Wednesday, August 15, 2012

On Prices, Narratives, and Market Efficiency

The fourth anniversary of the Lehman bankruptcy has been selected as the release date for a collection of essays edited by Diane Coyle with the provocative title: What's the Use of Economics? The timing is impeccable and the question legitimate.

The book collects together some very thoughtful responses by Andrew Haldane, John Kay, Wendy Carlin, Alan Kirman, Andrew Lo, Roger Farmer, and a host of other luminaries (the publishers were kind enough to send me an advance copy). There's enough material there for several posts but I'd like to start with the contribution by John Kay.

This one, as it happens, has been published before; I discussed Mike Woodford's reaction to it in a previous post. But reading it again I realized that it contains a perspective on market efficiency and price discovery that is concise, penetrating and worthy of some elaboration. Kay doesn't just provide a critique of the efficient markets hypothesis; he sketches out an alternative approach based on the idea of prices as the "product of a clash between competing narratives" that can form the basis of an entire research agenda.

He begins with a question famously posed by the Queen of England during a visit to the London School of Economics: Why had economists failed to predict the financial crisis? Robert Lucas pointed out in response that the inability to predict a financial crisis was in fact a prediction of economic theory. This is as pure a distillation of the efficient markets hypothesis is one is likely to find, and Kay uses it to evaluate the hypothesis itself:
Lucas’s assertion that ‘no one could have predicted it’ contains an important, though partial, insight. There can be no objective basis for a prediction of the kind ‘Lehman Bros will go into liquidation on September 15’, because if there were, people would act on that expectation and, most likely, Lehman would go into liquidation straight away. The economic world, far more than the physical world, is influenced by our beliefs about it. 
Such thinking leads, as Lucas explains, directly to the efficient market hypothesis – available knowledge is already incorporated in the price of securities. And there is a substantial amount of truth in this – the growth prospects of Apple and Google, the problems of Greece and the Eurozone, are all reflected in the prices of shares, bonds and currencies. The efficient market hypothesis is an illuminating idea, but it is not “Reality As It Is In Itself”. Information is reflected in prices, but not necessarily accurately, or completely. There are wide differences in understanding and belief, and different perceptions of a future that can be at best dimly perceived. 
In his Economist response, Lucas acknowledges that ‘exceptions and anomalies’ to the efficient market hypothesis have been discovered, ‘but for the purposes of macroeconomic analyses and forecasts they are too small to matter’. But how could anyone know, in advance not just of this crisis but also of any future crisis, that exceptions and anomalies to the efficient market hypothesis are ‘too small to matter’?
The literature on anomalies is not, in fact, concerned with macroeconomic analyses and forecasts. It is rather narrowly focused on predictability in asset prices and the possibility of constructing portfolios that can consistently beat the market on a risk-adjusted basis. And indeed, such anomalies are often found to be quite trivial, especially when one considers the costs of implementing the implied strategies. The inability of actively managed funds to beat the market on average, after accounting for costs and adjusting for risk, is often cited as providing empirical support for market efficiency. But Kay believes that these findings have not been properly interpreted:
What Lucas means when he asserts that deviations are ‘too small to matter’ is that attempts to construct general models of deviations from the efficient market hypothesis – by specifying mechanical trading rules or by writing equations to identify bubbles in asset prices – have not met with much success. But this is to miss the point: the expert billiard player plays a nearly perfect game, but it is the imperfections of play between experts that determine the result. There is a – trivial – sense in which the deviations from efficient markets are too small to matter – and a more important sense in which these deviations are the principal thing that matters. 
The claim that most profit opportunities in business or in securities markets have been taken is justified.  But it is the search for the profit opportunities that have not been taken that drives business forward, the belief that profit opportunities that have not been arbitraged away still exist that explains why there is so much trade in securities. Far from being ‘too small to matter’, these deviations from efficient market assumptions, not necessarily large, are the dynamic of the capitalist economy. 
Such anomalies are idiosyncratic and cannot, by their very nature, be derived as logical deductions from an axiomatic system. The distinguishing characteristic of Henry Ford or Steve Jobs, Warren Buffett or George Soros, is that their behaviour cannot be predicted from any prespecified model. If the behaviour of these individuals could be predicted in this way, they would not have been either innovative or rich. But the consequences are plainly not ‘too small to matter’. 
The preposterous claim that deviations from market efficiency were not only irrelevant to the recent crisis but could never be relevant is the product of an environment in which deduction has driven out induction and ideology has taken over from observation. The belief that models are not just useful tools but also are capable of yielding comprehensive and universal descriptions of the world has blinded its proponents to realities that have been staring them in the face. That blindness was an element in our present crisis, and conditions our still ineffectual responses. 
Fair enough, but how should one proceed? Kay suggests the adoption of more "eclectic analysis... not just deductive logic but also an understanding of processes of belief formation, anthropology, psychology and organisational behaviour, and meticulous observation of what people, businesses, and governments actually do."

I have no quarrel with this prescription, but I'd also like to make a case for more creative and versatile deductive logic. One of the key modeling hypotheses in the economics of information is the so-called Harsanyi doctrine (or common prior assumption), which stipulates that all differences in beliefs ought to be modeled as if they arise from differences in information. This hypothesis implies that individuals can only disagree if such disagreement is not itself common knowledge: they cannot agree to disagree. It is not hard to see that such a hypothesis could not possibly allow for pure speculation on asset price movements, and hence cannot account for the large volume of trade in financial markets. In fact, it implies that order books in many markets would be empty, since a posted price would only be met by someone with superior information.

The point is that over-reliance on deductive logic is not the only problem as far as financial modeling is concerned; the core assumptions to which deductive logic has been applied are themselves too restrictive. To my mind, the most interesting part of Kay's essay suggests how one might improve on this:
You can learn a great deal about deviations from the efficient market hypothesis, and the role they played in the recent financial crisis, from journalistic descriptions by people like Michael Lewis and Greg Zuckerman, who describe the activities of some individuals who did predict it. The large volume of such material that has appeared suggests many avenues of understanding that might be explored. You could develop models in which some trading agents have incentives aligned with those of the investors who finance them and others do not. You might describe how prices are the product of a clash between competing narratives about the world. You might appreciate the natural human reactions that made it difficult to hold short positions when they returned losses quarter after quarter.
There is definitely ongoing work in economics that explores many of these directions, some of which I have surveyed in previous posts. But the idea of prices as the product of a clash between competing narratives about the world reminded me of a paper by Harrison and Kreps, which was one of the earliest models in finance to shed the common prior assumption.

For anyone interested in developing models of heterogeneous beliefs in which trading occurs naturally over time, the Harrison-Kreps paper is the perfect place to start. They illustrate their model with an example that is easy to follow: a single asset provides title to a stream of dividend payments that may be either high or low, and investors disagree about the likelihood of transitions from high to low states and vice versa. This means that investors who value the asset most in one state differ from those who value it most in the other. Trading occurs as the asset is transferred across investors in the two different belief classes each time a transition to a different state occurs. The authors show that the price in both states is higher than it would be if investors were forced to hold the asset forever: there is a speculative premium that arises from the knowledge that someone else will, in due course and mistakenly in your opinion, value the asset more than you do. The contrast with the efficient markets hypothesis is striking and clear:
The basic tenet of fundamentalism, which goes back at least to J. B. Williams (1938), is that a stock has an intrinsic value related to the dividends it will pay, since a stock is a share in some enterprise and dividends represent the income that the enterprise gains for its owners. In one sense, we think that our analysis is consistent with the fundamentalist spirit, tempered by a subjectivist view of probability. Beginning with the view that stock prices are created by investors, and recognizing that investors may form different opinions even when they have the same substantive information, we contend that there can be no objective intrinsic value for the stock. Instead, we propose that the relevant notion of intrinsic value is obtained through market aggregation of diverse investor assessments. There are fundamentalist overtones in this position, since it is the market aggregation of investor attitudes and beliefs about future dividends with which we start. Under our assumptions, however, the aggregation process eventually yields prices with some curious characteristics. In particular, investors attach a higher value to ownership of the stock than they do to ownership of the dividend stream that it generates, which is not an immediately palatable conclusion from a fundamentalist point of view.
The idea that prices are "obtained through market aggregation of diverse investor assessments" is not too far from Kay's more rhetorically powerful claim that they are "the product of a clash between competing narratives".  What Harrison and Kreps do not consider is how diverse investor assessments change over time, since beliefs about transition probabilities are exogenously given in their analysis. But Kay's formulation suggests how progress on this front might be made. Beliefs change as some narratives gather influence relative to others, either though active persuasion (talking one's book for instance) or through differentials in profits accruing to those with different worldviews. While Kay is surely correct that a rich understanding of this process requires more than deductive reasoning, it is also true that deductive reasoning has not yet been pushed to its limits in facilitating our understanding of market dynamics.


  1. " is also true that deductive reasoning has not yet been pushed to its limits in facilitating our understanding of market dynamics."

    I doubt that further work of this kind will offer new insights into what causes bubbles and crashes. The result you describe is one in a long line of negative results (going back at least to Kreps et al 1982) that show that standard neoclassical conclusions are fragile when extremely strong assumptions about unlimited information processing capacity, mistake-free behavior, etc., are weakened even slightly.

    More realistic models have to incorporate, at a minimum, agents with heterogeneous information-processing capabilities that are bounded above, and with heterogeneous resources. But it seems quite likely that small tweaks to these models could result in very different outcomes. Part of the reason models like those favored by Lucas survive in the discourse, despite being rather obviously inconsistent with events, is that it really is hard to build better models.

  2. I don't know Som. You seem to be arguing that standard models are fragile but still survive in the discourse because models that could replace them would also be fragile? Not sure I buy this. I don't see the harm in taking an intuitive argument and trying to explore it's logical consistency formally, while keeping in mind both the data and the institutional structures in place. Aside from Harrison/Kreps, there's work by Abreu/Brunnermeier and Hong/Stein on bubbles that I think is really insightful and quite robust. But Kay is right that we need much more than deductive reasoning.

  3. All of this disregards game theory in that winning market and trading strategies involve provide false information about intentions and valuations. In fact, the bulk of the current crisis is the result of parties profiting by the use of ratings agencies, risk insurers and simple complexity to hide the actual, and actually rather obvious to anyone watching wages and housing prices, risks.

    Economists pay too much intention to the fundamentals. That really isn't where one finds the big money in a free market.

  4. Kaleberg, economists can be accused of a lot of things but neglecting game theory is not one of them. The literature on strategic information transmission is huge. The two papers I mentioned in my reply to Som both try to account for non-fundamental volatility and are also part of an enormous literature. I have discussed ratings agencies, incentives in organizations, and the betrayal of trust in previous posts. The efficient markets hypothesis is just one of several perspectives on markets among economists.