This paper illustrates how commonly used measures of risk preferences can lead to biased inferences regarding risk attitudes and firm survival. When risk attitudes are elicited only from current entrepreneurs, the measures are subject to a selection bias which excludes unsuccessful entrepreneurs. When the risk attitudes are collected after entrepreneurial entry, the risk preferences will endogenously reflect the realized success of the business. We illustrate that prior findings of an inverted U-shaped relationship between willingness to take risk and firm survival probability are a result of these biases. Selecting on entrepreneurial success causes an upward bias in measured effects of risk aversion on firm success, while endogenous ex post measures of risk aversion bias the measured effect toward zero. Correcting for these two sources of bias, we find that less risk averse entrepreneurs are more likely to survive, a result consistent with experimental findings.