A colleague passed this link on to me today - Insurers use credit scores for premiums, but is it fair? It's a nice well-balanced article on the use of credit data in insurance risk assessment. There is a fair amount on risk assessment in the insurance section of the blog including this one on using math to replace bad judgment and this one more generally on predictive analytics. There was also a great series of discussion on this over on the blog of the editor of National Underwriter - check out this post, this response and this re-visting.
This debate is interesting at one level because it impacts how consumers' behavior in one arena should be used to impact what they are charged in another. It also illustrates an interesting point for any of you building predictive analytic models. You may be able to predict interesting things about your customers with models designed for something else. If there is some underlying quality of a customer that impacts a model you build it might mean that the model also predicts some other things about your customers.