Over-Identification Tests in Earnings Functions With Fixed Effects
推导了面板数据固定效应模型的整体拟合优度统计量，并将其应用于检验工会工资效应和学校教育回报率，发现固定效应假设在人力资本收入函数中可能不适用。
The fixed-effects model for panel data imposes restrictions on coefficients from regressions of all leads and lags of the dependent variable on all leads and lags of right-side variables. In the standard fixed-effects model, the omnibus goodness-of-fit statistic is shown to simplify to the degrees of freedom times the Rz from a regression of analysis of covariance residuals on all leads and lags of right-side variables. This result is applied to test models for the union-wage effect using data from the National Longitudinal Survey of Youth (NLSY). Identification and estimation of the return to schooling in models with fixed effects is also discussed. Although schooling is often treated as time-invariant, schooling increases over a five-year period for nearly 20% of continuously employed men in the NLSY The analysis of covariance estimate of the returns to schooling is precisely estimated and roughly twice as large as the ordinary least squares estimate. Unlike in the union-wage effects equation, however, the omnibus goodness-of-fit test suggests that the fixed-effects assumption may be inappropriate for human capital earnings functions.