ABSTRACTTHREE ESSAYS IN APPLIED ECONOMETRICSByMonthien SatimanonThe three essays are self-contained and are the combination of applied and empirical econometrics. They are "Comparisons of Approaches in Measuring Willingness to Pay for Environmental Services", "Comparisons of Approaches in Measuring Causes of Wage Inequality", and "Estimation of Binary Response Model with Endogeneity and Hetero- skedasticity." The first essay proposes a comparison of both parametric and semiparametric estimation of willingness to pay (WTP) for environmental services. In order to solve for problem of inconsistency of estimation since heteroskedasticity, several conventional and new methods are used in the analysis. The methods are Probit (Probit), Heteroskedasticy Probit (HP), Turnbull (T), Watnabe (2010), Ahn (2000), and sieve semiparametric estimator (S). The comparison includes the estimated parameters as well as the estimated standard errors since the WTP is derived from these parameters. Monte Carlo simulations have been used to compare finite sample properties of each estimating methods. The empirical application comes from a study of the demand for payment for environmental services, water quality preservation, in eastern Costa Rica. By Monte Carlo Simulation, we found out that neglecting heteroskedasticity could lead to over estimation of WTP by almost 100 percent. In the empirical study, we found out that WTP for water conservation program in villages in eastern Costa Rica is about 2400 Colones that is about two times of the current monthly water cost program. This estimate is consistent with previous studies in the water conservation program.The second essay proposes a comparison of both parametric and semiparametric estimation of causes of income equality. In quantile regression setting, this paper analyzes the determinants of wage inequality with endogenous categorical regressors. The framework of Lee (2007) has been extended to cover the case where control function comes from generalized residuals of ordered probit models. We found out that the proposed method yields not only consistent but also efficient estimated parameters when there is a present of both endogeneity and heteroskedasticity while the conventional estimators led to overestimated parameters. In addition, we use all the proposed to estimate the return to education on wage using the data from Current Population Survery (CPS). We found out that returns to education are not monotonically increase throughout the wage distribution. The third essay analyzes the binary response model that encounters the problems of endogeneity and heteroskedasticity that lead to inconsistent estimated parameters. Our model allows both heteroskedasticity in structural and reduced formed equation. To handle both problems, we employ control function estimation with sieve estimator. The first step generates control function with flexible form of multiplicative exponential heteroskedasticity. The second step employs the use of maximum likelihood method to find the average partial effects with the adjusted standard errors based on Ackerberg et.al. (2012). Monte Carlo simulations are conducted to study the performance of this new estimator comparing to standard binary choice models. Then, the estimator is applied to estimation of female labor supply.
Read
