Partial Mean Processes with Generated Regressors: Continuous Treatment Effects and Nonseparable Models
2013/08/13
研討會日期 : 2013-08-13
時間 : 14:30
主講人 : Dr. Ying-Ying Lee
地點 : Conference Room B110
演講者簡介 : Ying-Ying Lee received her Ph.D. in Economics from University of Wisconsin-Madison in 2013. She will be Postdoctoral Research Fellow at Oxford University in Fall, 2013. Her research fields are econometric theory and applied econometrics.
演講摘要 : The unconditional distribution of potential outcomes with continuous treatments and the quantile structural function in a nonseparable triangular model can both be expressed as a partial mean process with generated regressors. I propose a multi-step nonparametric kernel-based estimator for this partial mean process. A uniform expansion reveals the influence of estimating the generated regressors on the final estimator. In the case of continuous treatment effects, an unconfoundedness assumption leads to regression on the generalized propensity score (Hirano and Imbens, 2004), which serves as the generated regressor in the partial mean process. Analogous to the binary treatment effect case, my results suggest that the generalized propensity score reduces the dimension of nonparametric regression in estimation, but does not improve first-order asymptotic efficiency. Nonseparable triangular models commonly include a conditional independence assumption that yields a control function approach to deal with endogeneity (Imbens and Newey, 2009). In a preliminary step, the control variable is estimated nonparametrically as a generated regressor. My general partial mean process results can then be applied to provide the asymptotic distribution of the nonparametric estimator for the average and quantile structural functions. By extending my results to Hadamard-differentiable functionals of the partial mean process, I am able to provide the limit distribution for estimating common inequality measures and various distributional features of the outcome variable, such as the Gini coefficient. Monte Carlo results demonstrate the finite sample behavior of my estimator. In addition, a substantive empirical application using data from a Colombian conditional cash transfer program illustrates the usefulness of the current findings for the estimation of continuous treatment effect models.
