Predictive Analytics: Bringing Price Elasticity Concepts to P/C Insurance

Published By : Carrier Management

The concept of price elasticity of demand has not received enough attention in the world of property/casualty insurance.

Regulatory hurdles, a tradition of cost-plus pricing and maybe even an outright aversion to the word “discrimination” combine to hold insurers back from reaping the full benefits of applying some of the basic economic principles that are commonly used in other industries—industries that charge consumers different prices for the same product or service.

Microeconomic theory teaches us that thoughtful selection of prices, or price discrimination, is a key to maximizing revenue and profit. Our research, in fact, reveals that if P/C insurers adopt advanced pricing strategies that consider customer elasticity differences, they can boost their revenues by roughly 3 percent and returns-on-equity by 1 percent, on average.

Price elasticity of demand (PED) essentially refers to the responsiveness—elasticity—of a customer in terms of the quantity of a product he or she will buy when the price of that product changes.

The airline industry provides a recognizable example of how a pricing strategy can be developed based on an understanding of price elasticity. Airlines get higher prices from business travelers with inelastic demand than from vacationers who shop for deals and show flexibility on departure times.

Likewise, all types of retailers offer discounts to senior citizens for their products, because seniors with more flexible schedules are price-elastic.

Some P/C insurers exploit the relative inelasticity of policyholder demand (on a limited basis). Insurers, for example, may account for elasticity differences when applying single-digit discounts to multi-year auto policyholders, even when their loss experience suggests steeper discounts are possible. Business insurers also may offer modest package discounts when the data suggests their loss experience could support a more substantial package-policy bonus.

In other words, they’re increasing their own revenues and profits rather than sharing savings with customers.

Such practices, however, are intuition-led and not typically based on rigorous analytics.

We believe that Big Data and predictive analytics offer new avenues for research and application of PED concepts in insurance. In addition, we will propose one way in which P/C insurers can make PED concepts work within the existing regulatory framework.

Price Discrimination Anathema to Insurance

“Discrimination” is a loaded word, but what we’re discussing here is segmenting consumers based on their behaviors, not by characteristics that insurance regulations identify as unfair, such as race or religion.

Still, P/C carriers are rooted in a cost-plus pricing world—borne from the data we capture, the relative size of variable costs over fixed costs, and most notably from actuarial and regulatory principles that emphasize rates that are “reasonable and not excessive, inadequate or unfairly discriminatory.”

Outside of the United States, price differentiation practices are more advanced than the package and multi-year discounts discussed above. In fact, in the U.K., car insurers are permitted to change rates at will—on a daily basis if they so desire. Only race and gender are considered off limits, although insurers can and do market by gender. One British insurer is even called “drive like a girl.”

Despite regulatory hurdles, we are pleased to report that PED does have a role in the U.S. personal and commercial insurance marketplace. Measures of PED cannot be applied to the individual as they might in the U.K. market, for instance, where the time of day a purchase is made might result in different rates for otherwise identical risks.

In the United States, we must apply these principles at the cohort level in a manner consistent with accepted rating variables.

The Power of Big Data

To begin to understand their customers’ price elasticity, insurers will need to capture new data. While it is true that most insurers today don’t have the same robust customer-behavior data that an online retailer might have, that picture is changing as more consumers buy—or at least shop for—their auto policies online before making a purchase.

The data now being captured at the insurer’s website can provide useful insight into consumers’ behavior.

Once it is captured, price elasticity will manifest in at least two traditional ways:

  • Switching behavior exhibited when inefficiencies in pricing enable customers to find lower premiums or more favorable combinations of prices and coverages.
  • Reduction in consumption in response to a change in price. A policyholder who raises his deductible or lowers his limit in response to a price increase is revealing something about his price elasticity. We refer to this behavior as “rate avoidance.”

Readers will recognize the latter as a common aspect of the property-catastrophe reinsurance market, where insurers regularly change their property-catastrophe program in response to rate changes.

Insurers will—and many already have—grasped these concepts. But through the power of predictive analytics and our ability to marry these behaviors with other forms of data, we can systematically look for patterns and correlations that can provide insights not visible through intuition or informed judgment alone.

Another benefit available to insurers that study PED on their own books of business relates to improved business and financial forecasting. Underwriters may reason that a 10 percent rate increase, for instance, will drive away some business. But predictive analytics may help to refine the likely percentage, or even to examine the impacts of other changes that may mitigate policy attrition.

Regulation Limits But Doesn’t Preclude PED Application

It is possible to use PED concepts while working within the limits established by insurance regulation. We turn to the example of personal auto insurance pricing to illustrate how price elasticity principles can be applied.

Most auto insurance rates are derived by applying rate relativity factors to a base rate. Those relativity factors will vary based on gender, age, credit score and so on. Importantly, the statistical analysis used to estimate the relativity factors produces a range of outcomes that within a certain confidence interval should be equally acceptable to the regulator. We believe that auto insurers will benefit by applying PED in their relativity selections. Our work shows that PED relativity selections from relativity ranges estimated with 95 percent confidence can improve revenue and profit outcomes by single-digit percentages.

Assume that through careful study of an insurer’s data, we can determine that females of a certain age exhibit less price elasticity than male drivers of the same age or females in adjacent age categories. Without being overly prescriptive, we could tweak the established rating relativity variables—within their accepted confidence intervals— to arrive at a different rate for those drivers.

In other words, carriers can apply PED at the cohort level—working within the existing framework to incorporate PED into already accepted pricing categories.

For a business, attributes including credit score, the industry it competes in and the number of years it has been in business are among those that might provide insight into PED. In turn, these could be used in rating algorithms to improve an insurer’s revenues.

While we have provided two examples of behavior that could be used to measure price elasticity—switching behavior and rate avoidance—there are actually many more.

We believe the incorporation of PED principles into insurance ratemaking is an idea whose time has come.