This paper uses data on detected misstatements—earnings restatements—and a dynamic model to estimate the extent of undetected misstatements that violate GAAP. The model features a CEO who can manipulate his firm's stock price by misstating earnings. I find the CEO's expected cost of misleading investors is low. The probability of detection over a five‐year horizon is 13.91%, and the average misstatement, if detected, results in an 8.53% loss in the CEO's retirement wealth. The low expected cost implies a high fraction of CEOs who misstate earnings at least once at 60%, with 2%–22% of CEOs starting to misstate earnings in each year 2003–2010, inflation in stock prices across CEOs who misstate earnings at 2.02%, and inflation in stock prices across all CEOs at 0.77%. Wealthier CEOs manipulate less, and the average misstatement is larger in smaller firms.

We estimate linguistic‐based classification models of deceptive discussions during quarterly earnings conference calls. Using data on subsequent financial restatements and a set of criteria to identify severity of accounting problems, we label each call as “truthful” or “deceptive.” Prediction models are then developed with the word categories that have been shown by previous psychological and linguistic research to be related to deception. We find that the out‐of‐sample performance of models based on CEO and/or CFO narratives is significantly better than a random guess by 6–16% and is at least equivalent to models based on financial and accounting variables. The language of deceptive executives exhibits more references to general knowledge, fewer nonextreme positive emotions, and fewer references to shareholder value. In addition, deceptive CEOs use significantly more extreme positive emotion and fewer anxiety words. Finally, a portfolio formed from firms with the highest deception scores from CFO narratives produces an annualized alpha of between −4% and −11%.

We construct a novel measure of disclosure choice by firms. Our measure is computed using linguistic analysis of conference calls to identify whether a manager's response to an analyst question is a “non‐answer.” Using our measure, about 11% of analyst questions elicit non‐answers from managers, a rate that is stable over time and similar across industries. A useful feature of our measure is that it enables an examination of disclosure choice within a call. Analyst questions with a negative tone, greater uncertainty, greater complexity, or requests for greater detail are more likely to trigger non‐answers. We find performance‐related questions tend to be associated with non‐answers, and this association is weaker when performance news is favorable. We also find analyst questions about proprietary information are associated with non‐answers, and this association is stronger when firm competition is more intense.

Formal theory and empirical research are complementary in building and advancing the body of knowledge in accounting in order to understand real‐world phenomena. We offer thoughts on opportunities for empiricists and theorists to collaborate, build on each other's work, and iterate over models and data to make progress. For empiricists, we see room for more descriptive work, more experimental work on testing formal theories, and more work on quantifying theoretical parameters. For theorists, we see room for theories explicitly tied to descriptive evidence, new theories on individuals' decision making in a data‐rich world, theories focused on accounting institutions and measurement issues, and richer theories for guiding empirical work and providing practical insights. We also encourage explicitly combining formal theory and empirical models by having both in one paper and by structural estimation.