The assumption no table can give you
Through this rung you have built products out of three classic levers — mortality, interest, and expenses — and you met each as something you could, in principle, look up: a mortality table, a yield curve, an expense study. There is a fourth lever that quietly sits beside them and behaves nothing like the others, because it depends not on biology or markets but on what a free human being decides to do. It is lapse: a policyholder simply stops paying premiums and walks away. Its cousin is *surrender*, where the holder of a cash-value policy cashes out for the surrender value. Together they form policyholder behavior, and it can make or break a product more violently than mortality ever will.
Why does it bite so hard? Recall from the universal-life guide that an insurer spends real money up front to put a policy on the books — commission to the agent, medical underwriting, paperwork — and books those costs as deferred acquisition costs, expecting to earn them back slowly over many years of premiums. A policy that lapses in year two has soaked up the whole acquisition cost but returned almost none of it. So the actuary must *guess*, at pricing time, what fraction of policyholders will quietly leave each year. There is no table to look up; there is only experience, judgment, and the humbling certainty that real humans will surprise you.
Why lapse can make or break a product
Here is the part that trips up newcomers: lapses are not always bad for the insurer, and not always good. Their effect *flips* depending on the product and the year. Picture a term policy still in its early years. The insurer has not yet recouped its acquisition cost, so an early lapse is a straight loss — the policy left before it paid its way. But for a profitable block of business late in its life, a lapse can actually *help* the insurer by releasing the reserve held against a policy that will now never claim. The same event — a customer leaving — can be a wound or a windfall depending entirely on where the policy sits on its profit signature.
Worse than getting the *level* of lapse wrong is getting its *correlation* wrong. The dangerous case is when the people who leave are systematically different from those who stay — the same antiselection logic you met in underwriting, now running on the way *out*. On a policy that has become cheap relative to the holder's worsening health, the healthy lapse and the sick keep paying, so the surviving block is sicker than the table assumed and mortality costs climb. This is sometimes called the *lapse-supported* trap: a product priced assuming heavy lapses can look wonderfully cheap, right up until policyholders stubbornly keep their policies and the assumed profits never appear.
Policy loans: borrowing from yourself
There is a gentler way for a policyholder to get at their money without surrendering, and it is woven right into the contract: the policy loan. Because the cash value of a permanent policy belongs, in substance, to the policyholder, the contract lets them borrow against it — no credit check, no application, because the policy itself is the collateral. Interest accrues, and the loan plus interest is simply subtracted from the death benefit (or the surrender value) if it is never repaid. In effect you are borrowing from your own future claim, and the insurer is never really at risk, because it is holding your collateral the whole time.
But a policy loan is a behavioral lever too, not just a kindness. When loan interest rates are attractive, waves of policyholders borrow at once — *disintermediation* — and the insurer must find cash to fund those loans, sometimes by selling bonds at a loss in exactly the high-rate environment that triggered the borrowing. A loan also quietly erodes the protection: an unrepaid loan that grows faster than the cash value can push a policy to lapse, leaving the family with far less than the headline benefit they thought they had. So the modest-sounding policy loan is, for the actuary, one more strand of behavior to model — coupled to interest rates, and capable of hollowing out the very benefit it sits inside.
The illustration: a projection, not a promise
When a person sits across from an agent, they almost never see the formulas you have learned. They see a policy illustration: a multi-year table showing, year by year, the premium they pay, the cash value building up, and the death benefit — often in two columns, a *guaranteed* set and a *non-guaranteed* set that assumes today's generous crediting rates and dividends carry on forever. The right-hand, non-guaranteed numbers are usually the ones that close the sale, and they are precisely the ones that are not a promise.
Whole-life illustration, age 40, $100,000 cover (sketch)
end of premium GUARANTEED NON-GUARANTEED (current
year paid cash value dividend scale assumed)
------ ------- ----------- ------------------------
1 1,500 0 150
5 1,500 3,900 5,200
10 1,500 11,400 17,800
20 1,500 31,000 58,000 <- the number
that sells
The left column is contractually guaranteed. The right column
assumes today's dividends/crediting continue for 20 years -- it
is a forecast, and the future is free to ignore it.An illustration is the products of this whole rung — the mortality, interest, and expense assumptions, the dividend scale, the lapse pattern — projected forward and dressed for a customer. It is genuinely useful: it shows the *shape* of a policy, and the guaranteed column is a real contractual floor. But it is a model, and a model is not reality. If interest rates fall, the non-guaranteed cash value can come in far below the picture; a policy illustrated to be "paid up in year 10" can demand premiums for years longer when the assumed dividends fail to materialize. Honest practice — and, in many countries, regulation — requires the non-guaranteed columns to be clearly labeled and a lower scenario shown alongside the rosy one.
Closing the loop: behavior, products, and reserves
Step back and the whole rung connects. The reserve the insurer holds against a policy is computed from assumptions — and lapse is one of them. Reserve too little because you assumed heavy lapses that never came, and the company is short of cash when those persistent policyholders all eventually claim. Reserve as if everyone stays when many will leave, and you tie up capital you could have deployed. The same surrender value that comforts a customer is a *liability* the insurer must always be able to pay on demand, which is part of why a reserve is never idle cash — it is money standing ready against promises, including the promise to hand the cash value back.
So the honest summary of this rung is that a life insurance product is never just its formula. It is a formula wrapped in human choices — to keep paying or stop, to borrow or not, to believe a glossy table or read its footnotes. The actuary's craft is to price and reserve for those choices with humility, label projections honestly as projections, and remember that the most expensive mistakes in this field come not from getting the mortality table slightly wrong, but from assuming people will behave the way the spreadsheet wishes they would.