Life insurance medical underwriting (UW) is the practice of understanding how medical conditions, medical testing, and physical measurements impact mortality risk. Traditionally, in life insurance underwriting, for fully underwritten cases, underwriters have relied on underwriting manuals with predefined guidelines and debit/credit rubrics for medical test results to classify individuals into insurable risk classes and to identify and segment substandard risks. Although these manuals determine the thresholds at which to decline coverage entirely, they do not explore the potential severity of mortality for these declinable risks.
In our previous white paper, Decline Mortality: shape and severity of mortality for declinable life insurance risks, we used Milliman IntelliScript’s Irix® Risk Score 3.0 as a proxy for traditional underwriting to help define the average relative risk and timing of mortality for declinable risks. Following that analysis, this article explores the implied distribution of relative risks for a single declinable individual and highlights the potential volatility introduced by low-frequency, high-severity declinable risks entering a life insurance risk pool.
In our white paper analysis of declinable risk mortality, we leveraged the Risk Score to identify a declinable risk population from a large data set of life insurance applicants. In that analysis we set a threshold using a combination of drug severity and scores of 2.0 and above to define a population of declinable risks. While this is not a definitive distribution of declines, it serves as a proxy for traditional underwriting and allows us to examine the respective actuals to expecteds (A/Es) for these risks compared to the 2015 VBT Unismoke, ALB industry table. Figure 1 below details the total declinable risk life insurance applicant count, exposure years, and death count contained within that analysis.
| Milliman Data | Decline Count | Exposure Years | Deaths | |
| Total | 2,192,564 | 10,011,143 | 120,049 |
Figure 1: Total declinable risk life insurance applicant count, exposure years, and death count
The white paper went on to use this robust sample of declinable risks to calculate relative risks at a credible level for the first 10 policy years at which point the slope of mortality appeared to align best with the 2017 Simplified Issue (SI) S&U table. Using this relative SI slope in durations 11+, we proceeded to calculate the impact of declinable mortality on a present value (PV) basis for given combinations of gender and policy duration resulting in Figure 2 below.
Present Value A/E vs 2015VBT of Decline Mortality
| Term Period | Male, Age 45 | Female, Age 45 | |
| 10 Years | 702% | 837% | |
| 15 Years | 590% | 682% | |
| 20 Years | 521% | 585% | |
| 30 Years | 432% | 466% |
Figure 2: Impact of declinable mortality on a present value (PV) basis for given combinations of gender and policy duration
Further details of the methods and analysis used to define declinable risks and create the above table can be found in the white paper referenced above. For the purposes of this article, we will assume an average present value relative risk of 550% for our reference declinable risk distribution. This corresponds roughly to the unisex PV A/E impact on a 20-year Term product based on Figure 2 above.
To understand the distribution of A/Es for a declinable risk we must first put the mortality impact on a present value basis so that our analysis is not skewed based on the relative duration of the sample data by Risk Score bin. This was accomplished by using a net single premium (NSP) calculation by bin for both expected 2015 VBT death rates by applicant, as well as actual death rates by bin. We then took the ratio of the calculated NSPs to determine the PV A/E by Risk Score bin. The PV A/Es were then vertically shifted such that the weighted average PV A/E was 550%. The results of these transformations can be seen in figure 3 below.
Figure 3: Declinable risks by Risk Score bin (0.2 bin size)
Figure 3 shows a histogram of the declinable risks by Risk Score bin (0.2 bin size), with the recentered PV A/Es shown in grey on a secondary axis along with a best fit trendline based on a power function.
It is clear from this graph that the Risk Score continues to effectively segment mortality risk well into the tail of the declinable distribution and, while the policy counts become small, there are still hundreds of deaths per bin out to scores of 25.0 and at least 25 deaths per bin out to scores <= 40.0, where we have cut off the graph for visual clarity. While the PV A/Es do begin to get choppy near the tail of the graph, they clearly approach 4,000% or 40x the expected NSP cost of mortality of the equivalent Standard risk applicants.
Now let’s assess the volatility introduced by declines in a pool of insured risks. To start, a PV A/E probability density distribution for a single declinable risk was developed by first taking the log of the Risk Scores and re-binning to create more credible data points in the tail of the declinable risks. The resulting points were fitted to an exponential function by minimizing the mean squared errors weighted by application count. Using this fitted exponential function the declinable risk density by PV A/E was generated, as seen in Figure 4 below.
Figure 4: PV A/E probability density distribution for a single declinable risk
With an estimated distribution of the severity of a declinable risk, we can now assess the potential volatility of such risks entering the mortality pool. The current industry average decline misclassification frequency is ~2% within AUW programs, based on our current PRe EMPT™ industry benchmark and other analysis of Post Issue Audits (PIAs) and Random Holdouts (RHOs).
If an insurer experiences a 2% declinable risk misclassification rate and audits 5,000 Accelerated Underwriting (AUW) policies, they will find roughly 100 declinable cases. Using the declinable density distribution in Figure 4, we generated 10,000 random samples of 100 individuals each from this declinable population, and then calculated the average PV A/E for each sample. The observed range of average declinable risk mortality varies significantly based on this heavy right tail distribution. The table in Figure 5 below shows the range of observed average PV A/Es.
| 100 Random Decline | Min | Max | Mean | 5th% CI | 95th% CI | |
| Average PV/AE | 368% | 833% | 554% | 443% | 642% |
Figure 5:Range of observed average PV A/Es for 100 random declines
Based upon the 90% confidence interval in Figure 5, for 100 simulated declines the change in the average relative risks combined with a 2% decline frequency can produce anywhere from -2% to +2% change in the total slippage estimated for an AUW program. This 400bps swing in AUW mortality A/E based on just a small group of declinable risks entering the pool means that insurers must use caution when assessing the impact of declinable risks on their AUW program. They should consider not just the AUW mortality slippage on average but also the volatility of mortality for the resulting insurance block of business. This is particularly true for short-duration term life insurance products with thin profit margins, where failing to account for this increased volatility can quickly lead to unexpected claim patterns and underpriced products. Put differently, the volatility from the long relative risk tail of declinable risks in addition to the inherent variability within AUW programs due to misclassifications may change the underlying actuarial credibility of the mortality used in pricing assumptions today. Additional sensitivity testing, discloser of volatility risk to management, and appropriate credibility weighting of the AUW pricing mortality assumptions should be considered.
The recent evolution of life insurance underwriting has benefits for consumers, streamlines the sales process, and facilitates automation for carriers; however, it also introduces new challenges and brings into question mortality credibility assumptions. Understanding the severity of declines, adjusting mortality expectations based on new data, and addressing increased mortality volatility are critical steps in ensuring accurate pricing and risk management. As underwriting continues to evolve, insurers must remain adaptable—leveraging data-driven insights while also recognizing the inherent increase in risk associated with introducing a new underwriting process.
PartnerRe continues to develop our PRe EMPT™ suite of AUW monitoring and pricing tools that explore the mortality consequences from the rise of accelerated underwriting. Given the limitations of current UW manuals, it is difficult to underwrite and quantify the true severity of individual declinable risks today. As part of PartnerRe’s PRe EMPT™ research, we continue to develop new guidance for quantification of declinable mortality risk that can be shared with our partners. Get in touch with our team to learn how we can support your goals.
Ehren Nagel, Head of Actuarial Innovation, US Life
Abu Bhalla, Associate Actuary, Accelerated Underwriting and Analytics, US Life