A1437
Title: Estimation bias in local projections with heterogeneous panels
Authors: Benjamin Alexander - Federal Reserve Bank of Chicago (United States) [presenting]
Peter Pedroni - Williams College (United States)
Abstract: Over the past two decades, local projections have become an important tool for estimating dynamics in macroeconomic data. While their properties in the time series context have been well studied, the behavior of local projections on panel data is not fully understood, despite numerous empirical implementations. The focus is on the econometric properties of local projections in a heterogeneous panel context. An analytical expression is derived for the asymptotic bias when heterogeneous dynamics are pooled across units of a stationary panel. This bias is shown to arise from a lagged dependent variable specification, and is shown to depends on the degree of dispersion in dynamics across units. On the basis of this asymptotic expression, finite sample biases with Monte Carlo simulations are documented. A consideration of feasible solutions are concluded with. Findings suggest the importance of appropriately accounting for heterogeneity in panel analysis with local projections.
