【主题】 | Endogenous High-Dimensional Quantile Regression: A Control Function Approach |
【报告人】 | 张凯夕（博士候选人，香港科技大学） |
【时间】 | 2024年5月30日周四10:00-11:30 |
【地点】 | 经济学院701会议室 |
【语言】 | 中文 |
【主持人】 | 金泽群讲师 |
【摘要】 | In this paper, we propose a two-step double selection estimator to estimate a high-dimensional quantile regression model in the presence of endogeneity based on the control function (CF) approach. First, we extend the idea of double selection to quantile analysis. We perform a l1 penalized quantile regression on the residuals of the reduced-form equation for the endogenous explanatory variable in the first step, and then select control variables to predict the tau-quantile of the conditional distribution of outcome with residuals included as an additional control. Second, we find that under general conditions, our model is computationally simpler than instrumental variable quantile regression (IVQR). Monte Carlo simulations also show that our estimator performs well under high-dimensional controls. Third, we employ our model to investigate the impact of compulsory schooling on earnings using 1530 instruments for education based on Angrist-Krueger data, and we find that high-dimensional quantile estimates are smaller than 2SLS and OLS estimates. |
【报告人简介】 | Kaixi Zhang is a PhD candidate at the Hong Kong University of Science and Technology. Her research field is econometric theory, with particular interests in quantile regression, high-dimensional models and machine learning. |
