Defending the Efficient Market Hypothesis

The efficient market hypothesis (EMH) posits that market prices accurately reflect all available information. There are three forms of the EMH: weak, semi-strong, and strong. Each form supposes different sets of information that are fully reflected in market prices. The weak EMH information set consists of historical prices, the semi-strong form includes all publicly available information, and the strong form encompasses information privately held by individual investors or groups.

Essentially, no one believes that the strong form holds completely, and most economists believe that the weak form generally holds. Academic articles on the EMH center typically focus on whether the semi-strong form holds or to what degree the strong form holds. I believe that the semi-strong EMH holds and that private information can only allow for above-market return traders for a very limited amount of time.

Behavioral Economics

Attacks against the EMH founded on behavioral economics findings do not hold water. Behavioral economics aims to find systematic biases or errors in people’s behavior in markets. The problem with trying to assert the presence of systematic error driving market prices is that systematic error can be systematically exploited. One person’s bias is everyone else’s profit opportunity. Professional traders have a strong incentive to identify systematic errors and arbitrage opportunities, and therefore will not let systematic bias drive prices.

 Furthermore, findings in behavioral economics often struggle to withstand replication. In a special issue replicating famous behavioral economics studies, about one-third of the replications found similar effects, a third found significant effects but with dramatically smaller effect sizes, and a third did not replicate, which is only marginally better than the replicability rate in psychology. 

Conflicting Interpretations

A piece of information can have multiple interpretations, and any given interpretation can give rise to various judgments on proper action. How can information be fully reflected in market prices if that information leads to multiple competing interpretations and judgments? 

Conflicting interpretations are reflected by both the variation of interpretations and the strength of the belief. If I believe an asset is undervalued, I am inclined to buy it; if I believe it is overvalued, I am inclined to sell it. However, my willingness to get involved in the market will also depend on the strength of my conviction about the incorrectness of the valuation. 

Betting markets offer a suitable environment for testing how efficiently markets reflect available information, even when people have biases and conflicting interpretations of the same information. Brier scores indicate how well something predicts an event, and the closer a score is to 0, the better a predictor it is. Strong weather forecast systems receive Brier scores ranging from 0.05 to 0.12 for 12-hour rain forecasts. In contrast, Polymarket betting markets with a market cap of one million dollars or more have Brier scores of 0.0256 and 0.0159 12 hours and a day before, respectively. Although the Polymarket predictions don’t perfectly match the frequency of events, they are incredibly close. Some of the inherent errors may be due to the operation of Polymarket’s market-making algorithm, so it is possible that improvements in market-making could yield even better results. 

That is an incredible level of forecasting strength. That means that given the choice between listening to a cutting-edge forecast model or a sufficiently capitalized Polymarket betting line, you would be better off listening to Polymarket, despite the biases and conflicting interpretations of information by the bettors.

Beating the Market

Renaissance Technologies’ Medallion fund has consistently outperformed the S&P 500. Is this proof against the strong EMH? Possibly, but if so, not in any meaningful way. 

Over the past 1, 5, and 15 years, 72%, 86%, and 88% of large-cap investors have underperformed relative to the S&P 500, respectively. That means as time goes on, investors are more likely to fall behind the market than catch up and exceed market returns. That could mean that, given enough time, all firms fail to beat the market, but not necessarily. Firms like Renaissance Technologies have beaten the market by such a large margin and so consistently that it seems very unlikely that they are doing so randomly. 

However, an examination of how they beat the market offers more support for the strong EMH than it does against it. Renaissance Technologies employs a few hundred of the best PhD mathematicians, physicists, engineers, and computer scientists. They have created an algorithm that creates information about the market that they can trade on to make above-market returns. The information can only be profitably traded on for a few minutes.  

However, they must limit the size of their trades due to slippage. If their trades are too large, they work too far through the order book, which moves the price to an efficient level where they can no longer profitably trade on the stock. Essentially, Renaissance Technologies creates private information to beat the market, but that information makes the market more efficient by trading on it. 

If the only way to consistently beat the market requires getting hundreds of PhDs to write an algorithm that runs on supercomputers to find stock prices that are slightly mispriced for a few minutes, then the strong EMH is essentially true. It means that the blackboard version of the strong EMH is not perfectly descriptive of markets, but prices can only be made more efficient by the most cutting-edge algorithms and only for a few minutes.