Stock return predictability in rational markets: on the impossibility of informationally inefficient markets

Understanding the debate between behaviouralists and neo-classicists pays dividends for all researchers

By Bruce D. Grundy

The apparent predictability of share returns has led many to conclude that share prices are driven by investor behavioural biases. But are they? My research shows that, even when returns are apparently predictable, the final price set after a sequence of trades can fully and rationally reflect all available information and can therefore serve as a clear signal of value.

The primary goal of this paper is to demonstrate how research in finance is influenced by both the minutia of competing theoretical and empirical approaches and the broad sweep of market bubbles and crashes. The secondary goal is to acknowledge the social element of a research career: friendships with co-authors, admiration for past teachers, joy when a former student publishes their own work, and the respect one feels for all protagonists in a debate you find fascinating.

The neo-classical approach

Neo-classical finance has long referred to share markets as ‘efficient’ without always being clear as to what the word meant. In its earliest incantation, ‘weak-form efficiency’ was described as a situation where apparent patterns in past share prices and trading volumes could not be used to improve one’s prediction of future returns. The logic seemed compelling: if some past pattern did imply a high future payoff, then everyone would buy that share, bid up its price and any high return would have already occurred. In the 1970s, a view of markets as efficient was revolutionary and largely confined to academia. The profession of technical analysts, also known as ‘chartists’, was vigorously attacked by academics for claiming to be able to use past returns to predict the future. Academics viewed charting as a bygone relic, no more relevant than the 1830s political movement of the same name.

It was an exciting time to be an undergraduate student of Finance. The subject lecturer could demonstrate his purported proof of weak-form efficiency by inviting a group of technical analysts to view what was described as a recent history of values and then asking them to predict what would happen next. The chartists would obligingly identify ‘head-and-shoulders’ and other patterns in the data and give their predictions. Only then would the lecturer reveal that the set of values was not a series of past share prices, but had come instead from a random number generator. The duped share analysts were under-whelmed by the academic’s attempt to educate them through humiliation. Although the students loved it, such encounters hardly encouraged donations from the brokerage industry to fund further academic finance research.

The term ‘semi-strong-form efficiency’ was used to mean that all publicly announced information was fully reflected in stock prices. ‘Strong-form efficiency’ was viewed as the holy grail of capital markets – all public and private information impounded into prices. In other words, shares prices would be equal to the values that would prevail if insider information were instead posted on Internet bulletin boards. Since information is costly to acquire, a conundrum arises if markets are strong-form efficient. If prices do reflect all information, no-one will want to bear the cost of producing information.

Everyone will prefer to free-ride on the efforts of others. Less information will then be produced. Perhaps everything available might be reflected in prices, but there will be little information reflected. Share prices will prove poor guides for investment decisions.

The Grossman-Stiglitz challenge to market efficiency

A rigorous view of market efficiency was published in the American Economic Review in 1980, by Grossman and Stiglitz. That seminal paper’s title, ‘On the Impossibility of Informationally Efficient Markets’, inspires the contradictory title of this paper. The Grossman-Stiglitz model demonstrates that traders with the lowest costs of generating information will find that their information is only partially impounded into prices and that their profits from trading do cover their information costs. Less-efficient information producers will chose not to acquire information, but will condition their trades on the fact that they are at an informational disadvantage. They still trade for liquidity and rebalancing reasons even though they lose money on average to the better informed.

Thus prices in a Grossman-Stiglitz setting are noisy signals of value. Some equilibrium level of noise is necessary for the informed traders to hide behind. For most economists, the elegant Grossman-Stiglitz model has established that markets cannot be strong-form efficient. Joseph Stiglitz has since gone on to win the Nobel Prize in Economics and publish many important public policy studies based on his understanding of the flaws in markets. Sandy Grossman has left academia for the millions he now makes as a currency trader.

The empirical challenge to market efficiency

For some, the nail in the coffin of efficient markets was sunk on Black Monday, October 19, 1987 when $500 billion dollars was wiped off the Dow. How could so much wealth simply evaporate overnight if prices had been set efficiently on the prior Friday? And even the faith of stubborn believers happy to describe the ‘87 crash as just one data point, began to falter with the 1993 Journal of Finance publication of an empirical study by Jeegadesh and Titman. This study demonstrated that a simple long-short strategy of buying shares with prior-year winning performance in the top 10 per cent of all firms and selling the prior losers with realised prior-year returns in the bottom 10 per cent was shown to produce subsequent six-month profits of around six cents per one dollar, long and short. Such profits are both statistically and economically significant. A strategy of buying past winners and selling past losers is known as a ‘momentum strategy’. A myriad of academic studies into other long-short strategies have helped underpin the boom in the hedge fund industry.

The behavioural challenge to market efficiency

As the empirical evidence against market efficiency mounted, the field of behavioural finance blossomed and focused directly on individual behaviour as the source of the noise in share prices. Pioneering research was bought together in a 1993 volume edited by the field’s founder, Richard Thaler. Behaviouralists view the battle between their ideas and neoclassical believers in market efficiency as a paradigm shift of the type identified in Kuhn’s classic text, The Structure of Scientific Revolutions. Kuhn has argued that scientific advances are not incremental, but instead involve heated scientific revolutions in which one paradigm supplants another before the profession settles back down to incremental ‘normal science’.

And yet some still believe

Some financial economists (often Chicago-trained) remain more enamoured with the beauty of unimpeded capital markets than is fashionable among behaviouralists. As an admirer of markets, I do not see efficiency and costly information as inconsistent. I believe capital markets can in fact solve the twin problems of providing a reward to those who generate information and a guide to those who must make investment decisions. After doctoral studies at Chicago, I took my first academic position at Stanford’s Graduate School of Business. Most of my new colleagues were heavily influenced by their reading of Grossman-Stiglitz and considered efficient markets as a passé 1970s view of the world.

In co-teaching a course with an accounting academic, Maureen McNichols, we discovered a joint interest in just how accounting information came to be reflected in prices. We took Grossman-Stiglitz as given – in other words, we recognised that all information cannot be revealed through just one round of trade. Yet, we found it interesting to ask what happens if the market re-opens after an initial trading round.

A re-evaluation of the Grossman-Stiglitz contribution

One apparent answer is that nothing happens: the quoted price will be unchanged from the prior price and no further trade will occur. This possibility is incorrectly believed by many economists to be a simple reflection of the Milgrom-Stokey ‘no-trade theorem’ published in the Journal of Economic Theory in 1982. The no-trade theorem states that no trade can occur when traders agree on how information should be interpreted and start from an initial pareto optimum allocation given their information.

Grundy and McNichols showed that how different traders interpret a given piece of information and whether an allocation is pareto optimal given their information are endogenous properties of an equilibrium and not something to be assumed. They also demonstrated that further trade can occur at new prices when the market re-opens. And when this does happen, the sequence of prices is fully revealing – the final price is equal to the value that would prevail if any insider information had been posted on Internet bulletin boards. At the end of the first round of trade, traders need not agree on how to interpret the information in the first round’s price and need not have traded to a pareto optimum. Our work, published in 1989 in the Review of Financial Studies, demonstrated that the widely-cited no-trade theorem is itself vacuous and should in fact be expressed as: ‘In an equilibrium in which there is no trade, there will be no trade.’ Grundy and McNichols established that producers of costly information can earn their rewards in the early, noisy rounds of trade, while corporations’ investment committees can rely on the final set of share prices as accurate guides for investment purposes. When interpreting the final prices, all traders will be cognisant of the entire price history and all rational traders are in fact chartists (though the reverse need not be true).

A re-evaluation of the empirical evidence

Grundy and McNichols went on to show that rare events like the crash of ‘87 can be the rational reflection of existing information not previously reflected in prices. It is interesting to note that the 1988 Securities and Exchange Commission Report into the ‘87 crash pinned the blame squarely on the portfolio insurance industry. Portfolio insurance was a then-novel product that had not yet established its own lobby group in Washington. The actual evidence underlying the Report’s conclusion is as convincing as that underlying the conclusion in the earlier 1963 New York Stock Exchange Report into the crash of May 1962. That earlier crash wiped 27 per cent off the Dow and was supposedly due to the fact that the pace of selling by women far out-stripped their purchases! By 1987 women were no longer powerless and portfolio insurers were to be burnt in their place.

What about the apparent profitability of momentum strategies? Momentum strategies that pick stocks based on their past returns do on average make money.

If share prices appropriately reflect past prices, prices are at least weak-form efficient, the average profit to a momentum strategy must be an appropriate reward for the risk inherent in the strategy. In joint work with Spencer Martin, a former doctoral student I had supervised after moving to the Wharton School, I documented that a momentum strategy is risky and makes losses 40 per cent of the time. Still, in our 2001 Review of Financial Studies publication, we concluded that the strategy’s average profit is too large to be explained by current state-of-the-art risk measures. As true neo-classicists, we concluded that the fault lies in the finance profession’s present inability to accurately define and measure risk, rather than in a failure of markets to accurately reflect past returns.

Since joining the Department of Finance at the University of Melbourne, I have continued my research into how future returns are related to past returns. In joint work with Wei Li and Joe Zhang, researchers at Louisiana State University and Singapore Management University respectively, I have established that when new information signals both higher profits in this period and an additive increment to next period’s profits, the shares of companies with higher realised returns in this period will be less risky in the future. These shares will be expected to earn lower returns in the future commensurate with their reduced risk. Stock returns should then display contrarian rather than momentum behaviour. But if good news in this period presages a multiplicative rather than additive effect on future profits, returns are likely to display momentum behaviour. Our ongoing work empirically investigates the importance of distinguishing between additive and multiplicative uncertainty.

Research puzzles continue to fascinate

The debate between behavioural and neoclassical finance is misdirected. Just how individuals make investment decisions is important and must affect quantities. For example, an individual’s often biased predictions will affect whether he or she buys life-insurance. But an individual’s biases need not affect the price they will be asked to pay for that insurance. The real promise of behavioural finance lies not in understanding stock prices, but in understanding how best to market financial products to those who must finance their own retirement. The real promise of neo-classical finance is the pleasure inherent in a perpetuity of fascinating research puzzles. With the aid of our colleagues we can solve these puzzles and, in doing so, improve the standard of living for all future retirees.

Top ^

A condensed version of his Inaugural Lecture delivered on 3 May 2005.

Professor Grundy is Professor of Finance in the Department of Finance at the University of Melbourne.

Volume 1 April 2007