THE STOCK-PRICE EFFECTS OF ALTERNATIVE TYPES OF MANAGEMENT EARNINGS FORECASTS

This paper examines the stock price effects of alternative types of management earnings forecasts. Beyond deciding whether to disclose forecasts, managers must decide whether to issue a point projection or a more qualitative estimate (e.g., a bounded range), and whether to project interim or annual earnings or both. Our empirical tests assess differences in the information content of management earnings forecasts that differ by form and horizon. Our tests provide a comprehensive investigation of the price effects of these alternative forecast disclosure types. While an extensive literature exists on the relation between management forecasts and stock prices, most previous studies examine only point and range forecasts of annual earnings (e.g., Penman 1980; Ajinkya and Gift 1984, Waymire 1984; McNichols 1989; Pownall and Waymire 1989). Exceptions include Lev and Penman (1990), Patell (1976), and Baginski et al. (1993). Lev and Penman (1990) include lower and upper bound forecasts for part of their sample period, but do not examine these disclosure forms separately. Patell (1976) provides evidence on mean price changes associated with a sample of annual minimum and maximum forecasts. Baginski et al. (1993) examine alternative forecast forms. Prior analyses of managers' disclosure incentives speculate that investors may condition their assessment of forecast information on disclosure form and horizon. For instance, King et al. (1990) suggest that forecast disclosures emerge as voluntary managerial actions to reduce costly information asymmetry in capital markets. Under the ''expectations adjustment'' hypothesis, managers have incentives to acquire and maintain a reputation for credible disclosure. Rational investors recognize that disclosure quality varies systematically by disclosure form and will discount qualitative projections or those issued with longer horizons. Policy debates on mandatory disclosure of qualitative information, such as the recent SEC debates over the content of ''Management Discussion and Analysis'' disclosures, and deliberations on forecast disclosure in the 1970s (see King et al. 1990), also suggest a need for evidence on the information content of qualitative prospective disclosures and alternative forms of forecasts.1 Our primary tests are based on a sample of 1,252 forecasts disclosed by 91 firms between July 1, 1979 and December 31, 1987. Several conclusions emerge from these tests. First, forecast disclosures remain highly informative even when including other disclosure types not analyzed in prior studies. Second, forecasts are less informative than earnings announcements for our full sample, a finding that is inconsistent with earlier results in Pownall and Waymire (1989). Third, differences across forecast forms are not significant at conventional levels. Fourth, interim forecasts are significantly more informative than annual projections. This result is driven largely by maximum forecasts, which are highly informative and more frequent in the interim forecast subsample. We document several additional regularities that may be of interest to researchers. First, point and range annual forecasts comprise less than 20 percent of our sample. This suggests that the incidence of voluntary management forecast disclosure is possibly far greater than suggested by previous studies. Second, range forecasts tend to be quite inaccurate ex post. Actual earnings per share (EPS) fell outside the forecasted bounds in more than 50 percent of our range forecasts. Third, forecasts that are more qualitative tend to be issued over longer horizons. Minimum forecasts are issued over the longest horizons for our sample, and interim point projections have the shortest horizons. Finally, extensions to our primary tests provide some evidence that maximum forecasts have significant negative price effects, and that for point forecasts, forecast revisions are highly informative.