Pindyck and Rubinfeld, Chapter 17 Sections 17.1 and 17.2 Asymmetric information can cause a competitive equilibrium allocation to be inefficient. A market has asymmetric information when some agents know more than others in the market. Sellers of a good often know more about the good than buyers. For example, workers know more about their abilities than a firm that is considering hiring them. Also, managers know more about the firm s costs and investment opportunities than the owners of the firm. Some arrangements can be explained by asymmetric information. For example, car companies offer warranties and service for new cars. Firms and workers sign contracts that specify incentives and rewards. Shareholders of corporations design specific incentive structures for their managers (CEO). Quality uncertainty and the market for lemons Why does the resale value of a car drop so dramatically after it is driven only a few miles? Even if a car is in perfect condition, the seller can get only much less than they paid for it. This is partly because people think that the seller may be selling the car because it is of bad quality (a lemon). They cannot be sure about its quality because of asymmetric information. Even if the prospective buyer hires a mechanic to inspect the car, they still can t know as much about it as the previous owner. In a paper called The Market for Lemons: Quality Uncertainty and the Market Mechanism (1970), George Akerlof explained how asymmetric information can lead to an inefficiently low amount of trade in the good. This analysis works for insurance markets, financial markets and the labor market. Consider the market for used cars. Suppose there are two kinds of used cars: high-quality and low-quality. First suppose that buyers and sellers can both tell the quality of a car. Then there will be two markets, one for high-quality and one for low-quality cars. The demand curve D H for the high-quality cars will lie to the right of the demand curve D L for low-quality cars, because buyers are willing to pay more for a given quantity of high-quality cars. The supply curve S H for high-quality cars lies to the left of the supply curve S L for low-quality cars, because to be willing to sell a given quantity of high-quality cars, sellers must receive a higher price. Therefore the equilibrium price for the high-quality cars will be higher than the equilibrium price for the low-quality cars. In reality, sellers of used cars know more about their quality than buyers. Suppose that half the cars are high-quality and half are low quality. Sellers know their car s quality. Buyers might start out believing that there is a 50:50 chance 1
a car is high-quality. The demand curve for cars corresponds to the demand for a medium-quality car. It is higher than the demand curve for low-quality cars with full information, but lower than the demand curve for high-quality cars with full information. Its intersection with the supply curve for high-quality cars occurs at a point lower then the full-information equilibrium point. So fewer high-quality cars are sold relative to full information. The intersection of the demand curve with the supply curve for low-quality cars occurs at a point higher than the fullinformation equilibrium point. So more low-quality cars are sold than with full information. At some point, consumers will realize that the proportion of low-quality cars to high-quality cars is higher than 50:50, say 75:25. The demand curve for cars will shift further to the left, and this process may continue until only low-quality cars are sold, and no high-quality cars are sold. This is called unraveling. When applied to the insurance market, it is called a death spiral. This situation is extreme. The markets might come to equilibrium at a price such that some high-quality cars are sold. But the number of high-quality cars sold will certainly be lower than the full-information number, and the fraction of high-quality cars will be smaller. It is uncertain whether the number of lowquality cars sold is lower or higher than the full-information number (it depends on the shape of the demand curve). This situation is a kind of market failure. There are owners of high-quality cars who value their car less than potential buyers. But the trade between them is not made, which is inefficient. Adverse selection exists when lack of information causes goods of different quality to be sold indistinguishably, at the same price, leading to too much of the low-quality good and too little of the high-quality good. Other examples of asymmetric information: The market for insurance. People over 65 have trouble buying health insurance at any price. It is true that people over 65 have a higher risk of illness than the general population, but the price could reflect that higher risk. However, such insurance is often difficult to find at all. This is due to asymmetric information: Individuals know more about their health status than an insurer does. So the insurer has to guess about someone s health status based on the probabilities of different sicknesses in the general population. If there is a unique price for health insurance that does not depend on health status, some very healthy people may find the price too expensive based on their perceived probability of getting sick (we are talking about people over 65 because they may have better knowledge about their health than younger people, having lived longer). They will not buy insurance. The insurer will after a time realize that the probability of a person buying insurance being sick is higher than the probability for the general population, and will raise the premium. Then again the healthier ones will drop out of the market, and the 2
premium rises again. This process continues until some equilibrium is reached. At that point, either insurance is very expensive, and is bought only by very unhealthy people, or insurance is not sold at all. In reality, equilibrium may be reached with some healthier people buying insurance, because people can not perfectly predict their chances of getting sick, and some relatively healthy people may be very risk-averse. In California, just before the vote on the health bill, one of the big health insurance companies raised the premium for individuals by 39%. They justified it by saying that after the financial crisis, healthier people stopped buying insurance at all, and they were left with an unhealthier pool. This seems to be leading to a death spiral. A solution to the adverse selection problem is to pool risks. Instead of insuring individuals, an insurance company offers a plan to a group. For example, much of health insurance in the United States is sold to employers, who use it to cover their employees as a group. This grouping makes the average probability of someone getting sick closer to the population average. Unless people choose their job based on the health insurance policy it offers (which is unlikely), the average probability of someone getting sick is not likely to be very different from that of the general population if the employer has a relatively large number of employees. If anything, the pool of workers in a firm is likely to be healthier on average than the general population, since it is difficult for people with chronic health problems to work. In most firms that offer health insurance (about half of firms, mostly the bigger ones), workers are offered such a favorable price for health insurance that it is advantageous for them to accept it. This is partly due to the tax-exempt nature of payment by health insurance, and partly due to the pooling. Then the adverse selection problem becomes much smaller. The market for credit. High quality buyers of credit are those who pay their debts, and low-quality buyers are those who don t. Lenders want to charge a higher interest rate to the low-quality buyers and a lower interest rate to the high-quality ones. If they can t distinguish between them, they have to charge an average interest rate based on the expected quality of a buyer. But the same problem will happen as in the insurance market: The high-quality ones, who are less likely to want credit, will drop out, leaving the lower-quality buyers in the market. This problem can be alleviated to some extent if the credit card companies check credit histories. This allows the companies to offer different prices to different types of customers. Other markets where asymmetric information exists include: retail stores will the item purchased be defective or not? The store knows this 3
better than the buyer. dealers of rare stamps, paintings and other art objects are they real or counterfeit? The dealer is likely to know this better than the buyer. roofers, plumbers and electricians before the work is done, the customer does not know how good a job they will do. the labor market in general potential employees know their skills and how hard they are going to work better than the potential employers. The market for unemployment insurance suffers from both adverse selection and moral hazard. The adverse selection that would exist in a private unemployment insurance market already causes that market not to exist (people know very well how likely they are to lose a job). Government insurance covers a large fraction of workers, avoiding the adverse selection problem, but there is still a moral hazard problem. Moral hazard refers to actions taken by one side in a contract that are hidden from the other side but affect the quality of the good contracted for. While the unemployed person is receiving income replacing work income, they tend to work less hard at finding a new job. In all these markets, unless sellers can provide information to buyers about the quality of their product, low-quality goods will drive out high-quality ones. If firms can develop a reputation, buyers can use this to decide on their demand for the firm s product. Then they will pay more for the high-quality products, and the low-quality products will not drive out the high-quality ones. The recent credit crisis was to some extent caused by asymmetric information about the risk of lending money to financial firms. Banks had thought that overnight loans to firms like Lehman Brothers were risk-free. When Lehman Brothers defaulted on a 2 billion dollar loan from another branch of the company, banks realized that some financial firms were risky, but they couldn t tell which ones. They then stopped lending to anyone. In some cases, businesses cannot develop a reputation. For example, most customers of motels stop there only once or not very often. In that case standardization has the purpose of making quality known (it may not be very high quality, but at least the quality is known). Restaurant and hotel chains benefit from standardization. Market signaling is another way in which buyers and sellers can get around the asymmetric information problem. Michael Spence showed that in some markets sellers send buyers signals that tell them about the quality of the product. Consider a labor market. A firm doesn t know how hard people are going to work until after it has hired them and let them work for a while. So why doesn t the firm hire workers at random, see how well they work, and then fire the ones that don t work well? Such a policy would cost the firm a lot, because firms often have to give severance pay to anyone fired after working for more than a few months. Also, on-the-job training can be very important, and firms have to 4
invest a lot in the workers it does hire. So firms have an interest in hiring the right people to begin with. A signal that actually conveys information about the worker s quality has to be more difficult for low-quality workers to send (otherwise it would be impossible to distinguish between the types). Education is considered to be such a signal, because it is easier for people with high ability to acquire education (although people whose parents have little money might not be able to get an education regardless of their ability). Even if education did nothing to improve a person s abilities or productivity, it would still be valuable as a signal. A simple model of job market signaling. Suppose that workers can be divided into low-productivity (group I) and high-productivity (group II) workers. Group I workers average and marginal product is 1, and group II workers average and marginal product is 2. The firms in the market are competitive, the products sell for $10,000 per year, and the firms expect the employees to work 10 years on average. Half the workers are in Group I and half the workers are in Group II. So the average productivity of all the workers is 1.5. The expected revenue generated by Group I workers is 1 10, 000 10 = 100, 000, and the expected revenue generated by Group II workers is 2 10, 000 10 = 200, 000. If firms could identify what group a person belonged to, they would pay them their marginal revenue product. Group I workers would get 10, 000 per year and group 2 workers would get 20, 000 per year. (We have assumed that each worker s marginal product equals their average product because this avoids the problem of having to figure out entry of firms into the market or exit from the market when there are positive or negative profits. Assume that labor is the only input. If average product were less than marginal product, firms would be making negative profits when they pay workers their marginal revenue product. If average product were greater than marginal product, firms would be making positive profits when they pay workers their marginal revenue product. When average product equals marginal product, a firm s profits are revenue minus cost, which is zero no matter how many workers are employed when the firm pays the workers their marginal revenue product.) If firms cannot identify a worker s productivity before hiring, they pay workers an average marginal revenue product per year, which is 15, 000. Group I people would be better off relative to full information, and Group II people would be worse off. Now suppose that signaling is possible through education. Let y represent years of higher education (suppose that y summarizes all attributes of education). There is a cost of education that depends on y. C I (y) = 40, 000y is the cost to group I people, and C II (y) = 20, 000y is the cost to group II people. Notice 5
that the cost of education is lower for the higher-productivity workers getting education is easier for them. Suppose that education does nothing to improve one s productivity. Its only purpose is as a signaling device. We want to find out whether there is an equilibrium in which the different types buy different levels of education. Consider a possible equilibrium in which firms use a decision rule: Anyone with education y or above gets paid $20,000, and anyone with education level below y gets paid $10,000. For this decision rule to be part of an equilibrium, those with education equal to or above y must actually be Group II, and those with education below y must be Group I. Otherwise, the firm would want to change its decision rule. If a firm pays $20,000 to a Group I member, it will be losing money. If a firm pays $10,000 to a group II member, another firm with a better decision rule can take that worker away. Therefore, given that firms are using this decision rule, we must determine how much education each group gets. Group I members will get education y if 200, 000 40, 000y 100, 000, that is, if their payoff over the ten years is greater with education y (where firms think they are high-productivity) than without any education. They will buy no education if 200, 000 40, 000y < 100, 000, that is, their payoff from getting education y is less than their payoff from getting no education. So they get education y if y 2.5 and 0 otherwise. Notice that there is no reason to get an education level between 0 and y, because the firms will think the worker is Group I if they choose such an education level. So the choice is between 0 and y. Group II members will get education y if and only if 200, 000 20, 000y 100, 000. So they get education y if y 5. So as long as y is between 2.5 and 5, Group II will get education y and Group I will get no education. So the firm s decision rule will be compatible with an equilibrium for those values of y. Guarantees and warranties are a way that sellers of durable goods can signal their product s quality. A warranty will cost more to a seller of a low-quality good than to a seller of a high-quality good, since low-quality goods are more likely to break. 6