Corporate Governance and Risk-Taking: A Statistical Approach

Recent bank failures have spurred widespread demands to impose greater penalties on corporate managers that engage in excessive risk-taking. This is not surprising; politicians and the media tend to attribute virtually every dramatic business failure to excessive risk-taking or fraud.

In part, these responses reflect at least two cognitive biases: hindsight bias, the tendency to believe that a past event was predictable or inevitable; and ultimate attribution error, in this context, the tendency to assign responsibility for a failure to individuals, as bad actors, rather than to external factors. The result is that legislatures, including Congress, often react to failures by enacting laws that focus on preventing excessive risk-taking (and fraud) by imposing harsh managerial performance standards, without addressing the actual causes or consequences of the failures.

My article, Corporate Governance and Risk-taking: A Statistical Approach (available at http://ssrn.com/abstract=4542464), makes three related claims about corporate risk-taking. Prudent corporate governance requires managers to take business risks, much of which is data-driven and statistically based. Although excessive risk-taking and fraud cause some corporate failures, even good faith statistically based risk-taking can result in failure. The article’s first claim, therefore, is that managers should not automatically be presumed to be responsible for corporate failures that result from risk-taking decisions based on statistical methodologies that reasonably justify the decisions ex ante.

Conceptually, the business judgment rule (“BJR”), which generally exempts non-conflicted managers from liability for making good faith decisions that have a reasonable basis, already should protect managers from liability for these corporate failures. Jurisdictionally, however, the BJR has a very limited scope, applying primarily to state-law shareholder derivative lawsuits. Furthermore, even where the BJR otherwise applies, it is uncertain whether, in addition to corporate directors, it also protects corporate officers. The BJR thus leaves a large protection gap.

To fill that gap, my article’s second claim is that corporate managers should also be protected by a “statistics-based governance” rule that exempts officers and directors from both federal and state liability for making risk-taking decisions based on statistical methodologies that reasonably justify their decision making (assuming good faith and no managerial conflicts of interest or fraud). Protection under this rule would require full transparency of that evidentiary record, including the statistical methodology and the underlying data and assumptions. A statistics-based governance rule thus would be more objective, and therefore less subject to criticism, than the BJR.

Expected-value analysis is the most generally accepted and widely used statistical methodology for assessing risk-taking outcomes. The article explains expected-value analysis in user-friendly terms and uses it to exemplify statistics-based governance. For most decisions, the expected-value calculation should only take into account the firm and its investors. The article’s third claim, however, is that managers making expected-value decisions should ask, Expected value to whom? For decisions that could significantly impact the public, the answer to “Expected value to whom?” should strive to additionally include the public.

Next, the article tests a statistics-based governance rule by applying it retrospectively to two risk-taking examples: Enron’s risk-taking that resulted in the firm’s bankruptcy, and Ford’s risk-taking that resulted in the exploding gas tank on its “Pinto” car. These applications demonstrate how managers could make statistically based governance decisions.

In the case of Enron, for example, the applications indicate that the managers acted reasonably when deciding, ex ante, to engage in special-purpose-entity hedging transactions in compliance with reasonable corporate processes and with the help of the firm’s outside counsel and accountants. The applications also suggest that the liability standards of congressional legislation enacted in reaction to Enron’s failure were unrelated to the actual causes and consequences of that failure.

In the case of Ford, the applications show that the managers may have contorted statistics-based governance to avoid making the requisite safety modifications. Furthermore, Ford’s managers improperly equated human life with dollars. Even though governments sometimes weigh the value of lives when balancing costs and benefits, there should be an ethical distinction between government and private action that harms third parties.

The foregoing applications also help to explain other limits of statistics-based governance. For example, although such governance can help to balance the outcomes that may result from a decision, it does not necessarily take into account the possibility that alternative decisions might yield more favorable outcomes. Business managers and other decisionmakers cannot, however, realistically be expected to take all alternatives into account. They often must decide based on the choices before them. The possibility of missing preferable alternatives is thus a widely known and accepted imperfection of any decisionmaking, including when making cost-benefit-analysis decisions.

Another limitation can occur where a decision could lead to an action that violates law. Clearly, managers should not consider actions that violate criminal law. But should they consider actions that merely require paying civil monetary penalties as a cost of doing business? The article examines that question, taking into account such indirect costs as reputational cost, lower worker morale, potentially weaker investor demand, and the costs of defending against an investigation and prosecution.