The panel/longitudinal data analysis course covers most of the traditional panel data estimation techniques for micro panels in which the number of individuals (or firms etc.) is large, but the number of time periods is quite small. It focuses on the treatment of unobserved individual specific heterogeneity and discusses the difference between random and fixed effects model specifications.
Attention is given to the estimation of models with explanatory variables that are not strictly exogenous. This means that there can be feedback from the process to be explained to the explanatory variables (for example outputs and inputs in a production function, the effect of previous cigarette consumption on current consumption), or simultaneous determination. In these cases models in first differences can be estimated with instrumental variables estimation techniques. For non-linear count data models the treatment of unobserved heterogeneity is less straightforward. Various ways of dealing with it will be discussed with special focus on how to interpret the results.
The course is a mixture of lectures and applied sessions. Course participants will apply the various techniques using real data on their computers. Software applications used are Stata and PcGive, and no prior knowledge of these is assumed. It is assumed that participants have a basic knowledge of econometrics, to a similar level to that taught in the Introductory Microeconometrics course.
Lecture notes, data and papers from this course can be found in the Resources section of this website