A Case Study of Unfair Discrimination within Auto Insurance Pricing Models

Session Number

Project ID: CMPS 36

Advisor(s)

Dr. Frank Quan, University of Illinois Urbana Champaign

Discipline

Computer Science

Start Date

17-4-2024 10:00 AM

End Date

17-4-2024 10:15 AM

Abstract

In the insurance industry, pricing algorithms are often closely guarded as trade secrets. So, when Allstate Insurance publicly disclosed detailed information about its new auto insurance pricing algorithm, it drew significant attention. It is worthwhile to meticulously examine the data, and further uncover if there are disparities that disproportionately affected consumers. Considering the national presence of Allstate Insurance, inspecting its pricing model could provide valuable insights into the broader landscape of auto insurance pricing models. Using the state-of-the-art correlation method, Chatterjee’s Correlation Coefficient, and Python programming, we conducted an analysis of Allstate’s public data in conjunction with Maryland Census data. This analysis revealed that there is a moderate positive correlation between the race of the consumer, their assigned risk level, and their suggested price increase, as determined by the pricing model. These results suggest that pricing bias is potentially a concern for the auto insurance industry and that greater regulation and oversight should be put into creating pricing algorithms to mitigate bias.

Share

COinS
 
Apr 17th, 10:00 AM Apr 17th, 10:15 AM

A Case Study of Unfair Discrimination within Auto Insurance Pricing Models

In the insurance industry, pricing algorithms are often closely guarded as trade secrets. So, when Allstate Insurance publicly disclosed detailed information about its new auto insurance pricing algorithm, it drew significant attention. It is worthwhile to meticulously examine the data, and further uncover if there are disparities that disproportionately affected consumers. Considering the national presence of Allstate Insurance, inspecting its pricing model could provide valuable insights into the broader landscape of auto insurance pricing models. Using the state-of-the-art correlation method, Chatterjee’s Correlation Coefficient, and Python programming, we conducted an analysis of Allstate’s public data in conjunction with Maryland Census data. This analysis revealed that there is a moderate positive correlation between the race of the consumer, their assigned risk level, and their suggested price increase, as determined by the pricing model. These results suggest that pricing bias is potentially a concern for the auto insurance industry and that greater regulation and oversight should be put into creating pricing algorithms to mitigate bias.