Panel data enable the modeler to explore and analyze latent heterogeneity across individuals in ways that cannot be revealed by cross sections. Models for panel data incorporate individual effects in many different ways to account for these characteristics. This course will develop models that extend beyond the traditional linear random and fixed effects linear regressions.
The major theme of the course is how to estimate and interpret models that incorporate latent heterogeneity in nonlinear frameworks such as discrete choice and nonlinear regressions. A range of recently developed techniques as well as applications in the literature will be discussed. Topics will include: (to set the stage) panel data and fixed and random effects linear regression models; nested random effects models; nonlinear models with fixed effects including counts and discrete choice conditional vs. unconditional estimation and the incidental parameters problem; random effects and practicalities of estimation quadrature and simulation based estimation; hierarchical and random parameters models; latent class models and finite mixture modeling; dynamic discrete choice models and the initial conditions problem; Bayesian and classical approaches to estimation. Participants will also be introduced to a variety of advanced econometric methods including using quadrature vs. simulation to estimate random effects models, cluster corrections for stratified data, unconditional fixed effects estimation, maximum simulated likelihood estimation, the EM algorithm and Gibbs sampling for Bayesian estimation.
The course will include lectures that develop the relevant theory and extensive practical, laboratory applications. Emphasis in the laboratory sessions will be on estimation of nonlinear models, interpretation of results and using the models for analysis of individual heterogeneity. Course participants will apply the techniques on their own computers using the LIMDEP computer program and real panel data sets. Prior knowledge is assumed to include a course in econometrics at the level out of the textbook Greene, W., Econometric Analysis, 5th edition. Familiarity with LIMDEP will be helpful, but is not necessary.