The Stock Market
Part of the the investment sequenceIs the stock market efficient? The answer is challenging, and experts have strong, contradicting opinions Efficient-market hypothesis. In Wikipedia.
Discourse isn't helped by the fact that some people think the hypothesis is true by definition, while others think it's unfalsifiable, and therefore outside the realm of science.
What follows is my attempt to separate the wheat from the chaff.
Expert Opinion
Warren Buffet arguably stands as a strong counterexample against the EMH. He's argued that some investors predictably beat the market The Superinvestors of Graham-and-Doddsville and has historically done so himself by an average of more than 9% since 1964. However, what is less often touted is that his edge has shrunk dramatically over time and has disappeared since the early 2000s. Overall Berkshire Hathaway has outperformed the S&P 500 37 of 55 years; the odds of 37+ heads from 55 coin flips is 0.72% - rare, but certainly not implausible in a world of millions of investors. More recently, Buffet himself is skeptical that active management can beat broad based index funds, even making a bet with numerous hedge fund managers that they won't beat thee S&P 500 index - a bet he is currently winning.
Only a minority of actively managed mutual funds beat the S&P index Active management Liu - though critics point out half of invested dollars, by definition, must do worse than the market.
Roughly half of "mutual funds" are actually "closet indexers" - essentially index funds that are technically "actively managed" to justify higher fees Active management - though its unclear if this demonstrates experts think the stock market is efficient or that some mom and pop investors are just rubes.
The general idea that actively managed investments generally performs no better than passively managed ones is widely accepted by economists (absent insider information) Stock Prices.
This is all to say, it is widely believed that professional investors aren't worth their salaries, suggesting the semi-strong-form of the EMH is accurate.
The Literature
First, much of the literature is biased, because it ignores the role of risk. The EMH does not state that you can't predict price movements or expected real returns better than chance - it claims that you can't given some risk paradigm. The failure of many papers to do so makes them worthless.
What risk model to use? Well, you might say the Markowitz model, but there's a free-floating risk-aversion parameter in there with no true value, because people have genuinely different levels of risk-aversion (especially the same person as they age).
Therefore, to really demonstrate the EMH is false, you have to show that given two strategies with the same risk profile, the expected returns are not equal, or, equivalently, that two strategies have the same expected return but different risk profiles.
I'm going to ignore empirical work from before 1990, since it seems likely the EMH was false for stocks prior to that.
TODO
- 1973 - Malkiel
- 1992 - Fama
- 2003 - Chan
- 1995 - Overreaction, underreaction, and the low-P/E effect
- 2003 - Hirshleifer
- 2008 - Contrarian investment strategies: The next generation
- 2011 - A non-random walk down Wall Street
- 2017 - Marwala
One particularly interesting paper showed that stock price changes to federal reserve actions conform to the CAPM model Bernanke, which provides some evidence of efficiency.
My Analysis
I looked at 100 stock's daily returns from 1984 to 2017 Marjanovic and modeled each daily price move as
Y_it = B_0 + B_i1 + B_i2 * X_t + B_i3 * Norm()
where Y_t is stock i's daily return, X_t is the normalized change in the overall stock index, B_0 is the average index return, and the three B_i parameters represent alpha, beta, and idiosyncratic risk, respectively.
Of the 100 stocks, 4 rejected the null-hypothesis that B_i1=0. One would expect 5 to reject the null-hypothesis by chance alone, so this basically means I found no statistical evidence that any stock reliably beats the market.
This is actually puzzling because 89 of the stocks rejected the null-hypothesis that B_i2=1, which means that many of these stocks did have different levels of risk. Moreover, B_i3 correlated with B_i2, which means stocks with high systematic risk (beta) also tended to have high idiosyncratic risk.
This raises the question: why does anyone invest in high-beta stocks if there's no evidence they compensate their owners with higher returns or even lower idiosyncratic risk?