Case Study 2 — The Impaired Life: How a Table Rating Turns a "Decline" Into a Profitable Policy

This is a clearly-labeled composite. It is built from real, standard life-underwriting practices — table ratings, time-graded ratings for treated conditions, postponement pending stability, the whole-person read — assembled into one teaching file. The applicant is fictional; the techniques are industry-real. No figure here is a real person's or insurer's data; the round numbers are illustrative. The complementary angle to Case Study 1: where that study showed the frontier and its dangers, this one shows the century-old craft working exactly as intended on a hard file — and where even good craft reaches its limits.

Background: the applicant a naive system declines

"Marcus Bell" [composite] is 52, applies for \$750,000 of 20-year term to cover a mortgage and protect his family, and on paper looks like a decline to anything that simply tallies negatives. The application and the ordered evidence assemble a file with real impairments:

• Type 2 diabetes, diagnosed roughly six years ago.
• A heart attack four years ago, treated with a stent, followed by cardiac rehabilitation.
• A build on the heavy side of the chart.
• A former smoker who, by the records, quit more than three years ago.

Run through a box-checking lens, that is four marks against — diabetes, a cardiac event, build, a smoking history — and a thin-skinned underwriter (or a crude model with a low threshold) might decline outright or quote a rate so high it amounts to a decline. Many applicants in Marcus's position believe, wrongly, that they are simply uninsurable. The case is about why that belief is usually false, and what the disciplined underwriter does instead.

The insurance issue: classify the trajectory, not the diagnosis list

The chapter's central craft — read the whole person, weigh trajectory and stability, and price rather than refuse an elevated risk — does all the work here. The underwriter orders an APS (this is exactly the file that justifies the cost) and reads it for the things the diagnosis list cannot show:

• The diabetes. What matters is not the label but the control: the APS shows an HbA1c that is elevated but reasonably managed, an age at onset that is not alarmingly young, and no evidence of the serious complications (kidney, eye, nerve, vascular) that turn diabetes from a ratable condition into a decline. A controlled diabetic with no complications is a textbook table rating, not a decline.
• The cardiac event. Here the governing variables are time since and recovery: a heart attack four years ago, successfully stented, with completed rehab, good current cardiac function, normal blood pressure on therapy, and no events since, is a very different risk from one six months old and still being worked up. Life impairment manuals are explicitly time-graded — the debit for a cardiac event declines as the survival horizon lengthens and stability is demonstrated. The four years of uneventful recovery are not a footnote; they are the single most favorable fact in the file.
• The smoking history. Quit more than three years ago, with a negative cotinine, Marcus very likely earns nonsmoker rates — the residual mortality of a former smoker is far below a current one, and the durable, demonstrated cessation (§17.1's "credit durable change, not the brand-new turnaround") is exactly the kind underwriting rewards.
• The build adds a debit, but the underwriter asks the §17.5 question — is it part of one metabolic cluster with the diabetes? Likely yes, which means it should be weighed with the diabetes as one story, not stacked as a wholly separate strike.

What it shows: the table rating is the tool that says "yes, at the right price"

Summing the manual with judgment — not blindly — the file lands at a substandard table rating (in illustrative terms, somewhere in the Table 4-to-Table 6 range, perhaps with a temporary flat extra that steps down as the cardiac history ages further), at nonsmoker rates. Read that outcome against the chapter's themes:

• Pricing follows risk (Theme 4), literally. Marcus is not refused the pool; he is admitted at a premium that reflects his genuinely elevated — but quantifiable and stable — mortality. At, say, roughly two to two-and-a-half times the standard premium, the policy is real protection for his family and a properly priced risk for the insurer. Declining him would have forfeited a profitable, fairly-priced policy to a competitor willing to do the work.
• Adverse selection cuts both ways (Theme 2). The table-rating system is precisely what prevents the opposite error — writing Marcus at standard rates, which would let his elevated mortality drag down the experience of the standard lives and quietly subsidize his risk at their expense. Fair classification protects the standard pool as much as it serves the impaired applicant.
• Underwriting is judgment (Theme 1). Every favorable read here — control over diagnosis, time-since over label, durable cessation, build-with-diabetes as one cluster — is a judgment a crude tally misses. This is the craft the whole book teaches, applied to a file a naive system throws away.

Outcome and the limits even good craft hits

Written well, Marcus's policy is a steady, adequately-priced risk and a family protected — the system working as designed. But the case also honors the book's discipline of naming limits:

• Some files are a postpone, not a rate. Had the cardiac event been six months old and still being worked up, the right answer would have been postpone — wait for stability — not a table rating, because the mortality is not yet calculable (Chapter 1's insurability criterion). Knowing the difference between "ratable now" and "not yet assessable" is itself the skill.
• Trajectory can reverse. The level premium is locked for twenty years; if Marcus's diabetes decompensates or a second cardiac event occurs, the insurer cannot re-rate — which is exactly why the current control and the demonstrated stability had to be real, not hoped-for, before binding.
• The model-and-judgment boundary. A modern mortality model would likely rate this file reasonably well, because the variables (controlled diabetes, stented MI four years out, ex-smoker) are well-represented in the data. The human value-add here is smaller than in the David Okafor case — and recognizing that, too, is judgment: knowing when to trust the model and when the case turns on context it cannot see (Theme 5).

The lesson

The table rating is the single most underappreciated tool in life insurance. It is the mechanism that lets the industry say "yes, at the right price" to a vast middle range of human health that a box-checking system would simply decline. The applicant who believes a diabetes diagnosis or a past heart attack makes them uninsurable is usually wrong — if an underwriter reads the trajectory, weighs control and time-since and stability, and prices the mortality that is actually there. And the same discipline that writes Marcus also knows when to postpone instead, and when a model has already done most of the work. Good life underwriting is not the art of saying no to impaired lives; it is the art of pricing them correctly.

Discussion questions

1. List the four "negatives" in Marcus's file and, for each, name the single piece of evidence that determines whether it is a decline, a table rating, or nearly neutral. Why does the diagnosis label alone decide none of them?
2. Explain why writing Marcus at standard rates would be an adverse-selection failure that harms the standard pool — not a generous favor to him.
3. The underwriter used a temporary flat extra that steps down as the cardiac history ages, plus a table rating. Why is a stepping-down flat extra the right tool for the recent-cardiac-event component specifically, rather than simply a higher permanent table?
4. Change one fact: the heart attack was six months ago and still being worked up. Why does the right answer flip from "table rating" to "postpone," and which Chapter 1 insurability criterion explains the change?
5. Compare this file to David Okafor's (Chapter 17 main text). In which file does human judgment add more value over a predictive model, and why? What does your answer say about where underwriters should spend their scarce attention (Theme 5)?