Underwriting: Which Risks an Insurer Takes, and the Price

Underwriting is the decision an insurer takes on each proposal: accept it, decline it, or accept it on stated terms, and at what price. The insurer sorts the proposal into a group whose expected cost it believes it can estimate, then quotes a price carrying that estimate, the cost of running the arrangement and a margin for being wrong. The price is set before the cost is known.

Seven things about underwriting are somebody else's to decide. An insurer may ask a proposer only certain questions, and may put only certain answers into a price. The consequences of an answer once it is given are set elsewhere, and so is how long any of it stays open to challenge. Who is allowed to sell the contract is set elsewhere. So is what has to be set aside against a policy already written, and where a complaint travels when somebody disagrees with a decision. All seven belong to the Insurance Regulatory and Development Authority of India (IRDAI), and all seven move. The mechanism of underwriting holds whatever those seven values happen to be on the morning they are needed.

Every part of the decision follows from one asymmetry, and the asymmetry runs in two directions at once. The insurer is quoting a number today for something whose cost turns up later, sometimes decades later. The party proposing knows more about their own circumstances than the insurer is ever going to know. Underwriting is the set of moves an insurer has against both of those at the same time.

The pair of asymmetries explains everything an underwriter does. Every move is one of three things: an attempt to improve the estimate, an attempt to narrow what the estimate has to cover, or an attempt to make sure that being wrong about the estimate is survivable.

What does underwriting actually decide?

Ask most people what an underwriter does and the answer comes back as a price. The price is half of it. There are three possible answers to a proposal, and the price is attached to whichever one is chosen rather than being the choice itself. The insurer can accept the proposal as it was made. The insurer can decline it. The third answer is to accept it on stated terms, and the third answer carries most of the work.

Accepting on terms means changing what is being covered before agreeing to cover it. Cover less than was asked for, by naming a loss the contract does not carry. Cover it later, by starting a stated part of the cover after a stretch of time has passed. Or cover it with a share of each loss left on the other side, keeping the smallest and most frequent claims out of the arrangement altogether. Each of those is a different contract from the one that was proposed, and each carries its own price.

Declining is a decision about this insurer's own pool, not a verdict on the party who proposed. That distinction matters more than it looks. An insurer declines when it cannot place the proposal in any group whose cost it believes it can estimate, or when taking it would distort a group it has already priced. A risk that one insurer will not carry at any price is a risk another may carry at a price, on a different set of groups, with a different bookThe whole set of policies an insurer has written and is still carrying, taken together rather than one at a time. behind it. The difference between one insurer's book and another's is a large part of why more than one insurer exists.

Why is the price set before anybody knows what the claims will cost?

In almost every other business the order runs the other way. A shop knows what the stock cost before it puts a label on it. A caterer knows what the vegetables cost before quoting for a wedding. Even a builder who quotes for a repair before opening the wall is guessing about a job that will be finished, and paid for, within months.

An insurer quotes now and finds out later, and on cover for a life the later can be decades away. The premium is fixed at the front of the contract. The claims that test whether it was the right premium arrive over the whole length of the cover, and some of them arrive after the person who wrote the price has left the job. The reversal is not a quirk of the industry. The reversal is the reason underwriting exists as a separate discipline with an actuaryThe specialist who builds the basis on which expected cost and reserving are estimated, using mathematics of probability and of long dated obligations. behind it rather than being a pricing clerk's job.

Three consequences follow, and between them they explain most of what an insurer does.

The first is that the price has to carry a margin for the estimate itself being wrong, and the margin is a separate thing from the expected cost. Expected cost is the middle of a range. The margin is there because the actual result can land anywhere in that range, and the insurer has already promised the cover.

The second is that a mistake stays invisible for a long time. If a price is too low, nothing happens on the day it is set. Nothing happens the following month either. The proposals keep arriving, the underwriter keeps saying yes on the same basis, and the error is repeated across a whole book before a single claim contradicts it. An underwriting mistake is a mistake that gets copied thousands of times before it announces itself.

The third is that capital has to sit behind the possibility. The claims of a group that turned out to be underpriced still have to be met, and they are met out of the insurer's own money once the premiums of that group run out. The possibility of an underpriced group is why an insurer is required to hold a margin above what it owes its policyholders. How much of a margin, measured how, is set by IRDAI at irdai.gov.in.

What is risk classification, and why does the arithmetic need it?

Classification is the step where proposals are sorted into groups that are similar enough in expected cost to be priced together. Classification looks like filing. Classification is not filing. Sorting proposals into groups creates the pool the arithmetic is done on, and that puts it inside the mathematics rather than inside the paperwork.

Pooling only does its work inside a set of exposures that resemble each other in expected cost. Put fifty similar exposures together and the average outcome of the group is far steadier than the outcome of any one of them, and that steadiness is the whole basis on which a premium can be quoted at all. Put fifty exposures together that do not resemble each other, and the average is still an average, but it no longer describes anybody in the group.

Here is the everyday version. A courier's delivery van on the road eleven hours a day and a scooter parked in a compound most of the week are not two versions of the same exposure. One price charged across both has not been even handed with either of them. A single price charged across two groups with different expected costs is not a compromise between two right answers: it is one answer that is too high for one group and too low for the other at the same moment.

What happens to the pool when the price does not match the group?

Pricing against a pool that moves is the least intuitive part of underwriting, and it is worth going slowly. Consider two groups of five thousand members each, on an illustration built for teaching and belonging to no insurer. One group has an expected cost of Rs 800/- a year. The other has an expected cost of Rs 2,000/- a year. The two groups differ in expected cost and in nothing else at all: no age, no health, no occupation, nobody in particular. The mechanism is about prices and information, and it works exactly the same whatever the two groups happen to be.

Now charge everybody the same price. The obvious price is the blended average of Rs 1,400/- a year, and it is what most people reach for first. Under the acceptance rule stated on the drawing below, that price is a fine deal for the higher cost group and a poor one for the lower cost group, so take up differs between them: 1,250 of the 5,000 lower cost members take the cover, and all 5,000 of the higher cost members do.

Count what that leaves. There are 6,250 members in the pool. Their combined expected cost is 1,250 multiplied by Rs 800/- plus 5,000 multiplied by Rs 2,000/-, or Rs 1,10,00,000 for the year. Dividing that by 6,250 members gives Rs 1,760/- a member a year. The price of Rs 1,400/- has attracted a pool that is expected to cost Rs 1,760/- a member a year, a gap of Rs 360/- a member a year that the price does not cover.

The reflex is to raise the price to meet the cost. Watch what that does. At Rs 1,600/- a year the lower cost group has gone entirely, the pool is 5,000 members expected to cost Rs 2,000/- each, and the gap has widened to Rs 400/-. Price and cost only agree at Rs 2,000/-, and they agree there because there is nobody left in the pool except the group that always cost Rs 2,000/-. The price did not catch up with the pool; the pool shrank down to meet the price. That is why finer classification, and not a higher single price, is the move that actually repairs the arithmetic.

What can an insurer change other than the price?

An underwriter who can only move the price has one instrument and, as the illustration above shows, it is the instrument that fights back. There are four others, and they are worth naming separately because they behave differently.

Classify more finely. Each group is then charged closer to its own expected cost, and the blending problem shrinks. Cover less, by naming the losses the contract does not carry. Start the cover later for a stated part of it, keeping a loss already in motion when the contract began off the pool. And leave a share of each loss on the other side, keeping the smallest and most frequent claims out of the arrangement altogether.

All four change what is being priced rather than the price of it. That is the useful way to hold them: the underwriter is redrawing the thing being covered so that the estimate it needs is one it can actually make. And what none of them can do is repair an estimate that was wrong about the group as a whole. If the expected cost of a group has been misjudged, a longer list of exclusions moves the boundary of the contract without moving the misjudgement.

Where does the insurer's information about the risk come from?

Almost all of it arrives on one document. The proposal is the insurer's view of the risk, and on most contracts it is very nearly the whole of that view. Around it sit three smaller sources: what the insurer already holds on similar risks it has written before, whatever it is able to verify independently, and whatever it can observe over the life of the contract once the cover has begun.

The party proposing knows more about their own circumstances than the insurer does, and the contract is built on the expectation that what is asked and what is answered closes most of that gap. That single sentence explains why the proposal form looks the way it does, why an intermediaryAn agent or a broker standing between the insurer and the party proposing, carrying the proposal in one direction and the contract in the other. is regulated as carefully as the insurer, and why so much of the law around insurance is about questions and answers rather than about money.

IRDAI sets what an insurer may ask about, what it may use in setting a price, what follows from an answer given on a proposal, and the window inside which any of that may be raised at all, and publishes all of it at irdai.gov.in. One point about the mechanism holds whatever those values are. An answer that later turns out to have been incomplete is something the contract has terms about, and the contract terms settle it. An incomplete answer is not evidence that somebody set out to mislead.

What did one year of underwriting look like at one insurer?

Chandrika Life Insurance Limited, a life insurer made up for teaching, gives the shape of one stated year below. The base a ratio is struck on differs between a life insurer and a general insurer, so being a life insurer matters for every ratio struck on it.

The divisions can be redone rather than taken on trust: Rs 6,720 crore over Rs 15,600 crore is 43.08 per cent of total premium, and Rs 2,496 crore over Rs 15,600 crore is 16.00 per cent of total premium. Renewal premium is exactly 2.0 times new business premium. Most of the year's underwriting decisions were taken in earlier years and are simply still running.

The amount left after claims and expenses is not profit, and calling it profit is the fastest way to misread an insurer. Nothing in that table has touched what must be set aside against the policies still running, and beside those figures sit policyholder fundsThe money an insurer holds that is owed to policyholders rather than to itself, standing against the contracts it has already written. of Rs 72,000 crore against net worth of Rs 7,200 crore, a ratio of ten to one. A year's premium of Rs 15,600 crore is a small movement beside an obligation of that size.

When does anybody find out whether the underwriting was any good?

Later than anybody would like, and the delay has a shape worth seeing. A policy written this year has had only a few months in which to produce a claim, and one written eight years ago has had eight years. The claims arriving in any one year therefore came overwhelmingly from policies written in earlier years.

A low ratio of claims to premium in a young book is a statement about the age of the book, not about the quality of the underwriting, and the two are indistinguishable inside a single year's figures. Two things push a claims ratio down that have nothing to do with the underwriting. Policies written recently have not yet had time to claim, and a book growing quickly carries a large slice of premium in the denominator from exactly those policies.

Separating the two would need the claims split by the year the policies were written. Each year of underwriting is then judged on the claims that year produced rather than on the claims that happened to land. The split by year of writing is not in the table above. One stated year for one insurer, with no breakdown by how old the policies are, cannot produce it, and a number put in its place would be made up rather than derived.

How is deciding a claim different from underwriting it in the first place?

Underwriting and claim decisions are two decisions, taken at different times, by different people, under different questions, and the two are merged constantly. Underwriting asks whether to take this risk and at what price, and it happens before anything has occurred. A claim decision asks whether what did occur sits inside the contract that was already written, and it happens afterwards.

A claim decided against the policyholder is not an underwriting decision arriving late. It is a reading of terms that were fixed at the outset, taken under conditions IRDAI sets at irdai.gov.in, and there are stages by which such a decision is escalated and reviewed. The stages belong to the authority rather than to the insurer.