Forget What You Know About Picking the Best PE Funds
Machine learning algorithms can be used on fund documents to detect “patterns that could help you untangle the good managers from the bad ones,” says Reiner Braun of the Technical University of Munich.
During the selection process for private equity managers, institutional investors often evaluate quantitative metrics such as fund size and past performance. But new research suggests that they may need to look beyond these factors to identify managers who can generate outsized returns.
According to a new research paper, investors may have a better chance of choosing top performing PE funds by analyzing qualitative information contained in investment strategy documents than they do by evaluating quantitative factors, such as past performance and the pace of fundraising.
To conduct the research, the authors sorted one group of private equity funds in three different ways: one by past returns, one by fundraising speed, and one by qualitative information found in documents. The researchers then compared the TVPI (total value to paid-in capital) ratio of the three different groupings.
The top quartile of funds in the group sorted by past returns had a TVPI of 1.72, according to the paper. The top quartile of funds in the second group, which were ranked by the rate at which they raised capital from investors, had a slightly higher TVPI of 1.95. For the third group, researchers used a machine learning algorithm to analyze the investment strategy documents provided by fund managers and then rank funds based on their predicted probability of success. The researchers found that the top quartile of funds in the last group had the highest TVPI of 2.09.
The paper is based on a study of 395 PE funds in Europe and the United States. The researchers collected the funds’ quantitative information and the descriptive text of their investment strategies from their private placement memorandums.
“The best investors in private equity – the institutional investors – do react to quantitative information they read in PPMs,” said Reiner Braun, professor of entrepreneurial finance at Technical University of Munich in Germany and one of the paper’s five authors. It is reflected in the fact that larger funds and funds with higher past returns have an easier time raising capital from investors, according to the paper.
But the fact that using the machine learning-based approach to select managers produces a higher TVPI suggests that the text in PPMs could be more valuable than the quantitative information, according to Braun. “It is a source of information that has not been exploited by the majority of investors,” he said. “If investors want to gain some competitive advantage in selecting great fund managers…paying attention to this might be a viable strategy.”
“It’s the idea of market efficiency,” said Ludovic Phalippou, professor of financial economics at University of Oxford and one of the authors of the paper. “It’s like the same thing in public markets. You will never make money by having information that everybody has.”
Braun said that institutional investors can either go through PPMs themselves or use machine learning algorithms to decide which funds to invest in. Either way, it’s important to look beyond the quantitative metrics that most investors have been using to evaluate managers.
“An indication from our research for [institutional investors] is that…there are sections of the PPMs where there are patterns that could help you untangle the good managers from the bad ones,” Braun said. “Whether you believe in the potential of machine learning or you want to read it yourself, you should definitely look at it and take it seriously.”