SOEP Brown Bag Seminar

Pseudo-event studies promise to overcome the limited availability of panel data by estimating event-study effects from widely available cross-sectional data. Originally developed and increasingly used for child penalty estimation, the approach has been proposed for broader applications. We show that pseudo-event study estimates are systematically biased for both pre-trends and treatment effects and analytically characterize this bias. The similarity to panel estimates in child penalty applications reflects a special case and need not generalize. Pseudo-pre-trends are mechanically concentrated around zero irrespective of actual pre-trends and therefore uninformative about the identifying assumptions. Simulations and an application using the German Socio-Economic Panel confirm the theoretical predictions.