Private assets—including private equity, private real estate, and private credit—represent an increasingly large and growing portion of the global investment universe. As their popularity continues to rise, many clients are asking us for our views on the potential benefits of incorporating private assets in diversified portfolios.

Private assets often have unique characteristics that differentiate them from publicly traded investments, creating potential opportunities to improve the overall risk/return profile of a multi-asset portfolio. However, combining public and private assets in a portfolio can present a difficult challenge for portfolio construction—particularly when it comes to modeling the appropriate allocation weights for both categories.

To assist asset allocators in this effort, T. Rowe Price has developed a comprehensive process for evaluating private assets alongside public investments, one that can be used consistently across portfolio construction projects. Using this framework, we analyzed the historical performance of three key private asset categories, taking into account the significant differences between the way private and public asset indexes are constructed, how their returns are calculated, and how frequently those returns are reported.¹

Based on this analysis, we examined the impact of incorporating a 20% allocation to private assets in a hypothetical 60/40 U.S. public stock/bond portfolio. We found that while private assets would have improved the portfolio’s risk/return profile, the optimal allocations were extremely sensitive to the assumptions used—including the assumed ability of private asset managers to generate positive excess returns. Under these conditions, manager selection becomes an especially critical step in the portfolio construction process.

Comparing private and public assets

Historically, many private asset categories offered attractive return potential, driven by factors such as illiquidity premia, deal-sourcing advantages, and lower market transparency. Reported volatilities in many private asset categories also were lower than for their public market counterparts, and their low reported equity correlations appeared to create attractive diversification opportunities.

However, unlike indexes for public assets—which typically only include investible securities and have transparent guidelines for index construction—private asset indexes represent an aggregation of active manager results (often self-reported), with meaningful dispersion between the highest and lowest performers. This dynamic can make it difficult to separate alpha from beta.

Additionally, reported returns on private assets also are usually cash flow-based (e.g., represent internal rates of return), require appraisals of the underlying investment values (not market prices), and typically are only reported quarterly.

Smoothed, appraisal-based valuations and a lack of mark-to-market pricing can produce reported performance results that are not comparable with public assets. Some investors welcome this smoothing effect as it reduces apparent maximum drawdowns and can mitigate negative behavioral reactions. However, from a multi-asset perspective, the smoothed results do not accurately represent the actual volatility and correlation characteristics of private investments. We find evidence for this in several places.

• Analysis over longer holding periods washes away the impact of smoothing and increases comparability. Private-equity volatility, for example, was comparable with U.S. large-cap equity over rolling one-year periods and higher over rolling three-year periods (Figure 1).
• Many private investments can be sold on the secondary market, and these transactions may provide market-based prices that are more reflective of their intrinsic value. This can be observed in the discounts to net asset value (NAV) required to sell private investments in secondary markets, particularly in periods of market stress.
• During the 2008 global financial crisis, some private-equity investors were forced to sell assets at up to a 60% discount to NAVs, which for the most part had not been marked down. While this may have been due in part to lower secondary market liquidity, it does suggest that the intrinsic value of these assets was lower than their reported value.

Observed private asset volatility consistently appears artificially low when compared with similar public assets. To directly compare the two, then, we need to adjust reported returns to estimate the “true” risk of private assets. We can do this by taking into account the artificial smoothness of appraisal-based prices. Without such adjustments, private assets will tend to dominate multi-asset portfolios in a traditional mean-variance optimization exercise.

Adjusting risk estimates on private assets

Given that private assets rely on appraisal-based valuations while comparable public assets typically are marked to market daily, price movements for public assets tend to be reflected in private assets but only with a time lag—an impact that can persist over several quarters and that also tends to smooth reported returns.

This tendency can create high levels of autocorrelation in private asset returns, meaning there is a strong positive relationship between the return in the current quarter and the return in the prior quarter and a likelihood of a similarly strong relationship in subsequent quarters (Figure 2).

While smoothed returns and lower reported risk can make private assets look highly attractive on first impression, we think it’s prudent to adjust for smoothness and autocorrelation to arrive at their “true” risk characteristics—i.e., estimates that are more comparable with public assets. One common technique adjusts each quarter’s reported return for a private asset to reflect a weighted average of the “true,” unobserved return in that quarter and the return in the prior quarter.³

While this technique adjusts for much of the smoothing in reported volatility, we prefer an even more robust approach that unsmooths returns based on optimal lag lengths, rather than a single lagging period. There are several reasons for this.

• Because of the appraisal-based nature of many private asset valuations, the impact of autocorrelation can persist for more than just one quarter. Private real estate indexes, for example, are only required to update property values on an annual basis. This produces an optimal lag length of four quarters for that asset class (Figure 3).
• Other private asset classes also have their own intrinsic characteristics and therefore different lag structures, which we think should be taken into account.
• By unsmoothing private asset return data with our more targeted approach, we think we can obtain more accurate measures of true economic risk. Once we do this, we typically find that the volatility of private assets is roughly comparable with their public asset counterparts (Figure 4).

Impact of return unsmoothing

Although our analysis confirmed that incorporating private assets can improve diversification in multi-asset portfolios, those benefits were significantly lower when the reported risk characteristics were unsmoothed.⁴ To illustrate this effect, we looked at the impact of adding a 20% allocation to three private asset classes—private equity, private credit, and private real estate—to a hypothetical 60/40 U.S. public stock/bond portfolio. Based on the historical performances of the asset classes involved, we then derived risk and return outcomes for the hypothetical combined public/private portfolios, using as inputs both reported private asset volatilities and our own unsmoothed adjustments.⁵

Figure 5 illustrates the impact on hypothetical portfolio performance. The line represents the optimal risk and return frontier for a set of hypothetical portfolios with differing public equity/public fixed income weights. The dots each show the results of adding a 20% allocation to each private asset class to the 60/40 hypothetical portfolio.

Based on reported volatility, inclusion of private assets appeared to meaningfully reduce overall risk in our hypothetical portfolios. All three private asset classes moved the risk/return trade-off toward the upper left-hand corner of the chart—a clear improvement in portfolio efficiency, as measured by the Sharpe ratio.

However, when we replaced the reported volatilities with our unsmoothed estimates, the diversification benefits were reduced. This was largely due to the higher volatilities and equity correlations that unsmoothing revealed for both private equity and private real estate.

These key portfolio metrics are summarized in the table in Figure 6.

Sensitivity analysis

As discussed above, we believe that incorporating private assets in multi-asset portfolios can enhance diversification and improve both absolute and risk-adjusted return potential—provided that the private asset managers employed are skilled enough to take advantage of that potential.

The required return that the manager of a private asset strategy would need to generate to justify inclusion in a multi-asset portfolio depends on the strategy’s expected risk (higher volatility implies a higher required return) and its correlation to equities (higher correlation also implies a higher required return).

Figure 7 illustrates our estimates of the returns that were needed for a private asset to improve risk-adjusted results for our hypothetical 60/40 portfolio over the historical period we examined. Not surprisingly, if the expected risk and/or equity correlation was relatively low for a particular private asset strategy, so was the return hurdle needed to justify its inclusion in the hypothetical portfolio, given the potential reduction in overall risk.

For example, if we consider a private asset strategy with 12% expected annualized volatility and a 0.5 correlation to U.S. large-cap equities, the strategy only needed to clear a hurdle rate of return no more than 2.2 percentage points below the total equity return in order to improve risk-adjusted returns in the hypothetical portfolio.

Because even higher-risk private assets can offer at least some diversification benefits as long as their equity correlations are less than one, our analysis found that the rates of return required to justify inclusion in a hypothetical 60/40 portfolio were lower than what skilled managers historically have been able to deliver in the private asset space.⁶

However, we want to stress again that the forward-looking potential for private assets to enhance portfolio diversification is sensitive to the risk and return assumptions incorporated in an optimization analysis. Because the universe of private assets is extremely heterogeneous, the appropriate allocation also can depend heavily on the style and expected performance characteristics of the specific strategies under consideration—not least the active skill of the managers used.

Appendix

Autocorrelation methodology⁷

The econometric model used in our process for adjusting reported private asset volatility assumes that the observed (i.e., reported) returns can be modeled as a moving average⁸ of the recent history of actual⁹ but unobserved returns given by the following equation:

Where r_obs,t is the observed return, r_t is the unobserved actual return, Q is number of lags and ω _j are weights reflecting the impact of past realized returns on current, observed returns. Further constraints are applied, so that

Thus, a maximum likelihood estimation on observed returns gives an estimate of the ω _j and the appropriate number of lags (Q) is selected based on their statistical significance.¹⁰

Since the returns of any asset may be expressed as a linear combination of risk factor returns:

this can be re-expressed as

where N is the number of risk factors. These components are estimated using linear regression and, because the error terms are auto-correlated, Newey-West correction is applied to assess the statistical significance of the coefficients (β _i )

Important information regarding hypothetical results

Hypothetical results: The information provided above reflects data for hypothetical portfolios based on the theoretical blending of the indicated benchmarks. It does not reflect the actual returns of any portfolio or strategy. For the applicable hypothetical portfolios, the assumption of constant benchmark weights has been made for modeling purposes and is unlikely to be realized. Results shown for blended portfolios are hypothetical, do not reflect actual investment results, and are not a guarantee of future results. Hypothetical results were developed with the benefit of hindsight and have inherent limitations. Hypothetical results do not reflect actual trading or the effect of material economic and market factors on the decision‑making process. Results do not include management fees, advisory fees, trading costs, and other related fees. Results have been adjusted to reflect the reinvestment of dividend and capital gains. Actual returns may differ significantly from the results shown above. It is not possible to invest in an index. Different time periods would yield different results.