• instrumental variables
• local projections
• VAR

The identification of the dynamic causal effect of a structural shock is a central topic in empirical macroeconomics. In the past decade, the use of external instruments to estimate impulse responses has gained increasing popularity. Quasi-experimental variables, correlated with the macroeconomic shock of interest but uncorrelated with other structural shocks, are often employed as instruments to isolate the impact of the unobserved macroeconomic shock. The dynamic causal effect can be estimated through an external instrument, either using Structural Vector Autoregressions (SVAR-IV) to identify the impact matrix or, without any auxiliary VAR step, employing Local Projections Instrumental Variables (LP-IV).
While the properties of SVAR-IV inference have been extensively investigated, the LP-IV literature lacks a deep analysis of the estimator performance. The aim of this work is thus to conduct a Monte Carlo study to assess and compare the properties of Local Projections IV and Structural Vector Autoregressions IV in estimating impulse response functions in finite samples.