Many accounting variables represent estimates of future outcomes. For example, the bad debt expense, pension liabilities, and depreciation expense all reflect managers’ estimates of the future. In fact, many accounting variables are labeled as estimates, such as earnings forecasts and estimated contingent liabilities. Therefore decision-making under conditions of uncertainty is of fundamental importance in accounting. When individuals make predictions they must form beliefs about unknown parameters (Von Neumann and Morgenstern 1967). This entails estimation risk. Estimation risk matters because it alters the behavior of risk-averse individuals (Klein and Bawa 1976). This study shows that estimation risk has a large impact on the degree of conservatism exhibited by risk-averse auditors. We focus on the estimation risk associated with predicting bankruptcy because auditors can incur substantial losses from litigation and reputation impairment when their clients file for bankruptcy (Palmrose 1987; Carcello and Palmrose 1994; Lys and Watts 1994; U.S. House of Representatives 1985, 1990; Firth 1990; Weil 2001; House of Lords 2011). These losses mean that it is important for auditors to form accurate expectations about the likelihood of bankruptcy. Estimation risk arises because the probability that a client will file for bankruptcy is not known. Instead, the probability must be estimated. The level of estimation risk captures the degree of imprecision in the point estimates of bankruptcy. We argue that estimation risk is likely to have a significant impact on auditor behavior because auditors are risk averse. To clarify the distinction between bankruptcy risk and estimation risk, it is helpful to introduce some notation. Let B* represent the risk that a company will file for bankruptcy in the future, where B* is uncertain and therefore must be estimated. Let B^* represent the best point estimate of B* . We measure estimation risk using the standard errors of the point estimates, (SE B^*
), because the standard errors reflect the precision of each point estimate. In particular, B^* is a less precise measure of the company’s bankruptcy risk when SE B^*
is larger, i.e., when there is more uncertainty surrounding the point estimate. When estimation risk (SE B^*
Þ is higher, it is more probable that the point estimate (B^* ) will turn out to be incorrect ex post, which imposes incremental risk upon risk-averse auditors. To illustrate, consider two companies, Y and Z, that have the same point estimates of bankruptcy risk (i.e., B^*y = B^*z), but company Z has higher estimation risk, i.e., SE B^*z
> SE B^*y
. The higher estimation risk implies greater uncertainty in predicting Z’s future bankruptcy status than predicting Y’s future bankruptcy status, even though both companies have the same point estimates of bankruptcy. This uncertainty is expected to be important in explaining auditor behavior because auditors are risk averse. 1 Risk-averse auditors will respond differently to companies Y and Z because Z has greater estimation risk, even though they have the same point estimates of bankruptcy risk. We expect that estimation risk causes risk-averse auditors to behave more conservatively. In particular, we predict that auditors respond to greater estimation risk by a) issuing more going-concern opinions, b) resigning more often from audit engagements, and c) charging higher audit fees. We test these predictions after controlling for the point estimates of bankruptcy (B^* ) to ensure that our estimation risk variable (SE B^*
) captures the incremental effect of estimation risk on auditor behavior.