Semiparametric Estimation of Consumer Demand Systems with Micro Data

This article proposes a semiparametric two-step procedure for estimating a censored consumer demand system with micro data. The semiparametric estimator considered in the first step is suggested by Klein and Spady (1993). This estimator, used as a counterpart of the probit estimator in a conventional two-step model, does not make any distributional assumptions about the disturbances and so is exempt from model misspecification and plausible heteroscedasticity. In the second step, we motivate the choice of the Almost Ideal Demand System (AIDS) as an economic representation of consumers' demand behavior. Implementing our proposed semiparametric two-step procedure as well as Shonkwiler and Yen (1999)'s two-step model to a household meat consumption dataset from China generates the price and expenditure elasticities of demand. We also conducted the Horrowitz and H01rdle (1994)'s specification test to our data and reject the null.