Smooth Panel Local Projections

Abstract

This paper extends smooth local projections to panel data and uses Monte Carlo methods to explore the bias and variance properties of smooth panel local projections (SPLP). SPLP allows researchers to penalize the impulse response toward a polynomial, while standard panel local projections (PLP) are nonparametric but result in theoretically unappealing impulse response functions (IRFs) because they are too lumpy. Relative to PLP, SPLP has appealing properties in smaller samples. We demonstrate the utility of the estimator through two applications: oil shocks from Arezki, Ramey, and Sheng (2017) and democracy shocks from Acemoglu et al. (2019).

Keywords: Local Projections, Semiparametric Estimation, Impulse Response