|
|
|
In My Opinion
Intelligence Isn't Everything
by Paul E. LackoMy mother-in-law gave me a copy of a book called The Bell Curve: Intelligence and Class Structure in American Life a few onths ago. She saw it at a book sale and thought I might find it interesting. I do, and I think other CAS members will, too. I don't mean to write a full-blown book review here; that would require much more space than this column affords me. Here's a thumbnail sketch: it's a thick book, well-documented, full of tables and charts and appendices. The authors, psychologist Richard J. Herrnstein and political scientist Charles Murray, examine the relationships between intelligence (as measured by generally recognized standard tests of cognitive ability) and education, sex, race, income level, and other characteristics of the broad U.S. population. The book raised quite a commotion for a short while after it was published in 1996, because the authors argue that we should question many strongly held beliefs about domestic social policy, particularly with respect to the U.S. educational system.
The authors hypothesize that intelligence is the single most important explanatory variable for some important and significant changes in society over the last hundred years. They posit that our society has richly rewarded people of high intelligence as the economy has grown more technical and more technology-dependent, and that the distribution of income in the broad population is now determined primarily by intelligence.
CAS members can appreciate this book on both personal and professional levels. We're an intelligent group, probably averaging above two standard deviations above the mean intelligence level of the U.S. population. We actuaries are compensated well for the work we do, which reflects the cognitive and technical requirements of the job. We tend to associate with other people of high intelligence at work and at home. We tend to live in neighborhoods where other highly intelligent people live. Since we see people like ourselves everywhere we tend to go, we can draw the erroneous conclusion that we make up a good-sized, representative sample of the U.S. population. To the contrary, we are members of the privileged small minority of a society that has become segregated by intelligence.
As you and I well know, however, intelligence isn't everything. In a group of highly intelligent people, such as the CAS membership, intelligence is not such an important distinguishing characteristic, the way it was, say, in high school. Intelligence alone does not guarantee a passing grade on any of the CAS exams, for example. Highly intelligent people do not necessarily have highly developed people skills. Of the best and the brightest among us, the best aren't always the brightest, and vice versa.
The authors put considerable effort into explaining and clarifying this phenomenon. The authors show that intelligence correlates highly with exam scores, managerial success, and attained level of corporate responsibility across the entire population. At any given level of intelligence, however, we will see a great deal of variation among individuals. This is the residual variation that cognitive ability does not explain. If we restrict our sample to people who score at the upper end of the distribution of cognitive ability, then almost all the variation we observe among individuals is the residual variation that cognitive ability cannot explain.
You knew that, of course, having been thoroughly grounded in statistics, credibility theory, and risk classification principles.
Can we run across this sort of situation in property/casualty insurance ratemaking? Suppose your underwriting department wants to specialize and market itself as a niche writer with valuable expertise. Let's say that the personal lines auto underwriters are graded very positively if new business is heavily weighted towards married, white-collar drivers, ages 25-35, who live within ten miles of their workplaces, and have good credit records. In other words, the underwriters are writing as much as they can in one class, as little as they can in all other classes. Is there any point to setting up rate tables by age, marital status, distance from work, and credit score? Can you simply file one rate per territory plus a merit rating plan? Is the one big class the only one with sufficient credibility for establishing stable rates? Will the underwriters complain that their loss ratios vary wildly from one quarter to the next, and who cares about long-term averages when my annual salary review depends on quarterly results? What kind of underwriting expertise will your company gain if the rating variables cannot explain the variation in losses among the individual insureds?
An example relating to capital adequacy also comes to mind. I have read reports in the insurance trade press recently about calls to revise or increase Risk Based Capital (RBC) standards. But what if current RBC standards are effectively flagging the insurers who could be running into financial trouble for reasons that the RBC formula measures? Then the insolvencies we've seen over the last few years happened for reasons that the RBC formula cannot measure or explain. How can raising the RBC requirements reduce the number of insolvencies that the RBC formula cannot identify?
Ultimately, what I like most about reading The Bell Curve is that it gets me thinking of questions about the limits on actuarial knowledge and how we can recognize the boundaries when we bump against them.
