Cliff Asness – The Past, The Present & Future of Quant [Invest Like the Best, EP.111]

Episode Summary

Executive Summary: Patrick O'Shaughnessy interviews Cliff Asness of AQR about the state of quant investing: factor efficacy, whether value is crowded or broken, how AQR balances research openness with proprietary work, machine learning’s role, fees, private markets, and the future of active management. The core message is that quant edges still exist but require realistic expectations, diversification, and discipline through bad cycles.

Main Topics: How institutions should use factors (Priority: 5/5): Asness argues for matching factor implementation to client constraints on leverage, shorting, and stickiness. Why bad factor years do not imply failure (Priority: 5/5): He says value’s drawdowns are explainable by spread compression or factor mix, not necessarily strategy decay. Crowding, capacity, and factor robustness (Priority: 5/5): He says too much money can matter, but the evidence does not yet show value or related factors have been arbitraged away. Research culture and sharing (Priority: 4/5): AQR distinguishes between widely known styles and proprietary alpha, while keeping a strong academic, open research process. Machine learning at AQR (Priority: 4/5): ML is treated as evolutionary and interpretable, used to find nonlinearities without abandoning economic intuition. Market timing, asset allocation, and private markets (Priority: 4/5): He extends factor logic to asset allocation cautiously, and sees private markets as partly a behavioral choice to avoid mark-to-market. The role of fees and active management (Priority: 3/5): He expects more indexing and lower fees, but not a world without active managers or public markets.

Key Arguments: Best use of factors depends on client constraints; the better strategy you can’t stick with isn’t better. Too much money can hurt factors, but crowding would make them look random, not simply bad. Value remains within historical valuation bands except the tech bubble; it is not obviously broken. AQR diversifies across many value measures and mostly within industries, improving robustness. ML should be evolutionary and interpretable, with economic explanations required at AQR. Three-sharp-ratio strategies are usually tiny or non-scalable; real capacity matters more. Public markets still have enough stocks; active managers remain necessary for price discovery. Private markets often smooth returns and help investors tolerate risk, but opacity is part of the appeal.

Data Points: AQR assets under management: $226 billion - Describes AQR’s scale 20 years after founding. AQR founding year: 1998 - Historical reference for the firm Cliff Asness founded. Initial Goldman/AQR-style start: 1994 - Asness says systematic value and momentum started at Goldman in late 1994. Value spread norm: three to six times - Historical band for expensive stocks relative to cheap stocks on price-to-book. Tech bubble value spread: 12 to 14 times - Late 1999/2000 extreme for price-to-book spread versus cheap stocks. Mutual-fund indexing share: mid 30% - Asness says mutual-fund indexed assets were in the mid-30s. Typical target Sharpe range: 0.5 to 1 - Range Asness says many real-world institutional strategies target. Bad periods at AQR: three tough periods - He says AQR has had about three major difficult stretches over 20 years. Time spacing of bad periods: about 10 years apart - He notes the tough periods occurred roughly a decade apart. Two sets of twins: 2 - He mentions raising four children born as two sets of twins. Children: 4 teenagers - Family discussion at the end of the interview. Relationship between children birth timing: a year and a half apart - He says the two sets of twins were born a year and a half apart. Stock return timing horizon in CAPE work: next 10 years - He references tests of CAPE versus subsequent 10-year equity returns. Tech bubble era valuation anomaly: late 99, 2000 - The only major exception to normal price-to-book spread bands.

Pivotal Quotes: "The better strategy you can't stick with is not the better strategy." — Cliff Asness: He explains that implementation must fit an investor’s behavior and constraints. "If it's in the data, write the paper." — Gene Fama (quoted by Cliff Asness): Asness recalls Fama’s response to his momentum research as an act of intellectual honesty. "We think you're the most attractive thing in our portfolio right now." — Unnamed client: A client said this while still redeeming because recovery was expected to take too long.

Implications: The unresolved question is not whether quant works, but how much capacity and complexity investors can truly tolerate; the next test is whether newer tools like ML add durable, interpretable edge.