In this project, we study expenditure patterns among poor households in rural Colombia with three main goals. First and foremost, we want to characterize the demand patterns for different commodities for very poor households. The extreme poverty of the households in our sample makes some facts taken for granted among other populations, for instance that the food income elasticity is less than one, questionable. Moreover, when looking at detailed systems, such as components of food, one can quantify the income (and price) elasticity of different food components. Such elasticities are relevant for the design of policies aimed at improving the nutritional status of children and other poor and vulnerable individuals. Second, as the data was collected for the evaluation of a welfare program, one can assess the extent to which, given the effect on total expenditure predicted by the quasi-experimental variation in our sample, our model can predict the changes in expenditure patterns observed in the data. Specific inadequacy of the demand system we estimate in predicting how the structure of consumption changes with the policy interventions might be suggestive of the channels through which the policy operates and of richer behavioural models that can be fitted to the data. Finally, by addressing the various methodological problems we will be dealing with and exploring different modelling choices, the paper gives a methodological contribution to the study of demand patterns. In particular, we will be addressing the issue of the functional form for the demand system, the appropriate instrumenting of total expenditure, how to control for price differences across clusters when prices are not observed, and the importance of measurement error.

Sample: Approximately 11,500 very poor households in rural Colombia.