How to Specify Panel Data Analyses?
Within-estimation with panel data (e.g. fixed-effects regression) has the advantage that it automatically controls for all time-constant unit-specific confounders. Therefore, within-estimation is meanwhile widely used in social research. However, available textbooks are largely silent on how to specify a within analysis and therefore it is no wonder that many published studies use problematic specifications that have the potential to spoil the advantages of within-estimation. In this workshop we will discuss several of these specification issues for within-estimation: How to model impact functions? How to model the age effect? Should one use a control group? How to use hybrid models? How to specify interaction effects? Should one use lagged outcome/treatment variables? How to get rid of the parallel trends assumption? How to deal with reverse causality?
Instructor: Prof. Dr. Josef Brüderl (Ludwig-Maximilians-Universität, Munich)
(for registered participants only)