Volume of Trading and the Dispersion in Financial Analysts' Earnings Forecasts

• Published in: The Accounting review Vol. 66; no. 2; pp. 389 - 401
• Main Authors: Ajinkya, Bipin B., Atiase, Rowland K., Gift, Michael J.
• Format: Journal Article
• Language: English
• Published: Menasha, Wis American Accounting Association 01.04.1991
• Subjects: Analytical forecasting; Earnings; Earnings forecasting; Earnings per share; Empirical evidence; Employment forecasting; Financial advisers; Forecasting models; Forecasting standards; Forecasts; Investors; Mathematical models; Securities trading; Shorter Articles; Statistical forecasts; Statistical significance; Studies; Trade volume; Variables; Volume
• ISSN: 0001-4826; 1558-7967

Varian (1985) and Karpoff (1986) showed analytically that trading volume is positively related to the degree of differing beliefs. This study provides empirical evidence on the postulated relationship. The extent of disagreement or dispersion in financial analysts' forecasts of annual EPS for a firm is employed as the proxy for agents' differing beliefs about the firm's prospects. The revision in analysts' mean EPS forecasts, from one month to the next, is used to control for the volume effects of the net information signals emanating during the period. While some researchers have addressed the relation between the level of trading and earnings announcements (Beaver 1968; Morse 1981), and between trading volume and the magnitude of earnings forecast errors (Bamber 1986, 1987), the impact of discordant expectations on trading volume has been the subject of more recent empirical examination (Comiskey et al. 1987; Ziebart 1990; Lang and Litzenberger 1989). The hypothesis tested in this study posits that the fraction of outstanding common shares traded is a positive function of both forecast dispersion and mean forecast revision. While most previous studies of volume have concentrated on specific accounting disclosures, this study examines the trading associated with an almost continuous flow of information about the sampled firms that analysts implicitly use in their periodic revisions of earnings forecasts. Hence, tests of the model become possible at several times during the year, independent of formal accounting events/disclosures. Monthly observations in each of the four years 1978 to 1981 are used for the sample of 420 calendar-year firms; total number of observations is 16,747 over 48 months. Generalized least squares (GLS) estimation is applied to observations pooled over time and cross-sections and for each of the four years separately. Ordinary least squares (OLS) estimations are also performed and reported for comparative purposes and also to assess the stability of the models in monthly cross-sections. The results indicate a significant positive association between the dispersion in analysts' forecasts of annual EPS and the volume of trading. A relatively stable and positive association is found even after controlling for the volume effects of the magnitude of monthly revisions in the mean analysts' (annual) EPS forecast. The evidence corroborates the theoretical result that the degree of heterogeneity in beliefs is a determinant of the intensity of trading.