Swedroe: Crowded Trades & Tail Risk

Over the last 20 years, the hedge fund industry has grown from a few hundred funds with an estimated $400 billion in assets under management (AUM) to several thousand funds with an estimated $3.2 trillion in AUM.

Hedge funds are fairly unique investors because they: (1) tend to be particularly active in seeking alpha; (2) often take highly concentrated positions; (3) are often run by managers who were trained on similar investment styles and therefore tend to seek similar investments in factors uncovered by academic research; and (4) use leverage and short sales.

Worth noting is that over the last 20 years, the number of publicly listed companies in the U.S. has declined by about 50% from nearly 8,000 in 1997 to under 4,000 in 2017. The result is that the universe of opportunities for hedge fund trades has shrunk significantly even as the number of funds and size of assets has grown.

These facts all make the study of hedge fund positions a good laboratory for understanding the role of active traders in creating risks associated with crowded trades.

Crowded Trades

We can define a crowded trade as a collection of similar positions in a particular asset by investors who share a common investment thesis and may therefore want to trade in the same direction at about the same time. Such trades run the risk of “fire sales,” which occur when a large investor(s) wants to sell significant amounts of an asset, and there are insufficient buys near the current market price. Financial crises, such as the failure of Lehman Brothers, often lead to fire sales.

Gregory Brown, Philip Howard and Christian Lundblad, authors of the February 2019 study “Crowded Trades and Tail Risk” sought to understand how concentrated positions held by hedge funds create risks and how they affect asset and portfolio returns. Their focus was on long positions in U.S. equities held by about 1,500 hedge funds from 2004 through 2016. Because crowdedness does not have a single widely accepted definition, they measure it in four ways.

• The value of holdings by hedge funds. For the average stock hedge fund, holdings grew more than threefold, from about $160 million in 2004 to over $500 million in 2016.
• The number of hedge funds invested in each stock. The number of funds holding a typical stock has roughly doubled from 15 in 2004 to 28 in 2016.
• The percentage of shares outstanding (PSO) held by hedge funds. Average PSO increased from about 8% in 2004 to 12% in 2016.
• Hedge fund holdings as a percentage of average daily trading volume (days-ADV). This is likely the best measure of the ones we consider, because it measures crowdedness in terms of the size of hedge fund holdings as well as the underlying liquidity of the stock. The days-ADV increased from about 18 days in 2004 to 26 days in 2016.

Quintile Analysis

Brown, Howard and Lundblad then sorted stocks into quintile portfolios based on the cross-sectional variation of each measure. As is common practice, “To limit the contaminating effects of difficult to trade micro-capitalization stocks, the smallest 20% of securities over the sample period are removed from the sample,” they wrote. The data sample constituted almost 6,000 securities. Following is a summary of their findings:

• In each measure, the most crowded portfolios have the highest returns, and the least crowded portfolios have the lowest returns. On a value-weighted basis sorted on days-ADV, the low crowdedness portfolio (Q1) average return is 8.5% per annum versus 11.3% for the high crowdedness portfolio (Q5). The difference of 2.8% was highly significant (t-stat of 6.2). The difference was much larger on an equal-weighted basis: 9.1% with a t-stat of 8.0.
• The crowdedness factor generates, on average, statistically significant alphas even after controlling for other traditional risk factors, including the relative illiquidity of the more crowded positions.
• At a particular point in time, some hedge funds are disproportionately invested in relatively crowded positions, whereas others are not.
• The difference between high and low illiquidity portfolios is positive and significant for the equal-weighted portfolio (even with the removal of the smallest 20% of stocks). Illiquidity matters.
• Constituent stocks are relatively sticky—transition probabilities of firms across portfolios are reasonably low. Taking the equally weighted days-ADV-based portfolios as an example, conditional on a firm being in portfolio Q1 (Q5), the likelihood of remaining there in the next period is 77% (82%).
• High PSO or (Q5) firms are significantly smaller and less liquid, on average, than firms in the low crowdedness portfolios (Q1).
• Some hedge funds disproportionately load on our crowdedness factor, others less so. For example, 58% of equity funds have significant exposures to crowdedness, while only 18% of global macro funds have significant exposures.
• Firms with relatively higher exposures to crowdedness exhibit greater realized volatilities during the financial crisis. And they become relatively more correlated with peers holding similar investment positions. In bad states of the world (times of market turmoil), funds holding crowded securities, on average, become riskier.

Crowds During Crisis

Brown, Howard and Lundblad then examined the extent to which returns are disproportionately affected by crowdedness during the financial crisis by calculating cumulative abnormal returns (CARs)—returns in excess of the returns implied by a factor model, such as the CAPM or Fama-French—for the crowded and uncrowded portfolios. They found:

• Hedge fund distress may transmit to the prices of the securities in which they disproportionately invest—there were significantly negative CARs for the crowded portfolios during 2008, which largely reversed in 2009.
• CARs for uncrowded positions were closer to zero (or positive in some cases).

They concluded: “These results suggest that during the financial crisis, firms with both significant hedge fund ownership and relatively low levels of daily liquidity can be characterized by exit frictions that exhibit significant, temporary price dislocations (i.e., fire sales).”

The authors note: “While the results described suggest that returns of individual stocks are subject to crowdedness risk, this does not mean that funds would necessarily have significant exposure to crowdedness. Funds may only hold a small part of their portfolio in crowded trades or may undertake other types of risk mitigation with short positions or derivatives unobserved in 13F filings.”

Therefore, they examined monthly hedge fund returns to see if crowdedness is an important risk factor. They found: “The days-ADV risk factor is statistically significant for about 26% of the hedge fund universe (and certain subsets of hedge fund strategies are highly exposed to crowdedness). These results indicate that crowdedness risk is important at the hedge fund portfolio level as well as the individual security holding level.”

Finally, Brown, Howard and Lundblad examined the relation between crowdedness and fund downside risk. They found that: “Funds with a higher average investment weight in the least crowded days-ADV portfolio experienced less severe drawdowns during the financial crisis, while funds with a higher allocation to the most crowded days-ADV portfolio experienced more severe drawdowns. This indicates that crowded trades are associated with relatively worse fund performance in particularly bad states of the world.”

Unique Risk Or Skill (Alpha)?

The findings from the study raise an interesting question: Is the average return difference documented due to risk compensation separate from that embodied in the other common risk factors used in asset pricing models, or is this an indication that hedge funds are, on average, picking the right securities and delivering bona fide outperformance?

Unfortunately for hedge fund investors, even if the results are an indication of skill (Q5 positions predict abnormal excess returns), the evidence on hedge fund performance demonstrates that the return to the skill has gone to the managers and not to investors. As reported in my February 2019 column, their performance over the last calendar decade was dismal.

Conclusion

It appears that while hedge funds are, on average, able to identify securities that will outperform (either because they have identified a unique risk factor or because they have securities selection skill), the high fees they impose (typically 2% plus 20%, and often significant other incremental fees), as well the trading costs incurred in implementing their strategies (including market impact costs), result in poor net returns—the only kind that matter to investors.

Another important problem identified by Brown, Howard and Lundblad is that because these crowded trades tend to perform poorly at exactly the wrong time (during crises), hedge funds are contributing to the tail risks incurred by investors—risks they are generally seeking to reduce. In other words, hedge funds in general are not hedging and narrowing the dispersion of portfolio returns but exacerbating the tail risks, which raises the question, “Why they are called hedge funds?”

Larry Swedroe is the director of research for The BAM Alliance, a community of more than 130 independent registered investment advisors throughout the country.