The repeated observation of economic agents revealed in panel (or longitudinal) data gives insight into the dynamic aspects of agents’ responses to changes in situation and environment. Understanding the dynamics of response is often crucial to understanding the impact of policy intervention.
Panels typically contain observations on many agents but usually capture just a short history of their responses. Agents differ in tastes and preferences and the environment in which each one operates is imperfectly measured. Understanding variation in response is important in assessing the distributional impact of policy interventions.
Panel data affords the opportunity to gain understanding of the variation across agents in their dynamic responses but many unsolved problems arise in making inferences of this sort. Unless data are generated by processes satisfying very restrictive conditions, the “incidental parameters problem” set out in Neyman and Scott (Econometrica, 1948) results in identification and consequent estimation problems. In reality these restrictive conditions, involving substantial amounts of linear structure, are unlikely to prevail.
These MasterClasses explore this issue and present solutions for cases in which nonlinearities are intrinsic to the problems considered.