In the wake of the 2008/2009 financial crisis, a number of policy reports (Vickers, Liikanen, Volcker) proposed to separate investment banking from commercial banking to increase financial stability. This paper empirically examines one theoretical justification for these proposals, namely attention distortion under multiple tasks as in Holmstrom and Milgrom (1991). Universal banks can be viewed as combining two different tasks (investment banking and commercial banking) in the same organization. We estimate pay-performance sensitivities for different segments within universal banks and for pure investment and commercial banks. We show that the pay-performance sensitivity is higher in investment banking than in commercial banking, no matter whether it is organized as part of a universal bank or in a separate institution. Next, the paper shows that relative pay-performance sensitivities of investment and commercial banking are negatively related to the quality of the loan portfolio in universal banks. Depending on the specification, we obtain a reduction in problem loans when investment banking is removed from commercial banks of up to 12 percent. We interpret the evidence to imply that the higher pay-performance sensitivity in investment banking directs the attention of managers away from commercial banking within universal banks, consistent with Holmstrom and Milgrom (1991). Separation of investment banking and commercial banking may indeed be associated with a reduction in risk in commercial banking.