Understanding Corporate Governance Through Learning Models of Managerial Competence

The central focus of research in corporate governance has historically been on the problems of controlling managers’ actions. Without minimizing the real-world importance of such control problems, in our paper, Understanding Corporate Governance Through Learning Models of Managerial Competence, which was recently made publicly available on SSRN, we argue that such a focus is incomplete and ignores important factors affecting corporate governance. In particular, it overlooks the crucial element of career concerns: managers care about the inferences that current and future employers draw over time about their abilities from observing their performance.

Career concerns create powerful incentives for managers. Sometimes they are valuable, as Fama (1980) argues, but sometimes they can prove counterproductive, as Holmstrom (1982) and others demonstrate. In addition, because what others will ultimately infer about them is uncertain ex ante, managers are necessarily exposed to risk, which leads them to demand greater compensation. Further, because what is left to learn decreases over a manager’s career, the corresponding uncertainty likewise decreases, which has numerous implications. Finally, how managers are assessed and by whom are endogenous decisions; decisions that are fundamentally about the design of governance structures. In short, there are a number of reasons to expect career concerns to be a major factor in corporate governance.

Of particular note is the importance of career concerns—a learning and assessment perspective more generally—for understanding the dynamics of corporate governance. Such a perspective seems essential for gaining insights into phenomena such as executive selection and turnover; the evolution of executive salaries and other forms of non-contingent compensation; changes in the balance of power between boards and CEOs; and the changes in the volatility of stock returns over time.

Although by no means comprehensive, our discussion of the extant literature nonetheless amply illustrates that this approach has already yielded considerable insights into governance. Yet it is not a spent force: we believe it remains a rich approach going forward. This is especially true for empirical work. A striking feature of the models in this literature is how well suited they are to testing: as a rule, they offer clear time-series and cross-sectional predictions. These predictions occur, in large part, because beliefs about ability are, almost by necessity, more volatile to the current performance of newcomers than of old-timers. In addition, other observables are closely correlated with factors such as the precision of the relevant prior estimate of ability or the precision of the signals available to the market. While some of these insights have already been explored empirically, many remain to be investigated.

One potential shortcoming of the current literature is that it has ignored the well-documented fact that most people are “bad Bayesians”—they fail, in systematic and predictable ways, to update their beliefs in a manner consistent with Bayes’ Rule. As we discussed in detail above, recognition of this fact opens a number of crucial avenues of research, both empirical and theoretical. No doubt the most critical issue in this regard is determining empirically how far real-world belief formation deviates from the Bayesian ideal in the context of governance.

In sum, learning and assessment offers a complementary approach to agency theory for the study of corporate governance; in particular, it is a means to provide a unifying framework to a number of disparate phenomena in governance. Further, it is an approach amenable to empirical analysis and, indeed, a rich empirical literature has emerged that builds implicitly and explicitly on it. Finally, it offers natural ways to incorporate behavioral insights into the study of governance.

The full paper is available for download here.