Contents - Previous - Next

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

Household distribution of energy intake and its relationship to socio-economic and anthropometric variables

Nazmul Hassan and Kamaluddin Ahmad
Institute of Nutrition and Food Science, University of Dhaka, Dhaka, Bangladesh


National averages conceal the actual state of consumption of energy-providing foods at disaggregated household levels. We undertook to study, by household, the distribution of energy intake and its adequacy or inadequacy in relationship to recommended allowances and to socio-economic and anthropometric variables.


The data for the present study were obtained from the nutrition survey of rural Bangladesh from 1981 to 1982, which was carried out in 12 statistically selected locations and involved 4,000 persons. Information on dietary intake from the selected households, obtained by trained investigators using the 24-hour food-weighing method and a prescribed pre-tested questionnaire, was converted into calories using a Bangladesh food value table [1] . To estimate per capita values, the calories of each food consumed by each household were totalled and divided by the total number of consumption units in the household as defined below. An individual present at all meals within the 24-hour period was considered a full-time consumer and counted as one unit. All children, including breast-fed infants, were included as consumers, valued in the same way as adult members of the family. Arithmetical adjustments were made for absent members and for visitors at family meals [2] .

The household energy requirement was calculated in kilocalories (kcal) using the 1973 FAD/WHO report on energy requirements [3]. Then the ratio of energy consumption (ECj) to energy requirement (ERj) was generated to find out the household adequacy of energy intake (AEj = ECj, where i = 1,2,3 ... n households). The basal metabolic rate was calculated using the linear regression equations of the 1985 FAO/WHO/UNU report on energy and protein requirements [4].

The anthropometric data on height and weight, also obtained by personal visits during 1981 and 1982, were analysed in terms of weight-for-height and height-for-age classifications of nutritional status. Since height does not change significantly beyond age 18, analysis of nutritional status on the height-for-age classification was performed only up to that age. Harvard values were used as the standard for the comparison of height data against age and of weight data against height [5] .

Information on various socio-economic variables was also obtained during the 1981-1982 survey from the same sampled households studied dietarily and examined anthropometrically. Only the relevant information on socioeconomic indices (land holding, income, and expenditure on food) were related with energy adequacy through linear and log-linear function.


The results are presented in three sections: dietary intake, interrelationships between energy intake and socioeconomic indices, and the relationship between anthropometry and energy adequacy.

Dietary Intake

Results of the dietary studies are presented in tables 1, 2, and 3. It can be seen from table 1 that about 48 per cent of the study households had intake below 1,800 kcal per capita per day; 22 per cent consumed 1,800 to 2,200 calories, another 22 per cent between 2,200 and 2,800 kcal, and 8 per cent had intake above 2,800 kcal per capita per day. Energy requirement, on the other hand, was calculated to be 1,800 kcal for 2 per cent of the households, between 1,800 and 2,200 kcal for 35 per cent, 2,200 to 2,800 kcal for 58 per cent, and 2,800 kcal or more for the remaining 5 per cent. These figures show the alarming energy situation of our rural households on the one hand and maldistribution of intake among households on the other. In other words, the vast majority of households had an energy intake that fell far short of their requirements, while quite a good number of them had an intake that was more than adequate.

TABLE 1. Per capita energy intake and average requirement by household

Energy intake requirement range (kcal) Intake Requirementa
No. of households % No. of households %
Below 1,600 197 33 3 1
1,600 - 1,800 89 15 7 1
1,800 - 2,000 71 12 51 8
2,000 - 2,200 58 10 161 27
2,200 - 2,400 62 10 173 29
2,400 - 2,600 39 6 130 22
2,600 - 2,800 36 6 41 7
2,800 - 3,000 17 3 23 4
3,000 and above 28 5 8 1
Total 597 100 597 100

a. Based on FAO/WHO/UNU recommendations [4].

TABLE 2. Adequacy of energy intake expressed as percentage of requirement by household

Adequacy of energy intake (%) No. of
Below 50 48 8
50 - 60 57 9
60 - 70 83 14
70 - 80 103 17
80 - 90 93 16
90 - 100 72 12
Over 100 141 24
Total 597 100

TABLE 3. Per capita observed and expected basal metabolic rate (BMR) by household

BMR range (kcal) Observed BMR Expected BMR
No. of
% No. of
Below 900 98 17 34 6
900 - 1,000 163 27 69 12
1,000 - 1,100 136 23 166 28
1,100 - 1,200 117 20 133 23
1,200 - 1,300 57 10 109 18
1,300 and above 20 3 80 13
Total 591 100 591 100

Table 2 shows the adequacy of energy intake expressed as a percentage of requirements by household. It can be seen that only about 24 per cent of the study households had an adequate calorie intake and the majority (76 per cent) fell short of the required level. Eight per cent of them could not even meet 50 per cent of their requirement.

Table 3 shows the per capita observed and expected basal metabolic rates (BMR) of our study population by household. The observed BMR was calculated on the basis of the existing weight and the expected BMR on the basis of the expected weight for existing height. There was a discrepancy between observed and expected BMRs. The majority of the households (67 per cent) had a per capita BMR of 1,100, although only 46 per cent were expected to have a BMR below this level. Similarly, 23 per cent of the households were expected to have BMRs between 1,100 and 1,200 kcal and 31 per cent 1,200 or above; the observed figure was 20 per cent for 1,1001,200, and 13 per cent for 1,200 or more. This suggests that the majority of our consumers were underweight for their height, and, as a result, a greater proportion of the households had a per capita BMR that was lower than expected.

Interrelationships between Energy Intake and Socio-economic Indices

Energy intake is usually influenced by a large number of factors, e.g. land holdings, income, expenditure on food, level of education, and tax category. In this study, attempts were made to examine some of these interrelationships rigorously through the use of the ordinary least-square multiple-regression technique. However, since formal education bears no direct or regular relationship to energy intake [6], this variable was dropped from the analysis. Similarly, tax category was left out because tax is levied on income, and households are classified on the basis of income; thus income is a proxy for tax category, and the amount of income spent on energy intake is reflected in the tax category. In the final analysis, the regression analysis related adequacy of energy intake with land holding, income, and expenditure on food.

It should be emphasized that the multiple-regression analysis employed in the study is not per se a technique for policy selection but rather a useful means of providing insight to facilitate and improve upon such selection. The fact that an independent variable in an equation is statistically most significant and has the highest coefficient is not, in and of itself, sufficient grounds for selecting that variable as an instrument of policy. In actuality, other instruments may prove more practical and feasible or less costly. The value of such analysis is that is provides an effective way to stratify society for purposes of an intervention policy, and provides important guidance toward rational selection of policy instruments

TABLE 4. Regression co-efficients (b) of per capita land holding, income, and expenditure on food and co-efficient of determinations (R2) for different models studied

Model no.

Constant (a)

b co-efficients

b1 b2 b3 R2
1 70.0869 0.2344 0.0991 0.0541 0.11
(0.0366)a (0.0209) (0.0360)
2 1.8221 0.2481 0.0841 0.0880 0.12
(0.0002) (0.0001) (0.0002)
3 34.3612 0.2318 0.1262 0.0158 0.10
(1.6503) (7.4627) (7.7304)
4 1.6770 0.2617 0.1346 0.0291 0.12
10.0081) (0.0366) (0.0379)

a. Figures in parentheses indicate standard error.

It is difficult to demonstrate that a single functional form can describe the logic and mechanism of relationships between energy adequacy and some indices of socioeconomic status; rather, several functions are competent for initial approximation of the true form. A suitable choice can be made from among such functions on the basis of theoretical implications, a good fit to data, and computation manageability.

To explore which type of functional relationship would best explain the changes in energy intake, the following four types of functions were tried:


2. (semi-log),

(semi-log), and

(double log), where y is the adequacy of energy intake; x1, x2, and x3 are per capita land holding, income, and expenditure on food respectively; and b is constant.

The regression co-efficient (b) of different independent variables, the standard errors of these co-efficients, and the co-efficient of determination (R2) for each of the above models employed are given in table 4. In all the models, 10 to 12 per cent of the variation in adequacy of energy intake is explained by the variables included. The co-efficient of per capita land holding (x1) is positive in all the models and is statistically significant (at 5 per cent level of significance). The co-efficients of per capita income (x2) and per capita expenditure on food (x3) are also positive, but they have become unreliable due to the presence of multi-colinearity between them. The correlation between them was observed to be 0.70.

This requires that the regression be run separately, excluding either of the variables (x1 or x2). (Full tables of regression co-efficients are available from the author.) The results of such an exercise indicated that when either x3 or x2 is excluded, models 2 and 4 can explain the variation in AK' caused by the variation in the included variables better than models 1 and 3. This can be seen better from the R2 of the models. The first set of models could explain about 12 per cent of the variation in AEj, while the second set takes note of only 9 to 11 per cent of the variation. Again the standard errors of the b co-efficients are smaller in the first set of models than in the second, implying that the b co-efficients are more reliable in the former set than in the latter, although the coefficients are positive and statistically significant in all cases.

Among the first set of models, again, the standard error of the b coefficients is lower in model 2 than in model 4 (R2 remains the same). Therefore one may conclude that the differences in the included variable are better explained by this model, according to which an increase of per capita land holding by one decimal increases the energy adequacy by 0.26, and an increase of 1 take in per capita income or expenditure on food increases the same by 0.14. The results of the regression analysis, therefore, seem to suggest that energy adequacy in rural Bangladesh is more significantly affected by land holding than by income or expenditure on food. Since the scope of increasing the per capita land holding is very limited and increased expenditure on food primarily depends on increased income, the levels of income of the rural population rather than land holding should be raised as a policy alternative. Studies in India similarly identified income as an important determinant of calorie intake, even though its contribution was lower (0.07) than that of other variables [7] .

Relationship between Anthropometry and Energy Adequency

This relationship was established between the anthropometric indices of height and weight of the population survey of 1981 and 1982, and their energy adequacy obtained from the data on dietary intake. The anthropometric data were analysed in terms of weight-for-height and height-for-age classifications of nutritional status; this part of the study was limited to households in which 50 per cent or more of their consumers were considered undernourished according to these criteria. They constituted 509 households as against 597, the total number of study households. This implies that, as a consequence of severe inadequacy of energy intake, the majority of the household members suffered from various degrees of malnutrition.

There was no direct or definite relationship between household energy adequacy and individual nutritional status according to the weight-for-height classification. Household adequacy did not necessarily ensure individual energy adequacy and hence better nutrition, because food was often distributed unequally within the family. Even in families having energy adequacy below 50 per cent of requirement, consumers were normal or marginally nourished. Similarly, in families having AEj 100 or more, consumers suffered from moderate to severe malnutrition. Ideally, there should be no well-nourished (healthy) consumer in families having AEj less than 50 and no malnourished consumer in families having AEj 100 or more. However, with the increase in adequacy of household energy, relatively more consumers were healthy or marginally nourished, and fewer of them suffered from moderate to severe malnutrition. But the trend toward reducing moderate to severe malnutrition with an increase in adequate energy was not very significant. A similar relationship could be noticed with respect to energy adequacy and classification of nutritional status by height-for-age. With the increase in energy adequacy, there was a tendency of persons with moderate and severe malnutrition to shift toward normal nutrition categories. Again, the shift was not very significant.


The study revealed an alarming energy situation in the rural population of Bangladesh. About 48 per cent of the study households had intake below 80 per cent of their requirement and 76 per cent of them lived below the poverty level. Maldistribution within and among the families seemed to aggravate the situation. As a consequence of chronic energy shortage, the majority of consumers became underweight for their height, and as a result a greater proportion of them had a per capita basal metabolic rate that was lower than expected.

Socio-economic factors such as land holding, income, and expenditure on food have a positive influence on energy adequacy. Regression analysis revealed a higher contribution by land holding than by income or expenditure on food. Since the possibility of increasing the per capita land holding is very limited and expenditure on food depends primarily on income, even a small increase in income would seem to be effective in increasing the energy adequacy.

No regular relationship between anthropometry and household energy adequacy was immediately seen, probably because household adequacy did not ensure better nutrition for individual consumers. Unequal distribution within the family left many of the members malnourished.


1. Institute of Nutrition and Food Science, Nutritive Values of Local Food Stuffs (University of Dhaka, Dhaka, Bangladesh, 1980).

2. Food and Agricultural Organization (FAO), Manual on House-hold Food Consumption Surveys, FAO Nutritional Studies, no. 18 (FAO, Rome, 1967).

3, FAO/World Health Organization (WHO), Energy and Protein Requirements, Technical Report Series, no. 522 (WHO, Geneva, 1973).

4 FAO/WHO/UNU, Energy and Protein Requirements, Technical Report Series, no. 724 (WHO, Geneva, 1985).

5. D. B. Jelliffe, The Assessment of Nutritional Status of the Community (WHO, Geneva, 1966).

6. K. Ahmad and N. Hassan, eds., Nutrition Survey of Rural Bangladesh 1981-82 (Institute of Nutrition and Food Science University of Dhaka, Dhaka, Bangladesh, 1983).

7. F. J. Morinda Levinson, An Economic Analysis of Malnutrition among Young Children in Rural India, Cornell-MIT Inter national Nutrition Policy Series (Cambridge, Mass., 1974).

Contents - Previous - Next