Symmetry in Pay for Luck

Naveen Daniel is the Denis O’Brien Research Scholar in Finance at Drexel University’s LeBow College of Business; Lily Li is Research Assistant Professor at the Temple University Fox School of Business; and Lalitha Naveen is Associate Professor of Finance at Temple University. This post is based on their recent article, forthcoming in the Review of Financial Studies. Related research from the Program on Corporate Governance includes Lucky CEOs and Lucky Directors by Lucian Bebchuk, Yaniv Grinstein and Urs Peyer (discussed on the Forum here) and Golden Parachutes and the Wealth of Shareholders by Lucian Bebchuk, Alma Cohen, and Charles C. Y. Wang (discussed on the Forum here).

Are CEOs of public corporations rewarded for good luck but not penalized to the same extent for bad luck? Previous studies have found this to be the case, and have termed this “asymmetry in pay for luck.” Some studies find that this asymmetry in pay for luck is stronger in firms with weaker corporate governance, and take this as evidence that CEO compensation is not optimal. Given the intense scrutiny and debate on CEO compensation, it is critical for researchers to understand the extent to which contracts are set optimally. In a recent article titled Symmetry in Pay for Luck (forthcoming in the Review of Financial Studies), we re-examine this issue. We find no asymmetry in pay for luck, either on average or in subsamples of firms with poor governance.

Our interest in revisiting the prior result of negative asymmetry stems from our observation that researchers have tremendous degrees of freedom in choosing the appropriate specification to test for asymmetry. To better understand this, consider the 2-step methodology used in prior literature to examine asymmetry.

In the first step, regression analysis is used to separate firm performance into performance attributable to “luck” (for example, firm is doing well simply because the industry is doing well) and performance attributable to CEO skill. In the second step, regression analysis is used to determine whether CEO pay changes asymmetrically with luck (estimated from the first step). Specifically, is the pay increase that CEOs receive for performance improvements due to good luck greater than the pay decrease they receive for performance declines due to bad luck? As noted above, some prior studies find such negative asymmetry. Further, this asymmetry appears to be more pronounced in weakly governed firms, leading researchers to conclude that CEOs in such firms are rewarded for good luck but partially insulated from bad luck.

We believe that this conclusion is premature. The problem here is that, across the two regressions, there are at least 17 decisions that researchers need to make. For each decision, there are several reasonable choices. In our study, we consider several specifications based on: (i) alternative methodologies to decompose performance into luck and skill (executive-specific regressions, pooled regressions), (ii) alternative luck factors (industry return, market returns, Carhart four factors), (iii) alternative pay measures (pay change, level of pay, or rate of change in pay), (iv) alternative performance measures (stock returns, accounting returns, combination of both), (v) alternative performance evaluation periods (current year’s performance, 12-month performance preceding the largest equity grant, 3-year performance), and (vi) alternative subsamples (based on governance, compensation, time-periods, industry groups).

Overall, fewer than 2% of the 205 specifications we explore show significant negative asymmetry (at   the 0.10 level). This result is in contrast to prior literature that finds strong evidence of negative asymmetry.

How do we reconcile our results with prior literature? While we are able to replicate earlier studies that document asymmetry in pay for luck if we use their exact specification, we find that this result does not hold out of sample, is not particularly robust to changes in specification, and seems to be caused by extremely large (top 1–2%) firms.

Overall, our primary contribution in this paper is to document that there is no asymmetry in pay for luck. Our finding is important given that the literature widely accepts the idea of asymmetry in pay for luck (in weakly governed firms), and typically points to this as evidence of rent extraction. Testing for asymmetry in weakly governed firm is intuitively appealing. Researchers need to be careful, however, about using the right specification and test for robustness given the wide latitude of choices available.

The complete article is available for download here.

Both comments and trackbacks are currently closed.