There is limited large-sample empirical evidence in the U.S. that CEO compensation contracts use the accounting performance of peer firms as a benchmark to evaluate the CEO’s own-firm accounting performance (i.e., accounting-based relative performance evaluation, RPE). Given the anecdotal and empirical observation that firms do use their own accounting numbers in determining CEO compensation, the lack of evidence of accounting-based RPE presents a puzzle. My study questions whether the lack of evidence is attributable to a limitation of the conventional empirical research design used to test for RPE. I propose that efficient relative evaluation using peer-firm accounting performance requires that the peer firm has a comparable financial reporting system. Thus, I refine the selection of the peer group by matching the firm of interest with industry-size peer firms with high financial reporting comparability. I find strong empirical evidence of accounting-based RPE in determining CEO’s total compensation when financial reporting comparability is taken into account in peer selection. Consistent with expectations, I also find that accounting-based RPE is used more in determining cash compensation than equity compensation and its use increases in the fraction of CEO cash compensation to total compensation. I further find that, when the fraction of equity compensation declined during the recent financial crisis, the use of own-firm stock returns and price-based RPE in determining CEO’s total compensation declined as expected, but the use of own-firm accounting performance and accounting-based RPE continued to be significant. Overall, my results resolve the apparent inconsistency between the substantial anecdotal evidence and the lack of prior empirical support for accounting-based RPE in determining CEO compensation in the U.S.