SCORviews article shared potential considerations when resetting existing underwriting criteria cutoffs within a preferred class underwriting knock-out methodology. While this is a worthy goal for companies using this approach, some companies have switched to either debit-credit systems or more sophisticated predictive modeling schemes.
These ever-evolving nontraditional underwriting methodologies have overcome many of the deficiencies inherent in simple knock-out systems. But from a practical standpoint, is the industry ready for an
en masse movement to a new way of classifying preferred risks?
How We Got Here
Much of today’s preferred criteria came from clinical research studies into the causes of cardiovascular disease, primarily the Framingham Heart Study. Using these studies, pricing actuaries developed theories relating insurable population distributions to the mortality associated with the definitions of preferred underwriting classes (Figure 1).
Using knock-out criteria, one could determine the proportion of applicants qualifying for a class and then calculate the relative mortality for that class, which allowed us to develop reasonable mortality pricing assumptions before actual life insurance claims experience emerged.
As preferred experience began to emerge in the late 1990s and early 2000s, it turned out that the theories were good at predicting the relative mortality for preferred and residual standard classes. An example from SCOR’s propriety reinsurance mortality experience database is shown in Figure 2.
Figure 2 - Underwriting Class Differentials
Mortality A/E Ratio*|
*A/E ratios are based upon the SOA 2001 VBT
Based upon male and female fully underwritten nontobacco business with exposures from 2004-2010, experience illustrates that knock-out classification systems using mostly cardiovascular criteria are clearly effective in segmenting mortality risk.
Why Does it Work?
Most preferred classification systems, when removed from the sanitary controls of the theoretical environment, suffer at the hands of “the real world.” Market forces skew our expected theoretical distributions. For example, evolving distribution channels, agent selection criteria and underwriting business decisions can have a measurable impact on the A/E ratio. This can be seen in Figure 2 above.
Companies can compensate to a degree by designing underwriting classifications that reflect the company’s agency structure and marketing approach. However, the effects of these remedies remain limited. Therefore it is natural to question why and how these traditional underwriting schemes have served the industry so well over the years.
Many factors contribute to the answer, but let’s focus on two major considerations – one statistical and one underwriting. From a statistical standpoint, preferred classes must contain enough lives to assure emerging experience is credible and capable of providing predictive value.
This is especially important given that pricing margins among preferred risks are relatively thin. In the real world, this often requires on average 30% of the entire pool to be classified as “best-preferred,” which may require some relaxation in our selection criteria thresholds.
From an underwriting perspective, failure to pass the threshold for a single criteria in a knock-out system may disqualify the applicant from the best class. Any individual risk factor’s knock-out threshold in isolation (i.e., keeping all other risk factors constant), may seem overly generous: 90%-95% of the applicant population would pass.
However, the knock-out system introduces the concept of “conditional probability” into underwriting. For example, a single risk factor may pass 90% of applicants. Assuming the same 90% pass rate, the applicant pass rate for the next criteria will be 81% (90% of 90%). The third risk factor passes 90% of the remaining 81% (i.e., 73%) and so on, until we have the desired 30% to 40% of the applicants in the “best-preferred” class.
Additionally, underwriting provides for judgment calls.If an applicant misses a threshold by just a few points the underwriter nevertheless may decide to approve the applicant if other correlated hurdles are well surpassed. Furthermore, other relatively uncorrelated criteria are designed to catch less favorable risks that happen to slip through other risk thresholds.
Can We Underwrite Better?
While we continue to rely on the knock-out system as an industry, research is ongoing in alternative approaches, such as debit-credit systems, predictive modeling, and non-fluid underwriting. However, many of these efforts remain as ongoing research, and companies are hesitant to roll out new risk platforms without sufficient validation.
From a pricing perspective, the relative mortalities based on the theoretical mortality distributions line up well against emerging experience. Still, we used theoretical mortalities because we had little knowledge of how cardiovascular risk factors from general population studies translated to a fully insured population. Today, individual company and industry experience studies reflect true life-insured relative mortality and drive new product premium calculations. An entirely new classification method might mean having to repeat this theoretical/experience process again.
Any movement away from the current (mostly knock-out) preferred classification system could adversely affect the competitive landscape. This is especially true if the new system dramatically alters which risks are assigned to which underwriting class.
Currently, most companies perform basically the same process to classify preferred risks and in the process largely minimize antiselection across the industry. A change in the status quo could produce a dramatic (albeit temporary) shift in sales distributions and also create market confusion among agents and consumers. No company wants to be the first to make a change – or the last.
As American patriot John Hancock is quoted in the musical 1776, “Either we all walk together, or together we should stay where we are.”