| This dissertation studies the firm's forecast management behavior by taking into consideration the effect of market learning over time. I find that the firm's forecast management behavior and the market reaction to forecast management when I take market learning into consideration differ from those in a static setup. This may occur because the market learns from a firm's history about its propensity to manage forecasts and earnings surprises are the result of the manager's forecast management. Although the market may treat a firm that manages its forecast downward for the first time as an honest one and respond positively to the earnings surprise, the market response may be less positive when a firm's earnings surprise is followed by a downward revision when the firm has done this several times in the last few years. Expecting that the market learns over time, firms will adjust their forecast management behavior over time as well. A rational firm will react to this market learning by not managing sequential forecasts. Using forecast management measures defined below, I find evidence supporting the hypothesis that the market learns about a firm's forecast management behavior from the firm's history.; In this dissertation, I employ several methods to find accurate measures of forecast management. Related prior research has been criticized for using a biased measure of forecast management: the consensus forecast revision. This measure contains three components: forecast management, analyst effect and new information. I develop empirical procedures to disaggregate these components. First, I control for analyst effect by explicitly estimating it. Second, I employ two methods to control and correct for the new information bias: adding a control for new information and an instrumental variables (IV) specification. These methods can mitigate the biases to a different degree and generate better estimates for the effects of forecast management on market reactions. Statistical tests using these measures of forecast management generate results that are consistent with existing conjectures in the literature. |
