The problem no spreadsheet can solve alone
Imagine you retire at 65 with a pile of savings and one job: spend it across the rest of your life without ever hitting zero. The cruel twist is that you do not know how long that life is. If you budget to age 85 and live to 95, your last decade is spent in poverty; if you budget to 100 and die at 78, you scrimped for years and left a fortune unspent. This is longevity risk — not the risk of dying, but the risk of *living longer than your money* — and no clever withdrawal spreadsheet can erase it, because the one number you most need, your date of death, is exactly the number you cannot know.
An annuity is the contract that flips this risk onto an insurer. You hand over a sum of money, and in return the insurer promises to pay you an income for as long as you live — stopping only when you die. You can no longer outlive your money, because the payments do not run out; they run out *with you*. The magic ingredient is the same one from the pooling rung: across thousands of buyers, the ones who die early subsidise the ones who live long, and the insurer only has to fund the *average* lifespan, not everyone's worst case. That is why an annuity can pay you more per year than you could safely draw from the same sum alone.
From the symbol you already know to the product on the shelf
You have already met the machinery. In the contingencies rung the whole-life annuity symbol adot_x gave the value today of paying 1 at the start of each year *for as long as a life aged x survives* — it is just an annuity-due in which each future payment is weighted by the probability of still being alive to collect it. Pricing a real annuity is that symbol with two real-world layers added: scale it to the income the buyer wants, and load it for the insurer's expenses and profit. The gap between the pure value adot_x and the actual price you pay is where the commercial product lives.
Compare this with the annuity-certain from the interest-theory rung, which paid for a fixed number of years no matter what. A life annuity is the same present-value sum, but every term is multiplied by a survival probability that fades toward zero in the far future. That single change is what makes a life annuity safe to over-spend from: the certain annuity must reserve fully for year 40, while the life annuity barely reserves for it, because almost nobody is alive to be paid. The contrast between actuarial present value and ordinary present value is precisely this survival weighting.
Why a life annuity pays more than "safe" self-withdrawal
(illustrative, single 65-year-old, sum = 100,000, interest ~4%)
drawing it down yourself, planning to age 95 (30 yrs):
safe income ~ 100,000 / a(30-yr certain) ~ 5,400 / yr
...and you must STILL fear living past 95.
a life annuity (priced on adot_65, survival-weighted):
income ~ 100,000 / adot_65 ~ 6,500 / yr
...and it NEVER stops while you live.
The extra ~1,100/yr is the "mortality credit": the pooled
subsidy passed from those who die early to those who live on.The family tree: when, how, and for whom it pays
Real annuities differ along a few clean axes. The first is *when payments start*. An immediate annuity starts paying almost at once — you buy it the day you retire and income begins next month. A deferred annuity sits dormant first: you pay now (or over years), it accumulates, and income switches on at a chosen future date. A radical, cheap variant of deferral is the longevity annuity, bought at 65 but not paying until 85 — it is pure insurance against the tail of old age, costing little precisely because many buyers will not live to collect.
The second axis is *how you pay in*. A single-premium annuity is bought with one lump sum — the classic move of rolling a lifetime's savings into guaranteed income on retirement day. Others are funded by years of contributions. The third axis is *what happens around death*, and here the buyer trades income for protection of their heirs. A pure life annuity pays the most but stops dead the moment you die — buy it, get hit by a bus next week, and your estate gets nothing. That sting is why most people choose a softer form.
The two softer forms are the workhorses. A period-certain annuity guarantees payments for, say, ten years *even if you die early* — if you pass in year three, your beneficiary collects the remaining seven years, after which, if you are still alive, it simply continues for life. A joint-and-survivor annuity covers two lives, typically spouses: it keeps paying (often at a reduced rate, say two-thirds) until the *second* person dies, so the survivor is never left destitute. Each guarantee costs income: the more you protect against early death, the lower your monthly cheque, because you have given back some of the mortality credit that made the annuity generous in the first place.
Variable and indexed: chasing growth without losing the floor
Everything so far assumed a *fixed* payment, where the insurer bears the investment risk and you get a flat, predictable cheque. Buyers who want their income to grow — to beat inflation, or simply to feel they are still in the market — drove the rise of variable and indexed annuities. In a variable annuity your money sits in investment sub-accounts (essentially mutual funds), and your future income rises and falls with markets; the upside is real, but so is the chance your retirement income shrinks in a crash. The insurer here is more custodian than guarantor — unless you buy a rider.
An indexed annuity sits between fixed and variable: your credited interest is tied to a stock index, but capped and floored. A typical contract might credit you, say, 60% of the index's annual rise up to a 7% ceiling, while guaranteeing you never lose principal in a down year — a zero floor. It sounds like the best of both worlds, and that is precisely the warning. The caps, participation rates, and spreads are levers the insurer adjusts to fund that guarantee; the floor is paid for, in slices, out of the upside you are quietly handing back. There is no free lunch here, only a repackaged one.
The headline feature of modern variable annuities is the guaranteed-benefit rider, names like GMWB (guaranteed minimum withdrawal benefit) or GMIB (guaranteed minimum income benefit). These promise a floor — you can withdraw at least X% for life *even if your sub-accounts crash to zero*. For the policyholder this is genuinely valuable; for the insurer it is a frightening obligation, because it is essentially a long-dated put option on the stock market combined with a bet on longevity. Insurers must hedge it dynamically in financial markets, the subject of dynamic hedging of guarantees, and several firms were badly burned in 2008 when markets and policyholder behaviour both moved against their models at once.
The annuity puzzle: the product nobody buys
Here is the deep strangeness economists call the annuity puzzle. Theory says a rational retiree facing longevity risk should pour much of their savings into a life annuity — it is the cleanest possible solution to outliving your money, and decades of models conclude people *should* annuitise heavily. Yet in practice almost nobody does. Voluntary annuity markets are tiny. The very product engineered to solve retirement's scariest problem sits on the shelf, largely unsold. Explaining the gap is one of the most fertile arguments in all of pension economics.
- Loss of control and the bequest motive: handing a lump sum to an insurer feels like losing it, and many people genuinely want to leave money to children — a pure life annuity leaves nothing, which feels like the worst case if you die early.
- Adverse selection raises the price: because the people who buy annuities know they are healthy and expect to live long, the pool skews long-lived, so insurers must price for that — making annuities look expensive to an average buyer. This is adverse selection in plain sight.
- Existing annuities crowd it out: most retirees already hold annuity-like income from state pensions and any defined-benefit pension, so their longevity is partly insured already, dulling the urge to buy more.
- Framing and complexity: people see the headline price as a gamble ("what if I die next year?") rather than as insurance against the opposite, scarier risk, and the product's riders and jargon make it hard to trust what you are buying.
None of these alone fully resolves the puzzle, and honest economists admit a chunk of it remains a mystery — which is exactly why partial annuitisation, deferred longevity annuities bought cheap at 65, and default-annuitised pension designs are all live attempts to capture the benefit while soothing the fears. As you climb into the pension rungs ahead, keep the puzzle in mind: a defined-benefit pension is, in effect, an annuity that the worker never had to *choose* to buy, and that quiet automatic-ness is a large part of why such pensions delivered secure retirements that voluntary markets cannot.