Contents - Previous - Next

This is the old United Nations University website. Visit the new site at http://unu.edu

Continue - 1

TABLE 5. Estimates of three methods modelling energy intake (kcal)

Numbers in parentheses are t statistics

* Significant at the 5% level.
** Significant at the 1% level.

However, the high-income regimen results are substantially different from the one-regimen and the low-income regimen. This result indicates that the nutrient consumption of 15% of the individuals would be misinterpreted or obscured if all individuals from different income levels were mixed.

Large differences in statistical significance occurred. Of the eleven explanatory variables, only three are significantly associated with fat and caloric intake for both low- and high-income groups: education, family size, and northern or southern residence for fat intake; and gender, urban or rural residence, and heavy physical activity for caloric intake. Age, gender, smoking, alcohol use, residence, physical activity, and income are significantly associated with the fat intake of the low-income regimen but not of the high-income group. Age, education, alcohol use, family size, northern or southern residence, and moderate physical activity are significantly associated with the calories of the low-income group but not the high-income group. Of them, the effect of income on the fat intake of the low-income group is significantly different from that on the high-income group.

Income significantly increases the absolute fat intake and percentage of calories from fat of the low-income group, but for the high-income group it has little effect on absolute fat intake and tends to decrease percentage of calories from fat. This is consistent with the previous results. Most of the explanatory variables are significantly associated with fat and caloric intakes of the low-income regimen but are not significantly associated with that of the high-income regimen. This lack of explanatory effect in the high-income group does not appear to result from a lack of sample size, because the group had 15% of the study sample, 805 individuals. It may suggest that a set of unmeasured variables affects the consumption behaviour of the high-income group.

The factors affecting nutrient consumption are different for people at different income levels. Interpretation of consumption behaviour should capture different factors, and promotion of healthful eating behaviours should consider and focus on different factors, based on income level.

Of equal or greater importance, the effect of many key factors is very different for low- and high-income households and also between the one-regimen and two-regimen models. Use of the one-regimen model could lead to misleading policy decisions. Table 9 presents the simulated effects of some variables on the fat and caloric intake. The differences in the effects of variables between the two groups indicate differences in the factors affecting their nutrient-consumption behaviour. For example, among the low-income group, when household income increases by Y1,000, fat and % fat intake significantly increase 3.4 g and 1 percentage point respectively, and caloric intake increases 2.8 kcal, although this increase is statistically insignificant. Yet among the high-income group, fat and calorie intakes decrease 0.2 g and 0.3 kcal respectively, but the decreases are statistically insignificant. The calories consumed from fat significantly decrease 0.1 percentage point.

TABLE 6. Estimates of three methods modelling the percentage of calories from fat

Numbers in parentheses are t statistics.

*Significant at the 5% level.
**Significant at the 1% level.

TABLE 7. Summary of the significance and sign of regression coefficients

- or +, significant at the 10% level.
- - or + +, significant at the 5% level.
- - - or + + +, significant at the 1% level.

A blank space indicates no significance.

Continue

Contents - Previous - Next