Using a financial reporting and valuation model, we investigate the construct validity of Basu's (1997) asymmetric timeliness (AT) regression coefficient as a measure of conditional conservatism in corporate financial reporting. We predict that the AT coefficient will be positive even in the absence of conditional conservatism, and it will vary with non-accounting factors even if the degree of conditional conservatism is held constant. Our empirical analysis shows that AT coefficient estimates vary in directions predicted by our theory. Specifically, we find that AT coefficient estimates increase with expected returns and asymmetry in the distribution of returns, and decrease with cash flow persistence. Importantly, we identify the spread between the variances of bad news and good news accruals as an alternative measure of conditional conservatism that is free of the effects confounding the AT coefficient. Consistent with a key implication of conditional conservatism, we find that the variance of bad news accruals is significantly higher than the variance of good news accruals primarily due to conditionally conservative accruals related to inventory write-downs, long-term asset write-downs, and goodwill impairments. A series of placebo tests provides additional support for the construct validity of our alternative measure of conditional conservatism.