Prior research finds that agents who are compensated on an imperfect measure of performance tend to distort their operational decisions and believe that the measure is more reflective of reality than it actually is (i.e., they surrogate). I find that agents distort decisions less and surrogate less when they can provide unverifiable narrative explanations for their actions. In my experiment, experienced chess players place bets on and write explanations about in-progress chess games. I manipulate whether participants give their explanations to their boss (who allocates a subjective bonus) when working as an agent. Participants who give explanations are more likely to make bets that reflect all dimensions of a chess position, rather than making bets that favor only the objective measure. They are also less likely to surrogate and are more likely to write about unmeasurable aspects of the games when subsequently making bets on their own behalf.

We investigate whether and how a “critical audit matter” (CAM) disclosure affects managers' real operating decisions in two contexts (issuing a loan that decreases versus increases the average risk profile of loan portfolios, or choosing to hedge versus speculate on commodity risk). We expect that a CAM disclosure increases disclosure costs and implies expanded auditor support for both types of activities, but we expect implied auditor support to be valued more highly for risk-increasing than for risk-decreasing activities. As a result, we predict that a CAM disclosure decreases managers' risk-decreasing activities (due to increased disclosure costs) more than managers' risk-increasing activities (as the implied auditor support counteracts the increased disclosure costs). We find evidence consistent with our prediction across multiple experiments. Our study sheds light on the unintended consequences of a CAM disclosure and provides insight to relevant parties as the new standard goes into effect.

Researchers frequently proxy for managers’ non‐GAAP disclosures using performance metrics available through analyst forecast data providers (FDPs), such as I/B/E/S. The extent to which FDP‐provided earnings are a valid proxy for managers’ non‐GAAP reporting, however, has been debated extensively. We explore this important question by creating the first large‐sample data set of managers’ non‐GAAP earnings disclosures, which we directly compare to I/B/E/S data. Although we find a substantial overlap between the two data sets, we also find that they differ in systematic ways because I/B/E/S (1) excludes managers’ lower quality non‐GAAP numbers and (2) sometimes provides higher quality non‐GAAP measures that managers do not explicitly disclose. Our results indicate that using I/B/E/S to identify managers’ non‐GAAP disclosures significantly underestimates the aggressiveness of their reporting choices. We encourage researchers interested in managers’ non‐GAAP reporting to use our newly available data set of manager‐disclosed non‐GAAP metrics because it more accurately captures managers’ reporting choices.

Agents often inflate measured performance by distorting operating decisions (e.g., real earnings management) and/or reporting decisions (e.g., accruals management). Across four studies, we find that public judgments of distortion’s acceptability largely reflect assessments of how harmful and norm-violating the distortion is. Judgments of operating distortion primarily reflect assessments of harm, whereas judgments of reporting distortion primarily reflect assessments of norm violation. These results are consistent with the Theory of Dyadic Morality (Gray, Waytz, and Young 2012; Schein and Gray 2018). We also find that those who perceive an accounting system as more unfairly withholding an agent’s bonus assess distortion (especially reporting distortion) to be less norm-violating. Those who perceive the performance measure as less appropriate for capturing the value of performance to stakeholders assess distortion (especially operating distortion) to be more harmful. Assessments of distortions’ harm and norm violation explain a substantial portion of the variation in acceptability judgments.

Advisors frequently have an interest in the decisions their advisees make, forcing advisees to distinguish their advisors' unbiased beliefs from their self-interested bias. This task is likely to be especially hard when psychological forces distort advisors' beliefs to make some of their bias sincerely held. In our first experiment, we show that advisors bias both their recommendations and their own actions toward their persuasion goal, and that advisees are better at distinguishing between the unbiased, sincerely biased, and insincerely biased parts of their advisor's recommendation when they meet face-to-face to discuss, compared with when they receive only a written recommendation. Our second experiment shows that advisees distinguish their advisor's bias from their advisor's unbiased beliefs more accurately when the advisors are asked to provide fact-based information about their own actions. Both experiments show that post-report interactions are more helpful for identifying insincere bias than sincere bias. Data Availability: All raw data (excluding identifiable information), data processing code for tabulated analyses, and full experimental materials are available from the authors.