Testing for Serial Correlation in Fixed-Effects Panel Data Models

In this paper, we propose three new tests for serial correlation in the disturbances of fixed-effects panel data models. First, a modified Bhargava, Franzini and Narendranathan (1982) panel Durbin-Watson statistic that does not need to be tabulated as it follows a standard normal distribution. Second, a modified Baltagi and Li (1991) LM statistic with limit distribution independent of T , and, third, a test using an unbiased estimator for the autocorrelation coefficient to achieve robustness against temporal heteroskedasticity. The first two tests are robust against cross-sectional but not time dependent heteroskedasticity and the third statistic is robust against both forms of heteroskedasticity. Furthermore, all test statistics can be easily adapted to unbalanced data. Monte Carlo simulations suggest that our new tests have good size and power properties compared to the often used Wooldridge (2002)-Drukker (2003) test.