Contents - Previous - Next

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


The aim of this study was to obtain more insight into the effects of household food insecurity on individual food consumption. This was done using both qualitative and quantitative indicators [1]. Since food insecurity may not affect all household members equally, an intra-household analysis was carried out, making a quantitative and qualitative comparison between the food consumption of husbands, wives, and their 2-5-year-old children. Although the number of subjects was small, which limits any generalizations from the study, our results add important information to a topic on which data are lacking.

Caution also should be used in interpreting these data because of methodological difficulties with food-consumption studies, especially in developing countries. Bias occurs when subjects are forced to serve themselves separately, when normally they share one bowl with other family members. Subjects may change their usual dietary pattern when an observer is present. Food-composition tables contribute to errors in calculations, especially with respect to indigenous foods for which composition data are few. Observation of more than one subject at the same time, as is the case in the intra-household food-distribution survey, carries additional difficulties. Children who are used to eating from one bowl with their mother have to eat separately to be able to weigh their consumption. Children who are not participating in the study may want to eat with the participants.

The total energy intake for the study of women's food consumption was higher than that of the women measured in the intra-household food-distribution study (2,605 and 1,854 kcal respectively in April 1991; 2,633 and 1,755 kcal in August). However, the former study was concerned with total daily food intake, while the latter considered only the intake from meals eaten at home. Snacks and food eaten else-where-including beer, fruits, roasted groundnuts, and bambara groundnuts were not recorded and thus not taken into account during the intra-household study. It seems probable that 30% of the food contributing to energy intake was missed, since they were eaten outside the home. The survey on women in April 1991 showed beer contributing 10% of total energy intake, and fruits and vegetables contributing 6% more and pulses 2% more than for the women in the intra-household study. Since the activity patterns of men and women are different in that men spend about one and a half to two times as much time on agricultural work in the fields away from home or on travel, men seem to have more possibilities to eat outside the home. Especially in August, when they participate in work parties, where local beer is served, men are likely to get even more than 30% of their total energy intake outside the home. For children 2-5 years old this seems less likely, since they are not yet able to gather foods in the bush and are mostly carried around by their mothers.

However, the interest of this study was mainly the relative contributions of food groups and food sources, and not absolute levels of intake. Therefore, conclusions are drawn cautiously and do not regard absolute intake levels, which are indicated for comparisons between groups only.

Assumptions with regard to adequate intakes and the requirement levels of individuals were made in the intro-household food-distribution study. Energy requirements for adults were estimated with regard to the work load for each season [24]. A study of the physical activity level in the same study population of Beninese women found an average of 1.63 x resting metabolic rate (RMR) for intermediate seasons and 1.77 x RMR for pre-harvest seasons [18]. These values hardly differ from the values given by the WHO [24] for moderate and heavy workloads respectively. Since no data were available on the requirements of Beninese men, the WHO values were used to estimate requirement levels for both men and women in this study. As indicated in a study on the distribution of protein and energy intakes in New Guinean households, the use at the individual level of energy requirements and safe levels of protein intake meant for groups demands careful interpretation [25]. Less attention should be paid to the degree of inequality than to its direction. Protein requirement fulfilments were over 100% and even up to 186% for children in the pro-harvest season of 1991. Although high, they were in line with other reports-for example, 135%-158% [4] and 125%-195% [15] for protein requirement fulfilments of adults and children.

Dietary changes due to seasonal food insecurity

In periods of food insecurity a shift to less preferred foods is expected. The dominant contribution of cereals to energy and protein intake as reported in this study also was mentioned by Rosetta [4]. But when cereal stocks became depleted, the first change in consumption patterns was a shift to the consumption of pulses. For children this change was even more drastic than for adults: they received an additional 11% of energy from pulses. Adults, however, during the pro-harvest season consumed more tubers, which provided 6% of their energy intake, whereas children obtained only 2% of their energy from tubers. An increase in the consumption of pulses coincided with an increase in the consumption of fats. Beans are preferably eaten with shea butter.

In periods of food stress, a change may take place not only in food groups but also in food sources. In this study, gifts and purchased foods constituted a larger part of energy and protein intake in the pre-harvest season when the household's own cereal stocks were running low. Wild foods contributed a little more to energy intake during the pro-harvest seasons than in the other seasons, but their contribution to protein was negligible. Gathered foods are mostly leafy vegetables, fruits, and shea nuts. A comparison of energy intake from wild foods between the post-harvest and the lean seasons in a rain-forest community in Ghana found that a decrease in consumption of farm products from 80% to 65% in the lean season corresponded to an increase in consumption of bush foods by 10% and gifts by 4%, and that consumption of purchased foods did not increase [7]. In Kenya, a decrease in the availability of own production was mainly compensated for by the purchase of foods and by gifts [8].

Intra-household food distribution

The level of fulfilment of the energy requirements of the parents decreased and that of the children increased between seasons, although not significantly. Interpretation of these data is complicated by the facts that food sources differ between seasons and that it is not known how much is eaten outside the household in each season. However, the fulfilment of the requirements of the children in August, the hungry season, was significantly larger than that of the men. The growth rate in Beninese children was higher in the intermediate season and lowest in the pro-harvest period [18]. Unfortunately, growth rates in these two seasons were not compared statistically. Infectious diseases also play a role in children's growth rate [26]. The prevalence of some diseases such as malaria and gastroenteritis is highest in the rainy pro-harvest season and may counteract the effect of a privileged food distribution.

No indications were found for a difference among men, women, and children in the contribution of different food groups to energy intake. The men were not privileged in meat or fish consumption in this study, in contrast with what has been reported in same Asian studies [11, 27]. The children consumed slightly more pulses and less tubers than their parents during the pro-harvest season.

In summary, the consequences of seasonal food insecurity (i.e., shortage of cereals) for the consumption patterns of individuals in this study were in higher consumption of pulses and tubers and greater contributions of food gifts, purchased foods, and wild foods to total intake in the pro-harvest season. These dietary changes correspond to coping behaviour at the household level, as determined in another study in this population. The coping behaviour of north Beninese households with seasonal food insecurity included cultivation of early crops such as yarns, sweet potatoes, and early bean varieties; gathering of wild foods such as shea nuts, leafy vegetables, and fruits; reliance on social networks; and seasonal migration or other income-generating activities, such as handicrafts and sale of processed foods.

This study did not provide clear evidence for sex differences in food allocation between and within seasons. The food distribution in our study families seemed to favour children over their parents during the hungry season.


The financial support of the STD programme of the European Community, contract nos. TS2-0150-NL and TS3*-CT91-0026, is gratefully acknowledged. The cooperation of the Department of Nutrition and Food Sciences, Université National du Bénin, is highly appreciated. In particular we thank Dr. M. C. Nago, Vice-Dean of the Faculty of Agricultural Sciences, and Dr. F. L. H. A. de Koning for their support.


1. Maxwell S. Frankenberger TR. Household food security: concepts, indicators, measurements-a technical review. New York: United Nations Children's Fund, and Rome: International Fund for Agricultural Development, 1992.

2. Corbett J. Famine and household coping strategies. World Development 1988;16(9):1099112.

3. Asibey EDA. Wildlife as a source of protein in Africa south of the Sahara. Biological Conservation 1974;6(1): 32-9.

4. Rosetta L. Sex differences in seasonal variations of the nutritional status of Serere adults in Senegal. Ecol Food Nutr 1986;18:231-44.

5. De Garine I, Koppert G. Coping with seasonal fluctuations in food supply among savanna populations: the Massa and Mussey of Chad and Cameroon. In: De Garine I, Harrison GA, eds. Coping with uncertainty in food supply. Oxford: Clarendon Press, 1988:210-59.

6. Pagézy H. Coping with uncertainty in food supply among the Oto and the Twa living in equatorial flooded forest near Lake Tumba, Zaire. In: De Garine I, Harrison GA, eds. Coping with uncertainty in food supply. Oxford: Clarendon Press, 1988:175-209.

7. Del GJS. Hunting and gathering in a Ghanaian rain forest community. Ecol Food Nutr 1989;22:22543.

8. Neumann C, Trostle R. Baksh M, Ngare D, Bwibo N. Household response to the impact of drought in Kenya. Food Nutr Bull 1989;11(2):21-33.

9. Wandel M. Household food consumption and seasonal variation in food availability in Sri Lanka. Ecol Food Nutr 1989;22:169-82.

10. Zinyama LM, Matiza T. Campbell DJ. The use of wild foods during periods of food shortages in rural Zimbabwe. Ecol Food Nutr 1990;24:251-65.

11. Gittlesohn J. Opening the box: intrahousehold food allocation in rural Nepal. Soc Sci Med 1991;33(10):114154.

12. Kaiser LL, Dewey KG. Household economic strategies, food resource allocation and intro-household patterns of dietary intake in rural Mexico. Ecol Food Nutr 1991 ;25:12345.

13. Gopaldas T. Saxena K, Gupta A. Intra-familial distribution of nutrients in a deep forest-dwelling tribe of Gujarat, India. Ecol Food Nutr 1983;13:69-73.

14. Hassan N. Ahmad K. Intra-familial distribution of food in rural Bangladesh. Food Nutr Bull 1984;6(4):34-42.

15. Van Steenbergen WM, Kusin JA, Nordbeck HJ, Jansen AA. Food consumption of different household members in Machakos, Kenya. Ecol Food Nutr 1984;14:1-9.

16. Leonard WR. Household-level strategies of protecting children from seasonal food scarcity. Soc Sci Med 1991;33(10):1127-33.

17. Abdullah M, Wheeler E. Seasonal variations and the intro-household distribution of food in a Bangladeshi village. Am J Clin Nutr 1985;41:1305-13.

18. Ategbo EA. Seasonality in food availability and adaptation of individuals. PhD dissertation, Wageningen Agricultural University, Wageningen, Netherlands, 1993.

19. Cameron ME, van Staveren WA. Manual of methodology for food consumption studies. London: Oxford University Press, 1988.

20. Woof Tsuen WL. Table de composition des aliments à l'usage de l'Afrique. 1st ed. Rome: FAO/WHO, 1968.

21. Platt BS. Tables of representative values of foods commonly used in tropical countries. London: Her Majesty's Special Office, 1979.

22. Norusis MJ. The SPSS guide to data analysis for SPSS/ PC+. Chicago: SPSS, 1988.

23. Siegel S. Castellan NJ. Nonparametric statistics for the behavioral sciences. 2nd ed. Singapore: McGraw-Hill, 1988.

24. World Health Organization. Energy and protein requirements. Geneva: WHO, 1985.

25. Ferro-Luzzi A, Norgan NG, Paci C. An evaluation of the distribution of protein and energy intakes in same New Guinean households. Nutr Rep lot 1981;24(1): 153-63.

26. Rowland MGM, Cole TJ, Whitebead RG. A quantitative study into the role of infection in determining nutritional status in Gambian village children. Br J Nutr 1977;37:441-50.

27. Senauer B. Garcia M, Jacinto E. Determinants of the intrahousehold allocation of food in the rural Philippines. Am J Agric Econ 1988;70(1):170-80.


Food policy

Income and food-consumption behaviour in China: A structural - shift analysis

Haijiang Ma and Barry M. Popkin


The nutrition transition in China has proceeded to the extent that the food -consumption behaviour of low-income and high-income groups is different. Failure to consider these differences could lead to inappropriate assumptions about some basic food policy issues. This analysis was undertaken using a sample of adults 1850 years old from the 1991 China Health and Nutrition Survey. Low-income families have a greater propensity to increase or decrease fat and calorie intakes than high-income families. This difference has important, policy implications and shows why it should be considered in making decisions regarding consumption behaviour. Selecting the appropriate income switching or cut-off point is central to capturing this structural difference.



A substantial body of literature attempts to explain patterns of food consumption or nutrient intake with respect to income. These studies frequently involved regression analyses that hypothesize that food consumption or nutrient intake is "produced" as a function of income combined with other structural variables such as personal characteristics (age, sex, smoking) and socio-economic characteristics (urban or rural residence, occupation) [1-5]. The usual approach is to pool data from the whole sample and run a single regression. In this case, the model is the same for all income levels, and it is assumed that the relationships among income, prices, and other explanatory factors and food intake outcomes is the same for all income levels.

If the behaviour of higher- and lower-income groups is different, traditional statistical estimates of an income and diet relationship would be biased [6]. More important, the effect of the key relationships for the target population of low- or high-income groups might be incorrect. The behaviour of the two groups probably will vary in countries such as China, where food markets are not constrained, dietary adequacy has been achieved for a significant segment of the population, and disposable income allows higher-income groups some flexibility in their consumption patterns.

This literature does not necessarily assume that the income and consumption relationship is linear, but it generally assumes that there are not shifts in overall decision-making processes. Few researchers assume that every increment of increased income is associated with the same amount or proportionate consumption change. This literature often explores complex non-linear income and consumption relationships. A wide range of functional forms for the independent income and dependent consumption outcomes are used. The literature on this topic is vast. What is rarely examined is the way the overall set of parameters changes with income [7].

After reviewing the literature, however, one finds that little attention has been devoted to estimating regression models in which the variables are not necessarily assumed to have positive or negative values in perpetuity, and to the exploration of the specific form of the relationship between income and food consumption. In Brazil there is strong reason to question this assumption [4]. In a national analysis of the income and body mass index (BMI) relationship among the 30% poorest Brazilian families, increases of income were significantly and positively related to increases of women's BMI (r=.11, p<.001). The same applied to the next 40% of families (r=.06, p < .001) but not to the richest 30% of families, where income and women's BMI wore significantly and negatively related (r=.05, p < .001).

Little else is done to examine this income and nutrition status relationship in low-income countries. The consideration for a shift in between them appears to be reasonable when we analyse the food-consumption patterns in many developing countries where income is highly constrained for lower-income households compared with higher-income ones.

Economic theory posits a positive relationship. Essentially, the proportion of income expended on food is reduced as income increases. Moreover, the absolute level of food expenditures increases as income increases. In addition, several other associations between income and food consumption have been observed.

With increasing income, the proportion of energy in the diet &am various sources changes in the following ways:

» Unseparated animal fat and animal protein increase.
» Unseparated vegetable fat and vegetable protein decrease.
» Carbohydrate decreases.
» Sugar and separated edible fat increase.

This well-known relationship is based mainly on cross-country comparisons but is buttressed with a number of within-country time-series studies [8]. The responsiveness of total dietary energy, total and saturated fat, and other macro- and micro-nutrients to income change depends on the nature of the demand for particular foods as well as overall eating patterns. For example, in China, pork consumption is highly responsive to increases in income, which results in large increases in the proportion of energy from fat [9]. As many have shown, food demand is much more price- and income-elastic among the poor than among higher-income groups [10,11].

In contrast, when income increases are spent on more elaborate packaging and processing or on higher quality of specific foods rather than on larger quantities of food or shifts in types of foods, changes in income have little effect on dietary structure. In other words, beyond the point at which total food energy needs are met, people spend more per food item, partly to obtain higher quality [12]. But they also combine these changes in quality with increased consumption of higher-priced goods that have undergone more processing [13]. In other words, they purchase food items that take much less time to prepare. The net effect does not have to be an increase in total energy intake. It might mean the reverse under many circumstances.

Over the last 15 years China has achieved remarkable economic progress. From 1979 to 1987, income per capita quadrupled in rural areas and tripled in urban ones [14]. Accompanying these changes was a rapid transition in dietary patterns, so that now undernutrition and overnutrition coexist [15].

It would be reasonable to assume that food-consumption behaviour may differ in households with a low income and those with a high income. The propensity (mainly in the economic sense) to choose foods high in fat, protein, and calories is different for people with low and high incomes because they face different price and budget constraints. In addition, as is shown in descriptive statistics, high- and low-income populations engage in different physical activities at both work and home. A low-income family has little chance to increase fat and caloric intake because of limited affordability and availability. But as income improves, the propensity to increase these intakes might well be expected to be large. In contrast, a high-income family has much more leeway for discretionary consumption, depending on its own standards [12].

Dissimilar income has different effects on the structure of the function in terms of the direction and magnitude of the effects of the explanatory variables. This makes usual (structural) methodology difficult, if not impossible, to apply. Failure to account for this possibility may bias our perception of the contribution income makes to dietary intake. Appropriate estimation methods for explaining food consumption must be developed.

Food-consumption behaviour is very difficult to describe and predict. In our study, we observed a mixture of old and new habits growing from expectations or preconceptions about a "good" meal and healthful setting. The reactions of income are varied. Trends associated with income include increases in animal product and total fat intake, the quality of the foods purchased, and the proportion of processed and away-from-home consumption. The net effects on nutrient intake levels depend on a wide range of factors. Apart &am income, demographic characteristics such as gender, ethnicity, education, and employment status, and other factors such as residence, market price, women's participation in the labour force, and food policy can also be important because of the way they relate to consumption or concurrently coaffect consumption with income.


Materials and methods

Survey design

The China Health and Nutrition Survey was designed to provide comprehensive information on individual nutrition status, individual and family economic status, and community conditions. It included food-consumption data, physical examinations, and household and community characteristics. The sample frame covered eight provinces with varied geography, economic development, public resources, and health indicators. In each of the provinces, a multi-stage, random cluster process was used to draw a sample consisting of two cities (usually the provincial capital and a less-developed city) and four counties stratified by low, middle, and high income. The total sample contained 3,780 households and about 16,000 individuals. The detailed description of the sample design is presented elsewhere [16].

The data used for the present analysis came from the second round of the survey (1991). Three waves of data were collected (1989, 1991, 1993), but the most recent is unavailable for analysis. For the purpose of reducing variability caused by age, we selected individuals 18-50 years old, giving a subsample of 5,975 individuals. After the elimination of those for whom some data were missing, an analysis sample of 5,932 persons was available.

Dependent variables and explanatory variables

The dependent and explanatory variables included in this analysis are shown in table 1. Three nutrients were selected as dependent variables for their importance to health and their ability to reflect consumption behaviour: fat intake, caloric intake, and percentage of calories from fat (% fat). Fat intake and % fat were chosen because of their potential contribution to obesity, diabetes, cancer, and cardiovascular risk and because their intakes are indicative of consumption of specific, more preferred food groups such as animal products. For each individual, control variables of age, sex, education, smoking, intensity of labour activity, and region of residence were selected as independent variables because they are linked directly or indirectly to income and nutrient consumption. For example, findings in China [16] and the United States [17] suggest that people with high incomes usually drink more alcohol than those with low incomes. In China, drinking alcohol not only contributes directly to caloric intake but also often accompanies meat intake and thus increased fat intake. People living in northern and southern areas have different food habits, including alcohol consumption.

Higher-income persons in this study consumed less energy but had a much higher fat diet. The lower-income sample consumed over 3.5 percentile points less energy from fat than the higher-income sample, but also 176 more kilocalories of energy. The two groups were remarkably different in education, area of residence, and patterns of physical activity.

Ethnicity is presumed to be associated with food consumption. In our study sample, almost 90% were Han; this was not a significant factor. Control variables were height, weight, number of children under the age of seven years, and job type, all of which were dropped because of lack of statistical significance. The variables in the final specification were all significant at the 10% level or better.

TABLE 1. Means of variables used in the analysis

Variable Total (N= 5,932) Low income (N=5,127) High income (N = 805)
Dependent variables
energy (kcal)
fat (g)
calories from fat (%)
Independent variables
age (yrs)
sex (male = 1, female = 0)
education (yrs)
smoking (yes = 1, no = 0)
alcohol use (yes = 1, no = 0)
family size
dwelling size (urban = 1, rural = 0)
locale (northern = 1, southern = 0)
moderate physical activity
heavy physical activity
household income (yuan/yr)

Source: China Health and Nutrition Survey, 1991.

Dietary data

Individual dietary intake data were collected for three consecutive days, randomly allocated from Monday to Sunday. Each individual was interviewed daily to report all food eaten at home and away from home on a recall basis.

In addition, weighing and inventory techniques were employed for collecting household food-consumption data for the same three days. All food in the house-including raw food, processed food, and edible oils and salt-was weighed and recorded at the beginning of the survey. All food entering the house and all food discarded was weighed and recorded during the three days. All remaining food was again weighed and recorded at the end of the last day.

Collecting individual and household dietary data allowed the immediate calibration in the field of nutrient intake calculated from two data sources. The type and amount of food consumed by any household or individual could be cross-checked for significant discrepancies between the two sources, and those respondents could be revisited for correction. (Because all the calibration was done in the field by the investigator, we do not have the number of individuals or households whose food data were calibrated.)

The nutrient intakes used in this analysis were calculated from the individual data of the three consecutive days. We averaged the intake, which could reduce intra-individual variability and more closely measure "usual intake" [18, 19]. Almost the entire sample had a full set of dietary information. It was necessary to use intake data from one or two days because of missing data for only a very small sample.

TABLE 2. Nutrient and food consumption patterns (grams per capita per day) by income tertile-adults 20 years old or older, 1989

Food group Income fertile
(N = 1,972)
(N = 1,989)
(N = 1,925)
Rice and its products 331.2 341.0 324.6
Wheat flour and its products 207.2 177.0 154.0
Other cereals 107.0 37.5 26.9
Cookies, cakes 1.5 5.1 6.4
Starchy tubers 137.0 59.0 395
Red meats 32.35 3.7 64.2
Organ meats 3.7 4.7 7.8
Poultry 4.3 6.4 7.7
Milk and milk products 0.5 0.6 6.6
Eggs 6.3 10.3 16.4
Aquatic products 11.3 25.5 38.4
Vegetable oil 12.8 16.4 18.9
Energy (kcal) 2,801.5 2,6065 2,567.8
Fat 45.5 54.4 62.8
Protein 80.6 77.8 79.0
Energy from fat (%) 14.8 18.7 21.8

Income data

Of the explanatory variables, the most important for this analysis is income data. Income could be produced by cash and non-cash, market and non-market activities. Household income was estimated in detail by including all cash and non-cash income components (food and other subsidies, commodity coupons, private farm crops).

Income and consumption relationship

Table 2 presents the relationship among income, nutrient outcomes, and several foods for this adult sample in 1989. The sample is stratified into income tertiles to provide some sense of the shift in diet with income changes.

Statistical methods

The relationship between nutrient intake and income and other explanatory variables was estimated in three multiple regression models.

The first model used for the description hypothesized linear relationship without any interaction between income and all other variables [20]:

Yt -It x B1t + Xt x B2t + Et, (1)

where Y is the dependent variable of the tth observation, B1 is a coefficient for the variable of the income, I is income, B2 is a vector of unknown coefficients, X is a vector of other explanatory variables, and E is an error term, which is a distributed random variable with (0,s 2).

The second model hypothesized a similar linear relationship but with interaction between the income and all other explanatory variables:

Yt = It x B1t + Xt x B2t + It x Xt x B3t + Et, (2)

The third model was designed to explain structural difference [17, 21]. Nutrient intakes are governed by different sets of parameters, and the set that governs a particular nutrient intake is determined by the level of the income. By specifying only two sets of parameters, it is assumed that two equations describe two different behavioural regimens in food consumption. These regimens will be referred to as high level of income and low level of income. The latter consists of those whose income is less than 1' and the former is 1' or more. The following regression equations are specified:

Y1t =Xa + et1 if 0 < I < I*, (3)

Yt2 = Xß + et2 if I > I*, (4)

where Yt1 and Yt2 are nutrient intake for regimen 1 and regimen 2 respectively, I is the income that defines the regimen, and I* is an unknown switching point that could be obtained by means of special technique (e.g., switching regression technique).

Ideally, we would have used switching regression software to select the appropriate income cut-offs. Such software was not readily available, so we tested a range of arbitrary income cut-offs as part of this work. It is important to note that an alternative approach is to explore the non-linear effects of income in model 1. This would produce a better fit for the income effect but would not consider the central theme of this paper, that the effects of the other covariates or independent variables also change as income changes.

These three models allow us to test systematically for different relationships between income and other socio-economic variables and food consumption.

Selecting an income cut-off

The first step is to determine the appropriate switching point or cut-off for income. To search for the switching point at which income breaks the sample into two distinct groups in nutrient intakes, we first split the sample at 50%, 75%, 80%, and 85%; these percentages correspond to incomes of about Y4,000, Y5,500, Y6,000, and Y7,000 respectively (Y=yuan renminbi). Then we ran two separate regressions for each subsample in the two income groups.

To check whether a structural difference exists between the two groups, the Chow statistic was calculated [6]. The Chow statistic tests whether the null hypothesis that coefficients for two regimens" regression are identical, a = ß, is accepted. If the null hypothesis is accepted, the two equations yield identical coefficients, and there is no structural difference between the two regression equations. Consumption behaviours are the same in the two regimens; otherwise, the two sets of coefficients differ, and it is accepted statistically that consumption behaviours differ for dissimilar levels of income. If the Chow tests are significant, the null hypothesis is accepted, and it will be determined that a structural difference is present in the determinants of food consumption behaviour for different income levels. In this case, it will be essential to consider this relationship if one wants to examine income and consumption relationships accurately. Ordinary-least-squares statistical methods were used for all the equations.

Contents - Previous - Next