of the adult population at the lowest end of the distribution of human capital (poor health and body composition, no skills, etc.), the same improvements in energy status may represent more human capital formation than in employed workers.

Internal rate-of-return analysis is possible with longitudinal studies when supplementation is involved. When no supplementation is involved, daily food intake needs to be measured over time and costed using market prices. In cross-sectional studies, body composition indicators have mostly been used, which does not allow for an internal rate-of-return analysis.

Clearly longitudinal studies with a long measurement period are preferable to cross-sectional designs, but are also a great deal more costly and complex.

5. Economic behavioral studies

5.1. Methodological aspects
5.2. Analytical limitations

These studies attempt: a) to provide an understanding of how households (or individual members) allocate time and energy to different productive and leisure activities under conditions of chronic or short-term deficiency in energy availability, and b) to measure the economic (and biological) consequences in terms of work output, income, expenditure patterns and nutritional status (body composition). Relevant studies cited previously have employed as the "experimental variable" short- or long-term changes in food availability, and/or in energy requirements (or work demands), or have attempted to establish statistical associations between levels of energy intake (or body composition indicators) and economic consequences. Again the final purpose of these studies is policy and program advocacy with the aim of making more food available to populations suffering from CED.

Economic behavioral studies have a theoretical basis in the economic literature. The efficiency-wage theory postulates as a central argument a direct relationship between food consumption and work productivity of rural workers in developing countries (LEIBENSTEIN, 1957; 1958). A comprehensive review of the theory may be found in BLISS and STERN (1978). The postulated energy intake-productivity relationship assumes that after daily energy intake (at normal level) covers the basal metabolic energy requirements of the worker, the excess is then allocated first to work activities. More than proportional productivity returns in response to higher levels of daily energy intake are expected over a certain range. Beyond this range of daily energy intake, less than proportional productivity returns result. The strength of the energy-productivity intake relation will depend on the types of work performed. The supplementation studies in India, Kenya and Guatemala to which reference was made in section 3 can be regarded as empirical tests of the efficiency-wage hypotheses (BELAVADY, 1966; IMMINK, BLAKE and VITERI, 1986; IMMINK and VITERI, 1981; WOLGEMUTH et al., 1982).

A number of proposed hypotheses are listed in the Appendix to this paper, which may indicate the future direction of economic behavioral studies. In formulating these hypotheses, worker productivity is not limited to the narrow definition in the efficiency-wage hypothesis. The study design presented in section 3 which involves seasonal changes in food availability and labor require meets lends itself to testing some of the proposed hypotheses. Furthermore, these hypotheses explicitly take the household as the unit of analysis, and test for interactions of biological and behavioral adjustment processes. Acute decreases in food availability and increases in energy requirements, superimposed on chronic energy deficiency conditions (which are likely to be marginal in the majority of the population in developing countries), may result in significantly higher social costs than just CED itself.

Table 3. Methodological aspects of economic behavioral studies

Methodological aspects:

1. Study design:

a. Cross-sectional, with comparison groups.
b. Longitudinal (repeated measurements), with or without quasi-experimental design.

2. Study subjects:

Households which are known to be energy deficient.

3. "Experimental variable":

a. Seasonal variation in household food availability.
b. Seasonal variation in opportunities for productive activities

4. Key economic variables:

a. Activity patterns: time shares among classified activities.
b. Individual and household income and expenditures.
c. Home production (including food).
d. Intrahousehold economic decision-making.

5. Other variables:

a. Energy intake of household, and individual members (index members).
b. Nutritional status/body composition; physical growth pattern of individual household members.

5.1. Methodological aspects

Some methodological aspects of economic behavioral studies are summarized in Table 3. Cross-sectional as well as longitudinal designs have been employed. In the studies discussed in the previous section, households which ex ante were thought to be energy-deficient, or individuals belonging to such households, were selected for inclusion as study subjects. In cross-sectional designs, a statistical association between energy status and certain outcomes is established, or a comparison between groups with different energy status is made to establish differences in economic outcomes. A few studies have employed energy supplementation as the experimental variable, as we have previously indicated. Seasonal variation in food availability and/or in opportunities for productive activities have only occasionally been employed. Development programs or projects which create new opportunities for productive activities on a long-term basis have not been used as experimental variables. However, within the context of what is stated in the introductory section of this paper, development projects are perhaps the most relevant experimental variables.

Activity patterns are often expressed as time-shares among activities, which are classified by their energy costs or by type of productive and leisure activity. Time allocation data may be converted to energy expenditure estimates applying energy cost coefficients (DE GUZMAN, 1985; GROSS and UNDERWOOD, 1971; VITERI, 1971). Time allocation data are usually obtained by recall surveys or by direct observation by external field workers (EVENSON, 1978; MINGE-KLEVANA, 1980). JOHNSON (1975) has applied a random time-sampling technique, which renders data for a community or a sample of households as a whole. This technique permits pre- and postintervention comparisons in the same sample of households or individuals. However, it cannot be applied when long-term measurement of complete daily activity patterns of specific households or individuals is involved.

Another outcome variable is household or individual income. Changes in activity time-shares will be reflected in total income as long as the income measure includes the market value of home production activities. As was indicated in the previous section, the value of the utility generated by leisure activities ("psychic income") is difficult to quantify and incorporate in the complete income measure. In order to quantify additional economic effects, expenditure patterns related to income changes may be included. Key intermediate parameters are then aspects related to intrahousehold decision-making, such as: which household member receives what type of income and at what frequency?

Additional variables which may complement economic variables are energy intake measures and body composition or physical growth indicators obtained by means of anthropometry. These indicators may be included to measure the impact of energy supplementation on total energy intake and body composition (IMMINK, BLAKE and VITERI, 1986; WOLGEMUTH et al.), seasonal or short-term variation due to external factors (IMMINK et al., 1987), or to classify groups of sample subjects (VITERI, 1981).

Contents - Previous - Next