A0779
Title: Specification analysis in quantile regression models with control functions
Authors: Xiaojun Song - Guanghua School of Management, Peking University (China)
Jen-Che Liao - National Chengchi University (Taiwan) [presenting]
Abstract: Nonparametric consistent tests are proposed for correct specifications in a quantile regression model with a control function. Specifically, the tests include the specification test of a parametric conditional quantile function, the specification test of a parametric control function, and the test of exogeneity of explanatory variables. Using a transformation of quantile restrictions to mean restrictions under the null hypothesis, the proposed tests are constructed based on the mixed residuals that combine the null and alternative models. The asymptotic properties of the proposed tests are established. To improve inference accuracy, a simple wild bootstrap approach is suggested, with its theoretical justifications provided. A simulation study examines the finite-sample properties of the proposed tests. For empirical illustration, the proposed methods are applied to testing quantile Engel curves with endogenous total expenditure.
