Chapter 15
Understanding the Dangers of Asymmetric Information
In This Chapter
Identifying asymmetric information in action
Encouraging people to behave “badly”
During the banking crisis of 2008–9, one of the key financial markets — the interbank lending market — screeched to a halt. The problem was that any bank needing to borrow had become a signal to a lender that the bank was in a poor financial state. A bank trying to engage in the routine behavior of borrowing from another bank was telling the potential lender that it needed credit, and so inadvertently identified itself as a risk. As a result, until more normal trading conditions returned to the market, no one would risk lending to anyone else.
The only lenders left in the game at that point were the central banks. Amid cries for bailouts, the central bankers had to ponder a difficult question: If you establish the principle that a central bank will bail out a bank in distress, aren’t you saying to those banks that they can always count on you? And doesn’t that mean that banks will have less incentive to keep themselves honest and healthy, financially speaking? After all, if they don’t, they can always come to the central bank for a bailout.
These two cases — signaling that you’re risky by your need to borrow and giving someone a bad incentive by bailing them out — are aspects of an important economic problem called asymmetric information, where one side in a trade has more information than the other side. The first case is an example of adverse selection: The ones who want to borrow the most are the bad risks for lenders. The second is an example of moral hazard: The terms of the bailout provide an incentive to behave badly. Whereas previous models (for instance, the ones in Chapters 9–14) assume perfect information on both sides of a deal, sellers and buyers, asymmetric information in a trade can make markets break down and “fail,” leading to no trading taking place at all.
Asymmetric information can be a problem every time a seller knows more about the quality of her product than a buyer can pre-sale. It also lies at the heart of labor market problems — when candidates know more about their abilities than hirers. And insurance problems — when buyers know their own risk levels better than insurers. They’re also the bane of gambling industries, which rely on betters not knowing more about the market than bookies.
In the case of asymmetric information, what you don’t know can hurt you — which is why you need to read this chapter.
Seeing the Effects of Asymmetric Information
The market models developed in Chapters 9–14 are built on the assumptions that both parties to a trade have full knowledge of all information relevant to the trade or, alternatively, that they’re equally ignorant. When you change this core assumption, things start to look very different indeed, as this section makes clear.
Visiting the market for lemons and plums
One way in which asymmetric information can lead to market failure is when it creates a negative externality (a cost imposed on a third party not included in the transaction — check out Chapter 14 for more about externalities).
In 1970, George Akerlof published an extremely important article on the economics of asymmetric information. “The Market for Lemons” shows very simply how asymmetric information affects markets — even shutting them down because bad trades drive out good.
We begin by assuming that a used car market has two kinds of car:
- Lemons: Real clunkers — cars of poor quality.
- Plums: Cars of good quality.
Suppose 100 people are selling their cars, and all that buyers know is that 50 of the cars are plums and 50 are lemons. Each seller knows the quality of its car, but the buyers don’t know until they buy. Caveat emptor, folks. (“Let the buyer beware.”)
Now we put in some prices. Suppose the sellers are willing to sell a plum for $2,000 and a lemon for $1,000. That would, fairly, show that their price is related to the quality. Imagine that the buyers are willing to pay a little more than that: $2,400 for a plum and $1,200 for a lemon. What’s the equilibrium for this market?
Well, if you could immediately tell whether you’re buying a plum or a lemon, you’d just have two separate equilibria for each kind of car where a plum went for $2,000–2,400 and a lemon for $1,000–1,200.
But buyers can’t tell here. So their best strategy is to offer the expected value of the car — which you get by multiplying the buyers’ offer by the probability of getting that type of car. If they know that 50 percent of the cars are lemons, then the expected value (EV) of a used car is as follows:
That’s the maximum that a buyer would bid for a car of unknown — some may say dubious! — quality. But now suppose the sellers know that buyers would only pay $1,800 for a used car. What do they do?
Sellers of plums leave the market, because buyers are willing to pay only $1,800 for a car, and plum sellers won’t sell for less than $2,000. But wait, it gets worse. Now buyers know that only lemons are left in the market. They won’t pay more than $1,200 for a car that is a lemon, so the market equilibrium drives out good cars and leaves only the market for lemons. The externality makes bad cars drive out good, and the market for plums fails.
Signaling your risk by buying cover: Adverse selection
Sometimes the externality arising from asymmetric information gets so bad, it can destroy an entire market. In the case of interbank lending during the financial crisis, it very nearly did. If lending banks knew that the only borrowers were poor risks, they couldn’t very well lend at the prevailing rate. The problem was that some banks were in real trouble, and as long as they had enough effect on the lenders’ view of the average likelihood of failure in the market as a whole, the lenders couldn’t lend to them.
This problem relates to hidden information: The lenders were unable to directly observe the quality of the borrowers, but they could read the need for capital borrowing as a signal (Chapter 19 talks more about signals). Alarming signals eventually made lenders almost disappear from the market.
Demonstrating adverse selection with a quality-selection model
A simple way of seeing adverse selection is with a quality-selection model in which producers must choose whether to produce a high-quality item as opposed to a low-quality item. They choose high quality with a probability q, and what you want to know is what size of q will keep the market working.
We start by putting some prices into the model so that you can see it in action. Suppose the product is a tradable good (that is, rival and excludable — see Chapter 14 for definitions), such as a basic cellphone. A better-quality one would cost you $100, and a lower-quality one $64. Suppose for now that manufacturers can make either type of phone for $85. We make everything nice and easy by assuming that the industry is in perfect competition — so that the phones aren’t differentiated by obvious features, and the price equals marginal cost.
Starting off with that information, three things can happen:
- If high-quality manufacturers are the only ones in the market, the trades will occur somewhere between $85 and $100, but perfect competition means that the price gets competed down to $85.
- If low-quality manufacturers are the only ones in the market, customers are willing to pay only $64. The manufacturing cost is $85, so no phones are produced, and no market exists.
- If high- and low-quality manufacturers are in the market, the market clears only if q is big enough to ensure that the expected value (EV) of customers is also higher than the cost of production. So you need to know what value of q leads to EV being $85.
Finding the value of q that clears the market therefore means solving the expected value equation:
The answer is 7/12.
Graphing the equilibrium result allows you to see how this market failure manifests (check out Figure 15-1). Putting price on the vertical axis and q on the horizontal, you can show the market failure as a blank area where no items are sold. The supply is horizontal, as in perfect competition, but only above q = 7/12. Above that, the diagonal shows consumers’ willingness to pay — that is, the expected value as q goes up (it can’t be bigger than 1). Between the two is a shaded area of consumer surplus — excess benefit the consumers get when they’re willing to pay more than the $85 that producers end up charging under perfect competition.
© John Wiley & Sons, Inc.
q = Probability of high quality product; p = Price
Figure 15-1: Showing the quality selection model in a diagram.
Note that the shaded oblong — no market equilibrium — reflects the external cost of producers not producing a high quality product. If enough of them exist, they destroy the entire market. But if each individual producer doesn’t believe that they, individually, contribute to the average quality in the marketplace, they have an incentive not to choose to produce at a high quality. Thus, poorer products tend to drive out better ones, and consumers lose out.
Seeing what happens when producers choose product quality
Adapting the model to allow producers to choose the quality of the product they’ll produce shows how adverse selection can destroy both high- and low-quality markets. Suppose now that the cost of making a high-quality phone is a little more — say, $90 — and the manufacturer can choose which to make.
Consider the position of a high-quality manufacturer. It knows that it has only a small fraction of the market, and so it too has an incentive to produce a lower-quality phone. Unfortunately, the same reasoning is true of all the higher-quality manufacturers. Because a consumer will pay only $64 for a low-quality phone, no equilibrium exists, and high-and low-quality markets disappear.
Dealing with adverse selection through compulsory coverage
In practice, adverse selection is generally associated with health insurance markets, where consumers differ according to the risk they face of requiring costly medical treatment. If one population has a higher risk than another, and the insurance company does not know who is in which risk category, it faces a problem of adverse selection. If it tries to price insurance (that is, set premiums) using the average risk in the overall population, lower-risk consumers may not find insurance to be worth it, and so they wouldn’t buy, and higher-risk people would make more claims. Therefore, the bad risks drive out the good — adverse selection in action, again.
The provision of health insurance is such an important policy issue, it is worthwhile spending some time understanding the economics of health insurance. As an example, suppose each consumer belongs to one of four different risk categories, where a risk category is the probability of needing a medical treatment that costs $10,000. There are a large and equal number of consumers in each group. Each consumer knows her risk, but insurers cannot tell which group a consumer belongs to.
The willingness to pay (WTP) of consumers in each group is in the following table. We also give the actuarially fair premium, which is the expected loss a company faces when it insures a consumer. It is equal to $10,000 times the risk or probability of needing treatment, and it is always lower than the willingness to pay since consumers exhibit risk aversion.
|
Risk |
20% |
40% |
60% |
80% |
|
WTP |
2,500 |
5,200 |
6,800 |
8,500 |
|
Actuarially fair premium |
2,000 |
4,000 |
6,000 |
8,000 |
|
Potential surplus |
500 |
1,200 |
800 |
500 |
If everyone has insurance, then the company would face a 50% chance of a loss in the population and would have to charge at least $5,000 to avoid a loss. In a competitive market, this would be the price of health insurance. However, at this price, consumers in the lowest-risk 20% category would not buy insurance because they would find the premium too high. However, if only the 40%, 60%, 80% risks are in the market, then the chance of a loss is 60%, and a company would need to charge $6,000 to avoid losses. If the price is $6,000, only the 60% and 80% risks will want to buy insurance, but then the overall risk rises to 70%. A company would need to raise the premium to $7,000, driving out the 60% group and hence the equilibrium premium for insurance will be $8,000, and only the highest-risk agents will insure.
This is not an efficient outcome — if insurance could be provided on actuarially fair terms to each group individually, this would generate a surplus for each group. However, the adverse selection problem leads to the collapse of the insurance market, leaving many consumers uninsured.
Could we do any better? If there is no way to obtain the private information, then the policy maker faces the same constraints as the market, and the insurance market creates only $500 of surplus per high-risk person.
Making participation mandatory can create more total surplus. In this case, insurers would know that the aggregate risk is 50%, and the competitive price of insurance would be 5,000. At this price, the surplus of the agents would look like this:
|
Risk |
20 |
40 |
60 |
80 |
|
WTP |
2,500 |
5,200 |
6,800 |
8,500 |
|
Actuarially fair premium |
2,000 |
4,000 |
6,000 |
8,000 |
|
Pooled premium |
5,000 |
5,000 |
5,000 |
5,000 |
|
Surplus |
–2,500 |
200 |
1,800 |
3,500 |
Note, however, that even though the total surplus has increased, this is not a Pareto improvement — the low-risk consumers are worse off with mandatory participation, and there is no way to compensate them because that would require that we can somehow identify them, which by the assumption of private information is impossible.
One way of dealing with the problem of health insurance is to require everyone to buy a product with the average-risk profile:
- High-risk people are better off, because they can get insurance that they’d have otherwise been priced out of getting.
- Lower-risk people are better off only insofar as they can get an insurance policy that they wouldn’t otherwise be offered.
Changing Your Behavior because of Asymmetric Information
In October 2008, recapitalizing the banks by providing huge amounts of capital was part of the policy makers’ solution to the crisis in the financial markets. The principal aim was to prevent the financial system from descending into an even greater crisis. One side effect of this policy, however, was the need to manage moral hazard.
Moral hazard, as mentioned earlier, is a situation that gives an incentive to someone to behave “badly.” Suppose, for instance, that you have superb all-coverage home insurance. You may not have an incentive to fit a good-quality lock on your back door, because you know that if someone gets into your house, the insurer, and not you, is going to pay for the loss.
In the case of the bank bailouts, many commentators saw the bailouts as an invitation for banks to lend recklessly. If the banks knew that the government would pick up the bill, they had no incentive to stop lending recklessly — when things worked out well, they made large profits, and when they didn’t, they weren’t taking the losses.
Managing moral hazard
Policy makers and companies seek to manage moral hazard in different ways, including the following:
- Banking: The U.S. government has confronted the problem of moral hazard in the recent financial crisis. The government introduced the Troubled Asset Relief Program (TARP) as a mechanism for purchasing the assetsof troubled financial institutions and putting them on the path to recovery. Other bailout programs included the federal takeover of Fannie Mae and Freddie Mac and the public ownership of GM and AIG. Whether bailing out the banks and other institutions will rise to the problem of “too big to fail” is still an open moral hazard question.
- Insurance: Companies cover perhaps only 90 percent of the losses to ensure that the insured customer bears some of the risk. The risk born by the customer is the deductible in the plan and provides at least some incentive to take care.
If an insurer could observe accurately how much care a customer would take, no moral hazard problem would exist, and the market would work as efficiently as it should. But because the insurer doesn’t know, the outcome of moral hazard is under-insurance — each consumer would like to buy more insurance, but the insurance companies would be less likely to provide it because rational customers who would buy the full insurance product would have the incentive to take less care. That’s why insurers insist on the type of deductible clauses just mentioned.
Incentivizing: Asymmetric information in contracts
One classic problem where asymmetric information and moral hazard pop up is when you want to get someone to do something for you and the other person knows more about whether she’s going to do it or not.
Considering the principal-agent problem
One problem of this kind is the principal-agent problem. It concerns a principal wanting to get an agent to do something for him and an agent having her own interests that she wants to take care of too.
A good example is a Hollywood agent who gets a percentage of the fees paid to an actor she represents. The actor — the principal here — has an interest in maximizing his lifetime career standing and is therefore particular about what kind of jobs he wants to take. The agent, however, just wants to maximize her percentage and tries to get the actor signed up for as many jobs as possible, regardless of quality.
More seriously, the principal-agent problem also exists in the structure of an ordinary company. Shareholders own the company and are the principals, but managers are their agents. Management may want to act in their own interests — manage a big company instead of a smaller one — rather than in the principals’ interests. Thus shareholders need a scheme for making the management act in their interests. The simplest, though not always most effective way, is to pay managers in company stock, so that they take the same risks and gain the same benefits as do the owners. But that too may not always work.
Modeling incentives for contracts
In a contract, an owner (the principal) wants a manager (the agent) to put in an effort or care level, x*, which is known to the manager but not the owner. The manager wants to choose her own utility-maximizing level of x. What should the owner do?
Simple: The profit-maximizing level of x has to be the level where the marginal product to using x amount of effort is equal to the cost of performing the effort. So the owner wants to choose x* on that basis.
The owner needs to get this level of effort into an incentive-compatible contract — one that assures that the utility to the manager of doing anything other than x* is less than the utility of doing x*.
Here are three methods the owner can use to achieve this aim:
- Lump sum fee: The owner can charge the manager a lump sum fee, R, so that after the manager has paid R — you might think of this amount as a franchise fee — any profit belongs to the manager. In that way, the manager chooses x* in the same way as if she were the owner.
- Wages: Set wages so that the marginal product of the manager’s work equals the wage, and the manager will want to choose the level of effort that maximizes this relationship.
- Fixed fee: The owner can pay the manager a bonus if x* is achieved and nothing if it’s not.
What won’t work is a share scheme where manager and owner divvy up the output. The manager won’t be trying to maximize marginal product but only the share of the product that she gets. Calling that share α, the manager would be trying to set α times the marginal product equal to marginal cost, and that won’t give the optimal output x*.
Now suppose because of random events the owner is never able to infer from the output whether the manager has put in the correct amount of effort. This certainly makes the model more realistic and in that case, we can deduce that
- If the manager is more risk averse than the owner, and output is random, the lump sum franchise fee must be reduced so that the manager is willing to bear the risk of uncertain output. (This is generally a problem with franchises.)
- Wage setting depends on owners being able to assess the amount of labor input. If the owner can’t do that, using wages becomes inefficient.
- In a take-it-or-leave-it case, the manager is at risk of losing everything if the output can’t be controlled from her desk, which places all the risk on the manager and is unlikely to be agreed to as a contract.
- If the output fluctuates, the profit-sharing solution means both parties take on some of the risk, though not all. In turn, the manager has the incentive to produce some output but isn’t carrying all the risk.
The value of an incentive contract depends crucially on the degree to which the principal is able to monitor the efforts of the agent. If the principal can find out at some point, even if not up front, what the agent is doing, then franchising, wages, and bonus payments will ultimately all lead to the same outcome. If risk is present, so that neither the principal nor agent knows how things will turn out, or if the principal is never able to monitor the activity of the agent, profit-sharing may be the most efficient solution.