I estimate CEO pay-for-performance schedules for two purposes. First, the predictions of several agency and sorting models are tested. Second, the validity of a common observation/complaint about CEO compensation policies is examined. The principal empirical finding is that CEOs of firms that are prone to high (stock-market) performance volatility receive compensation schedules that lie entirely above the schedules of other CEOs. This shows that the high levels of pay cannot be compensation for bearing more risk. Hazard models show CEOs of high volatility firms also have lower probabilities of turnover.