Predicting, Portending or Analytical Forecasting?

As the title indicates, this month we are going to examine the practice of prediction, and one kind in particular: mortality prediction. Until now, this obscure skill was of interest only to actuaries at life insurance companies and attorneys battling in tort cases. But with the current drumbeat of data reports regarding Covid-19, it’s worth understanding a little bit better. 

Forecasts predicting the total number of deaths from COVID-19 have varied wildly because we lack solid information about the virus. We simply don’t know how many people have had it, how well people are observing social distancing measures - and how long they will be willing to continue this safety precaution. 

"Even though there's a huge amount of resources being poured into modeling... [the forecasts are] going to be wrong," said Irwin Redlener, Professor of Health Policy and Management and Pediatrics at the Columbia University Medical Center, who is currently working on the coronavirus pandemic. "The problem of forecasting is increasingly inaccurate because of the variables, there are many uncertainties with COVID-19."[1]

Jeffrey Shaman, Professor of Environmental Health Sciences at Columbia, who works on developing COVID-19 models, said: "They are projections, not forecasts; they are done for a continually evolving situation for which there is little information on transmission over the last two weeks.”[2]

Projecting the trajectory of how this virus will affect the populace raises serious, important, and expensive questions for which we all seek equally serious answers. Models forecasting death tolls are based on incomplete and rapidly changing information rendering them inaccurate, but they serve a purpose by helping to inform policymakers about resources needed and when lockdown measures might be reduced.

What these scientists and most every government on the planet is trying to predict is the “mortality risk” represented by Covid-19.  For any one person the risk is “X,” which is fine as long as you are not the person who actually contracts it. For any one person it’s very difficult to say how much at risk they are. But the present situation reminds me of just what a miracle plain old life insurance really is. 

On an individual basis, “mortality risk” is hard to calculate, but when working with the general population, it’s much easier. This problem was solved a long time ago through the miracle of risk spreading – more commonly known as “insurance.” 

It may be an unglamorous industry, but the invention of insurance stands as one of the true wonders of the financial world, indeed, perhaps even of modern society. All people walking the earth face risks and there is no way to know upon which particular individual bad luck will visit. But thanks to this mechanism, we don’t have to predict any one individual’s fate, we can employ a “risk pool” to manage that risk.

And for this particular task, we rely on (appropriately-titled) “life insurance” companies. What can they do that we can’t? How is their math any better than ours? Well, their math isn’t any better than ours, but they have two big advantages: they can draw on the “law of large numbers” and they have a gigantic data base of actual lives lived upon which to draw.

The “law of large numbers” lets them change the question from “when will this particular person die” to “how many people this person’s age will die in the next year, the year after that, etc.” Analyzing the duration of lives actually lived by a million people gives them a pretty good idea of how many on average out of a pool of 10,000 won’t survive the next twelve months. Instead of trying to predict the fate of one specific individual, they only have to know the approximate number who will die out of a much larger pool. And this converts their math from “WAG” (wild-ass guess) into a highly reliable algorithm. 

As it turns out, predicting how long a given percentage of the population will live is a surprisingly predictable and low-risk activity – assuming average health. Winners offset losers and the average will reign.

But how do we offload the mortality risk of a single person in poor health? That is a different matter altogether. From a risk management standpoint, the only prudent thing for a life insurance company to do would seem to be to decline that business (as they regularly do with standard life insurance policies). But the math flips in our line of work because settlement annuities are not life insurance. What presents itself as an unacceptably high risk for a life insurance product becomes a potential opportunity when writing settlement annuities. 

This dynamic translates directly into pricing. One of the most powerful benefits of a structured settlement is that we can provide tax-free income guaranteed for the rest of the claimant’s life - they cannot outlive the benefits of the settlement plan.

But, using standard annuity rates to provide a lifetime income for a person with a serious medical condition equates to statistical overpricing. Life insurers know this and employ a correction method known as “substandard underwriting.” Here, they request current medical records describing the person’s condition and have highly trained specialists review them. If they see evidence of a condition which tends to shorten life, they will assign that person a different age for pricing purposes. This adjusted age is called a “rated age” and it allows us to reduce annuity prices accordingly.

Example: Jane, a healthy 20-year-old female living in the United States can expect to live another 60 years to age 80.[3] $ 1,000 per month to Jane for the rest of her life might cost $346,324 and pay out $720,000 in tax-free income over this expected lifetime. This plan has an internal rate of return of 3.02% (which will of course increase if she lives longer). 

But if Jane has some medical condition or illness which diminishes her life expectancy by 20 years, Jane would have the life expectancy (or “rated age”) of a 40-year-old female. That same $1,000 per month for life would now cost only $300,442, a 14% discount which raises the internal rate of return of this adjusted plan from 3.02% to 3.69%. 

Taking this further, if Jane’s condition was grave enough to diminish life expectancy by another 20 years, it would give her the life expectancy of a 60-year-old. $1,000 per month for life now costs only $232,971. This represents a 33% discount off the initial cost of the settlement annuity and bumps the internal rate of return up to 5.06%. 

These examples illustrate cost reductions, but we can also use the rated age to increase benefits at a given level of cost. Given the original annuity cost of $346,324 and “rate-ups” of 20 and then 40 years, we could increase Jane’s lifetime income as follows:

Standard Life Expectancy (age 20)    Rated Age of 40                   Rated Age of 60

$ 1,000 per month                               $ 1,154 per month                $ 1,487 per month

As you can see, rated ages can substantially increase the monthly benefit to the claimant without increasing cost one penny. 

The following conditions may prompt a life company to assign a rated age: alcoholism, cancer, cardiac problems, chemotherapy, cigarette smoking, diabetes, drug addiction, high blood pressure, hepatitis, kidney failure, lung disease, multiple sclerosis, obesity, paralysis, stroke, severe burns. Note: the claimant’s medical condition need not have anything to do with the claim at hand and pre-existing conditions are fully acceptable. 

Do you want to try using a rated age to increase the value of settlement dollars? Contact Frank C. Kilcoyne, CSSC, [email protected] or call 800-544-5533. I am here to help.

[1] https://www.newsweek.com/why-covid-19-death-forecasts-are-wrong-1499243

[2] ibid

[3] Social Security Actuarial Tables https://www.ssa.gov/oact/STATS/table4c6.html. Figure rounded down; exact life expectancy 61.72 years.