Abstract
This paper investigates the relationship between household income and the share of food expenditure in South Africa using clustered household income and socio-economic data published in the National Dynamics Income Study. To achieve this objective and because of the clustered nature of the data, a multilevel linear model, in particular, the random intercept model, was employed as a suitable method of investigation. Fixed effect methods were also used to compare results. The results confirmed that the share of household food expenditure is inversely related to household income in a district. This finding is consistent with Engel's law, such that an increase in household income is related to a decrease in the proportion allocated to food items by a typical household within a district. It was found that the relationship between household income and food expenditure does not vary across districts. The finding also pointed out that the district context matters because the district's average household income significantly explains the share of food expenditure at the household level.