A computationally simple bias correction for linear dynamic panel data models is proposed and its asymptotic properties are studied when the number of time periods fixed or tends to infinity with units. The approach can accommodate both fixed-effects random-effects assumptions, heteroskedastic errors, as well higher-order autoregressive models. Panel-corrected standard errors that allow robust ...