Recent studies and some policy experts have posited that dividends indicate higher-quality earnings. In this study, we test this conjecture by comparing the dividend policies of firms accused of accounting fraud to those of firms not accused of accounting fraud. Specifically, we examine whether alleged fraud firms are as likely to be dividend payers as non-fraud firms, and whether managers of dividend-paying fraud firms increase dividends at the same rate as managers of non-fraud firms. Our data reveal that dividend paying status is negatively associated with the probability of committing accounting fraud. In addition, we also find that, during the alleged fraud period, the earnings–dividends relation is weaker for the alleged fraud firms relative to firms not accused of fraud. Finally, using propensity score match tests, the data provide evidence that managers of alleged fraud firms increase dividends less often than managers of firms not accused of fraud, consistent with the alleged fraud firms not being able to match the dividend policies of firms not accused of fraud. Overall, our results suggest that dividends, especially dividend increases, are associated with higher earnings quality.