Contents - Previous - Next
This is the old United Nations University website. Visit the new site at http://unu.edu
Urbanization can affect the nutritional status of a formerly agrarian population in many ways. Some effects operate through changes in food production and distribution patterns, while others depend on consumers' responses to changes in income, family size and composition, and the availability and prices of foods that accompany the process of urbanization. One way to assess the effects that depend on consumer behaviour is to compare urban and rural food consumption and their determinants at the same point in time. This study carried out such a comparison of urban and rural data on north-east Brazil collected by the Estudo Nacional de Despesa Familiar (ENDEF) from 1974 through 1975.
Most of what is known about household food consumption in Brazil, as well as most of what is known about the nature and prevalence of malnutrition, is based on the large sample used in the ENDEF. There have been numerous other household budget studies in various parts of the country and also many small studies of malnutrition and its correlates , but none of these studies used samples comparable in size to that of the ENDEF. Moreover, the smaller studies did not consider food consumption and malnutrition jointly. Studies of nutritional status, in particular, have tended to focus on anthropometry to the exclusion of any economic explanatory variables. On the other hand, budget surveys have excluded biomedical information because it greatly increases the complexity and cost of a study.
The disaggregated, family-level information from the ENDEF has never previously been used to study how food consumption is related to such likely determinants as household income or expenditure, food prices, and family size and composition. Because of the quantity of data and their complexity, it took three years to publish many of them, and they were aggregated across families by income-class and/or geographic region . Subsequent research on food consumption  and nutritional status [10, 11] has used these data or similar aggregated results to examine the effects of income and prices on food consumption. There has been no comparable research on how household size affects food consumption, because families of different sizes were combined in the data sets within classes of total income or expenditure. Even evaluations of nutritional status or of the effects of food and nutrition programmes on it have typically not considered family size .
Previous ENDEF-based research, in addition to being incomplete in these respects, has produced two somewhat paradoxical findings: The first is that income has been identified as the most important factor in explaining nutritional status, but the income elasticity of food consumption is quite low [4, 11]. If these are both true, either food consumption and nutritional status are not so closely associated as one would expect, or poor households do not recognize malnutrition and therefore do not devote additional income to overcoming it. Either of these conclusions would suggest that disease or knowledge of food habits and preferences also have large effects on nutritional status . The second paradoxical finding is that poor households appear to be extremely sensitive to small price changes, at least for certain combinations of basic foodstuffs . Some of these results imply unlikely degrees of superiority or inferiority of particular foods; others are inconsistent with the apparent income responses.
There are a large number of federal government programmes in Brazil that either donate fixed amounts of one or several foods to poor families or to particular members of those families, usually young children and sometimes pregnant or nursing women as well. Other programmes subsidize a variety of foodstuffs, or try by interventions in the food-marketing system to reduce or stabilize food prices. The effectiveness of such programmes obviously depends on how the beneficiaries react to income transfers and price reductions, but previous evaluations of the programmes have dealt almost exclusively with anthropometric outcomes and have not tried to estimate effects on food intake .
The present study tries to explain how food consumption depends on income, prices, and household size, and how the relationships between consumption and these variables differ in cities and rural areas. The purpose of this investigation is to provide a basis for evaluating food and nutrition programmes and for testing some of the conclusions drawn from previous research with the ENDEF data. In order to avoid problems attributable to regional differences in food preferences, the study is limited to the North-east Region of Brazil (region V in the ENDEF classification), which comprises seven states. Sixteen staple foods are studied. Because no fruits or vegetables, except beans, are included, the consumption findings refer to calories and protein but not to vitamin and mineral micronutrients. The unit of observation in this study is the household, because the ENDEF data are exceptionally complete and reliable at this level. There is, however, no information in the ENDEF data on how food consumption is distributed among the individual members of a household. For this reason, the nutritional status of individuals is not considered, and the analysis explains only household food intake.
Parameter estimates and poverty levels
Equation 1 in the appendix to this paper was estimated for each of 16 basic foodstuffs, once for the three North-east metropolitan areas included in the ENDEF (Fortaleza, Recife, and Salvador) and once for rural areas. A separate estimate was not made for cities of intermediate size.
There are two reasons for estimating the functions separately instead of introducing another variable to distinguish urban and rural areas. The first is that high incomes are concentrated in cities, particularly in metropolitan areas. If the model does not represent consumer behaviour equally well at different income levels, there may be systematic differences between estimates that include higher-income families and others that do not. The second reason for separate estimates is that production of food for domestic consumption is almost exclusively a rural phenomenon. The determinants of food purchases could, therefore, differ between urban and rural areas, at least for those foods produced by many households in the Northeast. The results presented here refer only to food purchased, not to total consumption including domestic production. Since the dependent variable is a logarithm, only households that consumed the food in question are considered.
The sample sizes, R2 statistics, and parameter estimates are presented in table 1. The mean price for each foodstuff for those families purchasing it is also shown; these prices are in August 1974 cruzeiros (at that time the US dollar was valued at approximately Cr$6.90) and enter the variable 1/q, for calculation of the income and price elasticities at mean paces.
TABLE 1. Consumption functions for 16 basic foodstuffs in metropolitan and rural areas, north-east Brazil: estimated parameters and descriptive statistics Parameter estimates (standard errors in parentheses)
|No. of||Mean price||Parameter estimates (standard errors in pharentheses)|
|Foodstuff and location||observations||R2 statistic||(Cr$)||B0 x 10-3||B1||B2||B3||B4|
|metropolitan||1,504||0.19||26.7||13.4||(0.40)||-0.34||(0.06)||-0.39||(0 09)||0.85||(0 07)||-13.10||(5.74)|
|rural||5,133||0.31||19.3||10.0||(0.25)||0.37||(0 03)||-0.91||(0.06)||0.45||(0 03)||-9.16||(1.65)|
|Beef (bone in)|
|rural||212||0.06||112.9||4.7||(2 43)||0.03||(0.25)*||0.70||(0.55)*||0.45||(0.27)||-4,43||(3 89)*|
*Not different from zero with 95% confidence. Source: Original calculations, from disaggregated ENDEF data for region V (North-east).
As is evident from equations 4-7 in the appendix, consumption responses to changes in the explanatory variables cannot be determined from the parameter estimates alone. These estimates, nonetheless, deserve a few comments. First, there are some significant urban/rural differences. Income responses, as indicated by B1 are nearly always higher in rural areas. Family-size responses, indicated by B3, in contrast, are typically larger in big cities. Consumption does not change the same way everywhere when income per person changes; it depends on whether the change arises from income or from family size.
Second, cornmeal and manioc flour, two sources of calories, are considered inferior goods, except perhaps among the very poor. In metropolitan areas, dried beef and beans are also considered inferior. These results are consistent with the finding from the ENDEF [10, 11] that malnutrition is more of an urban than a rural problem at a given income level and that in urban areas more additional income is spent on improving the quality of the diet than on increasing the quantity eaten.
Finally, the term in 1/q is usually significant, at least in large cities; it is more often indistinguishable from zero in rural areas. With a few exceptions, the parameter estimates have the expected signs-positive for B3, negative for B2 and B4, B1 can be negative for inferior foods. For normal foods, expenditure elasticities start out higher than B1 and decline as income rises, while price elasticities start out below B2 and rise. The only significantly incorrect signs for a parameter are the inexplicable positive price effects on pasteurized milk and soybean oil consumption in metropolitan areas.
Since consumption elasticities (or discrete changes) depend on q, they must be calculated for particular values of expenditure X and family size N. The combinations of expenditure and family size should correspond to poverty, because it is the behaviour of poor households that matters for combating malnutrition. So far as family size is concerned, we consider values of N- = 5, 6, 7, and 8 members when calculating changes in consumption according to equation 7 in the appendix; the reference family consists of N= 4 members. The middle value in this range, six members, is close to the mean of the ENDEF sample, so it is used when choosing levels of expenditure or poverty lines.
TABLE 2. Poverty lines per person and for a family of six (income, in August 1974 cruzeiros, needed to supply 2,000 calories per person per day on a diet of only rice and beans)-north-east Brazil, 1974-1975
|% of income spent on food||
Recife and Salvador
Sources: Ref. 4, tables 15 and 16; and original calculations from disaggregated ENDEF data.
The commonest method of classifying families as poor or not in Brazil is by reference to the legal minimum wage, sometimes in terms of income per person, but often just in terms of total household income, irrespective of size. The legal minimum is supposed to bear a relation to the cost of an adequate diet for a family of four, but in practice it often has not done so-the real value of a minimum wage varies considerably among regions of Brazil and has fluctuated greatly through time. For her estimates of consumption elasticities, Gray  used relative poverty lines, defined either by a family's position in the social distribution of income or by the relative satisfaction of calorie requirements. Poverty was defined to correspond to the 15th or 30th percentile of these distributions, separately for urban and for rural households.
Contents - Previous - Next