As Japanese life insurers begin to develop new preferred risk products, tobacco usage is one criterion where separate risk classifications can lead to marked differences in the underlying mortality assumption. Although heretofore not distinguished in the underwriting process, and therefore not used in establishing appropriate premium levels, companies are now looking at ways to offer more competitive plans. Distinct tobacco/nontobacco rates is one issue currently being investigated.

In the U.S., smoker/nonsmoker policies have been commonplace for over 25 years. Companies use urine or oral fluid modalities to check for, among other things, cotinine, which is a metabolite of nicotine. Testing limits can be set relatively low because of the cost effective value, both perceived and real, of the tests in identifying impairments which indicate nonstandard mortality. In Japan, on the other hand, testing is not commonplace and is used only when very large amounts of coverage are applied for. With the advent of smoker distinct pricing practices, however, this could change.

A protective value study determines the theoretical breakeven amount at which laboratory or other underwriting tests become cost effective under a given set of assumptions. Essentially, we solve the equation:

Cost Savings

Cost = cost of test, including the kit, collection services, processing and analysis

Savings = (policy face amount) x (savings/1000)

Savings/1000 = R·S·T·PVB

R=impairment prevalence in the insurance-buying population

S=sensitivity of the test; the ability to detect impairments properly in a population of known impairment risks

T=attribution ratio (varies by impairment); how much weight can be given to the test to be the exclusive identifier of the impairment

PVB=present value of excess mortality due to impairment

Persistency starts at 85% in year 1 and grades up to 92% over 4 years. We computed differences in present values of death benefits over 20 years for issue ages in five-year intervals from 25 to 65 for each sex by discounting annual mortality excesses at 8% interest. [A discount rate is required to collapse a series of future mortality savings back to policy issue where the test costs are charged.]

Assuming that the sensitivity of the cotinine test is 100% (or very close to it), and that the prevalence rate of smokers is 55% for males and 20% for females in each age group, then the only other variable left is the attribution ratio, T -- the percentage of applicants who willingly, but temporarily, forget they smoke and deny it on the policy application. There is no way to know in advance what this percentage might be, so we calculated breakeven thresholds based on T=10%, 20%, and 30%. The results are shown in Table 1. Test costs are assumed to be $25 for the purpose of this illustration.

It is interesting to note that the cotinine test is only necessary to give to admitted nonsmokers. It has no value by itself if given to someone who admits his or her habit; in fact, there would be a cost to do so. Thus, in Japan, the total number of tests a company may need to employ for males, for example, might be only half of what would be expected in the U.S.

Note also that the breakeven amounts can be calculated for other values of 'cost' and the variables R and T by simple interpolation or algebraic formula. For example, if 'cost' is doubled, so are the breakeven amounts. If 'prevalence' (R) is halved, breakeven amounts are doubled. If T=15%, the breakeven amount is equal to that shown for T=10.

There is a caveat: the breakeven amounts shown in Table 1 are best used when gauging potential testing thresholds for term insurance plans. For cash value products, the breakeven thresholds would be higher than those shown in Table 1 because the policy's net amount at risk (subtracting the policy's cash value from its death benefit) would get subtracted for the level death benefit. This might increase a whole life policy's theoretical breakeven amount from 5% to 25% between ages 25 to 65. However, I know of no companies in the U.S. that vary testing limits among various policy series.

PVB's are calculated over 20 years and discounted at 8%.