Is Conservatism Demanded by Performance Measurement in Compensation Contracts? Evidence from Earnings Measures Used in Bonus Formulas

What informational properties of earnings are demanded by compensation contracting for performance measurement? The answer to this question has important implications for accounting standard setting. While some expect conditional conservatism to be a desirable feature that helps to alleviate agency conflicts between shareholders and managers, others suggest that firms trade off conservatism for other important properties such as persistence. We provide a direct answer to this question using a novel dataset of the realizations of earnings that are used in bonus formulas to calculate payout to executives (hereinafter, Compensation Earnings), the disclosure of which is mandated by the SEC after 2006. Compensation Earnings can either be GAAP Earnings or deviate from GAAP Earnings at the board's discretion; considering both cases, we examine the average properties of Compensation Earnings to infer the board's preferences in performance measurement. On average, Compensation Earnings deviate from GAAP Earnings in a distinctly anti-conservative fashion, such that Compensation Earnings are completely removed of conditional conservatism. We find similar results in subsamples with strong corporate governance, suggesting that the removal of conservatism is unlikely driven by poor governance. There is also similar evidence in subsamples with high ex ante agency costs, where one could argue that the demand for conservatism is especially high. In none of the subsamples do we find Compensation Earnings to be conservative. Last, analyses indicate that Compensation Earnings are more persistent than GAAP Earnings. Overall, our results suggest that the board trades off conservatism for persistence in measuring accounting performance for executive compensation