Research suggesting the existence of the accrual anomaly runs into the issue that risk serves as a competing explanation for abnormal returns. This paper proposes a novel approach to distinguish between risk and anomaly explanations for the negative association between accruals and returns. The intuition is that high-risk stocks should experience relatively high and low returns more often than low-risk stocks. Thus, a variable that has the opposite correlations with high returns than with low returns is unlikely to capture risk, which points toward an anomaly. The paper implements this perspective via two logistic regressions predicting relatively high and low returns. Controlling for standard risk measures, we document that low accruals increase the probability of large positive returns, but reduce the likelihood of large negative returns. This finding is inconsistent with the prediction that accruals reflect risk and supports the hypothesis that the accrual “anomaly” is indeed an anomaly.