Random Subspace Local Projections

We show how random subspace methods can be adapted to estimating local projections with many controls. Random subspace methods have their roots in the machine learning literature and are implemented by averaging over regressions estimated over different combinations of subsets of these controls. We document three key results: (i) Our approach can successfully recover the impulse response function in a Monte Carlo exercise where we simulate data from a real business cycle model with fiscal foresight. (ii) Our results suggest that random subspace methods are more accurate than factor models if the underlying large data set has a factor structure similar to typical macroeconomic data sets such as FRED-MD. (iii) Our approach leads to differences in the estimated impulse response functions relative to standard methods when applied to two widely-studied empirical applications.