Rationality of analysts’ earnings forecasts: evidence from dow 30 companies

Abstract

We test the rationality of analysts’ earnings forecasts for Dow 30 companies using an improved statistical methodology that accounts for non-stationarity in time-series data, non-normality in co-integrating regression, and serial correlation of forecast errors. Using one-quarter-ahead forecasts from 1984:Q4–2000:Q1 and analyzing firm-by-firm for Dow 30, we find that the earnings forecasts for at least two-third of our sample firms are consistent with the prediction of rational expectations hypothesis (REH). The most important implication of this finding is that it is premature to conclude that analysts’ estimates are irrational and systematically biased.

Acknowledgements

We thank Terence Lim, Bill Francis, Arnie Cowan, Christian Wolff, Raj Aggarwal, Rene Stulz, Peter Pope, Susan Malley, George Papaioannou, Steven Krull, Steve Bolten for providing many useful comments, which helped improve our paper significantly. We also thank Richard Green, John Campbell, two anonymous referees, and seminar participants at University of South Florida, St. John's University, Hofstra University, Grand Valley State University, and University of St. Thomas for providing useful comments on earlier version of this paper. We are also thankful to Christopher Pantzalis (the discussant) and other participants of the session at the FMA 2001 in Toronto, where the earlier version of this paper was presented. We are grateful to I/B/E/S Inc. for providing earnings per share forecast data. We acknowledge research support from the Zarb School of Business at Hofstra University, the College of Business at University of St. Thomas, and the Deutsche Asset Management. However, we are solely responsible for any remaining errors.

Notes

¹ Abarbanell and Bernard (1992), Brown et al. (1985), and Stickel (1990) using Value Line, I/B/E/S and Zacks data sources, respectively, found analysts’ earnings forecasts are systematically biased.

² In April 2003, the 10 largest Wall Street Firms reached a landmark settlement with NY state Attorney General, the SEC and state regulatory authorities on the issue of conflicts of interest faced by security analysts. These firms agreed to pay $1.4 billion in penalties to settle the charges that securities analysts employed by these Wall Street firms routinely issue optimistic research reports to win investment banking business from the companies they analyse.

³ Analysts employed by an investment bank with no underwriting relationships with a given firm are referred as non-underwriter analysts.

⁴ Prior studies document that analysts exhibit herding behaviour which may cause one analyst's over-estimate to influence others (Scharfstein and Stein, 1990; Huberts and Fuller, 1995; Brown, 1996; Olsen, 1996; Welch, 2000). Earlier studies on analysts’ forecast errors also suggest that individual estimates tend to cluster (Cragg and Malkiel, 1968; Richards et al., 1977). Prior literature provides evidence that analysts make biased estimates by irrationally extrapolating recent earnings information (Lakonishok et al., 1994; La Porta, 1996).

⁵ We have chosen not to extend our data set beyond the year 2000:Q1 due to the downturn in the economy, bear markets associated with extreme earnings volatility, corporate scandals, accounting frauds, and earnings manipulations during the past three years.

⁶ Earlier studies on analysts’ forecast errors also suggest that individual estimates tend to cluster (Cragg and Malkiel, 1968; Richards et al., 1977).

⁷ Unlike the forecasts reported in the I/B/E/S database, whisper forecasts are unofficial and may contain information which is impounded in stock price prior to the earnings release (Bagnoli et al., 1999).

⁸ For additional details on the advantages and limitations of using cointegration and analysis to assess time series data, see for example, Campbell and Perron (1991), Banerjee and Hendry (1992), and Engle and Granger (1992).

⁹ We would like to point out that Keane and Runkle (1998) used a different time period (1983:Q4–1991:Q4) with only six industries and they rejected the hypothesis of forecast rationality for the airline industry due to unpredictable earnings and losses associated with the industry.

¹⁰ Citigroup was created after the merger of Travelers Group with Citicorp. Travelers Group, formerly a Dow-30 company, had gone through several mergers and acquisitions since the 1990s prior to the merger with Citicorp in 1998. Our finding of bias in Citigroup is similar to O’Brien (1988) related to firm-specific as well as aggregate industry shocks in earnings.

¹¹ Even though when analysts’ forecasts are biased and conducting further tests of efficiency is pointless, we performed the test of efficiency for the second sub-period for Microsoft to confirm if there was any trend in the technology sector for the second sub-period. We found that the results for the second sub-period were similar to the first sub-period.