How Risky Are Private Assets?

 

Patrick Warren and Luis O’Shea, PhD

Key Takeaways

  • When returns are calculated at short time horizons, volatility is biased toward zero because of valuation smoothing; this bias dissipates as return horizons increase, but estimates become noisier. We believe a six- to eight-quarter horizon offers the best tradeoff.

  • Risk-adjusted returns are critical for asset allocation decisions, but these require accurate volatility estimates. Using the approach described in our research brief, we find a sensible ordering for volatilities of the major private capital asset classes.

  • Using our preferred methodology to account for valuation smoothing, Venture Capital and Real Estate return volatility is double what naïve approaches would suggest.

Despite the importance of risk-adjusted returns for asset allocation, several aspects of private markets make estimating risk a perennial challenge. Unlike public markets, private markets operate at a relatively low frequency, have irregular cash flows and subjective valuations, and have data that is generally opaque. The Burgiss Manager Universe (BMU) can allay concerns around industry coverage and accuracy, with complete cash flows from over 11,000 funds; however, issues such as subjective GP valuations are unavoidable when working with private capital. While private market volatility is challenging to estimate, LPs need a practical sense of risk across asset classes.

There are more involved methods of estimating risk, but in this post, we lay out a simple, model-free way to measure volatility using the BMU. In both this post and our research brief, we address valuation smoothing, which understates volatility, especially at shorter time horizons. As time horizons increase, volatility grows while smoothing remains constant—thus volatility dominates the effects of smoothing at longer horizons. We analyze log returns to estimate annualized volatilities at a range of time horizons and graph the results below. Notably, the amount of smoothing differs across asset classes, resulting in some volatility curves being steeper in the short run—smoothing halves the reported one-quarter volatility of Venture Capital and Real Estate returns, hides a third of Buyout volatility, and understates Debt’s true volatility by almost 25%. While Real Estate and Debt look comparably volatile based on one-quarter returns, after accounting for smoothing, Real Estate is almost twice as volatile, highlighting the importance of evaluating private market volatility at a longer time horizon. In contrast, public equity valuations are not subject to smoothing, resulting in an almost flat line.

 
 

But what is the right horizon over which to calculate volatilities? At longer time horizons, we mitigate smoothing, but confidence intervals also become worryingly large. At shorter time horizons—five quarters or fewer—results are more subject to valuation smoothing and are thus understated. In our view, volatility estimates of around six to eight quarters offer the best tradeoff between accuracy and precision. We believe these results are qualitatively appropriate—i.e., Venture Capital is the most volatile, followed by a tight clustering of Real Estate, public equities, and Buyout, and then lastly Debt. As we discuss in our article, in this application it is critically important to calculate volatility using log returns, though the details are beyond the scope of this blog post. Lastly, even at appropriately long horizons, smoothing still modestly understates volatility, so it is prudent to add an extra 10-15% to our estimates at those horizons.[3]

In conclusion, while private market risk is difficult to evaluate, the quality of the Burgiss Manager Universe allows us to estimate the volatility of private capital accurately. However, there remains a tradeoff between accuracy and precision. Using short-term returns results in volatility estimates that seem precise (because we have many observations) but are in fact inaccurate because they ignore smoothing. At the other extreme, using very long-term returns (such as 5-year IRRs) eliminates most of the smoothing, but results in an imprecise volatility estimate. Somewhere in between is a return horizon that is a reasonable compromise between the two. We believe a return horizon of six to eight quarters best balances these competing concerns—this overcomes most of the smoothing while still allowing sufficiently precise estimates to rank the asset classes. Our confidence intervals should also serve as a guide when comparing the volatilities of private asset classes with their public counterparts.

[1] Based on our preferred time horizon of eight quarters, we add a 10% adjustment to account for residual smoothing to Venture, Buyout, and Real Estate; see the full article for further discussion.

[2] MSCI USA Index

[3] The background for this calculation is beyond the scope of this post, but the explanation is available in our research brief.

Additional Reading

Warren and O’Shea, “How Risky Are Private Assets?”

O’Shea and Jeet, “Estimating Public Market Exposure of Private Capital using Bayesian Inference”

 
 
Guest User