CEO ownership and rm value: Evidence from a structural estimation

I estimate a dynamic model in which a board of directors and a CEO interact to set the levels of CEO incentives and e¤ort. The intent is to understand the e¤ect that CEO incentives have on …rm value. In the model incentive levels are the result of CEO risk aversion, the cost of CEO e¤ort, the e¤ect of CEO e¤ort on …rm value, the volatility of shocks to …rm value, and the preference a board has over giving a CEO equity ownership. Model estimates show that CEO e¤ort is an important component of …rm value, and that CEO’s exert a substantial amount of e¤ort, on average 94.7% of the possible maximum. A one percentage point increase in CEO ownership is found to increase …rm value 7.6 basis points, with the net (of the increase in CEO ownership) bene…t to shareholders being 4.6 basis points. Keywords: CEO Compensation, Dynamic Principal-Agent Model, Structural Estimation * I wish to thank Toni Whited for her guidance and encouragement. I also thank Michael Raith, Boris Nikolov, Cli¤ Smith, Ed Owens, and Paul Nelson for their comments and criticisms, which have greatly improved this paper. Additional thanks to Chet Reeder and Ivan Ivanov for their feedback. I thank the participants in the Simon School AEC 510 Ph.D. student seminar for their comments and discussion. From 1992 to 2007 the average CEO held 3.6% of the equity of his …rm, worth $108 million 2000 dollars. CEOs acquired most of this ownership as a part of their compensation package intended to align their incentives with those of shareholders. This study seeks to quantify the gain shareholders receive from incentivizing CEOs. It also seeks to understand the economic mechanisms behind any such gains. Identifying these e¤ects is di¢ cult because CEO e¤ort is unobservable, and the relation between CEO ownership levels and CEO e¤ort is endogenous. To address these issues, I develop and estimate a dynamic model of interaction between a CEO and a …rm’s board of directors. The …ndings indicate that CEO e¤ort makes an important contribution to …rm value, which is sensitive to changes in the level of CEO ownership. On average, a 1% increase in CEO ownership increases shareholder value by 0.049% after taking into account the increase in CEO compensation. The model estimates imply that CEOs exert a substantial amount of productive e¤ort, so that much larger ownership stakes are required to incentivizing them further. Connecting the level of CEO ownership and …rm value is hard because CEO e¤orts to increase or decrease …rm value are unobservable and because incentive pay is set in equilibrium. The link from ownership to …rm value goes through e¤ort, and not measuring this e¤ort leads to a downward bias of the estimate of the causal relation. For example, consider two …rms, one with an lazy CEO and one with a hard-working CEO. With everything else held constant, because incentive pay is set in equilibrium, in the data we would observe the lazy CEO with high incentive pay, and the hard-working CEO with low incentive pay. However, because the two …rms are otherwise identical, the two CEOs end up exerting the same amount of e¤ort. In the data we would see identical values, and thus no relation between incentive pay and value, even though incentive pay is clearly working in this example. By developing and estimating a structural model, I can identify the e¤ect of ownership on