The rank and model specification of demand systems: an empirical analysis using United States microdata

A rank-three demand system is estimated with United States Consumer Expenditure Survey microdata. A unique price data set is also used, which permits the analysis of effects of systematic errors in price variables. It is found that errors in price variables bias test results for the rank-three hypothesis, in particular towards rejection. Other test results are affected to a lesser extent. Estimating smaller systems of demand equations, even when conditioning on excluded goods, yields significantly different results. Another important conclusion is that model specification is statistically significantly different for households of varying family sizes and housing tenure statuses.