Dynamic panel data: a useful technique in experiments

Abstract

Numerous experimental studies use a panel approach to analyze repeated experiments involving a large number of periods. They use “static” panel techniques and do not incorporate any temporal dependency (lags) of the dependent variable. This paper introduces dynamic panel data techniques to experimental economists. This is a standard tool in many other fields of economics and might also be useful in our discipline. It uses the lags of the dependent variable as explanatory variables. Although the coefficients on lagged dependent variables might be far from our interest, the introduction of these lags becomes crucial to control for the dynamics of the process. To show the advantages of this technique, we have compared two datasets using static and dynamic panel data. We conclude that the use of dynamic panel data models in the context of experiments allows to unravel new relationships between experimental variables and highlighting new paths in behaviors.