CEOs, brokers shed light on “the Black Box” of underwriting

Companies using predictive models for underwriting may be making it more difficult for the public to understand how their premiums are derived, and thus making it more difficult for brokers to explain the rating process to consumers.
CEOs of insurance companies discussed the issue, which was raised by brokers, during the Insurance Brokers Association of Ontario (IBAO)’s 90th Annual Convention in Niagara Falls on Oct. 22.
Referred to within the industry as the “black box” phenomenon, insurance companies are using more sophisticated modeling techniques and individualized ratings variables to price premium more accurately. But as insurers input rating variables into their proprietary algorithms to price their product, it has been more difficult for brokers to explain to consumers how the final output – the consumer’s premium calculation – is derived.
“How does a broker deal with this in the field?” Randy Carroll, CEO of the Insurance Brokers Association of Ontario (IBAO), asked the CEO panel. “You’ve got these tools, you’ve got the algorithms, so you’re working on your back end. But you haven’t taken into consideration the workers in the front end….
“I don’t think the issue is what you’re trying to accomplish, more than it is the issue of how you’ve accomplished it, because you forgot about them (the brokers).”
Carroll’s comment came on the heels of a debate about whether the variables themselves needed to be more transparent, or whether the problem was that the companies’ algorithms were proprietary and hence confidential.
“The transparency part of this is absolutely critical,” said George Cooke, CEO and president of The Dominion of Canada General Insurance Company. “Not the transparency of the algorithms … but the transparency of what exactly is being manipulated inside the algorithm.”
Cooke mentioned this is particularly important in the case of disputed variables such as credit scoring, which are banned in some lines and allowed in others.
Jean-Francois Blais, president and CEO of Axa Canada, said the variables insurers are using are already well-known and “intuitive.” He noted insurers still most commonly use factors such as the age of the driver, how many kilometers a car is driven, the make of the car, etc. – all of which are easily understood by consumers.
“The data has not changed,” said Blais. “What’s changing is the model. Why? What we are trying to do is to fix the price for the years to come, so basically we are trying to predict the future. It’s not like we are building a table based on known costs, build a margin and then sell a policy.”
Louis Gagnon, president of Intact Insurance Company, suggested that brokers in some way stood to gain by the choice created by the complexity of the premium calculations.
“Those algorithms make life complicated, I agree, because they are different [between companies],” Gagnon said. “So when you are talking to a customer, they have different possibilities. It’s difficult to explain all of those things…
How do we marry the science of insurance with the art of setting insurance [rates]? As a broker, I think that’s the most important thing, to have different choices for people, having different companies and different ways of looking at things.”