Cover Image
close this bookWho's Hungry? And how do we know? Food Shortage, poverty and deprivation (United Nations University - UNU, 1998, 199 pages)
close this folder2. Measuring hunger
View the document(introductory text...)
View the documentInput: Enough to eat?
View the documentOutput: Nutritional outcomes
View the documentRelations among the hunger indicators
View the documentNotes
View the documentWorks cited

(introductory text...)

Sara R. Millman and Laurie F. DeRose

Given the definition of hunger as consumption of a diet inadequate to sustain good health and normal activity, growth, and development, an ideal measure of hunger would involve a comparison between the diet actually consumed and that required for these purposes. In fact, one set of hunger indicators is based on this principle, focusing on the question of whether people are getting enough to eat. Practical implementation of the ideal, however, encounters significant difficulties both in measuring or estimating the diet and in defining the requirements against which it should be compared.

A second set of indicators focuses on outcomes of malnutrition. This approach has the advantage of identifying people whose intake is poor enough to have measurable consequences; it avoids the necessity of measuring either intake or need. But because some of the manifestations of hunger have other causes, it is not always clear that inadequate intake accounts for the outcome measured. To take an extreme example, death rates have been interpreted as aggregate indicators of hunger, and ratios of males to females in the surviving population as evidence of less adequate nutrition for one sex than the other. The possibility of other causes must also be kept in mind, even when using outcomes more directly related to food intake: anthropometric measurements may be influenced just as heavily by access to health care as by access to food.

An additional complication inherent in the use of outcomes of malnutrition as indicators of hunger is that these measurements must be compared with some standard of physical normality. This is most problematic with respect to growth. While the standards for what constitutes normal blood sugar levels are defined fairly precisely and uncontroversially, "normal growth" encompasses a wide range of alternatives: unhealthy growth patterns are more difficult to identify. This particular controversy will be explored in detail later in this chapter. For now, it is important to note that whether food intake or nutrition outcome is used as a measurement of hunger, standards are necessary.

Despite the great individual variability in intakes that would support good health and normal activity, cut-offs below which most would be functionally impaired can be defined. In measuring hunger and even in answering the more specific question of who is hungry, the use of standards allows us to identify types of individuals who are functionally impaired, even if some of those above the cut-offs also suffer from hunger and some below function normally. Chronic shortfalls in caloric availability in particular communities, or high proportions of underweight in particular sex and age groups, provide enough information to target interventions effectively. Nevertheless, we devote a great deal of attention to measurement issues in order to (1) identify the types of data that would be needed for future, more reliable, hunger estimates, (2) identify the biases inherent in some commonly used measurement techniques in order to inform interpretation of the statistics generated by them, and (3) explore the extent to which different indicators of hunger function interchangeably.

Input: Enough to eat?

To determine whether people are getting enough to eat requires the answers to two separate questions: what are people eating, and how much do they need? The former question is conceptually straightforward, but presents daunting data collection and estimation problems in practical terms. The latter compounds these problems. A diet that is adequate for some purposes is inadequate for others, since varying activity levels and patterns of growth can reconcile good health with a range of dietary intakes for the same individual. Thus, normative judgements as to desirable activity levels and growth patterns are implicit in any definition of dietary requirements. Uncertainties as to the range within which cost-free adaptation can occur and the factors underlying variation in requirements complicate the picture still further. Our review of hunger-estimation procedures that assess intake relative to need considers first the measurement of food availability and then the standards for need.

Food supplies

The discussion of estimates of food supply or dietary intake below is organized according to the three levels of social organization: these are the geographic region (in particular, the nation), the household, and the individual. Common approaches used to estimating food availability at each of these levels are described, with particular attention to measurement difficulties as well as sources of error and bias.

Food available to national populations

Estimates of national food supplies for most countries are maintained and regularly updated by the Food and Agriculture Organization (FAO) of the United Nations. These are based on food balance sheets, which involve detailed analyses of national food systems (see, e.g., FAO 1983,1984). The accuracy of these accounts has been extensively questioned. The FAO estimates of per capita, per day food availability for national populations are an essential part of any attempt at either assessing hunger globally or exploring its variation across countries.

These estimates are obtained by first calculating the amounts of specific foods available for human consumption within the country. This is done by taking the sum of amounts produced, imported, or withdrawn from carry-over stocks and then subtracting amounts lost in processing or transport, used for purposes other than human food consumption (seed, feed for animals, industrial raw materials), exported, or stored for later use. Estimates of the nutrient content (calories, protein, and fat) per unit of each food are then multiplied by the amount of the food available to obtain estimates of the total amounts of each nutrient available from each foodstuff. These are summed across all foodstuffs to obtain total supplies of each specific nutrient. The final step is to divide national totals by the number of days in the period to which the supply data apply, and by the total number of people in the population. Thus, the results are commonly expressed as per day, per capita nutrient availability.

The data requirements of the food balance sheets are enormous. Precise values are often lacking for at least some of the necessary inputs. In other cases the situation may be one of no data at all for certain parameters, rather than of imprecise data. To the extent necessary, the blanks are filled in with approximations, assumptions, and informed guesswork. The resulting estimates, therefore, should not be viewed as unduly precise.

According to the FAO (FAO 1984), statistics on non-food utilization of food supplies are perhaps most widely problematic, while estimates of loss between harvest and household typically rest on reliable local expert opinions. And "... even the production and trade statistics on which the accuracy of food balance sheets depends most are, in many cases, subject to improvement..." (FAO 1984: ix) Uncertainties on withdrawals from and deposits to carry-over stocks are limited by preparing food balance sheet estimates as three-year averages.

The degree of imprecision may be greatest for exactly that set of countries in which the aggregate food supply is likeliest to be short, since general statistical systems are less well developed in poorer and less-industrialized countries than elsewhere and even population data may be lacking. In 20 of the 125 counties covered in detail in the World Bank's World Development Report 1992 (World Bank 1992), the most recent census occurred in the 1970s or even earlier. Estimates of current populations extrapolated from past trends, or from old data that may have been inaccurate even when new, may be far from the truth. Nigeria's 1991 census, for example, counted a total of 88.5 million people, as compared with a UN estimate of 120.5 million people (APS 1992). Since dietary energy supply is divided by population to calculate per capita dietary energy supply, these alternative figures result in estimates that vary by 27.4 per cent. In this particular case, the food supply in Nigeria looks considerably more favourable when (presumably more accurate) data from the 1991 census are used instead of the older population estimates.

Food-production data may also be subject to extreme inaccuracies, especially in areas where much of the food produced is consumed directly by growers without ever entering the market. When such subsistence production occurs in remote regions it may be missed in national statistics. Even aerial surveys are likely to miss food production when it occurs as part of a system of shifting cultivation, with crops grown in small clearings in the jungle. Underestimation of available food is also particularly likely where uncultivated or lesser-known foods are commonly consumed. Some foods are omitted from food balance sheets entirely; in certain settings these may make up a significant portion of the total diet. Governments that tax and procure foods at below-market prices also present significant incentives for understatement of production by producers. These considerations will affect estimates of food production more in some areas than others, and therefore could bias the comparisons of relative food availability using food balance sheets. Svedberg (1991) has argued that the FAO's food balance sheet approach significantly underestimates food availability in sub-Saharan Africa relative to other regions. Despite their real shortcomings, however, food balance sheets provide the best information available as to national food supplies.

Household surveys

Surveys of representative samples of households are an alternative source for estimating national food supplies per capita. In addition, such surveys yield direct measures of household access to food. The FAO (1983) provides a useful comparison of several approaches.

Income/budget/expenditure surveys, with a broad focus on the economic situation of households, report amounts spent on food or the value of food consumed. Food quantities may be obtained as part of the process of estimating the value of food consumed, but are usually not shown in reports. This kind of survey omits food consumed outside the household and losses of food within the household; food gathered wild, received as a gift, and sometimes even produced by the household may also be omitted.

In contrast, food-consumption surveys focus more on amounts and nutrient composition of foods than on their economic value, and attempt to include food consumed away from home. Amounts may be ascertained by survey (recall), by consumption records maintained by the respondents, or by food weighing at meals (minus plate waste). Data collection (typically over a period of one to seven days for each household) is time-consuming and needs to be supervised for accuracy. Not surprisingly, food-consumption surveys of samples large enough, and sufficiently dispersed geographically, to yield national statistics are largely unavailable. India is the only country that conducts such surveys routinely. Nevertheless, this strategy is used more frequently in local-area studies and yields valuable information.

Finally, items typical of either income/budget/expenditure surveys or food-consumption surveys may be incorporated in multiple-purpose surveys. The information about intake from these is less detailed than that from studies focused on food and nutrition, but they significantly contribute to what is known about household food availability. One example is the Living Standards Measurement Survey that the World Bank has sponsored in a number of developing countries.

Despite their advantages, survey data of the types described above are flawed, as indicators of either national food supplies or the distribution of food across households, by the short time span to which they typically apply. Since household food consumption and food expenditure do vary over time, a household's apparent food security may be very much affected by the duration and/or the timing of data collection. Even from one day to the next, household food consumption and expenditure are often highly variable. As recognized in some methodological treatises (e.g. National Research Council 1986), this short-term variation suggests that the shorter the period covered, the more frequently extreme values of calories per capita per day will be found. Therefore, shorter-term data collection will tend to find higher proportions of households at extremely low (or high) levels of dietary adequacy. The sensitivity of proportions observed at the lower extreme to measurement duration matters when measuring hunger. Although statistical adjustments are possible (National Research Council 1986), this problem often goes unrecognized and therefore uncorrected.

Temporal variations in food intake will have different effects on survey results, depending on whether they have a common cause that affects households similarly or whether the causes are more household specific. If patterns of temporal variation in household food security are unrelated across households (for example, job loss for some families in the absence of recession or widespread lay-offs), surveys taken at different times might find different sets, but similar proportions, of households unable to meet the needs of their members, and food consumption or expenditure totalled or averaged across households might be unaffected. If many households experience the same temporal pattern of food security, however, even the proportion of households falling below any specified cut-off and the total of consumption or expenditure across households may be very much affected by the timing of data collection.

In subsistence agriculture, for instance, a food-consumption survey taken just before harvest might well show a dire situation indeed, but one a few weeks later a situation of abundance. Neither snapshot reasonably characterizes the population's usual access to food. Even the per day, per capita caloric availability over a three-year period reflected in a food balance sheet would not capture this situation; the more compelling reality for understanding who is hungry may be exactly the wide swings between scarcity and abundance rather than the level to which these two extremes average. Only longitudinal data can capture this reality. Enough evidence exists to suggest that temporal variation is an important dimension of hunger in some settings, but the research investment required to document such variation and its effect on the people who experience it leaves its details unexplored for most populations.

CONSUMPTION SURVEYS VERSUS SUPPLY METHODS. Overall, the consumption surveys versus supply methods of estimating national level food availability yield very different results. Comparing estimates of national per capita dietary energy supply based on income/budget/expenditure surveys and on food balance sheets, the FAO (1983) finds sizeable discrepancies. For developed countries, the survey results are consistently lower by hundreds of calories per day, at least partly owing to their omission of food consumed away from home. For developing countries, estimates from the two different sources also differ substantially, but the differences are smaller and not in a consistent direction. It is likely that the closer agreement of the two approaches for developing countries may reflect the fact that both surveys and food balance sheets underestimate food availability. Food balance sheets more accurately reflect availability in developed countries where most food is marketed than they do in developing countries. The more complete food balance sheet data for developed countries may eliminate a downward bias that would otherwise bring estimates into closer agreement with those based on income/budget/expenditure surveys.

In the FAO's assessment:

Annual food balance sheets tabulated regularly over a period of years will show the trend in the overall national food supply, disclose changes that may have taken place in the types of food consumed... and reveal the extent to which the food supply of the country, as a whole, is adequate in relation to nutritional requirements... (T)he food balance sheets... while often far from satisfactory in the proper statistical sense, provide an approximate picture of the overall food situation in the countries which may be used for economic and nutritional studies, the preparation of development plans, and the formulation of related projects. (FAO 1984)

Despite their difficulties, the food balance sheets are the best information we have to quantify national food supplies for most countries of the world. Because they are collected yearly, using the same methodology, they can provide valuable information on the relative magnitude of year-to-year fluctuations in food availability (Atwood 1991). Although shifts in production (especially shifts to marketed crops) may change the proportion of production recorded by the FAO even when total production remains unchanged, this type of change typically takes place over a number of years and is therefore unlikely to bias comparisons seriously between adjacent years.

Chapter 3 on food shortage in this volume draws on food balance sheets to assess the adequacy of national food supplies. They are also an essential input into global and regional estimates of numbers of people living in households that cannot afford the food their members need.

Food available to households

Even where the estimates of national per capita dietary energy supplies based on food balance sheets are quite accurate, they provide no clue about variable access to supply within nations. We know that some people go hungry, even in countries in which per capita total food supplies are far in excess of requirements, and that others are well fed, even in countries with too little food to meet the needs of their populations. We must have information on the distribution of food, not just the total amount available, in order to assess the prevalence of hunger. Surveys such as those discussed above provide one means of measuring variation across households in access to food.

With household-level food consumption or acquisition data, we can go beyond estimating the numbers of households, or of people in households, in food poverty. It is also possible to contrast the characteristics of households falling above or below some cut-off, or to compare access to food across different types of households. Both exercises help us to understand which households are hungry, not just how many. Household-level data also are indispensable for exploring associations between a household's food poverty status, health, education, and any other characteristics. Because nationally representative surveys gathering such information are scarce, our picture of household food poverty is necessarily incomplete. Many findings emerge from the literature on hunger and poverty; how widely these can be generalized is limited because there are so many settings for which the necessary data are not available.

A household's access to food is measured directly or estimated, and then compared with some standard of need to judge adequacy. The choice of cutoffs for minimally adequate food intake influences both the estimates of food poverty and the comparability across samples. The FAO and the World Bank now accept common standards, but the thresholds of undernutrition which they used even in the recent past were defined differently,1 and both organizations regularly readjusted their methodologies (Uvin 1994). Therefore, comparative estimates of food poverty across time or across data sources both need to check which methodology was being used, to standardize reporting.

Household food surveys are used to help to define what income levels are necessary for households to enjoy different levels of access to food. Income levels measured in other data sets then can be translated into access to food, to estimate the adequacy of the diet for households for which no actual food data were gathered. The widely cited estimates of world hunger produced by the World Bank and the FAO estimate indirectly the distributions of household income and proportions of households falling below food adequacy. These are calculated country by country but published only for regions, in the expectation that errors in single-country estimates will tend to balance out in aggregation.

Single-country estimates of numbers of people living in households that cannot afford to feed their members have been published by the United Nations Development Programme (UNDP 1992) for many countries. These rely on national reports of numbers in absolute poverty, defined as inability to meet their needs for food and other basic necessities. In some instances, figures are given separately for rural and urban areas, permitting within-population comparison of food poverty along at least this one dimension. Operational definitions of absolute poverty are country specific; the situations captured by these country-specific operationalizations are not comparable across countries without further standardization to adjust for different underlying conditions and methodologies.

Food available to individuals within households

DIRECT MEASURES. Ultimately, we measure the hunger of individuals, whose access to food is regulated by intra-household allocation rules and processes. We know that some individuals go hungry in households enjoying food security, while certain individuals are well nourished even in households that are food insecure overall. Solid conclusions regarding intra-household allocation of food require direct measurement of individual food consumption. This form of data collection is even more demanding in terms of time and skill than the household food-consumption survey. In addition to ascertaining and recording kinds and quantities of foods prepared for, and left over after, every meal, the researcher must also keep track of which household member gets how much of each dish. Collecting this kind of data within households is never unobtrusive: even if it is not resented, it may result in some change in the actual consumption, especially if those observed believe that their ordinary behaviour is subject to disapproval. Understandably, few surveys collect individual food-consumption data for a large or nationally representative sample of households over an extended period of time.

Even when the effort is made, differences in the completeness of dietary information across household members can still confound comparisons. Some are likelier than others to do some of their eating outside the household (food may be provided on the job or at school, or may be purchased from street vendors) and such meals or snacks tend to be underreported. Children also "forage" - a practice that has led some anthropologists to insist that "child following" is the only reliable method to record their intakes (Wilson 1974; Laderman 1991).

One particularly vexing instance of such differential completeness of dietary data relates to the difficulty of measuring amounts of breastmilk consumed by small children. Many food-consumption studies report only on children over 12 months of age, but intake of children past infancy will be underestimated where breast-feeding extends into toddlerhood, as it commonly does in developing countries.

Breastmilk is often completely omitted from studies on individual food consumption; this is hardly surprising, given that some manuals on food-consumption study methodology (e.g. Cameron and Van Staveren 1988) give no recommendations for estimating its intake. Jelliffe and Jelliffe (1989) outlined considerable problems in measuring breastmilk intake and concluded that all of the methods in use produced inaccurate estimates. Nevertheless, even large-scale surveys can question women as to how frequently their children nursed, rather than just whether they nursed. More labour-intensive surveys can observe women going about their daily tasks and count breast-feeding bouts over a full 24 hours.2 These methods could provide a better indication of how important breastmilk is in the diet of children in different cultures. An indication - rather than a precise measurement - is all that is going to be available from dietary-survey data. More accurate methods, such as isotope dilution of labelled water, are too costly to be used in large-scale surveys.

The frequent omission of breastmilk from food-consumption data may qualify the common finding that the youngest are worst fed. While there are undoubtedly times and places in which children do receive less than a fair share of household food supplies, the downwards bias on estimates of food consumption for the youngest, resulting from omission of mother's milk, may both exaggerate such a pattern where it exists and create one where it does not.

INDIRECT MEASURES. The difficulty of collecting data on individual food consumption has led some researchers to use less-direct measures. For example, the order in which different household members are served at meals often is interpreted as an indicator of relative access to the common food supply, and the conclusion is drawn that those served last receive less than a fair share (e.g. den Hartog 1973; Katona-Apte 1975; Maher 1981; Papanek 1990).

Reports that certain foods are normatively forbidden to, or reserved for, people in particular age, gender, or status categories provide a second set of indicators of within-household differences in access to food and the relative adequacy of diets. These observations suggest differences in dietary adequacy within a household. However, without further measurement of actual intakes compared with some nutritional standard they are not conclusive. More effort has to be made to validate these kinds of intra-household consumption indicators; where individual food-consumption data are available, it would be useful to explore the extent to which inferences that could be drawn from eating order and other aspects of dietary tradition are borne out.

Nutritional requirements

Measures of food supply available to national populations, accessible to households, or consumed by individuals always must be compared with amounts required, in order to permit any conclusion as to adequacy. Our discussion of nutritional requirements, that here focuses on energy, covers the same three levels of social organization as the preceding discussion of food availability; however, because the requirements for larger aggregations of people are conceptually based on the sum of individual requirements, we proceed from the individual, to the household, to the population.

Individual requirements
According to the United Nations:

The energy requirement of an individual is the level of energy intake from food that will balance energy expenditure when the individual has a body size and composition, and level of physical activity, consistent with long-term good health; and that will allow for the maintenance of economically necessary and socially desirable physical activity. In children and pregnant or lactating women, the energy requirement includes the energy needs associated with the deposition of tissue or the secretion of milk at rates consistent with good health. (WHO 1985: 12)

These requirements vary across individuals and within individuals over time. They are influenced by gender, body size, physical activity level, age, reproductive status, and disease. In addition, cross-population variation unexplained by these factors appears to reflect some influence of climate as well as other factors, such as dietary composition. But individuals who appear comparable on all these dimensions and who are in energy balance, as indicated by absence of weight loss or gain, still vary widely in their caloric intake. Consequently, accurate determination of an individual's caloric requirements would necessitate very intensive data collection under laboratory conditions. This is clearly infeasible on any large scale. Thus, studies of individual dietary adequacy most commonly compare actual intake with expected or average per capita requirements for persons similarly classified on the same background variables.

The most contentious and difficult aspect of caloric requirements relates to adaptations to a constrained diet. Desirable patterns of growth for children and desirable patterns of physical activity at all ages are an important part of this controversy. Some analysts (e.g. Sukhatme 1988) have argued that intake and expenditure of dietary energy function as a self-regulating homoeostatic system in which the efficiency of energy use is increased as intake decreases and decreased as intake increases. The argument is a rather technical one, but its real-world implications are far from trivial. It has been taken as demonstrating "cost-free" adaptation to low intake - the possibility of maintaining not only energy balance and health but even usual activity levels on severely restricted consumption.3

Others have responded (see, e.g., Scrimshaw and Young 1989; Waterlow 1989) that adaptation to low intakes does not occur without undesirable limitation of physical activity. A proportion of the reduction in energy use that occurs as a result of reduced intake is apparently innocuous: weight loss itself causes some decline in caloric requirements, and less energy is used in digestion, absorption, and storage of nutrients when less food is consumed. However, the major mechanisms for reduced energy expenditure are behavioural. Therefore, most nutritionists do not view the adaptation to low intakes as cost free.

Although controversial, the concept of cost-free adaptation may well underlie decisions by some major nutrition-monitoring organizations (e.g. FAO 1977, 1985; NNMB 1981) to define cut-offs for undesirably low caloric intake at levels two standard deviations below average requirements. Setting such low cut-offs for minimal intake guarantees that the prevalence of hunger will not be overestimated. But unless the people with the lowest intakes also have the lowest requirements, it also guarantees that a high proportion of cases of genuinely inadequate intake will go unrecognized.

The controversy about the degree of adaptation to nutritional stress is also a policy debate about resource allocation. Few would argue that adaptation to extreme deprivation could be cost free, but those who interpret adaptation to lower levels of nutritional stress as cost free favour allocating resources to more narrowly focused nutritional programmes that would benefit only those in extreme need. If their premise is correct, this would reduce expenditure while increasing effectiveness. Those who have higher estimates of the costs of more moderate deprivation emphasize the needs of marginally nourished people. Their arguments tend to focus on the difficulty of emerging from poverty when productivity is limited by intake (e.g. Dasgupta and Ray 1990). If this is the case, government expenditure on broader nutrition programmes represents a better long-term policy choice, since it will improve both national productivity and national income distribution.

Household requirements

To estimate requirements for aggregations of individuals, whether these be households or national populations, it is necessary to consider what kinds of individuals make them up. Households are commonly defined - even outside the literature on hunger - as groups of individuals who share cooking facilities. Some individuals have regular access to food in more than one household (e.g. polygynous men whose wives do not cook together), and this complicates the estimation of household need.

Household need is determined by the number of members as well as their age, gender, and other determinants of individual need listed above. For households, caloric requirements are often estimated as the product of household size and national average per capita requirements. The implicit assumption is that each household's composition mirrors that of the nation as a whole. In fact, of course, households differ in their composition: the nutritional needs of a woman living with three children under five differ substantially from those of a childless couple living with two of the husband's brothers. Although estimates of household requirements reflecting each household's own composition are preferable in principle, detailed analysis of Indonesian data (see table 2.1) suggest that the additional analytical effort involved makes little difference to the proportions of households estimated to consume less than they need. Approximately equal numbers of households are misclassified in each direction by relying on national average per capita requirements. Thus, although misclassification does occur for many households, aggregate results are virtually unaffected, at least for this population.

Table 2.1 Percentage of households in food poverty by definition of household caloric requirements, Indonesia, 1980

Household requirements defined by product of household size and national per capita requirement

Households below cut-off

Households above cut-off


Household requirements defined as sum of individual requirements

Households below cut-off




Households above cut-off








Source: Data from the Indonesian National Socio-Economic Survey of 1980; tabulation by Mark M. Pitt.

Caloric consumption requirements for households are the sum of the requirements of their individual members. However, if requirements are to be compared with measures of food supplies available to a household, rather than with those actually eaten within it, some allowance should be made for loss or wastage of food within the household. Such losses occur, for example, owing to spoilage, to infestation by insects or other pests, to losses in preparation, or to "plate waste" - leftovers discarded. While plate waste may be low in situations of scarcity, poor storage facilities make it difficult to avoid other forms of food loss within the household. More-affluent households may have the storage facilities needed to minimize spoilage but are less likely to make sure every bit of edible food is consumed.

Thus, for different reasons, some loss of food is expected in both poorer and wealthier households.

National requirements

Determining caloric need for national populations is somewhat less complicated, because the fluidity of boundaries between households is not an issue and the gender and age composition of national populations is generally well known. Estimates of average per capita caloric requirements for national populations incorporate information on typical body sizes and physical activity levels, as well as age and sex composition, and may also allow for effects of childhood disease and desired growth patterns for children.

Several sets of average per capita caloric-requirements estimates for national populations have been published by various organizations and applied widely. Unfortunately, requirements estimates for the same population may vary substantially. More unfortunately still, some statistical compendia present caloric availability data only as a proportion of requirements, without specifying what standards for requirements are being used. Thus, the potential for confusion and contradiction is substantial. The remainder of this section briefly reviews some of the reasons for the differences in sets of requirements estimates.

Estimates of national average per capita caloric requirements produced by the FAO were published in the Fourth World Food Survey (FAO 1977) and have been used since in preparation of several kinds of national-level estimates of hunger. FAO publications from 1990 (e.g. FAO 1990a, 1990b, 1990c), however, show a different set of caloric-requirements estimates, in most cases substantially lower than the old ones.

The difference between these sets of estimates - a median decrease of 280 calories per day - was enough to result in drastically different pictures of world hunger. For example, the total 1987 population of countries with per capita caloric availability below requirements was 1,603 million relative to the old requirements, but 152 million relative to the new ones (Millman and Chen 1991). The difference turns out to be attributable to two main factors. First, the new estimates incorporate new (and presumably more accurate) data on body sizes of national populations. In most cases, this new information contributed to a reduction in requirements estimates. Second, and more significant, the reduced FAO caloric-requirements estimates eliminate an allowance for food losses within the household or retail establishment. The FAO's food balance sheets do account for losses of food prior to harvest and during food storage, processing, and delivery to retail establishments, but they no longer allow for losses within households or institutions. In the past, their caloric requirements estimates were inflated by 10 percent to allow for such losses.

Not accounting for food loss within households amounts to making the unrealistic assumption that its value is zero. Some allowance would seem essential in assessing the adequacy of supplies of food available to households or to national populations. Comparisons of hunger prevalence at different points in time should not be made without attention to the standards applied, or misleading conclusions as to trend may result. Comparisons made using data published since 1992 are less problematic because the FAO and the World Bank have adopted a common methodology for producing their estimates.

The balance of food supply and requirements

At each level of social organization that we consider, conclusions about hunger are seriously affected by the measurement issues raised in this chapter. We draw attention to the methodological difficulties in the study of hunger, not to create pessimism about our ability to measure hunger but to inform comparisons between results derived from different practices. Assessments of food shortage may underestimate production and underestimate waste, but, as we have emphasized, the most commonly used methods are likely to yield reliable data about trends in food security. Employing even faulty methodologies consistently gives us a picture of how the balance of food supply and requirements is changing, and can even help identify what factors are influencing the changes.

Assessments of food poverty may be flawed by lack of attention to household composition or the changeable nature of household boundaries, but we have shown that these issues are unlikely to affect estimated levels of food poverty profoundly. In chapter 5 we will address how these and other issues affect our understanding of which households are in food poverty.

Comparisons across countries, of the extent of food poverty or the determinants of food poverty, may be flawed or reach varying conclusions if they fail to account for different types of data and definitions employed within countries. Studies that measure household food availability by documenting consumption cannot be directly compared with those that estimate food poverty on the basis of low income. But important insights into the determinants of hunger can result from comparing methodologies and outcomes in a single setting that can then guide additional research. If household income seems to be adequate and household food consumption is unacceptably low, the determinants of spending patterns then deserve more attention than the determinants of income. If certain households seem to do a better job than others, of balancing intake with requirements despite income constraints, then focusing on nutritional strategies and sources of food and income in those households might be productive.

Assessments of food deprivation based on intake must be careful to consider individual requirements when drawing conclusions, especially where judgements of discrimination leading to inadequate intake are involved. Some studies of intra-household food allocation have concluded that women and children received less than their fair share because they consumed fewer calories per day than others. If everybody had the same caloric requirements, differences in consumption could, indeed, be interpreted as showing patterns of advantage and disadvantage. But caloric requirements do vary and, typically, are lower for women and children than for others. To reach meaningful conclusions on adequacy or equity, dietary comparisons must examine individual intake relative to individual need. This can be done by dividing each household member's consumption by the corresponding estimate of requirements and comparing the results, often referred to as indices of dietary adequacy or caloric adequacy, which can be interpreted to show patterns of discrimination or equity. Only where the ratio of consumption to requirements is lower for women than for men, and for children than their elders, can we then conclude that women and children are at a disadvantage in the intrahousehold allocation of food. The additional complication of gender bias built into standards is discussed further in the section on choice of standards.

Another set of studies bases conclusions about discriminatory allocation of food within the household on the observation that certain household members are eating less than they need, without demonstrating that others are doing any better. In these instances, we usually have solid evidence of inadequate diets for those identified as victims. Again, however, a comparison with the situation of others is necessary to support a conclusion of relative disadvantage: other household members may be equally underfed, and, if so, our interpretation of the situation should be quite different from that if we find that others in the same household consume a more adequate diet. If all are underfed, the household is clearly in food poverty and food deprivation is being experienced by all its members. If only some are food deprived, this may or may not result from food poverty.

Output: Nutritional outcomes

We now turn our attention to the second set of hunger indicators, those based on measuring outcomes of malnutrition. This section focuses almost exclusively on anthropometric measurements, both because they are widely used and because they are the subject of considerable controversy. However, it is important to note that micronutrient deficiencies are also important because of their significant functional consequences, described in the introductory chapter, and also because they can signal other problems in dietary adequacy. Correctly diagnosing clinical signs of micronutrient deficiency requires well-trained staff and more extensive data: making comparisons between populations is only now becoming possible. If means of identifying mild-to-moderate deficiencies become available, these could provide increasingly important tools for measuring hunger.

Nutritional status is most commonly measured, especially in young children, by anthropometry - measurement of dimensions of physical size, such as height or weight, and comparison with distributions of the same measurement in a presumably healthy and well-nourished reference population. Children whose weight falls below the range of normal variation for children of the same age observed in a reference population are identified as underweight. Underweight may reflect small stature, excessive thinness, or both. These two dimensions are differentiated in two more refined anthropometric measures - weight for height and height for age. If the child's weight falls below the range of normal variation for children of the same height, it is considered wasted. If its height falls below the range of normal variation for children of the same age, it is considered stunted. Wasting is generally interpreted as an indicator of acute malnutrition - a current or recent crisis involving extreme weight loss. Stunting, in contrast, indicates early malnutrition. Either a past episode (or episodes) of acute malnutrition, or a routinely limited diet over an extended period, has resulted in growth impairment, even though current nutrition may be adequate.

Anthropometry is used to assess adults as well as young children, but this is done less widely. Shortness, although it may be nutritionally caused, provides clues only to the individual's experience during childhood, so that adult heights are uninformative regarding current or recent nutrition. Thinness, however, implies current undernutrition for adults as it does for children. Thus, weight for height is relevant across the range of ages, as are such other measures of fatness as mid-upper arm circumference (MUAC), skinfolds, or the body mass index. Although it is thus possible to measure people of all ages using the same anthropometric indicators, this rarely occurs in practice. As a result, evaluation of commonly stated conclusions regarding age patterns of variation in nutritional status is often quite difficult.

Anthropometry as an indicator of individual malnutrition

Anthropometry alone is not sufficient to diagnose nutritional problems in individuals, although it identifies children whose situation should be examined further in making such a diagnosis.4 If the lower bounds of normal variation were set so low as to exclude all cases of healthy small size, much actual malnutrition would not register. In order to obtain a reasonable degree of sensitivity, cut-offs are set high enough that a small proportion of individuals fall below them, despite good health and adequate nutrition. At the same time, some who are naturally larger may fall above the cut-offs, even when they are, in fact, malnourished.

The lower limit of the range of normal variation in anthropometry has been variously operationalized. One common practice has been to define a cut-off at some set percentage of the median from the reference population. For weight for age, 80 per cent has been the most widely used cut-off, but milder or more severe underweight has been defined in terms of higher or lower percentage cut-off points. For other anthropometric measures, different percentage cut-off points have been identified. Recent work has more consistently used a cutoff two standard deviations below the mean of the reference population. Using standard deviations is preferred to using percentage cut-offs because comparability across measures and for the same measure at different ages is not compromised by greater or lesser variability. Roughly 3 per cent of healthy children will be more than two standard deviations below the mean of a healthy reference population, but most children this small are correctly identified as being at nutritional risk. Those more than two standard deviations below the mean are usually identified as moderately malnourished, while those falling three standard deviations or more below the mean are severely malnourished. Some researchers have suggested using a cutoff of one standard deviation in order to target nutritional supplementation programmes to those at risk of undernutrition (see comments in Popkin 1994). While it is desirable to enhance marginal diets of children who are showing no clinical signs of growth faltering, it is preferable to target children at greater risk, if resources are limited. It is worth repeating that none of these cut-off points has any necessary functional significance, despite their utility in helping to identify hungry individuals.

Meaningful individual diagnosis involves repeated measurement over time. Repeated measurement allows a child's growth trajectory to be compared with that of normal growth, rather than simply relying on a one-time measurement relative to a cut-off point. The repeated measurement process is referred to as "growth monitoring" and can be used to determine if small children are growing normally; it can also identify larger children whose growth has become compromised, even before their size drops below some cut-off. Weight loss, as opposed to unusually low weight, is a more reliable indicator of nutritional crisis. Similarly, a period during which no increase in height occurs tells us more than does a one-point observation of unusual shortness, which might have resulted either from such a crisis (or repeated crises) at any time up to the present or from a pattern of uninterrupted slow growth. Growth monitoring has been promoted as part of the UNICEF/WHO "GOBI" (growth monitoring, oral rehydration therapy, breast-feeding, and immunizations) initiative for child survival. Its utility in this context is in alerting the mother and the health practitioner to developmental problems at an earlier stage, and thus encouraging intervention before much damage has occurred.

Anthropometry as an indicator of malnourished groups

Cross-sectional anthropometric measures are less ambiguous indicators of nutritional problems for groups than they are for individuals. While the lesser growth potential of some individuals may cause them to register as underweight, wasted, or stunted, even when they are in good health and weld fed, such individuals will be rare. If the proportion of individuals so identified in any group is high, we can be more sure that malnutrition is a significant problem for this group than that each individual identified is necessarily malnourished.

This assumes, however, that the reference population defining the range of normal variation is appropriate - that the target population would, in fact, show the same distributions of weights for age, weights for height, and heights for age if it, too, were healthy and well nourished. The most commonly used growth standards for international research are US based, and there is real controversy as to whether US growth standards are necessarily applicable everywhere. It has been argued that, in populations facing nutritional constraint over the long term, small size may, in fact, represent a healthy adaptation rather than indicating a problem (see review in Osmani 1992). Since nutritional requirements are, in part, a function of body size, stunted growth helps keep requirements low for individuals and populations and may permit good health and normal functioning at intake levels that would be too low if size were larger.

This controversy has clear implications for measuring hunger. If populations vary with respect to their proportion of healthy individuals whose anthropometric measurements fall far below the mean of the reference population, using any absolute cut-off would bias comparisons of hunger prevalence between populations. Such an argument, in fact, makes cross-country comparisons of hunger virtually impossible: any variation in achieved stature or even in caloric intake could be explained as adaptation rather than evidence of hunger. Although we argue that cross-country comparisons are both desirable and useful, we also consider the arguments against our position very carefully. The remainder of this section reviews three major arguments for why US-based growth standards might be inappropriate. We then address the question of how much difference the choice of a growth standard makes to our understanding of who is hungry.

Natural selection for small body size?

"Healthy adaptation" to nutritional constraint may occur at the population rather than the individual - level. Evolutionary pressures may have favoured small size in populations facing a very constrained diet. This variant of the "small but healthy" argument suggests that genetically determined potential size is actually less in those populations (such as most of South Asia) where average body size is small by Western standards.

However, numerous studies have demonstrated that elite groups in such populations, less constrained in terms of both diet and health care, show growth patterns (at least in early childhood) that are quite similar to those reflected in the US-based growth standards. Genetic potential for these elite groups within populations of small average size does not appear strikingly less than for their Western counter

parts. Marked increases in stature from one generation to the next are commonly observed when individuals from populations of small average stature migrate into countries with larger average stature. These increases, which are generally associated with changes in the diet and health, are difficult to reconcile with the notion that small statures in the areas of origin result from genetically determined limited growth potential. Similar very rapid increases in average stature within populations undergoing dietary enrichment (Japan, China), often associated with modernization or increasing affluence, also suggest that growth potential is comparable in populations between which actual attained body sizes vary widely.

Individual adaptation to nutritional constraint?

Even if the small body sizes observed in nutritionally constrained populations are not genetically determined, an adaptive mechanism rooted in individual experience remains a possibility. We know both that nutritional constraint during childhood is a cause of permanently small body size and that small body size in turn reduces lifetime nutritional requirements. The question then becomes, at what cost is this lifelong economy won? We can separate the answer to this question into two parts - costs of being, and of becoming, small.

To be small in stature means, ceteris paribus, to be less powerful physically than if one were taller. Since the poorest, who are likeliest to suffer growth impairment, are also probably likelier than others to have to earn their living through hard physical labour, this cost may be significant in terms of lost productivity and earnings. In contrast, small size itself may not be a problem for those whose work is mental rather than physical.

Small mothers are also at higher relative risk of bearing low birth-weight babies. The physical mechanisms through which maternal stunting is linked to birth weight are not completely understood. It has also not been conclusively demonstrated that stunting in the absence of other nutritional problems (e.g. underweight, anaemia) or deprivation during pregnancy adversely affects birth weight (Osmani 1992). Nevertheless, the correlation between maternal height and birth weight is strong across a large number of populations. Even if maternal stunting serves only as a proxy for the conditions that place children at high risk, it is still a useful indicator of who is likely to be hungry.

Some nutritionists (e.g. Beaton 1989) argue that, while there is nothing wrong with being small, the process of becoming small is damaging. Increased risks of morbidity and mortality are among the major concerns. As Martorell (1995) argued:

Although growth retardation does not cause depressed immunocompetence, the factors that cause growth faltering, such as infection and inadequate intakes of specific nutrients, also result in immunodepression... children may get infected for reasons largely determined by their environment, but, once they are infected, the course of the infection will be influenced strongly by nutritional status, reflected by the degree of growth retardation.

Even if increased risks of morbidity and mortality could be ruled out, the advantages of small size should not be counted as cost free if the constraints resulting in small size also cause significant functional impairment. Associations of growth impairment with poor intellectual and social development are well documented, and there are plausible mechanisms by which the same processes of malnutrition that cause impaired growth may also cause developmental impairments.

The small size associated with nutritional constraint has often been equated with increased risks of mortality - so strongly so, in the minds of some, that the infant mortality rate has been suggested as "one of the best tools available for measuring the extent of hunger in a society" (THP 1983), and growth has been called "the most important single indicator of health" for a child (Grant 1990). Others (Mosley and Chen 1984) have recommended that analysts combine growth impairment and mortality into a single variable, with survivors classified according to the Gomez scale of weight for age (relative to the median from an applicable growth chart, 89-75 per cent is grade I malnutrition, 74-60 per cent is grade II, less than 60 per cent is grade III) and then creating a fourth (grade IV) category for non-survivors. An assumption implicit in each of these suggestions is not only that underweight is strongly associated with elevated risks of mortality but also that the form of this relationship is relatively invariant.

There is certainly reason to expect malnutrition to increase risks of morbidity and mortality, as discussed in chapter 1. Malnutrition does reduce resistance to some infectious diseases, with different aspects of the immune system affected by deficiencies of varying degree with respect to specific nutrients. Many studies demonstrate cross-sectional associations between malnutrition and morbidity (for a useful review, see Tomkins and Watson 1989) or mortality (Puffer and Serrano 1973). In cross-sectional studies, however, effects of illness on nutritional status confound the effect of nutritional status on illness, and effects of nutritional status itself on survival chances are confused with effects of illness on both. Longitudinal research linking anthropometry to subsequent morbidity or mortality is relatively scarce.5 When this is done, reverse causation can be ruled out, although surrounding circumstances (such as crowding or poor sanitation) that cause frequent illness could be responsible for both malnutrition and other adverse outcomes. At the extremes of malnutrition, there is no doubt that a range of adverse outcomes become increasingly likely.

However, some poor areas considered to have extraordinarily low infant and child mortality for their level of economic development (e.g. Sri Lanka, the state of Kerala in India) also have very high prevalences of underweight in small children. In contrast, some countries with the very highest rates of mortality in infancy and early childhood (including much of sub-Saharan Africa) exhibit relatively moderate prevalences of underweight. Since disease is a major cause of both malnutrition and death, a positive association between aggregate indicators of underweight and mortality would be expected even if malnutrition did nothing to increase risks of death. Deviations from the expected pattern at the aggregate level require further investigation. Further research might show that factors such as access to health care or maternal education mediate the relationship between body size and mortality; small stature may be a significant but surmountable mortality risk factor. If other variables, in fact, play an important mediating role in the body size/survival relationship, high proportions of underweight, wasting, or stunting might accurately reflect important variations in nutritional adequacy rather than simply measurement problems.

In summary, two additional conclusions about the desirability of individual adaptation to low food availability deserve emphasis. First, as outlined in chapter 1, limitations on physical activity due to lethargy during childhood have both physical and intellectual consequences. Second, stronger manual labourers may be able to earn more than enough to compensate for the increased caloric needs associated with their larger size. Although it seems likely that individual adaptation to low intake occurs regularly, anthropometric measurements are still useful hunger indicators.

Distinctive growth patterns of breast-fed and formula-fed infants

Applying US-based anthropometric standards to infants is another controversial area, since the US standard reflects the experience of mostly formula-fed infants who were supplemented with solid food fairly early in life (Akin and MacLean 1980). Furthermore, at the time the data for the standards were collected, more nutrient-dense infant formulas were used than is now the case. Under circumstances where adequate amounts of formula can be given and hygiene maintained, the use of modern infant formula leads to more rapid weight gain, at least after the first few months, than does breast-feeding (Ritchie and Naismith 1975; Stuff and Nichols 1989). This divergence in growth trajectories has been conventionally interpreted as showing that unsupplemented breastmilk is sufficient only for the infant's first four to six months, the period before the growth paths diverge. It is therefore recommended that other foods should be added to the infant's diet even if breast-feeding continues beyond the first four or six months (see, e.g., Underwood and Hofvander 1982).

Nevertheless, a number of researchers have argued that the standard reflects a less-than-optimal growth pattern (Akin and MacLean 1980; Huffman 1991; Whitehead and Paul 1984). Whitehead and Paul argued that there was no reason for concern when children fall behind standards based on "inappropriately constituted and administered formulae." This conclusion seems especially appropriate in view of medical evidence that breast-fed children who show growth faltering relative to the standard can be shown to be just as healthy or healthier, according to other measures. They have fewer respiratory infections and less incidence of diarrhoea (Chandra 1982), and their energy requirements are lower because of their smaller body size, lower heart rates, and lower metabolic rates - not because of lower levels of physical activity (Garza and Butte 1990). They are also unlikely to be undernourished since their caloric intakes do not increase when their diets are supplemented with solids (Garza and Butte 1990; Stuff and Nichols 1989).

Such findings lead researchers to question whether all infants ought to follow the growth patterns that can be achieved with infant formula. If more rapid growth is not advantageous, application of standards based on the experience of bottle-fed infants may overstate the prevalence of underweight in populations where most children are breast-fed and may also lead to a perception that supplementation is needed at ages at which breastmilk still fully meets the infant's needs. Where surrounding conditions make it difficult to maintain good hygiene, unnecessarily early introduction of supplementary foods that are likely to be contaminated may increase rather than lessen risks of malnutrition.

Growth standards still need to be developed that reflect the experience of exclusively breast-fed children and children receiving non-formula supplements. Until that time, cross-country comparisons using anthropometric measurements should be interpreted with special caution for the youngest children. The main difference in growth patterns between breast-fed and formula-fed children is faster weight gain in the formula-fed group (Garza and Butte 1990), and the greatest difference in weight gain is associated specifically with formula feeding, not other methods of artificial feeding. Therefore, comparisons of stunting (height for age) are less likely to be affected by use of the US-based standards than are comparisons of underweight. The greatest caution needs to be applied when comparing weight gain in populations where breast-feeding is common versus those where commercial infant formula is widely used.

How much difference does choice of growth standard make?

As the discussion above suggests, selection of an appropriate reference standard for anthropometric measurements is a contentious issue. For international purposes, use of the National Center for Health Statistics (NCHS) standard (based on the experience of children in the United States) has been recommended by the World Health Organization and is now widely accepted. Some countries, however, have chosen to develop and use local standards instead. Typical growth patterns do vary across populations, and the issue is which deviations ought to be viewed as problematic and avoidable.

Implications of this choice for our understanding of hunger are not trivial (Millman et al. 1991). Different anthropometric standards can yield very different estimates of the prevalence of malnutrition in the same population. For India, for example, estimates of the prevalence of underweight among children based on the NCHS or the local (Hyderabad) standard differ by 25.7 percentage points for 1989 (NNMB 1989). If local growth standards reflect the experience of less-than-healthy adaptation, their use will define real nutritional problems out of existence (Messer 1986).

Less obvious, but equally problematic, is that the choice of standard can also affect the analyst's understanding of which groups within a population are worse off. In particular, the contrast between the prevalence of malnutrition observed for males and females can be very much affected by the standard employed. Within each standard, separate reference values are defined for males and females, a complication necessary to capture typical healthy growth patterns for boys and girls. Any standard embodies the pattern of gender contrast that typifies the population on which it is based. A standard based on a population in which treatment of boys and girls differs in nutritionally consequential ways essentially defines the resulting differentiation of developmental paths as the norm. For example, the Hyderabad standard used in India, which is based on the experience of a population of urban middle-class children in southern India, incorporates a pattern of male advantage as compared with the US-based and internationally used NCHS standard. Table 2.2 contrasts median weights by age and sex in the two standards. While the Hyderabad standard in general defines lower weights as normal than does the NCHS one, the point here is that the downward shift in median weight associated with the use of the Hyderabad standard is greater for females than for males.

Table 2.2 Median weights (kg) for age and sex compared between Hyderabad and NCHS standards



Age (years)



































Source: for the Hyderabad standard, NNMB (1989); for the NCHS standard, Dibley et al. (1987).

Distributions of weight for age that imply the same prevalence of malnutrition for boys and for girls as compared with the NCHS standard would inevitably show higher rates of malnutrition for boys than for girls if the Hyderabad standard were employed. Conversely, a situation that appears to be one of gender equality in malnutrition relative to the Hyderabad standard would show a female disadvantage if the presumably non-gender-biased NCHS standard were employed.

Table 2.3 shows the sharply different patterns of gender contrast that are observed when the same situation is viewed through the lens of one or the other weight-for-age standard. Tabulations of 1989 data for seven states of India (NNMB 1989) based on the Hyderabad standard show an apparent nutritional advantage for girls, startling in view of the frequency with which one hears that boys are favoured in that country. This surprising result is at least partly due to the fact that individual data are being measured against a standard that has a male advantage built into it. When the same data are measured against the presumably non-gender-biased NCHS standard, the apparent female advantage tends to disappear, although the expected male advantage still fails to become visible. We will return to the question of gender differences in nutrition in chapter 5. For present purposes, the important point is that the choice of reference population itself strongly affects the gender contrast we witness in anthropometry.

Table 2.3 Gender comparisons Underweight Indian children according to the Hyderabad and NCHS standards, 1989


Percentage of boys underweight

Percentage of girls underweight

Female advantage (boys - girls)









Source: NNMB (1989).

Relations among the hunger indicators

The assumption is sometimes made that patterns of variation or change observable in child anthropometry indicate variation or change in nutritional status for the population of all ages. The prevalence of underweight among small children is used as a leading indicator of malnutrition in famine early warning systems, and cross-sectional variations in underweight among small children are taken as indicators of likely concentrations of malnutrition at other ages as well. Little attempt seems to have been made to validate this wider application of the findings on child anthropometry by exploring its association with hunger indicators pertaining directly to other age groups. Given the crucial importance, for malnutrition among small children, of processes such as weaning and childhood diseases that are irrelevant to others, the use of childhood anthropometry as an indicator of nutritional conditions for adolescents and adults seems questionable.

As Heyer (1991) observed in her analysis of Kenyan data:

Child malnutrition is not at all closely linked with... poverty (whether measured in terms of income or expenditure)... or even food intake estimates. This is consistent with micro-level evidence on the role of health and other factors.

Similarly, for a low-income sample in the Philippines, Pinstrup-Andersen (1990) found only low correlations between child anthropometry and a wide range of household-level indicators - per capita household income, per capita food acquisition, per capita calorie consumption, household calorie adequacy, total household food acquisition, and total household calorie consumption. The highest correlation was only.22. The very weak relationships between child anthropometry and other hunger indicators suggest that children's growth impairment is not a useful indicator of household food security. The authors were actually asking the opposite question - whether household food-security indicators would serve to identify households in which malnourished children were located. The answer to this question was also negative. In contrast, caloric adequacy of pregnant and lactating women was reasonably strongly related to that of their households, suggesting that nutritional problems for this group are more a function of household food insecurity and also discounting the interpretation that poor measurement of household data could account for the lack of relationship with child anthropometry.

To identify linkages between individual and household hunger, which is essential for diagnosing nutrition problems and setting priorities for interventions, empirical work using data covering a broad range of ages and including multiple indicators of hunger needs to be given a high priority. Such work might explore, for example, the extent to which underweight children are concentrated in households with low access to food, and the covariation of anthropometry for children and adults within the same households. If it turns out that underweight among small children acts as a reliable proxy for hunger of others in the same household, location, or social group, the wide availability of childhood anthropometry could be exploited more systematically to enhance our overall understanding of hunger in entire populations. If, on the other hand, variations in childhood anthropometry diverge sharply from those reflected by other hunger indicators, the temptation to generalize widely from data pertaining directly only to small children should be resisted. In the meantime, it is safer to interpret changes in child anthropometry within a region or social class as indicative of changes in the hunger status of entire families, but not without first considering whether changes in infant feeding practices or in the disease environment might provide an adequate explanation for the trends.


1. The World Bank most commonly used a cut-off set at 90 per cent of the calorie requirements estimated by the FAO/WHO/UNU committee in 1971; the FAO defined its cut-off as 1.4 times the basal metabolic rate (BMR). In neither case do the cut-offs employed allow for more than minimal physical activity for adults, and the newer common cut-off of 1.54 BMR still allows only for light activity (Uvin 1994).

2. Counting breast-feeding bouts over shorter periods is problematic because daytime consumption may or may not be reflective of night-time consumption: when children sleep with their mothers, nursing may follow a similar pattern around the clock; where they do not, there may be little or no night-time nursing.

3. Sukhatme and Margen (1982) interpret interindividual variation observed cross-sectionally as reflecting intra-individual variability; they also interpret the autocorrelation of daily individual intakes as evidence of a homoeostatic, self-regulating process. Although neither of these interpretations is implausible, interindividual variation could reflect stable differences across individuals and autocorrelation of intakes could result from external influences that vary cyclically over a span of days (such as different eating patterns on weekends and weekdays). Even if the evidence for energy intake and efficiency of use as a self-regulating process were definitive, the conclusion that intake levels observed only as the low point in a fluctuating series could be maintained indefinitely without damage seems questionable.

4. Smallness on any of the anthropometric indicators may result from illness rather than from compromised nutrition, though in most cases it is likely to be a combination of the two. Smallness may also result from normal variation or genetic potential, and one of the challenges this presents is to set cut-off points for anthropometric measurements that identify nutritional problems without also including children who are simply small.

5. Mid-upper arm circumference has been shown to predict risk of death better than either weight for height or height for age (Briend et al. 1987).

Works cited

Ahn, Chung Hae, and William C. MacLean. 1980. "Growth of the Exclusively Breast-Fed Infant." American Journal of Clinical Nutrition 33: 183-192.

APS. 1992. Nigeria's Census Confounds Experts. All Africa Press Service, APS News Bulletin, 30 March. Nairobi, Kenya: All Africa Press Service.

Atwood, D. A. 1991. "Aggregate Food Supply and Famine Early Warning." Food Policy 16, No. 3: 245251.

Beaton, G. H. 1989. "Small But Healthy? Are We Asking the Right Question?" Human Organization 48, No. 1: 30-38.

Briend, A., M. G. M. Rowland, and B. Wojtyniak. 1987. "Measures of Nutritional Status." Lancet 1, No. 8541: 1098-1099.

Cameron, M. E., and W. A. Van Staveren. 1988. Manual on Methodology for Food Consumption Studies. Oxford: Oxford University Press.

Chandra, R. K. 1982. "Physical Growth of Exclusively Breast-Fed Infants." Nutrition Research 2: 275-276.

Dasgupta, Partha, and Debraj Ray. 1990. "Adapting to Undernourishment: The Biological Evidence and its Implications." In: Jean Drèze and Amartya Sen, eds. The Political Economy of Hunger, Volume I. Oxford: Clarendon Press, pp. 191-246.

den Hartog, A. P. 1973. "Unequal Distribution of Food Within the Household (A Somewhat Neglected Aspect of Food Behaviour)." FAO Nutrition Newsletter 10, No. 4: 8-15.

Dibley, M. J., J. Goldsby, M. Strehling, and F. L. Trowbridge. 1987. "Development of Normalized Curves for the International Growth Reference: Historical and Technical Considerations." American Journal of Clinical Nutrition 46: 736-748.

FAO. 1977. The Fourth World Food Survey. Food and Nutrition Series. Rome: Food and Agriculture Organization of the United Nations.

_____. 1983. A Comparative Study of Food Consumption Data from Food Balance Sheets and Household Surveys. FAO Economic and Social Development Paper No. 34. Rome: Food and Agriculture Organization of the United Nations.

_____. 1984. Food Balance Sheets 1979-81 Average. Rome: Food and Agriculture Organization of the United Nations.

_____. 1985. The Fifth World Food Survey. Food and Nutrition Series. Rome: Food and Agriculture Organization of the United Nations.

_____. 1990a. "Action Programmes to Overcome Specific Nutritional Deficiencies in the Asia-Pacific Region." Twentieth Regional Conference for Asia and the Pacific in Beijing, China. Rome: Food and Agriculture Organization of the United Nations.

_____. 1990b. "Malnutrition in the Latin American and Caribbean Region: Causes and Prevention." Twenty-First Regional Conference for Latin America and the Caribbean in Santiago, Chile. Rome: Food and Agriculture Organization of the United Nations.

_____. 1990c. "Strategies for Combating Malnutrition in Africa." Sixteenth FAO Regional Conference for Africa in Accra, Ghana. Rome: Food and Agriculture Organization of the United Nations.

Garza, C., and N. F. Butte. 1990. "Energy Intakes of Human Milk-Fed Infants During the First Year." Journal of Pediatrics 117, No. 2: S124-S131.

Grant, James P. 1990. The State of the World's Children 1990. New York: Oxford University Press for UNICEF.

Heyer, J. 1991. "Poverty and Food Deprivation in Kenya's Smallholder Agricultural Areas." In: J. Drèze and A. Sen, eds. The Political Economy of Hunger, Volume 111. Oxford: Clarendon Press.

Huffman, Sandra. 1991. "Maternal Malnutrition and breast-feeding Is There Really a Choice for Policy Makers?" Journal of Tropical Pediatrics 37, Suppl. 1: 19-22.

Jelliffe, Derrick B., and E. F. Patrice Jelliffe. 1989. Community Nutritional Assessment: With Special Reference to Less Technically Developed Countries. Oxford Medical Publications. Oxford: Oxford University Press.

Katona-Apte, Judit. 1975. "The Relevance of Nourishment to the Reproductive Cycle of the Female in India." In: Dana Raphael, ed. Being Female: Reproduction, Power, and Change. The Hague: Mouton, pp. 43-48.

Laderman, Carol. 1991. "Where the Wild Things Are." In: Anne Sharman, Janet Theophano, Karen Curtis, and Ellen Messer, eds. Diet and Domestic Life in Society. Philadelphia: Temple University Press, pp. 15-32.

Maher, Vanessa. 1981. "Work, Consumption and Authority with the Household." In: Kate Young, Carol Wolkowitz, and Roslyn McCullagh, eds. Of Marriage and the Market: Women's Subordination in International Perspective. London: CSE Books, pp. 69-87.

Martorell, Reynaldo. 1995. "Promoting Healthy Growth: Rationale and Benefits." In: Per Pinstrup-Andersen, David Pelletier, and Harold Alderman, eds. Child Growth and Nutrition in Developing Countries. Ithaca: Cornell University Press, pp. 15-31.

Messer, Ellen. 1986. "The 'Small But Healthy' Hypothesis: Historical, Political, and Ecological Influences on Nutritional Standards." Human Ecology 14, No. 1: 57-75.

Millman, Sara Ruth, and Robert S. Chen. 1991. Measurement of Hunger: Defining Thresholds. RR-91-1. Providence, Rhode Island: Alan Shawn Feinstein World Hunger Program.

_____,_____, J. Emlen, V. Haarmann, Jeanne X. Kasperson, and Ellen Messer. 1991. The Hunger Report: Update 1991. HR-91-1. Providence, Rhode Island: Alan Shawn Feinstein World Hunger Program.

Mosley, W. Henry, and Lincoln C. Chen. 1984. "An Analytic Framework for the Study of Child Survival in Developing Countries." Population and Development Review 10 (Suppl.): 25-48.

National Research Council. 1986. Nutrient Adequacy: Assessment Using Food Consumption Surveys. Washington, D.C.: National Academy Press.

NNMB. 1981. Report for the Year 1980. Hyderabad, India: National Institute of Nutrition.

_____. 1989. Interim Report of Repeat Survey (Phase I) 1988-89. Hyderabad, India: National Institute of Nutrition.

Osmani, S. R. 1992. "On Some Controversies in the Measurement of Undernutrition." In: S. R. Osmani, ed. Nutrition and Poverty. Oxford: Clarendon Press, pp. 121-164.

Papanek, Hanna. 1990. "To Each Less Than She Needs, From Each More Than She Can Do: Allocations, Entitlements, and Value." In: Irene Tinker, ed. Persistent Inequalities: Women in World Development. New York: Oxford University Press, pp. 162-181.

Pinstrup-Andersen, P. 1990. "Data on Food Consumption by High-Risk Family Members: Its Utility for Identifying Target Households for Food and Nutrition Programmes." In: B. L. Rogers and N. P. Schlossman, eds. Intra-Household Resource Allocation: Issues and Methods for Development Policy and Planning. Tokyo: United Nations University Press, pp. 164-175.

Popkin, Barry M. 1994. "The Nutrition Transition in Low Income Countries: An Emerging Crisis." Nutrition Reviews 52, No. 9: 285-298.

Puffer, R. R., and C. V. Serrano. 1973. Patterns of Mortality in Childhood. Pan American Health Organization, Scientific Publication No. 262. Washington, D.C.: PAHO.

Ritchie, C. D., and D. J. Naismith. 1975. "A Comparison of Growth in Wholly Breast-Fed Infants and in Artificially Fed Infants." Proceedings of the Nutrition Society 34, No. 3: 118A.

Scrimshaw, N. S., and V. R. Young. 1989. "Adaptation to Low Protein and Energy Intakes." Human Organization 48, No. 1: 20-29.

Stuff, Janice E., and Buford L. Nichols. 1989. "Nutrient Intake and Growth Performance of Older Infants Fed Human Milk." Journal of Pediatrics 115: 959-968.

Sukhatme, P. V. 1988. "Energy Intake and Nutrition: On the Autoregulatory Homeostatic Nature of Energy Balance." In: T. N. Srinivasan and Pranab K. Bardhan, eds. Rural Poverty in South Asia. New York: Columbia University Press, pp. 365-388.

Svedberg, P. 1991. "Undernutrition in Sub-Saharan Africa: A Critical Assessment of the Evidence." In: I. Drèze and A. K. Sen, eds. The Political Economy of Hunger. Volume III. Oxford: Oxford University Press.

THP. 1983. "Where We Stand Today" (Editorial). The Hunger Project 17: 6-7.

Tomkins, A., and F. Watson. 1989. Malnutrition and Infection: A Review. United Nations, Administrative Committee on Coordination, Subcommittee on Nutrition, ACC/SCN State-of-the-Art Series Discussion Paper No. 5. Geneva: World Health Organization.

Underwood, B. A., and Y. Hofvander. 1982. "Appropriate Timing for Complementary Feeding of the Breast-fed Infant: A Review." Acta Paediatrica Scandinavica Suppl. 294.

UNDP. 1992. Human Development Report 1992. United Nations Development Programme. New York: Oxford University Press.

Uvin, Peter. 1994. "The State of World Hunger." Nutrition Reviews 52, No. 5: 151161.

Waterlow, J. C. 1989. "Observations on the FAP's Methodology for Estimating the Incidence of Undernutrition." Food and Nutrition Bulletin 11, No. 2: 8-13.

Whitehead, R. G., and A. A. Paul. 1984. "Growth Charts and the Assessment of Infant Feeding Practices in the Western World and in Developing Countries." Early Human Development 9: 187-207.

WHO. 1985. Energy and Protein Requirements. Report of a Joint FAO/WHO/UNU Expert Consultation. World Health Organization, Technical Report Series 724. Geneva: WHO.

Wilson, Christine S. 1974. "Child Following: A Technique for Learning Food and Nutrient Intakes." Journal of Tropical Pediatrics and Environmental Child Health 20: 9-14.

World Bank. 1992. World Development Report 1992: Development and the Environment. New York: Oxford University Press for the World Bank.