The present paper investigates the nutrition demand pattern for rural households in India. The non-parametric approach of quantile regression is applied to characterize the entire distribution of calorie consumption. This technique has an advantage over the traditional ordinary least square technique. It relaxes the assumption of a constant effect of the explanatory variables over the entire distribution of the dependent variable. These effects are allowed to vary over the entire distribution of dependent variable i.e., in this case the distribution of calorie consumption. The results show that indeed, the responsiveness of calorie consumption to various factors differs across different levels of calorie consumption. A comparison of the quantile regression results with OLS results suggests conclusions and policy suggestions based on OLS results are unlikely to be ideal. Some further light is also shed on the debate on calorie income elasticity as the magnitude is observed to be different for the under-nourished and the over-nourished households.
ASARC Working Paper 2005-2, 55 pp.