We examine whether transient institutional investors (i.e., institutions that trade actively to maximize short‐term profits) have information that allows them to predict a break in a string of consecutive quarterly earnings increases and thereby avoid the economically significant negative stock price response associated with the break announcement. We show that transient institutions predict the break at least one quarter in advance of the break quarter. We also provide evidence that is consistent with transient institutions obtaining information regarding the impending break from private communications with management.

Using a sample of privately held C corporations and S corporations from the motor carrier industry during 1984–92, we assess the effect of the 1986 Tax Reform Act on the amount of corporate income shareholders of privately held C corporations shifted to their personal tax bases. We estimate that the C corporations shifted a mean of $130,587 taxable income each year to shareholders (representing 29% of their mean accounting earnings before income shifting) after the 1986 tax law change. The C corporations used deductible managerial compensation and rent expense, but not interest expense, to shift income to shareholders.

This study offers evidence on the earnings forecast bias analysts use to please firm management and the associated benefits they obtain from issuing such biased forecasts in the years prior to Regulation Fair Disclosure. Analysts who issue initial optimistic earnings forecasts followed by pessimistic earnings forecasts before the earnings announcement produce more accurate earnings forecasts and are less likely to be fired by their employers. The effect of such biased earnings forecasts on forecast accuracy and firing is stronger for analysts who follow firms with heavy insider selling and hard‐to‐predict earnings. The above results hold regardless of whether a brokerage firm has investment banking business or not. These results are consistent with the hypothesis that analysts use biased earnings forecasts to curry favor with firm management in order to obtain better access to management's private information.

We assess the impact of Regulation Fair Disclosure (Reg FD) on the trading behavior of transient institutional investors in the quarter prior to a bad news break in a string of consecutive earnings increases. Bad news breaks are defined as breaks that are by growth firms, preceded by longer strings of consecutive earnings increases, followed by longer strings of consecutive earnings decreases, and associated with larger declines in earnings. Pre–Reg FD transient institutions have abnormal selling of stocks in the quarter immediately preceding a bad news break. This abnormal selling is confined to firms that hold conference calls in the pre–Reg FD period. However, in the post–Reg FD period transient institutions do not exhibit similar abnormal selling of stocks in the quarter before a bad news break. Furthermore, after Reg FD transient institutions allocate less of their stock portfolios to conference call firms relative to non–conference call firms in the quarters prior to a bad news break. These results demonstrate that Reg FD has had an impact on management's selective disclosure behavior and significantly changed the trading behavior of transient institutions.

We develop a state‐of‐the‐art fraud prediction model using a machine learning approach. We demonstrate the value of combining domain knowledge and machine learning methods in model building. We select our model input based on existing accounting theories, but we differ from prior accounting research by using raw accounting numbers rather than financial ratios. We employ one of the most powerful machine learning methods, ensemble learning, rather than the commonly used method of logistic regression. To assess the performance of fraud prediction models, we introduce a new performance evaluation metric commonly used in ranking problems that is more appropriate for the fraud prediction task. Starting with an identical set of theory‐motivated raw accounting numbers, we show that our new fraud prediction model outperforms two benchmark models by a large margin: the Dechow et al. logistic regression model based on financial ratios, and the Cecchini et al. support‐vector‐machine model with a financial kernel that maps raw accounting numbers into a broader set of ratios.