Predicting within country household food expenditure variation using international cross-section estimates

Abstract

There is a long and distinguished literature involving demand analysis using international cross-section data. Such models are widely used for predicting national per capita consumption. However, there is nothing in this literature testing the performance of estimated models in predicting demands across the income spectrum within a single country. This paper fills the gap. We estimate an AIDADS model using cross-section international per capita data, and find that it does well in predicting food demand across the income distribution within Bangladesh. This suggests that there may be considerable value in using international cross-section analysis to study poverty and distributional impacts of policies.

GTAP-POV: A Framework for Assessing the National Poverty Impacts of Global Economic and Environmental Policies

“…AIDADS has been shown to perform well in out of sample forecasting of international food consumption patterns (Cranfield et al 2003). In addition, Verma et al (2011a) show that, in the case of Bangladesh, a version of AIDADS estimated using international cross-section data performs quite well in predicting food consumption behavior across the entire income spectrum. This latter finding is particularly important in the context of GTAP-POV, as we will use the demand system, estimated based on international cross-section data, to predict consumption shares at the poverty line in each of the poverty regions.…”