Motivated by competing theories on the properties of earnings required for compensation performance measurement, we provide direct evidence on the properties of actual accounting earnings that are used in determining compensation payouts (Compensation Earnings). Using a large sample of manually collected Compensation Earnings for U.S. firms, we show that firms make economically significant adjustments to GAAP Earnings in arriving at Compensation Earnings. While GAAP Earnings exhibit conservatism, we fail to detect conservatism (either by statistical significance or by magnitude of coefficient) in Compensation Earnings using the same sample and the same research design. The absence of conservatism in Compensation Earnings is also documented in various subsamples partitioned on market-to-book ratio, leverage, firm size, and corporate governance. Further analyses indicate that the adjustment from GAAP Earnings to Compensation Earnings involves the removal of less persistent components of GAAP Earnings, resulting in Compensation Earnings that are more persistent than GAAP Earnings.