Using Confidence Levels to Quantify the Uncertainty in Management Range Forecasts

ABSTRACT We use statistical properties of interval forecasts to show that management range forecast width is a function of two distinct factors: uncertainty about future earnings and the statistical confidence level (i.e., the ex ante likelihood that earnings will fall within the forecasted bounds). Conceptually, the notion of forecast confidence levels is missing from the management forecast literature. To demonstrate the importance of confidence levels when interpreting forecast widths, we develop a measure of the implied confidence level associated with each forecast and show that confidence levels, as opposed to uncertainty, explain the majority of variation in widths. We also identify several managerial traits and incentives that affect forecast widths through confidence levels. Finally, we show that conditioning width on confidence levels improves forecast width as a measure of uncertainty about future earnings. Our study demonstrates the importance of confidence levels for understanding the information content of forecasts widths. Data Availability: Data are available from the public sources cited in the text. JEL Classifications: G41; M21; M40; M41.