Bad behavior: personality tests can help underwriters identify high-risk drivers.

Elaine is 45 years old, married with two young sons. She makes $60,000 a year, pays all her bills on time, and thinks of herself as a "pretty good" driver. Elaine has had 10 accidents in the past three years.

Three years ago, none of the conventional rating criteria would have indicated any problem. Her credit score was in the safe range, she lived in an affluent suburb, and her lifestyle was strictly "red state." The insurer rated her as a good risk, but she was an accident--actually 10 accidents--waiting to happen.

By her own admission, she's impatient and aggressive on the road. She often talks on her cell phone and sometimes gets distracted while drinking, eating or even reading while she drives. Nonetheless, she blames other drivers for her misfortunes: They're too slow; they get in her way; they "just don't know what they're doing."

Tom, in contrast, has never had an accident or gotten a citation. He's a patient, cautious driver who never speeds or forgets to buckle his seat belt. Tom pays a sky-high rate. He's single and 21 years old.

Elaine's personality traits are typical of those of bad drivers surveyed in focus groups. Research shows that personality tests of as few as four questions can significantly improve the predictive power of standard underwriting variables such as credit score, gender and age. Not only do personality tests spot drivers like Elaine, who look better than they are, they also identify good risks such as Tom in high-cost classifications.

Correlation vs. Causation 

The history of automobile insurance classification is one of successive refinements. The first wave came alter World War II, when direct writers offered discounts to rural drivers. Cheap computing power and sophisticated data-mining techniques have further refined classification and raised the stakes even more in recent years. Insurers who fall behind in this race risk losing out to competitors whose class boundaries are better drawn.

This search for the perfect classification scheme might be expected to increase rate equity among insureds, but doubts remain. A good system contains relatively homogeneous classifications; that is, the drivers in each classification should generate similar loss costs. The current system works fairly well most of the time, but high-cost classifications are less likely to satisfy this requirement. For example, many young men are safe drivers and can be underwritten profitably at lower than their existing rates, but unless the classification system provides a way to differentiate these young men, they'll be charged the same rate, and one group will subsidize the other.

Correlation does not imply causation. Crashes aren't caused by credit scores--or even by age or gender--but by the way drivers behave behind the wheel. Leonard Evans, former safety director at General Motors, summarizes the results of a half-century of research on this topic when he writes: "There is a large body of evidence showing that behavior is of crucial importance in traffic safety." An Indiana University study cited by Evans in his groundbreaking book Traffic Safety and the Driver shows that more than 90% of crashes are caused at least in part by driver behavior, as opposed to poor highway design.

The first study to provide, as Evans puts it, "specific evidence of a strong link between broad personality" characteristics and crash involvement" was published in 1949. Since then, hundreds of other researchers have confirmed this link. Note that these researchers were not measuring driver performance (a measure of skill), but driver behavior. Evans maintains, "Driver performance is not the driver characteristic which has the largest influence on driver safety." One study shows that presumably highly skilled race car drivers have higher crash and violation rates than other drivers. Traffic safety is a function not of what a driver is able to do, but of what he or she chooses to do.

Another researcher in the field, Dr. Steve Stradling, spent seven years studying driver attitudes and behavior and their link to accident involvement, while part of the Driver Behavior Research Group at the University of Manchester in England. In their research, Stradling and his colleagues concluded that driver attitude is a critical factor in determining future accident involvement. Stradling aptly summed this up in releasing one of the reports of their studies: "We think it's drivers with bad attitude that are most likely to be crash-involved and most likely to act as crash magnets. Insurance companies knew about the age and gender relationship. But what we think we've done is identify the psychological mechanism involved."

Of course, the knowledge that behavior causes crashes is of little value to insurers unless they can identify crash-causing behaviors and somehow single out those insurance applicants most likely to display them. Research in this field has produced convincing evidence that this can be accomplished using standard personality tests. An article in the journal Accident Analysis and Prevention reported on the results of a study using personality tests to predict accident involvement by young men conscripted into the Finnish Army. The authors found that adventurous and impulsive behavior were emblematic of high accident rates, and demonstrated a correlation between accident proneness and worrying, poor self-confidence and moodiness. Good drivers were conscientious, Self-controlled and persistent.

Taking off from the work of Stradling and the other academics, the authors of this article set out to bring that research into the day-to-day world of property/casualty underwriting and rating, and more specifically to bring causation into that process. As part of its research and development work, RightFind Technology administered a series of questionnaires to thousands of drivers in the United States and ran a series of focus groups designed to elicit personality traits associated with drivers who had high insurance claim rates. The results are remarkably consistent and revealing: drivers who display a certain set of personality traits are more likely to have accidents and file insurance claims than those without those traits.

Based on the research, a number of underlying attitudes or personality traits were identified that correlate with claim involvement:

* Entitlement. Bad drivers think there's no reason they shouldn't speed or run red lights "as long as we're not hurting anyone."

* Impatience. Drivers with high claim frequency were "always in a hurry." They often feel stressed and pressured in a "too crowded" life.

* Shifting blame. A majority of drivers with bad records thought themselves better-than-average drivers, and blamed others' lack of skills for their own accidents.

* Aggressiveness. These drivers want to be the first to move when a red light turns green, the last to stop when a green light turns red. They'll cut others off and are much more likely to be involved in arguments with other drivers.

* Distractibility. These drivers admit that loud music, cell phones, eating, drinking, smoking and reading often are distracting.

Implementing Personality Testing

Building on the research, tests were developed to measure these traits and add predictive power to standard underwriting criteria. There are many ways such tests could be simply and efficiently administered by insurers to their applicants. A test with as few as four questions has been shown to add significant value to the underwriting process, and these questions could be asked and graded in less than two minutes by a call-center employee. A brief questionnaire might be given in pencil-and-paper form as part of a written application for insurance, but these tests will be more effective when computerized since the answers would be transmitted to the insurer in real time and combined with the rest of the applicant's information to arrive at a rate quote.

The computer also makes test construction easier. Unlike a pencil-and-paper test, for which the questions and their order are determined in advance, a computer can choose questions based on an applicant's prior answers, allowing a few questions to do the work of many more. Since questions will not be directly related to driving behaviors, and since many of the personality markers for high risk driving are not intuitively related to bad driving, tests can be constructed to avoid "gaming."

Personality testing will never replace traditional rating variables; however, it may very well represent the next level of refinement, giving a competitive edge to early adopters and leaving the doubters scrambling to survive.

Key Points

* Personality tests can significantly improve the predictive power of underwriting variables such as credit score, gender and age.

* Drivers who display a certain set of personality traits are more likely to have accidents and file insurance claims than those without those traits.

* Entitlement, impatience and aggressiveness are among the attitudes or personality traits that correlate with claim involvement.

Contributors Donald Bashline, Peter Hiam and Peter Robertson are principals of RightFind Technology Co. LLC in Boston. Bashline is an actuary, Hiam is a former Massachusetts insurance commissioner and Robertson is an industry lawyer.