The study on the potential for "non-traditional" data in insurance risk modeling described in this article will be very interesting and it makes perfect sense to test this data. Indeed the need to find ways to predict risk in groups with little or no credit history is something being worked on in banking as well as insurance. That said, not only do I doubt it will move the insurance industry to any substantial change in its current underwriting and pricing strategies, I think the study is ill-timed given the combination of data breach / privacy bills pending at both the state and national levels. While credit data is shown to be a good predictor of risk (see this post) and is used as a key part of many companies risk assessment, it gets a bad press. Recent legislation in California to restrict what data can be used to estimate risk also mitigates against this study’s success. The study will clearly lack data from California (as Prop 103 does not allow credit as a key pricing variable) and perhaps other key states as well. This means the study could have a hard time determining whether specific immigrant and protected classes could benefit from the inclusion of alternate data. If this were not enough, EPIC (Electronic Privacy Information Center) recently announced a $50 million settlement against a bank for using Drivers License Data in marketing. EPIC and others may well scream about this study also.
So I hope this will deliver proof that one can use good math to trump bad judgments about those with little credit history but I fear it will not make it out of the starting gate.