RDP 9501: Modern Approaches to Asset Price Formation: A Survey of Recent Theoretical Literature 4. Other Informational Inefficiencies

The EMH assumes that economic agents make correct inferences about an asset's intrinsic value by fully and accurately aggregating all relevant private and public information. In standard models, economic agents are usually assumed to observe private signals, (e.g. about prospects for the asset in question), as well as acting on publicly available information about observed market prices and volumes. In determining their investment decisions, individual investors rationally combine these sources of information, giving correct weight to the private and market information. These individual rational assessments take the form of orders to buy or sell the asset, which lead to movements in its price. Accordingly, through the collective action of rational individual traders, complete and proper aggregation of widely dispersed information occurs, resulting in an equilibrium market price which coincides with the asset's fundamental value.

The third broad strand of literature which challenges the EMH focuses explicitly on incomplete or imperfect aggregation of information as the main factor explaining deviations of observed asset prices from their intrinsic values. Participants in asset markets frequently ignore or neglect relevant information and focus instead on some piece of extraneous information or variable, the relevance of which is highly tenuous. This can lead to substantial departures of asset prices from their intrinsic values.

Recent theoretical innovations on the microeconomic foundations of the trading process advance novel explanations for the presence informational inefficiencies in speculative asset markets. This literature, referred to as the microstructure of financial markets, provides a variety of plausible arguments as to why asset prices will not always fully and immediately reflect pertinent information available in a market. It also helps explain why asset prices can change sharply even without the arrival of new relevant information.

4.1 Asymmetric Information Models of the “Crash”

The distinctive feature of this literature is the explicit modelling of interactions between rational but asymmetrically informed traders.[27] It focuses on the structure of information flows between such market participants and examines how the actions, and expectations, of an individual investor are influenced by the perceived actions, and expectations, of other investors in the market. This structure has implications for the trading environment and for the size and speed with which pertinent new information affects asset prices. Models of this kind play a prominent role in recent theoretical analyses of financial panics, such as the stock market crash of 1987.

Gennotte and Leland (1990) explain “crashes” in terms of the inability of investors to perfectly distinguish between “information-based” trades (i.e. transactions due to arrival of new information) and “informationless” transactions (i.e. those arising from dynamic hedging strategies such as portfolio insurance). A critical assumption underlying this analysis is that a large segment of market participants has no independent information about the outlook for an asset, and instead makes investment decisions based on their observations of the current price. Only a small proportion of investors (“informed investors”) actually possess relevant private information about future economic fundamentals. With such a structure, Gennotte and Leland demonstrate that erroneous inferences made by the “uninformed” segment of the market, in the face of an unobserved supply shock, can lead to precipitous falls in asset prices that cannot be justified in terms of the fundamentals.

In this model, the following chain of events leads to a crash: some negative information enters the market, initially triggering a small fall in asset prices. This prompts sales by portfolio insurers, which magnify the initial fall in the asset price. “Uninformed investors”, who do not know that the sell orders have come from execution of portfolio insurance programs, assume incorrectly that the significant price fall reflects actions of (better informed) investors acting on the receipt of new information. This causes the larger “uninformed” segment of the market also to sell, eventually prompting massive selling and a “crash” in asset prices.

Similarly, Grossman (1988) in a paper discussing the effects of program trading and dynamic hedging strategies on stock prices (presented at a conference four months before the Crash of 1987) prophetically warned of the possibility of financial crashes in speculative asset markets characterised by significant information asymmetries. He argued that heavy selling due to portfolio insurance, if the selling were erroneously interpreted by uninformed investors, would make the market highly vulnerable to a large fall in asset prices.[28]

4.2 Role of the Trading Process

Examination of mechanisms by which asset prices move, in the absence of new information about fundamentals, is at the forefront of the latest theories on speculative markets. Contrary to the predictions of the EMH, this literature explicitly acknowledges that new information is not immediately reflected in asset prices and that trading activity itself plays an essential role in both the release and dissemination of privately held information. It demonstrates that even with totally rational economic agents, asset prices need not adjust instantaneously to new information. Nor will sharp asset price changes necessarily reflect the contemporaneous arrival of significant “news” on fundamentals. The relevant literature remains relatively fragmentary; the remainder of this section discusses some representative samples.

Bulow and Klemperer (1994) develop a model which shows how “frenzies” and “crashes” in asset prices can arise without the arrival of significant news about fundamentals. The starting point for their analysis is that, in reality, asset markets are not cleared by a Walrasian auctioneer who aggregates asset demand and supply before posting an equilibrium price. Rather, in speculative markets, investors make inferences about demand and supply by observing a sequence of actual transactions.

Such a trading structure encourages certain strategic behaviour by rational investors. Prospective buyers are confronted by the problem of determining the best time to purchase an asset: should they buy now at the current price or postpone their purchase to a later point in time in the hope that they can get a better price. The decision to purchase an asset hence depends not only on the buyers' reservation values, but more importantly on their expectations about future market-clearing prices. So there may exist a large number of buyers with very different reservation values for the asset, all having the same “willingness to pay”.

This state of affairs can make market participants extremely sensitive to even the slightest piece of new information. Bulow and Klemperer demonstrate that mere observation of a purchase by a single buyer will be sufficient to generate a “frenzy” of buying activity: such marginal information is all that may be required to alter buyers' perceptions about the prevailing price from “slightly too high” to “sufficiently attractive”. The resultant large increase in trading will reveal information about asset prices and either reinforce initial buying, possibly leading to a situation in which demand “feeds on itself”, or to a “crash” where the asset price falls precipitously.

Romer (1993) contends that some changes in asset prices are quite rational even if they are not in response to “external news”. This view is based on two propositions:

  • new information about an asset's fundamental value is dispersed among a large number of investors and asset markets are unable, immediately and fully, to aggregate all of this information.
  • the trading process itself can reveal pre-existing, but not yet fully “processed”, information about fundamentals, especially investors' assessments about the relevance of new information for an asset's intrinsic value. Accordingly, Romer conjectures that an important part of actual movements in an asset's price can be explained by “internal news” generated by actual trade in the asset.

Romer advances two simple models to illustrate the general idea that the trading process can convey relevant information. These mechanisms are neither mutually exclusive nor exhaustive.

The first is a scenario in which information is heterogeneous and each investor is uncertain about the quality of information other investors have. Individual investors will learn from observing actual trading – say, the market response to a large transaction. This response to such a “non-informational” change conveys to individual investors information about market “sentiment”, possibly leading to a re-weighting of the investor's own views about the asset's fundamental value. Such re-assessments will result in discrete, and possibly substantial, movements in relevant asset prices.[29]

The second scenario involves information of uniform quality, which is widely dispersed among a large number of active investors in the market, and where there are costs to trading. Under these conditions, rational investors might not have sufficient incentive to trade immediately upon receipt of relevant new information. Such information might only gradually be incorporated in asset prices as trade (motivated by other considerations such as liquidity needs) is undertaken at a later point in time. Thus asset markets might initially fail to reveal all new information possessed by different investors in the market. Observed movements in asset prices, unrelated to the arrival of new information or “external news”, are interpreted by Romer as reflecting the market's response to “internal news” – the continuing refinements of investors' assessments of previously released information relating to an asset's fundamentals.

In a similar vein, Caplin and Leahy (1992) develop a model in which they demonstrate that markets would be highly susceptible to crashes and collapses if the trading environment were characterised by widely-dispersed and idiosyncratic information and investors altered their behaviour infrequently. This model has two important assumptions:

  • individual investors incur fixed costs when they change well established behavioural patterns; and
  • individual investors' private information is conveyed to other market participants by alterations to their routine behaviour.

As a result of the fixed costs associated with changing routines, relevant information may be trapped “locally”, to be released only when a large number of individual investors decide to alter their routine behaviour in response to a substantial change in economic fundamentals. “Routine” behaviour thus impedes the dissemination of relevant information held by individual investors and creates the potential for large swings in market sentiment when a sufficiently large number of investors decide to alter their “routines”.

In Caplin and Leahy's model, only a small piece of additional information is required to precipitate a market crash or collapse. Such a small piece of additional news may be sufficient to cause an individual investor to alter his/her routine behaviour, releasing private, idiosyncratic information. This might prompt other investors and traders to do likewise. Through such a process, the market will observe rapid disclosure of a large body of previously accumulated and relevant – but previously “hidden” – information, which might lead to a radical change in market sentiment, and possibly a market collapse. In this framework, a market “crash” is viewed as a mechanism by which markets aggregate previously received, but widely dispersed, private information, rather than being the product of irrational behaviour. The “post-crash” market price is the one which would accord with all relevant information.

Dow and Gorton (1991) attempt to capture the notion that interpreting new information may be a complex task. They emphasise the idea that decisions based on an individual's private information may be fundamentally different from those that might be made if they were based on the aggregate of the new information received by all investors in the market. In these types of models, the trading process is a vehicle for releasing privately held, or hidden, information

4.3 Herding Behaviour and Informational Cascades

Herding behaviour – a situation in which investors ignore their own information and imitate the actions of other investors – has often been cited as one of the factors that can generate serious asset price inefficiencies and misbehaviour. The recent theoretical literature has also advanced new rational models of “herding” and “herd-like” phenomena .

Froot, Scharfstein and Stein (1992) examine the implications for speculative asset prices when rational investors and traders possess short-term trading horizons. In a simple theoretical model, the authors demonstrate that when speculative asset trading is dominated by short-term considerations, a particular type of informational inefficiency may arise: rational speculators may focus attention, or “herd”, on certain types of information, while neglecting others. It may be rational for speculators to trade on, and “choose to focus on very poor quality data, … even on completely extraneous variables that bear no relation at all to fundamentals”.[30]

These results are driven by positive informational spillovers – as the focus on a particular variable becomes increasingly widespread, it is more likely to be incorporated into the actual asset price. As a result, there are strong incentives to study this information at the expense of other variables, including information which may be far more relevant to determining the asset's fundamental value. A type of positive feedback therefore operates in the process of acquiring information – the greater the attention paid to a particular variable, the more valuable new information about that variable becomes, and so the focus on it will intensify and news about it will become more valuable still.

To illustrate, suppose that there are two variables, a and z. In a model in which trading horizons are short term, if a critical mass of market participants was to focus on variable a, the positive informational externality would result in a herding equilibrium in which everyone in the market would study only information about a and base trading activity on it. Information about z would be completely ignored, even though z might be particularly important to the performance of the asset concerned over the longer-run. Froot et al. note that this is one rationale for “chartism” and technical analysis in asset markets: “… the very fact that a large number of traders use chartist models may be enough to generate positive profits for those traders who already know how to chart. Even stronger, when such methods are popular, it is optimal for speculators to choose to chart. … Such an equilibrium can persist even if chartist methods contain no relevant long-term information.”[31]

Froot et al. also demonstrate that such herding equilibria do not imply that the market participants will always focus on the same variable. The structure can accommodate a situation in which market participants show intense interest in one variable for a short period of time (say, the current account), and then switch their attention to something else (e.g. monthly employment statistics).

In this framework, asset prices depart from fundamentals not because of a malfunction at the pricing stage of the market process, but because of imperfections emanating at the information acquisition stage. In other words, asset prices will immediately and fully reflect information, once a given set of information has been acquired. Inefficiencies arise, however, when the information set is not “relevant” (i.e. it is not related to economic fundamentals) or is incomplete.

A principal contribution of Froot et al. is that they develop a “microstructure” for the speculative bubble process. By providing a theoretical explanation as to why rational investors and traders might focus on totally extraneous information, the authors describe one mechanism by which bubbles and bubble-like episodes of financial instability might emerge and be propagated.

Another innovative contribution to this literature is offered by Bikhchandani, Hirshleifer and Welch (1992).[32] Bikhchandani et al.'s contribution involves a theory of “informational cascades[33] which explains why behaviour converges, how this situation is maintained, and why it can be fragile (in the sense that small shocks can lead to dramatic changes in behaviour).

An informational cascade occurs where it is optimal for individuals, after observing the actions of previous agents, to ignore their own information and mimic the decisions of their predecessors. Informational cascades therefore tend to produce uniform actions by a large number of rational economic agents, who may possess disparate and conflicting private information. They are likely to be most evident in a sequential trading environment in which individuals have limited private information. This class of model shows that, when individuals base their decisions on private information and observations of their predecessors' behaviour, informational cascades develop with virtual certainty.

Bikhchandani et al. emphasise that the decision to ignore private information and participate in the “herd” arises from a process of rational decision making. It should not necessarily be interpreted as representing irrational or foolish behaviour. The actions of predecessors provide valuable information, since this history reveals information about private signals that other agents have received. A rational decision maker will, therefore, combine this evidence with his private information. If he observes other investors repeating a particular action, it is rational to conclude that these investors have all received the same signal. Accordingly, he will give added weight to the implications of his predecessors' actions, even if his own information has a contrary implication. Once the information implicit in the actions of other investors accumulates to the point where it marginally outweighs his own information, the optimal response is to imitate unconditionally the actions of other investors. The next individual will then find that the accumulated evidence from his predecessors actions begins to dominate his information, and find it optimal to join the cascade. Such a chain of reasoning extends to all subsequent individuals and the cascade takes hold.

One feature of Bikhchandani et al.'s approach, is that informational cascades are inherently fragile.[34] Since they can develop on the basis of very little information, and their “depth” does not increase with the number of individuals in the cascade, informational cascades can also collapse suddenly with the arrival of a small item of news. This “knife-edge” property arises because uniformity of actions occur when everyone in the cascade just barely prefers to ignore their own information and imitate the actions of others. This makes the cascade brittle.

Informational cascades hinder the proper aggregation of widely dispersed relevant information since, once a cascade develops, actions of successive individuals do not reflect relevant information which they have received; by definition, individuals become part of a cascade when it is optimal for them to ignore their own private information and follow the actions of others. As a result, informational cascades may lead to outcomes which are sub-optimal or inefficient, given all the relevant information.[35]

4.4 Rational Beliefs

The final seminal contribution to be discussed is by Kurz (1992). He develops a model in which asset mispricing arises from the heterogeneous beliefs of rational economic agents, reacting to the same piece of information. Kurz introduces the notion of rational beliefs which are defined as “probability beliefs about future economic variables which cannot be contradicted by the data generated by the economy”. He argues that the set of such rational beliefs will generally be very large and their diversity is the most important cause of serious overvaluations (and undervaluations) in asset prices. The view that even rational economic agents do not and cannot possess complete “structural” knowledge of a dynamically evolving economic system underlies Kurz's analysis; as he states “… the assumption that agents possess complete structural knowledge has no empirical support … there is much in the dynamic structure of the economy which cannot be learned with certainty.”[36] Consequently, it is possible that two rational economic agents who receive the same piece of information, will hold totally different views about its likely impact on asset prices.

Kurz's framework explicitly allows for the possibility of inaccurate assessments of available information by rational economic agents individually, as well as by the market as a whole. In this model, the source of asset price inefficiencies does not lie in imperfect aggregation of relevant information, but in the unavoidably imperfect evaluation of already processed information by rational economic agents.[37]


The theoretical literature on the problems posed by heterogeneous and asymmetric information in speculative asset markets has its genesis in the seminal contributions of Sanford Grossman. See in particular Grossman (1976, 1977, 1978, 1981) and Grossman and Stiglitz (1980).

One of the central propositions that emanated from these theoretical contributions is what is sometimes referred to as the “paradox of fully revealing rational expectations equilibrium”. (A fully revealing rational expectations equilibrium is defined as being one where the equilibrium market price reveals and aggregates all diverse information). Grossman and Stiglitz (1980) demonstrated that the notion of market efficiency is incompatible with competitive equilibrium in the presence of information costs: with costly information there cannot exist a competitive equilibrium price which fully reflects all relevant information. The logic of the argument is relatively straightforward. If market prices were informationally efficient (i.e. they reflect and convey all relevant information available in an economy), there would not exist any incentives for economic agents to outlay the resources in acquiring and collecting new information; it could be costlessly obtained by observing the prevailing market price. However, if no individual economic agent has an incentive to collect and acquire new information, how does new information get aggregated or impounded into the market price in the first place? [27]

Central to Grossman's argument is the important informational distinction between a real security and a synthetic security – a set of dynamic trading strategies which attempt to replicate the payoff profile of an actually traded security. “Portfolio insurance”, a trading strategy in risk-free securities and index futures which replicate the payoff profile of a European Put Option, is an example of such a synthetic security. Grossman argues that even though dynamic trading strategies such as portfolio insurance can synthesise the payoff of a real security (a European put option), they impose quite different informational demands on actual market participants. Grossman argues that the price of a real security such as an option conveys important information concerning future trading plans and price volatility which is not transmitted when investors utilise a dynamic trading strategy. The increasing substitution of synthetic securities for real securities therefore renders asset markets far more susceptible to erroneous assessments and inaccurate inferences. [28]

This first model of Romer's is not dissimilar to the arguments presented in Gennotte and Leland (1990). Romer actually advances this structure to explain episodes characterised by large asset price movements without the arrival of significant outside news (viz. the stock market crash of 1987). [29]

Froot, Scharfstein and Stein (1992, p. 1463). [30]

Froot, Scharfstein and Stein (1992, p. 1480). [31]

See also Hirshleifer (1994). [32]

The notion of an “informational cascade” was first formalised by Welch (1992) in his theoretical discussion on the optimal pricing of Initial Public Offerings (IPO). [33]

Birckhcandani, Hirshleifer and Welch (1992) identify four other mechanisms/theories of conformity that have appeared in the recent literature. These are (i) sanctions against deviants; (ii) positive payoff externalities; (iii) conformity preference; and (iv) communication. The authors point out that these processes, unlike informational cascades, induce patterns of mass conformity which are resilient to small shocks or perturbations. The fragility of an informational cascade is what distinguishes it from these other models of mass conformity. [34]

See also Banerjee (1992), Kirman (1993) and Lee (1993a,b) for further contributions to this literature. [35]

Kurz (1992, p. 1). [36]

Some further theoretical contributions which have not been discussed but are worth drawing attention to are: Allen and Gorton (1993), Banerjee (1993), Friedman and Aoki (1992), Topol (1991), Wang (1993) and Ziera (1993). [37]