Testing for Selectivity Bias in Panel Data Models
针对面板数据中常见的缺失观测问题，提出了几种检验回归估计中是否存在选择性偏差的方法，包括变量添加检验和准豪斯曼检验，并通过蒙特卡洛模拟比较了它们的检验功效。
Missing observations are a rule rather than an exception in panel data.In this paper we discuss several tests to check for the presence of selectivity bias in regression estimates based on panel data.One approach to test for selectivity bias i n these estimates is to specify the missing data mechanism explicitly and to estimate the response mechanism and the regres-equation jointly.Alternatively, one can derive the asymptotically efficient Lagrange Multiplier test once an assumption on the response mechanism has been made.Both approaches are computationally demanding as e.g.multivariate probit models have to be estimated.We propose the use of simple variable addition and (quasi) Hausman tests to test for selectivity bias and compare the power of these tests with the asymptotically efficient tests using Monte Carlo methods.Keynesian and New Classical Models of Unemployment Revisited