With regulators’ renewed focus on potentially discriminatory factors affecting high rates for automobile insurance, the American Property Casualty Insurance Association released a white paper, Behavioral Validation of Auto Insurance Rating Variables, which explains the validity of the most commonly used factors for rating auto policies in terms of their correlation to telematics derived data.
APCIA, the primary national trade association for home, auto and business insurers, noted that auto insurance pricing and underwriting has substantially evolved over the past century, with many of the changes tracking innovations in motor vehicle design and technology, as well as changes in road and traffic management infrastructure, driving patterns and behaviors. The innovations continue to evolve, noted APCIA; telematics, which allows insures to collect data on driver behaviors in real time, represents one of the more recent developments in automobile insurance underwriting and pricing.
Despite the ability to drill down to individual driving behavior, insurers utilize large pools of drivers to share in losses. Noted insurance and risk economist, Robert Hartwig authored the white paper for APCIA. He said, “Because the population of drivers is large and diverse, the best and fairest way to manage the inherent complexity and uncertainty associated with auto insurance pricing is to use a large combination of actuarially sound and independently predictive rating variables.” Hartwig, formerly the chief economist of the Insurance Information Institute, is currently the clinical associate professor of finance, risk management and insurance at the Darla Moore School of Business, University of South Carolina.
Hartwig identified the most commonly used rating factors, which have been used for decades because they have proven highly predictive of future losses. These factors fall into four major categories: policy attributes, driver characteristics, driving environment and vehicle characteristics, and all appear to be confirmed as reliable by telematics data.
Policy attributes include number of drivers, number of vehicles, limits, deductibles, prior lapse and coverages chosen.
Driver characteristics used by most insurers include age, gender, credit behavior, marital status, occupation, education, driving record, moving violations, claims history and miles driven.
Driving environment includes territory or location, whether the vehicle is garaged or street parked, repair costs, medical costs and weather exposures.
Vehicle age, make and model make up the vehicle characteristics.
“Insurers want to make the most accurate risk assessment of each driver, and the use of highly accurate and predictive data helps to achieve that goal,” said Robert Gordon, senior vice president, policy, research and international for APCIA. “If you are a low-risk driver, you shouldn’t have to subsidize insurance rates for high-risk drivers. Those who pose a greater statistical risk should pay more than those who pose a smaller risk.”
“Additionally,” Gordon continued, “we want our products to be affordable and accessible to the largest possible number of people. That starts by doing what’s fair, which is using a large combination of accurate variables that help predict risk. In this way insurers maximize pricing accuracy and assure that no single rating variable has a disproportionate impact on an individual’s premium. This approach to pricing also allows insurers to offer their products to a broader range of consumers and to include incentives that promote safe and responsible driving behaviors.”
Actuarial standards of practice and state insurance regulators also mandate that all rating factors used by insurers comply with stringent requirements documenting strong statistical correlations between rating variables and loss outcomes, said APCIA. While most people have an intuitive understanding of why most variables are used to develop auto insurance rates, there are also some less intuitive variables that have proven to be highly accurate. No matter how intuitive a variable is to the consumer, they all must be strong predictors of future loss.
Among the key findings in Hartwig’s white paper,
Telematics data prove that population density is a highly accurate predictor of insurance cost. Population density can be viewed as a variable that reflects territory, such as urban vs. rural driving. Drivers in areas where population density is the highest are associated with insurance claim costs that are approximately 20 percent higher than the overall population of drivers. Conversely, drivers in areas with the least population density tend to have claim costs that are 20 percent below the overall population of drivers. Telematics data also demonstrate that drivers in areas with high population density are far more likely to engage in hard braking behaviors. Frequent hard braking and hard acceleration have proven to result in more frequent accidents.
Education is another driver characteristic that is highly predictive of loss. Drivers with lower levels of education attainment are associated with insurance claim costs that are approximately five to 10 percent higher than the overall population of drivers. Conversely, drivers with higher educational attainment have claim costs that are at least five to nearly 20 percent below the overall population of drivers.
Telematics data show there is a very clear association between educational attainment and hard braking behavior. Drivers with high degrees of educational attainment engage in hard braking behaviors at least five percent less frequently than the overall population of drivers, while those with lower educational attainment engage in hard braking approximately five percent more frequently than the overall driver population.
Hard acceleration, another risky driving behavior that is correlated with education and loss, occurs at an increased frequency among drivers with lower levels of educational attainment. Hard acceleration behaviors ultimately lead to higher relative costs, says Hartwig.
Occupation has also been found to be highly predictive of loss and is supported by telematics data.
Drivers with certain occupations are associated with approximately five percent to 10 percent higher insurance claims costs than the overall population of drivers. Conversely, drivers with other occupations have claim costs that are five percent to 10 percent below the overall population of drivers.
Occupation is predictive because it influences vehicle usage and the regularity of commutes. For example, realtors generally drive more frequently and often to areas they aren’t as familiar with compared to a schoolteacher who tends to drive the same route to and from school every day. Occupational risk groups are not based on income; they are based on actual claim costs. Low-income and high-income occupations exist within each risk group.
Additionally, U.S. Census data shows that each risk group is represented by people from every race and ethnicity. For example, data from one insurer show that: Bank tellers have lower claim costs compared to psychologists; teachers have lower claim costs compared to financial advisors; enlisted military have lower claim costs compared to pharmacists; firefighters have lower claim costs compared to dentists; receptionists have lower claim costs compared to veterinarians.
Marital status is highly predictive of loss. Married drivers are associated with insurance claim costs that are approximately 20 percent lower than the overall population of drivers. Conversely, single drivers have claim costs that are approximately 15 percent above that of the overall driver population.
Telematics data also demonstrate a very clear association between marital status and hard braking behavior, with married drivers hard braking 10 to 15 percent less frequently than the overall population of drivers. Hard braking behavior is far more prevalent among single drivers than married drivers.
Credit based insurance scores
Credit based insurance scores first began being used approximately 20 years ago because they are highly predictive of loss. Insurer use of credit-based scores has been extensively studied by insurers, regulators, the federal government and academics. Every serious and reputable actuarial study on the issue, including a seminal study in 2007 by the Federal Trade Commission (FTC), has reached the same conclusion: There is a high correlation between insurance scores and the likelihood of insurance claims.
Telematics data show that hard braking behaviors are strongly associated with the credit-based insurance scores and higher risk drivers. The cost to insure the highest risk drivers, those with the lowest credit-based insurance scores is approximately 28 percent higher than for the overall driver population. Conversely, the cost to insure drivers with the strongest credit scores is about 30 percent less than the overall driver population.