Household Food Security -- Principles for Measurement

The following sections are mainly concerned with the measurement of current household food security (4). This relates directly, in terms of decisions and interventions, to the methods of improving household food security put forward in the ICN Theme Paper by Von Braun et al., 1991, and as summarized in their Tables 2 and 3 (5). Household food security measurement refers to food consumed (or available for consumption) at the household level (6). This is difficult to measure. The implications of dietary intake need to be understood in deciding approaches to measurement, which are introduced below. First, the connection with other aspects of food security are noted. The methods here have been introduced in Table 2.

Food insecurity may affect households constantly, transitorily, or periodically; the latter may be seasonal, and occur most years, or may deteriorate into food crises in times of particular hardship, e.g., from drought or conflict (food crises have been covered in the previous section on timely warning). Chronic food insecurity is dealt with here, as part of chronic or endemic undernutrition; it is recognized that very often hunger and inadequate access to food is seasonal, certainly in rural areas. As mentioned above, the aspect of ensuring households' future access to food is not covered here because of need to limit the scope, vitally important though this is.

Methods of assessing and monitoring food security -- which may be referred to as food security information systems (FAO, 1992a, para 11) -- may focus on the need for cross-country comparisons and thus national-level data, or on within-country information, distinguishing and tracking vulnerable groups. This section is primarily concerned with in-country data (although clearly compilation of such may eventually give more global results) and emphasizes assessment for population sub-groups, defined by location, socioeconomic status, etc. The link with such indicators is along the following lines.

Implications of Dietary Intakes

These are illustrated in Figure 6, which raises several points. Here, average kcal intakes are shown from "low" to "high" without specifying the denominator or reference period. At the very low end, people lose weight reduce activity, indeed starve. At the high end they may gain weight. The majority, and the assessment problem, lies between these extremes. With inadequate food, people may reduce physical activity, favoring productive activity; much income (or expenditure) is then on food, with the cheapest foods being chosen; whatever the precise definition, such people are "food insecure" and often hungry. Particularly in agricultural societies, this food insecurity may be primarily seasonal, before harvests and at slack labor times of the year.

Problems in practical assessment of current household-level food consumption derive from complexities in both measurement and interpretation. One important aspect to which attention is now turning involves the time dimension (FAO, 1990). If it could be measured, daily household energy consumption would probably be seen to vary considerably day by day. Cross-sectional measurements of 24 hour intakes would pick up some well off households who happened to eat little that day, and vice versa. Daily intake is thus too fine a measure: one low intake day may not be meaningful, while a low-intake week probably is, and a low-intake month definitely is. For some purposes, such as global assessments, averaging over a period of one year is appropriate (because food balance sheet estimates are given by year) (FAO, 1992c).

For interpretation of household survey data for within-country planning, a reference period of a month is probably best. In terms of Figure 6, this would seek to position households along the horizontal axis of low to high dietary energy intake, defined as the average per caput intake for the household over one month. For interpretation, a scale is necessary to give an estimate of adequacy, that is: what constitutes "low", or "adequacy for what?" This relates to scaling the kcal intake as a factor * BMR (Basal Metabolic Rate). Figure 6 illustrates effects, responses, and controls over the range of kcal intakes.

Undernutrition could be defined as levels of intake below which productive activity is reduced because of food energy constraints; and food security (or adequacy) considered to be achieved within the area of discretionary activity, with appetite control over intake (as happens when food is plentiful). The "currently food secure" point is likely to be indicated by a number of behavioral factors, such as where percentage of income spent on food starts to drop (see Figure 6 B), behavior is less food-directed, activity is not constrained, and levels of living begin to take off from absolute poverty. All of these are reasons for improving household food security, and may also be useful for assessment (notably percent of income on food). In practice, the actual figure for kcal intake representing the transition between insecurity and security is probably around 1.6 - - 1.8 * BMR, which in turn is typically around 1800 - 2100 calories/caput/day. In global terms, a definition of the cut-off defining low kcal intake is "calorie intake on average over a one year period sufficient to maintain light activity and body weight" (FAO, 1992c).

Assessment of current household food security may be thus regarded as locating households in the range shown in Figure 6. Essentially, assessment methods involve direct or indirect estimates of food availability and consumption, in relation to need, primarily for activity.

It should be noted that we have so far considered estimates of cut-off points at individual household

level, below which the household will be food-insecure. If the average intake in a population is at this cut-off point, food-insecurity will still be prevalent depending on the distribution (and assessing this distribution is a major issue). As food becomes more available and the mean exceeds the cut-off, other control mechanisms take over -- notably appetite -- such that the intake begins to correlate increasingly with the requirement. Finally, when food is plentiful as in most developed countries, the balance between intake and expenditure for the individual and household is almost exact, and the correlation is nearly 100%, probably over quite short periods (days).

Should Levels or Changes in Household Food Security be Measured?

An important distinction is between levels of food intake, and changes in these. By levels is meant: actual kcal intakes averaged over, for example, one month, in relation to need. This is also referred to as household food security status (HFS). Changes means changes in these levels, again averaged over suitable periods, e.g. months. However, it is not always essential to know levels accurately in order to interpret changes. If it is reasonably likely that levels are marginally inadequate (from other indicators), then a change in the level towards deterioration is clearly cause for concern, and an improvement a sign of progress. This distinction is important, since it is probably considerably easier to measure changes in food intake than levels. This distinction is recognized in food security information systems, which should provide data for "identifying problems of vulnerable groups, and for following the evolution of the food situation over time" (FAO, 1992a, para 11).

How Necessary is it to Measure Levels of Food Intake?

There is a spectrum of opinion on this, but it comes down to, yet again, the implications for decisions that would be made on this basis. Essentially, if major decisions concerning resource commitment are to be made on the basis of food security considerations (for example, who is food insecure and why), then the cost of getting the necessary information on the HFS levels should not be an overriding constraint. For example, millions of dollars in government expenditure have been more efficiently allocated by changing food subsidy schemes based on such data, expensive though they were, and this has been regarded as well worthwhile. On the other hand, there are plenty of examples of expensive household food consumption surveys which have not led to any resource commitments or changes.

As a rule of thumb, it is likely that where there is a persistent significant prevalence of underweight children, HFS is inadequate; in other words, there is endemic undernutrition. This would indicate the need for interventions aimed at improving HFS. At this point, if the decision is made in principle to take action, specific designs of such interventions may require estimates of levels of HFS; these may however only be needed relatively infrequently, perhaps every five years or so.

It should be stressed again that whether information is expensive or cheap, whether it refers to HFS or other matters, it is only useful if used. Thus description of methods here does not constitute advocacy for their application. If for policy decisions one needs to assess levels of HFS (that is, how much people are eating) there is no alternative but to initiate an enquiry to measure these.

The assessment of the status of HFS or of its trend is unlikely by itself to indicate the necessary policy choice, for which additional research will usually be required. Perhaps in contrast to health, specific actions are

not always obvious. Many of the actions affecting household food security involve rather substantial investment of resources, so such research is well-justified. The degree of analysis may be less extensive when the assessment and decisions are made at local level, if the resources are under the control of the community (cf. "Triple-A", UNICEF, 1990a).

Vulnerable Groups, Risks, and Policy Options

Households and people at food security risk, and the risks they face, are described in the theme paper by Von Braun et al., (1991, see Table 2); they then give examples of the policy choices applicable to these risks and groups (summarized in their Table 3). Measuring household food security feeds into this in several ways. First, measuring household food security levels and differentials for vulnerable groups may indicate relative priorities and urgency for intervention for these groups. The definition of such groups -- e.g. small-holders growing a certain crop -- should thus be included in surveys. Second, exposure to risks varies by group (food producers versus net food purchasing households, with respect to risks from changes in food prices). This guides choice of indicators for assessing change in HFS -- the indicators are of risks. Some examples, based on Tables 2 and 3 of Von Braun et al., 1991, may be helpful.

One example might apply in rural areas, where small-holders may specialize in certain export crops, or be diversified between food and export crops; there may also be those with very small landholdings, or landless. Measurements of general nutritional status, for example by anthropometry, may give some initial indication of the probability of household food insecurity in such groups. If the anthropometric status is satisfactory, there may be less concern that household food insecurity exists. However, let us assume that there is a certain prevalence of underweight children in these groups, and therefore that interventions to improve household food security may be indicated. For effective policy planning, it may then be necessary to assess, at one time, the level of household food security by such groups. It might be found that the current food security situation for farmers growing a high value export crop (e.g. coffee) was good, whereas that for the more diversified farmers was moderate, but likely to be relatively stable. For the smallest landholders and the landless the situation might be of greater insecurity. Further considerations might then be that the coffee-growers, whilst presently doing well, were particularly vulnerable to change in producer prices, and these prices might then be monitored. For the diversified farmer, the main risk might be drought, which should be tracked. For the near-landless, longer term food security might be enhanced by credit schemes, public works, etc.

A second vulnerable group whose household food security status might require assessment could be the urban poor. Again, initial indications might come from prevalences of underweight children. The groups could be defined in terms of employment, occupation, female-headed households, etc. Assessment of levels of household food security could show relative priority amongst these groups, and point to possible policy interventions. Risk factors, such as food prices in the market, would be indicated for monitoring changing HFS.

A reasonable combination of measuring HFS status and changes in this might be as follows. HFS status could be estimated from large scale household budget surveys, perhaps every five years. If feasible, it would be useful to include actual weighing of food as purchased, produced, or taken from stock, at least on a subsample. In the intervening years, a much smaller sample might be used for updating the information, and/or reliance might be placed on indicators of changing HFS (e.g. food price-based), depending on the circumstances. When there is substantial investment in improving HFS, such expenditures are likely to be readily justified, in terms of improved efficiency of utilization of resources (including targeting), and evaluation of impact to guide long-term policy. Proposals for measurements of HFS status and change follow in the next two sections (the organization into these two sections is rather arbitrary, because repeated measurement of the status measurements can estimate changes.)

Assessing HFS Levels

Ideally, measuring HFS status or levels would involve measurement of household food availability and average household food consumption levels over a period of time, in relation to need (ACC/SCN 199la). As discussed above, there are considerable constraints in terms of measurement and interpretation.

This has led to exploration of other indicators of current HFS, often as a set. It is suggested that the following indicators and methods should be considered when it is necessary to assess levels of HFS.

• dietary energy intake (kcal/caput/day, averaged over the household, with specified reference period, preferably one month)
• percentage of total expenditure on food
• dietary patterns/choice of food
• experiences of hunger and food insecurity, by interview
• adult anthropometry (mainly thinness).

Dietary Energy Intake

Surveys of dietary energy intake (kcal/caput/day) are the basic way in which HFS status has been measured in the past, and will remain important whenever it necessary to make such measurements. As put forward below, these should not be the only method used. Still, if one needs to know whether people are eating enough, it makes sense to try to measure it directly. The principles also apply to surveys of other factors, such as proportional expenditure and dietary patterns (points 2 and 3 below).

Many different methods have been employed (with different names), often with much similarity (7). Here relevant characteristics are briefly described, and some recommendations made. Authoritative reviews of household survey methods, including income/expenditure and food consumption, are given in Casley and Lury, 1987 (p.167-173), and specifically for food consumption, in Cameron and van Staveren (1988). Food consumption surveys and their methods are regularly compiled by FAO (FAO, 1986). The characteristics can be seen as follows.

1. Such surveys have to be done at household level. Estimates of household consumption cannot be derived from national data, like food balance sheets, alone: an estimate of the distribution as well as the mean is required, which dictates household level data. Results should be expressed averaged over household members; accuracy can be improved by taking account of the household composition in terms of age and gender, expressing results as per "adult equivalent".
   
2. Household surveys can be done at one time, or be repeated over several rounds; an approximation to repeating for the same households over several rounds (which has technical problems of response) can be achieved by spreading the survey over the relevant period, e.g., one year. Assessing food consumption over at least a years’ period is very important, so repeated measurements, preferably as a panel of households revisited, are recommended.
   
3. The reference period to which the data apply is a central issue, needing to be carefully thought out and specified. As indicated above, it is suggested that a reference period of a month could be used; this means that the average intake over a month should be estimated. Adjustments can be made analytically if the reference period is shorter, as long as households record consumption in the previous week or day, but this is not ideal. A useful compromise would be to assess the average kcal intake for each household during one month at a time, repeated every three months for a year, i.e., four times. This would give estimates of typical consumption at four points in time, generally covering seasonal changes. These data would be so useful that it would probably be worth some sacrifice of sample sizes to achieve it (see point (e) below). Experience exists for example with the valuable studies carried out by a number of countries in collaboration with IFPRI (Pakistan: Garcia and Pinstrup-Anderson, 1987; Bouis and Haddad, 1990; Pakistan: Alderman and Garcia; Kenya: Kennedy, 1989).
   
4. The basic measurements on the questionnaire from which kcal estimates may be made (9) can be the following, included (for i - iii) within existing household survey procedures (most usually household expenditure or budget surveys):

1. expenditures by food item, with prices paid; this allows calculation of dietary energy intake, but better is:
2. recording at interview of quantities purchased, harvested, and/or taken from stores (if necessary in local measures);
3. a further alternative is to add a special module to a survey procedure to actually weigh the quantities of food, either as purchased/harvested/taken from store, or as prepared for eating; these may be beyond the scope of routine household surveys, thus if this detail is needed:
4. special surveys may be required to allow sufficient enumerator time to measure food quantities as prepared, or eaten.

For many purposes, i) is adequate, and may be all that is possible retrospectively. Prospectively, it is little additional burden in a household budget survey to include ii), recording of quantities purchased by interview (see Bouis, 1986), and generally this may be recommended.

e) The survey and sample design require particular attention here, because estimates of means and distribution parameters pose somewhat different problems. Distinguishing mean kcal consumption levels between groups may be achieved as discussed under (d), using the sample and interviewers from an existing household survey. However there will be much random error, although not too serious in affecting the estimate of the mean. But answering the question "how many people are not getting enough to eat?" will not be possible on this basis, because the estimate of the spread around the mean will be exaggerated. It may be worthwhile to do a separate study, with a smaller sample and more careful measurement, to establish the characteristics of the distribution (shape, SD, etc). This is a crucial point in connection especially with food availability or consumption estimates, where the level of random error is particularly large, and where there is a practice of reporting prevalences below a cut-off. Trade-offs in sample size will again need to be considered also, in order to achieve comparable repeated measurements within the year.

Food Expenditure Proportion

Data on percentage of total expenditure on food may be obtained fairly readily from household budget surveys. Since the poor spend a large part of their income on food -- precisely because of food insecurity -- this indicator may be particularly useful if interpreted well. The shape of the relationship between kcal intake and income (or expenditure, expressed as % on food) is usually following Engel's Law; that is, as income increases the proportion on food remains steady to begin with, often as high as 80% (Lipton, 1983). At a certain point the % of expenditure begins to fall, presumably as food needs begin to be satisfied -- this could be seen as entering the range where food security begins. (The observation that more expensive kcals are purchased around this point could also be used, but anyway does not greatly affect the argument.) Eventually, % food expenditure tends to plateau out at around 30% or so, when food is no longer a constraint. This relationship can be used to interpret food expenditure proportion as an indicator of the level of HFS.

The proportional expenditure for the food insecure (the flat portion on the left of the Engel’s Law graph) will be location-specific, so that some initial analysis will be required to establish the shape of the relationship. With this completed, it should be possible to then locate different groups (in terms of their probable HFS status) and compare them, as well as monitor change.

The proportional expenditure on food itself indicates the cost of achieving HFS. As pointed out by Jonsson and Toole (1991), "there is some distinction between households (a) that achieve food security at great cost, (b) those that achieve it at less cost, and (c) those that despite using a large proportion of available resources on food remain food insecure." The indicator of % expenditure on food identifies but does not distinguish (a) and (c), which may be adequate for many purposes. Together with food consumption data, these three can be distinguished.

Related data that can be obtained from household budget surveys would assess the variety of foods used. Under some circumstances useful indicators may be derived thus. A similar assessment could be obtained from the proportional expenditure on staple foods to the overall food expenditure; from the contribution of less preferred staples (e.g. cassava, sorghum) to the diet (see next section), and so on. These would merit investigation when access to household budget data is established. However, the indicator of % expenditure on food is likely to be the most generalizable.

In practical terms, these indicators are obviously most applicable when the surveys are anyway being carried out -- which applies periodically in many countries at the present time. Requirements are thus to have access to the data and adequate analytical capability.

Dietary Pattern and Food Choices

Further indicators of food security may be obtained from observations of dietary patterns, although these are less easy to discuss in general terms. Data may be obtained from household surveys as discussed in section 2 above. However, dietary patterns are particularly suitable for observation in qualitative or semi-quantitative terms by rapid assessment methods.

In this case, those most food insecure may be defined by such factors as: (a) buying the cheapest and least preferred staple, and (b) only buying small amounts of preferred foods, such as animal products. A similar indicator of numbers of meals per day cooked and/or eaten can be obtained by such methods, but is probably more relevant to monitoring change, and will be discussed later.

Rapid assessment procedures encompass a range of semi-quantitative and qualitative approaches to

efficiently gaining relevant information, less formal than by interview within a household survey. In particular, they do not usually use a defined sample, in line with their semi-quantitative nature. They are particularly relevant to gaining information, in the present context on topics such as dietary patterns, experiences of hunger (see next section) behavioral responses, etc. They are therefore also important in relation to timely warning, as discussed earlier.

The basic premises of RAPs have been described as: the participatory approach, methodological pluralism, and action orientation (Ndolamb, 1991). These have been applied to many subjects, notably health and nutrition. They are described in Scrimshaw and Gleason, 1992, which includes the proceedings of an international conference on the topic.

Experiences of Hunger and Food Insecurity

Hunger is closely related to household food insecurity. Although the precise meaning may be debated, the significance is clear to those who experience hunger, for themselves, or their families. Preventing hunger is the major humanitarian objective of addressing food security. And hunger, or fear of it, drives food-seeking behavior, leads to major sacrifices to obtain food, and may control the reduction of activity when food is scarce and people are hungry. Thus a direct measure of hunger would be very significant in assessing food security. In terms of the construct used here, it would locate household members on Engel’s Law graph just below the point at which appetite control takes effect.

Achieving freedom from hunger probably means much the same as attaining food security. In describing development of indicators to assess hunger, Radimer et al., 1990, first noted that "hunger definitions vary widely, and direct measures are uncommon" but based on their research justify a definition of hunger which is almost identical to that of food security. This research is very useful in demonstrating and testing an approach that could be much more widely used.

A considerable amount of the hunger in poor societies is seasonal. This is difficult to capture by quantitative survey methods, requiring more frequently repeated measurements than are usually possible, so that again direct assessment of the experience of hunger, retrospectively in relation to season, can help to fill this gap. The approaches now coming to be used are straightforward, using carefully designed questions. A shift to a direct approach in this area seems overdue, and has some similarity to the successful demonstration that reasonable estimates can be made of farmers' production by directly asking, compared with more complicated methods of crop cutting, etc. (Scott, Marchant and Verma, 1989). The questionnaires can be included in rapid assessment procedures, and within more formal household surveys where the sample is defined -- the latter has some advantage in that prevalence estimates of hunger can thus be produced on a population basis.

The components of a definition of hunger that have been used in designing questions include quantity, quality, psychological factors (leading to food anxiety in the household, and deprivation among individuals) and social dimensions (unacceptable means for acquiring food, disrupted eating patterns, etc -- see Radimer et al., 1991, Table 1). This has led to development of questions to identify hunger in households, among women, and for children. For example: whether the household is anxious that food will run out, whether women consider they cannot afford to eat they way they should, whether mothers can give their children a balanced meal, and whether they consider they are not eating enough because they cannot get access to enough food. Already such inquiries provide results showing, for example, that children tend to experience hunger later than their mothers, because their mothers will sacrifice their own intake first. The third National Health and Nutrition Examination Survey (NHANES) in the U.S. now includes specific questions to detect hunger in families, for example to determine whether meals were skipped because of lack of food or money. These approaches emphasize the involuntary nature of food restriction, an important concept in developed countries (quoted from Campbell, 1990).

The approach of directly asking people whether they are hungry has been used in India, in the National Sample Survey in 1983 (as described by Minhas, 1990). A single question was included in the questionnaire: whether all members of the household got two square meals a day throughout the year. The results indicated that around 19% of the sample replied "no" to this question. There was considerable variation between states, up to 40% reporting inadequacy in West Bengal, for example, and the results appeared of considerable interest.

A further recent example of such direct questioning is from a survey in the U.K. of low income families. Questions were on parents going short of food in the last 12 months, or last month, to ensure enough food for their children. Answering negatively to "do you usually find you have enough money for food all week?" and similar questions gave important results (NCH, 1991).

There is important potential for such direct estimates of the experience of hunger, which would need

to be developed taking account of local circumstances. Very probably greater inclusion of such questions in household surveys, both formal and by rapid assessment, would give useful information for planning interventions to prevent hunger and food insecurity.

Adult Anthropometry

Anthropometry in children has been described above as a general nutritional outcome indicator. It has been suggested that changes in adult anthropometry are more specifically related to changes in household food security (James, 1991). Whilst it is likely that an association of adult anthropometry with household food security exists -- although this has not been investigated very extensively -- practical considerations also intervene in recommending on the extent of its future use. To begin with, since child anthropometry is much more common, adult anthropometry would have to be seen to have a distinct advantage over that of children to shift attention to it.

Specifically, thinness in adults measured by weight in relation to a function of height (e.g. weight/height, known as Body Mass Index, or BMI) has been put forward as a definition of chronic energy deficiency, itself seen as a stable state where low weight is maintained with energy balance between a deficient intake and reduced activity (James et al., 1988a). Initially, it was suggested that with moderate thinness estimates of food intake and/or activity would be necessary; however, it is now suggested that thinness measured alone could define chronic energy deficiency (James, 1991, p.i). It is also expected that reduced activity imposed by food problems is generally accompanied by loss of body weight (Payne and Lipton, 1989), again implying that thinness should locate individuals on the food security scale shown in Figure 6.

It should be noted that the issue is thinness, rather than the precise index of thinness. This can be assessed as weight in relation to height, in adults as absolute weight sometimes, by measures such as arm circumference, etc. Frequently the index of weight/(height squared), or BMI, (also known as Quetelet's) has been adopted; if adequate standards were available, an index equivalent to weight/height, which is familiar in children, could be used.

Thinness in adults has been convincingly shown to be associated with functional disadvantage, including lowered work output, increased vulnerability to infection, poor pregnancy outcome, etc. (James, 1991). There is also good evidence that weight changes in adults (which must be related to thinness, because height does not change) fluctuates seasonally, and is related to socioeconomic status, age, etc. (Seasonal examples are given in ACC/SCN, 1989b; associations are given in James, 1991). It is thus likely to be a reasonable measure of HFS, but data to show this are still scarce. The major problem with including adult anthropometry in HFS assessments is the relative difficulty of its measurement particularly where weight and height are concerned.

The first issue is therefore whether adult anthropometry is practical and offers advantages over other indicators; the second is which would be the most practical indicator. This remains a matter for research. It would be particularly appealing if measurements such as arm circumference, to which there is little resistance, and adequately measured thinness, could be standardized in a way similar to BMI. There is some evidence for this, e.g., Huffman et al., 1985. Nonetheless, it would be necessary to calibrate alternative measures such as arm circumference in terms of causes and functional consequences, perhaps in line with BMI, for which the cut-off points of 16.0 meaning severe thinness, and 18.5 meaning moderate thinness, are now well established (James et al., 1988a). However, the main priority for investigation in this context remains how far thinness can pick up changes in kcal availability and consumption. Evidently, the reason it may not is because of buffering by reduced activity, as indicated in Figure 6.

Monitoring Changes in HFS

Changes in HFS are probably easier to assess than status, at least insofar as different proxy indicators of change may be used. The relatively simple methods for assessing change -- compared to status -- offer a particular opportunity for improving the information available for decisions on improving HFS.

To begin with, clearly changes in HFS can in principle be satisfactorily monitored by repeating the measurements discussed in the previous sections; this should evidently be done when the capability exists for periodic measurement of HFS status. Such monitoring of HFS changes can be achieved through repeated household surveys, re-estimating indicators such as household intakes of kcal/caput/day averaged over one month. Such assessment of trends in kcal intakes from repeated surveys needs to take careful account of seasonal factors, changes in household composition, etc. Changes in proportion of expenditure on food may equally be determined from repeated surveys, and again be informative when changes are viewed by, for example, expenditure bands. Similar results can show changes in dietary patterns, again generally interpretable in terms of improving or deteriorating HFS by viewing relative proportion of staples, etc. (although account needs also to be taken of food prices in this).

The main opportunity however is to set up systems, often based on data already being collected, to track likely changes in HFS. This concerns especially two categories of information: based on prices, sometimes with wages; and based on changes in anthropometry, often derived from clinics. Importantly, these are now known to be related. In fact large food price-based indicator changes sometimes precede major changes in anthropometry, thus allowing prediction. Some of the empirical data has been summarized in Tabatabai, 1989, and ACC/SCN, 1992b; possible applications put forward by ACC/SCN, 1989a. Recent results from Sudan show that deteriorating terms of trade between livestock sales and grain purchases preceded increases in child wasting. (Teklu et al., 1991; von Braun, 1991). Similar results have been observed in Sahelian countries (e.g. Khan et al., 1991).

Experience to date is that the best predictions of anthropometric changes are obtained from food prices as a ratio to the general price index (Tabatabai, 1989; Kelly, 1989); or to minimum wages when these are adequately recorded and actually vary. This relationship suggests some innovative low-cost systems that could be set up. First however, price-based and anthropometric indicators are discussed individually.

Price-Based Indicators

These have been reviewed in the context of timely warning, where it was suggested that some indicators derived from food prices in relation to general prices seemed to give short-run (e.g. three months) predictions of changes in nutritional outcome, at least under conditions of potential food crisis.

Food price indicators may be analyzed in relation to the minimum wage, and results may be expressed as, for example, number of days work required to purchase a food basket, or other standard food measure. One drawback in using minimum wages itself, is that this tends to vary infrequently, but when it does produces a step in the indicator.

Deriving such indicators usually depends on relatively simple calculations from routinely reported government statistics such as the consumer price index. This is available at least from urban areas in many countries, and is generally reported quite rapidly. It may be worth considering when HFS is marginal, establishing additional sites for reporting market prices in vulnerable areas.

A recent example of using such an indicator is from Malawi, where a sharp increase in the hours of work required to purchase a standard bag of maize was used to trigger an increase in the minimum wage (Government of Malawi, 1991).

Changes in Anthropometric Indicators

Whereas long-term (e.g. year-to-year) changes in anthropometry are regarded as a general outcome, not specific necessarily to household food security, rapid changes in anthropometric indicators (unless there is an epidemic) are likely to reflect household food security changes at community level. Rapid change in prevalence of underweight can only represent wasting. The seasonal patterns detected on analysis of clinic-derived data appeared to confirm their relationship to food security, as discussed in ACC/SCN 1989b, pp 182-183: seasonal patterns always coincided with the pre-harvest hungry periods; longer-run trends have been observed to coincide with known food shortages; and the correlation of underweight prevalences with food price derived data supported the argument.

Thus it is suggested that monitoring month-to-month or quarter-to-quarter changes in underweight prevalences in pre-school children, from clinics or sentinel sites, could provide reasonably up-to-date assessments of changes in HFS. (This idea is elaborated in the context of structural adjustment in ACC/SCN 1989a).

Sentinel Site Systems

This possibility has already been put forward in the context of timely warning, for which it has the advantage of rapidity of reporting. In the context of longer term planning the approach also has merits, namely for tracking changes in HFS through price and anthropometric indicators. In this case it may be broadened to allow follow-up to investigate changes, through rapid assessments or small-scale surveys of some of the factors discussed earlier, such as changing dietary patterns, and experiences of hunger.

(4) Near and distant future food security are also important, but beyond the present scope, see Gillespie and Mason, 1991, table 1.1

(5) More details of intervention options and relations to measurement as discussed by the ACC/SCN are in ACC/SCN, 1991a, and Gillespie and Mason, 1991, p. 30-54.

(6) "An ideal measure of household food security includes the measurement of household food availability and average household food consumption levels over a period of time, in relation to need" (ACC/SCN, 1991a, p.7)

(7) e.g. "... there is very little that has to he included in a food consumption survey that is not found in many budget surveys ..." Casley and Lury, 1987, p.171.

(8) This issue is elaborated in the context of national and global estimates by FAO in the methodology for the 6th world food survey, and for the global assessment for the ICN.

(9) Much detail is given in Cameron and van Staveren, 1988.