Cliff Asness on How Markets Got Dumber in the Last 10 Years

Episode Summary

Executive Summary: The episode centers on AQR co-founder Cliff Asness arguing that markets have become meaningfully less efficient, driven by retail trading, passive investing, social media, gamification, and AI-era dynamics. He explains why prolonged valuation extremes make both value and momentum more compelling yet harder to trade, and why drawdown length can be more painful than depth. The conversation also covers AI in quant investing, multi-strat models, and emerging opportunities in sports/prediction markets.

Main Topics: The 'less efficient market hypothesis' (Priority: 5/5): Asness argues markets now experience larger, longer bouts of mispricing than in the past, especially around expensive-vs-cheap stock spreads and bubble regimes. Behavioral drivers: retail, passive, and social media (Priority: 5/5): He attributes greater market dislocation to more retail participation, the rise of passive investing, and social-media-driven herd behavior that weakens crowd independence. Value vs. momentum and the persistence of bubbles (Priority: 4/5): The discussion revisits why value and momentum work, why they can coexist, and how extreme periods like the dot-com era and COVID amplified valuation distortions. Pain of long drawdowns in client money management (Priority: 4/5): Asness emphasizes that prolonged underperformance is often more damaging than deeper but shorter losses because clients and employees react to duration as much as magnitude. AI, machine learning, and interpretability (Priority: 4/5): He describes AQR’s use of NLP/ML to improve signal extraction from earnings calls while acknowledging the loss of intuition and explainability compared with traditional quant methods. Multi-strat investing and the limits of talent (Priority: 3/5): Asness contrasts AQR’s internally built multi-strategy approach with multi-manager platforms, noting diversification benefits but also constraints from scarce alpha and talent. Sports betting and prediction markets (Priority: 3/5): The conversation closes by linking sports analytics to market inefficiency, with Asness suggesting betting markets may still be mispriced and increasingly similar to retail gamification.

Key Arguments: Markets are not perfectly efficient, and in recent years they have become more susceptible to extreme mispricings and bubbles. The cheapest-vs-most-expensive stock spread became wider than any period in 50+ years during the dot-com bubble and again during COVID, supporting the idea of repeatable regime shifts. Retail investors tend to lose on average, so their larger market presence can amplify dislocations even if individuals sometimes win. Passive investing improves investor welfare overall, but a market cannot be fully passive because prices still require active price discovery. Social media reduces the independence of decision-making, turning the 'wisdom of crowds' into herd behavior and making markets more prone to extremes. Long drawdowns are often more painful than larger but shorter losses because investors and clients experience psychological fatigue and business pressure over time. AI/ML can improve predictive power in text analysis and other signals, but it often sacrifices intuition and interpretability relative to older quant approaches. Quant factors like value and momentum do not have a single accepted explanation; both rational risk-based and behavioral stories may be partly true. Sports betting/prediction markets resemble other gamified financial products and likely contain persistent inefficiencies, but Asness is cautious about overstating the edge.

Data Points: Odd Lots anniversary: 10 years - The hosts frame the episode as part of their 10-year anniversary celebration. Historical valuation extreme: 125th percentile - Asness jokes that the cheap-vs-expensive spread during COVID exceeded the previous maximum, which he calls a '125th percentile'. Prior extreme valuation period: 50+ years - He says the dot-com valuation spread was the craziest in more than half a century of data. Market comparison: 77th percentile - He says the current cheap-vs-expensive spread is wider than average but far from bubble territory. AQR launch year: 1998 - He references AQR's founding after a strong run at Goldman Sachs. Value-market stress periods: Two major episodes - He highlights the dot-com bubble and the COVID era as major times when value factors were punished and then recovered. Down period at AQR: About 3 years - He says AQR fell by roughly half over about three years during a painful stretch for investors. Value recovery characteristic: Round trip - He notes that in extreme bubbles the firm has often lost and then later made back the losses. Modeling horizon: 50 years - He repeatedly references 50+ years of data when discussing spreads and empirical patterns. Sports betting example: 5 to 6 minutes left - In the hockey goalie example, Asness says the optimal pull point is when trailing by one with five or six minutes left.

Pivotal Quotes: "Markets are not arbitrage mechanisms. They are voting mechanisms." — Cliff Asness: He explains why mispricings can persist when many participants share the same view and arbitrage is costly or risky. "The bigger disconnect from reality is a two-edged sword: it's a bigger opportunity for people who can stick with it, and it's harder to do." — Cliff Asness: He describes how extreme mispricing can improve expected returns while increasing the difficulty of holding positions through pain. "You have to shrink when you lose money for a while, and you grow when you make money for a while. It's the ironclad rule of this business, and sometimes it's just backwards." — Cliff Asness: He captures the business consequences of performance cycles for asset managers.

Implications: Listeners should expect more persistent market dislocations, greater importance of behavioral effects, and growing use of AI in investing. For managers, the challenge is less spotting mispricing than surviving long enough to monetize it.