Essays on estimating consumer demand: Semiparametric and nonparametric approaches

In this thesis, I apply nonparametric/semiparametric methods to estimating consumer demand. In particular, I focus on the correction of sample selection bias and possible nonlinear functional relation in the consumer demand (expenditure) functions. Chapter II examines the effect of smoking choice on the demand for medical care when the smoking status is treated as endogenous. The smoking status is possibly endogenous in that the decision to smoke or not is correlated with unobservables relegated to the error term in the demand specification. Parametric and semiparametric approaches are employed to accommodate endogeneity of the smoking decision and enable consistent estimates of a demand function for medical care. I have concluded that consistent estimates from the two-step estimation procedure uncover a distinctive and somewhat surprising pattern in medical expenditure by smokers. It is clear that the different behavior of smokers and non-smokers can be partly explained by the simultaneity of smoking status and medical expenditures, and accounting for this when estimating the demand for medical care is critical.; Chapter III investigates the patterns of Food-Away-from-Home (FAFH) expenditure in China. Using China's urban household-level survey data from 1992 and 1998, I study household food-away-from-home (FAFH) expenditure across two time periods and across regions. I use a popular parametric linear specification and a newly developed nonparametric estimation method to estimate the FAFH expenditure function. The nonparametric estimation results reveal some interesting FAFH consumption patterns which are not detected by the parametric method. Goodness-of-fit analysis and a model specification test suggest that the nonparametric method gives a better description of the data than the parametric approach.; Chapter IV models credit card borrowing with Type I Tobit and Type II Tobit approaches using 1998 Survey of Consumer Finance data. The Type II Tobit approach used here overcomes a shortcoming of the conventional Type I Tobit model, which restricts the coefficients of the binary borrowing decision (the first step) and balance level decision (the second step) to having the same sign. By separately estimating the first and second steps as reduced form functions, in which the overall effects of supply and demand are incorporated, the Type II Tobit procedure yields notably more consistent results with other studies and quantifies the marginal effect (elasticity) of a particular determinant for practical uses.