Identification by Laplace transforms in nonlinear time series and panel models with unobserved stochastic dynamic effects

Abstract

We consider nonlinear parametric and semi-parametric models for time series and panel data including unobserved dynamic effects. These regression models have an affine specification with respect to lagged endogenous variables and unobserved dynamic effects. We derive conditional moment restrictions based on suitable Laplace transforms. We show how to deploy these nonlinear moment restrictions to identify the parameters of the affine regression model, and the parametric or nonparametric distribution of the unobserved effects. This approach is appropriate for studying identification in (nonlinear) latent factor models encountered in macroeconomic and financial applications as well as in panel models with stochastic time effects.