Can your credit score and other financial “big data” be used to assess suicide risk, or even predict an attempt or death? #SPSM chats, 11/12/17, 9pCT.
There are multiple historic examples about the relationship between suicide and finances. Consider the US Great Depression, the Asian Financial Crisis of 1997, or the most recently suicide rate increases during the Great Recession of 2008. Financial losses may be the triggering event in a decline in health that ends in suicide. It may also be that changes in credit scores are a digital signal about deteriorating mental health.
What’s interesting about about the relationship between finances and suicide, is that financial matters involve a LOT of digital signals, and it is entirely plausible that these signals could be used to assess suicide risk or predict a death. Of course, like many “big data’ innovations discussed on SPSM, there are a LOT of ethical and legal issues to consider. Join us for the chat, and watch us LIVE here:
Insurers are already using your credit score to predict risk and future behaviors, like your risk of a car wreck. Because it involves large markets, a lot of money, and a LOT of data, the insurance industry is on the front edge of using big data to assess risk and predict behaviors and events that seem rare and “random” (like car wrecks). Check out some choice quotes from the link above:
“The evidence presented in this empirical study of credit scoring and automobile insurance losses is clear: Credit scores predict insurance losses in automobile insurance at a statistically significant level. In fact, they are among the most useful predictor variables available to underwrite and price automobile (and homeowners) insurance. Rationales as to why these predictors work are socio-psychological, behavioral, and biological/biochemical.”
“To produce a “credit score” for an individual for predictive use in insurance, an individual’s credit history file is examined, and a subset of variables is selected from a total array of approximately 450 variables collected in the credit record. Different insurance companies may use different subsets of these behavioral and financial variables and develop different statistical credit score models; however, all generally contain from 10–50 credit history variables that are incorporated into statistical models using insurance losses as the dependent variable.”
“Considerable additional data are being collected by both insurers and others (e.g., credit-scoring firms, GPS firms, social media firms, store loyalty programs) that, combined with new predictive modeling techniques, have the potential to uncover “nontraditional” underwriting variables providing enhanced risk assessments. As the insurance industry advances beyond traditional classification and underwriting variables, the need will increase to justify why accurate prediction works for these “nontraditional” variables and to go beyond simply complying with Actuarial Standard 12 and to verify that any correlation discovered has a basis in fact and is not without an economic, socio-psychological, or behavioral underpinning. The development of a theoretical foundation for why a predictor (such as credit scores or occupation and educational achievement) works can also provide a path for new underwriting variable discovery beyond an ad hoc search.”
“Credit scores are, as we demonstrate empirically, strongly associated with future losses and can be incorporated as an underwriting and classification variable to improve underwriting and loss prediction. They contain behavioral information predictive of loss propensity not duplicated by traditional underwriting variables, yet they remain controversial.”
Controversial enough, that using credit scores actuarial models and pricing is banned in some states. And, of course, because financial data is even more heavily regulated and protected than most health care data, doing peer reviewed research on the relationships between credit scores and insurance risk is pretty hard to do. Just imagine doing this kind of peer reviewed research related to suicide risk assessment and prediction…So, of course, people are finding a way around this:
“According to a 2011 Celent report (Beattie and Fitzgerald 2011, p. 15Beattie, C., and M. Fitzgerald. 2011. Using Social Data in Claims and Underwriting. Celent Industry Trends Report. http://www.celent.com/reports/using-social-data-claims-and-underwriting. [Google Scholar]), “Just as insurers recognize a link between credit health and risk in auto insurance, social data may offer similar insights for insurers who set out to crack the data.” They predicted that social media data use will be incorporated into core underwriting activities in the future. Data mining of social networks and social media are already used in certain areas of insurance. For example, insurers data mine social media to discover fraud in workers’ compensation (NAIC 2012b) and for subrogation negotiations (Kenealy 2013Kenealy, B. 2013. Insurers Finding Ways to Use Social Media in Underwriting, Claims Handling. Business Insurance June 3. http://www.businessinsurance.com/article/99999999/NEWS070109/130609971?tags= |332|65|342#full_story. [Google Scholar]).”
What do you think? Fascinating? Disturbing? Orwellian? Join us 9pCT to chat!