Mispresenting the Risks of Illiquid Funds

Spurious autocorrelation is a false or misleading correlation between two variables that appears to be related but is not. This can happen due to chance (such as when two variables exhibit similar trends over time even when there's no direct causal link) or because there's a third, hidden factor—a confounding variable—that's influencing both variables (correlation doesn’t imply causation). A classic example of spurious correlation is that ice cream sales increase when shark attacks rise. It might seem like people eat more ice cream when sharks are around, but the real reason is that both ice cream sales and shark attacks tend to increase in the summer, when people are more likely to be at the beach. The summer is the confounding variable.

Spurious autocorrelation can lead to misleading conclusions if not carefully considered (see here and here). Funds (such as private equity and private real estate) that invest in illiquid assets report returns with spurious autocorrelation, which can lead to misleading conclusions if not carefully considered. Understanding the autocorrelation leads to the conclusion that the fund’s volatility is higher than the reported returns show. Unless an adjustment is made to account for the autocorrelation, investors will underestimate risk and overestimate risk-adjusted returns, leading to overallocation to such assets. AQR’s Cliff Asness has referred to the misleading presentations of the returns of illiquid investments that do not address the issue of autocorrelation (caused by stale marks-to-market) as “volatility laundering.”

To help investors see the real risks and rewards of a such funds, tests like the Durbin-Watson test or the Breusch-Godfrey test have been developed to detect autocorrelation in time series data. Having identified the autocorrelation, an “unsmoothing” process reveals the true performance of a fund. The traditional method is a simple one-step unsmoothing that takes the reported returns and adjusts them based on the average difference between the reported returns and the returns we would expect if there were no smoothing. And if the fund is consistently reporting returns that are 2% higher than what we'd expect, we should subtract 2% from each reported return to get a more accurate picture.

Spencer Couts, Andrei Gonçalves, and Andrea Rossi, authors of the February 2024 study “Unsmoothing Returns of Illiquid Funds,” found that common sources of spurious autocorrelation are not fully resolved by traditional one-step unsmoothing methods, leading to underestimation of systematic risk. Thus, they proposed a three-step unsmoothing technique that tries to account for different smoothing patterns. Their three steps are:

1. Identifying smoothing periods: Look for periods where the reported returns seem unusually good or bad compared to what we would expect.

2. Estimating smoothing intensity: Calculate how much the fund might have been smoothing during those periods.

3. Adjusting returns: Adjust the returns for those periods based on the estimated smoothing intensity.

They then applied their three-step process to hedge funds and private commercial real estate funds. They divided the hedge fund universe into three groups based on the degree of illiquidity of the underlying assets: low, medium, and high. Their final hedge fund dataset was based on merging the Lipper Trading Advisor Selection System database with the BarclayHedge database. It covers a total of 5,069 funds with at least 36 uninterrupted monthly observations over the period from January 1995 to December 2017. Their final private commercial real estate (CRE) dataset is from the National Council of Real Estate Investment Fiduciaries (NCREIF). It covers a total of 66 funds with at least 36 uninterrupted quarterly observations over the period 1994-2017. Following is a summary of their findings:

· Fund-level autocorrelations are high in private CRE and hedge funds—funds with similar illiquid investments have a common source of spurious autocorrelation not fully resolved by traditional unsmoothing methods, leading to underestimation of systematic risk.

· Volatility strongly increases after unsmoothing and the fraction of volatility due to systematic risk increases when the three-step unsmoothing process was used. Thus, the three-step method significantly improved the measurement of funds’ risk exposures and risk-adjusted performance, especially for highly illiquid funds. (Note that as shown in the charts below the autocorrelations of returns were high in the low- and mid-liquidity categories, but very low in the high liquidity funds, which should be expected.)

Real Estate and Hedge Fund Return Autocorrelation: Fund-level vs Aggregate Panel (a) plots the average 1st order autocorrelation coefficient for returns of private commercial real estate (CRE) funds (quarterly returns from 1994 to 2017) and Hedge Funds (monthly returns from 1995 to 2017), with the latter sorted on strategy liquidity (see Subsection 2.2 for the details on the strategy liquidity sort). Panel (b) plots the analogous measure, but for average returns (i.e., first taking the equal-weighted average of fund-level returns and then calculating the autocorrelations). We consider three definitions of returns: observed returns, 1-step unsmoothed returns (Geltner (1991, 1993) for private CRE funds and Getmansky, Lo, and Makarov (2004) for Hedge funds), and 3-step unsmoothed returns.

· After three-step unsmoothing, the average r-squared of funds in the low liquidity strategies increased by 14.9% (from 34.3% to 39.3%) relative to reported returns and by 21.3% (from 32.4% to 39.3%) relative to one-step unsmoothing. This suggests that the three-step unsmoothing method better uncovers the true systematic risk exposure of hedge funds, which is usually partially concealed because observed returns are smoothed.

· The average illiquid hedge fund Fung-Hsieh 8-Factor model alpha is lower than previously thought. The Fung-Hsieh 8-Factor model also explains a significantly higher fraction of the volatility of hedge fund returns than suggested by looking at observed returns (or at one-step unsmoothed returns)—the improvement coming from the three-step unsmoothing method stems from better measuring exposures to market risk and emerging market risk in the underlying illiquid assets held by hedge funds.

Average Sharpe Ratios

Hedge Fund Risk and Performance by Strategy: The figure plots average fund-level results by hedge fund strategy using observed returns, 1-step unsmoothed returns (as in Getmansky, Lo, and Makarov (2004)), and 3-step unsmoothed returns. The sample goes from January 1995 to December2017.

· Past alphas estimated using the three-step unsmoothed returns provided a better signal for future alphas than past alphas obtained from observed returns or one-step unsmoothed returns.

· There results for private commercial real estate (CRE) funds were about the same as for hedge funds. The three-step unsmoothed returns displayed little autocorrelation both at the fund level and aggregate-level. However, the degree to which the three-step unsmoothing process improved upon one-step unsmoothing was much higher given the extreme illiquidity of real estate assets. For example, the average beta (systematic risk exposure) of private CRE funds to the public CRE market increased from 0.07 to 0.34, driving the 4.3% annual alpha of private CRE funds (measured with observed returns) to 1.6% after three-step unsmoothing.

Their findings led Couts, Gonçalves, and Rossi to conclude that applying their new return unsmoothing method to hedge funds leads to substantially improved measurement of risk exposures and risk-adjusted performance relative to what is obtained from returns using traditional unsmoothing methods. The improvement was even more pronounced for CRE funds due to the high degree of illiquidity in their underlying assets.

Investor Takeaways

Couts, Goncalves, and Rossi contribute to the literature on the returns of funds that invest in illiquid assets, providing a more accurate picture of the risks and returns of such assets, allowing investors to make more informed asset allocation decisions. Importantly, they noted that their same unsmoothing methodology could be applied to other illiquid assets such as private equity, venture capital, private credit, and highly illiquid assets such as collectible stamps and art investments.

The takeaway then is to make sure that before you invest in any illiquid asset you have addressed the issue of autocorrelation of returns. The fund sponsor and/or your advisor should be able to provide the information.

Post-Script

In a related December 2023 research paper entitled “Cliffwater Corporate Lending Fund (CCLFX) NAV-timing Analysis,” (CCLFX has about $23 billion of assets) Spencer Couts examined the performance of the fund, which invests in floating-rate, private credit that is senior, secured and backed by private equity sponsors, to determine the degree of autocorrelation of returns. He found that while the fund’s returns are positively correlated with its own lagged returns as well as the lagged returns of other debt-based and market-based return series, leading to a degree of predictability in the CCLFX returns based on lagged CCLFX returns and lagged returns of other factors, the cross-correlations became insignificant (or even turned negative) beyond one month and was non-existent in quarterly returns (when redemptions can occur). He also found that even without the quarterly redemption restrictions, the cost of not being invested in the CCLFX was greater than any potential wealth transfer benefit associated with trying to implement a NAV-timing strategy—while the timing (switching) strategies generated alphas in the regressions, the returns were less than those of a buy-and-hold CCLFX investor. As to risk-adjusted returns, while CCLFX’s monthly and quarterly alphas decreased slightly when one-month lagged and two-month lagged variables were added, the alpha values were still economically and statistically significant.

The findings of a lack of any significant autocorrelation beyond one month should not be surprising because Cliffwater does provide daily pricing (to allow for daily purchases), there is virtually no duration risk since all loans are floating rates with either 30- or 90-day resets, and the loan quality is higher (the higher the loan quality, and the lower the duration risk, the lower will be the fund’s volatility, reducing the risk of autocorrelation in returns due to stale pricing).

Full Disclosure: I own shares of CCLFX.

Larry Swedroe is the author or co-author of 18 books on investing. His latest is Enrich Your Future.