Combine statistical data with experienced analysis for the best risk predictions

An article in the Sloan Mgt. Rev., Fall 2010 at 51, contrasts what it calls the “frequentist” method for risk management (also called “objectivist”) with the Bayesian method (also called “subjectivist”). Frequentists predict risks based on probabilities obtained from repetitive historical data. For example, in a legal department an objectivist would assess the likelihood of risk by looking at the incidence of cases of a certain kind, or billings by law firms, or claims that are resolved before litigation. They believe the physical world gives you all the data you need to quantify risk.

Bayesians “consider risk to be in part a judgment of the observer, or a property of the observation process, and not solely a function of the physical world.” A Bayesian takes trend data and complements it with judgment and experience. On trends of particular cases, this might mix in understanding of changes in the law or rules of civil procedure or an aggressive cabal of plaintiff’s lawyers. For billings of law firms, the Bayesian would consider seasonality, changes in partners at major firms, and sales of units of the company.

According to the authors, the objective view has three major shortcomings. With its reliance on historical data it does poorly when addressing issues where historical data is lacking or misleading. Second, it allows little input for judgment. Third, it can produce a false sense of security because it seemingly implies scientific accuracy, and thus may encourage excessive risk taking.

The subjectivists more readily adjust to changed circumstances. The downside of the Bayesian approach, however, is that cognitive (unintentional) and motivational (intentional) biases distort probability judgments. The authors mention two techniques for increasing the accuracy of judgments: expert interviews and prediction markets. Most general counsel are closet Bayesians.