We study a model of managerial incentive problems where a manager chooses the first two moments of his firm's profit distribution - mean and volatility - along an efficient frontier. Assuming that managers differ with respect to their marginal cost of effort and their risk aversion we explore our model's comparative statics predictions in full detail. If managers' preference parameters are commonly known and associated, then a positive correlation between expected returns, volatility of profits, and incentives is the natural outcome. Allowing in addition for adverse selection with respect to the managers' preference parameters does not change the predicted correlation if the variation in observed contracts is not too large. Moreover, observed incentive schemes reflect exclusion of some manager types. Neglecting the endogeneity of risk in empirical studies biases estimates towards zero.