Mortality slippage can originate from changes in risk selection and risk classification. Assessing the slippage from the two sources involves very different assumptions. By understanding where assumptions matter most – and where they matter less – insurers can evaluate underwriting innovation more transparently and manage mortality impact more effectively.
In insurance, “slippage” is a deceptively simple word. It suggests something small, incremental – a modest deviation from expectations.
But when underwriting processes change, such as through accelerated underwriting or the incorporation of non-traditional evidence, slippage can represent something far more consequential: the gap between what we expect mortality to be and what it may ultimately become.
Because credible experience data rarely exists at the outset of new underwriting programs, insurers rely on estimation methods that compare new decisions with traditionally derived ones. These methods often produce precise‑looking results – yet those results depend heavily on assumptions that are not always well understood or consistently applied.
This article examines how mortality slippage — also known as mortality impact – is commonly estimated, the assumptions embedded in those calculations, and why understanding those assumptions is essential to evaluating modern underwriting approaches responsibly. If insurers treat mortality slippage estimates as hard facts instead of assumption‑driven signals, they will misprice risk and misjudge the effectiveness of underwriting programs.
The construction of a confusion matrix is based on applying new and traditional underwriting methods simultaneously.
Table 1 presents an example of the cross-tabulated matrix. For demonstration purposes, the risk classifications have been reduced to two categories: low and high. The letters a–i within each cell represent the number of cases assigned to each classification.
Each underwriting class is assigned a relative mortality factor, also known as relative risk (RR), which indicates the anticipated mortality rate for individuals within that group. For example, the standard non-smoking category generally carries an RR of 1, while a substandard Table B classification may have an RR of 1.5.
In the example of Table 1, the relative risks for the three categories are denoted as RR_l, RR_h, and RR_d, respectively. It is important to recognize that these RRs are usually based on historical data from traditional underwriting methods and serve as assumptions intended to reflect actual mortality rates under traditional underwriting.
The most commonly accepted method for quantifying mortality impact is as follows:
For further clarification, referencing the example provided in Table 1, this calculation can be expressed as:
Please note that the computation does not include cells labeled g, h, or i. This represents a particular assumption, which will be addressed in detail later in the paper.
Equations 1 and 2 imply that traditional underwriting outcomes match the actual mortality rates for each group (∑RR Traditional), while new underwriting results correspond to the expected mortality rates (∑RR New). The ratio of these values represents slippage; for instance, a ratio of 110% indicates that the actual mortality rate is 10% higher than what would have been expected under traditional underwriting. This formula and its interpretation are commonly used and cited in the field.
The purpose of this article is to conduct an analysis of the formula and to explicitly identify its underlying assumptions. A thorough understanding of these assumptions offers valuable insight into how slippage may vary should these assumptions be modified.
While this article compares mortality between new and traditional underwriting methods, it is worth noting that factors such as policy placement and lapse also play significant roles in the overall business impacts of modernized underwriting approaches. However, those topics are outside the scope of this discussion.
Underwriting can be broken down into two main parts. The first is risk selection, where underwriters decide to approve or reject cases. The second is risk classification, which involves assigning accepted cases to suitable risk categories. By separating these steps, the confusion matrix becomes simpler and can be split into two distinct cross-classification tables, as shown in Tables 2 and 3.
As in Table 1, the letters A through H in Tables 2 and 3 represent the number of cases in each risk category. RR_accept and RR_decline indicate the expected mortality rates for accepted and declined cases, respectively, while RR_L and RR_H correspond to the low- and high-risk groups. By dividing the underwriting process into two distinct components, it is possible to evaluate the mortality assumptions used in slippage calculations independently.
Mortality impact is defined as the ratio of actual mortality between traditionally accepted cases (A+C) and those accepted under new underwriting (A+B):
Here, A represents the cases common to both methods, so impact hinges on the mortality difference between groups B and C. Group B includes cases rejected by traditional underwriting but accepted with the new approach – the so-called "slippage" cases. It is assumed that their mortality matches all declined cases (RR_decline), reflecting strong confidence in traditional practices. However, the mortality assumption for group C is less straightforward; it varies depending on the assumption selected, each indicating a different level of trust in the new underwriting. Two possible interpretations are detailed below.
For risk selection, mortality impact assessment depends largely on assumptions about the true mortality rates of newly declined cases – that is, cases rejected by new underwriting methods but not by traditional ones, labelled as C in table 2. Two assumptions can be made:
For instance, in our earlier published studies (The Danger of Keeping it Too Simple with Digital Underwriting Evidence and Assessing Mortality Impact of Digital Underwriting Evidence) examining the mortality impact of different digital underwriting evidence, we applied both assumption options and presented the resulting mortality impact as a range.
Table 3 demonstrates how class distributions vary between traditional and newer underwriting methods. Despite these changes, actual mortality remains the same because the pool of cases is unchanged. The difference between the two underwriting approaches lies in their expectations of mortality. From the perspective of risk classification, mortality slippage does not mean actual mortality is rising; instead, it indicates lowered expectations for mortality. Mortality impact is determined as follows:
If a new underwriting method identifies more cases as low risk than traditional methods, overall mortality expectations or premium income may decrease. This effect mirrors higher actual mortality rates, since in life insurance, increased actual mortality can have a similar financial impact to lower expected mortality.
When viewed solely from a risk classification standpoint, the calculation of mortality impact does not rely on any mortality assumptions. This should enhance confidence in mortality slippage assessment.
The calculation of mortality impact shows that changing the risk classification rule or cut-off point in a new underwriting process will also alter the resulting mortality impact. This means that mortality impact is influenced by adjustable classification criteria. For example, using stricter criteria could reduce mortality slippage or even create mortality savings. This differs from risk selection, where missing a decline-able case (such as B in Table 2) is usually an unavoidable part of new underwriting and cannot be changed. In practice, it is important to assess mortality slippage from the perspective of risk selection first, then develop risk classification rules to help adjust the overall mortality impact.
As modern underwriting continues to evolve, so does the challenge of understanding mortality slippage. What appears to be a small deviation often reflects deeper shifts in how applicants are selected and classified.
This analysis shows that when the slippage mostly originates from who is accepted or declined, the industry must rely on assumed mortality for newly accepted and newly declined cases, and reasonable changes in those assumptions can produce a wide range of outcomes.
Separating risk selection from risk classification gives insurers a clearer view of where slippage truly arises and what can be done about it. It also brings needed transparency to the assumptions driving today’s mortality-impact calculations – assumptions that matter more than ever as underwriting practices diversify and experience data remains limited.
RGA works closely with carriers around the world to refine these methods, strengthen assumption frameworks, and design underwriting programs that balance innovation with sound risk management. Contact us today to explore how your organization can put these insights into action and improve the way you evaluate mortality impact.
