Analysis of household demand patterns using household data : re-thinking the use of unit values or community prices

Analysis of household food consumption patterns and welfare requires knowledge of household demand responses to changes in price and income. Estimation of the price and expenditure elasticities requires detailed data on household purchases and prices, which are often not available in many developing countries. To overcome constraints on the availability of price data, two approaches are mostly used: community prices and unit value (obtained by dividing household expenditure by quantity purchased). However, prices from these approaches are most likely measured with error. Also, they fail to account for quality variation in consumer demand. If these limitations persist, then price and expenditure elasticities estimated using these data are most likely to be biased, with negative implications for policy-making. In this paper, I assess the differences in price and expenditure elasticities for food between the unit value and community price approaches using household data from Tanzania. Estimations are done using the Quadratic Almost Ideal Demand System augmented to deal with issues of censored expenditure shares. My analysis also offers evidence on the food demand patterns in the country. Findings from the paper reveal no significant differences in the expenditure elasticities obtained from using unit value and community prices. The estimates from both approaches, however, indicate substantial discrepancies of price elasticities. These deviations mostly might have resulted from community prices if enumerators did not consider the probability of consumers to bargain. I suggest that before using household data, researchers should carefully account for quality variations and measurement errors when they derive price elasticities.