Why Nobel Economists Disagree About Market Efficiency

Eugene Fama, Lars Peter Hansen and Robert Shiller shared the 2013 economics prize "for their empirical analysis of asset prices". Fama and Shiller accept the same two findings about stock returns. They explain the second one in different ways.

A label such as "efficient" against "not efficient" hides how much the two accept in common. This guide sets out the shared findings first, then the main question they answer differently. It uses the committee's 2013 prize documents and the written lectures of Fama and Shiller, who read the evidence differently. Hansen's prize work is on statistical methods and is not covered. Nothing after 2013 is covered, so every statement below is the position as it stood then.

Fama's prize lecture states the central question as "whether asset prices reflect all available information". That proposition is called the efficient markets hypothesis.

It cannot be tested alone. A test needs a model of the returns investors should expect. Those include a risk premium, the extra return an asset must offer over a risk-free asset to compensate for its risk.

Fama's lecture states the consequence: "If the tests reject, we don't know whether the problem is an inefficient market or a bad model of market equilibrium." This is the joint hypothesis problem. It is why one pattern in returns can be read in two ways.

The committee reports that Fama's work in the 1960s found stock returns very hard to predict from past returns. It sums up the later research this way: the predictability is so low that "hardly any unexploited trading profits remain, once transaction costs are taken into account".

Shiller's lecture agrees: "we know it cannot be easy to make a lot of money trading".

The two differ on how much this proves. For Fama, early tests were evidence of efficiency. For Shiller, they "tell us only the obvious" and say little about whether markets are rational.

The committee credits Shiller's work with the finding that at a horizon of a few years "the overall market is quite predictable".

On average, the market tended to fall after periods when prices were high relative to dividends. It tended to rise after periods when they were low. Fama's lecture accepts the pattern: "stock returns are somewhat predictable from dividend yields and interest rates".

Three terms are needed first. The dividend-price ratio is the dividends paid over the past year divided by the share price. An excess return is the return above that of a risk-free asset. R-squared is the share of the variation in returns that a predictor accounts for.

The chart reproduces a table in the committee's document, taken from Cochrane (2001). The dividend-price ratio accounts for 15% of the variation in one-year excess returns, 23% at two years, 37% at three years and 60% at five years. The table does not give its sample period. The committee's footnote warns of statistical problems, including small-sample bias.

Fama adds a caution about the rise with the horizon. He writes that it is "just a consequence" of persistent dividend yields. Forecast errors also grow with the horizon, so "in this more important sense, longer horizon returns are less predictable".

Fama's lecture states his reading directly: "Efficient market types (like me) judge that predictable variation in expected returns on stocks and bonds is rational, the result of variation in risk or willingness to bear risk."

He cites Fama and French (1989). Expected returns on stocks and bonds move together and track business conditions. The lecture summarises it: "expected returns are high when business conditions are poor and low when they are strong". A high price relative to dividends, he explains, "likely signals a lower expected return".

He adds an empirical point. He reads claims about bubbles as claims that a price rise implies a predictable fall. His lecture says "the available research provides no reliable evidence that stock market price declines are ever predictable".

The two claims fit together. A lower expected return is not a predictable fall: a high price signals weaker returns ahead, not a decline anyone can forecast.

He also sets a test for the alternative. The behavioural literature, he writes, "has not put forth a full blown model for prices and returns that can be tested and potentially rejected".

Shiller's excess-volatility test checks an efficient markets model with a constant required return. In that model, the price is the best forecast of the discounted value of the dividends paid afterwards. A forecast should vary less than the thing it forecasts.

A 1980 working paper with the same title as his 1981 paper finds the opposite.

With the long-run trend removed from both, that present value is "a very stable and smooth series" next to the actual price series. For the 1981 paper, the committee reports that the price variance was "much larger than the variance of the discounted sum of future dividends".

The abstract adds a condition. The model is inconsistent with the data "if the long-run trend of real dividends is assumed given".

Shiller's lecture states that "the basic notion of efficient markets does not necessarily require that discount rates are constant or that returns are not forecastable".

His answer is that attempts to improve the consumption-based model have not yet given "an inspiring vindication of efficient markets theory". That model is a rational one. It ties prices to how investors value spending now against spending later.

He states his own reading: "It is far more plausible that the aggregate stock market price changes reflect inconstant perceptions".

His alternative is a model of behavioural finance with two groups. Ordinary investors' demand does not respond to expected returns. He calls this demand "animal spirits". Smart money investors respond rationally but are limited by their wealth.

In this model "price may have low predictability from day to day or month to month, consistent with efficient markets theory, even if animal spirits dominate the broad movements". The first shared finding fits his reading too.

Shiller writes: "Bubbles are not, in my mind, about craziness of investors."

The committee's public summary states the open question: "It is too early to say to what extent predictability reflects natural swings in the amount and (rational) perception of risks and to what extent it reflects mispricing."

Fama's lecture describes the evidence in similar terms: "No available empirical evidence resolves this issue in a way that convinces both sides."

The committee's press release says current understanding relies in part on risk and in part on behavioural biases and market frictions.

Both lectures single out one word. Shiller writes that he suspects "some of the apparent split is illusory", because there is no widely accepted definition of "bubble". Fama writes that the word needs "more careful definition and empirical validation".

A claim to beat the market has a horizon, a cause and a record.

Horizon: over days and weeks, a claim runs against the committee's finding that hardly any unexploited profit remains after transaction costs.

Cause: over several years, the shared pattern supports two meanings. If expected returns rise after prices fall because risk is higher, the higher return is payment for bearing that risk.

If the cause is mispricing, the higher return is a gain from errors that limited smart money has not removed. Higher past returns do not by themselves say which.

Record: Shiller's lecture says a test can have very little power to tell a random walk (unpredictable price changes) from slow mean reversion (prices drifting back to a trend). Where the two are close enough, he writes, that holds "even with millions of minute to minute observations".

A backtest is one fixed span of history. More frequent observations within it do not lengthen it.

Fama and Shiller agreed on what returns do over days, weeks and years. They parted ways on why. On the 2013 record, a claim to beat the market inherits that open question.

This explainer is part of the YX Insights Academy. Each one takes a single idea and checks it against real data.

The same approach runs through everything else we publish:

Good places to start on the website:

It is the proposition that asset prices reflect all available information. Fama's prize lecture says it cannot be tested alone, because a test also needs a model of the returns investors should expect. A failed test therefore leaves open whether the market or the model is at fault.

Yes. Both accept two findings. Stock returns are very hard to predict over days and weeks, but show some predictability over several years. They differ on the cause of the second finding. Fama's lecture reads it as a changing risk premium, while Shiller's lecture reads it as mispricing driven by investor behaviour.

Eugene Fama, Lars Peter Hansen and Robert Shiller shared it, "for their empirical analysis of asset prices". The committee announced the award on 14 October 2013. The prize is formally the Sveriges Riksbank Prize in Economic Sciences in Memory of Alfred Nobel.

It is the extra return an asset must offer over a risk-free asset to compensate for its risk. In the committee's account, an asset that does badly when investors value money most must command a higher one. Fama reads the long-horizon predictability as variation in this premium.

DISCLAIMER: This article is strictly educational. Any information or analysis in this note is not an offer to sell or the solicitation of an offer to buy any securities. Nothing in this note is intended to be investment advice and nor should it be relied upon to make investment decisions. Any opinions, analyses, or probabilities expressed in this note are those of the author as of the note's date of publication and are subject to change without notice.