Three routes up the same mountain
In the last guide you met the policy reserve as an idea: the moment a level premium is set by the equivalence principle, the balance between the two sides starts to drift, and the reserve is the money the insurer must hold to cover that drift. That guide told you *why* the reserve exists. This one answers a sharper question — *how much, exactly, is it at the end of year five?* — and reveals something quietly beautiful: there are three different ways to compute that one number, and if your arithmetic is honest, all three land on the same figure.
Picture the policy's whole life as a road, with "now" — say the fifth anniversary — as a marker partway along it. To find the reserve at that marker you can stand on it and stare *forward* down the road that remains; you can turn and stare *backward* at the road already travelled; or you can simply take one careful step from last year's marker to this one. Forward-looking is the prospective method, backward-looking is the retrospective method, and step-by-step is the recursive method. Three views of the very same spot.
Looking forward: the prospective reserve
The prospective reserve is the most natural definition, and it reuses a tool you already trust completely. Stand at the marker and ignore the past entirely — water under the bridge. Look only at what is *still to come*: the benefits the insurer may yet have to pay, and the premiums it has yet to collect. Each of those is a life-contingent cash flow, so each has an actuarial present value. The prospective reserve is simply the first minus the second.
Why does the reserve climb as the years pass? Two forces, both intuitive. The benefit creeps closer and grows more likely — an 80-year-old is far nearer to a claim than the 40-year-old who bought the policy. Meanwhile the level premium stays flat, even though the true cost of one more year of cover is rising fast. In the early years the flat premium *overcharges* relative to that year's real risk; the surplus is set aside. In the later years it *undercharges*; the reserve is drawn down to make up the shortfall. The reserve is that stored-up overcharge, sitting ready.
Looking back: the retrospective reserve, and why it agrees
Now turn around. The retrospective reserve asks a different but equally fair question: of the premiums collected so far, how much is left over after paying for the cover already provided — all rolled forward with interest and survivorship to today? In words: take the accumulated value of past premiums, subtract the accumulated value of past benefits, and what remains is the fund built up by this group of policyholders. It is pure bookkeeping of the road already travelled — no peering into the future at all.
Here is the quietly beautiful part: under the *same* assumptions — the same mortality, the same interest rate, and a premium set by the equivalence principle — the prospective and retrospective reserves are *always equal*. The proof is almost a one-liner once you see it. The equivalence principle says, at issue, that the APV of *all* premiums equals the APV of *all* benefits, where "all" means past-plus-future. Split each side at today's marker into a past piece and a future piece. Rearrange. The future-benefit-minus-future-premium block (the prospective reserve) drops out equal to the past-premium-minus-past-benefit block (the retrospective reserve). The whole equation was balanced at the start, so any way you slice it, the two halves must still tally.
Stepping forward: the recursive formula
The prospective and retrospective methods each rebuild the whole present value from scratch at every marker. The recursive method is lazier and, frankly, more illuminating: it says *if you already know this year's reserve, here is exactly how to get next year's.* It tracks the reserve fund through one single year, accounting honestly for everything that happens to it. The result is a tidy bridge — the recursive reserve formula — and once you write it down, you can roll the reserve forward one rung at a time like climbing a ladder.
Follow the money through one year in four honest steps. (1) Start with the reserve you are holding at the beginning of the year. (2) Add the year's premium — fresh cash coming in. (3) Grow the combined pile by one year of interest, since the fund earns a return while it waits. (4) Pay out the year's expected death benefit — the benefit times the probability someone in this group dies during the year. Whatever survives all four steps, divided among the people *still alive* at year-end, is exactly next year's reserve. That is the entire idea, in plain bookkeeping.
One year in the life of a reserve (per surviving policy):
( reserve_now + premium ) * (1 + i) <- grow with interest
- death_benefit * q <- pay expected death claims
-------------------------------------------------
= reserve_next * p <- shared among survivors
so: reserve_next = [ (reserve_now + premium)(1+i) - benefit*q ] / p
i = interest rate, q = chance of dying this year, p = 1 - q = chance of survivingThree things make this formula a gift. First, it is fast: knowing one reserve hands you the next with a single line of arithmetic, no full present value to rebuild. Second, it is transparent — every term is a real event (premium in, interest earned, claims out, survivors share), so it doubles as a teaching tool and an audit trail. Third, it must reconcile with the other two methods: roll the recursion all the way from the issue reserve of zero and you will trace out exactly the same reserve curve the prospective and retrospective methods draw. Three routes, one mountain, confirmed once more.
Thiele's equation: the recursion made continuous
The recursive formula moves the reserve forward in whole-year jumps. But premiums can be paid continuously, benefits can be due the instant of death, and interest can compound smoothly rather than once a year. Thiele's equation is what the yearly recursion becomes when you shrink the step down to an instant — the same story told moment by moment instead of year by year. You met its ingredients earlier: the force of interest is interest acting continuously, and the force of mortality is the instantaneous rate of dying. Thiele weaves both into a single statement about how the reserve changes from one heartbeat to the next.
In words, Thiele's equation says: at each instant, the reserve is *growing* from two inflows — the interest it earns on the fund it already holds, and the premium streaming in — and *shrinking* by one outflow — the cost of the death benefit it must stand ready to pay at that instant, which is the benefit (minus the reserve already set aside) times the force of mortality. Inflows lift the reserve; the outflow pulls it down; the net of the two is the rate of change of the reserve at that moment. Strip away the calculus and it is the very same four-step story as the yearly recursion, just told without ever rounding to whole years.
Why bother with the continuous version? Because real product designs rarely fit neat annual boxes, and because Thiele generalises gracefully: the same shape of equation extends to multiple-state models — disability, recovery, lapse — where a policy can move between several conditions, not just "alive" and "dead." For your purposes today the takeaway is modest but real: Thiele's equation is not a fourth, mysterious method. It is the recursion you just understood, written in the language of instants. The yearly recursion and Thiele are the *same idea* at two zoom levels.
What the reserve is — and is not
Carry one honest correction out of this guide, because it is the misconception that haunts the whole subject. A reserve is *not* idle cash sitting in a vault, and it is *not* the insurer's profit waiting to be spent. It is a measured liability — the present value of a promise already made and not yet kept. The money behind it is invested in bonds and other assets precisely so it can earn the interest the recursion and Thiele's equation assume. Calling the reserve "the company's money" is like calling your mortgage balance "the bank's money it owes you" — it gets the direction of the obligation exactly backwards.
And remember the modesty stamped on every figure here. The reserve computed by all three methods is the net premium reserve — it backs the benefits only, using a chosen mortality table and a single fixed interest rate, exactly the assumptions the net premium reserve inherits from the net premium. Real reserves held for regulators and accounts add margins, allow for expenses and lapses, and may use different bases entirely. The three methods you learned are the honest skeleton; the flesh that supervisors require is layered on later in this ladder. But the skeleton is sound, and it is the same skeleton under every one of those richer reserves.