Cemmap Working Paper (CWP53/17)

Estimating dynamic panel models: backing out the Nickell Bias

Date: 30 November 2017
Publisher: The IFS
JEL classification: C01, C22, C23

We propose a new estimator for the dynamic panel model, which solves the failure of strict exogeneity by calculating the bias in the first-order conditions as a function of the autoregressive parameter and solving the resulting equation. The estimator does well in a wide variety of situations where other estimators do not perform well: stationary initial condition, predetermined but not strictly exogenous regressors, and the presence of correlation between the error terms and the fixed effects. We also propose a general method for including predetermined variables infixed-effects panel regressions.