Estimating Production Functions with Fixed Effects

I propose an estimation procedure that can accommodate fixed effects in the widely used proxy variable approach to estimating production functions. The procedure allows unobserved productivity to have a permanent component in addition to a (nonlinear) Markov shock. The procedure does not rely on differencing out the fixed effect and thus is not restricted to within-firm variation for identification. Finally, the procedure is easy to implement as it only entails adding a two stage least squares step using internal instruments.