Although lauded for their ability to outperform broader market indices, hedge funds’ “Best Ideas” do not excel relative to the rest of their portfolios, according to a study produced by Epsilon Asset Management, a quantitative investment firm, in collaboration with New York University.
Although lauded for their ability to outperform broader market indices, hedge funds’ “Best Ideas” do not excel relative to the rest of their portfolios, according to a study produced by Epsilon Asset Management, a quantitative investment firm, in collaboration with New York University.
Epsilon partners Faryan Amir-Ghassemi and Michael Perlow co-authored the report – “Hedge Fund Alpha and their Best Ideas” – with National Science Foundation award-winning assistant professor of finance and risk engineering Andrew Papanicolaou of NYU.
The study focused on Form 13-F regulatory filings data submitted by nearly 1,500 hedge funds between 1999 and 2018. The aim was to quantify the stock picking skill of these vaunted managers, and to see whether their Best Ideas truly added a higher degree of alpha to their funds’ overall performance.
Their findings suggest this is not the case. While the amount of alpha generated through long-equity stock picks by hedge funds is meaningful over a twenty-year period – some 250 basis points – when it comes to Best Ideas they have made little difference, performance-wise, to the overall portfolio.
Investors who are pitched “Best Ideas” from hedge fund managers should, therefore, drill deeper to substantiate such claims before being seduced and parting with their capital. How does the manager construct their portfolio? What risk parameters do they use and how do they right size positions?
This is a key consideration when investors have to justify spending fees on active managers when they can easily buy ETFs and tracker funds at a fraction of the cost.
Epsilon’s own research process focuses on bottom-up stock selection through a systematic process, predominantly in US domestic equities. Explaining the purpose of the study, Amir-Ghassemi says: “We thought that producing some research as a corollary to our investment strategies would be a good way to let the world know there are thousands of ways to extract structural risk premia from the data.
“Our approach is to see if there are risk premia associated with this filing data in the same way that quantitative investors look at fundamental data to construct portfolios using value, quality or growth signals.”
Epsilon uses statistical techniques that have been recently popularized by Wisdom of the Crowds to try to improve forecasting i.e. how can you look at a huge variety of different forecasts (which is essentially what each hedge fund strategy is) and come up with meta signals that can resiliently find higher probabilities of success?
“Regulatory filings are one of the few sources that allow you to do this in an unbiased fashion,” adds Amir-Ghassemi.
Papanicolaou said that while hedge funds can rightly boast of stock picking skill, “The notion that ‘Best Ideas’ show any statistical outperformance against the rest of their portfolios was simply not supported by our findings.”
Persistence of skill
At Epsilon, the firm’s goal is to try to capture top quartile stock picking alpha, just in a more consistent fashion than allocating directly to managers. Managers change, their returns mean revert, but skill sets are more persistent and can be boosted across all managers to develop more resilient alpha signals.
“What we try to do internally is find those signals we think are persistent for alpha generation; it’s a lot more complicated than just eliciting Best Ideas,” explains Amir-Ghassemi.
“The research study we produced showed that that naïve approach hasn’t worked in a decade and doesn’t generate the results investors might otherwise have expected. We think the robustness of our process leads to differentiated results.”
Peeling the onion
As every fund investor is warned, ‘past performance is not a predictor of future performance’. In their study with NYU, Epsilon tried to peel back the onion. If performance alone is not predictive, they asked, are the building blocks of performance?
“We do this as broad as possible, across every type of manger, to understand what is the true systematic behaviour of their funds’ performance. If we find predictive signals in the data, we try to factorise them through a canon of existing quantitative investment strategies and impart them into our investment strategies,” states Amir-Ghassemi.
By better understanding predictive signals in the behaviour of portfolio managers, it allows Epsilon to create a quantitative metric to describe that particular skill set. “Take something like stock selection,” says Amir-Ghassemi. “If you have a theory that managers who do a better job of selecting stocks in the past will do so in the future, you can test that with the data. We can measure the alpha versus beta component of their portfolio, and we can test that metric across our entire universe of fund managers, for persistence of skill.
“Then, you can think about factorizing it, in the same way that a stock might have higher value or higher momentum. We look for investment management skill and see if we can translate it into a risk premium we can harvest.”
‘Best Ideas’ work in mutual funds
Unlike hedge funds, empirical analysis of mutual fund portfolio positions (Cohen, Polk, Silli), suggests that Best Ideas in mutual funds do outperform the rest of their portfolios. In Epsilon’s study, it found little evidence of deteriorating alpha as it removed larger baskets of Best Ideas from its analysis, contrary to what would be expected in the event of ‘Best versus Rest Idea’ outperformance.
“We know their performance is less transparent and they don’t react as sensitively to daily capital flows than mutual funds. And we also know that hedge funds charge performance fees, not just management fees. The question is, with these structural differences, would we notice any difference in the behaviour of their Best Ideas?
“Our research suggests that hedge funds are less incentivised to stuff their portfolios with closet index exposure, which is why the bottom of their portfolio, as defined through these best idea methodologies, performs similarly to the top of their portfolio. This is contrary to the Best Idea outperformance in mutual funds, which respond to capital inflow by closet indexing,” argues Amir-Ghassemi.
Alpha performance deteriorating
One final point to observe in Epsilon’s study with NYU is the fact that alpha generation, while better than in mutual funds, has nevertheless fallen. Between July 1999 and June 2009 (Period 1), using a three factor model, Epsilon found that hedge funds returned 350 to 400 basis points of alpha per year. During that period, the average aggregate reported assets, was USD700 billion. In contrast, between July 2009 and December 2018, monthly alpha generated was as low as 5 basis points per month, gross of fees. At the same time, average aggregate reported assets grew to close to USD2 billion.
Central bank intervention following the calamitous ’08 financial crash led to an unparalleled period of market growth for investors, where beta performance became so dominant that it effectively quashed hedge funds’ ability to maintain stock-picking outperformance.
“This means manager selection is more important today than it was historically. The tailwind is less prominent,” says Amir-Ghassemi, who offers a final thought in conclusion:
“When doing due diligence on investment managers, although the stock ideas they use to illustrate their research process and their investment philosophy can be useful from a qualitative standpoint, understanding the guts of the portfolio construction process is probably more important; and oftentimes tends to get overlooked by investors.”
The full report can be read here.
