Contemporary Financial Intermediation

close.

Basic Concepts

Practical men, who believe themselves to be quite exempt from any intellectual influences, are usually the slaves of some defunct economist. Madmen in authority, who hear voices in the air, are distilling their frenzy from academic scribbler of a few years back. I am sure that the power of vested interests is vastly exaggerated compared with the gradual encroachment of ideas.

John Maynard Keynes: The General Theory of Employment, Interest and Money, 1947

Introduction

The modern theory of financial intermediation is based on concepts developed in financial economics. These concepts are used liberally throughout the book, so it is important to understand them well. It may not be obvious at the outset why a particular concept is needed to understand banking. For example, some may question the relevance of market completeness to commercial banking. Yet, this seemingly abstract concept is central to understanding financial innovation, securitization, and the off-balance sheet activities of banks. Many other concepts such as riskless arbitrage, options, market efficiency, and informational asymmetry have long shaped other subfields of finance and are transparently of great significance for a study of banking. We have thus chosen to consolidate these concepts in this chapter, to provide easy reference for those who may be unfamiliar with them.

Risk Preferences

To understand the economic behavior of individuals, it is convenient to think of an individual as being described by a utility function that summarizes preferences over different outcomes. For a wealth level W, let U(W) represent the individual’s utility of that wealth. It is reasonable to suppose that this individual always prefers more wealth to less. This is called nonsatiation and can be expressed as U′(W) > 0, where the prime denotes a mathematical derivative. That is, at the margin, an additional unit of wealth always increases utility by some amount, however small.

An individual can usually be classified as being either risk neutral, risk averse or risk preferring. If risk neutral, the individual is indifferent between the certainty of receiving the mathematical expected value of a gamble and the uncertainty of the gamble itself. Since expected wealth is relevant for the risk neutral, and the variability of wealth is not, the utility function is linear in wealth, and the second derivative, denoted U″(W), will equal zero. Letting E(•) denote the statistical expectation operator, we can write U[E(W)] = EU(W) for a risk-neutral individual, where U [E(W)] is the utility of the expected value of W and EU(W) is the expected utility of W. For such an individual, changing the risk of an outcome has no effect on his well-being so long as the expected outcome is left unchanged.

The utility function of a risk-averse individual is concave in wealth, that is, U″(W) < 0. Such an individual prefers a certain amount to a gamble with the same expected value. Jensen ’s inequality says that

if U is (strictly) concave in W. Thus, risk-averse individuals prefer less risk to more, or equivalently, they demand a premium for being exposed to risk.

A risk-preferring individual prefers the riskier of two outcomes having the same expected value. The utility function of a risk-preferring individual is convex in wealth, that is, U″(W) > 0, Jensen’s inequality says that

if U is (strictly) convex in W.

Despite the popularity of lotteries and parimutuel betting, it is commonly assumed that individuals are risk averse. Most of finance theory is built on this assumption. Figure 1.1 depicts the different kinds of risk preferences.

FIGURE 1.1 Three Different Types of Utility Functions

In Figure 1.2 we have drawn a picture to indicate what is going on. Consider a gamble in which an individual’s wealth W can be either W1 with probability 0.5 or W2 with probability 0.5. If the individual is risk averse, then the individual has a concave utility function that may look like the curve AB. Now, the individual’s expected wealth from the gamble is E(W) = 0.5W1 + 0.5W2, which is precisely midway between W1 and W2. The utility derived from this expected wealth is given by U[E(W)] on the y-axis. However, if this individual accepts the gamble itself [with an expected value of E(W)], then the expected utility, EU(W), is midway between U(W1) and U(W2) on the y-axis, and can be read off the vertical axis as the point of intersection between the vertical line rising from the midpoint between W1 and W2 on the x-axis and the straight line connecting U(W1) and U(W2). Hence, as is clear from the picture, U[E(W)] > EU(W). The more bowed or concave the individual’s utility function, the more risk averse that individual will be and the larger will be the difference between U[E(W)] and EU(W).

FIGURE 1.2 Risk Aversion and Certainty Equivalent

We can also ask what sure payment we would have to offer to make this risk averse individual indifferent between that sure payment and the gamble. Such a sure payment is known as the certainty equivalent of the gamble. In Figure 1.2, this certainty equivalent is denoted by CE on the x-axis. Since the individual is risk averse, the certainty equivalent of the gamble is less than the expected value. Alternatively expressed, E(W) – CE is the risk premium that the risk averse individual requires in order to participate in the gamble if his alternative is to receive CE for sure.

The concept of risk aversion is used frequently in this book. For example, we use it in Chapter 3 to discuss the role of financial intermediaries in the economy. Risk aversion is also important in understanding financial innovation, deposit insurance, and a host of other issues.

Diversification

We have just seen that risk-averse individuals prefer to reduce their risk. One way to reduce risk is to diversify. The basic idea behind diversification is that if you hold numerous risky assets, your return will be more predictable, but not necessarily greater. For diversification to work, it is necessary that returns on the assets in your portfolio not be perfectly and positively correlated. Indeed, if they are so correlated, the assets are identical for practical purposes so that the opportunity to diversify is defeated. Note that risk can be classified as idiosyncratic or systematic. An idiosyncratic risk is one that stems from forces specific to the asset in question, whereas systematic risk arises from the correlation of the asset’s payoff to economy-wide phenomena such as depression. Idiosyncratic risks are diversifiable, systematic risks are not.

To see how diversification works, suppose that you hold two assets, A and B, whose returns are random variables.¹ Let the variances of these returns be σ²A and σ²B, respectively. Suppose the returns on A and B are perfectly and positively correlated, so that ρAB = 1, where ρAB is the correlation coefficient between A and B. The proportions of the portfolio’s value invested in A and B are yA and yB, respectively. Then the variance of the portfolio return is

     [1.1]

where Cov(A, B) is the covariance between the returns on A and B. Then, using

     [1.2]

we have

     [1.3]

Since ρAB = 1, the right-hand size of (1.3) is a perfect square, (yA σA + YBσB)². As long as yAσA + yBσB ≥ 0, we can write (1.3) as

     [1.4]

Thus, if ρAB = 1, the standard deviation of the portfolio return is just the weighted average of the standard deviations of the returns on assets A and B. Diversification therefore does not reduce portfolio risk when returns are perfectly and positively correlateci. For any general correlation coefficient ρAB, we can write the portfolio return variance as

     [1.5]

, that is, portfolio risk increases with the correlation between the returns on the component assets. At ρABab = 0 (uncorrelated returns),

     [1.6]

Example 1.1

. Calculate the variance of a portfolio of assets A and B, assuming first that the returns of the individual assets are perfectly positively correlated, ρAB = 1, and then that they are uncorrelated, ρAB = 0.

Solution

In the case of perfectly and positively correlated returns,

. With uncorrelated return, (. Thus, not only is this variance lower than with perfectly and positively correlated returns, but it is also lower than the variance on either of the components assets.

The maximum effect of diversification occurs when ρAB is at its minimum value of −1, that is, returns are perfectly negatively correlated. In this case

     [1.7]

so that

     [1.8]

This seems to indicate that the portfolio will have some risk, albeit lower than in the previous cases. But suppose we construct the portfolio so that the proportionate holdings of the assets are inversely related to their relative risks. That is,

     [1.9]

or

     [1.10]

Substituting (1.10) in (1.8) yields

indicating that in this special case of perfectly negatively correlated returns, portfolio risk can be reduced to zero!

Even when assets with perfectly negatively correlated returns are unavailable, we can reduce portfolio risk by adding more assets (provided they are not perfectly positively correlated with those already in the portfolio), while keeping fixed the total wealth invested in the portfolio.² To illustrate, suppose we have N assets available, each with returns pairwise uncorrelated with the returns of every other asset. In this case, a generalized version of (1.6) is

     [1.11]

where yi is the fraction of the portfolio value invested in asset i, where i = 1,…, N, and σi² is the variance of asset i. Suppose we choose yi = 1/N.

Then, defining σ²max as the maximum variance among the σi² (we assume σ²max < ∞, and permit σi² = σ² for all i in which case σ²max = σ²), (1.11) becomes

As N increases, σ²p diminishes, and, in the limit, as N goes to infinity, σ²p goes to zero. Thus, if we have sufficiently many assets with (pairwise) uncorrelated returns, we can drive portfolio risk as low as we wish and make returns as predictable as desired.

An obvious question is why investors do not drive their risks to zero. First, not all risks are diversifiable. Some contingencies affect all assets alike and consequently holding more assets will not alter the underlying uncertainty. This is the notion of force majeure in insurance. Natural calamities such as floods and earthquakes are examples, as are losses attributed to wars. Second, as the investor increases the number of securities held in the portfolio, there are obvious costs of administration. These costs restrain diversification, but in addition numerous studies indicate that a large fraction of the potential benefits of diversification are obtained by holding a relatively small number of securities. That is, the marginal benefits of diversification decline rapidly as the number of securities increases.

Finally, cross-sectional reusability of information diminishes the incentive to diversify. We shall have more to say in Chapter 3 about information reusability since this is a major motivation for the emergence of financial intermediaries. Suffice to say that if a lender invests to learn about a customer in the steel business in order to make a loan, it will see a potential benefit to lending to others in the steel business. The resulting concentration spreads the costs of becoming informed. Thus, we observe diversification within areas of specialization among most financial intermediaries. And when we speak of financial intermediaries processing risk, we mean that they are typically diversifying some, absorbing some, and shifting some to others.

The concept of diversification is used in this book in a variety of different contexts. We use it quite extensively in Chapter 3, for example, to explain economies of scale in the production of financial intermediation services.

Riskless Arbitrage

Arbitrage is the simultaneous purchase and sale of identical goods or securities that are trading at disparate prices. This opportunity for riskless profit is transitory because the exploitation of such opportunities eliminates the initial price disparities.

The term arbitrage is often loosely applied to situations in which objects of trade are similar, but not identical, and where the risk is thought to be small but not totally absent. Since such situations are often referred to as arbitrage, the redundant riskless arbitrage has emerged to describe arbitrage rather than limited risk speculation (a situation in which a profit can be had for a small risk). Thus, succinctly defined, riskless arbitrage is profit without risk and without investment. We shall later discuss risk-controlled arbitrage as an illustration of limited risk speculation. Consider the following illustration of riskless arbitrage.

Example 1.2

Suppose that there are two possible states of the economy next period: high (H) and low (L). Available in the capital market are two risky securities, R1 and R2, and a riskless bond, B. The state-contingent payoffs and current market prices of these instruments are presented in Table 1.1 below. Examine whether there are riskless arbitrage opportunites.

TABLE 1.1 State-Contingent Payoffs and Prices of Securities

Solution

Since you can combine R1 and R2 to get a payoff that is equivalent to that from B, you can see now that there is an opportunity for riskless arbitrage. If you buy one unit each of R1 and R2 for a total outlay of $80, you are assured of $100 next period, regardless of whether state H or L is realized. So you can sell two units of B for $86 earning a riskless profit of $6. You are obliged to pay the buyers of these two units of B a total of $100 next period, but this you can do from the cash inflows produced by the R1 and R2 that you possess. Since you can sell these two units of B before you even buy R1 and R2, your profit requires no investment on your part and no risk. You could of course sell an arbitrarily large number of units of B and buy the appropriate units of R1 and R2, giving yourself a veritable money machine. But as your purchases and sales increase in volume, it is reasonable to expect the prices of the securities to converge, thereby eliminating the opportunity for riskless arbitrage again. An important implication is that the prices of related securities cannot be determined independently of each other. This observation has provided a powerful way to price derivative securities such as options.

The notion that any capital market equilibrium should preclude riskless arbitrage has proved to be a powerful concept in many applications in finance, including financial intermediation. We will see this idea applied in other contexts, including the valuation of contingent claims such as loan commitments.

Options

An option is a contract that gives the owner the right to either buy or sell an asset at a predetermined price at some future time or over some fixed time interval. Consider an asset whose value at time t = 1 will be X. Viewed at t = 0 (the present), X is a random variable. A call option entitles its owner to buy this asset at a fixed price, Pc, at or before t = 1. If he does not wish to buy the asset, he can allow the option to expire unexercised. Thus, the value of the call option at t = 1 is

     [1.12]

The theory of option pricing explains C(t = 0), the value of the call option at t = 0. The basic idea is to construct a portfolio consisting of the underlying stock and a riskless bond in such a manner that it yields the same payoff as the option. Since there can be no riskless arbitrage in equilibrium, the prices of this portfolio should equal the price of the option. We can then price the option by using the observed prices of the stock and the bond. We will have more to say about option pricing in later chapters.

Symmetrically, a put option entitles the option owner to sell an asset at a fixed price, Pp, at or before t = 1. Thus, at t = 1 the value of the put option is

     [1.13]

In addition to being a put or call, an option can be either European or American. A European option can be exercised only at some predetermined maturity date, for example, at t = 1 in the above discussion. An American option can be exercised any time prior to maturity. Thus, an American option never can be worth less than its European counterpart.

An important property of options that we will use frequently is that the more volatile the value of the underlying security on which the option is written, the more valuable the option. The following example illustrates this property.

Example 1.3

Consider a European call option with an exercise price Pc = $100. At t = 1, X will be $110 with probability 0.5 and $90 with probability 0.5. For simplicity, suppose everybody is risk neutral and the discount rate is zero (so that future payoffs are valued the same as current payoffs). Then from (1.12) we have

Thus, C(t = 0) = 0.5(10) = $5. Now suppose we increase the variance of X, keeping its mean unchanged. Let X be $150 with probability 0.5 and $50 with probability 0.5. From (1.13) we have

Thus, C(t = 0) = 0.5(50) = $25. The call option is now five times more valuable! You should work through a similar example for put options to convince yourself that puts have the same property.

Option pricing theory is used in our later discussions of the valuation of off-balance sheet claims like loan commitments, and in our analysis of deposit insurance.

Market Efficiency

An efficient capital market is one in which every security’s price equals its true economic value. But what is true? In economics, it means a price that incorporates all the information available to investors at the time. In an efficient market, an appropriately defined set of information is fully and immediately impounded in the prices of all securities. The basic idea is that competition among investors and the resulting informational exchanges will lead to market efficiency. This implies that price changes in an efficient market must be random. If prices always reflect all relevant information, then they will change only when new information arrives. However, by definition, new information cannot be known in advance. Therefore, price changes cannot be predictable.

We speak of three forms of market efficiency, distinguished by the amount of information impounded in the price. A market is said to be weak-form efficient if prices impound all historical information. In a weak-form efficient market, if Pt is the price at time t, then the expected value (at time t) of the price at time t + 1 conditional on the price at time t, written as E(Pt+1|Pt), is the same as E(Pt+1|Pt, …, P0), the expected value of Pt+1 conditional on the entire history of stock prices up until time t (that is, Pt,…, P0). That is,

This means that you can do no better forecasting tomorrow’s price Pt+1 using the entire history of prices than you could using just today’s price Pt. The reason is that weak-form efficiency implies that Pt itself should contain all the historical information contained in the sequence {Pt–1 |Pt–2, …, P0}.

Semistrong form market efficiency requires that all publicly available information be contained in the current price. Since all historical information is in the public domain, a semistrong form efficient market is always weak-form efficient. However, there may be contemporaneous information in the public domain that became available after the most recent price was determined. Thus, semistrong form efficiency is a more demanding form of efficiency than weak-form efficiency.

A market is strong-form efficient if prices impound all information, including that possessed by insiders. Few economists believe that markets are strong-form efficient. Although there is a mountain of empirical evidence accumulated over nearly 2 decades suggesting that markets are semistrong form efficient, recent theoretical and empirical research has shown that the market may not even be weak-form efficient.³

If the capital market were strong-form efficient, there would be no role for financial intermediaries as information processors (unless intermediaries were crucial in making the market efficient). However, when strong-form efficiency fails to obtain, we can have different individuals primarily possessing different sorts of information. In Chapter 3 we will show that in such markets, financial intermediaries have a role to play. At many junctures in this book, we will discuss how the efficiency (or lack thereof) of markets affects the profits to be earned from financial intermediation. An example of this is financial innovation.

Market Completeness

The economic world we inhabit is complex and pervasively uncertain. It is often useful to think of this uncertainty in terms of the possible states of nature that can occur in the future. Each such state, call it θ, can be viewed as a possible economic outcome. For example, θ may correspond to different levels of gross domestic product. Although we do not know what θ will be tomorrow, we can assign a probability distribution over possible values of θ. For the theory, it does not matter how many values θ can take. For simplicity, suppose θ can take integer values from 1 to some arbitrary number N.

In evaluating problems of economic efficiency, an important consideration is the number of different financial securities available relative to the number of states of nature. Two financial securities are considered different if they do not have identical payoffs in every state. To see the implications of this, consider the following simple example.

Example 1.4

Suppose there are three states of nature and only two securities may be thought of as shares of stock issued by two different companies. The payoffs offered by these securities in the different states of nature are shown in Table 1.2.

TABLE 1.2 Example With Three States of Nature and Two Securities

Consider now an individual who owns 10 percent of security 1 and 20 percent of security 2. If θ = 1 occurs, his wealth will be 0.10(10) + 0.20(15) = 4. If θ = 2 occurs, his wealth will be 0.10(20) + 0.20(0) =2. If θ = 3 occurs, his wealth will be 0.10(15) + 0.20(25) = 6.5. Thus, the value of the individual’s portfolio can be described by the vector (4, 2, 6.5), where the first element corresponds to his wealth in state 1 and so on.

While the individual can achieve the vector (4, 2, 6.5), it is easy to see that one cannot achieve the vector (2, 6.5, 9.5). It is impossible for one to find ownership fractions in the two securities that will allow one to achieve this wealth vector. The reason is that there are fewer (independent) securities than there are states of nature. If we had a third security, we could have ensured that our individual could achieve any desired income vector. Of course, in reality individuals are also constrained by their budgets. The point is simply that when there are fewer securities than there are states of nature, it is generally impossible for the individual to attain any desired future wealth rearrangement. This is ultimately a limitation on the individual’s ability to insure against contingencies.

The securities depicted in our simple example are not really stocks or bonds or any of the other financial securities commonly found in the capital market. Rather, these securities are claims to income in different states of the world. We can nevertheless visualize a market where such claims are traded. We would then have a number of securities, one for each state of nature, promising to pay 1 dollar if that particular state occurred and nothing otherwise. Such securities are called primitive state-contingent claims or Arrow-Debreu securities after the economists Kenneth Arrow and Gerard Debreu, who first studied this issue and later went on to win Nobel Prizes in Economics. Such a market would represent an ideal way of organizing a securities exchange, since it would give individuals complete freedom (subject only to their own purchasing power limitations) in designing portfolios that deliver the desired distribution of income in different states of the world. That is, an individual can design any homemade security in such a market.

If there are as many Arrow-Debreu securities as there are states of nature, the market is referred to as complete. In a complete market, an individual can achieve any desired distribution of income, subject to the individual’s budget constraint. On the other hand, if there are fewer Arrow-Debreu securities than there are states of nature, we have an incomplete market, which places a limitation on the ability of transactors to manage uncertainty. The conceptual beauty of a complete market is that we can examine the market prices of securities that are currently trading and determine the market price of any new security we may wish to introduce. We can do this without knowing the preferences of individual investors in the economy. The key is that we can use the prices of existing securities to compute the prices of the (fictitious) Arrow-Debreu securities, and then use this information to price any new security we want to introduce. Suppose that in Example 1.2, we are given the prices of securities R1 and R2; recall that the price of each security is $40. Moreover, R1 pays off $100 in state H and 0 in state L, whereas R2 pays off 0 in state H and $100 in state L. Let PiH and PiL be the prices of the Arrow-Debreu securities in states H and L, respectively. Then, the market price of security R1 should be 100 times the price of the state H Arrow-Debreu claim, that is, 40 = 100 PiH, PiH = 0.4. Similarly, the market price of security R2 should be 100 times the price of the state L Arrow-Debreu claim, that is, PiL = 0.4. We are now ready to price any security in this two-state economy. For example, the riskless bond in Example 1.2, which pays $50 in each state, should be priced at 50 PiH + 50PiL = $40. A security that pays $1,000 in state H and $56 in state L should sell at 1000PiH + 56PiL = $422.40, and so on.

The concept of market incompleteness is used in Chapter 12 in connection with our discussion of financial innovation. Other applications can be found in chapters on off-balance sheet activities, securitization, and deposit insurance.

Asymmetric Information and Signaling

Economic transactions often involve people with different information. For example, the borrower usually knows more about its own investment opportunities than the lender does. Corporate insiders normally know more about the values of assets owned by their firms than shareholders. A doctor can be expected to be better informed about his or her own medical expertise than a patient.

The better informed economic agents have a natural incentive to exploit their informational advantage. Insider trading scandals on Wall Street illustrate how those with access to privileged information can profit, despite laws aimed at preventing such activity. Of course, those who are uninformed should anticipate their informational handicap and behave accordingly. It is this interaction between the inclination of the informed to strategically manipulate and the anticipation of such manipulation by the uninformed that results in distortions away from the first best (the economic outcome in a setting in which all are equally well-informed).

Problems of asymmetric information were brought to the forefront when George Akerlof (1970), who later went on to win the Nobel Prize in Economics for his contribution, sought to explain why used cars sell at such large discounts relative to the prices of new cars. The following example takes some shortcuts, but conveys the intuition of Akerlof’s analysis.

Example 1.5

Consider a used car market in which differences in the care with which owners use their cars lead to quality differences among cars that started out identical. It is natural to suppose that the owner of the used car knows more about its quality than potential buyers. As an example, assume that there are three possible quality levels that the used car in question can have, q1 > q2 > q3 = 0. If the quality level is q3, the car is a lemon. Such a car would be priced as being worthless if buyers could correctly assess its quality. If the quality is q2, the car has a value of $5, and if the quality is q1, the car is worth $10. Assume that all agents are risk neutral and a buyer does not want to pay more for a car than its expected worth. In like vein, the car owner does not wish to sell at less than what the car is worth. Suppose that each car owner knows his car’s quality, but buyers only know that cars for sale can be of quality q1, q2, or q3. Faced with a given car, they cannot identify its precise quality. However, they believe that there is a probability 0.4 that the quality is q1, a probability 0.2 that it is q2, and a probability 0.4 that it is q3. What will happen in such a market?

Solution

If all cars are offered for sale, risk neutral buyers will compute the expected value of a (randomly chosen) car as (0.4) × $10 + (0.2) × $5 + (0.4) × 0 = $5. Hence, if the market is competitive, we would expect $5 to be the market clearing price. However, at this price those who own cars with quality q1 will refuse to sell. Thus, only cars of qualities q2 and q3 will be offered at $5. However, buyers will anticipate this and revise their beliefs about the quality dispersion of cars in the market. They will now assume that if the selling price is $5, the probability is 0.2/(0.2 + 0.4) = 1/3 that the quality is q, and it is 2/3 that it is q3. Thus, the expected value of a car drops to (1/3)(5) + (2/3)(0) = $1.67. No cars will, therefore, be bought at $5 (it cannot be a market clearing price). Now if $1.67 is the price, those with cars of quality q2 will drop out and the only cars offered for sale will be lemons. This process is called adverse selection and it results in the market clearing price being driven to zero. In other words, the demand for cars at any positive price is zero, and the market breaks down, as depicted in Figure 1.3. You should note a key assumption made in this example. All market participants have rational expectations. That is, uninformed buyers rationally anticipate what informed sellers will do at any given price and informed sellers rationally anticipate the demand buyers will have at that price. Hence, we don’t need to go through a sequential process of price convergence to zero. No cars will be bought or sold.

FIGURE 1.3 A Pictorial Depiction of the Adverse Selection Process

The insight that asymmetric information can cause market failure was novel and striking. Its profound implications were quickly recognized to extend well beyond the used car market. Informational asymmetries were seen as being capable of causing markets to break down and thus possibly justify regulatory intervention by the government. Indeed, in the chapters that follow, we will examine banking regulation from this informational perspective.

However, calls for regulation based on Akerlof’s analysis were too hasty. Market participants have the capability and incentives to deploy mechanisms to prevent market failure, and in any case market failure is the most extreme form of distortion created by asymmetric information. To see this in the context of our used car example, consider the following extension of that example.

Example 1.6

Suppose that cars of different qualities have different probabilities of engine failure within a given time period, and that these differences are reflected in their values of 0, $5 and $10. Suppose the failure probability is 0.1 for the q1 quality car, 0.5 for the q2 quality car, and 1 for the q3 quality car. Do warranties have a role to play in this market?

Solution

To prevent market failure, the sellers of better cars must somehow distinguish themselves from the sellers of lower quality cars. One way to do this would be with warranties or guarantees. The seller of the q1 quality car can announce that he will reimburse the buyer $W1 if his car fails, and the seller of q2 quality car can announce that he will pay the buyer $W1 if his car fails. If buyers believe that only the owners of q1 quality cars will promise a $W2 payment upon failure and that only the owners of q2 quality cars will promise a $W2 payment upon failure, then they will make the appropriate inference and should be willing to pay prices that accurately reflect the qualities of the cars offered for sale. In order for such an indirect transfer of information to be effective, no seller should wish to mimic the strategy of a seller of a different quality car. Otherwise, buyers will eventually learn of the potential mimicry and the credibility of the signal will be destroyed.

Since the failure probability for a q1 quality car is 0.1, the buyer should be willing to pay $10 (the intrinsic worth of a q1 quality car) plus 0.1 times W1, the latter being the amount he expects to collect from the seller. Thus, the equilibrium price (P1) of a q1 quality car should be $10 + 0.1W1. Similarly, if the owner of a q2 quality car follows his equilibrium strategy, the equilibrium price (P2) of a q2 quality car should be $5 + 0.5W2. To ensure that the q2 quality car owner will not misrepresent himself as a q1 quality car owner, W1 should be set to satisfy

     [1.15]

The left-hand side (LHS) of (1.15) is the expected payoff to a q2 quality car owner misrepresenting himself as a q1 quality car owner; he receives a price P1 and has an expected outflow of 0.5 W1 to pay the liability under the warranty. The right-hand side (RHS) of (1.15) is what the q2 quality car owner gets if he follows his nonmimic strategy; he receives a price of P2 and has an expected cash outflow of 0.5W2. When someone is indifferent between telling the truth and lying, it is conventionally assumed that truth-telling will be chosen. Thus, (1.15), which is referred to as an incentive compatibility (IC) condition, can be treated as an equality and we can solve it to obtain W1 = 12.5. Incentive compatibility here means that the seller’s incentives to maximize personal profit should be compatible with truthful representation of the car’s quality.

The IC condition that ensures that the seller of lemons does not mimic the seller of q2 quality cars can be similarly expressed as follows

     [1.16]

Solving (1.16) as an equality yields W2 = 10. It is straightforward to verify that the seller of q2 quality cars will not mimic the seller of lemons under the described conditions, that is, q2 quality cars will be offered for sale.

You can easily verify that this scheme guarantees that the seller of lemons will not mimic the seller of q1 quality cars and that the seller of q1 quality cars will not mimic either the seller of q2 quality cars or the seller of lemons.

To summarize, we have produced a simple scheme of warranties that prevents market failure. The seller of q1 quality cars promises to pay the buyer $12.5 if his car fails; this enables him to sell his car for 10 + 0.1(12.5) = $11.25. The seller of q2 quality cars promises to pay the buyer $10 if his car fails; this enables him to sell his car for 5 + 0.5(10) = $10. The lemons are withdrawn from the market.

The warranty offered here can be viewed as a signal of quality. A (perfectly revealing) signal is one that enables the uninformed to infer which the informed agent knew privately a priori. For a signal to be useful it must be informative, and this requires that the signaling mechanism be incentive compatible. In turn, incentive compatibility requires that the cost of signaling must be negatively correlated with quality,⁴ Michael Spence too was awarded the Nobel Prize in Economics for his contribution to the economics of asymmetric information. That is, it must be less costly at the margin for a higher quality seller to emit a given signal. The higher cost of signaling serves to deter the lower quality sellers from mimicking their higher quality counterparts. In our context, you can see that a warranty of $12.50 imposes an expected liability of $1.25 on the q1 quality seller, $6.25 on the q2 quality seller and $12.50 on the seller of