Past performance is not… (helpful!)
Investing is sometimes similar to driving a car: a glance in the rearview mirror is not enough by itself to get you to your destination unscathed. Analogously, equity funds that have performed well in the past very rarely succeed at persistently earning above-average returns over the long run. Historical returns also foretell little about the future performance of private equity funds. Artificial intelligence could help financial analysts to sniff out future outperformers.
“Past performance is no guarantee of future results” – the information leaflet for nearly every investment product bears this warning or one like it. But in contrast to many other types of perhaps exaggerated disclaimers, the one above should not be construed as a legal safeguard (for the issuer of the investment product) against the unlikeliest occurrence conceivable. One can interpret it more as being a well-intended and above all fact-based word of advice for investors who may be somewhat overly naïve as well as for purported investment pros, because the statistical evidence is clear: equity funds that have performed well in the past very rarely succeed at persistently earning above-average returns over the long run. S&P Global supplies fresh numerical data on this phenomenon every year. According to its US Persistence Scorecard published in May, only 48.6% of top-quartile US equity funds in 2018 were still in the top 25% of all mutual funds one year later, and not a single former top performer was still in the uppermost quartile as soon as four years later. The findings are sobering, even when the bar is lowered significantly. Of the equity funds that ranked among the best 50% of all mutual funds in 2018, only 16.8% were still in the top half three years later and just 2.7% were still present there four years later. Observation over a longer period shows that these numbers weren’t distorted by the difficult pandemic years for fund managers: of the equity funds that ranked among the top 50% during the 2013–2017 period, only 37% managed to stay in the top half during the subsequent five-year period. If fund-manager performance were completely random, one would expect 50% of the winners in the first five years to post an above-average performance also in the second five years; if substantially more than 50% of the winners won again in the second interval, that might even be evidence of the existence of consistent fund-manager skill. The actual findings, however, clearly refute this.
“I don’t know any investors who shouldn’t act as if markets are efficient.” | Eugene Fama
Performance persistence of US equity funds ranked in the top 25% and top 50% in 2018
Sources: S&P Dow Jones Indices, Kaiser Partner Privatbank
The evidence in private equity
In contrast to the situation with regard to conventional equity funds, for a long time it was a common practice among investors in private equity funds to bet on funds (and managers) that were winners in the past. Even to this day it is a prevalent tactic to select top-quartile funds exclusively if possible in the hope that they will stay at the top of the pack in the future. This received wisdom about the performance persistence of private equity managers had for a long time also been underpinned by academic studies.1 However, earlier analyses were often based on data from the 1980s and 1990s. That data has proven, in the meantime, to be qualitatively substandard. Moreover, the earlier research, which is already outmoded today, does not reflect the massively enlarged size of the private equity universe since the turn of the millennium, nor does it reflect its greatly increased level of professionalization. Newer research in recent years has debunked the earlier assumptions in the meantime. For example, one study bearing the title “Has Persistence Persisted in Private Equity? Evidence from Buyout and Venture Capital Funds” 2 demonstrates on the basis of current data (through end-2020) drawn from the high-quality database from Burgiss that there is practically no performance persistence among classical private equity buyout funds. Past performance says nothing about future (out)performance, particularly when the database available to the investor at the time of the investment decision is the one used as the information basis. Since it ordinarily can take 12 to 15 years until the full redemption of a private equity fund, meaning that it can also take that long to determine the definitive final performance numbers, investors can only factor the interim “live” performance of private equity funds into their deliberations. Venture capital funds are an exception, though. The authors of the study find (compelling) evidence even today of performance persistence in the venture capital space – top VC fund managers remain more successful than average more frequently than pure random chance would have it.
Help from artificial intelligence
The quite important role that historical fund performance plays in the world of private equity owes in large part to the opacity of the industry to date. Whereas enormous databases on conventional equity funds exist, as does copious research on most publicly traded companies, the public availability of data on private markets is much scarcer. A very lopsided information asymmetry exists between investors and private equity managers, and managers have lots of leeway to frame their track record in a favorable light for themselves. This makes selecting private equity funds a particularly difficult undertaking. Artificial intelligence could remedy this situation in the future because even though ChatGPT cannot replace private bankers anytime soon, smart algorithms could definitely provide helpful assistance for financial analysts in the future. A quintet of scholars led by Oxford Professor Ludovic Phalippou applied just such machine-learning algorithms to the prospectuses of private equity funds.3 In their study titled “Limited Partners versus Unlimited Machines; Artificial Intelligence and the Performance of Private Equity Funds”, artificial intelligence proved very successful at extracting performance-relevant insights from the sections of fund prospectuses in which managers describe their investment strategy and at using that information to predict potential outperforming and underperforming funds. An algorithm trained on data from 2003 through 2013 sorted funds with vintages from 2014 to 2016 by their probability of outperformance. The top quartile of funds (signifying the highest probability of outperformance) achieved a total-value-to-paid-in-capital multiple of 2.09 through end-2022, 0.23 points higher than the average for all funds. Since private equity funds usually hold their investments for five years on average, this translates into an annual outperformance of 4 percentage points, which is a handsome excess return. Artificial intelligence was thus actually able to derive performance-relevant added value from qualitative information to make assertions about the future performance of private equity funds. In contrast, quantitative facts, such as investor interest in private equity fundraising for example, were much poorer indicators of a good future performance. Historical fund performance, which the study also put to the test, yielded added information value as well, though differently than one might think: to wit, (former) top-quartile funds underperformed significantly and did worse than the median of all funds. “Past performance is not indicative of future returns” therefore also goes for private equity funds. Meanwhile, “big is beautiful” is a much less valid tenet than many an investor might suppose.
The bottom line is that when investing in private-market assets, it is very important to work together with experts, possibly also with the aid of artificial intelligence in the future. Moreover, especially in this asset class, it pays to heed the advice to take diversification to heart and to avoid putting all eggs in one basket.
Looking back at past performance doesn’t help… | …but qualitative information and machine learning definitely do help
Performance of selected funds
Sources: R. Braun et al. (2023), Kaiser Partner Privatbank
1) Including, for example, Steven N. Kaplan, Antoinette Schoar (2005): “Private Equity Returns: Persistence and Capital Flows”
2) Robert S. Harris, Tim Jenkinson, Steven N. Kaplan, Ruediger Stucke (2022): “Has Persistence Persisted in Private Equity? Evidence from Buyout and Venture Capital Funds”
3) Reiner Braun, Borja F. Tamayo, Florencio López-de-Silanes, Ludovic Phalippou, Natalia Sigrist (2023): “Limited Partners versus Unlimited Machines; Artificial Intelligence and the Performance of Private Equity Funds”