Cite as Michael A. S. Guth, "Feedback Effects and Speculative Bubbles in Informational Price Theory," Chapter 7 in Michael A. S. Guth, SPECULATIVE BEHAVIOR AND THE OPERATION OF COMPETITIVE MARKETS UNDER UNCERTAINTY, Avebury Ashgate Publishing, Aldorshot, England (1994), ISBN 1856289850.

Feedback Effects and Speculative Bubbles in Informational Price Theory

Informational price theory (IPT) has become the main paradigm to explain how prices reveal information. In the older efficient markets literature, prevailing market prices were thought to impound all available information. Thus a price change could indicate that an asset was previously under or overpriced but not that it is currently mispriced.[1] In contrast, IPT implies that prices not only clear markets but signal future movements. Prices convey information about expectations for future earnings, so that a rise today may be confirmed by subsequent rises tomorrow - as more and more people learn the private information.

If prices fully reflect information, then no one has any incentive to buy information. Everyone can simply infer information from prices for free. But if no one purchases information, then market prices will reflect no information. Hence everyone will have an incentive to buy information. The obvious solution to this conundrum is that prices only partially reflect information. Grossman and Stiglitz (1976) developed a model that formalizes this logic.

In equilibrium, some fraction of the market pays to become informed and the remaining fraction chooses to infer information from noisy prices.[2] Prices reflect some, but not all, of the private information which the informed traders have purchased. The competitive advantage of knowing the private information compensates the informed for their acquisition cost. At the margin, prices aggregate just enough information such that the last individual to buy information is indifferent between paying for information or attempting to infer it from prices.

The sequence of events in the Grossman and Stiglitz model looks something like this. Informed traders receive good news about some asset, buy the asset in mass, and thereby raise its price. Uninformed traders witness the increase in price, infer that the informed must have received favorable news, and subsequently buy as well. A similar logic applies when the informed receive unfavorable news and sell some asset. The uninformed follow suit. In both instances, purchases or sales by the informed cause a feedback effect by the uninformed.

One example where the uninformed might benefit from mimicking the actions of the informed, without introducing a feedback effect, would be in selecting a restaurant in an unfamiliar city. Visitors (the uninformed) might patronize a restaurant that seemed crowded. They would reason that local residents (the informed) know the prices and quality of food at that restaurant and have chosen to eat there. In this restaurant selection example, the informed derive no benefit from the actions by the uninformed. But in financial markets, capital gains will usually depend on the purchasing activities of other market participants.

The literature on information and prices has identified three basic speculative problems with the IPT approach.[3] First, in judging the quality of goods or services by price, consumers face uncertainty over whether high prices actually indicate superior quality or merely an attempt by the seller to attract consumers. If a consultant charges a high hourly fee, is it because he is superior to other professionals with a lower fee, or merely a speculative attempt to persuade clients into thinking he must be good if he charges that high fee? Only in strictly competitive supply markets would price-taking behavior by sellers limit the opportunities of low quality vendors to fool new consumers with high prices.

Second, attempts at market timing may limit the extent to which the informed act on their information. If someone receives a favorable signal on a stock, should he rush out and buy it? How can he know whether the current price already impounds his private signal? If other traders are planning to sell the stock and realize short term capital gains, he might gain from waiting until the price has fallen with their sales. Hence, no feature of IPT guarantees the informed will immediately act upon their signal.

Third, feedback effects have been thought to unravel the foundations of IPT, as investors then have incentives

to acquire (and in some cases disseminate) information on the purchasing proclivities of the market as well as on the `real' factors of earnings, dividends and so forth. What we end up with is a model of a market in which even when favorable information on earnings is available to insiders or informed traders, there is uncertainty as to whether this will be acted upon; in which `tulip bulb mania' factors can be important so that rises in prices might reflect only a belief that the market will value the stock higher next period, independent of earnings; and in which the chances of actually ferreting out any information on earnings prospects from observed prices is miniscule (sic). (Burness, Cummings, and Quirk (1980), p.75)

At first glance, this line of criticism would seem to eradicate any application of IPT to financial markets. However, in moving from words to formal models, we find that generating speculative profits off the feedback effect is actually more difficult then it sounds. Uninformed traders are watching prices. In order to realize their speculative profits, informed traders must buy and resell, yet their subsequent sales must not dissuade the uninformed from buying as a feedback.

This criticism of IPT also involves questions about strategic interactions usually applied in noncompetitive environments. Grossman and Stiglitz developed their IPT model with perfect competition: traders take prices as given. If purchases by informed traders cause prices to rise, then prices are indeed a strategic variable under the control of the informed agents - even in the original theory. In financial markets, informed traders could then profit from manipulating the market price in order to fool the uninformed into buying or selling as feedback effects. The question then arises as to what modeling tool best captures the essence of IPT and the possibility for this strategic interaction.

One might envision casting the informed agents as Stackelberg leaders with the uninformed agents acting as followers in a game theory model. However, a Stackelberg equilibrium approach would entail some lack of intelligence or rationality on the part of one or more players.[4] The usual IPT dichotomy of intelligence holds that the uninformed traders `are smart enough'[5] to know the correlation between market price levels and news. Yet the informed traders are not smart enough to realize their purchases (sales) induce a feedback effect.[6] In a Stackelberg equilibrium approach, the leader (the informed) would be smart enough to realize his purchases trigger a feedback effect, but the follower (the uninformed fraction) does not consider the leader's strategy in deciding whether to buy. We need a model in which all players act intelligently.

The first solution developed in this chapter is a Bayesian Nash Equilibrium in a game of incomplete information. Informed traders play opposite uninformed traders, each taking the other's strategy as given. The informed group, having incurred the cost of acquiring private information, moves first. The informed group may try to fool the uninformed into believing that asset returns are better than they actually are. Sometimes the deception succeeds. A bubble results, and after the crash is over and the dust has cleared, the informed group turns out to have gained at the expense of the uninformed. Sometimes the deception fails, and the informed traders wind up losing more money than if they had followed a conventional, fundamental-based strategy.

The equilibria of this game, contained in Sections 7.3 through 7.6, indicate the probability that the informed traders acting as a group will collude (i.e., buy when they have received a low signal) as well as the probability that the uninformed, acting as a group, will buy as a feedback effect. To assure that some informed traders do not deviate from the group strategy, and thereby tip off the uninformed, we employ a self-enforcing[7] Nash equilibrium approach. By restricting individual behavior this way, we are limited to a two-player model, which cannot adequately describe the operation of densely populated, competitive markets.

To generalize the model beyond the two-group, two-player model of informational price theory, we examine subjective correlated equilibria in the final sections of this chapter. In this framework, the informed traders receive correlated - though not identical - signals, and all players act as individuals, rather than as members of an orchestrated group. The equilibrium characterizes the quantity of the risky asset that each individual selects, rather than the probability of purchasing.

Game theoretic analysis rarely applies to competitive markets. `Anony­mous games' are one class of `competitive' models where the outcome depends on the actions of each agent only through the proportion of agents that acts in a certain way. Like anonymous games, the pure strategy equilibria in this chapter might be interpreted as the frequency of each population that chooses to engage in this strategic play. Under this interpretation, we must consider each informed trader as having, e.g., firm-specific information. Thus one informed trader's private information and strategic use of that information will not affect another informed trader's strategy, because their strategies are focused upon different stocks. However, since no one individual's purchases should affect prices in a competitive market, this interpretation seems unrealistic at best. Thus we have reason to question the competitive market foundation of IPT regardless of the equilibrium concept chosen.

The significance of the results in this chapter are (1) the derivation of speculative bubbles from fully rational, payoff-maximizing behavior on the part of all market participants;[8] (2) the highlighting of an important feature of IPT: the two-player model; (3) a demonstration of the stringent conditions required for the informed traders to derive speculative profits off the feedback effect, thereby attenuating this criticism of IPT; and (4) utilization of some mathematical results by Nikaidô and Isodo that can be used in proofs of the existence of (subjective correlated) equilibria.

Section 7.2 compares our work to related articles in the speculative bubble and asymmetric information literatures. The players and extensive form game are presented in Section 7.3, the payoffs in Section 7.4, and the pure and mixed strategy equilibria in Section 7.5. A negative bubble is illustrated in Section 7.6. Section 7.7 offers an intelligent critique of the Bayesian Nash Equilibrium speculative bubble models. Section 7.8 generalizes the two-player, high-low demand model of the first six sections to allow for multiple strategies and multiple signals like Kyle (1989), and each player acting as an individual. Section 7.9 presents the conclusions and a discussion of the noncompetitive assumptions of the model. Table 7.1 in Section 7.3 conveniently lists the notation used in the Bayesian Nash Equilibria sections of this chapter.

7.2. The Speculative Bubble and Asymmetric Information

Literatures

The traditional view of speculative bubbles maintains that they result from irrational behavior by some or all of the market participants [Kindle­­berger (1989)]. Our model is distinguished from many previous derivations of speculative bubbles in that all the market participants are fully intelligent and rational. Blanchard and Watson (1982) and Leach (1991) illustrate bubble formation in which traders will refrain from participating unless they can rationally determine that the last (nth) trader will execute his trade, the n-1st trader will execute his trade, and so on. In these two articles, players may be induced to buy into the speculative bubble by the fact that the bubble may continue to grow indefinitely. Every `bubble' in our work, by contrast, bursts with probability 1 as the state of the world is revealed.

DeLong et al. (1990) have demonstrated convincingly that perfectly rational speculators can jump on the feedback effect bandwagon and not buck the trend. Speculators in the DeLong et al. model do not worry that some (nth) trader may get stuck holding the overvalued asset. From an expected profit-maximizing viewpoint, they would be foolish to be overly concerned with some distant collapse point, as long as they expect to realize their capital gains while the bubble still exists. Wang (1993) looks at the same basic model as DeLong et al.; however, `noise trading' is replaced by `supply shocks.' He finds the existence of investors with imperfect information increases the risk premia on stocks, and investors with different information will adopt different investment strategies.

Friedman and Aoki (1986, 1989) provide parametric examples of bubble formation when investors are oriented towards long-term gains and expectations; traders in their model ignore short-term capital gains and the potential ability to manipulate prices. Friedman and Aoki (1989) have illustrated a negative bubble arising from momentum in price trajectories over time. The Friedman and Aoki bubbles can result from overshooting, which tends to be self-correcting as the game is played over time and the players become more familiar with price escalations.

O'Flaherty (1987) looks at bubbles as nonconstant correlated equilibria in coordination games with asymmetrically informed agents. He finds a bias towards myopic investment and against public revelation of private signals. Similarly, Fishman and Hagerty (1992) show full disclosure of privately held information would lead insider traders to have zero profits. In the absence of a mandatory disclosure rule, security prices will generally not be efficient in transmitting information between informed and uninformed investors.

Fishman and Hagerty's work extends Kyle's (1985) noisy rational expectations model to study the impact of insider trading on price efficiency. In Kyle's model, informed traders strategically choose their transactions knowing their strategies will affect prices. The resulting Nash equilibrium quantity orders show that informed traders optimally choose to withhold some of their information. Kyle never satisfactorily explained why `noise traders' come into the market and routinely lose their wealth. Nevertheless, numerous authors have adopted Kyle's Walrasian auctioneer and noise trader framework to examine market issues such as intraday patterns in volume and price variability (Admati and Pleiderer 1988), diversification by the uninformed (Bhushan 1991), transaction costs to the uninformed (Subrahmanyam 1991b), and market liquidity and price efficiency (Subrahmanyam 1991a).

Smith, Suchanek, and Williams (1988) and King, Smith, Williams, and Van Boening (1993) have found evidence for bubble creation in experimental markets from `homegrown capital gains expectations,' which collapse when dividend and fundamental value information becomes common knowledge. Bubble creation in these experimental markets arise more from myopia than from an attempted manipulation of the uninformed based on feedback effects.

Allen and Gorton (1993) consider bubble creation resulting from incentive contracts paid to portfolio managers. The managers in their model receive no income if their portfolios have non-positive returns. However, they get to keep a portion of any positive return they achieve. This principal-agent contract leads portfolio managers to prefer taking risks and buying assets above their fundamental value on the chance that they will continue to rise.

Portfolio managers and institutional investors are often said to behave like herds. Banerjee (1992) has studied decision rules that depend on the investment strategies of previous decision makers. He shows that the resulting equilibrium is inefficient. People tend to be guided too much by other people's investments, even to the extent of ignoring their own private information. In Banerjee's Bayesian Nash Equilibrium, the order of play is important. The first few players' decisions have a much greater impact on subsequent investment patterns than those of players entering later in the game.

Our paper is most similar to work simultaneously derived by Gorton and Pennacchi (1989), who also develop a game theoretic model of strategic play between informed and uninformed traders. In the Gorton and Pennacchi model, the informed always try to manipulate and the uninformed always buy, whereas in our model this behavior would tend to drive the uninformed traders out of the market.[9] Gorton and Pennacchi focus on the rise of new securities that split the cash flows of underlying assets and on government intervention to protect the uninformed. Our paper focuses more on the differences between this strategic game and the predictions of informational price theory, as well an analysis of the competitive market assumptions.

The rest of the article can be found at http://michaelguth.com/economist/chap7.htm

Dr. Michael A. S. Guth, Ph.D., J.D. is a Professor of Financial Economics and Law for several universities with on-line degree programs and an attorney at law in Tennessee. He wrote his doctoral dissertation on topics in speculation theory, and this derivation of a fully rational speculative bubble was his most salient chapter. He can be contacted through http://michaelguth.com/economist.htm

In addition, Dr. Guth is a financial quant and former investment banker, having worked for Credit Suisse First Boston and Deutsche Bank in London and Frankfurt. He specializes in developing investment strategies and hedging techniques using derivatives. For five years, he consulted to the electric power and gas industry in the USA, even managing the Middle Office (financial risk control) function for two trading floors.

Dr. Guth has taught over 30 courses on-line at the undergraduate and graduate level on topics ranging from Managerial Economics to Strategic Management to Business Law. He is currently teaching Health Care Econ.

• Choosing A Financial Planner


• The Five Debt Triggers for Women


• Economic Outlook: Cause and Effect


• Buyer BEWARE


• Navigating Commercial Construction Financing


• Roth IRA secrets - 7 reasons why a Roth IRA trumps a Traditional IRA


• Thinking on Artificial Inflation


• Payroll Cards Improve Direct Deposit Participation


• Finance Tips


• How to Avoid Overdraft and Bounced Cheque Fees