Panel Sequential Group Estimation of Interactive Effects Models

ABSTRACT This paper proposes a novel procedure to identify latent groups in the slopes of panel data models with interactive effects. The method is straightforward to apply and relies only on closed‐form estimators when evaluating the objective function. This achieves substantial computational gains compared to alternative non‐linear methods and enables estimation even in large datasets. We establish consistency of the procedure and confirm through simulations its good finite sample properties and the validity of post‐grouping inference routines. The procedure's empirical usefulness is illustrated with applications to output growth and international trade.