Skip to content

← Money

Insurance

Chapter 19 · Moving Money Sideways

In this chapter 6 sections
  1. Many small payments for a few big losses
  2. Keep it, avoid it, shrink it, or pass it on
  3. What each side knows
  4. Rules first made at sea
  5. Insurers have insurers
  6. Try this

When someone dies in a Newar farming village of the Kathmandu Valley, the family doesn’t face the funeral alone. Its सी गुथी (si guthi), a funeral association, takes charge of the procession and the rites. In Jyapu villages membership is compulsory. It passes from father to son, and a son’s own place in it is settled once he has set up a house apart from his father’s, or after his father dies. These groups are much more than funeral associations: each has its own guthi house and its own god, and belonging to one is part of what makes a man a full member of his place.

Nobody in a सी गुथी pays a premium, and nobody calls it insurance. The comparison that follows is ours, not the guthi’s, but each household does get something from it. A death in the family is certain to come and impossible to time, and when it comes, the work and the grief land on one house at once. By joining, a household promises its hands to every other member’s funeral, and in return it knows that its own will never be faced alone. Many families each give a little, often; the few who are struck that year get a lot, at once. That is the idea of insurance, done in work instead of rupees.

Insurance does the same thing with money. Chapter 2 called it the sideways move: a little money taken out of every possible future, so that the unlucky ones are paid for. This chapter is about how that can possibly work, what it costs, and why some losses can be insured and others can’t.

Many small payments for a few big losses

Picture houses in a town where each has a 1 in 100 chance, every year, of a fire that costs Rs. 10 lakh. On average a house loses Rs. 10,000 a year, but no house ever loses Rs. 10,000. In 99 years out of 100 it loses nothing, and in the hundredth it loses everything a family has saved in ten years. The average is a number that never happens to anyone.

Now put a hundred such houses together. On average there will be one fire a year among them. But a hundred houses are still too few to count on the average: in about 37 years out of 100 there is no fire at all, and in about 8 there are three or more. A fund that collected the average would be empty in a bad year. Put ten thousand houses together and something changes. There will be about a hundred fires a year, and in 19 years out of 20 the number falls between 80 and 120. With a million houses it falls between about 9,800 and 10,200. Each house is as unpredictable as ever, but the whole crowd of them becomes almost as steady as a salary.

This is the law of large numbers. The more separate chances are added together, the closer their average comes to the true one. A coin tossed ten times may land heads seven times; tossed ten thousand times, it lands heads close to half the time, every time. Jacob Bernoulli, a Swiss mathematician, proved it in a book on chance published in 1713, after his death. It is the reason an insurer can promise something no single household could: to pay whenever the fire comes.

So an insurer collects a small, certain sum from many people, called the premium, and pays the large, uncertain losses of the few. It has to set the premium above the average loss, because it pays staff and agents, keeps money in reserve for the bad years that still come, and its owners want a profit. In our town, say it charges Rs. 12,500 a year, a quarter more than the average loss of Rs. 10,000. Chapter 15 of Thinking Clearly has more on how chance behaves in crowds, and a game’s average worth is worked out in the Going Further maths.

That raises an odd question. If the premium is more than the average loss, every household that buys cover loses money on average. Why would anyone sensible pay it?

Try it

A hundred lives

Each house is a household’s twenty years, all made up.

no disastera disaster, no loanhad to borrow
Average at the end
Rs. 17.9 lakh
Worst at the end
Rs. 0 lakh
Had to borrow
7 of 100

20 had a disaster and 7 had to borrow. The unluckiest ended with nothing.

In the widget, each household puts aside Rs. 1 lakh a year, a disaster costs it Rs. 10 lakh, and cover costs a quarter more than the average loss. Run the first hundred at 1 in 100 a year. Without cover, 80 households never have a disaster and end with Rs. 20 lakh. Nineteen have one and end with Rs. 10 lakh, and one has two and ends with nothing. Seven had to borrow along the way, because the disaster came before they had saved enough. The average is about Rs. 18 lakh. With cover, every household pays Rs. 12,500 a year and ends with Rs. 17.5 lakh. The average is lower with cover, just as it must be. But nobody is ruined and nobody borrows.

That is the answer. The last few rupees a family has matter far more to it than the twentieth lakh. Rs. 2.5 lakh less, spread over twenty years, is a smaller life; Rs. 10 lakh gone at once, when the family has saved only three, can mean a loan at the moneylender’s rate, a child out of school, a field sold. People buy insurance not to make money but to take the worst outcomes off the table. It is a lottery turned the other way round: a lottery ticket costs a little to buy a tiny chance of a big gain, while a premium costs a little to remove a small chance of a big loss. Both lose money on average. Only one of them makes a life steadier.

Some hundreds cost the insurer more than they paid it. At 1 in 100, in two of the five hundreds the claims came to more than the premiums. That is no failure. An insurer covers many thousands of households, and their quiet years pay for this hundred’s bad ones.

The steadiness insurance buys does more than protect one family. A business that is insured can open again after a fire, flood or earthquake, and banks lend more readily to a business whose losses are covered. Insurance keeps a business going, and helps it grow.

Keep it, avoid it, shrink it, or pass it on

A risk is the chance that something goes worse than expected. Not every risk should be insured, and textbooks of risk management sort the choices into four.

You can keep a risk: decide that the loss is small enough to pay for yourself if it comes. A broken phone screen, a lost umbrella, a week’s sickness. This is what Chapter 9’s fund for surprises is for, and it is often the cheapest choice, because no insurer’s costs and profit are added. You can avoid a risk: not build on the riverbank, not ride without a helmet on the highway, not lend the money. You can shrink it: put a lock on the door, an extinguisher in the kitchen, rods in the walls. And you can pass it on to someone else, by paying them to bear it. Insurance is the main way of passing a risk on. Deciding which of the four fits each risk, and doing it, is what risk management means, for a household as for a company.

A family might shrink the risk of fire with proper wiring, keep the small risks in a savings fund, and insure only the house. The rule of thumb follows from the widget: insure what would ruin you, keep what wouldn’t.

Only some risks can be passed on. The ideal one, as textbooks list it, comes in many similar copies, so the law of large numbers can work. Its losses are accidents, outside the control of the person insured. They are definite, clear in time, place and amount, so everyone can agree one happened. Their chances can be worked out, so a price can be set. They don’t all strike at once. And the premium is small beside the loss it covers. The risk should also be pure, in the sense that the event can only cost, never pay. A fire, a crash or an illness is a pure risk. Starting a shop or buying shares is a speculative risk, since it might make money as well as lose it, and no insurer will pay because your shop fails to sell. That is your bet, and the profit, if it comes, is yours too.

The losses people insure fall into three kinds: harm to a person (death, illness, injury), harm to property (a house, a vehicle, a crop, goods on the road), and harm someone causes to others, which the law may make them pay for. That last is liability insurance, and the part of a vehicle’s insurance that pays for the people a crash injures is one.

What each side knows

An insurer has one great weakness: the people it insures know more about themselves than it does. Two problems follow, and they have names.

The first comes before the contract is signed. Suppose an insurer charges everyone the same price for health cover, set at the average. Who is keenest to buy it? People who know they are often ill. Healthy people find it poor value and some stay away. Now the insurer’s customers are sicker than average, so its claims rise, so it raises the price, so more healthy people leave, and the circle goes round. This is adverse selection: the people who choose to buy are not the average people. The economist George Akerlof described this kind of market half a century ago, with used cars and with old people who find it hard to buy health insurance, and later shared a Nobel Prize for it.

Insurers fight it in two ways. They ask questions, and set prices by the answers: age, the kind of building, the driving record, a medical check. And they offer a choice. A policy with a deductible, where you pay the first part of any loss yourself, comes at a lower premium. People who know they are careful and rarely claim tend to choose it, and people who expect to claim tend to pay more for full cover. Each group sorts itself by its choice. Joseph Stiglitz, who shared the same prize, showed how such a menu of choices lets an insurer tell its customers apart without asking.

The second problem comes after. Once you are insured, you have less reason to take care, or to save on a loss you won’t pay for. This is moral hazard. The name is old and unkind; it means only that people respond to who pays. The largest test of it was in the United States, where in the 1970s more than 7,700 people were given health insurance of different kinds at random. Those who had to pay part of each bill made one or two fewer visits to the doctor a year than those whose care was free, and went into hospital about a fifth less often. For most of them their health was no worse. For the poorest and sickest it was, in blood pressure and eyesight and teeth: paying a share had made them skip care they needed.

So the deductible does a second job: it keeps you caring about a loss, because part of it is yours. Exclusions follow the same logic. A policy that paid for anything at all, including what you could have prevented, would be bought by the wrong people and used in the wrong way. The small print exists because of what each side knows.

Rules first made at sea

Insurance grew up with long-distance trade by sea, where a ship could carry a merchant’s fortune and never come back. It began in the trading cities of Italy around seven hundred years ago, and in the port of Genoa, in three weeks of 1393, a single notary drew up more than eighty insurance contracts.

In London, the business found a home in a coffee house. By the late 1600s a man named Edward Lloyd kept one in Tower Street, and it became the place where ship owners, captains and merchants went for news of ships: Lloyd gathered it from correspondents in other ports and printed it on a sheet. Men with money to risk went there too, and a ship owner who wanted cover for a voyage would find them. Each backer signed the policy below the others, taking a share of the risk, which made him an under-writer. It is still the word for anyone who decides whether to take on a risk and at what price. The market that grew from the coffee house, Lloyd’s of London, says plainly in its own history that it was the world’s center for insuring the British shipping that carried enslaved Africans, and in 2020 it apologized. Houses came to be insured there too: after the Great Fire of 1666 burned some 13,200 of London’s houses, the city’s first fire insurance office opened in 1680.

Out of these centuries of arguments in court came the principles that insurance still runs on, six in most lists. Each one exists to solve a problem, and the easiest way to see them is to follow one made-up shop.

Srijana’s shop in Bhaktapur is insured against fire. When she bought the policy, she had to tell the insurer everything that mattered about the risk, including the cooking gas she kept in the back room. That is utmost good faith: because the insurer can’t inspect everything, the person who knows must say, and a policy got by hiding something that mattered can be refused. It was her shop to insure, because she would lose if it burned. That is insurable interest. Without it, insurance would be a bet on someone else’s bad luck, and in 1774 Britain’s Parliament outlawed policies on the lives of people the buyer had no interest in, which had been taken out as wagers.

A short circuit starts a fire that damages Rs. 4 lakh of goods. The insurer pays Rs. 4 lakh, not more, even if her policy is for Rs. 10 lakh. That is indemnity: insurance puts you back where you were, never better off, or a fire would be worth starting. It holds for property, not for a life, which can’t be priced, so a life policy simply pays its agreed sum. If Srijana had insured the same goods with two companies, she still couldn’t collect Rs. 8 lakh; the two would share the Rs. 4 lakh between them. That is contribution.

Suppose the fire started in the faulty wiring of the shop next door, and its owner was to blame. Having paid Srijana, the insurer takes over her right to claim from him. That is subrogation, and it stops her being paid twice, once by the insurer and once by the neighbor. Last, the insurer pays because fire was the main and real cause of the loss. If the goods had been spoiled by rats and damp over months, it wouldn’t pay, even though they were found spoiled after the fire. That is proximate cause: the loss is paid when the risk insured against is what really caused it. Some lists add a seventh: Srijana must still try to put the fire out, or the insurer may pay less.

Every policy has two sides. The insurer, बिमक (bimak), takes the risk. The insured, बिमित (bimit), pays the premium, बिमा शुल्क (bima shulka). The written contract is the policy, बिमालेख (bimalekh): what is covered, for how much, for how long, and what is left out. Like any contract it needs an offer and an acceptance: the insured proposes, and the insurer may accept or refuse. Each side must also give something, the premium on one side and the promise to pay on the other, and the principles above are built into it. The two great families of policies are life insurance, which pays on a death or at the end of a set term, and everything else, non-life insurance: fire, vehicles, crops, cattle, health, travel. A non-life policy usually runs for a year and has to be renewed.

Insurers have insurers

The law of large numbers has one weak spot. It works only when the chances are separate, when one house burning doesn’t make the next one more likely to burn. Fires in different towns are like that. An earthquake isn’t. When the ground shakes, it shakes under every house in a valley at the same moment, and the insurer that covered ten thousand of them faces ten thousand claims in a day.

So insurers insure themselves. Reinsurance is insurance bought by an insurer: it passes part of its risks on to a bigger company, often one abroad, that holds risks from all over the world. An earthquake in Nepal and a storm in the Caribbean don’t happen because of each other, so a company that holds both can count on the law of large numbers again, at the scale of the whole planet. Reinsurance is very old too. The first contract on record, from 1370, passed on part of the risk of a voyage from Genoa to Bruges: the first insurer kept the Mediterranean stretch and passed on the rest, from Cadiz through the Bay of Biscay. Today it is arranged in two ways: a treaty, which covers a whole class of an insurer’s policies for a period, and facultative reinsurance, bought one large or unusual policy at a time. The reinsurer may take an agreed share of every policy and of every claim, or pay only the part of a loss above an agreed amount.

Nepal learned what this means on 25 April 2015. The earthquake and its aftershocks did damage and losses that the government put at about Rs. 706 billion. By June of that year insurers in Nepal faced claims of more than Rs. 16 billion, a large share of it reinsured abroad, and the government’s assessment still called the part left to Nepal’s own insurers substantial. The claims came to little more than two rupees in every hundred of the loss. Most of the loss had no insurance behind it at all.

The plain thing to take from all this is that insurance doesn’t make losses smaller or rarer; it moves them. A crowd of separate chances is far steadier than any one of them, so many people can each pay a small, certain amount and the few who are struck aren’t ruined. It costs more than the average loss, and it is worth that for the losses that would ruin you. The सी गुथी rests on the same core: nobody should face the worst day alone.

Try this

  • In the widget, keep the chance at 1 in 100 and switch between No cover and With cover for each of the five hundreds. Count how many households in each would have done better without cover, and how many would have been ruined.
  • Work out a fair premium on paper: a motorcycle worth Rs. 2 lakh has a 1 in 50 chance a year of being stolen. What is the average loss a year, and what premium would cover it with a quarter added?
  • List five risks your family lives with. For each, decide whether you would keep it, avoid it, shrink it or pass it on, and why.
  • Ask at home whether anything in the family is insured, and if there is a policy, find the page that lists what it doesn’t cover.

Found a mistake? Please tell us.