Managing Overconfidence of Actuaries
Are actuaries overconfident? Are you overconfident?
Take the `P/C confidence quiz' to find out!
by Robert F. Conger and Stephen P. Lowe

For years, behavioral scientists have been talking about the need for businesses to manage overconfidence. J. Edward Russo and Paul J. H. Schoemaker have explored this issue in depth in a number of publications, including the Winter 1992 issue of the Sloan Management Review. There they state, "Good decision making requires more than knowledge of facts, concepts and relationships. It also requires meta-knowledge, an understanding of the limits of our knowledge."

Routinely, in all businesses, executives make decisions based on the plans, projections, and forecasts made by their staff. Examples include the forecast for product sales revenue for the upcoming quarter, the projected date a new production facility will be ready to go on-line, the estimated savings from implementing a new software system, or the business plan for expansion into a new region. And because executives can't be involved in every detail of the business, they must rely on the ability of their staff to make realistic assessments in their projections, estimates and forecasts—otherwise the business will suffer the consequences of misinformed decisions.

There is abundant evidence that executives in all industries should be wary. Extensive research performed by behavioral scientists has consistently shown that people are almost universally more confident in their estimates and predictions than the outcomes warrant. For example, in one oft-repeated demonstration of this phenomenon, subjects are asked ten quantitative questions and asked to respond to each question with a numeric range such that they were 90 percent sure that the correct answer would fall within the range. (For example, one of the ten questions might be: "How many German automobiles were sold in Japan in 2001?" Obviously, respondents wouldn't know the precise answer to such a question; their response would be an educated guess expressed as a range based on general knowledge and reasoning.) While one would expect that people would be able to construct the ranges such that on average they get nine out of ten questions right on such a quiz, the reality is that most people fail miserably—typically getting only three or four questions right. Such tests repeatedly demonstrate that most people are inherently overconfident when it comes to their ability to make accurate estimates, forecasts, and predictions.

Behavioral scientists have found that overconfidence is rampant across a variety of industries, from advertising to data processing to security analysis. The relevance of the overconfidence issue to each industry is tied to its need for estimates, forecasts, and projections. For example, Russo and Shoemaker discuss the management of the overconfidence of geologists working to estimate the likelihood of finding oil and gas in a particular area. One can easily see how critical such estimates would be to the future success of energy companies.

Nowhere is the issue of overconfidence more important than the property/casualty insurance industry—particularly in the area of pricing and underwriting. Every day, actuaries and underwriters are required to make critical assumptions about the future costs of the business being written, based on imperfect knowledge. In order to make the best possible decisions, actuaries and underwriters must have a full appreciation of what they know and what they don't know. Overconfidence in the setting of pricing assumptions can have a devastating impact on underwriting results. Better management of overconfidence, by contrast, might help to reduce the underwriting cycle—or at least the degree to which a particular insurer is affected by the cycle.

In future articles, we intend to examine more closely the question of actuarial overconfidence and explore what types of tools can help to manage it, thereby improving underwriting performance. You can help us measure the current level of overconfidence among actuaries, and test your own overconfidence, by taking a ten question "confidence quiz" that draws its questions from the property/casualty insurance industry. Overall results, along with answers to the quiz questions, will be provided in a subsequent issue of The Actuarial Review.

For a lengthier article on this subject, and to take the "P/C confidence quiz" on-line, point your browser to www.tillinghast.com/tillinghast/surveys/confidence .