Introducing Granularity to Private Equity Performance Replication

Abstract

Over the past decade, the private equity (PE) asset class has generated superb returns that exceeded the S&P 500. However, this perceived outperformance comes at a price: the illiquid nature and the intricate and taxing fee structure. To overcome these challenges, this paper seeks to develop a quantitative, low-cost replication strategy for the performance of domestic PE returns with liquid U.S. equity markets. Specifically, as existing literature all overlook the fundamental process of how general partners (GPs) identify buyout targets, we introduce an unparalleled level of granularity to PE performance replication literature that addresses this process. We devote our efforts to predicting the likelihood of a public company getting bought out by a PE firm, which we denote as “PE-Selectabiliy,” using a variety of weighted regression as well as machine learning models. We then construct replicating portfolios by selecting stocks with the highest predictive power. We find evidence of returns outperformance in the replicating portfolios created by the weighted ordinary least squares (WOLS) model and the ensembles methods. Annualized returns of said models fall within the range of 15.07% and 19.24%, greatly exceeding the S&P 500 Total Return index’s same-period return of 11.15%. Adjusting for risk, the replicating portfolio consisting of the top 15 stocks with the highest “PE-Selectability” as determined by the WOLS model demonstrates strict dominance of all return and risk characteristics over the index. Our feature selection analysis reveals both features concerning a firm’s business operations and those concerning capital structure influence “PE-Selectabiliy,” suggesting that GPs place consideration into business fundamentals as well as potential return on capital when evaluating buyout targets.