Predictive analytics is basically the analysis of large data sets (big data) to make inferences or identify meaningful relationships, and the use of these relationships to better predict future events. Predictive modeling tools have the potential to enable insurers to address some of the concerns resulting in the low penetration of life insurance among millennials—complex and invasive underwriting process, costly and time-consuming process from time of application to when a policy is issued, etc.
While life insurers are noted among the users of statistics and data analytics, they clearly have not integrated predictive analytics to improve the insurance buying process in the same way that other businesses (e.g., property and casualty, banking and marketing) have. The property and casualty industry, for example, utilizes generalized linear models, credibility techniques and, more recently, credit scoring models as part of its modeling techniques for driving business decisions.
Life insurance plays the important role of protecting households and individuals from the financial effects of uncertain mortality. Over time, life actuaries have developed good estimates of life expectancy in the form of mortality tables that mirror aggregate insured population mortality, while underwriting techniques assist in assessing the relative risk of an individual. Though these traditional techniques have been widely accepted across the industry, the standard life insurance underwriting process is still time-consuming and quite costly.1A typical life insurer spends about a month and several hundred dollars underwriting each applicant, which clearly translates into higher insurance premiums.
In response to some of these long-standing concerns, the life insurance industry has clearly made efforts toward streamlining some of the underwriting and sales processes. A number of these improvements include simplified applications for smaller face amounts and the refinement of underwriting requirements based on protective value studies.
Though these are good efforts toward a more robust and streamlined life insurance sales and underwriting process, there’s clearly more to be done to embrace the revolution in predictive analytics and business intelligence. Property and casualty insurers are further along in developing analytics-based approaches to underwriting for better, faster and more efficient decision-making. Studies have shown that the life insurance industry is actually at the point where predictive modeling should be utilized more in its sales, underwriting and customer retention processes.2
The good news is that a few life insurers are beginning to explore the possibilities of using predictive modeling and automated life underwriting in their business processes,3 and we only expect the numbers to increase over time.
The late arrival of the life insurance industry on the scene was a result of a number of modeling challenges, which include the target model variable and the volume of data required for modeling. Because life insurance is mostly sold through long-duration contracts (e.g., 20 or 30 years, or even over the lifetime of the policyholder), it makes it difficult to have a clearly defined target variable. There’s also the possibility that some of the risk factors predicting mortality might change over time.
Also, life insurance claims usually have low frequency per year, and this poses significant challenges to modeling. Unlike for auto insurance, where approximately 10 percent of drivers make a claim per year, life insurers can typically expect about one death in the first year per 1,000 policies issued,4 making it very difficult to model any statistically significant variation in mortality.
To work around these challenges, a number of life insurers are beginning to work with a closely related and yet more immediately viable proxy target variable—the underwriting decision on a new policy issued having characteristics that address both concerns of modeling mortality.
Though costly and time-consuming, underwriting plays an important role in the life insurance process. To be able to generate more value from this process and break the barriers preventing enough penetration among the millennials, predictive modeling can be used to provide a streamlined process to give an early indication of the likely underwriting result. Thus, predictive modeling can be utilized not to make the final underwriting decision, but rather to triage applications and provide suggestions as to whether additional requirements are needed before making an offer.
- 1. According to the Deloitte 2008 Life Insurance Operations Expense Study (LIONS) benchmarking study of 15 life insurers, the median service time to issue a new policy ranges between 30 and 35 days for policies with face amounts between $100,000 and $5 million, and the average cost of requirements (excluding underwriting time) is $130 per applicant. ↩
- 2. Deloitte, Predictive modeling for life insurance—Ways life insurers can participate in the business analytics revolution, April 2010. ↩
- 3. An SOA sponsored study in 2009, Automated Life Underwriting, states that 1 percent of North American life insurers surveyed are currently utilizing predictive modeling in their underwriting process. ↩
- 4. This is an estimate based upon industry mortality tables. Mortality experience varies across companies with insured population demographics. In the 2001 CSO table, the first-year select, weighted average mortality rate (across gender and smoker status) first exceeds one death per 1,000 at age 45. ↩