MARKET EFFICIENCY

Market efficiency describes the degree to which asset prices reflect all available information at any given time.

The Three Forms of EMH

Eugene Fama formalized the efficient market hypothesis in 1970, distinguishing three forms based on what information is already priced in.

Weak form holds that current prices reflect all past trading data, meaning technical analysis cannot generate consistent excess returns. Semi-strong form holds that prices reflect all publicly available information, meaning neither technical nor fundamental analysis can produce persistent alpha. Strong form holds that prices reflect all information — public and private — making even insider knowledge unable to generate consistent excess returns.

The practical implication is direct: if markets are efficient, active management adds no value above index investing, and capital allocation becomes a question of cost minimization rather than security selection. If they are not, skilled analysis can generate alpha. The empirical record is complicated — most active managers underperform indices over long periods, but persistent anomalies exist, and markets clearly exhibit episodes of mispricing at scale (dot-com, 2008 housing).

For the AI Era

AI-driven trading and analysis are an interesting test case for market efficiency. If AI can process information faster and at greater scale than humans, it may push markets toward stronger efficiency — compressing the window between information release and price adjustment. But it may also create new instabilities: correlated strategies, flash crashes, and feedback loops that introduce volatility EMH doesn't account for. Whether AI is a force for efficiency or instability is an open empirical question with significant investment implications.

Where EMH Breaks Down Empirically

The anomalies that have survived rigorous out-of-sample testing tend to share a common feature: they require bearing risk or illiquidity that most investors rationally avoid. The value premium, the size premium, and momentum all have risk-based explanations that are difficult to definitively refute.

The more interesting challenge to EMH comes from behavioral finance: documented, systematic biases in how humans process information — overconfidence, anchoring, herding — produce predictable mispricings. The debate is not whether these biases exist but whether they are large enough and persistent enough to generate net-of-cost alpha after accounting for exploitation costs.

The Reflexivity Problem

Markets are social systems, not natural ones. When enough capital adopts a strategy based on an identified inefficiency, the strategy often arbitrages itself away. This is why market efficiency is better understood as a dynamic equilibrium than a static fact: efficient with respect to widely known, easily executable strategies, and inefficient at the frontier of new information, new analytical frameworks, and illiquid corners that well-resourced arbitrageurs haven't yet reached.

For investors, the implication is that the source of alpha is not superior information so much as superior willingness to bear specific risks that the market is currently mispricing.

The Practical Investor's Conclusion

The empirical evidence supports a nuanced position: markets are efficient enough that most active managers cannot consistently beat a passive index after fees, but not so efficient that all price discovery is instantaneous or that skilled fundamental analysis creates no value.

The implication is to default to passive for broad market exposure while concentrating active effort in markets, asset classes, or situations where information advantages are plausible — small-cap equities, private markets such as venture capital, credit, or situations requiring specialized domain expertise not yet reflected in consensus estimates. The durability of any such advantage often rests on identifiable economic moats that the market has not fully priced.

Open Questions

• Does passive investing's growth eventually undermine the price discovery that makes markets efficient in the first place, and if so, at what market share does this become material?
• Are behavioral anomalies (momentum, value, quality) genuine mispricings driven by cognitive bias, or risk factors with rational explanations — and can the distinction ever be settled empirically?
• Do AI-driven trading strategies converge on the same signals, producing correlated crowded trades that introduce systemic fragility rather than improving informational efficiency?

Part of the knowledge graph at The Best Blog Ever — reference definitions for ideas that matter.