Modeling Private Equity: A Machine-Learning Approach

Abstract

We introduce a flexible time-series estimator to measure factor exposures of individual private equity funds; our model estimates individual fund factor exposures with minimal assumptions about the underlying return process. Under varying autoregressive and moving-average simulations, we estimate systematic exposures with less bias and variance (thus, lower MSE) than competing estimators. We apply our model to a newly available commercial dataset, PitchBook, and find alphas that are higher than those reported in the recent literature. The large dispersion, across individual funds, of our estimated alpha and beta coefficients highlight the benefits of fund-specific analysis, compared to past cross-sectional estimation approaches.

Keywords: Private Equity, Buyout Funds, Time-Series, Supervised Machine Learning, Jackknife, Cross-Validation, Factor Modeling, Long Horizons, Overlapping

JEL Classification: C01, C22, G12, G17