No abstract

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.. Wiley and Accounting Research Center, Booth School of Business, University of Chicago are collaborating with JSTOR to digitize, preserve and extend access to Journal of Accounting Research. Existing empirical evidence indicates that analyst forecasts of corporate earnings do not meet the strict rationality standards prescribed by econometric tests of the rational expectations hypothesis.1 In this paper, we address the question of whether the expectation formation process underlying analyst forecasts is adaptive, or whether these forecasts are influenced by noninformational factors, such as incentives arising from the market for their forecasts (O'Brien [1988], Lin and McNichols [1993], Francis and Philbrick [1993], and Dugar and Nathan [1992]), and interaction between judgmental and/or statistical forecasting methods (Bunn and Wright [1991], Fan [1990], and Hoerl and Kennard [1970]).The rejection of rational expectations gives rise to two views of the forecasting process of analysts. The first, based on an informational argument, is that the forecasting process is adaptive (Marcet and Sargent *Carnegie Mellon University. We thank Andrew Alford, Larry Brown, Ravi Bhushan, Rick Green, Paul Healy, Shyam Sunder, and workshop participants at Carnegie Mellon University, M.I.T., SUNY at Buffalo, and the 1992 annual American Accounting Association meetings in Washington, D.C. The last author gratefully acknowledges the KPMG Peat Marwick Research Fellowship. 1 In particular, (i) analysts on average overestimate earnings (Fried and Givoly [1985], O'Brien [1988], and Abarbanell [1991]), and (ii) analyst forecasts appear to be inefficient relative to the information sets used to generate the forecasts (Klein [1990], Abarbanell [1991], and Lys and Sohn [1990]). However, Givoly [1985] reports that analyst forecasts exhibit properties implied by rationality. 103

A large body of literature has rejected rational expectations in relation to analyst forecasts. In this paper we start with a boundedly rational premise that analysts and managers adopt are influenced by the availability heuristic (Tversky and Kahneman [1973]). In the presence of business cycles, a rational growth forecast presupposes precise knowledge of the factors that cause a business cycle. In contrast, under the availability heuristic the current state is overweighted in the growth forecast and so whatever part of the business cycle the economy is currently in will be preserved by the investment decision. As a result, under the hypothesis of the availability heuristic we will observe ex-post, a systematic association between the drivers of this behavior (e.g., current state of the economy, industry, current performance) and subsequent forecast errors and (realized) growth when related to specific properties of the business cycle. We test for this condition and find evidence in support of it. The boundedly rational premise also provides a prediction for the nature of information that can improve analyst forecasts. The results from our analysis provide support for conclusion that a major driver of the boundedly rational behavior is ignoring the business cycle. Together these results provide direct evidence in support of Sargent's [2001] conjecture that a plausible reason for departures from rational expectations is the overly strong assumptions regarding the knowledge the agents are assumed to possess regarding the underlying laws of motion for the economy.

We study the performance of the rational expectations hypothesis in multiperiod experimental markets with multiple assets. We find that the markets are generally inefficient from the point of view of full information aggregation. However, arbitrage relationships hold, and it is not possible to detect the informational inefficiency by using some standard tests of market efficiency. These findings suggest that the lack of arbitrage opportunities and the failure of common tests to reject inefficiency are not sufficient to conclude that a market is informationally efficient.A GROWING BODY OF research has investigated whether the rational expectations hypothesis provides a reasonable description of behavior observed in laboratory markets. In this paper, we continue this investigation by examining information aggregation in fairly complex multiperiod markets with multiple stocks and experienced traders. We find strong evidence against information aggregation. However, we also find that conventional econometric tests are unable to reject the hypothesis of an informationally efficient market.Our study consists of two markets, one with three stocks and one with four. In each market, each stock has a two-period life. A stock pays a dividend at the end of each period, and the underlying distribution of endowments is common knowledge. Traders are given private information about the dividends.Important features of our laboratory markets are: (a) trading takes place in continuous time in a computerized market; (b) multiple units of the same asset can be traded in a single transaction; (c) the same dividends are paid to all owners of a stock; (d) the distribution of private information is not common knowledge; and, (e) borrowing and short sales are allowed.We analyze the data obtained from the experimental sessions in several ways. First, we study how efficiently our markets aggregate information by measuring the extent to which market prices reflect all available information. Since a stock pays dividends in both periods, the aggregation problem in the first period is more difficult than in the second (since the rational expectations price must correctly forecast the sequence of dividends). We find * Graduate School of Industrial Administration, Carnegie Mellon University. We thank without implicating two anonymous referees, seminar participants at Carnegie Mellon University, Duke University, MIT, and Washington University, the editor Rene Stulz, and Stan Zin for helpful comments. 1812The Journal of Finance that the rational expectations model does not perform well in the first period, but does better in the second period. This conclusion does not seem to depend on the number of securities in the market.Second, we ask whether some standard tests of market efficiency detect the deviations from rational expectations. In applying these tests, we assume the role of an econometrician who can only observe market data, not the underlying fundamentals.1 We find that, despite the lack of informational efficiency, the econometric tests d...

No abstract