How Insurance Pools Risk and Prices What It Takes On

Insurance is an arrangement where many parties each pay a known amount into a pool, and the pool meets the loss of the few who suffer one. The price is set before anybody knows what the losses will cost, the money is held in the meantime, and the promise runs for as long as the cover does. Pooling steadies the average. Each single outcome stays exactly as uncertain as it was.

A known payment from many meets an unknown loss suffered by few. Everything else an insurer does is machinery built on that one exchange.

What is the bargain, before any of the machinery is added?

Strip an insurer of its statements, its investments and its vocabulary and one exchange is left. A large number of parties, each facing the same small chance of the same painful loss, agree to hand over a small certain amount. Whoever among them suffers the loss is met out of the total. Everybody has swapped a small chance of something unbearable for a certainty of something manageable.

Nothing about the underlying event changes when that agreement is signed: the same number of losses still happen, to the same number of parties, for the same reasons. What changes is who carries them. A great deal of loose talk treats insurance as a way of making bad things less likely. It is not. Insurance moves the financial weight of a bad thing off one set of shoulders and spreads it across a great many.

Here is the everyday version. Ten shops sit in one market building, and each of them has a shutter that fails perhaps once every ten years, costing about Rs 30,000/- to replace. Any one shopkeeper faces a small chance of a bill that ruins a month. Put a tin on the counter, have each of them drop in Rs 3,000/- a year, and the tin meets the shutter that actually breaks. Nobody's shutter has become sturdier. The bill has simply stopped being a catastrophe for whoever draws the short straw.

An insurer is that tin, run as a business, at a scale where the arithmetic becomes reliable, by somebody who has to quote the Rs 3,000/- before knowing how many shutters will break. Every difficulty in this subject grows out of that last clause.

What has to be true before anybody can insure a risk at all?

Not everything frightening can be insured, and the reasons are not squeamishness. Four conditions have to hold, and a risk that fails any one of them cannot be priced by anybody at any price. The whole subject stands on these four, so take them one at a time.

First, there have to be many exposures of a similar kind, and they have to be independent enough that they do not all go wrong on the same day. Independence is doing more work here than the word suggests, and its failure is worked in full below. Second, the loss has to be definite, identifiable in time, in place and in amount. Two parties reading the same contract in good faith then reach the same answer about whether it happened and what it came to. Third, the event has to sit outside the control of the party buying the cover. A loss somebody can arrange at will is not a risk, it is a purchase.

Fourth, and this is the one that decides what is insurable rather than merely alarming, a price has to be quoted before anybody knows what happens, so the expected cost has to be estimable in advance. A risk nobody can put a number on is not refused for being too large. The reason is simpler. There is nothing to write on the invoice.

What does pooling make predictable, and what does it leave alone?

Pooling is the part of the subject people most often carry away backwards, and slow arithmetic is the cure.

Take a stated illustration. Each member of a pool faces a one in a hundred chance, in one year, of a loss of Rs 1,00,000/-. The expected costThe average of every outcome weighted by how likely it is. One hundredth of Rs 1,00,000/- is Rs 1,000/-. The figure is the average bill per member per year, and no member is ever billed exactly that. for each member for that year is one hundredth of Rs 1,00,000/-, which is Rs 1,000/-. Notice at once that no member will ever suffer a Rs 1,000/- loss. Ninety nine members in a hundred lose nothing and one loses the lot. Rs 1,000/- is a description of the group, not a prediction about anybody in it.

With one hundred such members together, the average cost per member for the year is still Rs 1,000/-, and it swings by about Rs 994.99/- either side of that. The swing is almost the whole of the average. A pool of one hundred is barely a pool at all. With the same members in a pool ten thousand strong, the expected cost per member is still exactly Rs 1,000/-. The swing has fallen to about Rs 99.50/-.

The pool has not made a single member safer, and no member's own chance of a loss has moved by a hair. Their chance was one in a hundred before they joined and it is one in a hundred afterwards. The pool has moved the insurer's ability to quote a price at the start of the year and be roughly right at the end of it. Pooling does those two things and nothing else, and reading the first without the second is where most confusion about insurance begins.

The settling of an average over many independent outcomes, as their number grows, is not a modern discovery. Jacob Bernoulli set the result out in Ars Conjectandi, published in 1713, and every price an insurer quotes leans on it.

What has to be inside a price set before the cost is known?

An ordinary business knows what a thing cost before it decides what to charge for it. A shop buys at Rs 80/- and sells at Rs 100/-, and the Rs 80/- is a fact before the Rs 100/- is a decision. An insurer works the other way round: it sets the price first and finds out what the year cost afterwards, sometimes decades afterwards. Almost everything that looks strange about the industry follows from that inversion.

Four things have to be inside the price, and a description that names fewer than four has hidden one of them.

Take them in turn. The expected cost of claims is the arithmetic worked above, applied to the particular group being covered. The cost of running the arrangement is everything from what the intermediaryWhoever stands between the insurer and the party buying cover, bringing the proposal and often servicing it afterwards. The authority named below sets what each type of intermediary may do. is paid for bringing the proposal to what it costs to assess a claim when one arrives. The margin for being wrong is a different animal from the first item: the first says what the claims are expected to cost, and the third accepts that the expectation itself may be mistaken. And the fourth is the money held in between. On a contract lasting decades it can be the largest of the four, and it is the one almost nobody names unprompted.

Because the price is fixed at the start and the cost arrives later, an insurer can be wrong for years without knowing it. A shop that misprices its stock discovers it within a month. An insurer that has underpriced a long contract may find out in the twentieth year, by which time it has sold the same contract a hundred thousand times.

How long does the obligation last after the premium has been paid?

A cover on a building runs for a stated period. At the end of it both sides decide again, and either can walk away. A cover on a life is a different creature: it may have been paid for in somebody's thirties and fall due in their seventies, and for the whole of the intervening forty years the insurer is holding money against a claim nobody has made yet.

The obligation outlives the premium, sometimes by decades, and everything difficult about running an insurer follows from that one fact. The money has to be kept somewhere and it has to be worth something when the claim arrives. The estimate made at the outset has to survive changes nobody could see coming. The party who signed is not the party who claims, in the sense that the person is forty years older and the world has moved. None of that is a complication added by regulation. The contract itself says as much.

Whose money is the insurer holding in between?

Now a worked case. Chandrika Life Insurance Limited, an invented life insurer, reports the figures that follow. Being a life insurer matters every time one of them is used. The base a ratio is struck on differs between a life insurer and a general insurer.

In the stated year, Chandrika Life Insurance took new business premium of Rs 5,200 crore from policies sold during the year and renewal premium of Rs 10,400 crore from policies sold in earlier years and still running. Add them and total premium is Rs 15,600 crore. Divide rather than quoting: Rs 10,400 crore over Rs 5,200 crore is exactly 2.0 times, so two rupees in every three of the year's premium came from business written before the year began and one rupee in three from business written during it.

Divide claims of Rs 6,720 crore by that same Rs 15,600 crore and 43.08 per cent of the year's total premium left as claims. Do it again with expenses of Rs 2,496 crore and the answer is 16.00 per cent, on the identical base. Both of those bases are stated because both matter: total premium received, earned premium and net premium are different denominators, and a percentage moves when the denominator does. Together the two lines take Rs 9,216 crore of the Rs 15,600 crore, leaving Rs 6,384 crore. Most of that Rs 6,384 crore is money set aside against promises that have not yet fallen due, so the remainder is not profit and must never be called profit.

Now the number this whole subject turns on. Chandrika Life Insurance holds policyholder fundsMoney the insurer is holding because it has been promised out to policyholders. The money sits on the side of the statement where obligations sit, not the side where the insurer's own resources sit. of Rs 72,000 crore against a net worthThe insurer's own capital. The amount left for the owners of the business after everything it owes has been met, and nothing more mysterious than that. of Rs 7,200 crore. Rs 72,000 crore over Rs 7,200 crore is exactly 10.0 times, so every Rs 10.00/- of promises has Rs 1.00/- of the insurer's own money standing behind it.

Say it the other way and one figure changes. Everything in the insurer's hands is Rs 72,000 crore plus Rs 7,200 crore, being Rs 79,200 crore. Of every Rs 100.00/- of that, Rs 9.09/- is the insurer's own and Rs 90.91/- is somebody else's claim on the future. Nine rupees and nine paise, not ten rupees, and the difference is not a rounding slip: the first division is promises against capital and the second is capital against the whole of what is held, and two different denominators give two different answers. A figure that cannot be reproduced from the other figures beside it is a defect, so both divisions are worked in full.

The large number is not the insurer's wealth. The figure measures what the insurer has promised. An insurer described as a pool of money, with nothing said about what is owed out of it, has been described as a manager of somebody else's money with a different sign on the door. Reading the pool as the insurer's wealth is the misreading worked below.

What is general insurance, and how is that cover arranged?

General insurance covers a loss to a thing, or a liability to somebody else, for a stated period. At the end of that period the cover is decided again by both sides, and a fresh contract is written or is not. Three properties follow from that arrangement and each one is worth stating in its own line.

The payout is measured by the loss actually suffered, rather than by a figure agreed at the start. A new contract is written at each renewal rather than a standing one continued, so the price can be reset every time. And the whole cycle from premium to claim is usually short. The insurer's estimate is tested quickly and can be corrected while the mistake is still small.

Cover on a life is arranged along the opposite lines on all three counts: it is written once, it may run for decades, and the amount is agreed at the beginning rather than measured at the end. How the two genuinely differ, criterion by criterion, is covered separately. The two are different shapes of contract, and a figure from one cannot be read as though it came from the other.

Who else is party to the arrangement besides the two who sign it?

Read a policy and it looks like a contract between two parties. In practice five roles act at different points on the same contract, and none of them is the insurer taking a second bite at the decision. Each role gets one line, and each is worked properly under its own subject.

An agent or a broker brings the proposal, and the two are not the same role. One acts for the insurer and one acts for the party seeking cover, and whose interest sits behind the advice changes with it. An underwriterThe person or process inside the insurer that decides whether a proposal is accepted and on what terms and price. The weighing behind that decision is covered separately. decides whether the proposal is accepted and at what price. An actuaryThe qualified professional who puts a value on what has been promised and works out what must be held against it. The valuation basis itself is set by the authority named below. values what has been promised and what has to be held against it. A surveyorThe assessor who establishes what a loss actually was after it has happened, so the amount payable rests on a finding rather than on an assertion. or assessor establishes what a loss actually was once one is reported. And a reinsurerAn insurer that takes part of a risk from another insurer. Some of the promise is then carried further back. The cost of that arrangement is covered separately. takes part of the risk from the insurer itself.

Every one of those roles is registered or qualified under conditions set by the Insurance Regulatory and Development Authority of India (IRDAI) at irdai.gov.in, and those conditions belong to the authority.

Who sets the conditions an insurer works under?

Eight separate requirements have already been touched on above. Every one of them is set by IRDAI, and every one of them moves.

Where does pooling stop working?

The arithmetic rests on an assumption underneath it. Every figure worked above assumed that the losses arrive independently of one another. One member's roof falling in says nothing about anybody else's roof. Independence is what let the spread narrow.

Now remove it. A flood through one district, or a storm along one coast, is a single event that hands the same loss to every member of the pool in the same week. When losses arrive for one shared reason, there is no longer anything for the pool to average, and adding members does nothing at all. On the illustration used above, the swing about the average stays at about Rs 9,949.87/- however many members join, which is exactly what a single member faces alone and exactly one hundred times the Rs 99.50/- a ten thousand member pool enjoyed when the losses were independent.

Ten shops in one market building can pool the cost of a shutter. The building closing arrives for all ten on the same morning, so its cost cannot be pooled. The size of the loss is not what decides it. The shared cause is.

Two consequences follow, and each is a subject of its own. An insurer facing losses that can arrive together has to hold capital against that possibility, and it has to pass part of the risk to somebody else who is not exposed to the same event. Both are covered separately.

There is a second limit and it is less obvious than the first: an insurer can simply be wrong about the expected cost for a whole group at once. If the one in a hundred was really one in eighty, more members does not help. More members multiply the error rather than averaging it away. Scale is a defence against variation around a correct estimate and no defence at all against an incorrect one.

How does anyone actually use this?