This is a very interesting conundrum. The amount of data now available on individuals allows each policy to be underwritten based on very specific information–placing each individual into their own personalized “risk bucket.” The irony here is that people that had previously benefited from being in a favorable risk pool (based on zip code, county, insurance score etc.) could now be paying much higher prices when individual data is taken into account. Mark my words, this will inevitably have a disproportionately negative affect on low-income individuals. I think a lot of people will be forced to admit that traditional insurance “rating buckets” were really not that bad. Nevertheless, we are in the information age and this form of rating is the future–like it or not.
As we gather more data on specific risks we can see that those traditional risk buckets in general were correct all along. The people in the more high claim, high risk zip codes already had worse driving records and worse credit. By using credit and claim based rating what we’ve done is further penalized those in the “risky” territories by having more data points in which to penalize them. If we’re talking about “fairness” in underwriting we need to be focused on past behavior that’s specifically insurance related to project future behavior. That means driving record on auto insurance and claims history across all lines as being the number one driving force. Unfortunately that isn’t going to happen because carriers have figured out how to cater to the people with the most to insure.
This is a very interesting conundrum. The amount of data now available on individuals allows each policy to be underwritten based on very specific information–placing each individual into their own personalized “risk bucket.” The irony here is that people that had previously benefited from being in a favorable risk pool (based on zip code, county, insurance score etc.) could now be paying much higher prices when individual data is taken into account. Mark my words, this will inevitably have a disproportionately negative affect on low-income individuals. I think a lot of people will be forced to admit that traditional insurance “rating buckets” were really not that bad. Nevertheless, we are in the information age and this form of rating is the future–like it or not.
As we gather more data on specific risks we can see that those traditional risk buckets in general were correct all along. The people in the more high claim, high risk zip codes already had worse driving records and worse credit. By using credit and claim based rating what we’ve done is further penalized those in the “risky” territories by having more data points in which to penalize them. If we’re talking about “fairness” in underwriting we need to be focused on past behavior that’s specifically insurance related to project future behavior. That means driving record on auto insurance and claims history across all lines as being the number one driving force. Unfortunately that isn’t going to happen because carriers have figured out how to cater to the people with the most to insure.
Regulators need to use AI to keep an eye on how the private sector is using AI.