Two promises that seem like opposites
By now you have built two machines that, on the surface, do opposite jobs. A whole-life insurance, with actuarial present value written A_x, values a promise to pay $1 *when the person dies* — it is money that flows at the moment a life *ends*. A whole-life annuity-due, written ä_x (an a with two dots), values a promise to pay $1 *at the start of every year the person is still alive* — money that flows precisely *while a life continues*. One pays on death; the other pays on survival. They feel like mirror images, almost like enemies.
And yet they are not two separate calculations that happen to live in the same chapter. They are two views of *one single financial object*, locked together so tightly that if you know one, you know the other exactly — no extra mortality table, no fresh integral. This guide is about that lock. By the end you will see why the insurance–annuity relationship is not a coincidence to be memorised but a near-obvious truth once you look at the right thought-experiment.
The thought-experiment: a coin you keep until you die
Picture a single $1 coin handed to a person aged x today. The deal is brutally simple: keep the coin invested, take out only the *interest* it earns each year while you are alive, and when you die hand the whole $1 back. Nothing has been created or destroyed — one dollar went in, and exactly its parts come back out. Now value each piece. The stream of interest you draw each year *while alive* is exactly an annuity on the life. The $1 returned *at death* is exactly a whole-life insurance. The single dollar today must equal the value of everything that flows out of it.
Here is the only subtle part, and it is where the rate of discount d earns its keep. If you hold $1 and want to spend its interest *at the start* of each year — matching an annuity-*due*, paid in advance — the amount available up front is not the full interest rate i but the discount rate d, where d = i/(1+i). Drawing d at the start of each surviving year is the same as drawing the year's interest. So the value of the interest stream is d times the value of the $1-per-year annuity-due: that is d·ä_x. The remaining piece, the dollar returned at death, is A_x. The whole dollar, viewed today, is worth 1.
The identity, and what it really says
Rearrange the conservation sentence and you get the headline identity that ties a whole-life insurance to a whole-life annuity: A_x = 1 − d·ä_x. Read it slowly. It says the value of paying $1 at death equals one whole dollar *minus* the value of all the interest you would have drawn off that dollar while alive. The longer the person is expected to live, the bigger ä_x is, the more interest gets skimmed off, and so the *smaller* A_x becomes — exactly right, because a death benefit far in the future is worth little today. A short life ahead means a small ä_x and an A_x pushed close to 1, because the dollar is coming back almost immediately.
Let's make it concrete with a tiny example. Suppose at age x the whole-life annuity-due value works out to ä_x = 15.0, and the effective interest rate is i = 5%, so the discount rate is d = 0.05/1.05 = 0.047619. Then A_x = 1 − 0.047619 × 15.0 = 1 − 0.714 = 0.286. So a $1 death benefit on this life is worth about 28.6 cents today, and a $1-a-year lifelong income is worth $15.00 — and we found the insurance value *without touching the mortality table again*, purely from the annuity value and the interest rate. That is the lock at work.
Conservation of a $1 coin held until death: 1 = (interest spent while alive) + (dollar returned at death) 1 = d * adotx + A_x so A_x = 1 - d * adotx <-- the headline identity Worked numbers: i = 5%, d = i/(1+i) = 0.05/1.05 = 0.047619, adotx = 15.0 A_x = 1 - 0.047619 * 15.0 = 1 - 0.71429 = 0.28571 Sanity check, turn it around: adotx = (1 - A_x) / d adotx = (1 - 0.28571) / 0.047619 = 0.71429 / 0.047619 = 15.0 (consistent)
Recursions: building one age from the next
The second great relationship is a way of building values *across ages* without re-summing everything from scratch. Stand at age x and ask: what happens in the next single year? With probability p_x the person survives the year, and with probability q_x they die during it. A recursion simply splits the whole-life value into "what happens this one year" plus "the value of standing in the same position one year older." For insurance, the one-step recursion reads A_x = v·q_x + v·p_x·A_{x+1}: discount one year, pay the benefit if death occurs now, otherwise carry the same kind of insurance forward on a life now aged x+1.
The annuity has its own twin recursion, and it is even friendlier: ä_x = 1 + v·p_x·ä_{x+1}. In words: a lifelong annuity-due pays you $1 right now (you are alive today, that is certain), and then, *if* you survive the year (probability p_x) and after discounting one year (factor v), it is exactly a lifelong annuity-due on a life aged x+1. The leading "1" is this year's guaranteed payment; everything after it is next year's problem, suitably discounted and survival-weighted.
These recursions are not just elegant; they are how the numbers are actually computed. You start at the oldest age in the life table, where the value is trivial (no one survives past the table's end, so the last A is just v·q and the last ä is just 1), and you sweep *backward* one age at a time, each step reusing the answer you just found. This backward sweep is the engine behind reserve calculations in the very next guides — and historically the same logic was frozen into commutation functions, precomputed columns that turned a lifetime of recursion into a single division.
Honest fine print
The identity A_x = 1 − d·ä_x is exact, but only for the *matching pair* you actually wrote down: a benefit paid at the *end of the year of death* and an annuity-*due* paid at the *start of each year*, both at the same interest rate, both on the same single life. Mismatch them and the clean form breaks. If the death benefit is paid at the *moment* of death (the continuous insurance Ā_x) it pairs with a *continuously* paid annuity through Ā_x = 1 − δ·ā_x, where δ is the force of interest rather than d. The shape of the truth survives, but you must keep the timing of payments and the matching interest measure consistent — a recurring discipline you will lean on when these functions go monthly or continuous.
Why this is the heart of the rung
Step back and feel what you have gained. Insurance and annuity are not two subjects to learn twice; they are one truth seen from two ends. That is enormously practical. The equivalence principle in the next track sets a level premium so that the value of premiums (an annuity the policyholder pays *while alive*) equals the value of the benefit (an insurance paid *at death*) — and the identity is exactly what lets you flip between those two sides of the balance. The recursions are what make reserves computable age by age. You have, in one short equation, the spine of life insurance mathematics.
And a closing honesty: every value here, A or ä, is an *expectation* — an average over a whole imagined crowd of identical lives, computed from a mortality table that is itself an estimate of the future. No single policyholder experiences the average; one dies next week, another lives to 102. The identity and the recursions are exact statements about those averages under a chosen model, not prophecies about any one person. That gap between the clean expected value and the messy individual outcome is precisely why pooling, reserves, and capital exist — and it is the thread the remaining rungs of this ladder pick up.