Estimating Semiparametric Panel Data Models by Marginal Integration

We propose a new methodology for estimating semiparametric panel data models, with a primary focus on the nonparametric component. We eliminate individual effects using first differencing transformation and estimate the unknown function by marginal integration. We extend our methodology to treat panel data models with both individual and time effects. And we characterize the asymptotic behavior of our estimators. Monte Carlo simulations show that our estimator behaves well in finite samples in both random effects and fixed effects settings.