Cliff Asness: Timely & Timeless Investment Wisdom | #528

Episode Summary

Executive Summary: Cliff Asness argues for broad diversification across countries, asset classes, and strategies, emphasizing that investors overweight recent winners like U.S. equities and bonds at their peril. He stresses long-term discipline over short-term performance-chasing, explains why alternative trend and fundamental momentum can complement price trend, and says AI/ML are useful but evolutionary rather than revolutionary in investing.

Main Topics: Diversification across assets and geographies (Priority: 5/5): Asness makes the case that concentrating in U.S. equities or any single winner is a backward-looking mistake; global diversification, bonds, and alternative strategies improve portfolio robustness. Short-term expectations vs long-term investing (Priority: 5/5): The conversation repeatedly returns to how investors mistakenly infer too much from 1-3 year results, causing them to buy high and sell low despite strategies needing long horizons. Bonds, nominal vs real returns, and portfolio-level thinking (Priority: 4/5): They discuss why bond drawdowns should be evaluated in the context of the whole portfolio and why nominal 'getting your money back' is not the key issue. Alternative trend following and fundamental momentum (Priority: 5/5): Asness explains AQR’s expansion beyond pure price trend into esoteric markets and fundamental signals such as earnings revisions, balance-of-payments data, and other economic trends. Machine learning and AI in investing (Priority: 4/5): He says AI/ML can improve research and factor allocation, but mostly as better statistics and higher computing power rather than a total transformation of asset management. Market efficiency and investor behavior (Priority: 4/5): Asness offers the contrarian view that markets may be less efficient now due to information overload, social media, and crowd behavior, creating both more opportunity and more volatility.

Key Arguments: Levering a diversified portfolio can be superior to concentrating in one 'best' asset; Buffett’s success does not invalidate diversification. Investors should evaluate the whole portfolio, not individual line items like bonds in isolation. The U.S. equity market’s long outperformance is partly valuation rerating ('richening'), which is unlikely to repeat at the same magnitude. Three years is often too short to learn anything meaningful about a strategy’s true edge; patience is essential. Alternative trend signals based on fundamentals can complement price trend because they are related but not redundant. AI/ML in investing is useful when constrained to known factor sets, but it is not magical and can easily overfit. A strategy’s practical value depends on whether clients can stick with it through drawdowns; the 'best' theoretical allocation is not optimal if behavior breaks it. Market efficiency may have declined over time because cheap, fast information can amplify overconfidence and crowding rather than wisdom.

Data Points: AQR age: 25 years - Asness notes AQR is now 25 years old when discussing the firm’s '20 for 20' summary piece. Sample period mentioned for U.S. outperformance: 30 years - Used to argue that U.S. stocks outperforming the world for decades does not justify assuming it will continue. U.S. valuation starting point: Two-thirds the price of non-U.S. stocks - Asness cites this as an approximate starting relative valuation in his paper, before U.S. stocks rerated. Current relative valuation: 1.5x the price of non-U.S. stocks - He describes the U.S. becoming much more expensive versus non-U.S. equities over time. Correlation between price trend and fundamental momentum: 0.5-0.6 correlated - He says this is the sweet spot: related enough to be real, but different enough to add diversification. Average hedge fund correlation to the S&P 500: About 0.8 - Used to argue that many 'hedge funds' are not very diversifying versus equities. Typical planning horizon that clients use: 1, 3, 6 months - Meb describes how many allocators judge a new strategy over very short windows. Historical portfolio drawdown reference: Value drawdown from 2018-2020 - Asness mentions a period in which AQR’s business shrank after a value drawdown. Market event example: S&P down about 7% in a day - Asness recalls October 1997 during the Asian debt crisis. Historical bond return context: 2% on a long-term TIPS - He says bonds looked much more reasonable once real yields moved back up to roughly this level. Factor/strategy performance benchmark: 0.35-0.4 - He compares the stock market’s risk-adjusted advantage over cash to this range. Hypothetical strategy edge: 0.5 information ratio / Sharpe-like advantage - Asness says even a modest edge can still underperform for a decade without being implausible. Optimization output example: 83% - He uses this as an example of an unrealistic solver output for allocating to uncorrelated alts. Alternative allocation upper bound for many institutions: 20s percent - He says the 99th percentile institutional allocation to true alts is likely in the 20% range. Own flagship trend allocation: Darn near half - Asness says AQR’s flagship trend product is close to 50%, which he calls unusual/crazy. Data span of dissertation: Out-of-sample longer than in-sample - He notes this humorous sign that he had been in the field a long time.

Pivotal Quotes: "We don't think AI, at least in our field, is as revolutionary as others do." — Cliff Asness: Introduced near the start as the quote that frames the conversation on AI and quantitative investing. "The long run is lying to you." — Cliff Asness: Referenced while arguing that recent long-term U.S. outperformance should not be extrapolated blindly. "The time horizon that matters for statistical inference is dramatically longer than the time horizon that matters in the real world." — Cliff Asness: Used to explain why investors overreact to short-term performance and abandon good strategies too soon.

Implications: Listeners are urged to diversify globally, judge investments on portfolio context, and set longer evaluation periods for strategies. For the industry, the message is that modest but persistent edges matter, and AI/ML should be used carefully as an enhancer, not a miracle cure.