Overview of Odd Thoughts Podcast — Episode with Cliff Asness
This episode (Odd Thoughts / OddLots) features Cliff Asness, co‑founder and CIO of AQR, discussing why markets feel "dumber" or less efficient over the last decade. Asness draws on decades of quant experience, his dissertation with Eugene Fama, AQR’s history (launching in 1998), and a long-form paper he wrote called “The Less Efficient Market Hypothesis.” He argues markets today show larger, more frequent bouts of disconnect from fundamentals — creating both bigger opportunities and greater behavioral/operational risks for investors.
Key takeaways
- Markets are showing larger episodic mispricings than in past decades — not permanently inefficient, but prone to bigger “bouts” of craziness.
- These bouts create bigger expected returns for investors who can identify and endure them — but they are harder to stick with because of long, painful drawdowns and client redemptions.
- Major structural drivers likely include: increased passive investing, social media / the loss of independent decision‑making (herding), greater retail & gamified participation (options, sports‑betting analogies), and prolonged low interest rates — though each is a conjecture, not statistical proof.
- AI/ML improves signal extraction (e.g., NLP on earnings calls) but comes with a loss of interpretability—trading off intuition for predictive power.
- Quant factors (value, momentum, etc.) still work empirically but their economic explanations remain contested (risk‑based vs behavioral) and can vary over regimes.
Main topics discussed
How Asness measures “less efficient”
- Primary quantitative measure: the spread between cheapest and most expensive stocks (value spread). Historically well‑behaved, it reached extreme levels during the dot‑com bubble and again around COVID (Asness joked about a “125th percentile”).
- Today the spread is elevated versus average (roughly the 77th percentile in his view) — attractive but not a bubble by his strict definition.
Why markets might be “less efficient”
- Passive investing: more capital not actively pricing individual securities reduces the marginal arbitrageur available to correct mispricings.
- Social media / network effects: when decision‑makers are linked (echo chambers, algorithmic reinforcement), crowd wisdom can flip into crowd madness; independence of votes matters.
- Retail & gamification: mass retail participation, zero‑day options and similar products amplify feedback loops and can transfer wealth to more sophisticated market players.
- Low interest rates: prolonged ZIRP may have loosened bounds of rational pricing (Asness estimates it explains only a small part, but likely contributory).
Behavioral & business implications for active managers
- The hardest part is not finding the misprice but sticking with it through long bad stretches; length of drawdown often hurts more (psychologically and via redemptions) than depth.
- Running client money is stressful — redemptions force shrinkage, layoffs or strategy changes even when the manager believes the approach remains sound.
- Markets are not pure arbitrage machines: moving price further toward “fair value” can be costly and risky; someone has to be willing to carry the trade.
AI and machine learning in investing
- ML/NLP can materially improve signals (e.g., parsing earnings calls via vectorized representations rather than naive word counts).
- Tradeoff: better predictive power but reduced interpretability — you may not be able to fully explain why a model signals what it does.
- Asness admits he initially slowed adoption at AQR because of the intuition loss, but accepts ML where it demonstrably helps.
Multi‑strategy / multi‑manager trends and talent
- AQR endorses multi‑strategy diversification (multistrat), but notes limits: alpha is finite and talent is the bottleneck.
- Multi‑manager firms and large multistrats have succeeded, but it’s unclear how scalable extra alpha is; selection and retention of talent is costly and competitive.
Sports betting / prediction markets
- AQR is exploring prediction markets and sports analytics (they have relevant academic and intellectual links), but Asness worries about social harms and the democratization/gamification effects.
Notable quotes and concise paraphrases
- “Markets are more susceptible to bouts of crazy than they used to be.”
- On crowds: when the crowd is independent the “wisdom of crowds” works; when the crowd is networked and amplified it can become “madness of crowds.”
- On AI: “If you’re going to use AI in your process, almost by definition you are going to lose a little intuition.”
- On drawdowns: length of a losing period (multi‑year pain) is often worse than a deeper but shorter drawdown.
Practical implications / recommended actions
- For allocators and investors:
- Expect episodic extremes; diversify strategies and consider time horizon: those who can endure long drawdowns capture larger opportunities.
- Beware gamified retail products and zero‑day options — they can amplify volatility and shift short‑term price dynamics.
- Don’t assume passive ownership will eliminate price discovery — it changes its character and can exacerbate mispricings.
- For quant/tech teams:
- Use ML where it demonstrably improves forecasting (e.g., NLP on text data) but pair it with process governance and post‑hoc interpretability checks when possible.
- Combine human intuition and ML outputs — accept some interpretability loss, but document and stress‑test models.
- For managers running client money:
- Prepare stakeholders for long pain periods; build client relationships that allow you to stick to processes when they’re out of favor.
Evidence, limits, and tone
- Asness’s arguments are mainly empirical and theoretical conjecture: he has long time series and firm experience but acknowledges you can’t “prove” causes with only a few bubble episodes.
- His paper “The Less Efficient Market Hypothesis” is the long form of these ideas — Asness frames them as informed hypotheses rather than definitive statistical laws.
Who should listen / value of the episode
- Useful for portfolio managers, quants, allocators, advanced retail investors and anyone interested in market structure, behavioral finance, and the practical effects of AI on investing.
- Offers both a veteran quant’s historical perspective (dot‑com, COVID) and present‑day diagnosis of market microstructure and social dynamics.
Concluding note: Asness’s view is nuanced — markets still have structure and predictable factor behavior, but social and structural changes (passive, social media, retail gamification, low rates) make occasional extreme mispricings larger and harder to arbitrage, creating a higher reward for endurance but much bigger behavioral and business risk.