Moral Hazard: How Protection Changes What People Do

Moral hazard is the change in behaviour that follows from being protected against a loss, and it shows up before anything happens rather than afterwards. Moral hazard works through prices. A funder who believes a claim will be made good whatever occurs stops charging for the possibility that it will not be, and a price that has stopped separating one institution from another has stopped disciplining either of them.

Deposit insurance settles what a protection scheme makes good and what it leaves alone, and it ends on one sentence worth picking up: a claimSomebody's right to be paid, which is an asset to the holder and a liability to the party that has to pay. Each of the two sides of a claim is settled separately. that is not covered still has a holder with every reason to keep watching. Turned around, that is the subject here. A claim that is covered has a holder with much less reason to watch, and once that has been said out loud it has to be followed wherever it goes.

The idea goes somewhere uncomfortable, and most treatments of this subject flinch at exactly the wrong moment. A treatment that flinches starts with a real idea and ends with an accusation, usually aimed at the people who had the least to do with it. The whole idea runs through one sentence: the extra a funder charges above what a safe claim earns is compensation for the possibility of not being repaid, so if that possibility is believed to have gone, the compensation has nothing left to compensate for and goes with it. Nobody has to decide anything. Nobody has to be dishonest. A belief moves, and a price follows it.

What is moral hazard, and why is it not an accusation?

Start with the name. The name is the worst thing about the idea, and it does real damage every time somebody meets it. The word moral makes this sound like a verdict on somebody's honesty. The term is nothing of the kind. Moral hazard is a term of art borrowed from insurance, where it has always described a change in behaviour rather than a change in character. Named today, it would probably be called something duller and far more accurate.

Moral hazard is what a reasonable party does when the consequences of an action change, and it would be strange if they did not. That is the whole definition. Nobody has to be greedy. Nobody has to be careless. A party choosing between two courses on what each one costs will choose differently once one of those costs has stopped moving. Nobody even has to be aware that anything has changed, and nobody has to be able to say why.

Here is the version to keep hold of. A shopkeeper installs a shutter on the front of the shop. Afterwards, the shopkeeper leaves the counter for a few minutes at a time in a way they never used to. There is nothing wrong with the shopkeeper. Nobody has become less careful as a person. The shutter has changed what leaving the counter costs, and a sensible person responds to that. Any impulse to disapprove of the shopkeeper runs into the fact that the shutter was installed precisely so that leaving the counter would cost less, and that the shopkeeper is doing what it was for.

Where does the whole mechanism sit, if not in a decision?

The mechanism sits in a price. Everything else hangs from the price, so it is worth going slowly. A funder lending to an institution charges a rate. The rate splits into two parts, and only one of the two matters. The first part is what a safe claim earns, and it compensates for parting with money for a period rather than for anything about this institution in particular. The second part is the extra above that, and it is there for exactly one reason: the possibility that the money does not come back.

Now change one thing and change nothing else. Suppose the funder comes to believe that the claim will be made good whatever happens to the institution. Not that the institution has improved. Not that it holds different things. Simply that if anything went wrong, somebody would stand behind the claim. Ask what happens to the second part of the rate.

The second part has nothing left to compensate for, so it goes, and that is the entire event: nobody decided anything, nothing happened to the institution, and a belief moved with a price behind it. This is why moral hazard is so slippery to talk about. There is no meeting. There is no announcement. There is no moment anybody could point at afterwards and call the beginning of it. There is a rate that used to have two parts and now has closer to one.

The household version is a rent guarantee. If a landlord knows that somebody else will pay the rent when the tenant does not, the deposit the landlord asks for goes down, and the landlord has not become careless. The deposit was compensation for a possibility, and somebody else has taken the possibility away. The deposit falls automatically, without anybody thinking about it as a change in behaviour at all.

Why does a price that stops separating stop disciplining?

Two institutions with different amounts of their own money behind what they hold are not the same risk to a funder, and a funder paying attention charges them differently. The difference in price is doing a job, and the job is discipline. The institution taking more risk pays more for its funding, so the cost of taking that risk arrives immediately rather than one day. A cost that arrives immediately is a constraint, whether or not anybody calls it one.

Take two institutions, both invented. Suvarna Commercial Bank Limited has net worth of Rs 24,000 crore against total assets of Rs 2,40,000 crore, so net worth is 10.0 per cent OF ASSETS. Deposits account for Rs 1,92,000 crore of what funds those assets, so Rs 24,000 crore of funding sits between the deposits and the net worth, and the contents of that layer are something a reader has to go and find out. Rukmini Finance Limited has net worth of Rs 3,600 crore against assets of Rs 18,000 crore, so net worth is 20.0 per cent OF ASSETS. Read the same fact as leverageWhat an institution holds, expressed as a multiple of the money its owners have put in. The same fact as the share of net worth, read upside down, and settled separately. and the bank is at 10.0 times equity against the finance company at 5.0 times. Same two facts, two very different looking numbers. The base has to be said out loud every single time.

A funder lending to each of them is standing at a different distance from the absorbing layerThe net worth standing between a loss and the people who funded the institution. The owners taking the first loss is settled separately, and only the thickness of the layer matters here., and that distance is what a price would be compensating for. Now suppose a protection is believed to reach both of them. The two prices converge. The cost of standing closer to the layer stops arriving. The constraint that used to bind has loosened without a single person voting to loosen it, and that is what makes this so hard to notice while it is happening.

The argument runs from what a price is for, and never from two prices set side by side.

What would a funder be pricing, if there were a price to look at?

The honest shape of this comparison is one that stops one step short. The same proportional illustration read against both institutions meets something different at each. A reading of 5.00 per cent OF ASSETS is Rs 12,000 crore at Suvarna Commercial Bank Limited, and Rs 12,000 crore is 50.0 per cent OF ITS NET WORTH of Rs 24,000 crore. The same 5.00 per cent OF ASSETS is Rs 900 crore at Rukmini Finance Limited, and Rs 900 crore is 25.0 per cent OF ITS NET WORTH of Rs 3,600 crore.

The multiple that carries a reading OF ASSETS into a reading OF NET WORTH is not a coincidence and it is not a second fact: it is that institution's leverage, 10.0 times at the bank and 5.0 times at the finance company. Leverage is why 5.00 becomes 50.0 at one and 25.0 at the other. Say what this is not, before anybody reads it as news. Neither institution has had a fall of 5.00 per cent in the value of what it holds. Neither has been protected by anybody. The reading is arithmetic about leverage on a balance sheet that stands, and it asserts nothing about anything having happened.

Now the absence. The missing price is deliberate rather than a gap. Comparing what the two are charged would need both prices. A cost of borrowing is available for one of them and nothing comparable for the other. The missing one cannot be built either: turning the bank's interest expended into a rate needs a complete base, and the contents of the layer between its deposits and its net worth are not known. So the two prices cannot be set side by side. The difference in the two layers still says what a price would have to be compensating for, and exactly which disclosure to ask for.

Whose behaviour changes, and does the institution have to intend it?

Three parties respond, and they are taken one at a time because running them together is how this subject goes wrong. The institution comes first. If the cost of funding stops rising with the risk being taken, then that risk is cheaper to take than it was, and more of it gets taken at the same price. The effect needs no intention at all: an institution comparing two courses on what each one costs will choose differently once one of those costs has stopped moving, and everybody involved can be behaving entirely properly throughout.

Notice what has and has not changed. Nobody at the institution has decided to take more risk. One of the two things being compared has stopped varying, so a comparison that used to come out one way now comes out the other. A shop that used to pay more for a riskier delivery route and now pays the same for both routes will find its drivers taking the shorter one, and no driver has become reckless.

The Reserve Bank of India at rbi.org.in sets what the institution has to hold against what it holds, how far it may lever it, and what liquidity it keeps in advance. The three appear as rows in the table below, with their labels on them and nothing inside them.

Why does the funder stop looking at exactly the same moment?

The professional funderSomebody who lends to an institution in size and as a business, rather than a household leaving a balance with it. Which parties do this and how they are organised is settled separately. had a job, and the job was to look. Looking costs time and money. Looking is worth doing exactly to the extent that what is found changes what is owed, and not one rupee further.

So follow the same belief through the funder rather than through the institution. If the claim will be made good whatever is found, then finding something has stopped changing anything. The monitoring stops for precisely the reason the price stops, and the two are not two effects at all: they are one effect seen twice, once as a number and once as an activity.

Readers push back at this point, and the pushback is worth answering. A funder who stops looking sounds lazy. The funder is not. The funder is doing a cost and benefit comparison that has been settled by one side of it going to zero. Any careful party would do the same. A person whose electricity bill is being paid by somebody else stops reading the meter, and nobody would describe that as negligence.

Is the depositor who stops watching part of this at all?

Here is where almost every treatment of this subject takes a wrong turn, and it takes it confidently. A depositor whose balance is good whatever happens has no reason to assess the institution holding it, and stops. The observation is true and it is easy to make. The error arrives in the sentence after it.

A depositor who stops watching is doing the entire point of the cover and not suffering a side effect of it. The whole purpose of the arrangement is that nobody should have to assess an institution before leaving money in it. A system in which every household has to read a balance sheet before opening an account is not a system anybody has ever wanted, and it would not work in any case. The assessment is difficult, the disclosure is technical, and the consequence of getting it wrong falls on the person least able to carry it.

So it is worth saying plainly. The depositor who has stopped watching is not the behaviour being described here, has failed at nothing, and is doing exactly what the arrangement was built to let them do. The Deposit Insurance and Credit Guarantee Corporation at dicgc.org.in settles what a scheme makes good, for whom, and how far, and the amount of coverThe making good of a claim by a scheme rather than by the institution that owed it. The corporation named alongside settles what is covered, for whom, and how far. is revised at that address rather than copied anywhere else. The behaviour worth examining belongs to the institution and to the professional funder, and the two groups stay firmly apart.

So is the answer simply less protection?

A reader who has followed the argument this far will reach for the obvious conclusion, and it deserves to be met head on rather than left to stand. If protection causes this, take the protection away.

The alternative to protection is not a system without this problem, it is a system in which the other outcome is always available. Deposit insurance sets out the two-outcome structure that belongs to Diamond and Dybvig, Bank Runs, Deposit Insurance and Liquidity, 1983: a rule that pays people in the order they arrive, reasoned about by people who all know the rule, has more than one stable answer, and which answer arrives depends on what everybody expects everybody else to do. Remove the protection and that second answer is back on the table every day, for every institution, whatever its condition and however carefully it is run.

So the question was never protection or none. The question is what stands alongside protection to do the work the price used to do. Which of the two arrangements is better turns on how often the second outcome actually arrives and what it costs when it does, and neither of those is a question about mechanisms.

What can replace the discipline a price used to do?

Three shapes. Each is described by what it attacks rather than by which is best.

The first is a requirement that does not depend on anybody's belief: what an institution holds against what it holds, how far it may lever it, and what liquidity it keeps in advance. None of those move when a funder's expectation moves, and that is the whole of their value. The second is a premium for the cover whose rate varies with the risk of the institution paying it. The cost of taking more risk then arrives as a bill rather than as a price a funder has stopped charging. The third is a class of claim that is explicitly not made good by anybody, held by parties who therefore keep looking, and the mechanics of such claims are settled separately.

All three attack the same joint from different sides: each one puts a cost back on the risk in a form that a belief cannot remove. That is the test for any fourth idea. Ask what happens to it when everybody changes their mind at once. If the answer is that it stops working, it was never a replacement for the price; it was another price wearing a different hat.

What is deliberately left unsaid, and what does leaving it unsaid cost?

There is a device that follows straight from everything above, and it is worth naming because it is the honest middle of this argument. If nobody knows for certain whether the claims of a particular institution would be made good, then no funder can fully stop charging for the possibility that they would not, and some of the discipline survives.

The gain is real, and so is the cost on the other side of it: uncertainty is expensive in itself, and it is exactly the uncertainty that keeps the second outcome available every day. One does not come without the other. A funder who is unsure keeps charging, and that is the thing wanted. A funder who is unsure can also change their mind quickly, and that is what the protection existed to prevent. Both halves are true at the same time.

So the arrangement holds a genuine tension rather than a solved problem, and how any particular authority resolves it is settled at that authority. The doctrine of standing behind an institution as a matter of last resortThe long standing idea that a central bank may stand behind an institution that cannot pay today, on stated terms. The doctrine itself is settled separately. belongs to Bagehot, Lombard Street, 1873, and it is settled separately rather than rebuilt here.

Why is the effect visible before anything has happened?

The property that makes the whole subject teachable is also the one readers find hardest. Moral hazard is not observed afterwards by comparing behaviour before and after some episode. There is no experiment of that shape available and none is needed.

An expectation causes the effect, and expectations are priced continuously, so the effect is present in the price of a claim today, in an institution to which nothing has happened and to which nothing may ever happen. A price is not a record of what has occurred. A price is what somebody is charging now for what they think might occur. The reading is therefore available on any ordinary working day, in a perfectly calm week, from an institution with nothing at all wrong with it.

The behaviour is reasoned out of what a price is doing rather than read off an experience, and no episode of a protection being extended is needed to reach it. Reasoning is not a weaker argument standing in for a stronger one but the argument the subject actually rests on, and an episode would add a story rather than a reason.

How does anybody actually use this idea?

Who sets each of these, and why is every value blank?

Five questions circled above belong to an authority and are revised, so a stated value would be wrong rather than merely stale on the morning it changed. The rows are drawn with their labels on and nothing inside them, and the authority is printed inside each row so that the value can be filled in from the source.

The whole argument turns on whether the premium for cover varies with the risk of the institution paying it, so the last row bites harder than the rest, and that value is still left to the source. That is not squeamishness. The answer is set at the address named inside the row, and it moves. A copy kept elsewhere would go quietly wrong rather than loudly out of date, and quietly wrong is the worse of the two failures. The extra requirements placed on an institution identified as systemically importantAn institution whose trouble would matter more than another's because of what depends on it. The qualifying test, and everything that follows from it, is settled separately. sit in the same block for the same reason.