Instrumental variables quantile regression for panel data with measurement errors

This paper develops an instrumental variables estimator for quantile regression in panel data with fixed effects. Asymptotic properties of the instrumental variables estimator are studied for large N and T when Na/T ! 0, for some a > 0. Wald and Kolmogorov-Smirnov type tests for general linear restrictions are developed. The estimator is applied to the problem of measurement errors in variables, which induces endogeneity and as a result bias in the model. We derive an approximation to the bias in the quantile regression fixed effects estimator in the presence of measurement error and show its connection to similar effects in standard least squares models. Monte Carlo simulations are conducted to evaluate the finite sample properties of the estimator in terms of bias and root mean squared error. Finally, the methods are applied to a model of firm investment. The results show interesting heterogeneity in the Tobin’s q and cash flow sensitivities of investment. In both cases, the sensitivities are monotonically increasing along the quantiles.