Cliff Asness: AQR diversification and portfolio resilience

Cliff Asness's AQR diversification research combines value, momentum, carry and defensive styles; its January 2026 assumptions put a global 60/40 portfolio's expected real return at 3.4% a year over 5 to 10 years. The research tests how much each return source depends on the same market risks.

Key takeaways

1. AQR's core idea is to combine return sources that do not fall together. In AQR's 2015 study, the four styles had correlations with global equities between -0.31 and 0.20 over 1990-2013, and value and momentum were -0.64 correlated with each other.
2. The combined numbers are simulations, not fund returns. The four-style composite showed a gross Sharpe ratio of 1.74, which the authors themselves said is "unlikely to be achievable in practice".
3. AQR's risk parity depends on leverage. Unlevered, its simple 2012 risk parity portfolio averaged 15% stocks and 85% bonds; leverage then lifted its risk to match the market's.
4. AQR expects modest market returns. Its 2026 estimates put US large caps at 3.9% a year after inflation, against 5.1% for emerging-market equities and 1.3% for US cash.
5. A higher Sharpe ratio is an intermediate measure. Copio evaluates whether the full allocation maximises expected geometric growth after costs; factor diversification and crash insurance have different roles.

What is AQR's investment strategy?

AQR's research builds portfolios from systematic, rules-based sources of return that it expects to be rewarded over time and to diversify each other. AQR, short for Applied Quantitative Research, was founded in 1998 by Cliff Asness, David Kabiller, Robert Krail and John Liew (AQR, About us). AQR lists Asness as Managing and Founding Principal.

The most recent firm-wide figure we found is in a third-party fund filing. A September 2026 prospectus filed with the SEC states that AQR and its affiliates had approximately $237.8 billion in assets under management at 31 May 2026 (Unified Series Trust, 485BPOS).

Returns come from a small number of styles, applied in many markets at once. AQR's risk-parity research balances estimated risk contributions rather than capital weights. And a long-short implementation of styles needs what the authors call the "three dirty words in finance": leverage, short-selling and derivatives (Investing with Style, 2015, p. 29).

What did "Value and Momentum Everywhere" find?

Asness, Tobias Moskowitz and Lasse Heje Pedersen reported positive historical value and momentum premia across the eight markets and asset classes they studied (Journal of Finance, June 2013). The sample covered stocks in the US, UK, continental Europe and Japan, country equity index futures, government bonds, currencies and commodity futures, with data ending in July 2011.

Value and momentum were negatively correlated, averaging about -0.60 within stock markets, so a 50/50 mix produced steadier returns than either strategy alone.

Annualised Sharpe ratios of high-minus-low portfolios to July 2011, gross of costs (Table I, pp. 940-943). Japan shows the limit of the idea: momentum barely worked there, and on this measure the mix only matched value (0.78 against 0.79), though the paper's signal-weighted version beat it (0.88 against 0.77). Government bonds show weak results for both.

The companion paper Fact, Fiction, and Momentum Investing (2014) answers ten common objections, including the claims that momentum only works on the short side, only among small caps, or does not survive trading costs. It cites US equity momentum evidence going back to 1801.

What happens when value, momentum, carry and defensive are combined?

Combining the four styles doubled the best single style's simulated risk-adjusted return over 1990-2013 and cut the maximum drawdown to 15%. In Investing with Style (2015), Asness, Antti Ilmanen, Ronen Israel and Moskowitz define carry as favouring higher-yielding assets over lower-yielding ones, and defensive as favouring low-risk, low-beta assets.

Simulated Sharpe ratios, January 1990 to June 2013: value 0.29; momentum 0.83; carry 0.87; defensive 0.58; all four combined 1.74. Source: Investing with Style, Tables 2 and 6, pp. 40 and 46; chart data (CSV). These are gross long-short backtests scaled to 10% volatility, before trading costs and fees. They are not the returns of any AQR fund.

Each style alone suffered a deep loss at some point, but those losses did not arrive together. The same paper notes that a 60/40 stock-bond portfolio was 0.99 correlated with global equities over the period, which is the problem the styles are meant to address. For the basics of correlation and the Sharpe ratio, see what diversification measures and our Sharpe ratio guide.

What is AQR's version of risk parity?

Risk parity gives stocks and bonds a similar share of portfolio risk, which means holding far more bonds than stocks and then using leverage to reach a useful level of return. In Leverage Aversion and Risk Parity (Financial Analysts Journal, January/February 2012), Asness, Andrea Frazzini and Pedersen argue that many investors cannot or will not borrow, so they overpay for risky assets, leaving safer assets with better returns per unit of risk.

Their simple version weighted each asset by the inverse of its volatility. Over 1926-2010 it averaged 15% stocks and 85% bonds before leverage. Scaled to the market's 15.08% volatility, it showed a Sharpe ratio of 0.53 against 0.40 for 60/40 and 0.25 for the value-weighted market portfolio (Table 2, p. 53).

The levered results do not deduct financing costs or the cost of forced deleveraging, which the authors say could be significant at high leverage. And the unlevered portfolio, the version that needs no borrowing, returned 2.20% a year over cash against 4.65% for 60/40. For a related approach to balancing economic risks, see Ray Dalio's All Weather portfolio.

Is value investing dead?

Asness's answer in May 2020 was "No it's not. Quite the opposite." (Is (Systematic) Value Investing Dead?). He argued that value needs investors to overreact, not stagnant technology, and that the spread between expensive and cheap stocks was extreme even after removing technology stocks or the largest companies.

In March 2021 he accepted that the value strategy's 2018-2020 period was "essentially a crash" (The Long Run Is Lying to You). His point was that past returns are distorted by valuation changes: a strategy that became cheaper looks worse in hindsight, while its expected return going forward rises.

AQR's 2026 capital market assumptions say the equity value style looked extremely cheap in the early 2020s, that spreads gradually normalised during 2023-24, and that they no longer point to an exceptional tactical opportunity (2026 assumptions, p. 8). Joel Greenblatt's Gotham fund shows how hard a lagging value strategy can be to hold; see the magic formula and Gotham's portfolio.

What returns does AQR expect from asset classes in 2026?

AQR's 2026 capital market assumptions, published on 14 January 2026, expect a global 60/40 portfolio to return 3.4% a year after inflation over the next 5 to 10 years, against 3.5% in the January 2025 edition. The estimates are dated 31 December 2025 and built mainly from current yields and valuations.

Expected local real returns, as of 31 December 2025: emerging-market equities 5.1%; non-US developed equities 4.9%; commodities 4.3%; US buyouts 4.2% (as of 30 September 2025); US large caps 3.9%; global 60/40 3.4%; US investment-grade credit 2.8%; US private credit 2.6%; US 10-year Treasuries 2.4%; US cash 1.3%.

Source: AQR, 2026 Capital Market Assumptions, Exhibit 1, pp. 7 and 10, Appendix Exhibits A1-A6; chart data (CSV). AQR stresses the uncertainty: there is a 50% chance that realised equity returns over 10 years miss its estimate by more than 3 percentage points a year. Local real returns are not directly comparable across currencies for a single investor.

US large caps have the lowest expected return among major equity markets in AQR's framework, with the US cyclically adjusted price-earnings ratio near 40, at the 96th percentile since 1980. For a European investor, AQR adds that a cap-weighted global 60/40 portfolio effectively holds a 40% position in the US dollar at the start of 2026.

For alternatives, AQR states assumptions rather than yield-based estimates. It assumes a diversified long-short multi-factor strategy could reach an expected Sharpe ratio of 0.7 after costs and fees, or 7% a year over cash at 10% volatility, against 0.2-0.3 for one style in one asset class.

On tail hedging, AQR researchers favour trend following over buying puts on cost grounds.

In Tail Risk Hedging: Contrasting Put and Trend Strategies (February 2021), Ilmanen, Ashwin Thapar, Harsha Tummala and Dan Villalon, not Asness, found that rolling 5% out-of-the-money one-month index puts lost money over January 1985 to March 2020, while simulated trend was up in nine of the ten worst S&P 500 months, October 1987 excepted.

They conclude that "the long-term cost argument tips the scales in favor of Trend, unless investors are explicitly focused on hedging fast market drawdowns." For a different view, see Mark Spitznagel's tail-hedging allocation; our put option guide checks a published 1986-2018 benchmark in which an S&P 500 portfolio protected with monthly puts about 5% out of the money reduced the maximum month-end drawdown but compounded at 6.64% a year, against 9.80% for the index.

Does a higher Sharpe ratio maximise geometric growth?

A higher Sharpe ratio does not by itself establish the highest expected geometric growth: position size, financing, costs and the loss distribution still matter. AQR's 2015 study measures diversification with simulated Sharpe ratios, and its 2026 assumptions describe long-short premia through expected Sharpe ratios at a chosen volatility. AQR also explicitly publishes annual compound-return estimates for major asset classes (2026 assumptions, PDF pages 3 and 8).

Copio's research objective is to maximise expected long-term geometric growth within the investor's constraints. Factor diversification can help by reducing shared losses, but it does not create a contractual crash payout. We believe a growth portfolio with cost-controlled convex protection offers another route: its payouts and reinvestment can support compounding even when the hedge loses money in most ordinary periods.

AQR's critique of recurring put costs is part of that test; a hedge fails it when the cost outweighs the expected compounding benefit. Our resilient portfolio allocation guide explains the difference between reducing variance and improving geometric growth.

What can an individual investor learn from Cliff Asness?

AQR's institutional strategies use leverage, shorting and derivatives. Its research also raises questions about simpler portfolios:

• If my stock and bond funds fell in the same month, what else in the portfolio would behave differently?
• Do my funds labelled "value", "quality" or "momentum" hold different positions, or the same large companies in different proportions?
• If a strategy I hold had three bad years, as value did in 2018-2020, would I know whether its price had become cheaper or its idea had stopped working?
• Am I judging an asset by its past ten-year return, or by what its current yield and valuation imply?
• Does any part of my plan rely on leverage, shorting or derivatives that I do not fully understand?
• Is my currency exposure a choice, or a side effect of a global index?