Performance Incentives within Firms: The Effect of Managerial Responsibility


  • Rajesh K. Aggarwal,

  • Andrew A. Samwick

    Search for more papers by this author
    • Aggarwal is at the McIntire School of Commerce, University of Virginia, and Samwick is at Dartmouth College and the NBER. We thank Sheri Aggarwal, Matt Clayton, Syd Finkelstein, Bob Gibbons, Jay Hartzell, Lorin Hitt, Leonhard Knoll, Margaret Meyer, Michael Weisbach, seminar participants at Boston College, Columbia, Dartmouth, Michigan, Oregon, Purdue, Vanderbilt, the American Finance Association Meetings, the American Economic Association Meetings, the NYU Accounting and Finance Conference, the UNH Spring Finance Conference, and the Western Finance Association Meetings, and especially Rick Green and the referee for helpful comments. We also thank Sarah Leonard, Yevgeny Neginsky, Evan Walsh, and Paul Wolfson for exceptional research assistance and Andy Halula for assistance with the ExecuComp database. This is a completely revised version of our 1999 NBER working paper of the same title. Financial support from the American Compensation Association and the Presidential Scholars Program at Dartmouth College is gratefully acknowledged. The views expressed are solely ours and do not reflect the views or opinions of the American Compensation Association. Any errors are our own.


We show that top management incentives vary by responsibility. For oversight executives, pay-performance incentives are $1.22 per thousand dollar increase in shareholder wealth higher than for divisional executives. For CEOs, incentives are $5.65 higher than for divisional executives. Incentives for the median top management team are substantial at $32.32. CEOs account for 42 to 58 percent of aggregate team incentives. For divisional executives, the pay–divisional performance sensitivity is positive and increasing in the precision of divisional performance and the pay–firm performance sensitivity is decreasing in the precision of divisional performance. These results support principal–agent models with multiple signals of managerial effort.