Abstract Research Summary Strategic management research often uses accounting data, despite well‐known concerns that earnings management could obscure the link between actual and measured performance. We apply methods from the econometric literature on bunching to estimate that around 15 percent of firm‐year observations in Compustat manipulate accounting earnings to achieve profitability. We show that cash‐based performance measures are less susceptible to manipulation and that the choice of accrual versus cash‐based measures “matters” for two classic strategy research questions: a decomposition of ROA variance and an analysis of persistence in firm performance. These findings underscore the importance of robustness testing and contribute to an emerging literature that reconsiders the link between theoretical constructs and empirical performance measures. Managerial Summary To understand what drives firm performance, researchers and practitioners often rely on reported accounting measures from annual reports. The most commonly used income‐based measures are strategically inflated or deflated through accounting and operational choices (i.e., earnings management). This article finds that approximately 15 percent of the time that ROA is reported for a fiscal year, a loss would have been reported instead of a gain if not for this manipulation. We then analyze the impact earnings management has on two types of profitability analysis by comparing accrual‐based or cash‐based profit.
Gibbs et al. (Wed,) studied this question.