Вы находитесь на странице: 1из 265

Principles of Microeconomics

Peter Thompson

Florida International University

0. 1. 2. 3. 4. 5. 6. 7. 8. 9. Introduction Auctions Three Principles Efficiency and Equity Supply and Demand Messing with Markets Messed-up Markets I: Externalities Messed-up Markets II: Public Goods, Common Property Resources, and Club Goods Messed-up Markets III: Monopoly Messed-up Markets IV: Oligopoly and Games of Strategy 2 11 32 78 98 130 162 191 208 237

Answers and Hints to Selected Problems


Chapter 0 Introduction
A quick search on the internet yields a good number of definitions of economics. Here are two examples, culled pretty much at random from popular information portals:
the branch of social science that deals with the production and distribution and consumption of goods and services and their management the science that deals with the production, distribution, and consumption of the worlds resources and the management of state income and expenditures in terms of money.

The Miriam-Webster dictionary offers a similar definition:

A social science concerned chiefly with description and analysis of the production, distribution, and consumption of goods and services.

These definitions are in line with the public perception of what economics is and what economists care about. I have lost count of the number times I have been asked by strangers what I think the stock market, the Federal Reserve, or the economy is going to do next week. My answer is always the same: I dont know what [it] did last week, I dont know what it will do next week, and I dont care. And yes, I am an economist. The problem with these popular definitions is that they define economics as a subject matter. Few economists would agree that this is the right way to think about what economics is. If it were, then it would be unlikely that the very best economics journals would willingly publish papers with such titles as The Causes and Consequences of Distinctively Black Names, Rotten Apples:

An Investigation of the Prevalence and Predictors of Teacher Cheating," or An Economic Theory of the Fifth Amendment." However, each of these papers was published in a top journal during the last two years.1 As one might expect, given that more than $100 is usually charged for them, Principles textbooks generally offer much better definitions. Bade and Parkin (2004) define it as the social science that studies the choices we make as we cope with scarcity and the incentives that influence and reconcile our choices. Krugman and Wells (2005) define it the branch of economics that studies how people make decisions and how these decisions interact. I like these definitions precisely because they are broad and vague. Their vagueness convinces the reader that economics is not equivalent to business studies, even though it so happens that many economists are interested in business. Economics is about decisions, not just business decisions. Their broadness suggests that there do not exist a set of topics that are economics and another set of topics that are not economics. If a topic involves decisionmaking, then some economist somewhere is probably working on it. But many psychologists, for example, also study how people make decisions. So what distinguishes the way economists study it from psychologists? The answer is that economics must be defined by how the people that do economics approach whatever problem happens to interest them. That is, economics is a methodology, not a subject. In fact, Frank and Bernanke (2004) limit themselves to the following mystically abstract definition: Fundamentally, it is a way of thinking about the world.

For my own part, I have in the last few years written papers on how husbands make their wives miserable, on the economics of bribing terminally ill patients to go to Hawaii and die, on how people choose whether or not to undergo diagnostic tests for illness, and why we shouldnt ask relatives permission to harvest organs from the dead.

1. How economists work

To begin to understand what economics is and how it differs from, say, psychology, it is useful to see how economists typically do their work. Most economic explanations of human behavior (i.e. human decision-making) consist of three components (see Figure 1). The first component is a statement of what people care about. For example, if the decision-maker is the CEO of a firm, she may care about making profits as large as possible. If the decision-maker is a commuter deciding whether or not to take an umbrella to work, we may assume that he cares about not getting wet and about not carrying an umbrella if it is a dry day. A good economist is one who has a knack for identifying correctly what is that people care about in any particular setting. A good economist also has a knack for understanding what can be left out. It may be true, for example, that that the commuter also cares about his train being on time. But as that has no bearing on whether or not he carries an umbrella to work, a good economist chooses to leave it out of the story. The CEO may also care about social status. If that influences the decision being studied, then a good economist knows it must be included. If caring about status does not affect the decision being studied, a good economist knows it can be left out. Deciding what to leave out of a model is a problem also faced by researchers in other disciplines. Consider physics. If you throw a bowling ball off the top floor of the Empire State Building, we can make two predictions. The first is that someone on the sidewalk is very likely to die. The second is that the ball will accelerate towards the unsuspecting pedestrian at a rate of 9.8 meters per second per second. In making this second prediction, we have ignored the effects of air resistance. Why? Because we know air does not really have much effect on a speeding bowling ball. But now drop a feather off the same floor. The prediction made by ignoring air resistance will in this case be wildly off:

A statement of what people care about

A statement of what contrains the choices people are able to make.

A statement of what people know at the time they have to make a decision.

FIGURE 1. Main components of an economic explanation of decision making.

the feather will slowly waft down in the wind. To predict the behavior of the feather, it is no longer reasonable to ignore air resistance. A good physicist, like a good economist, knows when certain factors can be ignored, and when they can not. The second component is a statement of what constrains the choices that can be made. The CEO cannot raise prices and sell more of a product. It may be the case that the commuter has no opportunity to buy an umbrella during the day, and so he must make his decision first thing in the morning, before leaving the house. These constraints are what Bade and Parkin refer to as scarcity. The third component is a statement of what people know at the time they make these decisions, and what they do not know. The CEO, in deciding to launch a new product may know that it is better than any of his competitors products, or she may not. The commuter doesnt know whether it will rain or not, but he may know the days probability because he listened to the forecast. Once the economist has precisely defined what people care about, what constraints they face, and what information they have, the next step is to make 5

predictions of their behavior using basic logic and, very often, mathematics. Simply put, what would a person do to make him or herself as well off as possible, given that s/he cares about x, is constrained by y, and knows z? Once we have worked out a set of predictions about peoples behavior, we can test our model by comparing these predictions with observed behavior. If the two do not coincide very well, then we should go back to the beginning and ask how we should alter the model. Did we not understand properly what people care about? Did we not correctly define the constraints they face? Did we make a mistake about they knew? Progress in economics is made in this way: construct a model that leads to predictions; then compare the predictions with observed behavior; and if the two are not consistent, go back to the beginning; repeat until observation and prediction are in line with each other. Once we have a model that we are satisfied with, there are many things we can do with it. Most important, we can use it to design policy. If we want to change peoples behavior, we can use the model to tell us how we need to change the constraints they face, or the information they have. If we want to change the constraints, we can use the model to predict how this would affect peoples well-being, and their behavior. Notice in this policy design that we do not ask how a policymaker can change what people care about. One of the hallmarks of economics is that we tend to take peoples preferences as given. This contrasts with some other fields, notably psychology, where the concern is often about how we can change what people care about. For example, an economist concerned with reducing pollution might think as follows: given that consumers (i) like clean air and big cars, and (ii) face certain prices for big and small cars, how should I change the prices of big and small cars to get more consumers to buy a small car? A psychologist might instead say the following: given that consumers (i) like clean air and big cars, and (ii) face certain prices for big and small cars, how can I make them care more about the environment and less about having a big car?

The big umbrella question. To pursue these ideas a little further, let us consider more formally the question of when a commuter is likely to take an umbrella to work. The first step is defining what the typical commuter cares about. Let us assume that the commuter doesnt like getting wet, but he also doesnt like carrying an umbrella. To capture these assumptions in a formal way, we can assume the getting wet imposes a cost, as does carrying an umbrella. Let w be the cost of getting wet, and let c be the cost of lugging an umbrella around all day. The second step is to define the constraints. We will assume that the umbrella must be taken from home, because the commuter never passes an umbrella store. We will also assume that if an umbrella is taken from home, it is too expensive to throw away during the day. That is, if the commuter takes an umbrella from home, he must pay the cost c regardless of how the weather turns out. The third step concerns information. Because the umbrella decision must be made in the morning at home, the commuter cannot know for sure that it will rain or that it wont. The most he can know is the forecast, which we can specify as the probability of rain. Let p denote this probability. Now consider choices. The cost of taking an umbrella is c. What is the cost of not taking the umbrella? If it does not rain, the cost is zero. But if it rains, the cost is w. As it rains a fraction p of the time when the forecast is p, then the average cost of not taking the umbrella is pw. A reasonable assumption about the commuters behavior is that he will listen to the forecast and then, knowing p for that day, will make the decision that minimizes his average cost. That is, he will take the umbrella if pw>c; he will not take the umbrella if pw<c; and he will be indifferent between the two actions if pw=c. Put another way, if the forecast, p, is greater than c/w, the commuter will take the umbrella, and if the forecast is less than c/w, he wont.

An important responsibility of an economist to communicate his or her ideas to others. Clarity is often increased by adding graphs to the words and math . Figure 2, for example, explains the decision that commuters make in graphical terms. The horizontal line, marked c, measures the cost of taking an umbrella. The line pw measures the average cost of not taking an umbrella when the forecast is p. For any forecast less than p*, the cost of carrying the umbrella exceeds the average cost of getting wet, so no umbrella is taken. If the forecast exceeds p*, the reverse is true.
Average cost of not taking umbrella


Cost of taking umbrella


Forecast probablity of rain

FIGURE 2. Solution to the umbrella problem.

You should be able to draw on this graph the effect of a rise or fall in w on an individuals decision, as well as a rise or fall in c. How do we test this theory? The first test is simple intuition. Do the predictions make sense? They do in this case. Formal testing involves more work, because we dont yet know what w and c are. But we can still do 8

something. Here is one way: Different people are likely to have different values for c/w. Logically, as the forecast p rises, the model then predicts that more and more people will choose to carry an umbrella because p exceeds their own personal ratio, c/w. This we can test. Observe the forecast each day and count the fraction of people walking around with umbrellas. If one rises with the other, there is our first indication that the model might be OK. Can we be more demanding in our test? Sure. On rainy days, we could sell umbrellas for $5. The fraction of people that will buy an umbrella for $5 must be the same as the fraction of people for whom the cost of getting wet exceeds $5. Another rainy day, sell umbrellas for $10. This way, we begin to get data information of the range of value of w. Similarly, we can buy umbrellas on sunny days to get information about c. One we have some idea of the values that w and c take, we can then test the theory quantitatively: if 40% of the population has a value of c/w greater than 0.5, then our model predicts that 40% of the population will carry umbrellas when the forecast is p=0.5.

2. Concluding comments
The umbrella example in the previous section is remarkably simple. But it does give a reasonable picture of how economist approach their work. In this course we are concerned with developing principles of microeconomics. Thus, we will concentrate more on developing ideas and models that economists already believe have stood the test of time, in the sense that the lead to predictions about behavior that are consistent with observed behavior. We will spend little time thinking about formally testing these models. Of course, we will ask whether the models make intuitive sense: unless one understands the intuition behind a prediction, one cannot really say one understand the behavior.

Bade, Robin, and Michael Parkin (2004): Foundations of Microeconomics, Second Edition. Boston: Pearson Addison Wesley.

Frank, Robert H., and Ben S. Bernanke (2004): Principles of Microeconomics,, Second Edition. New York: McGraw-Hill. Krugman, Paul, and Robin Wells (2005): Microeconomics. New York: Worth Publishers.


Chapter 1 Auctions and Economics

In class we held some auctions with different rules and with different goods for sale. We found that if we change the ground rules for the auction, we could get very different outcomes. We also found that differences in the nature of the good being auctioned affected the outcomes. In different settings people bid differently and they felt differently about the outcome of the auction. We found that some ground rules seem to be a lot better than others, in the sense that they produce what we feel are more desirable outcomes. And we found that in some settings bidders had a much harder time than others in working out what they should do. The questions that arose from these auctions touch on the most fundamental concerns of economists: When people make choices in different settings, can we predict what choices that they make, and what choices they should make? Can we devise ground rules that help people make choices leading to the best possible outcomes for themselves, or the best possible outcomes for society? If we change the ground rules, how will this affect the choices they do, or should, make? Having spent some time participating in auctions, it seems appropriate now to spend some time thinking about them. There are various auction designs, each of which sets its own ground rules, and we begin by reviewing some of them in Section 1. In Section 2, we then turn to some powerful results about one particular type of auction, the so-called Vickrey auction. We will see that, if 11

certain conditions hold, Vickrey auctions are best in a very precise sense:
they make the sum of everyones happiness as large as possible. We will also state what we mean by happiness in a very precise way. But we will also see that if these conditions do not hold, the Vickrey auction may not achieve this laudable goal. In such cases, we may want to intervene and change the rules. In Section 3, we will see that what we have learned about auctions sets us up very nicely for thinking about other types of markets. In fact, as we will see later in this course, many markets are just versions of Vickrey auctions.

1. Types of auctions
There are a number of different types of auctions, and it is useful to list and define them here. The first important distinction concerns how individuals value the item up for sale. Private-value auctions: Individuals know what the value of the item is to them, but the item has a different value to each potential buyer. Common-value auctions: Individuals would all have the same valuation of the item were they to obtain it, but they do not know in advance exactly what the value of the item is. In both types of auctions there is uncertainty. In private-value auctions bidders dont know how other people value the item up for sale, and hence they dont know what their opponents will bid. Auctions involving antiques, or vintage wine, where the bidders are not intending to resell the item but have it for their own enjoyment, are examples of private-value auctions. In common-value auctions, every bidder must hazard a guess at what the true value of the item is. A bidder then faces two types of uncertainty. First, she does not know precisely how much she should be willing to pay for the item. Second, she does not know what her opponents guesses (and hence their bids) will be. A classic example of a common value auction concerns the selling of drilling rights in oil fields. The value of the oil field will be the same for every firm: the field contains a certain amount of oil with a certain sale price and a 12

cost of extracting it. But every firm may have a different opinion about just how much oil there is, what price it can be sold for in the future, and what the extraction costs will be. The second distinction concerns the conduct of the auction. First there are oral auctions, in which everyones bid is visible to everyone elses. Second, there are sealed-bid auctions, in which every bid is secret. Each of these types of auctions has two common variants. English oral auction: In an English auction, the auctioneer begins with a low asking price, and continues to raise it until no one offers to beat the last bid. The last bidder wins the item at the price she last offered. This is how most of Sothebys auctions are held, and they are the type of auctions you usually see in movies. English auctions have been around a long time. Sothebys was founded in 1744 and has been using this auction format almost the entire time. Cassady (1967) provides evidence that this was essentially the format of auctions held in Ancient Rome.2 Dutch oral auction: In a Dutch auction, the auctioneer begins with a high price, and then gradually drops it until someone offers to take the item at that price. The name comes from their use in flower markets in the Netherlands. In contrast to English auctions, where the auctioneer cajoles people into making higher bids, Dutch auctions can be very fast. The price declines are called out rapidly until someone calls out Mine! If youre not on the ball, you miss out. First-price sealed-bid auction: A single, secret bid is made by each participant. The item for sale goes to whoever submitted the highest bid.

Cassady reports that in A.D. 193, after the Preatorian Guard had killed the emperor of the Roman Empire, they put the entire Empire up for auction. Didius Julianus was the highest bidder (in Latin he was the emptor), and bought the Empire for 6,250 drachmas per Guard. Good for him. But he was executed two months later when Septimus Severus conquered Rome. So caveat emptor.


Second-price sealed-bid auction: A single, secret bid is made by each participant. The item for sale goes to whoever submitted the highest bid. However, the winner only pays the value of the second-highest bid. These bids are also known as Vickrey auctions, after the Nobel Laureate, William Vickrey, who analyzed them in 1961.3 Even with these varieties of auctions to choose from, one can still find curious variations. Crawford and Kuo (2003) describe a fish auction in Taipei:4
In June 1996, in the Hu-Lin (Tiger-Forest) Street evening market near Taipei City Hall in east downtown Taipei, we observed a fish auction whose form was remarkable. First the auctioneer held up a basket and announced a price, which was fixed throughout the auction at 100 New Taiwan dollars, about US$3.65 in 1996. He then put a series of approximately identical fish into the basket, one by one, until some buyer signaled he was willing to pay the fixed price for the fish that the basket then held, at which point the sale was concluded. This process was repeated for as long as we watched, in a series of auctions with the same kind of fish . . . and the same fixed money price for the basket.

Crawford and Kuo point out that this is essentially a variation on the Dutch auction. As fish are put in the basket, the equivalent price per fish falls, and the winner is as usual the first to call out Mine! Yet another variation, apparently used sporadically in the 19th century, was to begin with Dutch auction. Once someone called Mine!, an English auction would break out, with what would have been the winning bid from the Dutch auction being used as the starting bid for an English auction. Note what this

A word of warning. You will often read that Vickrey designed this type of auction. He did not. Lucking-Reiley (2000) has reported that this type of auction was used to sell collectible postage stamps as early as 1893.
3 4

I am grateful to Jesse Bull for bringing this example to my attention.


type of auction can accomplish. If a bidder was half asleep and missed out on the rapid-fire Dutch auction, he had a chance to correct his omission as the price went back up, this time more sedately.5 Although these variations are interesting, they are also rare. Hence, most analysis of what happens in auctions has focused on the four standard types. We know quite a bit about them. For example, we know that the Dutch auction and the first-price sealed-bid auction are equivalent to each other, in the sense that everyone behaves the same in both of them, and the seller can expect to get the same price in both. Intuitively, in both cases, bidders must choose how much they are going to bid without knowing how much their opponents are prepared to bid. We also know that the English auction and the second-price sealed-bid (Vickrey) auction are also (approximately) equivalent to each other. In both auctions the winner pays the same (Vickrey auction) or just a little more (English auction) than the second-highest bidder. If the increments to the bid amount in the English auction are small enough, then these prices will be essentially the same.

2. Valuable properties of Vickrey auctions

Vickrey (1961) analyzed optimal bidding strategies and outcomes for different auction designs when the value of an item is known to each individual but that value differs across individuals (i.e. a private-value auction). Among other things, Vickrey proved the following important result:

In the Vickrey auction, the rational strategy is to bid exactly what the item is worth to you. Put another way, bidding honestly is at least as good as any other strategy in any scenario, and it is strictly better in some scenarios.
Because this is an important result, we will prove it.

I have forgotten where I read about this type of auction. If anyone comes across the source, please let me know.


Proof that the optimal strategy is to bid your true valuation. Let v be the value of the object to Mariela, and let b be the highest bid any of her opponents makes. If Mariela wins the bid, she makes a net gain of vb (which is negative a net loss if she bids more than the item is worth to her). If she doesnt win, then she receives 0. We want to find out what is the optimal bid Mariela should make under three scenarios (i) b>v. Someone else bids more than the item is worth to Mariela. If Mariela bids anything less than b, she does not win the auction, and she gets a payoff of 0. If she bids greater than b she gets vb, but this negative. Thus she is better off not winning and earning 0. Any bid less than b, including bidding v causes Mariela to lose the bid, Thus bidding v is a least as good as any other strategy, and strictly better than bidding more than b. (ii) b <v. The highest bid is less than the item is worth to Mariela. If Mariela bids less than b she get nothing. If Mariela bids anything more than b, including bidding v, she gets vb, and this is positive. Thus bidding v is at least as good as bidding anything greater than b, and strictly better than bidding less than b. (iii) b =v. In this special case, bidding exactly v results in a tie. Presumably, the winner will be selected at random from those that bid b=v, so a bid of v gives Mariela a chance of winning. If she does the payoff is vb, which is zero. if she doesnt the payoff is still 0. This is exactly the same as the payoff for bidding less than v and not winning. It is also exactly the same as the payoff from bidding more than v, which would cause her to win but again have a zero net gain of vb, Thus bidding v is just as good as any other bid. These are all three possible scenarios, and bidding v turns out to be at least as good as any other bid, and under some scenarios it is better. As Mariela does not know which scenario she faces until the auction is over, it makes sense to bid v. End of proof.


Economists have a name for a strategy that is at least as good as any other strategy under all possible scenarios: a weakly dominant strategy. If a strategy is strictly better than any other under all possible scenarios, it is called a strictly dominant strategy. Thus, bidding ones true valuation for an item in a Vickrey auction is a weakly dominant strategy. Why Vickrey auctions are good for the seller. You might wonder, why would a seller want to let his item go for the second-highest bid, when someone offered to pay a higher amount? Wouldnt a first-price auction get a better price? The answer is a resounding no. As a potential buyer in a first-price bid it make sense to strategize. Dont bid what an item is worth to you. Instead, bid just a penny more than what the second highest bid will be (assuming this is no greater than what the item is worth to you). That way you win the auction at the lowest possible price. Thus, in first-price auctions, people tend to make lower bids. Of course, you have to guess what the second-highest bid will be, and many times you will guess the second highest bid wrong. Sometimes you will guess too low, and you dont win the auction. Other times you guess too high and you win, but you pay more than you need to. How exactly you trade off these two types of mistakes depends upon what you believe the other bids will be like, and this can make for very complicated reasoning. Your best strategy, though, is always to make lower bids than your valuation. Vickrey (1961) showed that, under a set of reasonable assumptions about the number of bidders and the information they have, the average selling price of an item will be the same in all types of auctions. For example, in a first-price auction you bid lower than what you value the item, but you pay what you bid. In a Vickrey auction you bid exactly what you value the item but you pay less than you bid. The differences across auctions between what you bid relative to your valuation and what you pay relative to your bid tend to cancel each other out.


A Vickrey auction with multiple items for sale. Suppose a seller has 10 identical items he wants to sell. How can we devise an auction so that all 10 items are sold? One way is to have a modified Dutch auction or first-price sealed-bid auction. In this case, we would require that bidders not only state the price they wish to pay but also the quantity they wish to buy at that price. The seller would sell to all the highest bidders the quantities they ask for at the prices they bid, until the 10 items have gone. But we know now that first-price auctions cause people to bid less than they value the items. An alternative is to undertake a modified second-price sealed bid auction: deliver all the supply to the highest bidders and make everyone pay the price that is paid by the bidder who just failed to get one (i.e. who had the highest valuation among those who didnt win). Thus, if there is one item for sale, the price paid is equal to the second-highest bid. If there are five items for sale, the price paid is equal to the sixth-highest bid. We know that this type of auction gets everyone to bid their true values, so it guarantees that everyone with the highest valuations for the items will get them. The auction system will enable the seller to get rid of his stock of goods. This seems a good way to run an auction. In fact, when Google went public in 2004, it used this type of auction. There was a lot of news about this being a novel way to have an initial public offering, but in fact other companies including Overstock.com and Red Envelope had previously used it.6 Let us think about this type of auction in graphical terms. Figure 1 illustrates the outcome of a Vickrey auction for a particular numerical case. The seller has five units to sell. Mara values the item at $14 and wants only one. John values two at $10 each, Abu wants one for a value of $8, Fred values one at $4,

A word of warning. News media and financial advisors use the term Dutch auction when they are talking about what we have called here a second-price sealed-bid auction. Economists use the term Dutch auction for a very different thing the descending-price oral auction.


Juan one at $3, and Eli one at $2. The seller ranks the bidders in order of the bids (which are the same as their true valuations). She then finds that she must sell her units at a price of $3. Juan and Eli dont get any, because their bids were too low.
16 14

quantity available
12 10 8 6 4 2 0 Mara John John Abu Fred Juan Eli

p =3

FIGURE 1. The Vickrey auction with multiple items.

We can also work out how well off each successful bidder is. Mara earns a valuation of $14 put pays only $3. Her net gain is $11. Johns net gain is $14 two units at $7 each. Abus is $5, and Freds is $1. Juan and Eli have net gains of zero: they get none of the good, but they also dont pay anything. The excess of a persons valuation of a good over and above what they pay for it is called her consumer surplus. It is a measure of how well off they are from participating in this auction. If we add together each individuals consumer surplus -- $11 (Mara) + $14 (John) + $5 (Abu) + $1 (Fred) = $31, we get a measure of the total consumer surplus from this auction. It is easy to see that, graphically, total consumer surplus is given by the shaded area of the graph that lies above the price line (see also Figure 2). We can also get a measure of the revenues to the seller. The seller sells 5 units at $3 each, so her proceeds are $15. It is easy to see that the revenues to the seller are given by the shaded area of in Figure 2 that lies below the price line. 19

16 14 12 10 8 6 4

consumer surplus

p =3

Seller revenues (to be split between costs and profit

2 0 Mara John John Abu Fred Juan Eli

FIGURE 2. Consumer surplus and seller revenues in a Vickrey

Revenues may not be all profit, of course. She may, for example, have to spend money to make the items she is selling. For the moment, let us imagine that the seller just happens to have a stock of five units of the item in question, and they are worth $2 each to her if she consumes them instead of selling them. This $2 per unit is her cost of selling the items. As she can sell each unit for $3, she is happy to sell, and she makes profit of $5 on the five units. Economists also refer to this profit as producer surplus (see Figure 3). One can readily see that there is no way to change the allocation of the goods among the bidders to make them better off as a whole. If we take a unit of the good away from any of those who have them, and give it to either Juan or Eli, who dont, consumer surplus must fall at the current price. This is simply because Juan and Eli value the good less than do the others. If we keep the allocations among bidders unchanged but lower the price to $2, we can raise consumer surplus (every buyer gains by $1, for a total increase in consumer surplus of $5). However, the seller is made worse off by exactly the same amount her profits fall by $5 to zero so there is no net gain overall.


16 14 12 10 8 6 4 2 0 Mara John John Abu Fred Juan Eli

consumer surplus producer surplus costs

FIGURE 3. Producer surplus, consumer surplus and costs. An example

Moreover, Juan might now want one because the price falls below his valuation and we may have to devise some new rule to explain why he cant have one. If we raise the price to $5, we make the seller better off with a $10 increase in producer surplus, but only by reducing consumer surplus by the same amount. Moreover, Fred will no longer want to buy his unit, so we would have to force consumption on him. In summary, a Vickrey auction has several attractive features. At the price that enables the seller to offload all her goods: Everyone that buys the good finds it worthwhile to do so because their valuation exceeds the price. Everyone that doesnt buy the good finds it optimal not to do so, because the price exceeds their valuation. At this price, the sum of the welfare of consumers (measured by consumer surplus) and the welfare of the seller (measured by producer


surplus) is maximized. There is no alternative allocation scheme that makes society better off. Common-value auctions and the winners curse. We have just seen that in a Vickrey auction, consumer surplus is maximized. But this was a special type of Vickrey auction a private-value auction -- in which all consumers knew exactly what the item was worth to them. Knowing exactly what the item is worth enables consumers to bid exactly their valuations. Bidding exactly their valuations enables the auction scheme to allocate the items to all the participants with the highest valuations. This in turn maximizes the sum of consumer surplus and seller profits, which we have used as a measure of the well-being of society. But let us now consider an alternative scenario. The true valuation of the good in question is the same for everyone: $9. However, the bidders dont know this. Some think the item is worth only $8, some think it worth $10, and yet others worth $12. Imagine there are 9 buyers, each interested in one unit. Three of them overvalue the good, three undervalue the good, and the remaining three get it just right. There are five units for sale. If everyone follows the strategy of bidding what they think the item is worth, the outcome in a Vickrey auction is easy to see. The three bidders who most overvalue the item will bid $12. The three bidders who guessed $10 will bid $10, and two of them will get one of the items (perhaps we need a lottery to decide which two get the good). Everyone pays the lowest bid that clears the market -- $10. But once the successful bidders get ownership of the good, they will be sorely disappointed. They all find that they paid $10, only to discover, after all, that the item is only worth $9. This disappointment is known as the winners curse. The winners in an auction are those who made the biggest overestimates of the true value of the good. The term winners curse was introduced to the profession by Capen, Clapp and Campbell (1971), who studied bidding behavior in the oil industry. In the


1950s, firms were competing in auctions to win oil-drilling rights in the Gulf of Mexico. But every time a winning firm started drilling, they ended up losing money. The winners all suffered the winners curse. Here is an explanation from Marasco (2004):
Suppose you have four oil companies all interested in the same patch of offshore property. Back in the 1950s the methods available for determining the amount of liquid gold under a few hundred feet of water were primitive. This led to a large spread in the estimations of the value of the oil, and hence the drilling rights. Suppose that after all the costs of drilling were accounted for there was $10M worth of oil in the ground. Company A might determine that there was only $5M worth of oil. On the other hand, Company B might nail it at $10M, Company C could value the oil at $12M and Company D could make a gross overestimation and believe that there was $20M worth of oil in the ground. Now we come to the bidding for the drilling rights. At $5M, Company A drops out of the competition. At $10M, we lose Company B. At $12M Company C bows out, and Company D is very happy because they have paid $12M for property they believe is worth $20M. But since there is only $10M worth of oil in the ground, Company D is in the red for $2M. They have suffered the winner's curse. In an auction, the prize goes to whoever has the most optimistic view of the value of the object being bid upon. In many cases, this also means that the winner is the person who has overestimated the most.

Notice that Marascos explanation is in terms of an English auction. The winners curse applies to any auction where people dont know their true valuations. Let us go back to the Vickrey auction with Mara and friends. Their true valuations are the same as before, but now they make mistakes and their guesses (and bids) are different, as described in the following table.


True valuation Mara John Abu Fred Juan Eli 14 2 @ 10 8 4 3 2

Guess ( = bid) 11 2 @ 12 3 4 6 3.50

Remember there are five units up for grabs. Who gets them? We see now that Mara, John, and Fred still win. But instead of Abu (who mistakenly bids only 3), we find that Juan (who mistakenly bids 6) gets one of the goods. The price paid by everyone is $3.50, because this is the sixth-highest bid. Because people were ill-informed, the auction has misallocated a good to Juan, when it should have gone to Abu. It should have gone to Abu because it is really worth $8 to him, while it is only worth $3 to Juan. Who suffers the winners curse? It is not Mara. She will surely be pleased. She was expecting a value of $11, but the good she got turned out to be better than she expected. John will surely be disappointed. He was expecting goods with a value of $12 each, but they turned out to be worth only $10 to him. But John does not suffer the winners curse. He got two units of the good, and he is glad he did: they are still worth more than he paid for them. Juan is the only one who suffers the winners curse: he paid more for the good than it turns out to be worth to him: his consumer surplus is negative, and he would have been better off not winning. So, when participants are imperfectly informed about the value of the goods, the Vickrey auction does not maximize consumer surplus. If what we care about is making society as well off as possible, then as economists we will be disappointed by this outcome. We found earlier that a market that is organized as a Vickrey auction will make society as well off as possible only under the assumption that participants in the market have perfect information about 24

their valuation of the good. When that assumption is violated, the very same market no longer guarantees society will be made as well off as possible. What can people do to avoid or reduce the problem of the winners curse? If you know that mistakes in valuation are possible, you should bid less than your guess. How much less is a complicated question, and we will study how people make decisions when they have imperfect information towards the end of the course. Economists refer to this inability of the market to work as well as we would like as a market failure. Under certain assumptions a Vickrey auction achieves an allocation of goods among buyers that maximizes consumer surplus. If that is what we most care about, we can just leave the market alone. But if any of those assumptions are violated, perhaps we can change the rules of the game to make people better off. We could, for example, allow everyone to try the item in question for a week. After trying the good, everyone learns their true valuation, and so they can make accurate bids. Of course, it is not costless to allow everyone to try the good. We have to get it to them, and while they are trying the good, we cant go ahead with the auction. But as long as the cost is not too high, it might be worth doing. In our example, the total consumer surplus with perfect information was $26. The consumer surplus with imperfect information was $21 (in both cases profits to the seller are unchanged, so we can ignore them). So as long as it costs less than $5 to let everyone try the good, it would be worth doing.7 Love disappoints. In David Henry Hwangs play, The Golden Child, a character comments that the fact that something is new means only that it

As a rule maker, our challenge of course is that we wouldnt really know in advance exactly what the two consumer surpluses would be. But were not idiots. Perhaps we can devise some sensible rules for markets in which we believe imperfect information is a problem. We may have to accept that these rules will not always make us better off but, if they are welldesigned, they will make us better off more often than not.


hasnt had time to disappoint us. This is the insight of the winners curse, and it explains why most of your relationships will turn out to be a less than you hoped for. Imagine you are single and looking for a partner. You meet people from time to time, and you get some rough indication of their quality as a potential partner. When one meets your no doubt very exacting standard, you start dating. But then, over time, you discover their true quality. The winners curse tells us that the person you are most likely to start dating is the person you whose quality you had initially most overestimated. It is likely therefore, that whoever you date is the person who will disappoint you most. But when you dump him or her, you shouldnt be angry. After all, you almost certainly proved to be a disappointment as well.

3. Concluding comments
Vickrey auctions are much like many markets we see all around us. Of course, we dont submit sealed bids at the grocery store and wait to see if we get what we bid on. Instead, the grocery store posts the price at which it will sell its stock, and we will only buy a good if the posted price is less than our valuation. The person with the lowest valuation that actually buys the good has a valuation just equal to the posted price. People who dont buy the good have a lower valuation than the posted price. If the grocery store can guess the right price, it will be able to sell all the units it has for sale. The details of the grocery stores problem is different, but the outcome is the same. It seems logical to presume, then, that many markets therefore have just the same desirable features as Vickrey auctions. Thus, if the market is working well, we will prefer to leave it alone and let it function. One the other hand, if a market failure (such as imperfect information) stops the Vickrey auction from having the desirable properties we have identified, it is likely that these same problems will stop other markets from having desirable features. In these cases we might prefer to intervene to change the outcomes of the market.


Finally, some markets may not take the form of a Vickrey auction at all. It is then reasonable to assume that they will often not have the desirable features we are looking for. In such cases we might again want to devise a policy to change the way the market work.

Concepts introduced in this chapter

private-value auction oral auction English auction Vickrey auction producer surplus winner's curse first-price auction common-value auction sealed-bid auction Dutch auction second-price auction consumer surplus profits costs market failure


1. Buying goods on the internet opens up consumers to the risk of the winners curse. Think of a couple of products for which you think the winners curse might be an especially important problem. Explain why this is the case. For such products, one would expect firms to encourage trying a product by offering a very easy way to return goods. For the products you listed above, check on the internet for return policies. Does it appear that sellers of such products make returns easier than sellers of other products? 2. There are three bidders for a painting at an English auction. The initial bid is $1 million, and the bid increment is $100,000. There are three bidders. Bidder A values the painting at $1.2 million, bidder B at $1.5 million, and bidder C at $1.7 million. The bidders do not know each others valuations. What two possible prices will the painting be sold for, and who will buy the painting at each of these prices? Explain your reasoning.


3. For the same three bidders as in the previous question, what price will the painting be sold for if the auction is a) A first-price sealed bid auction? b) A second-price sealed bid auction? If you can provide an exact number do so. If you cannot provider an exact number, explain why not. 4. Manuel, a Miami fisherman, is planning on selling a used fishing reel by auction. He knows there are two types of potential buyers. There is the committed fisherman to whom the reel is worth $200. Then there the casual fisherman, to whom the reel is worth only $100. On the day of the auction, only two people turn up. a) If both are committed fisherman, what price will Manuel sell the reel for if (i) he sells by an English auction with bid increments of $10 and a starting bid of $50? (ii) he sells by a Vickrey auction? b) If one is a committed fisherman and the other a casual fisherman, what price will Manuel sell the reel for if (i) he sells by an English auction with bid increments of $10 and a starting bid of $50? (ii) he sells by a Vickrey auction? c) If both are casual fisherman, what price will Manuel sell the reel for if (i) he sells by an English auction with bid increments of $10 and a starting bid of $50? (ii) he sells by a Vickrey auction? 5. Assume again that two people turn up to the auction. Manuel happens to know that it is equally likely that either one is committed or casual. Then, there are four possible outcomes: Both are committed, Both are casual, Person A is committed, and person B is casual,


Person B is committed and person A is casual. Each of these outcomes is equally likely. If Manuel were to run many auctions, one after the other, and two people always show up to each one, (i) How much on average would Manuel get for his reel with an English auction using the same rules as in question 1? (ii) How much on average would Manuel get for his reel with a Vickrey auction? 6. Now, assume that Manuel has 10 used reels to sell. 15 people turn up to his auction, and they have valuations of 250, 240, 200, 180, 160, 160, 150, 150, 130, 120, 100, 100, 100, 90, 60. Manuel decides to sell via a Vickrey auction. (i) What is the consumer surplus from the auction? (ii) If Manuel values each reel at $50, what is his producer surplus? (iii) Assume now that, in response to lobbying efforts by manufacturers of new fishing reels, the government imposes a law requiring all used reels to be sold at a price no less than legal minimum of $140. How many reels are sold? Compared with your answers in (i) and (ii), what is the dollar loss in economic welfare, as measured by the sum of consumer and producer surplus? (iv) Assume now that, in response to lobbying efforts by consumers of used fishing reels, the government imposes a law requiring all used reels to be sold at a price no greater than legal maximum of $90. How many reels are sold? Compared with your answers in (i) and (ii), what is the dollar loss in economic welfare, as measured by the sum of consumer and producer surplus? 7. Henry is planning to sell a desk at an auction. Ten people are interested in buying the desk. The valuations of the 10 potential buyers are shown in the following table,
Ann 45 Bill 53 Colin 92 Dave 61 Ellen 26 Frank 78 Gale 82 Hal 70 Irwin 65 Jim 56

(i) Henry decides to hold an English auction. The initial bid is $20, and the bid increment is $10. The bidders do not know each others


valuations. Who will win the auction, and how much will this person pay? (Hints: You need to consider the order of the bidding)
(ii) Is there a way for Henry to get a better price for the desk? If yes, what is it, and what is the price? 8. Consider a Vickrey auction in which there are 2 units available for sale. Four people participate in the auction, and each of them is interested in purchasing one unit. The true valuation of the item is $10 for each person. However, they do not know this and instead make mistakes, as indicated in the following table. For each of the auctions listed, who will suffer the winners curse? Explain your reasoning.

Individual Guesses at the True Value of the Good Person A Auction 1 Auction 2 Auction 3 20 9 11 Person B 19 10 10 Person C 18 13 9 Person D 17 15 8

Capen, E., R. Clapp, and W. Campbell (1971): "Competitive bidding in high risk situations." Journal of Petroleum Technology, 23:641-653 Cassady, R. (1967): Auctions and Auctioneering. Berkeley: University of California Press. Crawford, Vincent P., and Ping-Sing Kuo (2003): A dual Dutch auction in Taipei: the choice of numeraire and auction form in multi-object auctions with bundling. Journal of Economic Behavior & Organization, 51(4):427-442. Lucking-Reiley, D. (2000): Vickrey Auctions in Practice: From Nineteenth-Century Philately to Twenty-First-Century E-Commerce. Journal of Economic Perspectives, 14(3): 183-192.


Marasco, D. (2004): The winners curse. Online at http://www.thediamondangle. com/marasco/opan/wincurse.html. Accessed 26 October 2004. Vickrey, W. (1961): Counter-speculation, auctions, and competitive sealed tenders. Journal of Finance, 16(1): 8-37.


Chapter 2 Three Principles

Economists adhere to a core set of ideas that govern how we approach problem-solving. While the details of any particular problem may become quite complex, underlying much of the analysis are three simple principles that we will do well to think about now. Economics is about happiness Incentives matter The indifference principle We will come across many ideas that could also be called general principles during the course. But these three are critical and they suffice to get us started.

1. Economics is about happiness

Going to work five days a week and getting a paycheck is a terrible thing to have to do. I do it because I like my house, and to keep it I have to pay the rent. But Id be a lot happier if I could go to work just three days a week and still pay the rent. Id be even happier if I could be unemployed and still pay the rent. Many of my colleagues are different. Some of them work seven days a week. They also say they love their job so much they would do it for nothing. I doubt that, but I am willing to accept that teaching a Principles class is something that gives them pleasure. Me? Id rather have a margarita by the pool. Some of my other colleagues have gone the other way. They already only work three days a week. Perhaps they have a nicer pool. However we juggle work and leisure, we are all doing it in pursuit of happiness. I work five days a week because that is the amount of work that


makes me happiest. But you dont like to work!, you may say. True. But if I worked only three days a week I would have to move to a smaller house, and working three days a week while living in a cardboard box would make me less happy. I could work more, but working seven days a week so that I could have a bigger house would also make me less happy than I am now. The pursuit of happiness is such a fundamental idea that it is enshrined in the Declaration of Independence (its omission from the Constitution must just be an oversight). It is such a fundamental idea that all economists not only understand it, they have made it central to the profession. As economists we start from the assumption that whatever we observe people doing, they must be doing it because any other behavior will make them less happy. So why do I drink beer while my wife drinks wine? Easy! We are each drinking what makes us happier. Why does Warren Buffett work so hard even though he is so rich? Easy! Because thats what makes him happier. Why does Jimmy Buffet not work hard, even though he is rich? Easy! Because thats what makes him happier. Why did Martha Stewart allegedly lie to the Securities and Exchange Commission about allegedly illegal trades in ImClone stock and go to prison. Easy! Because thats what made her . . . . OK, not so easy. Because at the time she thought she would get a way with it and given the information she had she believed her alleged actions would make her happier. Why did I write alleged three times? Easy! Because thats what made me happy. Economics is concerned with just two things. First, we try to understand why peoples behavior makes them happier. There is a mirror image to this challenge. If we know what people care about (i.e. what makes them happy), we can predict how they will behave in different environments. Our second concern is to try and understand how we can enact policies that help them become even happier. To do this, we of course need to know what people care about, and this in turn will enable us to predict how they will respond to policies we may design and implement.


All this is harder than it sounds, because economists impose restrictions on themselves. We want explanations for phenomena that teach us something about the wider world. If every explanation we gave for every behavior we observe were nothing more than Oh, they do that because it makes them happier we would essentially learn nothing. These restrictions involve putting structure on the problems we analyze. A few years ago, my former colleague Nick Feltovich of the University of Houston, came up with a puzzle: why do so many of the smartest students, like Bill Gates, drop out of college? A simple explanation would be that smart students dislike college more than mediocre students, so they are more likely to drop out. But then, why do the least smart people not go to college? Would we also argue that the least smart dislike college more than the mediocre? That would certainly be one explanation, but it is far from satisfying. Could there be another explanation, that does not involve asserting that both the least smart and the most smart dislike college more than those with middling ability? Feltovich had one, which he published in 2002 in a paper with Rick Harbaugh and Ted To. Imagine there are three types of people: a lot of dumb people, some mediocre, and a small number of smart people.8 Employers like ability. They are willing to pay a higher wage to mediocre employees than they are to dumb employees, and they are willing to pay a higher wage to smart employees than they are to mediocre employees. But the challenge for the employers is that they have difficulty identifying a potential employees ability. They can conduct interviews, but this only provides so much information. Through the interview process they can tell the difference between the dumb and the smart, but they cannot tell the difference between the dumb and the mediocre, and they cannot tell the difference between the mediocre and smart.

This is a slight variation on the story told in the paper.


Now, smart people want to distinguish themselves from the mediocre so they can get paid more. So what do they do? They go to college. Unfortunately, the mediocre will want to distinguish themselves from the dumb and confuse themselves with the smart. Thus, they also go to college. The dumb would like to go to college to confuse themselves with the mediocre. Unfortunately (and here is the catch) if they did they would fail out. Knowing that, they dont go to college. The mediocre are thrilled. They have distinguished themselves from the dumb and confused themselves with the smart. But the smart people are upset because they are confused with the mediocre. They decide to drop out of college. The interview process enables employers to distinguish them from the dumb, and the fact that they dropped out of college enables employers to distinguish them from the mediocre. Finally, the mediocre dont drop out to copy the smart. Because the dumb are more numerous, it is more important for the mediocre to distinguish themselves from the dumb than it is to confuse themselves with the smart. This outcome is consistent with the idea that everyone is doing what makes them happiest given the constraints of everyone elses behavior. The mediocre would be less happy if they were to drop out of college and get confused with the dumb. The dumb would be less happy if they were to go to college because they would fail out anyway. And the smart would be less happy if they were to go to college because they would be confused with the mediocre. Of course, some of these people might be happier in a parallel universe. The smart might be happier if there were no mediocre people. Then they could enjoy attending college. The mediocre might be happier if there were no dumb people. Then they would not have to attend college. But, sadly for these people, they dont live in a parallel universe, and they must do the best they can with the environment in which they live. Why do Feltovich et al. (and economists in general) prefer this more complicated story to the simple one that college always makes the dumb and


the smart more unhappy than it makes the mediocre? There are a number of reasons: It is more interesting, and that makes it attractive to economists. It generates some predictions that can be tested. For example, imagine you could conduct an experiment in a controlled environment where you knew that people had the same preferences. Then if you observed the smart and the dumb behaving the same as each other, but differently from the mediocre, the simple happiness story could not be true. Feltovich et al. conducted just such an experiment, and they found exactly what their story predicted. Finally, once one understands the reason why different people would behave in this way in this particular context, it opens the door to understanding similar behavior in quite different contexts. Feltovich et al. offer a few examples:
The nouveau riche flaunt their wealth while the old rich scorn such displays. Mediocre students answer the teachers easy question, while the best students cant be bothered. Minor officials prove their status by being petty, while the truly powerful show their strength by being helpful. A person of middling reputation angrily refutes accusations against his character, while a highly-respected person does not dignify the accusations with a response.

A final comment is worth making in light of this example. Feltovich et al. assumed there are three types of people. In reality, of course, there are many types of people with equally many levels of ability. Feltovich et al. made the simplification in order to make it easier to tell their story (and make the math they had to do easier). This is standard practice in economics, because the world is a far too complicated place. But is it bad practice? It can be, sometimes. The hallmark of a good economist is that he or she knows what


can be simplified without making the story incorrect, and what cannot be simplified. Not all economists are good at this, but some excel at it. How do we as a profession decide when a story an inevitably simplified representation of the real world is a good story and when it is bogus? One is by following our instincts. Is it plausible? Does it help me understand other phenomena? If a story passes the instinct test, that gets us as a profession to take it seriously. But that is only the first step. The second is more formal. We can collect data and see if the predictions of the model are true. Only when a story passes both tests and this can take many years as other economists explore the idea and test it does the profession begin to accept the story as part of our received wisdom. It is too early to tell whether the story given by Feltovich et al. will pass these tests. Maybe it is the reason why so many smart people drop out of college, maybe it isnt. The early reaction has been that it passes the instinct test. We will see over the years if it passes the data test. In the meantime, though, we can have fun thinking about it, and that makes economists happy.

Modeling happiness with utility functions

Economists, being economists, use a different word than happiness. They use the term utility. We say that if a person is happier driving a Porsche than a Honda, he gets more utility from the Porsche. I get more utility sitting by the pool than I do standing in the lecture hall. You no doubt get more utility from an hour by the pool than you do from an hour sitting in this lecture hall. Utility is not a concept we can ever expect to measure. Can we say whether I would get more utility from an extra dollar of income than you would? Sure, I earn more than most of you (otherwise why are you wasting your time sitting in this class?), and so you might think that a dollar would be worth more to you than it is to me. But it turns out that I have an unusual love for money, so I should have the dollar.


Although we dont expect to be able to measure utility in practice, it is nonetheless useful to us to think about happiness in terms of utility, and to imagine we can measure individual utility with a unit of measurement that we will call a util. The concept of utility will help us think through the logic of individual decision-making. As we will soon see, the concept of utility will predict that when the price of something rises people buy less of it. We cannot measure utility, but we can observe prices and quantities. Hence, although we cannot measure utility, we can test the predictions that the concept generates. Table 1 provides an illustration. In this example, Mariela gets utility from drinking soda and from drinking water. If she drinks just one soda, she gets a benefit of 95 utils. If she has three sodas and two waters, she gets 495 utils, 255 from soda and 240 from the water. Total utility for any combination of water and soda can similarly be obtained by finding in each column the appropriate utility and adding the utils from the two columns together. Before seeing what we can do with this concept, we should note a simplifying assumption in the table. The amount of utility from water is assumed not to depend on how many sodas Mariela has. In reality, the world is likely to be a more complex, and her utility from drinking water will depend also on how much soda she has. To capture this additional complexity, we would have to create a much larger table: we would need a separate column tabulating the utility from water for each possible number of soda bottles, and we would need a separate column tabulating the utility from soda for each possible number of water bottles. We wont bother with that additional complexity here. Later, we will have more convenient ways to capture this likely interaction between soda and water consumption. Information such as that in Table 1 allows us to make predictions about the purchases that Mariela makes. Imagine Mariela has a budget of $10 to spend on soda and water. Soda and water both cost $1 per bottle. How will Mariela allocate her budget between soda and water?



Marielas Utility from Soda and Water

# OF BOTTLES 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 SODA 0 95 180 255 320 375 420 455 480 495 500 500 500 500 500 500 500 500 500 500 500 WATER 0 130 235 330 420 500 575 639 700 755 800 840 875 910 940 960 980 990 995 1000 1000

Let us begin by assuming Mariela spends all of her money on soda. With $10 to spend, she can buy ten bottles of soda, and if she does so she will get 500 utils. Can she do better? Sure. If Mariela buys only 9 bottles of soda, she will get 495 utils from the soda. But then she has a dollar left for water, and one bottle of water gives her 130 utils. Thus, 9 sodas and one water gives her 625 utils. 8 sodas and 2 waters give Mariela 715 utils. If we continue to reduce the soda count by one and raise the water count by one, Marielas utility will keep rising until she arrives at a consumption bundle of 4 bottles of soda and 6


bottles of water. This combination gives her 895 utils, which cannot be beat by any other combination she can afford. Imagine now that Mariela has $15 to spend on soda and water. This increase in her budget obviously allows her to buy more of both. By the same process as before, we see that her utility is maximized when she uses her budget to buy 6 bottles of soda and 9 bottles of water. So we have predicted that an increase in Marielas income will raise the quantity of soda and the quantity of water she chooses to consume. Now return to the original budget constraint of $10. But assume now that the price of soda declines to $0.50. We can calculate Marielas new preferred consumption bundle in the same way, and we find that now her utility is maximized when she buys 6 bottles of soda and 7 bottles of water. So now we have predicted that a reduction in the price of soda will induce an increase in the consumption of soda, and an increase in the consumption of water.9 Although the calculations necessary to find Marielas preferred consumption are very easy, they are rather tedious. There is an easier way to go about this. Instead of writing down a table of utilities, let us instead write down a table that records the increase in utility gained by increasing consumption by one bottle. This measure is known as marginal utility, and the numbers for Marielas problem are given in Table 2. Consider the soda column. In Table 1 we saw that utility from consuming one bottle of soda is 95 utils. Thus, the marginal utility of going from zero to one bottle is 95 utils. In Table 1 the total utility from consuming two bottles of soda is 180 utils. Thus the marginal utility from changing from one to two bottles is 18095=85 utils, as recorded in Table 2. And so we can go on down the columns. Note that the total utility from consuming, say, four bottles of

This is not always going to be the case. A drop in the price of soda could also have induced Mariela to drink less water. However, it will always be the case that she would drink more soda.


soda in Table 2 is exactly the same as the sum of the marginal utilities for one, two, three and four bottles of soda. With a moments reflection, you will see that the sum of the marginal utilities must in this way always equal the corresponding total utility.

Marielas Marginal Utility from Soda and Water

# OF BOTTLES 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 SODA 95 85 75 65 55 45 35 25 15 5 0 0 0 0 0 0 0 0 0 0 WATER 130 105 95 90 80 75 64 61 55 45 40 35 35 30 20 20 10 5 5 0

Remember that soda and water cost $1 each. From Table 2 we can ask the simple question, at what point does an extra dollar spent on soda provide no more utility than what is lost by reducing spending on water by the same


amount? The answer is simple: when Mariela consumes 4 bottles of soda and 6 bottles of water. If Mariela were to deviate from this choice, she would be made worse off. Raising her consumption of soda by one unit (to 5) raises her utility from soda by the marginal utility, 55 utils. But to accomplish this she must reduce water consumption to 5, causing her utility from water to drop by 75 (she loses the marginal utility from the 6th bottle). The loss exceeds the gain and she is worse off. Alternatively, if she reduced water consumption by one bottle, Mariela loses 65 utils from soda consumption, but she can simultaneously gain one bottle of water, worth 64. The loss exceeds the gain and so she has no incentive to change in this direction either. The analysis is only a little more difficult when soda and water have different prices. Imagine soda is only $0.50, while Mariela still has $10 to spend. We have already seen that in this case she buys 6 bottles of soda and seven of water. If she were to reduce expenditure by $1 on water, she would lose one bottle, for a loss of 64 utils. In compensation she could go from 6 to 8 bottles of soda, but this would only raise her utility from soda by 60 utils, 35 of which come from going from 6 to 7 bottles, and 25 by going from 7 to 8 bottles. On the other side, if Mariela were to raise expenditure on water by $1, she would gain 61 utils as she goes from 7 to 8 bottles. But to accomplish this she must give up 2 bottles of soda. This reduces her soda consumption from 6 to 4 bottles, for a loss of 100 utils. Moving in either direction away from a starting point of 6 sodas and 7 waters involves a net loss of utility. Hence, this consumption bundle must be optimal. Diminishing marginal utility. You no doubt will have noticed a special feature about the numbers we have used in Marielas allocation problem. The greater the number of soda bottles Mariela is already consuming, the smaller is the marginal utility of consuming an extra bottle. Similarly, the marginal utility of water is smaller the greater the number she is already consuming. This property of utility functions is known as the law of diminishing marginal utility, and economists believe it is an extremely common, if not universal, feature of peoples preferences. 42

There are two simple and intuitive reasons why marginal utility declines with increasing consumption. The first is that you tend to get sick of consuming the same thing. You satisfaction from your first bottle of soda is greater than the satisfaction from a bottle of soda consumed after you have already drunk four bottles. The second reason applies when the items you are consuming differ from each other in some way. Consider the case of CDs. If you have enough money to buy just one, you will buy your favorite. If you have enough money to buy two, you will also buy your second favorite. As your second favorite CD does not give you as much satisfaction as your first, the marginal utility of the first CD is greater than the marginal utility of the second CD. Opportunity cost. When Mariela increases her consumption of water by one bottle, it costs her $1. But there is another way to describe the cost. In order to increase spending on water by $1, Mariela must give up some soda. If soda is $1 per bottle, then the cost of an extra bottle of water is equal to one bottle of soda. If soda is $0.50 per bottle, the cost is two bottles of soda. This notion that the cost of something is what you must give up to get it is referred to as opportunity cost. This is a very useful concept. When the price of soda falls, the opportunity cost of water increases. Even though the dollar price of water has not changed, in a very real sense the cost of consuming water has increased it implies an even greater sacrifice in terms of foregone soda. Marginal utility of income. When prices are different, it turns out to be more useful to tabulate, not marginal utility per bottle, but marginal utility per dollar spent. Calculating the marginal utility per dollar spent is extremely easy. Simply take the marginal utility per bottle and divide by the price of the bottle:
MU per dollar spent on soda = MU per bottle of soda . price of soda


These calculations are done in Table 3, for the case in which water is $1 per bottle and soda is $0.50 per bottle. Now, start out from the optimal bundle, of $3 spent on soda and $7 on water. Switching $1 from soda to water reduces utility by 39 utils. Switching $1 from water to soda reduces utility by 4 utils. Both are loss-making changes, so Mariela wont make them.

Marielas marginal utility of income, when spent on soda and water Soda = $0.50 per bottle Water = $1 per bottle
# OF DOLLARS 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 SODA 180 140 100 60 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 WATER 130 105 95 90 80 75 64 61 55 45 40 35 35 30 20 20 10 5 5 0


Some further insight can be obtained if we graph the numbers in Table 3. Figure 1 plots the marginal utility per dollar spent on soda and the marginal utility per dollar spent on water. The x axis plots the amount of money spent on soda. From Table 3 we see that marginal utility falls as expenditure rises, so the line plotting the marginal utility of expenditure on soda has a negative slope. At the same time, a rise in expenditure on soda implies a fall in expenditure on water. Thus, the marginal utility of expenditure on water rises as expenditure on soda rises. For example, when $2 is spent on soda, $8 is spent on water, which from Table 3 has a marginal utility per dollar of 61. This is indicated by the symbol marked with the letter a in Figure 1. When $5 is spent on soda, only $5 is available for water, and the marginal utility of a dollar spent on water at this expenditure level is 80. Clearly, at the optimal bundle ($3 spent on soda), the two marginal utilities are close to each other. But they are not identical. This is only because we have been assuming Mariela can only by complete bottles. In fact, if it were possible, Mariela would like to consume fractions of bottles. If she could, then she could do no better than consuming where the marginal utility of money spent on each good is identical. As Figure 1 shows, this corresponds to an expenditure of about $3.75 on soda, and hence of about $6.25 on water. Beginning at this consumption bundle, if Mariela raises her consumption of soda and moves to the right in the figure, the gain in utility from consuming more soda (the marginally utility of soda) is less than the loss from consuming less water (the marginal utility of water). If she moves to the left, then the gain in utility from consuming more water (the marginally utility of water) is less than the loss from consuming less soda (the marginal utility of soda). A move in either direction results in a net loss of utility, so the intersection of the two lines must be the optimal consumption bundle.


200 180 160 140 120 100 80 60 40 20 0 0 2 4 6 8 10



Expenditure on soda

Expenditure on water Mariela's budget ($10)

FIGURE 1. Marginal utility of income

In summary, then, Marielas happiness from consuming soda and water is maximized by equating the marginal utility of a dollar spent on each of the goods:
MU per bottle of soda MU per bottle of water . = price of soda price of water

When Mariela optimizes her expenditure in this way, the marginal utility per dollar of expenditure evaluated at the optimal consumption levels is also known as the marginal utility of income. The name is intuitive. If we give Mariela an extra dollar of income, the best she can do is increase her utility by this amount.


2. Incentives matter
People behave in certain ways to make themselves as happy as possible given the environment they have to live in. It follows that if we change the environment, we can change peoples behavior. Doing so may allow policymakers to achieve desirable goals. The key for the policymaker is to understand why people behave the way they do. If the policymaker can understand this, she can provide the right incentives to induce the behavior, and hence the social outcomes, that she wants. All creatures respond to incentives. It is how we train dogs, for example. In my house we have a parrot. When we first brought him home, he used to bite us quite regularly. To induce him not to bite, we would put the parrot in his cage as a punishment. What we didnt realize is that there was another incentive at work. The parrots food and water was also in the cage, and he soon learned to bite us every time he was hungry or thirsty. The parrot consequently gave us a strong incentive not to put him in the cage every time he bit us. We respond well to incentives and have changed our behavior in just the way the parrot prefers. Failure to think about incentives can induce policies that fail. Steven Landsburg (1994a) tells a revealing story about road safety. In 1965 Ralph Nader published a book, Unsafe at Any Speed, which argued that, although technologies existed to make cars safer at reasonable cost, manufacturers were not installing them. In 1968, the Federal Government responded with a set of regulations that required manufacturers to install a number of items, including safety glass, seat belts, collapsible steering columns, padded dashboards, and dual brake systems. No doubt the government expected to reduce deaths from automobile accidents. But thats not what happened. Because drivers felt safer in their newly-equipped cars, they drove faster and with less care. Doing so has a benefit its more fun. But it also has a cost in the form of an increased risk of death. The new regulations reduced the cost of


careless driving by reducing the risk of death if an accident occurs. Responding, as any animal will, to a change in costs and benefits, people did what any economist could predict: they had more accidents. The death rate of drivers and passengers per accident went down, just as the government predicted. But the number of accidents went up, as did the number of pedestrian deaths. The net effect on deaths could have gone either way. Whether the increase in accident rates and pedestrian deaths more than offset the reduction in motorist deaths per accident is a question that requires us to look at the numbers. And the numbers show that the 1968 safety regulations increased the number of motor vehicle deaths. After some complex analysis, Sam Peltzman (1975) concluded that, while the number of motorist deaths stayed about the same (the decline in the death rate per accident was just about offset by the rise in the accident rate), the number of pedestrian deaths rose beyond what it would have been had there been no regulations.10 There is no reason why this had to be the case. But because economists understand that incentives matter, any of them could have told the government that it was a distinct possibility. Interestingly, Peltzman also provided evidence on why it happened. He found that the increased deaths were driven (pardon the pun) by a number of behavioral changes, including increased speeding, an increase in the number of young drivers, and an increase in drunk driving. The numbers for drunk driving are particularly telling (see Table 4). Prior to the introduction of the 1968 regulations, arrest rates for drunk driving and general drunkenness had both been declining for at least fifteen years. But immediately after safety equipment began to be introduced, drunk-driving arrest rates rose. In fact, the drunk-driving arrest


The challenge is estimating what would have happened without the regulations, because many other things are going on to change behavior over the same time period. The statistical analysis of data to address such hypothetical questions is part of the specialized field of econometrics.


rate jumped by an astounding 63 percent between 1965 and 1971, while drunkenness arrests continued their declining trend.

Table 4. Annual Arrest Rates Drunk driving (per 1,000 drivers) 1953 1959 1965 1971 4.45 3.85 3.53 5.75 Drunkenness (per 1,000 adult population) 23.81 21.69 16.60 13.32

Source: Peltzman (1975), Table 10.

But wait! you should be screaming. If the government started getting serious about automobile safety in 1968, isnt it possible that they also started getting serious about drunk driving, so police were simply arresting more people?11 The answer is yes. So, perhaps a better test of the theory that the regulations encouraged drunk driving would be to look at drunk driving arrest rates on new cars (with the safety features) in comparison to older cars. Unfortunately, Peltzman did not have the data make this comparison. He was able to discover,

If you asked this question, you are already beginning to think like a critical economist. When faced with a an economic explanation for some observed data, the critical economists asks two questions: Is the story I am given plausible? Can I think of alternative plausible stories that explain the same observations? If the answer to the first is no, the economist focuses only on the alternative stories. If the answer to the second question is no, the economist is probably not thinking hard enough. If the answer to both questions is yes, the economist thinks about how to test which of the different stories is the better one.


however, that drivers of cars with the safety features were more likely to be involved in accidents.12 Peltzmans study of the effects of safety regulations on automobiles is just one of many thousands of economic studies showing that incentives matter. The death penalty has been shown to reduce murder, at the rate of about eight murders avoided per execution (Ehrlich, 1975). Divorce rates have risen since no-fault divorce has lowered its cost (Friedberg, 1998). And your attendance at class will be better given that I have unannounced in-class exams.

Modeling responses to incentives with utility functions

Let us return to Marielas water-soda consumption problem. With a price of $1 for both water and soda, and a budget of $10, we found that Mariela would prefer to consume 4 sodas and 6 waters. Perhaps the surgeon general thinks this is too many sodas. If so, she could easily induce Mariela to change her consumption habits. For example, the government could impose a tax of $1 on each bottle of soda. Doing this raises the price of a soda to $2, while leaving water at $1. We can easily predict how Mariela will respond. Given her budget of $10, if Marielas consumption of soda doesnt change from 4 she can only buy 2 waters, giving her a utility of 555. But she can increase her utility by decreasing her consumption of soda. Consuming 3 sodas and 4 waters gives her 675 utils; 2 sodas and 6 waters gives 755 utils; 1 soda and 8 waters gives 795 utils; and 0 sodas an 10 waters gives 800 utils. So Mariela quits drinking soda altogether. We could ask a related question. Suppose we wanted to reduce Marielas consumption of soda from 4 to 3 bottles. What is the smallest tax that could


Moreover, the timing of changes to drunk-driving laws does not seem to be right. Although drunk driving laws had been around a long time (the first law was introduced in New York in 1910), it was not until the 1980s that states began to enhance penalties and policy to step up enforcement, largely in response to pressure from interest groups such as Mothers Against Drunk Driving (Wikepedia, 2004).


accomplish this? We can proceed in the following way. At the current price of $1 per bottle, Mariela gets 420 utils from soda if she buys 4 bottles, and she has enough money left over to buy 6 bottles of water. We need to raise the price of soda, so that if she continues to consume 4 bottles she wont have enough money to buy 6 waters. In fact, if we put a tax of one cent on the soda, she will only be able to buy 5 bottles of water. She will then have spent $4.04 on soda, $5 on water, and is left with $0.96 that she cannot do anything with. Given this reduction in water consumption, would she be better off consuming only 3 sodas? Her utility from 4 sodas and 5 waters is 820. If she reduces her consumption of soda to 3 bottles, she has $6.97 left over to buy water. She can buy 6 bottles, giving her utility of 255 from soda plus 575 from water, or 830 in total. So it turns out that she would be better off buying only 3 sodas. Hence, a tax on soda of just one percent reduces Marielas consumption of soda by 25 percent.

3. The indifference principle

You have a choice between cleaning your apartment or watching television. The longer you spend cleaning the apartment, the cleaner it gets. Although you dont like cleaning, you like the end result. You may tell me you prefer watching TV to cleaning. I claim that either you are indifferent between the two, or your apartment is disgusting. If you really preferred watching TV, you would clean less and watch more. And you would keep increasing the amount of TV until you were indifferent. As long as you are doing some cleaning, you must be indifferent between the two. The only time this may not be true is when you dislike cleaning so much that you keep increasing your TV watching until you are doing no cleaning at all. Then you may still prefer watching TV to cleaning, and you may want to watch yet more TV and clean even less, but you cannot do a negative amount of cleaning. So if you tell me you prefer watching TV to cleaning, dont invite me over for dinner.


At Pittsburgh International Airport there is a point at which you must go from the first to the second floor in order to get to your departure gate. At this point you must make a decision do you walk up the stairs, or do you take the elevator, which is right next to the stairs? Some people choose one, and some choose the other. Why? One answer is easy some people prefer the escalator, others prefer taking the stairs. Some travelers have heavy bags to carry (even though at this point they should just have their carry-ons, but thats a whole other story), so they choose the escalator. Some people just like exercise, so they always run jauntily up the stairs. So let me ask about peoples choices in a more difficult setting. Imagine that everyone traveling is identical. You may guess that if everyone is identical they would all make the same choice. Either everyone takes the stairs or everyone takes the escalator. But this guess need not be correct. If identical people make different choices we can be sure of one thing: they must all be indifferent between the two choices. Put another way: if identical people make different choices, the choice they make is irrelevant for their happiness. To see why this is the case, we need to make use of our first two principles. We first need to think about what makes people happy what their utility depends on. We then need to think about how they respond to incentives. So lets think about what might affect peoples choices. A simple story might go as follows. People care about getting to the second floor quickly, but they dont like expending effort. Thus, everyone arriving at the foot of the stairs must trade-off speed against effort to maximize their utility. If someone arrives at the foot of the stairs, and both the escalator and the stairs are empty, we can confidently predict he will always choose the escalator. He can always walk up it at the same speed he would walk up the stairs. This undoubtedly makes him better off. First, he gets to the top faster. Second, he will expend less effort than going up the stairs (because the escalator is moving). The next person, finding only one person on the escalator is also likely to choose it. But as more and more people choose the escalator, it becomes congested. Walking up it 52

becomes more and more difficult until, eventually, everyone is standing still. It is then quicker to take the stairs. At some intermediate point, the escalator is sufficiently congested that the extra speed of taking the stairs just compensates for the extra effort involved. At that point, the next person to arrive is indifferent between the two choices. But then, as everyone is identical, everyone must be indifferent. Just as we can infer that you prefer watching TV only if your department is disgusting, we can also infer that travelers only prefer taking the escalator if no one is taking the stairs. Our story has some testable predictions. If anyone is using the stairs, they will get to the second floor more quickly than people taking the escalator. Fast escalators are more crowded than slow escalators (test these predictions next time you have an opportunity to observe). And our conceptual focus on identical people does not undermine the observation that, in addition, people with heavy bags that will always use the escalator. We can also make some policy predictions. If we think people dont get enough exercise and we would like to see more of them taking the stairs, we can accomplish this by slowing down the escalator motor. As long as we dont overdo it, so that the stairs themselves dont get congested, the only people we will make worse off are those with heavy bags. If no one has heavy bags, then there is a wonderful opportunity to make the world a better place. Slow down the escalator and induce more people to exercise. Everyone is indifferent between taking the stairs and the escalator so they wont mind switching, as long as the value of taking the stairs does not decline. Even if, as a policymaker, you dont care about exercise, you will at least have reduced electricity consumption. Steven Landsburg (1994b) uses the indifference principle to explain why, as a first cut, we can assume that all cities in the United States are equally good places to live:


Each year, the Places Rated Almanac and The Book of American City Rankings issue their reports on the best places to live in America. San Francisco gets credit for its cosmopolitan charms and Lincoln gets credit for the allure of its housing market. Weighing the importance of education, climate, highways, bus systems, safety and recreation, researchers rank cities in order of overall desirability. The implicit assumption is that researchers have identified features that most people care about, and that we all pretty much agree about their relative importance. If that assumption is correct, and if your tastes are not atypical, you can save yourself the expense of purchasing the manuals. When all factors are accounted for, all inhabited cities must be equally attractive. If they werent, nobody would live in any but the best. Landsburg (1994b, p. 31)

I have underlined the key assumptions. First, the cities must be inhabited. If everyone has left, then it is likely that everyone prefers not being there. Similarly, if no one is using the stairs, it is likely that everyone prefers being on the escalator. Second, your tastes should be typical. For example, if you like tropical weather much more than is typical, you will prefer Miami to Boston even though the average citizen is indifferent. If you are unusually good at software programming, you will prefer Silicon Valley to the Napa Valley, while if you have specific skills in winemaking, you will prefer the Napa Valley to Silicon Valley. If you are special in some way, you can be rewarded or punished for being different. Imagine you live in Boston but have an unusually strong preference for tropical weather. Then you will be unusually unhappy in Boston. But you may be able to do something about it: you can move, and become unusually happy in Miami. Similarly, if you have a heavy bag, you will be unusually unhappy going up the stairs. But you can do something about it: you can take the escalator, and be unusually grateful that the escalator is there.


Modeling the indifference principle

We return, once more, to Marielas allocation problem. If Mariela were to consume no soda and 5 bottles of water, she earns 500 utils. If she were to consume 10 sodas and no water, she would also earn 500 utils. When the utility a person gets from two different consumption bundles is the same, we say that person is indifferent between them. The concept of indifference gives us yet another way to think about Marielas decision problem. We can choose an arbitrary level of utility, say 500 utils, and find every consumption bundle that gives this level of utility. In a graph that plots the consumption of soda on one axis and the consumption of water on the other, we know that the graph must pass through the bundles [0 soda, 5 water] and [10 soda, 0 water]. We could choose another level of utility, say, 880 utils. From Table 1, we can see that this line must pass through [3 soda, 6 water]. Figure 2 plots a handful of these lines, along with some points (marked by circles) indicating consumption bundles from Table 1 that give the indicated level of utility. Each line is called an indifference curve, and the collection of lines is an indifference map.

Water (# of bottles)


1120 utils

895 utils 500 utils 0 0 5 Soda (#of bottles) 830 utils 10 15

FIGURE 2. Marielas indifference map


We have drawn Marielas indifference curves with four distinctive features: Higher indifference curves correspond to higher utility. They are downward sloping, They get flatter as we move from left to right. They do not cross each other. It is useful to take some time to understand why graphs of indifference curves for two goods have each of these properties.

Higher indifference curves correspond to higher utility. The reason for this is
very simple. If Mariela likes soda and water, then she can always be made better off by giving her more of both. That is, if we move Mariela in a northeast direction in the graph, so that she has more of both goods, she will be happier. Clearly, then, Mariela would like to get on the highest indifference curve possible.

They are downward sloping, The reason for this is that for Mariela a reduction
in water consumption must be offset by an increase in soda consumption if she is to maintain the same utility. Mariela likes both soda and water. If we were to give Mariela an extra bottle of soda while keeping the amount of water she has constant, she would be better off she would have higher utility. The only way to stop Mariela from being better off when we give her the extra soda is to simultaneously take away some water. That is, along an indifference curve, an increase in soda must be accompanied by a reduction in water. Similarly, an increase in water must be accompanied by a reduction in soda. How much water must we take away if we add a bottle of soda, in order to leave Mariela no better off than before? The answer to this is given by the slope of the indifference curve, as Figure 3 illustrates. Imagine we start from point a, where Mariela has four sodas and six waters. We now give her an extra bottle of soda, moving her to point b. Clearly, at point b, she is better off. In order to get Mariela back to the original indifference curve, we have to take 56

some water away, and this is indicated by the vertical distance between b and c, indicated by x. As Figure 3 shows, this is just the slope of the indifference curve at that point.

Water (# of bottles)

1 a


Soda (# of bottles)

FIGURE 3. The marginal rate of substitution of water for soda.

Economists, of course, have a special name for the amount of water Mariela must give up when she receives this extra bottle of water in order to be no better and no worse off than before.13 We call it the marginal rate of substitution of water for soda, which we denote by MRSw,s. The MRSw,s is equal to the slope of the relevant indifference curve at each point. Note that we write it as the MRS of water for soda because water is on the vertical axis and soda is on the horizontal axis. If we switched axes, the slope of the indifference curve would give us at each point the MRS of soda for water. A little bit of thought and you will quickly realize that the two MRSs are related by

13. Equivalently, one can say the maximum amount of water Mariela is willing to give up in order to get one additional bottle of soda.


MRSw ,s =

1 . MRS s ,w

How does the MRS relate to utility? It turns out that there is a simple relationship between the MRS and Marielas marginal utility for each of the goods. Too see this, let MUs denote Marielas marginal utility of soda, and let MUw denote her marginal utility of water. Now increase her soda consumption by one bottle. Marielas utility will increase by the amount MUs. Let x denote the amount we must change Marielas consumption of water of water so as to leave her on the same indifference curve . How big should x be? Well, MUw is the change in Marielas utility caused by changing her water consumption by one unit, so xMUw is the amount that her utility changes when we take x away. To leave Mariela on the same indifference curve, we require that xMUw= MUs. Solving for x:
x =
MU s . MUw

The quantity x is, of course the slope of the indifference curve (review Figure 3), and this is of course the MRSw,s. Thus, we have
MRSw ,s = MU s . MUw

That is, the marginal rate of substitution of water for soda is given by the ratio of the marginal utility of soda to the marginal utility of water.

They get flatter as we move from left to right. This is because of diminishing
marginal utility. When Mariela is consuming only a small quantity of soda, she requires a large amount of water to compensate her for any reduction in soda consumption. When she is consuming a large quantity of soda, she needs only a small increase in water to compensate for a reduction in soda consumption.

They do not cross each other. To show that indifference curves cannot cross
each other, we will do a proof by contradiction. That is, we will assume that


they can cross, and then show that we get a logical contradiction. This contradiction then allows us to conclude that they cannot cross. Consider Figure 4, which contains two indifference curves that cross. Four different consumption bundles, a, b, c, and d, are shown on the diagram. Because b lies to the northeast of a, the assumption that more is better implies that b is a better bundle than a. Because b and c lie on the same indifference curve, b and c are by definition equally good bundles. But if b and c are equally good, and b is better than a, it must be the case that c is also better than a. So we have just concluded that c is better than a. Now, note that d lies to the northeast of c so d is better than c. But d lies on the same indifference curve as a so by definition they are equally good. But if a and d are equally good, and d is better than c, it must be the case that a is better than c. This conclusion contradicts our first conclusion. This contradiction that a is better than c and c is better than a is unacceptable, and we are forced to conclude that indifference curves cannot cross.

Water (# of bottles)

a d c

Soda (# of bottles)

FIGURE 4. Proof that indifference curves cannot cross.


Do all indifference curves have the properties we have drawn here? The answer is no. For example, if Mariela hated soda, we would have to give her more not less water if we were to force her to consume more soda. In this case, the indifference curve would have a positive slope. You are asked to think about and draw some alternative possibilities in the problem set.

The consumer choice problem with indifference curves.

Mariela would like to get to the highest indifference curve possible. However, her budget constrains her ability to move to ever higher indifference curves. In Figure 5, this budget constraint is indicated by a straight line. Calculating where it should be on the graph is easy. Marielas budget is $10. If she spends all her money on water (at $1 per bottle), she can buy 10 bottles. If she spends all her money on soda she can also buy 10 bottles. We mark these two end points on the graph, and connect them with a straight line. This is the budget line. Mariela can also buy any combination of water and soda that lies on the budget line. She can also buy any combination of water and soda that lies below and to the left of the budget line, but if she does so she will have some money left over. Mariela cannot buy any combination that lies above or to the right of the budget line, because all these combinations cost more than $10. The best that Mariela can do then, is get on the highest indifference curve that has a point somewhere that does not lie outside the budget line. This is indicated at point a, where the budget line is tangential to the indifference curve. The indifference map not only tells us how well off (in terms of utility) Mariela is, but it also tells us the combination of soda and water that she will choose. In practice, we do not expect to be able to measure any individuals indifference curve. But because indifference curves have the same general shape, we can use them in a conceptual way, as an aid to thinking. The following examples illustrate how economists use them.


15 Budget line Water (# of bottles) 10

5 895 utils 500 utils 0 0 5 Soda (#of bottles) 830 utils 10

1120 utils


FIGURE 5. Marielas indifference map and budget constraint

The effect of price changes. A change in the price of one of the goods under consideration causes Marielas budget line to shift. The effect of a reduction in the price of soda is illustrated in Figure 6.14 The original budget line is indicated by Aa. When the price of soda declines, the budget line rotates outward, to Ab. If all Marielas budget is spent on water, she can buy no more than before, so the budget line must continue to pass through point A. But if all her budget is spent on soda, she can now buy more than before, and this is indicated by point b. The reduction in the price of soda allows Mariela to change her consumption bundle. With the new budget constraint, the best she can do is to switch from bundle 1 to bundle 2, on a higher indifference curve. Note that at bundle 2, Mariela is consuming more soda and more water. It seems intuitive that Mariela would consume more soda when it becomes cheaper. But it is perhaps less intuitive that she would also consume more water. Here is one way to think about why this is the case: a decline in the

14 . Once you understand this, you can easily work out the effect of a price increase.


price of soda has made Mariela better off, in much the same way that an increase in income would. Soda becomes cheaper, so Mariela buys more. But even after she has bought the additional soda, she has some extra money left over, which can be used to buy additional water.

Water (# of bottles)

2 1 I1 I0 a Soda (# of bottles) b

FIGURE 6. The effect of a reduction in the price of soda

If Marielas indifference curves had a different shape, it is quite possible that a reduction in the price of soda would have induced her to consume less rather than more water. Perhaps more surprisingly, it is also theoretically possible for Mariela to have chosen to consume less soda after its price declined.15 The problem set at the end of this chapter asks you to draw indifference maps that depict these outcomes. The effect of income changes. If Marielas income increases, her total budget for soda and water may also increase. If she allocates all her budget to water,


If the quantity of a good consumed declines when its price declines, the good is called a Giffen good (named for the person who first described the idea). Although one can draw an indifference map for a Giffen good, none has ever been observed in the real world. Thus, it remains a theoretical curiosity.


she can buy more after the income increase than before. This is captured by the movement from A to B in Figure 7. If she allocates all her budget to soda, she can also buy more after the increase in income than before. This is captured by the movement from a to b. Note that the proportional jump from A to B is the same as the proportional jump from a to b. Hence, the new budget line is shifted outward, but it remains parallel to the original budget line.
B Water (# of bottles) A

2 1

I1 I0 a b Soda (# of bottles)

FIGURE 7. The effect of an increase in Marielas income

In Figure 7, the increase in the budget enables Mariela to switch from consumption bundle 1 to bundle 2. She consumes more water and more soda and we know that she is better off because she is on a higher indifference curve. If Marielas indifference curves had a different shape, an increase in income could have caused to her to consume less soda or to consume less water (but you will not be able to draw an example where she consumes less of both!). If consumption of a good increases after an increase in income, we say that the good is a normal good. If consumption decreases, we say it is an inferior good. The problem set asks you to draw some of these possibilities. 63

The equations of budget lines and consumer choice. Imagine you have an amount M to spend on two goods, the quantities of which are denoted by x and y. The prices of these goods are px and py. Your budget constraint can then be written as the following inequality:
px x + py y M ,

which simply says that expenditure on the two goods must be less than or equal to your budget. Along the budget line, the entire budget is being spent, so the equation of budget line is
px x + py y = M .

Figure 8 illustrates this budget. Because good y is on the vertical axis, it is useful to rearrange this equation so that y is on the left hand side and everything else is on the right. Doing so gives
y= M px x. py py

This is the equation of a straight line that intercepts the vertical axis at M/py, and that has a negative slope equal to px/py. So for Marielas budget problem (Figure 5), where M=$10, and both prices are $1, the budget line equation intercepts the vertical axis at 10 units of water, and it has a slope of 1. As we have seen, a reduction in the price of x rotates the budget line upward. The intercept with the vertical axis (equal to M/py) does not depend on px and is unchanged. But the absolute value of the slope, px/py, declines as the budget line becomes flatter. In contrast, if income changes, the budget line shifts (because M appears in the intercept term; but in this case the slope does not change. Now consider the optimal consumption choice. We know that this is found where an indifference curve is tangential to the budget line. At that point, of course, the slope of the indifference curve is equal to the slope of the budget



Slope = -px/py


FIGURE 8. Budget line for p x x + p y y = M .

line. The slope of the indifference is given by -MU x / MU y , so at the optimal consumption bundle we have
MU x p = x . MU y py

which says that the ratio of the marginal utilities must equal the ratio of the prices. We can rearrange this equation to obtain
MU x


MU y


This equation should be familiar, because it is exactly the same equation defining optimal consumer choice that we had already worked out from Figure 1. More complicated budget lines. Not all budget lines are straight lines. Consider the following price schedule. The price of good y is $5. The price of good x is $5 for the first ten units you buy, and then $2.50 for each additional unit. As you already know that the slope of the budget line is equal to the ratio of


prices, you can immediately see that the slope of the budget line must change when the price changes. Figure 9 plots the budget line for a person with $100 to spend. Note the kink in the budget line at x=10. Up to that point, the two goods have the same price, so the slope of the budget line is 1. Beyond x=10, the market exchanges two units of good x for each unit of good y, so the slope of the budget line changes to 1/2. Budget lines can be as complicated as the pricing schedule that gives rise to them. In general, every time a price changes, the slope of the budget line changes.

20 Slope = -1

10 Slope = -1/2



FIGURE 9. A discount after making a minimum purchase of good x.

For some pricing schemes, there may also be jumps in the budget line. Consider the following pricing schedule: The price of good y is $5. The price of good x is $5 per unit if you buy any quantity up to 10 units, and $4 per unit if you buy any quantity in excess of 10 units. Figure 10 illustrates for a budget of $100. Up to 10 units of x, the budget line has a slope of 1. But then if 10.00001 units of x are bought, the price for all the units falls by a dollar. To the right of this point, then, the budget line becomes flatter. But there is also an upward jump. Buying 10 units of x at $5 leaves enough money to buy 10 units of y. But buying 10.00001 units of x at $4 leaves enough money to buy (very nearly) 12 units of y.



12 10

10 12.5


FIGURE 10. A bulk discount for good x.

Even though we do not know a buyers precise preferences (i.e. the shape of her indifference curves), we can still make an interesting prediction about her choices. Any consumer that would buy at least eight units without the bulk discount will choose instead to buy at least 10 units when the discount is offered. The reason is easy to see. With the bulk discount, a consumer can consume almost exactly 12 units of y by buying either 8 units of x or 10.00001 units of x. As more is better, why not buy the larger amount of x? This is, of course, the whole point of offering bulk discounts in the first place. Price changes when you already own a good. Imagine you are a farmer. You have 10 pounds of beef and 100 pounds of corn. You like consuming beef and corn. The market price of beef is $10 per pound; corn is $1 per pound. Figure 11 illustrates your budget line. If you consume the corn and beef that you have, you can use the money to buy 10 pounds of beef. This consumption bundle is indicated by point E, where the letter E denotes your initial endowment. But you can increase your corn consumption by selling some beef and using the income to buy more corn. The rate at which you can do this is indicated by the segment of the solid budget line to the right of E. If you sell


all your beef, you will be able to consume 200 pounds of corn. Alternatively, you can sell some corn to increase your consumption of beef. If you sell all your corn, you can consume a maximum of 20 pounds of beef.

20 15 10 E





FIGURE 11. Price changes when you begin with an endowment of goods.

What is new about endowments is that price changes behave differently than when you start out with cash. Imagine the price of corn drops to $0.50 per pound. You can still consume your initial endowment, so E remains on the budget line. But now if you sell all your beef, you can consume 300 pounds of corn; if you sell all your corn, you can only buy 15 pounds of beef. This new budget line is indicated by the dashed line. Note how the price change caused the budget line to rotate around your endowment point. Compare this with a price reduction when your endowment is only cash.

Who cares about a real-estate crash? Ask anyone that owns a house how they
would feel if the price of housing were to rise. They will tell you they would be pleased because the increase in price will make them better off. Ask them how they would feel if the price of housing were to fall. They will tell you it would make them worse off. On this second count, however, they would be


wrong. An increase in the price of housing makes homeowners better off, but so does a reduction in the price of housing. To see why this is the case, we need to make use of our indifference maps and budget lines. Figure 12 plots an indifference map depicting Daemons choice between housing (by which we mean more housing is a better house) and all other goods. Assume the original budget line is given by Aa. Given Daemons budget, he went ahead and chose a quantity of housing and other goods that put him at point 1. Because this was Daemons choice, it must be a point at which is budget line is tangential to his indifference curve. But now imagine that the price of housing rises. Because Daemon has already bought his house, even when prices change we know he can still afford to consume at point 1. That is, the house is paid for and he has enough income to buy the quantity of other goods indicated at point 1. The new budget line has a flatter slope, reflecting the rise in the price of housing, but it must pass through Daemons original consumption bundle. This is indicated by the line Bb. But with this new budget line, Daemon can now do better. He can sell this house at its higher price, buy a smaller house, and have money left over to buy more goods. Doing so allows Daemon to attain consumption bundle 3, on a higher indifference curve. Thus, an increase in the price of housing makes Daemon the homeowner better off. And so does a reduction in the price of housing. When housing prices fall, the budget line gets steeper, but it still passes through point 1. Daemon can now sell his house, buy a new bigger house, and consume at point B on the new budget line Cc. Again, Daemon attains a higher indifference curve, and is better off than at point 1.


C Housing A B 2

a Other Goods

FIGURE 11. The effect of housing price changes when Daemon already own a house

When Mariela has a fixed money budget, and must choose between buying water and soda, she is made better off when the price of one of the goods falls (moving from 1 to 2 in Figure 4), and she is made worse off when the price rises (moving from 2 to 1 in Figure 4). But for Daemon, both a price increase in and price reduction make him better off. The difference for Daemon is that after the price change, his original consumption bundle was still attainable. He didnt have to sell his current house and buy a bigger one (if housing became cheaper) or a smaller one (if housing became more expensive). Thus, he would only do so if it made him better off. In contrast, Marielas original consumption bundle is no longer feasible she cannot afford it when the price of one of the goods rises. Being imaginative: the escalator problem. To think about the escalator problem, we have to get inventive. People care about speed and effort. But, while speed is a desirable thing, effort is undesirable. Our previous indifference maps had two desirable goods on the axes, and diminishing marginal utility gave the indifference curves their shape. What would the


indifference map look like if we look at the choice between one thing that is good (speed) and one that is bad (expending effort)? No doubt we could work it out, but here is a trick. If one of the items you are thinking about is undesirable, simply reverse the scale! That is what we have done in Figure 12. On the horizontal axis we have speed, with speed increasing as we move from left to right. On the vertical axis we have put effort. But zero effort is at the top, and high effort is at the bottom. Thus, as we move up the axis, we have less effort expended, and this is a desirable thing. We can now plot indifference curves with their usual shape, and we have plotted two, labeled I0 and I1. Moving in a northeast direction involves more speed and less effort, both of which are desirable, so I1 involves a greater utility than I0. Next, we have to think about budget constraints. Consider first the constraint for taking the stairs. At zero effort, speed must be zero. At maximum effort the speed is positive. The feasible trade-offs between reduced effort and increased speed are traced out by the straight line Aa. Now consider the escalator constraint. Even at zero effort people have positive speed on the escalator. This is indicated by the horizontal distance AB, which must be equal to the speed of the escalator. For any given effort, the escalator will get you to the top quicker than the stairs by an amount that depends only on the speed of the escalator. Thus, the budget constraint for the escalator is a line parallel to Aa, shifted right by the escalator speed. This is indicated by Bb. The best you can do on the stairs is choose point 1, where Aa is just tangential to I0. On the escalator, you can do better, by choosing point 2 on I1. But now, given that the escalator makes people better off (putting them on a higher indifference curve), everyone chooses the escalator. This will create congestion. What will congestion do? It will make it impossible to walk up the escalator too quickly. Because other people are in the way, congestion will force you to expend less effort and move more slowly. The effect of congestion is to make the right part of the line Bb inaccessible. Congestion must increase


until people are indifferent between the stairs and the escalator. This is indicated by the vertical line Cc. Although people would like to expend more effort and move down the budget line to point 2, congestion stops them from doing so, and they are stuck at point 3. At this point, taking the stairs and taking the escalator both put you on the same indifference curve, I0.

0 Effort (reverse scale)

A B 3

Zero effort

2 1

I1 max 0 I0 c a b Speed

FIGURE 12. The escalator problem

This is the only equilibrium. If congestion were somehow to decline on the escalator, Cc moves to the right. But this allows people to get more utility by taking the escalator. Then, everyone chooses the escalator, congestion jumps up, and Cc moves left again. If Cc were to move to the left of point 3, people would switch to the stairs. This would reduce congestion, shifting Cc right again. Note that at the equilibrium, points 1 and 3, people expend less effort on the escalator while, just as we predicted, they move faster on the stairs.

4. Concluding comments
Despite its reputation as the dismal science, economists care only about peoples happiness. In many applications, this may involve giving them more money and more goods to consume. But the application of economics is far


more reaching than that. Nobel laureate Gary Becker has been particularly productive in applying economic principles to address many issues not normally thought of as the domain of economics, including, fertility, marriage, divorce, addiction, organ donation, philanthropy, discrimination, and crime. What makes a study economics, as opposed to, say sociology, is not the subject matter. rather, it is the methodology, and the basic principles that the researcher brings to bear on a problem. Central to these basic principles is that people respond to incentives. The fact that incentives matter leads us to a yet another basic principle, that if two otherwise identical people are doing different things, they must be indifferent between them. If not, one of them would have an incentive to change what he is doing. We have used these basic principles to explore how individuals might choose how to spend their income on consumption goods, and to predict how they will change their behavior if prices and income changes. This is what most people consider the normal domain of economics. But we have also used the exact same principles to study how people choose between taking the escalator and taking the stairs. We will see many more applications of these basic principles throughout the rest of the course.

Concepts introduced in this chapter

utility marginal utility diminishing marginal utility marginal utility of income indifference curve normal and inferior goods proof by contradiction budget constraint opportunity cost consumption bundle Giffen goods endowments marginal rate of substitution optimal consumption bundles


1. Agnes has $10 to spend on soda and water, and earns utility from consumption according to the following table. (i) If the price of soda and water is $1 each bottle, how much soda and water will Agnes buy? (ii) If the price of soda increases to $2 per bottle, how much soda and water will Agnes buy? (iii) If the price of soda is $1 per bottle, but water increases to $3 per bottle, how much soda and water will Agnes buy? (iv) What will Agnes purchases be if her budget falls to $6?

Agnes Utility from Soda and Water

# OF BOTTLES 0 1 2 3 4 5 6 7 8 9 10 SODA 0 260 470 660 840 1000 1150 1280 1400 1510 1600 WATER 0 180 360 510 640 750 840 910 960 990 1000

2. Consider the problem of allocating a budget between the purchase of CDs and DVDs. Draw indifference curves that yield the following responses to changes in prices and income: (i) A reduction in the price of DVDs causes an increase in DVD purchases and a decline in CD purchases. (ii) A reduction in the price of DVDs causes a reduction in DVD purchases (i.e. DVDs are a Giffen good). What happens to CD purchases?


(iii) An increase in income causes a reduction in CD purchases (i.e. CDs are an inferior good). Are DVDs a normal or an inferior good? Remember that indifference curves must have a negative slope, they become flatter as one moves from left to right, and they cannot cross each other. 3. Draw an indifference map for each of the following cases. Indicate which indifference curves represent higher utility: (i) Mariela loves soda and hates water. (ii) Mariela hates soda and water. (iii) Mariela loves water. She likes soda as long as she consumes less than 120 bottles. Thereafter, additional soda makes her sick. 4. Good y has a price of $1 per pound. Good x has a price of $1 per pound for the first 10 pounds you buy, and then $0.50 per pound for each additional pound. You have a budget of $20. Draw the budget line for this problem (be sure to indicate its location with some numbers), and show an indifference curve such that there are two optimal bundles. 5. Farmer Giles owns 10 bushels of corn, and he also has $100 in cash. The price of corn is $10 per bushel, and the price is the same whether Giles buys or sells corn. The price of the only other good Giles cares to consume is beef, which costs $10 per pound. Draw his budget line (be sure to indicate its location with some numbers). Then show how the budget line is affected by an increase in the price of corn to $20 per bushel. 6. Now imagine that there is a difference in the prices at which Farmer Giles can buy and sell corn. If he chooses to sell corn, he must sell wholesale at $8 per bushel. If he chooses to buy corn, he must pay retail, which is $10 per bushel. Draw his budget line for this case (Giles still must pay $10 per pound for beef, and he still starts out with 10 bushels of corn and $100). 7. John likes relaxing and dislikes studying. However, he likes to get good grades, and knows that studying helps him achieve them. John as 24 hours available per day to allocate between studying and leisure.


(i) Past experience tells John that if he does not study at all, he will get 20 percent on his final exam. But for every hour per day that he studies, his grade will go up by 10 percentage points. The maximum he can get on the exam is, of course, 100 percent. Draw Johns indifference curves and budget line, and show an optimal amount of studying of 5 hours. (ii) Now assume that John only cares about increasing his probability of getting an A. The probability of getting an A depends on the time spent studying according to the following formula: Prob John gets an A = 0.1h 0.01h 2 , where h is the number of hours per day spent studying. For example, if John spends 2 hours per day studying, his probability of getting an A is 0.1x2 0.01x22 = 0.2 0.04 = 0.16. Draw the budget line for this problem (it is a curve). Explain why John will never choose to study more than five hours per day, whatever his preferences are. (NOTE: when budget lines are curved as in this question, economists usually refer to the line as a production possibility curve. Whether we call it a budget line or a production possibility curve, it depicts the same thing, in this case the technical ability that John has to substitute between leisure and grades). 8. The Department of Water at the County of Kauai, Hawaii has the following rate structure for residential customers. Each residence is charged a flat $12 monthly fee. In addition there is a charge for the quantity of water consumed according to the following price schedule $2.75 per 1,000 gallons for the first 10,000 gallons consumed in a month. $3.20 per 1,000 gallons for consumption beyond 10,000 and up to 20,000 gallons. $4.50 per 1,000 gallons for consumption beyond 20,000. This type of schedule is called an increasing block rate. Using a graph with water on one axis and consumption of all other goods on the other axis, sketch out the budget line for this problem.


Ehrlich, Isaac (1975): The deterrent effect of capital punishment: A question of life and death. American Economic Review, 65(3):397-417. Feltovich, Nick, Rick Harbaugh, and Ted To (2002): Too cool for school? signaling and countersignaling. RAND Journal of Economics, 33(4):630-649. Friedberg, Leora (1998): Did unilateral divorce raise divorce rates? Evidence from panel data. American Economic Review, 88(3):608-627. Landsburg, Steven E. (1994a): The power of incentives. In The Armchair Economist, New York: The Free Press, ch. 1. Landsburg, Steven E. (1994b): The indifference principle. In The Armchair Economist, New York: The Free Press, ch. 4. Peltzman, Sam (1975): The effects of auto safety regulation. Journal of Political Economy, 83(4):677-726. Wikepedia (2004): Drunk driving. In Wikepedia. The Free Encyclopedia. http://en.wikipedia.org/wiki/Drunk_driving. Accessed, December 1, 2004.


Chapter 3 Efficiency and Equity

In chapter 1 we studied a Vickrey auction involving Mara and six friends. We concluded that the Vickrey auction set a price such that the sum of the welfare of consumers and the welfare of the seller is maximized, and that there was no other allocation that makes society better off. We concluded that this seems to be a desirable state of affairs. And of course it is: we would surely like everything in the world to be arranged so as to make society as well off as possible. Unfortunately, make society as well off as possible is too vague an instruction to be of any practical use. Economists, being practical people, have developed a far more useful guide as to how the world should be organized. The guide rests on two distinct ideas efficiency and equity that form the basis of how economists evaluate the desirability of different outcomes and different policies. In this chapter, we study the concepts of efficiency and equity separately, and in that order. Doing it this way is not simply a matter of didactic convenience. Economists not only teach these ideas separately and in that order, they also work in this way. To see why this is the case, consider the problem of providing pizza to two of your friends, whom (seeing as I dont know them) I shall call Greg and Elaine. For the sake of argument, let us assume that Greg and Elaine love pizza and never stop being hungry. One way to approach the problem is to divide it into two tasks. First, make the pizza as large as possible. No matter what sharing scheme you devise for any given pizza, you can always make both Greg and Elaine better off by increasing its size and giving both of them a little more. Only after you have made the pizza as large as you can do


you tackle the second task of deciding how to split it between your friends. The concepts of efficiency and equity are analogous to these two tasks. An efficient outcome is one that has made the pizza as large as possible. An equitable outcome is one that allocates the pizza to Greg and Elaine in a way that seems fair. Obviously, one could have an efficient outcome that that is entirely inequitable, and one could have an equitable outcome that is grossly inefficient. The socially optimal outcome is one that is both efficient and equitable.

1. Efficiency
Economists have a particular kind of efficiency in mind, called Pareto efficiency.16 It is defined as follows: A Pareto efficient outcome is one that cannot be changed so as to make someone better off without also making someone else worse off. As a natural corollary, A Pareto improvement is a change that makes at least one person better off without also making anyone else worse off. Any outcome that we can change to yield a Pareto improvement is said to be Pareto inefficient. When economists use the words efficient or inefficient alone, they mean Pareto efficient or Pareto inefficient. My wife tells me that I am inefficient at sweeping the floor. But she is wrong. I think what she means17 is that I am slow and incompetent, and that I leave parrot feathers under the bird cage. She is quite correct, but this does not mean I am inefficient. To the contrary, there is no way for me to do a better job (thereby making my wife happier) without making myself worse off. The


Named for Wilfredo Pareto (1848-1923), an Italian economist whom you need know nothing about. Sometimes its better not to ask.



feathers under the bird cage, annoying as they are to my wife, are part of a state of affairs that constitutes a Pareto efficient outcome. Ergo, I am an efficient sweeper. Of course, in my assessment of my sweeping prowess, I am ignoring equity. My personal tradeoff of feathers against relaxation time may not pay a fair amount of attention to my wifes preferences for a clean floor. But equity is a separate issue that well get to later. Being efficient is not the same as making society as well off as possible. But this does not make efficiency an empty concept. To the contrary, it is very powerful. If any current state of affairs is not Pareto efficient, the policy implication is straightforward: devise a policy to attain a Pareto improvement. Sadly, Pareto improvements are often hard to come by. The only way to give Greg more pizza without taking it away from Elaine is to provide a larger Pizza. But, in practice, you will have to pay more for a larger pizza, so you are worse off. If you dont buy a larger pizza, then making Greg better off requires that you make Elaine worse off. The only way for my wife to enjoy a cleaner floor without making me worse off is to discover a new technique that allows me to sweep better with no more effort than I already expend. We havent come up with such a magic bullet yet. In chapter 2 we saw that that under some special assumptions -- that everyone is indifferent between the escalator and the stairs, and that the stairs will never get congested then slowing down the escalator could save electricity without making anyone worse off. Unfortunately, these special assumptions are not likely to be met. The stairs are likely to get more congested, so everyone who was using them already will be made worse off. At the same time, people with heavy bags who had a strict preference for the escalator will continue to use it and be made worse off by its reduced speed.


But these examples do not mean that Pareto improvements are impossible to find. Here are two real-world examples: Gas Royalties. The Federal government sells leases for offshore oil wells to the highest bidder. The oil company then pays a royalty per barrel of oil, usually about 16 percent. As the well gets old, the cost of extraction rises, and when the extraction cost exceeds the price of oil net of the royalty, the well shuts down. For example, if the price of oil is $40 per barrel, the royalty will be $6.40, leaving a net income to the company of $33.60 per barrel. Hence, if the extraction cost rises above $33.60, the well closes down. Closing down the well once the extraction cost reaches this level is inefficient. The value of the oil still exceeds the extraction cost, while the government gets nothing when the well closes. One solution might be to charge a royalty that declines once the extraction cost rises sufficiently. Unfortunately, this provides an incentive for the firm to exaggerate its cost, and the government would have to get involved in the difficult and expensive task of auditing the cost of each and every well. A better solution is to lower the royalty rate on old wells, because it is impossible for the firm to exaggerate the age of the well. Then, whenever oil can still be produced from an old well at a cost below its market value, the firm can earn some profits and the government still gets some royalties. Both the firm and the government are made better off by continuing to operate these older wells, which is a Pareto improvement. Tradable SO2 emissions permits. Beginning in the 1970s, the US Federal government introduced regulations designed to reduce emissions of sulfur dioxide (SO2), especially from power plants. These regulations typically took the form of maximum emission levels common for all plants that effectively required every power plant to install expensive scrubbers or to switch to low sulfur-coal. The problem was that the installation of scrubbers was more expensive for some plants than for others. At the same time, some plants were located closer to sources of low-sulfur coal, and therefore could use it at lower cost. Plants with the most expensive compliance costs were forced to close


down, while plants with low compliance costs, which could have sustained an even greater reduction in emissions, reduced their SO2 output only by the minimum required of all plants. In 1990, the government introduced a tradable permit system. Each plant was allocated permits to emit a certain quantity of SO2, but they were then allowed to buy or sell these permits. Plant with low abatement costs reduced their emissions by more than they needed to and sold their excess permits. Plants with high abatement costs reduced their emissions by less than their permits required, but then bought extra permits from other firms. Both high-cost and low-cost plants were made better off by the switch from uniform standards to tradable permits. The total quantity of permits issued left aggregate SO2 emission unchanged. High-cost plants paid less for a permit to emit an extra ton of SO2 than it would have cost them to reduce their emissions by a ton, and low-cost plants earned more by selling a permit to emit a ton of SO2 than it cost them to reduce their emissions by the same amount. As profits increased for both types of plants without causing a deterioration in environmental standards, the tradable permit policy turned out to be a Pareto improvement over the previous policy. It is useful to think about what sorts of changes are Pareto improvements and what are not by means of a diagram. Figure 1 plots a new curve called the utility possibility frontier for two individuals. It traces out the maximum amount of utility that can be earned by individual A given the amount of utility that is earned by individual B. Equivalently, it traces out the maximum utility that B can earn given any level utility earned by A. Imagine, for example, that we are trying to allocate slices of pizza between A and B. As we increase Bs utility by giving him more pizza, we must make As utility lower by taking pizza away from her. One might think of the utility possibility frontier like a budget line, only it is expressed in terms of utility rather than quantities of goods. Like a budget line, the utility possibility frontier always has a negative slope.


Figure 1 also plots some particular allocations. Consider first, allocation x, which lies inside the utility possibility frontier. Clearly, it is possible to make both A and B better off by changing to an allocation that lies on the frontier and to the northeast of x. That is, any allocation along the segment ab of the frontier will make at least one of the two better off without making the other worse off. It follows that x is Pareto inefficient, and moving from x to anywhere on the segment ab is a Pareto improvement. Point x is equivalent to leaving some pizza on the plate. We have a pizza large enough to increase consumption (and hence utility) of both A and B, but at allocation x we are not using it all.

Utility of B

Utility possibility frontier a

b y

Utility of A

FIGURE 1. Allocations and the utility possibility frontier

Now consider allocation y, which lies on the frontier. It is not feasible to move in a northeasterly direction because that would move us outside the frontier. The only way to make B better off while satisfying the constraints of the utility possibility frontier is to make A worse off. Similarly, A can be made


better off only by making B worse off. Consequently, allocation y is Pareto efficient. In fact, any allocation that lies on the utility possibility frontier must be Pareto efficient, because it would never be possible to make A better off without making B worse off. At the same time, any allocation that lies inside the frontier must be Pareto inefficient because it is always possible to make both better off by moving towards the frontier in a northeasterly direction. Note that, although there always exist some efficient allocations that are Pareto improvements over any inefficient allocation, moving from an inefficient allocation to any efficient allocation need not be Pareto improving. For example, moving from inefficient x to efficient y makes B worse off, so it cannot be a Pareto improvement. Compensation criteria. Pareto improvements are always desirable. But many times, the Pareto improvement criterion for policy provides no insights about whether a policy change is desirable. For example, as a society we may prefer efficient allocation y to inefficient allocation x, but moving from x to y is not a Pareto improvement. We may also prefer a more egalitarian allocation b to the allocation y, but moving from y to b is also not a Pareto improvement. Under such circumstances, we need a more discriminating criterion for deciding between competing allocations. Two versions of an ethical criterion to use when there are losers as well as gainers were introduced by Nicholas Kaldor (1939) and John Hicks (1939). Collectively they are known as the compensation criteria, and they are defined as follows: Kaldor criterion. Implement a policy if, with the policy change, it is possible to fully compensate the losers while still leaving the gainers better off. Hicks criterion. Implement a policy if the potential losers cannot compensate the potential winners for not having the policy change, while


still leaving themselves better off than they would be with the policy change. The criteria do not require that compensation actually be made; it only needs to be hypothetically possible to make the compensating payments. Figure 2 provides a graphical illustration of the compensation criteria. Consider a government thinking about whether to undertake a project that involves a movement in the allocation of utilities from x to y. This makes B worse off and A better off. The Kaldor criterion asks whether, after the project is implemented, A can compensate B for his losses, perhaps by making a cash payment, so that both A and B are better off with the project than they are without. That is, can we use the project to move us from x to y and then devise some compensation payment to attain an allocation of utility at a point such as z? The Hick criterion asks whether B could bribe A to forgo the project. That is, starting from x, can B offer a cash payment to A to that both A and B are better off without the project? This would require that a payment is made that allows A and B to attain an allocation such as w.

Utility of B

z x w y

Utility of A

FIGURE 2. Compensation criteria


The compensation criteria help us make decisions about projects and policies when the concept of Pareto improvement provides no guidance. Both criteria always rank a Pareto efficient distribution of utility over a Pareto inefficient distribution. Figure 3 illustrates a prospective policy that could move society from x to y. Allocation x lies inside the utility possibility frontier and is inefficient, while y lies on the frontier and so is efficient. After moving from x to y, the frontier tells us that we can always redistribute utility to get to z, where both are better off than at x. Thus, by the Kaldor criterion, a policy that moves us from x to y is desirable. If we stay at x, B cannot bribe A not to have the project, because this would require attaining a point such as w, which is not feasible. Thus, by the Hicks criterion, the policy is desirable.

Utility of B

z x w y

Utility of A

FIGURE 3. Compensation criteria rank efficient policies over inefficient policies

The compensation criteria help us because they justify moves from inefficient allocations to efficient allocations, even when they are not Pareto improvements. Unfortunately, there are situations when the compensation criteria do not help us. As Figure 4 illustrates, they cannot help us when both x and y are efficient. If the initial distribution is x and the post-policy distribution is y, 86

then we cannot attain a distribution such as z with the policy, and the Kaldor criterion fails. In contrast, we cannot attain a distribution such as w without the policy and the Hicks criterion therefore recommends a movement from x to y. Thus, the two criteria give different answers, and we cannot reliably justify a policy (or decided against a policy) that moves us from x to y. When two allocations are Pareto efficient, they are non-comparable by the compensation criteria.
Utility of B

z x w y

Utility of A

FIGURE 4. Compensation criteria cannot rank two efficient policies

The compensation criteria also cannot rank two Pareto inefficient allocations, as Figure 5 illustrates. In this case, if we begin at x, the Kaldor criterion indicates that we should move to y, because then a transfer could be made to attain z. But the Hicks criterion indicates that we should not move to y, because it would be possible to for A to bribe B, attaining w and leaving both better off. A similar problem arises if we consider a movement from y to x. When two allocations are Pareto inefficient, they are also non-comparable by the compensation criteria.


Utility of B

z x w y

Utility of A

FIGURE 5. Compensation criteria cannot rank two inefficient policies

So, in summary: A policy that yields a Pareto improvement is always desirable. A policy that moves us from any Pareto inefficient allocation to any Pareto efficient allocation is always desirable under the compensation criteria, The compensation criteria cannot compare two Pareto efficient or two Pareto inefficient allocations. Even when the compensation criteria work, compensation is only hypothetical. If we know that compensation is not actually going to be made, do we really want to make a move that makes people worse off? To address these last two concerns, we need to develop some tools to think about how as a society we might feel about inequality. This we turn to next.

2. Equity
I understand that some people in this country think otherwise, but Kobe Bryant and Shaquille ONeal seem pretty much the same to me. So if you ask me whether I prefer that Kobe earns $10 million a year and Shaq $5 million or 88

earning levels that are reversed, Ill tell you Im indifferent. If you ask me whether I would prefer that Bill Gates keeps his money while Joe Homeless gets $1,000 a year, or Joe Homeless and Bill Gates swap positions, Ill also be indifferent, unless of course I am Joe Homeless. So ask me a serious question. How much of Bill Gates money would I be willing to take away in order to give Joe Homeless extra income of, say, $1,000? Id like to help Joe, and Im willing to hurt Bill to do it. I would certainly be willing to take a $100 from Bill to give $1,000 to Joe. In fact, that would make me happier than the status quo. Id certainly be willing to take $1,000 from Bill and give it Joe. Id even be willing to take $10,000 from Bill if that meant Joe could have $1,000 more. Whats the maximum amount I would be willing to hurt Bill in order to make Joe better off? Im not sure, because I have not thought hard enough about it. However, whatever the amount is, it must be just that amount that leaves me indifferent between the new income distribution and the status quo. These various distributions are plotted in Figure 6, which compares utilities rather than income. The figure plots an indifference map. This map is similar in principle to the individuals indifference curves that we studied in chapter 2. However, whereas the indifference map in chapter 2 illustrated one persons preferences about the consumption of two goods, the map in Figure 6 illustrates my preferences about the relative wellbeing of Bill and Joe. If my preferences are a good reflection of what society as a whole prefers, then this indifference map illustrates societys attitudes to inequality. Such an indifference map is known as a social welfare function. The status quo is indicated with Bill earning lots of dough and Joe having very little. Taking a little from Bill to give a lot to Joe implies a movement from s to a. Society prefers a to s, so a is on a higher indifference curve. The maximum amount we would be willing to hurt Bill in order to give Joe the same increase in utility as moving from s to a is the amount that leaves society indifferent between s and the new allocation, indicated by b. Any further damage to Bill,


say by moving to c, is less preferable from societys perspective, so this puts us on a lower indifference curve.
Bills Utility

Status quo s a

b c

Joes Utility

FIGURE 6. The social welfare function

The indifference map that plots out the social welfare function says nothing about our ability to trade off Bills happiness for Joes happiness. To compare our preferences about inequality with our ability to make adjustments, we need to superimpose the social welfare function on our utility possibility frontier. Figure 7 does so, and illustrates that the best society can do is to select point d, where an indifference curve is tangential to the utility possibility frontier. That is, the most preferred allocation of utilities will always be a Pareto efficient allocation. Society is indifferent between a and d, but a is not attainable. Society can attain both d and e, both of which are Pareto efficient. but d is strictly preferred to e. Thus, once we know what the social welfare function looks like, we can rank efficient allocations that the compensation criteria cannot rank. Similarly, we can see that b is preferred to c, even though both are Pareto inefficient. Thus, the social welfare function also allows us to rank pairs of inefficient allocations.


The compensation criteria suggested that we should always choose a policy that moves us from an inefficient allocation to an efficient allocation. However, if the compensation is not actually paid, society may not prefer such a move. For example, a policy that moves us from b to f in Figure 6 would be recommended by the compensation criteria, but f is clearly inferior to b. This example teaches us that we must be careful about our use of compensation criteria. However, if we were to move to f by means of a policy change, the utility possibility frontier tells us that we could subsequently do better by changing the allocation of utility from f to d.

Bills Utility

s a d b e c f Joes Utility

FIGURE 7. Socially optimal choices

The shape of the social welfare function. Figures 6 and 7 draw a social welfare function in which the indifference curves become flatter as we move from left to right. The reason for such a shape is intuitive. When Bill is rich and Joe is poor, we are willing to take a substantial sum from Bill to make Joe a little better off. Hence the curves are very steep at the left. But when Joe is already rich and Bill poor, the situation is reversed, and the indifference curves become very flat. Although this is the general shape that we expect the social welfare function to have, we are in a more difficult position if we want to discover the exact


shape and location of the indifference curves for a particular problem. They are not directly observable, so we cannot just go out and measure them. Perhaps we could go out and ask people what their preferences are, add everyones preferences up in an appropriate way, and thereby construct the social welfare function from these survey data. There are a number of problems with this. The first is that it would be exorbitantly expensive. The second is a more profound conceptual problem. People would state very different preferences about the distribution of utility if they are one of the people being affected. For example, if I were Joe Homeless, I would be strongly in favor of a massive redistribution of wealth away from Bill Gates and toward me. If I were neither Bill nor Joe, I would undoubtedly favor a more modest redistribution. Which of these stated preferences about equity is the correct one? The philosopher John Rawls (1971) has suggested that neither is correct. He argued that our preferences should be elicited under what he termed a veil of ignorance. We should ask people to imagine a situation in which they know that they could become Bill Gates or that they could become Joe Homeless. Then, given that they could be either one of these people, what sort of distribution would they prefer? Of course, people are not behind this veil of ignorance and so Rawls suggestion does not really get us anywhere. Rawls was aware of this, so he went one step further. He ventured the answer that people would give if they were truly behind a veil of ignorance. He argued that people in this situation would favor a distribution scheme that leaves the worst off person as well off as possible. Obviously, few governments have implemented policies based on Rawls ideas. Perhaps another scheme to discover the social welfare function is to allow people to vote on alternative polices and thereby reveal societys preferences through the ballot box. Unfortunately, democracy does not work well either. First, voting does not avoid the self-interest problem that Rawls was grappling with. But voting introduces yet another problem. Consider a vote for three


different tax policies A, B, and C, each of which impose different rates of income taxes on the poor, the middle class, and the rich. The population is evenly divided between the three income levels. Policies are offered in pairs, and the voting rule is to choose the policy that the majority prefers. Table 1 summarizes the options:

TABLE 1. Alternative tax proposals A Poor Middle class Rich 25% 31% 33% B 27% 27% 36% C 30% 30% 30%

The poor prefer A because they pay less in taxes. Similarly, the middle class prefers B while the rich prefer C. A majority (poor and middle class) prefers B to C, a majority (poor and rich) prefer A to B, and a majority (middle class and rich) prefers C to A. So, if we were to start with any policy, a majority would prefer to change policy. For example, if the status quo was policy A and policy C were offered as an alternative, the middle class and the rich would vote for a switch to C. Once at C, however, a majority would now vote for a switch to B. Finally, once we implement policy B, the poor and rich would vote for a switch to A! Our voting scheme creates an endless cycle of votes indicating a preference to switch. Is this outcome a peculiarity of the voting scheme adopted? The answer, unfortunately, is no. In an astounding paper published over 50 years ago, Nobel laureate Kenneth Arrow (1951) proved what has become known as the Arrow impossibility theorem. He started out with a short list of five desirable


properties of any voting scheme.18 Arrow then proved that there is no voting scheme including voting schemes yet to be invented! that satisfies these desirable properties. We do not have the math tools necessary to prove Arrows remarkable theorem. Table 1 gave an instance of the implications of the theorem: we devised a voting scheme and found an example in which it fails to produce a preferred outcome. Arrows theorem tells us that any voting scheme you care to think of will also fail at some point.

3. Concluding comments
Economists are enamored of the concept of Pareto efficiency for good reason. If the current situation is not Pareto efficient, we know immediately that we can make at least one person better off without making anyone else worse off. Obviously, this is a desirable thing to accomplish, so the first task for economic policymakers is to keep a sharp lookout for Pareto inefficient situations, and to devise a Pareto improving policy whenever one is found. The economists task becomes much harder when Pareto improvements are not available. To tackle these issues, economists developed, first, the concept of compensation criteria and, second, the more satisfactory concept of the social welfare function. We analyzed these with the help of the utility possibility frontier.


They are: 1) (Unanimity): if everyone prefers option A to option B, then the voting scheme should rank A above B. 2) (non-dictatorship): the outcome is not completely determined by the preferences of one individual, 3) (Transitivity): if A is preferred to B and B is preferred to C, then A should also be preferred to C. 4) (Independence of irrelevant alternatives): if A is preferred to B, and B is preferred to C, then A should be preferred to B whether or not option C is available. 5) (Completeness): the voting scheme should be able to compare all possible alternatives.


Because we cannot measure them, the utility possibility frontier and the social welfare function are more useful as a conceptual framework than they are as an empirical tool. Nonetheless there are some useful lessons to be learned from them. One is positive: when a policy changes a Pareto inefficient situation to a Pareto efficient situation, this is always good for social welfare as long as compensation is made to the losers. The other lessons speak more to the limitations of economic analysis. When the current situation is already Pareto efficient, or when compensation is not paid to the losers, we cannot say if a proposed policy is desirable without knowing what the social welfare function looks like. However, there is no foolproof way for economists or anyone else to identify societys preferences about different distributions. While economics can still tell us in these situations what the consequences of different policy choices will be, it cannot reliably inform us which of the choices is the best. Because selecting among different policies on equity grounds is fraught with so much difficulty, economists sensibly restrict their focus to the efficiency consequences of policy. If the status quo is inefficient and an efficient policy can be identified, we know that it is in principle possible to fully compensate the losers from the policy change and still leave those who gain better off. Whether society actually makes that compensation is a matter for the political process to decide. The Arrow impossibility theorem, of course, tells us that the political process is no more likely to choose the best policy than economists are, but wed rather that politicians be blamed for bad decisions. It is well known that the vast majority of economists are in favor of free trade. Critics contend that economists are obsessed with money, and that they have no empathy for the many people who may lose their jobs and their livelihood from allowing unfettered imports. These critics are revealing their ignorance of economists and economics. Economists generally favor unrestricted free trade because, as we will see later, it makes the economic pie as large as possible. That means that it is possible for those who gain from free trade to


fully compensate those who lose, and thereby leave everyone at least as well off as before. Economists are fully aware that if compensation is not paid, then we can no longer say that the move to free trade is an improvement. But economists dont get to decide how the economic pie is divided. Politicians choose not to make the compensation payments and the public vote against the tax increases necessary to finance them. Surely they are the ones to blame when people are made worse off by policy changes that make the economic pie larger.

Concepts introduced in this chapter

Pareto efficiency efficiency utility possibility frontier Arrow impossibility theorem Pareto improvement equity social welfare function Rawls' veil of ignorance

Hicks compensation criterion Kaldor compensation criterion

1. You have 8 units of a good to allocate between two individuals, A and B. Their marginal utilities are as follows:
Units of Good 1 2 3 4 5 6 7 8 MARGINAL UTILITIES Person A Person B 10 15 10 10 8 7 7 5 5 1 3 1 2 0 1 0

(a) For each of the following pairs of utilities, indicate whether it is efficient, inefficient, or infeasible.


A = 40, B= 37; A = 35, B = 32; A = 28, B = 37; A = 10, B = 39 (b) If social preferences are Rawlsian (that is, society prefers to make the worst off person as well off as possible), what should the allocation be? (c) If society does not care about equity, and is only concerned with maximizing the sum of all utilities, what should the allocation be? 2. Table 1 in the main text showed a set of tax proposals that generated cycles in voting. Assess whether the following sets of proposals would also generate cycles.
(a) Alternative tax proposals A Poor Middle class Rich 25% 27% 33% B 27% 28% 36% C 30% 30% 30%

(b) Alternative tax proposals A Poor Middle class Rich 17% 20% 28% B 18% 22% 24% C 16% 24% 26%

Arrow, Kenneth J. (1951): Social Choice and Individual Values. PhD Dissertation, Columbia University. Kaldor, Nicholas (1939): Welfare propositions in economics, The Economic Journal, 49:294-300. Hicks, John (1939): The foundations of welfare economics, The Economic Journal, 49:696-712.

Rawls, John (1971): A Theory of Justice. Cambridge, MA: Belknap.


Chapter 4 Supply and Demand

This chapter introduces some of the basic meat of the economic analysis of markets. In Section 1, we will see how a few small logical steps allows us adapt the concepts we learned in our study of Vickrey auctions for the study of other types of markets. In section 2 we see how changes in the economic environment cause prices and quantities bought and sold to change. Section 3 provides the tools we need to put numbers on our analysis -- to produce dollar values of the consequences of different policies. In Chapter 5 we will use these policy tools to analyze a variety of policy interventions.

1. From Vickrey auctions to demand and supply

Figure 1 is adapted from our study of Vickrey markets in Chapter 1. In this example, there were seven potential buyers and one seller. The seller had five units of the good to dispose of. We concluded that a Vickrey auction would entail all five units being sold at a price of $4. Our analysis also yielded some additional results: Everyone whose valuation exceeded the price would buy the good, while no one whose valuation was less than the price would do so. We measure the welfare of each buyer as the difference between their valuation of the good and the price they pay multiplied by the quantity each person buys (consumer surplus). We measure the welfare of the seller as the difference between the price received and the sellers cost multiplied by the quantity sold (producer surplus). We found that the equilibrium price arrived at through the Vickrey auction maximizes the sum of consumer surplus and producer surplus. 98

16 14 12 10 8 6 4

consumer surplus

p =3

Seller revenues (to be split between costs and profit

2 0 Mara John John Abu Fred Juan Eli

FIGURE 1. Consumer surplus and seller revenues in a Vickrey auction (replication of Figure 2 in Chapter 1).

Let us translate these ideas to the types of markets you see around you every day. Most markets are not auctions, of course. You go to a store and see an item for sale at a predetermined price. You decide whether or not you want to buy it. You will buy it only if your valuation of the item exceeds its price. Such markets are known as posted-offer markets. Although they appear to be very different from Vickrey auctions, posted-offer markets in fact have many similar properties. The demand curve. We begin with the consumers. Imagine that instead of seven potential buyers there are many of them. We again order each potential buyers valuation for a unit of the good in descending order from left to right. Remember, that the same consumer will appear more than once if he is potentially interested in more than one unit of the good. With many potential buyers, we would get a graph like that in Figure 2. The histogram bars plot each individuals valuation, or willingness to pay for a unit of the good. The line we have drawn is an approximation (when there are many potential buyers, we can always approximate the histogram with a line. There is, of


35 30 25
Valuations ($)

20 15 10 5 0 0 10 20




FIGURE 2. Consumer willingness to pay with many potential consumers.

course, no reason why the line should be a straight line, but to make our life easier, we will usually assume that the line is straight). Now, recall that everyone whose valuation exceeds the price they must pay will choose to buy the good, and everyone whose valuation is less than the price will not buy. For example, if the price everyone must pay is $20, the graph tells us that consumers will want to purchase a total of 23 units of the good. If the price falls to $10, then consumers will want to buy a total of 48 units. Consequently, the line traces out the quantity of the good demanded by consumers at each possible market price. This is the demand curve.19 Because we have ordered the potential buyers in decreasing order of their valuations, the demand curve has a negative slope. This negative slope tells us that the lower the market price, the greater the quantity that is demanded.


And we will call it a curve even though we will usually draw it as a straight line.


35 30 25
Valuations ($)

height = 10

20 15 10 5 0 0 10 20

base = 23




FIGURE 3. Demand and consumer surplus.

Let us now do away with the histogram. Imagine that the market price is indeed equal to $20, so that the quantity demanded is 23 units. Then, we also know that the triangular area under the demand curve and above the price line, shown by the shaded area in Figure 3, is consumer surplus. This triangular area corresponds to the sum of the difference between consumer valuations and the price paid for all consumers that actually buy the good. We can easily measure how much this consumer surplus is in dollars. The formula for the area of a triangle is

The base of the triangle is 23, which is the quantity demanded at the price of $20. The height of the triangle is 10: it is the difference between the price at which the demand curve intersects the y axis (which is the price at which the quantity demanded would be exactly zero) and the market price. Using the formula for the area of this triangle, we obtain



= $115. We therefore conclude that if the market price is $20, consumer welfare, as measured by the total consumer surplus, will be $115. Recall that if the price drops to $10, the quantity demanded rises to 48 units. You should be able to verify for yourself that consumer surplus will then rise to $480. The supply curve. Demand is only half the story. We also need to know how much producers are willing to sell at each price. Consider a market in which there are many potential producers. Each producer can manufacture a fixed quantity of the good, this quantity being determined by the production capacity of the firm. Some firms have a large capacity and are able to produce more than others. Each producer also has a cost per unit of output. If the market price is $20, every firm that can produce the good at a cost less than $20 per unit will choose to produce as much as its capacity allows. The difference between its unit cost of production and the market price is the profit it makes on each unit. Multiply that by the firms capacity, and you get the firms total profits. Other firms will calculate that their cost of production exceeds the market price. They conclude that participating in this market would cause them to lose money on every unit, so they decide to produce nothing. Figure 4 illustrates these ideas graphically. The graph orders firms according to their unit cost of production, with the lowest-cost producers at the left. The width of each bar is determined by the capacity of that producer, but for simplicity I have assumed in this graph that each firm can produce just one unit. For each producer, the vertical distance between their unit cost and the market price is their profit per unit (it is negative if their cost exceeds the market price). The firms total profits equal this vertical distance, multiplied by the width of the bar. If we add up the profits of all the firm that choose to


35 30
Unit Production Costs ($)

25 20 15 10 5 0 0 10 20




FIGURE 4. Producer costs and willingness to produce with many potential producers.

produce (i.e. those with costs less than or equal to the market price), we get total profits earned by firms operating in this market. This is a measure of producer welfare, that we call producer surplus. In Figure 5, we have again approximated the histogram with a straight line, which we call the supply curve. The supply curve traces out the quantity that firms are willing to produce for the market at each possible market price. Once we have the market price, the shaded triangular area (producer surplus), can again be measured using the formula for the area of a triangle. We have drawn the supply curve intersecting the y axis at $5. The supply curve also indicates that firms are willing to supply a total of 23 units of the good at a price of $20. If we assume again that the market price is $20, the producer surplus is readily calculated as

= $172.50.


35 30 25

base = 23

15 10 5 0 0 10 20

height = 15




FIGURE 5. Supply curve and producer surplus.

Market equilibrium. We say that the market is in equilibrium when three conditions are satisfied: (Utility maximization): At the market price, each consumer is voluntarily buying his preferred quantity. This is equivalent to saying that the quantity-price pair lies on the demand curve. (Profit maximization): At the market price, each firm is voluntarily producing its profit-maximizing quantity. This is equivalent to saying that the quantity-price pair lies on the supply curve. (Market clearing): At the market price the quantity demanded is equal to the quantity supplied. Figure 6 illustrates the market equilibrium. The only way all three conditions can be satisfied is for the market equilibrium to be at the point of intersection of the supply and demand curves. Thus, given consumers preferences and firm costs, the equilibrium is at a price of $20, at which 23 units of the good are traded. Total economic welfare from this market is the sum of consumer and



30 25 20 Consumer Surplus Producer Surplus

Supply b a

Equilibrium 15 price 10 5 0 0



20 Equilibrium quantity




FIGURE 6. Market equilibrium.

producer surplus. These areas have already been calculated, and economic welfare totals $115 + $172.50 = $287.50. Just as we saw for the Vickrey auction, it is easy to verify that there is no other market outcome that can increase economic welfare. Forget about price for the moment, and imagine that somehow we were able to increase the quantity above 23 units. The cost of producing the 24th unit exceeds the benefit to consumers to having an additional unit to consume. This is evident from the fact that at the 24th unit the maximum willingness to pay indicated by point a on the demand curve, is less than the cost of producing the 24th unit, which is indicated by point b on the supply curve. Put another way, the consumer is only willing to pay an amount for the 24th unit that is less than the cost of producing it. The same argument applies for the 25th unit, the 26th, and so on. In fact, forcing the quantity traded to exceed the market equilibrium quantity


induces a welfare loss, equal to the shaded triangular area given in panel (a) of Figure 7. Now imagine that it were possible to reduce the quantity traded to some level below the equilibrium quantity. Then there are some units for which consumer valuations exceed production costs, but the units are not produced. Restricting quantity in this way then implies that we forego some of the welfare benefits of production and consumption. This loss of welfare is indicated by the shaded area in panel (b) of Figure 7. Whether we raise or lower the quantity, it does not matter: any movement away from the equilibrium quantity reduces economic welfare. It is in this precise sense that economists are fond of saying that well-functioning markets are efficient.20 Finding the right price. The mechanism by which the equilibrium price is arrived at in an auction is easy to see. In a Vickrey auction, for example, everyone tells the auctioneer their true valuation, and then all those with the highest valuations pay the price necessary to clear the market. It is transparent how an auctioneer may tabulate the bids and then select the price that enables her to sell exactly the number of units that are available. There is no auctioneer in a posted offer market. So how, in practice is this magic equilibrium price arrived at? Imagine you are a firm deciding whether or not to produce a good. You know your own production costs. But to decide whether or not to produce, you need to know what the market price will be. This is easy to do if you are contemplating beginning production in an existing market, because you can simply observe the price. But if you are thinking about a new market, there is no price to observe. In this case, you have to


At a later stage we will explore in depth the many ways in which markets may not function well.


Panel A

30 25 20 15 10 5 0 0

Forcing quantity to exceed equilibrium induces a welfare loss because in this area production costs exceed valuations.








Panel B

30 25 20 15 10 5 0 0

Forcing quantity to be less than the equilibrium induces a welfare loss because in this area valuations exceed production costs.








FIGURE 7. Forcing quantity to exceed (panel A) or be less than (panel B) the equilibrium quantity induces an economic loss.


think about what the market price is likely to be. And to do this you need to have information about the demand curve and production costs of other potential producers. As you might expect, it is easy to make mistakes. Imagine, for example, that all potential producers expect market demand to be greater than it actually is. Then, too many firms will decide to produce, thinking that it will be easy to sell what they manufacture. But quickly they will discover their mistake. Firms will find that the only price they can sell at is lower than they had expected, and some firms those whose production costs are too high will give up. In response, the quantity supplied will fall. On the other side of the same coin, if firms underestimate the extent of demand, too few firms will enter. There will be a shortage of goods and the price will be high. Then firms that had earlier decided not to enter the market will revise their plans. The quantity supplied will expand and shortages will vanish. Through a process of firms adjusting their production plans, we see that the market will move toward equilibrium. If there is too much production, price will be low. Some firms will drop out of the market and price rises toward the equilibrium. If there is too little production, supply will increase and price will fall toward to its equilibrium value. The market equilibrium that we drew in Figure 6 is what will happen when most firms have got their decisions about right.

2. The location and slopes of demand and supply curves

The location of the demand curve for a good depends entirely upon individuals valuations of the good. The location of the supply curve depends upon firms production costs and production capacity. It follows that if consumer valuations change, then the demand curve must move. If costs or capacities change, then the supply curve must shift. Shifts in the demand curve. There are several reasons why consumer valuations might change:


Changes in preferences. As consumers, we are fickle. One year ipods are all the rage, the next you cant give them away. A colleague of mine tells me, on good authority, that this year flat shoes and vintage jewelry are all the rage among the fashion conscious. Im sure it will be heels and space-age adornments next year.
Price d D

Stronger preferences for the good shifts the demand curve upwards, raising both the equilibrium price and quantity traded.

Quantity FIGURE 8. Stronger preferences for a good shift the demand curve upwards.

When consumer preference for a good becomes stronger, we can assume that consumers valuations increase. Put another way, the number of consumers willing to buy at a certain price will increase when valuations rise. This is captured graphically by a shift of the demand curve upwards and to the right (i.e. in a northeast direction). Figure 8 illustrates. An increase in consumers willingness to pay shifts the demand curve from the line dd to DD. The equilibrium price rises, and firms are thereby induced to increase production. You should have no trouble figuring out what happens to price and quantity demanded when preferences for a good become weaker. 109

Economists take great care to distinguish between changes in demand and changes in quantity demanded. When economists say the demand for a good has increased (or decreased), what they mean is the demand curve has changed location. An increase in demand is an upward shift of the curve, while a reduction in demand is a downward shift of the curve. But when economists say that the quantity demanded has changed, what they mean is that equilibrium quantity has changed. A change in the quantity demanded could be caused by a shift in the demand curve (i.e. by a change in demand), or it could be the result of a movement to a new location on the same demand curve (perhaps as a result of a shift in the supply curve). Changes in income. A rise in income also affects the location of the demand curve. You might imagine that a consumers willingness to pay for a good will inevitably rise whenever her income increases, and hence that an increase in income will shift the demand curve upwards. However, this is not necessarily the case. As people become more wealthy, they do indeed increase their consumption of many items, but they also reduce their consumption of others. Rich people are more likely to eat steak and less likely to eat hamburger than poorer people. We would then predict that, beyond a certain level of income, any further rise in income will cause the demand curve for steak to shift up, and the demand curve for hamburger to shift down. If a rise in income causes an increase in demand for a good, the good is known as a normal good. If a rise in income causes a reduction in the demand for a good, it is referred to as an inferior good. Table 1 reports the effect of a change in income on demand for a variety of goods. Specifically the table provides the percentage change in quantity demanded caused by a one percent increase in income, under the assumption that the market price is held constant. Thus, a one percent increase in income induces only a 0.14 percent increase in the quantity of food demanded, but a much larger (5.8 percent) increase in the quantity of airline travel demanded. Note that all the entries in the table are normal goods. The entries also suggests that the basic necessities of life tend to have lower income elasticities of demand than do luxuries. It is easy for poor 110

people to avoid going to the movies, but not to avoid food. Thus, when peoples incomes rise, the demand for movies changes much more than the demand for food.

TABLE 1. Effect of a 1% Income Changes on Demand Food Telephone Newspapers Clothing Furniture Alcohol +0.14% +0.32% +0.38% +0.51% +0.53% +0.62% Cars Haircuts Restaurant meals Foreign travel Movies Airlines +1.07% +1.36% +1.61% +3.08% +3.41% +5.82%

Source: Bade, R., and M. Parkin (2004): Foundations of Microeconomics, p. 131.

The numbers in Table 1 are termed income elasticities of demand. Figure 9 illustrates precisely what it is that the numbers mean. A one percent increase in average consumer income shifts the demand curve upwards (for a normal good), from dd to DD. Holding price constant, this would induce an increase in the quantity demanded of the amount x. The income elasticity of demand is this number x, expressed as a percentage of the original equilibrium quantity, q. Income elasticities of demand are useful summary numbers for firms to know about. Each year, as technology advances, incomes rise in the United States by an average of about two percent (sometimes a little more, sometimes less). Knowing this basic fact, and knowing the income elasticity of demand allows a firm to forecast the progression of demand over time. A firm manufacturing food, for example, may reasonably predict that increases in the demand for food will be very small. Because there is little prospect for growth dues to rising incomes, food companies may decide that, in order to grow, they will


Price d

1% increase in income shifts demand curve up

Which induces an increase in x in quantity demanded when price is held constant.

d q

D Quantity

FIGURE 9. The income elasticity of demand is equal to x/q.

need to diversify into other activities. In contrast, airline companies might reasonably predict increases in the amount of airline travel at the rate of up to 10.64 percent per year. They can plan on growing by buying more planes to meet rising demand. Airport authorities similarly can project significantly growing demand for travel, and plan airport expansions accordingly. When should the airport plan to build that extra terminal? Income projections, in conjunction with the income elasticity of demand, will tell them. You may be wondering where exactly you might get this handy number. After all, in Figure 9 you can observe p and q they are simply todays price and quantity. But how do you find x? The answer, I am sorry to say, is through the use of statistical analysis. Specialists in this line of work collect many months or years of data on prices and quantities, along with additional information about incomes, and then use various statistical techniques to estimate the size of x. It is not particularly difficult to do, but we are not going to try and do it


in this course. But if you want to get in on this game of economics, at some point in your college career, a course or two in statistics is in your future. Changes in the price of related goods. I have it on good authority that many Americans are rather fond of peanut butter and jelly sandwiches. So what do you think happens to the demand for peanut butter when jelly goes up? First of all, we know that the quantity of jelly demanded goes down (simply because the demand curve for jelly, like all demand curves, has a negative slope). And because people like jelly with their peanut butter, they now value peanut butter less. Thus, the demand for peanut butter goes down the demand curve shifts downward. Note the distinctive use of the terms quantity demanded goes down and demand goes down. A rise in the price of jelly reduces the quantity of jelly demanded (the demand curve does not shift) and reduces the demand for peanut butter (the demand curve does shift). In contrast, what do you think happens to the demand for chicken when the price of beef goes up? A rise in the price of beef reduces the quantity of beef demanded. People who now decide not to eat beef must eat something else, so they turn to (among other things) chicken. Thus the demand for chicken rises in response to a rise in the price of beef its demand curve shifts upwards. If an increase in the price of good A induces an increase in the demand for good B, we say that A and B are substitutes. If an increase in the price of good A induces a reduction in the demand for good B, we say that A and B are complements. Thus, peanut butter and jelly are complements, while chicken and beef are substitutes. Most goods are neither substitutes nor complements. For example, an increase in the price of socks should have little measurable effect on the demand for, say, gasoline. Even among those pairs of goods that do seem to be related, either as substitutes or as complements, for many of them the strength of the relationship is likely to be small. The trick for a smart manager is to identify goods that are strong complements or substitutes for the good the firm is producing. In order to assess better the prospects for demand changes in his 113

own industry, a manager would do well to keep track of changes in the prices of these strongly related goods. There is a nice symmetry between goods that are substitutes or complements. If A is a substitute for B, then B is a substitute for A. Put another way, if we already know that an increase in the price of A increases the demand for B, then we also know that an increase in the price of B increases the demand for A. Similarly, if A and B are complements, then an increase in the price of either one of them will induce a reduction in demand for the other. Shifts in the supply curve. The two main reasons the supply curve shifts are from changes in the costs of production within existing manufacturing facilities, and changes in the distribution of production costs across firms caused by expansions and contractions of capacity. Changes in production costs. Production costs may rise or fall because of changes in the costs of inputs used in the production process. If silicon chips decline in price, then the cost of producing computers declines. Firms with reduced costs will be willing to sell all they are able to produce at a lower price (although they naturally hope they wont have to!). This reduction in the price necessary to justify production is manifested as a downward shift in the supply curve. Similarly, increases in the price of inputs shifts the supply curve upwards. Technological change is another major source of changes in production costs. Many firms engage in research and other forms of learning to discover better ways to produce goods and services. Firms may discover that the introduction of a new machine, or a rearrangement of the organization of the factory, reduces production costs even though the prices of the inputs have not changed. Technological advance is captured by a downward shift in the supply curve. Figure 10 illustrates the effect of a downward shift in the supply curve, from ss to SS. The equilibrium price falls while the equilibrium quantity rises. Contrast


Price s S

s S q Quantity

FIGURE 10. The effect of a reduction in production costs.

this with shifts in the demand curve that cause a fall in price. For demand shifts to reduce price, the demand curve must shift downwards. But when this happens, the equilibrium quantity traded also declines. Consequently if market data tell you price and quantity are moving in the same direction (i.e. both falling or both rising), you can infer that something has caused the demand curve to shift. You can then focus your effort on understanding why the demand curve has shifted is it a change in income, consumer preferences, or the prices of related goods? In contrast, if price and quantity are moving in opposite directions, you can infer that that something has caused the supply curve to shift. You can then focus your efforts on understanding what has made costs of production change. Changes in capacity. Imagine there are 100 firms with production costs of $1 per unit and each with a production capacity of one million units per year. There are another 100 firms, with inferior technology, whose production costs are $2 per unit, each with a capacity of one million units per year. At a price of



2.00 1.50 1.00

a s

100 120



Quantity (millions)

FIGURE 11. Supply curve changes with capacity adjustments.

$2, the consumers demand a total of 120 million units per year. At a price of $1, they demand 150 million units per year. What is the market equilibrium? Figure 11 provides the answer. There is a total capacity of 100 million units that can be produced if the price is above $1 but less than $2. Capacity rises to 200 million units if the price rises to $2. The corresponding supply curve is illustrated by the step function ss (because the step is so large, it does not make sense in this case to try and approximate the whole supply curve with a single straight line). The demand curve, however, is approximated by a straight line, one that we know passes through the two points (p=$1, q=150 million) and (p=$2, q=120 million). Market equilibrium is where supply intersects demand, and this turns out to be at a price of $2. Hence, production and consumption are equal to $120 million units per year. All firms with production costs of $1 produce at capacity. What happens to firms with production costs of $2? There is capacity of another 100 million units, but only demand for an additional 20 million. The answer must be that 116

only some of these firms get to produce. Which ones? The analysis does not tell us. Perhaps it is a matter of chance whoever gets in the market first. But now imagine the thought processes of the low-cost producers. They are making $1 profit on each unit they sell. They might well think to themselves, why, I can build another factory and increase my profits. So they go ahead and start building. After, say, a year, their new factory is up and running. Perhaps the new factory is not in quite as good a location as their first, so their production costs are higher, say $1.50. Nonetheless, if enough firms invest to expand capacity we will see a new supply curve that includes the segment aa. The equilibrium price drops to $1.50, and the high cost producers are forced out of business. This little story illustrates how the shape of the supply curve may differ depending on the time horizon we consider. The supply curve ss summarizes how production will respond to changes in price if there is not sufficient time for firms to invest in and build new factories. The modified supply curve saa summarizes how production might respond to price changes once sufficient time has elapsed to allow low-cost producers to expand their capacity. The story also provides a warning about a disastrous error often made by entrepreneurs starting up new businesses. Imagine the current market price is $2. You are considering entering with your new patented technology, and you know that your production costs are $1.75. It will costs you $500,000 to build a factor with a capacity of one million units per year. You conclude that you can make $250,000 profit per year, so that it will take you two years to recover your construction costs. After that, its all profit! So off you go. But at the same time, low-cost producers are spotting the same opportunities and building their own factories. By the time you get in the business, low-cost firms have expanded their capacity and the price has dropped to $1.50. You are immediately out of business, and you have lost the $500,000 you invested in the new factory. This is a very common mistake, and it largely explains why


one third of new business in the United States go bankrupt in the first five years. When firms with different costs can change their capacities, the current market price is unlikely to reflect the future price. To understand whether you can make profits and recover your investment, the current market price is only the starting point. You also have to think about what the production costs of other firms, yet to enter the industry or yet to expand their capacity, will be. This can, of course, be very hard to do. It requires that you be an expert on how your production technology compares with the production technologies of other firms.
300 250 6 Number of Firms Price (index) 200 150 100 50 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 5 4 3 2 1 0 1900 1910 1920 1930 1940 1950 1960 1970 1980 8 7

FIGURE 12. Number of Firms and Prices in the US Automobile Tire Industry. Source: B. Jovanovic and G. McDonald (1994): The lifecycle of a competitive industry. Journal of Political Economy, 102(2):327-347, figures 1 and 3.

Figure 12, which plots the number of firms and prices in the US automobile tire industry, illustrates a typical instance of this type of entry mistake. On the right hand side, we see that price is dropping dramatically in the early years of the industry. On the left, we see an equally dramatic rise in the number of firms up to 1922. But after 1922, there is a massive turnaround in the number firms, which are going out of business at an astounding rate. In fact, over two118

thirds of the firms that were in business in 1922 had gone by 1931. This decline in the number firms occurred during a period of especially rapid decline in prices, and (not shown) an equally rapid rise in the quantity of production. As output and prices are moving in opposite directions, we can infer that these changes were being caused by downward shifts in the supply curve. This was being caused by a massive expansion of capacity by the lowestcost producers: especially Goodrich and Firestone.

Chickens and Eggs. The article on the next page talks about circumstances that
brought about high prices for poultry in 2004, including increase in the price of beef, reductions in the price of inputs, and an upward shift in demand. It also talks about excess firm expansion in response to these changes in circumstances. In one market in one year, pretty much all of the factors affecting demand and supply that we have just talked about come into play. The slopes of supply and demand. Some demand curves are steep, others are much flatter. A steep demand curve implies that when prices change, the quantity demanded changes very little. This might be because there are few opportunities for consumers to change their consumption in the face of changing prices. Perhaps the good is an essential commodity, such as water. Perhaps there are few substitutes, such as with gasoline. Or perhaps consumers are addicted, such as with tobacco. In each of these cases, consumers have few options except to pay up when the price increases. A flatter demand curve can be expected when the good is a frivolous luxury, or when consumers can easily switch to substitute products. When a demand curve is very steep, economists say the demand is price inelastic. When it is very flat, we say that demand is price elastic. More formally, the price elasticity of demand is defined as the percentage change in demand induced by a one percent increase in the price of the good. It is, naturally, a negative number. If the price elasticity of demand is between 0% and 1%, we say demand is inelastic. If it is less than 1%, we say demand is


The Poultry Boom

It's never a good time to be a chicken, but now is really awful. Daniel Gross Posted Wednesday, Jan. 26, 2005, at 2:31 PM PT Food inflation is taking its toll everywhere from high-end restaurants to megamarketers like Sara Lee. Higher prices for essential inputs like milk and beef (along with the weak dollar) are chopping profits and tamping down growth in consumer demand. But there's one sector of the food world where the sky isn't falling: the chicken economy. The companies that buy, process, and peddle chicken parts are doing quite well. While many foods are getting more expensive, chicken seems to be getting cheaper. And while food fads come and go, chicken consumption has been rising consistently. Chicken businesses are clucking about their stellar performance. Buffalo Wild Wings, a 300-store wings chain (based, naturally, in Minneapolis), pre-announced higher earnings on Jan. 10. Earnings for the quarter were projected at as much as 50 percent above expectations because sales rose smartly and because "chicken wing prices average $1.30 per poundlower than the originally anticipated $1.40 per pound." On Monday, Pilgrim's Pride, the selfproclaimed "undisputed number-two poultry producer in America," likewise announced an excellent quarter. Each week, Pilgrim's Pride turns about 30 million chickens into nuggets, wings, drumsticks, and sundry other parts. CEO O.B. Goolsby cited "strong performance in our prepared foods and fresh chicken businesses, lower feed ingredient costs, rising exports and strong demand in the foodservice sector." Next week, the undisputed champion poultry producer, Tyson, will report earnings. Tyson diversified from white meat and dark meat into red meat by buying huge meatpacker IBP in 2001. But beef, the company's biggest unit by sales, is struggling with higher prices and apparently lower demand. Citing unfavorable operating margins, Tyson earlier this month temporary closed four beef plants. The company's most recent earnings release shows just where chicken stands in the pecking order. Chicken revenues rose 20 percent from the year before, while beef revenues fell. In the 12 months ended Oct. 2, 2004, chicken accounted for less than a third of sales but nearly 60 percent of total operating income. Beef, with 45 percent of total sales, contributed less than 14 percent of operating income. Some of chicken's unlikely flight can be explained by simple economics. As beef prices remain high, chicken seems a relative bargain to consumers. And with U.S. farmers reaping bumper corn and soybean crops, the price for the primary input into chickens (feed) has been falling. But it can also be attributed to diet trends. Chicken has increasingly become what's for dinner. According to the U.S. Poultry and Egg Association, U.S. chicken consumption per capita has risen from 68.8 pounds in 1995 to an estimated 85.6 pounds in 2004, up 24.4 percent. Meanwhile, turkey consumption has remained stagnant. And for all the hullabaloo surrounding the Atkins diet, beef consumption simply hasn't kept up. According to the U.S. Department of Agriculture, per capita beef consumption rose from a low of 64.4 pounds in 1993 to 67.5


pounds in 2002. And according to the Cattle Beef Board, beef consumption has fallen since 2002. Chicken's rise may have something to do with its malleability, relative healthfulness, and appeal to kids. (Beef nuggets, anyone?) Chicken, unlike other meats, tastes good when dipped into batter and dropped into vats of really hot oil. For burger joints like McDonald's and Burger King, fried chicken partswings, nuggets, strips, fillets, you name ithave become a significant business. KFC has no use for beefyou can't fry it. Chicken may be benefiting from the fading appeal of the Atkins diet. Indeed, the sector of the chicken economy that is the most Atkinsfriendly is struggling. A company with the unlikely name of Cal-Maine is the largest egg

producer in the country. Last year, it produced 605 million dozen eggs, accounting for 13 percent of domestic consumption. Egg consumption has soared since the late 1990s, perhaps spurred by Atkins and cholesterolfighting statins like Lipitor. Cal-Maine's stock soared along with it. In 2003, it rose more than tenfold. But the egg bubble may have burst. CalMaine's stock has struggled for much of the past year. Meanwhile, thanks to the surge in demand, egg producers increased ovo-capacity. But buyers for all those eggs didn't materialize. Last December, Cal-Maine reported that its second-quarter sales fell nearly 40 percent. For investorsat least for nowit's clear which comes first.

elastic. Price elasticity is a useful concept, just like income elasticity: it provides a summary measure of how we can expect demand to change if prices change. We will make use of the concept at various points throughout the course. Just as is the case with the income elasticity of demand, we need statistical techniques to estimate the price elasticity of demand for any particular product. Similarly, when a supply curve is very steep, we say supply is inelastic, and when it is flat we say it is elastic. What makes a supply curve steep? At given capacities, a supply curve will be steep if firms production costs differ significantly from each other. You should be able to plot a histogram of production costs to figure out why this is the case. In the longer term, when low-cost firms are in principle able to expand capacity, supply curves will tend to remain steep if it is very hard for firms to build new production facilities with production costs similar to their existing facilities.


3. The equations of supply and demand

Just at the point at which you were beginning think that this supply and demand business is easy, were going to spoil all that with some math. Fortunately, the math is easy. The demand curve plots a relationship between price and quantity that summarizes the amount that consumers are willing to buy at each possible price. We can also summarize this relationship by means of an equation. For example, the following demand equation,

q d = 225 6 p ,
states that for each $1 increase in price, the quantity demanded falls by 6 units. The supply equation,

q s = 25 + 4 p ,
states that for each $1 increase in price, the quantity that firms are willing to supply rises by 4 units. We will see in a moment what the numbers 225 and 25 tell us. In market equilibrium, quantity supplied must equal quantity demanded. That is qs=qd. So if we set the quantities equal to each other in the two equations, we find that in equilibrium it must be true that

25 + 4 p = 225 6 p .
This equation, which can easily be solved with a bit of high-school algebra, yields the result that the market equilibrium price is $25. Once we know the price, we can substitute this into the demand curve,

q d = 225 6(25) ,
or the supply curve,

q s = 25 + 4(25) ,
to find that quantity supplied and quantity demanded is equal to 75 units.

Even though we have solved for the equilibrium quantity and price using math, it is still useful to graph the market. To do so, we need to work out from each of the equations a couple of points that they pass through. The easiest way to do this is to rearrange the equations so that price is on the left hand sides and quantity is on the right hand side. For the demand curve this gives us

p = 37.5 q d ,
while for the supply curve we get

1 6

p = 6.25 + 0.25q s .
We already know that the demand curve passes through the point q=755 and p=25. By plugging in a quantity of zero in the demand curve above, we see that the demand curve also passes through the point q=0 and p=37.5. The supply curve similarly passes through the point q=755 and p=25, as well as through q=0 and p=6.25. We can simply mark these points on a graph and connect them with a straight line, as in Figure 13.
40 35 30 25 20 15 10 5 0 0 50 100 150

Supply: q=25+4p

Demand, q=2256p



So, we can now calculate consumer and producer surplus. Consumer surplus is the triangle above the price line and below the demand curve, and its area is CS = x (75) x (37.5 25) = $468.75. Producer surplus, which represents producer profits, is the area below the price line and above the supply curve, and its area is PS = (75) x (25 6.25) = $703.13. A real-world example. Hufbauer and Elliot (1994) produced some equations to approximate the supply and demand curves for softwood lumber in the United States. The equations the obtained were

q d = 117.5 291 p

q s = 24.3 + 47.9 p ,
where quantity is measured in units of billions of board feet of lumber, and the price is per board foot. The market equilibrium is easily found by setting quantity demanded equal to quantity supplied,

117.5 291 p = 24.3 + 47.9 p ,

and solving this equation for the equilibrium price. Doing so yields an equilibrium price of $0.275 per board foot. Plugging this back into the supply curve (or the demand curve) yields an equilibrium quantity of 37.475 billion board feet. Figure 14 illustrates this equilibrium. One feature in particular is worth noting. The supply curve appears to intersect the x axis at about 21 billion board feet. This seems to imply that firms would be willing to supply this extremely large quantity of softwood even at zero price! Doesnt seem plausible, does it? In reality, there must be some positive price at which supply would fall to zero. This curious implication comes from the fact that that we are approximating the supply curve with a straight line. This approximation in


turn came from observations on past market prices and quantities, and it is reasonable to suppose that those data were all quite close to the current market equilibrium.
0.4 0.35 billions of board feet 0.3 0.25 0.2 0.15 0.1 0.05 0 0 10 20 30 $ per board foot 40 50 60 D S

FIGURE 14. Market equilibrium in US software lumber.

Figure 15 illustrates what a more realistic supply curve might look like (dashed line): there is no reason why the real supply curve should be a straight line. However, as long as we restrict attention to market outcomes that are reasonably close to the current equilibrium, such as those within the circle, our straight line approximation will still work quite well. This does means of course, that we cannot get an accurate picture of what total producer surplus is. But that will be fine for most purposes, because we will still be able to analyze how policies change producer surplus, even though we dont know the total. Precisely how this will be done will become evident in the next chapter, when we revisit this example, among others.


0.4 0.35 billions of board feet 0.3 0.25 0.2 0.15 0.1 0.05 0 0 10 20 30 $ per board foot 40 50 60 D S

FIGURE 15. Approximation and reality in US software lumber.

1. For the following pairs of supply and demand curves, calculate the market equilibrium price and quantity, producer surplus, and consumer surplus. Graph the markets.


qd = 80 2p, qd = 80 8p,

qs = 20 + 2p qs = 16 + 4p

2. What are the equations of supply and demand that correspond to the following graphs?


p S 50 40 D 10 10 q

p 100

55 50

D 50 q

3. The market supply and demand curves are qd = 80 2p and qs = 20 + 2p. These are the equations from question 1(a), for which you have already calculated the market equilibrium. What happens to the equilibrium price and quantity, to producer surplus and consumer surplus if: (a) Preferences of consumers change, raising every consumers willingness to pay by $5 per unit? (b) A new invention reduces every producers cost by $1 per unit. 4. Consider the following supply and demand curves:

qd = 100 2p ,

qs = 50 + 3p.

(a) Calculate the equilibrium price, quantity, consumer surplus and producer surplus. (b) What happens to the equilibrium price and quantity if the cost of production rises by $10 per unit for every producer? 5. The following numbers summarize quantities sold and prices for several consecutive years. From one year to the next, the equilibrium changes, implying that either the supply curve or the demand curve must have moved. For each change, state


whether it was the supply or the demand curve that moved, and whether it was up or down.
Year 2002 2003 2004 2005 Quantity 8,300 9,300 8,766 9,030 Price $22.73 $30.00 $32.66 $27.00 Supply or Demand Shift? UP or DOWN?

6. Consider a market with demand curve qd=802p. The supply curve is either (a) qs=20+2p OR (b) qs=45+3p. In both cases the equilibrium prices and quantities are exactly the same. But for which supply curve, (a) or (b), is producer surplus larger? 7. Assume demand is perfectly inelastic. That is, regardless of the price, the quantity demanded is always equal to 10,000 units. (a) What happens to price when the unit production cost of every producer falls by $1? (b) How much does the reduced cost change consumer surplus? (iii) How much does the reduced cost change producer surplus? Illustrate your answers with a supply-demand diagram. 8. There are two types of producers. Type A producers each have a capacity of 100 tons of output per month and it costs them $100 per ton to produce, and Type B producers each have a capacity of 100 tons per month, and their costs are $200 per month. There are ten Type A producers and 15 Type B producers. If the demand curve is given by qd = 2500 10 p, where p is price per ton and q is quantity in tons, what is the equilibrium price and how much profit is made in the industry? (Hint

first draw the supply and demand lines; use this diagram to figure out what you need to calculate).


9. The demand curve for a product is given by qd =100 5 p. At any price below $10, no one produces anything. For every $1 increase in the price beyond $10, the amount produced increases by 1 unit. If the world price is $12, how much is imported?

Hufbauer, Gary, and Kimberly Elliot (1994): Measuring the Costs of Protection in the United States. Washington, DC: Institute for International Economics.


Chapter 5 Messing with Markets

Governments throughout the world intervene in markets all the time. In some markets, governments pay firms to produce more. In others, they force them to produce less. In some markets, governments use policy interventions to transfer wealth from consumers to producers. In others they do they opposite. In yet other markets, governments pay high prices to induce producers to increase output, and then impose quantity restrictions to prohibit them from doing so. In his 1961 novel, Catch-22 , Joseph Heller captured the apparent absurdity of these myriad government interventions:
Major Majors father was a sober God-fearing man whose idea of a good joke was to lie about his age. He was a long-limbed farmer, a God-fearing, freedom-loving, lawabiding rugged individualist who held that federal aid to anyone but farmers was creeping socialism. He advocated thrift and hard work and disapproved of loose women who turned him down. His specialty was alfalfa, and he made a good thing out of not growing any. The government paid him well for every bushel of alfalfa he did not grow. The more alfalfa he did not grow, the more money the government paid him, and he spent every penny he didnt earn on new land to increase the amount of alfalfa he did not produce. Major Majors father worked without rest at not growing alfalfa. On long winter evenings he remained indoors and did not mend harness, and he sprang out of bed at the crack of noon every day just to make certain that the chores would not be done. He invested in land wisely and soon was not growing more alfalfa than any other man in the county. Joseph Heller (1961): Catch-22, ch. 9

We saw in Chapter 4 how a well-functioning market, if left to itself, is efficient in the sense that the sum of consumer and producer surplus is


maximized. Inevitably then, a common theme of this chapter is that policy interventions that force price or quantity to differ from its free-market equilibrium will involve a decline in economic welfare. We will develop the tools to analyze how much these interventions cost us. Why, you may ask, would governments intervene in so many markets if their interventions always produce an economic loss? There are two answers to that, one we study in this chapter, and another that will take use several chapters to study. The first answer is that governments may not like how the market allocates welfare between consumers and producers. That is, it might like consumers or producers to have a larger share of the economic pie than the free market gives them, and to change the allocation the government might well be willing to make the pie smaller. By studying who gains and who loses from different policies, we can perhaps understand what drives many policies. We will also have the tools to be able to tell policymakers how much economic loss is caused by their desires to redistribute the economic surplus from one group to another. We might even be able to do more: our analytical tools might enable us to show how the government could achieve its aims at lower economic cost. The second motivation for policy intervention is that many markets are not well-functioning. In subsequent chapters, we will study various ways in which markets might not function well, the consequences of these market failures for economic welfare, and the options for policy interventions to improve matters. In this chapter, we will learn how to analyze a variety of policy interventions in markets that seem to be well-functioning.

1. Price Caps
A price cap is simply an upper limit on how much firms can sell a product for. For example, the equilibrium price might be $20, if the market is left to itself.


Price S A p* pcap C E B D

D qcap q* Quantity

FIGURE 1. Price Cap Legislation

The government might decide to prohibit any sales above, say, $15. Obviously, a price cap has no effect if the cap exceeds the fee-market market equilibrium. Figure 1 illustrates the effect of a price cap. We begin with the free-market equilibrium price of p* and quantity, q*, which are found at the intersection of the supply and demand curve. An effective price cap is one which sets a maximum price below p*, and this is denoted by pcap. At this legislated price, consumers would like to buy a lot more of the good than q*. Unfortunately, they cannot, because firms would not be willing to produce this much. In fact, the supply curve tells us that firms would only be willing to produce qcap, so there is a reduction in the quantity supplied as a result of the price cap legislation. Producers clearly lose as a result of this price cap. Not only is the price reduced, but the price reduction also causes firms to produce less. In the free market equilibrium, total producer surplus was given by the sum of the areas C, D and E (i.e., the area below the price line and above the supply curve). After the price cap, producer surplus is just E. Thus, firms lose profits equal to


the area C+D. If we had equations to help us locate the supply and demand curves, we could easily calculate the dollar loss in producer surplus. For consumers, the situation is less clear. On the one hand, they pay a lower price with the policy than without it. On the other, they are forced to consume less because production has declined. In the free-market equilibrium, consumer surplus was equal to A+B. With the imposition of the price cap, consumers lose B because of the reduction in quantity supplied, but they gain C because of the reduction in price. So do consumers gain or lose from the price cap? The answer, of course, depends on the relative sizes of areas B and C, and this in turn depends on the slopes of the supply and demand curves. Figure 2 illustrates two cases. On the left, the supply curve is steep supply is inelastic. It is easy to see that in this case, area C is larger than area B, so consumers gain from the price cap. When supply is inelastic, there is little reduction in the quantity supplied when the price cap is imposed, and hence consumers are better off. On the right, however, is a case where supply is elastic. A reduction in price induces a large reduction in quantity supplied, and this implies that area C is smaller than B. Consumers in this case suffer a loss as a result of the price cap.

Price S




A B p* pcap C E D S



D qcap q* Quantity

D qcap q* Quantity

FIGURE 2. Price Caps with different price elasticities. Left panel: inelastic supply and consumers gain. Right panel: elastic supply and consumers lose.


In summary, price cap legislation always hurts producers. It will benefit consumers if supply is relatively inelastic, but it will hurt consumers if the quantity supplied is sensitive to changes in the price. When supply is elastic, then, everyone loses, so there really can be no justification for the policy. With an inelastic supply, however, a government might be able to justify the policy if it cares enough about making consumers better off. There are a couple of informative ratios one can calculate from this and any other policy analysis. Notice that the price cap reduces total economic welfare by the area B+D. Thus, to increase consumer surplus by the amount CB, we must give up economic welfare equal to B+D. Put another way, each dollar of increase in consumer surplus has an economic cost equal to (CB)/(B+D). Another informative ratio is the cost to producers of each dollar of increase in consumer surplus. This is given by the ratio (C+D)/(CB). As the numerator of this ratio is greater than C and the denominator is less than C, the ratio is always greater than one. That is, to transfer each $1 of economic benefit to consumers costs producers more than $1. This is, really, another way of saying that using price caps to make consumers better off results in a loss of total economic welfare: what consumers gain will always be less than what producers lose. Politically, policymakers may decide it is worth shrinking the economic pie in order to redistribute it. As economists we cannot make that decision for them, although we can make sure policymakers understand what the tradeoffs are. We might also be able to suggest alternative ways to make consumers better off without shrinking the economic pie so much.

Price controls and the allocation problem. Clearly, price caps involve a loss of
economic welfare. Unfortunately, what we have seen so far is only the tip of the iceberg. This is because, once you directly control the price, you eliminate the role of price in ensuring that those who value an item most are the ones who get it. Our graphical analysis of the welfare consequences of a price cap implicitly assumed that the people who get the goods are those with the 134

highest valuations. But this need not be the case. There are many more people who value the item more than the regulated price than there are units available for sale. Any of these individuals may in fact get to consume, and any of them could be unlucky enough to be excluded. Hence, the actual gain in consumer surplus from a price cap is very likely to be less, possibly much less, than what we have drawn in the graphs. We are forced to conclude, then, that a price cap is a very inefficient way to try to make consumers better off.

Rent control. During World War II, New York state legislators introduced the
War Emergency Tenant Protection Act. The Act was designed to protect residents from shortages of apartments during the war, although how it was supposed to accomplish that aim was never entirely clear. Amazingly, many of the apartments that had rent caps imposed on them during WWII are still subject to rent control. But New York is not the only guilty party. Many other cities also institute rent control of one form or another. Our analysis of price caps shows unequivocally that the supply of rental housing will decline if rent controls are imposed. Owners of buildings will be made worse off. Consumers as a whole will be made better off only if the supply of apartments is sufficiently inelastic. But even then, an increase in the welfare of consumers is accomplished only by making some people (those who get the apartments) much better off, while making others homeless. The article on the next page from the New York Times describes the trials and tribulations of apartment hunters in San Francisco. Prices are high, but nonetheless every time an apartment is offered for rent, hordes of potential tenants arrive to beg for consideration. Capitalism run amok? Perhaps. But the article induced an immediate response from Princeton economist and NYT columnist Paul Krugman. [T]here was something crucial missing, he writes, specifically two words I knew had to be part of the story . . . . rent control. Although the first article makes no mention of it, the 1979 San Francisco Rent Ordinance subjects 170,000 rental units in that city to price caps.



In San Francisco, Renters Are Supplicants Evelyn Nieves, June 6, 2000

The open house for the $1,800-a-month four-room flat (1 bdrm, 1 bth) began at 9:30 a.m. sharp. By 9:35, a couple were perched on the wooden steps hastily filling out a rental application, and two more were about to do the same. Still, after three months of nonstop searching for an oxymoron -- an affordable, available apartment in San Francisco -- the couple, Kim Kowalski and Dylan Snodgrass, had nothing to lose by throwing their credit reports in the ring. ''Even though we have a combined income of $110,000, we're considered low-end,'' said Ms. Kowalski, a 31-year-old marketing representative. ''We've had places charge for the rental application. We've had people require that we give them sealed credit reports. Usually, if it's a good place for a good price -- our limit is $2,200 - there's usually around 30 applicants.'' In fact, after an hour at the open house, Gene Fong, who was showing the apartment in the four-unit house for its owners, had had enough. Just one little advertisement in an Internet community bulletin board on Thursday afternoon had yielded more than 40 e-mail queries and at least that many visitors to the Saturday morning showing. ''I guess the rent is pretty reasonable,'' Mr. Fong said, as though he thought it might be a little too reasonable. There would be no need for another open house. There never is. In San Francisco, which has a vacancy rate of less than 1 percent, compared with 2.57 percent in Manhattan, there are never enough apartments. The free-market horror stories about elderly longtime tenants being forced out so that young blood with new money can pay whatever it takes are all too common here. So are the tales of barely inhabitable rooms for $1,000 a month. The hunt for an apartment in almost any neighborhood in San Francisco is unlike the hunt anywhere else in the country, even overpriced Manhattan. A 300-square-foot unrenovated studio in the kind of neighborhood your mother taught you to avoid can cost $1,100 a month, with a three-month security deposit. A studio in a building you would not mind showing your boss can run $1,700. Rents for one-bedroom apartments with no amenities save a respectable address are at least $1,800. But it is not only the prices that make the search daunting. Any apartment listed in the classifieds or on rental bulletin boards will draw dozens, even scores of applicants. This leads to intense competition among prospective tenants. Individual appointments are almost unheard of; most apartments are shown through open houses where applicants elbow each other to audition as the best rental bet. Rental agents are suggesting that prospective tenants come armed with renters' resumes detailing their credit and job history, credit reports and references. They also suggest that renters wear professional attire and show enthusiasm for the apartment, as though the $1,800-a-month junior onebedroom walk-up overlooking the freeway were their dream pad, the place they wanted to make a home for life. ''Renting a place in San Francisco, people have to treat it as if they are applying to an Ivy League college,'' said Grey Todd, president of Rent Tech, an apartment-listing service. ''You have to kiss up to the landlords. The landlords can pick the creme de la creme of tenants, the absolute perfect person for their space. How many people are looking for a studio or a two-bedroom? As many as 50 to 80 for each place. The tenant has to stand out.'' 'I would suggest, if you're looking and you see something you're vaguely interested in, you take it,' Mr. DeShon said. Ms. Kowalski and Mr. Snodgrass, who have been renting a room in a friend's house for $1,200 a month until they find a place, have turned the task into a fulltime job. ''I basically spend at least six hours a day, six days a week looking,'' said Ms. Kowalski, whose real job is on hold while Mr. Snodgrass, a carpenter, earns their keep. The couple moved here from the Central Coast city of San Luis Obispo, where the living is easy and Ms. Kowalski paid $580 for a one-bedroom apartment. ''Dylan wanted to live in San Francisco because it's such a great city,'' Ms. Kowalski said. ''But we have no life. We do nothing but look for a place. We do nothing but compete with other people for the landlords' attention.'' With that, Mr. Snodgrass joined a group of applicants encircling Gene Fong at the open house on Saturday, hoping Mr. Fong would notice what a nice, clean-cut young man he was.


A Rent Affair Paul Krugman, June 7, 2000

Economists who have ventured into the alleged real world often quote Princeton's Alan Blinder, who has formulated what he calls ''Murphy's Law of economic policy'': ''Economists have the least influence on policy where they know the most and are most agreed; they have the most influence on policy where they know the least and disagree most vehemently.'' It's flip and cynical, but it's true. Consider, on one side, really tough issues -- where there are plausible arguments on both sides, where nobody really knows how to measure the tradeoffs. Should Microsoft be broken up and, if so, how? Should Britain adopt the euro? Let's ask the economists! And those economists who are prepared to express strong opinions on such inherently ambiguous questions command rapt attention. On the other side, consider an article that appeared in yesterday's New York Times, ''In San Francisco, Renters Are Supplicants.'' It was an interesting piece, with its tales of would-be renters spending months pounding the pavements, of dozens of desperate applicants arriving at a newly offered apartment, trying to impress the landlord with their credentials. And yet there was something crucial missing -- specifically, two words I knew had to be part of the story. Not that I have any special knowledge about San Francisco's housing market -- in fact, as of yesterday morning I didn't know a thing about it. But it was immediately obvious from the story what was going on. To an economist, or for that matter a freshman who has taken Economics 101, everything about that story fairly screamed those two words -- which are, of course, ''rent control.'' After all, the sort of landlord behavior described in the article -- demanding that prospective tenants supply resumes and credit reports, that they dress nicely and act enthusiastic -- doesn't happen in uncontrolled housing markets. Landlords don't want groveling -- they would rather have money. In uncontrolled markets the question of who gets an apartment is settled quickly by the question of who is able and willing to pay the most. And so I had no doubts about what I would find after a bit of checking -- namely, that San Francisco is a city where a technology-fueled housing boom has collided with a draconian rent-control law. The analysis of rent control is among the bestunderstood issues in all of economics, and -- among economists, anyway -- one of the least controversial. In 1992 a poll of the American Economic Association found 93 percent of its members agreeing that ''a ceiling on rents reduces the quality and quantity of housing.'' Almost every freshman-level textbook contains a case study on rent control, using its known adverse side effects to illustrate the principles of supply and demand. Sky-high rents on uncontrolled apartments, because desperate renters have nowhere to go -- and the absence of new apartment construction, despite those high rents, because landlords fear that controls will be extended? Predictable. Bitter relations between tenants and landlords, with an arms race between ever-more ingenious strategies to force tenants out what yesterday's article oddly described as ''freemarket horror stories'' -- and constantly proliferating regulations designed to block those strategies? Predictable. And as for the way rent control sets people against one another -- the executive director of San Francisco's Rent Stabilization and Arbitration Board has remarked that ''there doesn't seem to be anyone in this town who can trust anyone else in this town, including their own grandparents'' -- that's predictable, too. None of this says that ending rent control is an easy decision. Still, surely it is worth knowing that the pathologies of San Francisco's housing market are right out of the textbook, that they are exactly what supplyand-demand analysis predicts. But people literally don't want to know. A few months ago, when a San Francisco official proposed a study of the city's housing crisis, there was a firestorm of opposition from tenant-advocacy groups. They argued that even to study the situation was a step on the road to ending rent control -- and they may well have been right, because studying the issue might lead to a recognition of the obvious. So now you know why economists are useless: when they actually do understand something, people don't want to hear about it.


2. Price Floors
A price floor is a minimum price, below which it is not legal to sell. After thinking through the effects of a price cap, you should find it is easy to predict the effects of a price floor: consumers will lose, producers may win or lose depending on the slope of the demand curve, and there will be a net economic loss. Figure 3 illustrates these effects. A price floor is only effective if it exceeds the free-market equilibrium price. An effective price floor of pfloor is indicated in the figure. Firms would like to produce much more than the free-market equilibrium quantity at this price, but if they do consumers wont buy. In fact, consumers will buy only qfloor. Consumers lose surplus equal to area C+B. Producers may gain or lose. They gain C from the higher price, but lose D from the lower quantity demanded. C will be greater than D if the demand curve is steep, and it will be smaller if the demand curve is flat. You should draw these two cases as an exercise. Whether producers gain or lose from the price floor depends on the elasticity of demand. This is because it is consumer demand that determines the quantity change induced by the price floor. Contrast this with the price cap: in that case, consumers stand to gain or lose depending on the slope of the supply
Price S A pfloor C p* E B D

D qfloor q* Quantity

FIGURE 3. A Price Floor


curve because it is producer supply that determines the quantity change induced by a price cap. And yet again, Figure 3 illustrates the best that one can expect. With a price floor, there is no mechanism to ensure that only the lowest cost producers stay in the market. Any firms with unit production costs less than pfloor might be able to sell, and some high-cost firms might get their foot in the door at the expense of low-cost firms. Thus, the total welfare loss, which is given by B+D in Figure 3, may in fact be much larger than illustrated.

The Minimum Wage. Without doubt, the most widespread example of a price
floor is the minimum wage. In this case, the market under consideration is the demand for, and supply of, workers. Firms hire workers to produce goods and services. The cheaper workers are, the more workers firms will hire. Workers provide labor in exchange for wages, and the greater the wage the more workers enter the labor markets. Thus, in the labor market, firms are like consumers (because they have a demand for workers), and individuals are like firms (because they supply the labor). A free market equilibrium equates supply and demand. Now imagine the government imposes a minimum wage. If the minimum wage has any effect (i.e. if it exceeds the equilibrium wage), then its effect must be the following: it raises the wage of those who have jobs, while reducing the number of jobs. So some people will lose their jobs and hence their entire income, while more lucky workers keep their jobs and see a rise in their wage. Firms (who are on the demand side in this market) must be made worse off. Finally, the minimum wage reduces total economic welfare. Proponents of higher minimum wages are obviously willing to forego some economic efficiency in order to improve the standard of living of the working poor. The question that must be grappled with is how many of the working poor will become the unemployed poor after a rise in the minimum wage. This is of course, a question about the slope of the labor demand curve: how many of their lowest-paid workers will firms dismiss if they are required to pay them 140

more? The evidence is, unfortunately, hard to interpret. The labor market consists of many different types of workers doing different jobs and with different skills, and many different types of employers, so the statistical challenges of sorting out what changes in employment are due to changes in the minimum wage, and what are due to changes in other factors, is unusually daunting. Among the more prominent studies, Neumark and Wascher (1992), report that a 10 percent increase in the minimum wage is associated with a one to two percent reduction in teen employment. If their numbers are about right, it appears that the employment effects are quite modest. But it does mean the following. If there are 10 million teens earning the minimum wage, a 10 percent increase in the minimum wage will make 9.8 to 9.9 million of them 10 percent better off, at the price of putting 100,000 to 200,000 teens out of work. This is the sort of tradeoff that must be made. Unfortunately, these numbers are hardly reliable. Some researchers over the last decade, looking at changes in the minimum wages set at the state level have concluded that the effect on employment of increases in the minimum wage are negligibly small. Other researchers, looking at the exactly the same data, have concluded that there are much larger effects than those reported by Neumark and Wascher.

3. Production Quotas
The British love their potatoes, and their government loves the farmers who grow them even more. Thats why, from 1955 until the late 1990s, the government imposed massive fines on farmers who grew a lot of potatoes. New York loves its Yellow Cabs. Thats why the government there has mandated that there can be no more than 12,000 of them. Restricting things you love is not as crazy as it sounds because, if market conditions are right, you can make those you love better off by having fewer of them. The mechanism should be readily apparent. Pass a law to restrict supply, and price will get driven up. If price gets driven up enough, then total


Price S A pq C p* E B D

D qquota q* Quantity

FIGURE 4. Production Quotas

producer surplus may increase, despite the reduced quantity. Figure 4 illustrates just this scenario. The equilibrium quantity and price that would prevail in a free market are indicated by q* and p*. The government restricts production to qquota by issuing a limited number of licenses to produce. This drives the price up to pq. There is an area of lost producer surplus, given by D, and a gain, given by C. As drawn, C is larger than D, so producer surplus rises. That is total industry profits are greater with the quota than without. Clearly consumers lose: they get to consume less and they pay more for the privilege. Consumers lose area B+C. The net welfare change from the policy is given by the gain to producers, CD, minus the loss to producers, B+C. That is, the welfare change is (CD)(B+C)=(B+D). This is negative. In other words, the policy induces a net loss of welfare equal to B+D. Figure 5 illustrates a case where the quota reduces producer surplus. In this case, the increase in producer surplus, area C, is smaller than the loss in


Price S C


A p* E





FIGURE 5. A producer quota that reduces profits

producer surplus, area B. Industry profits are reduced by imposing a production quota. The difference between Figures 4 and 5 is visually obvious. Demand is much more price-elastic in Figure 5 than it is in Figure 4. When demand is price-elastic, price does not rise much when quantity is reduced, so profits tend to fall. One may expect, then, that quotas are used when the government wants to make producers better off and when the price elasticity of demand is small. We know that the demand for basic foodstuffs tends to be price-inelastic. This is why quotas are found in agriculture more often than anywhere else.

Selling quota rights. Imagine there are 100 producers, each capable of
producing one unit of output per month. The government decides that demand is elastic and it can raise producer surplus by restricting output to 50 units per month. How does it decide which of these fifty producers should get the quota? The government has a number of options. First, it could restrict each producer to half a unit per month. Then every producer gains from the rise in price. However, every producer is now only


producing at half capacity, so we are likely to see a lot of half-empty factories. This is surely a waste of resources. A second option is to choose fifty producers at random, and allow each of them to produce at capacity. This will have two undesirable effects. One is that fifty producers will now be very angry, as they have just lost their livelihood. The other is that production is not guaranteed to be undertaken by the fifty lowest-cost producers. This is a wasteful loss of efficiency. The government might try to identify the fifty cheapest producers and select them. This would be more efficient, but it would require that the government knows the production costs of every member of the industry, a virtually impossible task. Moreover, it would still leave the government with fifty angry former producers. An alternative is to allow there to be a market in quota rights. The government could allocate the rights to produce half a unit per month to every producer, and then allow them to trade these rights among themselves. Highcost producers will find that the quota rights are worth less to them than they are to low-cost producers. Consequently, they would sell or rent the rights to low-cost producers. To see how this would be more efficient and attractive to every producer, consider the following example. Imagine that when 100 units are produced the market price is $10. Fifty producers have costs of $5 per unit, and fifty have costs of $10 per unit. Hence, the fifty-low cost producers make profits of $5 each, while the fifty high-cost producers make zero profits. When production is reduced to 50 units, imagine that the price rises to $30. If everyone is allowed to produce half a unit each, then fifty producers make profit of $12.50 (equals $25 per unit x half a unit), while the other fifty make profits of $10. Every producer is better off. But by allowing a market in quota rights, producers can do even better. Take one high-cost producer and one low-cost producer. If the high-cost producer can rent out his quota rights out for at least $10 month, he will prefer to do so 144

than to produce. If a low-cost producer can obtain the quota rights for half a unit for anything less than $12.50 per month, he would like to do so and produce at full capacity. Hence, each high-cost producer will rent to a lowcost producer at some price between $10 and $12.50, and all production will be undertaken by low-cost producers. In many markets, producers do not rent their quota rights, they sell them, For example, the price of a medallion giving you the right to operate a yellow cab in New York now sells for around $250,000. Where does this price come from? It comes from calculating the capital value of the flow of profits you get from the quota rights. The method is much the same as the payoff options if you win the lottery: you can either choose a cash payment, or a flow of annual payments over twenty years. Because money today is more attractive than money in the future, the amount you get from a cash payment is much less than the sum of twenty years of annual payments. Nonetheless, most people opt for the cash payments.

Quota Prices in British Potatoes. Until recently, the British Potato Marketing
Board (PMB)21 operated a quota scheme with a couple of twists. While anyone was free to grow potatoes, the PMB charged a small license fee to growers who had been allocated a quota, and a much larger fee to growers who did not have a quota. Table 1 provides the relevant numbers of the period 1987 through 1991. In 1987, for example, growers with quotas paid a fee of 30.89 per acre, while growers without quotas paid a fee of 247.29. In addition, growers that had been allocated a permit for a certain acreage were told each year the maximum fraction of that acreage that they could actually plant without paying the larger fee. The limit was known as the quota cap. For example, in 1987, growers could only plant 95 percent of the acreage they were licensed for. Thus, in 1989, a grower with a one-acre quota paid a fee of 30.89 for 95


The Potato Marketing Board was replaced by the British Potato Council (BPC) in 1997. But the BPC really is too boring an organization to study.


percent of the acre, and 247.29 for 5 percent of it, for a total fee of 30.89x0.95)+247.29x(0.05)= 41.71. The value in 1989 of having a quota was therefore 247.2941.71=205.58, which is the amount of fees avoided in that year by virtue of having a quota. Although this policy seems complicated, the basic principle of the potato quota system can be drawn quite simply. Figure 6 illustrates. Imagine the PMB has decided on a total quota of qquota. For producers with quota rights, their production cost is, say, c, while for non-quota producers it is c+205.58. Hence the supply curve jumps up at quantity qquota because all production in excess of qquota pays the higher fee. Compare this with the case of no quota. Then all producers have the same production cost and, instead of the supply curve S, we would have the horizontal supply curve S/.22 Without the quota system the price would be much lower.

TABLE 1. Grower Fees Payable to the PMB Prices and fees in pounds sterling per acre WITHIN-QUOTA GROWERS 1987 1988 1989 1990 1991 30.89 30.89 31.60 33.08 34.66 OUTSIDE QUOTA GROWERS 247.29 247.29 349.82 363.85 381.22 QUOTA CAP 95% 99% 100% 100% 100% QUOTA SELLING PRICE 380 395 505 500 550


If some producers had lower costs than others because, say, they had better land, then the supply curve would slope upwards. Recall that it is the distribution of production costs among different producers that determines the shape and slope of the supply curve.



pq c+205.58 p* c


D qquota q* Quantity

FIGURE 6. The exciting British potato quota system

As Table 1 illustrates, the PMB adjusted the size of the excess fees and the quota cap from year to year. It did this for a very good reason. The PMB existed to raise producer profits. To do this, it needed to raise price. The ideal way, from the perspective of farmers, was for the quota limit and price to be set so that the jump in the supply curve at qquota lies just far enough to the left so that the demand curve only just intersects the supply curve at its highest level. If the quota were set too far to the right in the figure, the price would be lower, and if it were set too far to the left, too many producers would be needlessly paying the higher fees. As demand conditions changed from year to year, the PMB had to make annual minor adjustments to achieve its aim. The PMB also allowed farmers with quota rights to sell them to non-quota farmers. There are two factors that would make a quota more valuable. One is if the difference between the excess fee and the in-quota fee got bigger. The other is if the quota cap increased. In the final column of Table 1, average sale prices of quota rights are recorded. Note the rise in price in 1989, just when


the excess fee jumped significantly. Note also the further rise in price when the quota cap increased to 100%.

Taxis in New York. A market also exists for the rights to operate a yellow cab
in Manhattan. These rights are conferred by a license termed a medallion. The article on the following page suggests that efforts at arresting unlicensed cab drivers in 2003 had more to do with raising the value of a license than it had to do with consumer safety, because the city was just about to auction some new medallions. Why would this work? The value of the medallion is the difference between the profits you can make with it and the profits you can make without it. Reduce the profits you can make without the medallion (by arresting drivers who do illegal things) and you have succeeded in raising its value. Or do you think that author of the article is being just a little too cynical?

4. Taxes and Subsidies

The last set of policies that we study in this chapter are without doubt the most widely used. In the United States, most products we buy are taxed through the near-universal sales tax. Some products, such as gasoline, alcohol and cigarettes, are subject to additional taxes. Subsidies, where the government pays out money for each unit produced are much less common. However, because a subsidy is nothing more than a negative tax, we shall analyze them together. Let us get straight to the analysis, with Figure 7. The figure depicts a market equilibrium for a good in which the government has imposed a $1 tax on each unit of the good. In this case it must be the case that the price paid by consumers is exactly $1 greater than the price received by producers. In Figure 7, pc is the price paid by consumers and pp the price received by produers, and pc=pp+1. There is only one output level, denoted by qtax, where quantity demanded equals quantity supplied at the same time that the consumer price is exactly $1 greater than the producer price.


Taxi Busters Myron Magnet

14 May 2003
I walked out my door on the Upper-Upper West Side this morning into a posse of undercover cops. What crime were they busting? Terrorism? Drug trafficking? Nope: illegal pickups of passengers by livery car drivers, whom the law forbids to respond to street hails south of 125th Street in Manhattan. Well, you might think, perhaps this is the kind of low-level, quality-oflife enforcement that worked so well for the Giulianiera NYPD. Trouble is, this particular use of police will lower New Yorks quality of life, not raise it. And the Bloomberg administrations motives for using scarce police resources this way are intensely suspect. In a gradual, ad-hoc way, Gotham has developed a very efficient three-tier taxi system similar to that devised by wise policy in cities like Hong Kong. Since taxis tend to congregate in central business districts and at airports, the problem for regulators is to prevent excessive congestion in the center city, while ensuring adequate service in the outlying parts of town, where passengers are thinner on the ground. Hong Kongs solution is to issue three classes of taxi licenses, the most expensive allowing drivers to operate in the highest-volume areas, the cheapest permitting cruising in the periphery. New Yorks three-tier system turns over midtown and the airports to its 12,000-plus yellow cabs, who pay up to $250,000 for the medallion that licenses them to respond to street hails in those areas. This fleet is no bigger than the number of cabs on the street when the medallion system went into effect in 1937, and it is too small to provide Gotham with adequate service. But because medallion owners have resisted any efforts by the city to dilute their regulated monopoly by selling significantly more medallions, two additional kinds of service have grown up, called into existence by market demand. A fleet of about 8,000 black cars luxury sedans summoned by telephone and serving the corporate markethas effectively increased service to the business district. Much less opulently, some 30,000 livery cabs, or local car services, ply the outer boroughs and the northern reaches of Manhattan. Though the city licenses these vehicles to respond to street hails only north of 125th Street or outside of Manhattan, the livery cabs for at least a decade have in fact been picking up passengers on the streets of the Upper West and Upper East Sides, without interference from the police. As demand has inexorably increased beyond the capacity of the supply of yellow cabs to fill it, residents of those neighborhoods, thanks to livery cars, have had enough cabs to get to work or take kids to school. At the same time, a class of microentrepreneurs has flourished among the hardworking owner-drivers, mostly middle-aged, Hispanic, family men. This is the system, which benefits so many, that the new enforcement regime will destabilize. If the mayor werent an independent billionaire, you might think that the politically powerful taxi fleet owners had exerted pressure on himas they have on so many city politiciansto limit supply and so to increase the value of their monopoly. The real cause is probably no more savory, however. The war on livery drivers is of a piece with the administrations use of the cops to help fill the citys empty fisc by issuing a blizzard of tickets that will produce income from fines. Worse, since part of city halls cockeyed plan for raising revenue is to sell an additional 900 yellow-cab medallions (in itself, not at all a bad idea), one cant help wonder if the antilivery enforcement is aimed at raising the price of those medallions prior to the sale. In that case, the city would be not the tool of the special interests but just one more special interest itself.


Price S A pc C pp D D qtax Quantity $1 B

FIGURE 7. A $1 tax on each unit of a good.

Compare output and prices with what would prevail in a free (i.e. tax-free) market. You know where the free-market price and quantity would be, so I havent bothered drawing them in the graph. The tax has the effect of reducing the quantity produced and consumed; it raises the price consumers pay above what they would pay in the absence of the tax, and it reduces the price that producers receive below what they would get in the absence of the tax. I leave you to work out the areas that correspond to the loss of consumer surplus and the loss of producer surplus caused by the tax. Where does this lost economic welfare go? Part of it goes to the government in the form of tax revenues. Total revenues to the government are the tax per unit multiplied by the number of units produced. This is equal to the area C. But part of it comes from the loss in efficiency that results from forcing quantity to differ from the free-market equilibrium level. This loss is area B. When the government raises revenue from a policy, we define the economic value from the market as the sum of producer surplus, consumer surplus, and


government revenues. You should be able to work out that the sum of these three components must be less than the sum of producer and consumer surplus that would be obtained in the free-market equilibrium.

Tax Incidence. In Figure 7, we saw that producers and consumers are

adversely affected by the tax. Consumers pay more and producers receive less. Note that we have said nothing about who physically pays the tax. Politicians often make statements about who will be paying certain taxes. But as Figure 7 shows, any such statements are worthless. Under current laws (as of February 2005), social security our retirement benefitsare funded by a 12.4 percent tax on wages. The law explicitly states that half this tax , 6.2 percent of wages, is to be paid by employees and half by employers. The economic reality says different. Figure 8 contains four graphs, where we have plotted labor supply and labor demand. The price of labor is the wage. In each case we have identified the equilibrium quantity such that the wage paid by employers is 12.4 percent greater than the wage received by employees.23 We have also indicated in each case what the common wage would be if there were no tax. The burden of the tax on employers and employees is represented by the change in the wage paid or received. In panel A, we see that the burden is about the same for both employers and employees: they do indeed pay about half the tax. In panel B, however, employers pay much more than employees, in the sense that the increase in the wage they pay is much greater than the reduction in the wage received by employees. The opposite is true in the case of panel C, where employees pay the lions share of the burden. In panel D, finally, we have a case in which employees pay the entire burden.


To focus on social security taxes, we are of course ignoring all other taxes income tax, Medicare, etc. -- that must also be paid.



Panel A


Panel B

12.4% wf w* we

Labor Supply

wf w*


Labor Supply


Labor Demand

Labor Demand


Panel C




Panel D Labor Supply

Labor Supply

wf w* 12.4% we Labor Demand

wf , w* 12.4% we Labor Demand



FIGURE 8. The burden of the social security tax falls differently on employees and employers when the slopes of the labor supply and labor demand curves are different. wf is the hourly cost of labor paid by firms, we is the hourly wage received by employees; w* is the wage that would be paid and received in the absence of a tax.

The differences across these graphs are obvious: who pays the burden of a tax depends on the relative slopes of the supply and demand curves the price elasticities. When the elasticities of supply and demand are about equal, then the burdens of the tax are about equal. When labor supply is more elastic than labor demand, then employers (i.e. the people on the demand side) may the greater the share of the burden. When labor demand is more elastic than labor


supply employees (i.e. the people on the supply side) pay the greater share of the burden. In panel d, labor supply is completely inelastic: the quantity supplied does not change in response to a change in the wage received at all. In this case, the supplier of labor pay the entire cost of the tax. This is a general principal for all goods and services. If the supply curve for a good is steep relative to demand, then producers of the good pay most of the tax burden. If the demand curve is steep relative to supply, then consumers of the good pay most of the burden. The administrative detail of who physically collects the tax and pays it to the government is completely unimportant.

Revenues and Elasticity. Consider Figure 9. There is initially a tax equal to the
vertical distance aa. Assume now that the government raises the tax so that it equals the vertical distance bb. What happens to tax revenues? Under the original tax, revenues are equal to the area B+C. When the higher tax is imposed, revenues become A+B+D. The net change in revenues, then, is A+DC, which may be positive or negative. Tax revenues rise because more is collected on each unit sold. But they fall because the quantity traded declines.


b A a B C


a D b



FIGURE 9. Changes in the tax rate and changes in revenues


It should therefore be no surprise that tax revenues will rise after an increase in the tax level if quantity does not change too much, and revenues will fall if quantity changes a lot. The relationship between the change in the tax rate and the change in revenues therefore depends on the slopes of the supply and demand curves. The more elastic supply and demand are, the greater the likelihood that increasing the tax rate will reduce revenues. As an exercise, you should draw a graph in which an increase in the tax reduces revenues, and another in which an increase in the tax raises revenues. Whatever the slopes of the supply and demand curve are, two things are obvious. First, if the tax rate is zero, then revenues will be zero. Second, if the tax rate rises to a high enough level, indicated by the vertical distance cc in Figure 9, then quantity falls to zero and revenue must again be zero. At intermediate tax levels, greater than zero but less than cc, revenues will be positive. Put another way, as you increase the tax rate above zero, revenues must initially rise. At some point, as the tax rate keeps increasing, you will reach a maximum for revenues, after which any further increases in the tax will lead to a decline in revenues. This situation is illustrated in Figure 10. The figure plots two curves relating the tax rate to tax revenue. The curves are known as Laffer curves, for reasons that will be made clear below. Curve A corresponds to a situation of relatively inelastic supply and demand, while curve B corresponds to a situation of relatively elastic supply and demand. When demand and supply are inelastic, revenues rise more quickly as the tax rate increases, and they rise for longer. With elastic supply and demand, you dont get as much revenue as the tax increases, and you reach the point at which revenues are at their maximum at a lower tax rate than you do with inelastic demand and supply. In 1980, when President Reagan entered the White House, Laffer curves were at the center of a political storm. Reagans economic advisors had convinced him that a cut in taxes would stimulate so much increase in economic activity that revenues would actually rise. Thus, one could cut taxes, raise revenues,



B t Tax rate

FIGURE 10. The Laffer Curve

and use this increase in revenues to finance the massive military expansion deemed necessary to bring the Cold War to an end. Most economists thought that this was an unlikely scenario. They believed that a cut in taxes would reduce revenues, and if government expenditure were to increase, there would be extremely large budget deficits. But there were some exceptions, including Art Laffer, a professor who was at the University of Southern California at that time. In a meeting with Jude Wanniski, an editorial writer for the Wall Street Journal, Laffer drew a diagram just like Figure 10. Although Figure 10 was a familiar concept to all economists, and had been around for many years, the curve became popularized as the Laffer curve, as though it were a new concept. Laffer, Wanniski, and ultimately Reagans economic advisers believed that the US economy looked like curve B in Figure 10. If the current tax rate is t, then, a cut in the tax moves the economy up the Laffer curve, raising revenues along the way. In contrast, the majority opinion among economists was (and is today, should the debate arise again) that the economy looks more like curve


A. Given a current tax rate t, a cut in taxes moves one down the Laffer curve, thereby reducing revenues. In 1981, the Reagan administration instituted a series of large tax cuts. What happened to revenues? As the following table shows, revenues initially fell after the tax cuts, suggesting that Curve A was the more accurate representation of the US economy. But by 1984, tax revenues had recovered, suggesting that if one takes a long enough time perspective, Reagans advisers were right. In reality, however, it is very difficult to know what to make of the recovery in tax revenues. In 1981-2, the economy was in a recession so tax revenues would naturally be lower because there is less economic activity. By 1984 the economy, as it typically does, had recovered from the recession so revenues would have risen anyway. To this day, supporters and opponents of Reagans economic policy continue to debate the legacy. If you have not already done so, you will no doubt hear more about this in your macroeconomics class.

TABLE 2. Federal Tax Revenues, (billions of dollars, adjusted for inflation) 1980 1981 1982 1983 1984 728.1 755.5 738.2 584.3 730.4

Subsidies. A subsidy, a payment from the government per unit produced, is

essentially a negative tax. So, while a tax reduces quantity, a subsidy raises it. While a tax raises the consumer price, a subsidy reduces it. While a tax reduces the producer price, a subsidy raises. And while a tax reduces total


Price S A pp C pc B D E F $1

D qsubsidy Quantity

FIGURE 11. A Subsidy

welfare through inefficiency, a subsidy also reduces welfare. Thats right, taxes and subsidies both reduce welfare, and this is because they both induce a change in the quantity produced away from the quantity that would be produced in a free market equilibrium. Figure 11 illustrates the effect of a $1 per unit subsidy. The quantity supplied increases, by increasing the effective price received by producers and reducing the price paid by consumers. Producer surplus, given by the area B+C+F, increases compared with the free-market equilibrium, as does consumer surplus (given by A+C+E. This is brought about by government expenditure of $1 x qsubsidy on the subsidy, which is equal to area C+B+D+E. Showing that there is a net welfare loss (the gains in producer surplus and consumer surplus are less than the cost of subsidy) equal to area D is left as an exercise.

5. Concluding Comments
In Chapter 4 we saw that a well-functioning market, when left to itself, produces an equilibrium price and quantity that maximizes the sum of


consumer surplus and producer surplus. It is no surprise then, that policies that interfere with this equilibrium reduce economic welfare. We have analyzed in this chapter a small set of policies that governments may use. We looked at price floors, price caps, quotas, taxes, and subsidies. All of them resulted in a loss of economic welfare compared with the free-market equilibrium. Sometimes, I have shown this in the text, other times I have asked you to confirm it for yourself as an exercise. Governments are willing to engage in these policies despite the reduction in total economic welfare because it enables them to make one group of participants in the market better off. For example, a price floor raises the market price. This increases producer profits, while reducing consumer surplus. Although the loss in consumer surplus exceeds the gain to producers (and hence there is a net economic loss), a government that cares enough about producers may be willing to impose a price floor. Similarly, a government that cares enough about consumers may impose a price ceiling. The impacts of policies, especially production quotas, taxes and subsidies, depends critically on the slopes of the supply and demand curves. A production quota will raise producer profits only if demand is sufficiently inelastic. It will hurt producers if demand is too elastic. Regardless of the slopes of supply and demand, however, consumers will always be hurt by a production quota. With respect to taxes (and subsidies), the common public perception that either firms or consumers pay one tax or another was found to be highly misleading. The only sensible way to think about who pays the burden of a tax is to assess what price each group pays or receives in comparison to the price that would have prevailed in the absence of the tax. We found that the greater burden is paid by whichever group producers or consumers has the most inelastic curve.


1. For the following pair of demand and supply equations:

q d = 80 2 p , q s = 20 + 2 p ,
calculate the impacts on market price, quantity consumed and sold, producer surplus consumer surplus, and total economic welfare, of introducing the following policies. a) A production quota of 20. b) A price cap equal to $20 c) A price cap equal to $30 d) A price floor equal to $30 e) A tax equal to $5 per unit of the good. f) A subsidy equal to $5 per unit of the good. Remember that in (e) and (f) government revenue or government expenditure must also be figured in as part of total economic welfare. 2. Some policies are imposed in combinations. Here is one example. Let the freemarket price and quantity be denoted by p* and q*. The government offers a subsidy of $1 per unit, but at the same time it restricts output to be no greater than Q. Draw this policy for two cases: Q<q*, and Q=q*. Indicate which areas on your graph correspond to the changes in producer surplus and consumer surplus. What is the net welfare loss from the policy in each case? (The net change in welfare is equal to the change in producer surplus plus the change in consumer surplus minus government expenditure on the subsidy).

3. Let the demand and supply for a good be given by

qd = 1600 40p , qs = 2p + 400 .

Suppose a tax of $10 is imposed on each unit bought. How much of the tax is borne by producers?


4. For the following supply and demand curves,

qd=1002p ,


Draw the following policies and complete the table:

(a) A price cap of $10 (b) A

Free Market Price Quantity Consumer Surplus Producer Surplus Government Revenue Total Welfare $20 60 $900 $450 0 $1,350

subsidy of $5.

(c) A quota of 50 units

(d) A tax of $10

5. Consider the following supply and demand curves: qd = a 3 p , qs = A + 2 p . If we impose a tax on the product, who will suffer the greatest burden of taxation, consumers or producers? Make sure you provide an explanation for your answer. 6. Given the following demand and supply equations,
qs = 5 p , qd = 110 6p ,

(a) What tax on consumers will attain an output level of 40? (b) What subsidy will attain a price to consumers of $8? 7. The supply curve in an industry is given by qs=p. The market equilibrium is p=50 and q=50. The government now wants to impose a quota of Q=20 in order to raise producer profits. What is the smallest price increase in price that must result from the quota so that profits are indeed increased. Draw this case.


8. Assume supply is perfectly inelastic. That is, regardless of the price, the quantity supplied is always equal to 100 units. (i) What happens to price when the willingness to pay of every consumer rises by $1? (ii) What is the welfare cost of a $1 tax? Illustrate your answers with a supply-demand diagram. 9. There are many producers, each of which has capacity of 100 tons and production costs of $10 per ton. (a) What are industry profits? (b) Which of the following policies raise industry profits? (i) A quota restricting production below the equilibrium level. (ii) A tax. (iii) A subsidy.

Neumark, D. and W. Wascher (1992): Employment effects of minimum and subminimum wages: panel data on state minimum wage laws. Industrial and Labor Relations Review, 46(1):55-81


Chapter 6 Messed-up Markets I: Externalities

In Chapter 4, we saw that a well-functioning market will, if left to its own devices, arrive at an equilibrium price and quantity that maximizes economic welfare. Producers, caring only about their profits, will choose to produce every unit of a good for which the price exceeds their production costs. Consumers, caring only about their own benefits, buy every unit of the good for which their personal benefit their valuation exceeds the price. The market brings these two self-interested groups together, and ensures that every unit for which the benefit from consumption exceeds the cost of production is made, and also that every unit for which the benefit from consumption is less the cost of production is not made. But not all markets function well. As a society we should care about all the costs of production, not just those that are paid by the producers. And we should also care about all the benefits or costs of consumption, not just those that are considered by the buyer. If there are some production costs that are not paid by the producer, then too much of the good will be produced. The producer will produce goods if costs she must pay are less than the price. Doing so implies that for some of the units produced the full social cost of production, which includes the costs paid by the producer plus the additional costs not paid, exceeds the price. As a society, we would prefer that these additional units not be produced. If there is a benefit to society from consumption beyond the benefit enjoyed by an individual buyer, then too little of the good will be consumed.


Individuals will buy all goods for which the private willingness to pay exceeds the price. Doing so implies that some units of the good for which the full social benefit exceeds the cost will not be bought, and hence will not even be produced. As a society we would like these additional units to be produced and consumed. Costs and benefits not paid or received by the individuals making the production or consumption decisions are called externalities, and markets with externalities lead to market failure. Such markets have equilibria in which the quantity produced does not maximize economic welfare. In these cases, society can do better by devising policies to move the quantity in the right direction. A positive externality is one in which there are social benefits beyond those captured by producers and consumers, and in this case the socially optimal output level exceeds the market equilibrium quantity. A negative externality is one in which there are social costs beyond those paid by producers and consumers, and in this case the socially optimal output level is less than the market equilibrium quantity.

1. Examples of Externalities
Externalities can take many different forms, and it is useful to take a look at a few examples now. We will then consider what sorts of policies might be helpful in improving economic outcomes. Example 1: Pollution. Figure 1 shows a typical daytime scene in Pittsburgh at the beginning of the 20th century. Even into the 1950s, workers changed their shirts at lunchtime because they had become so dirty from the air pollution, streetlights stayed on all day, and the Pittsburgh Steelers developed a wicked running game because they couldnt always see the ball if it was in the air. In October 1948 in nearby Donora, PA, twenty people died and 6,000 people out of a population of 14,000 were sickened over a five-day period by pollution emitted from the local zinc factory. Remarkably, the factory continued operating for two days after people began dying, until its owner, American


FIGURE 1. The Schoenberger furnaces, Pittsburgh, 1906. Source: Smith, A.G. (1990): Pittsburgh Then and Now. University of Pittsburgh Press

Steel and Wire, temporarily closed it. In response, President Truman convened the countrys first national air pollution conference in 1950. Federal involvement in air pollution had been established, leading ultimately to the Clean Air Act of 1963 (Kiester, 1999).24 For the residents of Donora, clearly too much zinc was being produced on those dark days in October 1948. The air cleaned up completely when the factory closed in 1957, but the economic malaise that followed its closure caused the town to shrink to a third of its 1948 population. Many residents of Donora today think there is too little production of zinc. How do we work out the right amount of production for a polluting industry? Figure 2 illustrates the analysis we must go through. Producers have costs, and


We usually think of pollution as a problem of industry. But, as the article on the next page illustrates, pollution is also a rural problem.


hence a supply curve, given by S. Their production costs do not take into


Massive cow manure mound burns for third month

Monday, January 31, 2005
A mountain of cow manure smolders at a feedlot near Milford, Nebraska.

MILFORD, Nebraska (AP) -- Urban dwellers who enjoy dining on filet mignon at five-star restaurants would probably just as soon not know about David Dickinson's dilemma. Bad for the appetite, you know. But Dickinson, who makes his living in the cattle business, has an environmental problem on his hands that is vexing state officials: a 2,000-ton pile of burning cow manure. Dickinson owns and manages Midwest Feeding Co. about 20 miles west of Lincoln, which takes in as many as 12,000 cows at a time from farmers and ranchers and fattens them for market. Byproducts from the massive operation resulted in a dung pile measuring 100 feet long, 30 feet high and 50 feet wide that began burning about two months ago and continues to smolder despite Herculean attempts to douse it.

ment out a couple of times." And still it burns. Wilma Roth, who manages a restaurant along Interstate 80 about a mile north of the feedlot, said her customers have complained about the smoke, which wafts for miles. "I'd just as soon forget about it," she said. Dickinson said the smoke is not particularly malodorous -- although that comes from a man who works full-time around manure. "I guess it's just all perspective," he said. "To me, it just smells like smoke. I really don't know how to describe it." Dickinson has an average of 12,000 animals on hand, each eating about 25 pounds of feed daily, resulting in as much as nine pounds of manure a day per animal -- some 54 tons every 24 hours. Most big feedlots spread the manure over 166

Price C S

c p
Deadweight welfare loss from free market

D Q q Quantity

FIGURE 2. Pollution costs reduce the socially-optimal output level below the free-market level.

The Nebraska Department of Environmental Quality has informed Dickinson that his smoldering dung pile violates clean-air laws and is working with him to find the best solution to extinguish it, said agency spokesman Rich Webster. Simply dumping water on the heap is not the answer, Webster said, because of concerns about runoff to any nearby water source. Dickinson first tried using heavy equipment to spread out the smoldering pile and extinguish the fire. "But the problem was, it started in another spot," he said. "We've also had the fire depart-

farm fields or compost it to spread later or sell commercially to gardeners. Farmers in several states are experimenting with using the methane gas from livestock manure to produce electricity. The manure is heated and produces methane gas as it breaks down. The gas is collected and used to power a generator, which sends electricity onto a power grid.
Dickinson acknowledged that while some folks see the humor in his predicament, he takes the fire seriously. "It's a nuisance, and obviously we are trying to get it resolved," he said. "Everybody's been really patient."

account the social cost of the pollution their factories emit. Thus, the market established an equilibrium quantity q. However, each unit of production


Price C S

pfloor t pcap

D Qquota Quantity

FIGURE 3. Policies that attain the social optimum.

causes a certain amount of pollution. Lets assume that the social cost of this pollution is equivalent to a cost c per unit of good produced. Then, the real unit production cost is given by the line C, and the socially optimal quantity is found at Q. The industry produces too much of the good, and it consequently emits too much pollution. The market equilibrium generates lower economic welfare than the sociallyoptimal output level. Every unit produced beyond Q involves a social cost that exceeds the benefit from consumption. Compared with the social optimum, the free market generates a deadweight loss indicated by the shaded triangle. To raise economic welfare to its maximum possible level, we need a policy that induces firms to produce no more than Q. In Chapter 5, we saw four distinct ways we could do that, and all of them are illustrated in Figure 3. First, we could impose a price cap at just the right level (pcap in the figure). Second, we could impose a price floor (pfloor). Third, we could impose a production quota


(Qquota). Finally, we could impose a tax per unit of production, indicated by t. All have the same effect on output. Note that a special feature of the tax is that it needs to be set at a level exactly equal to the cost, c, of the pollution. This has the effect of turning the social cost of pollution into a private cost of production, inducing firms to freely choose to produce the socially-optimal tax. When a tax is set in this way, we say that the externality has been internalized. Another name for the same tax is a Pigouvian tax.25 Example 2: Education. Universities invest considerable effort in attracting students with the highest SAT scores, who are involved in many extracurricular activities, and who come from diverse backgrounds. This is in large part because universities believe that such students confer benefits on their peers. Because these benefits are not captured by the student (say, in the form of higher wages), the social value of the education exceeds the private value. More broadly, society appears to believe that the social value of education exceeds its private value for virtually everyone. We uses taxes to provide virtually free K through 12 education, and most state colleges are heavily subsidized for all residents (see Table 1). 26
TABLE 1. Average Costs and Prices of a College Education, 1996 COST PUBLIC 4-YEAR $12,416 PRIVATE 4-YEAR $18,387 PUBLIC 2-YEAR $7,916


Arthur C. Pigou (1877-1959) is one of the most important figures of early economics. He was the first to show how governments can use taxes (and subsidies, as we will see) to internalize externalities.


You may have wondered why MBA students often pay more tuition than students studying for a Masters degree in, say, economics. Here is my theory: MBA students have no social value beyond what they earn in the form of higher wages. Thus, there is no justification for subsidizing them.


Price S
external benefit


pconsumer D q Q SB Quantity

FIGURE 4. A subsidy is appropriate when the social benefit exceeds the private benefit. TUITION PERCENTAGE SUBSIDY Source: ACE (1998). $3,918 68% $13,250 28% $1,316 83%

Figure 4 illustrates a market in which there are social benefits that exceed the private benefits. Individual willingness to pay gives the demand curve, D, but this leads to too little consumption, at q, when the true social benefit per unit is given by SB. The policy solution is to provide a Pigouvian subsidy equal to the unit value of the external benefit. The subsidy lowers the price paid by consumers and raises the price received by producers, inducing an increase in quantity to the optimal Q. Example 3: Vaccinations. Some types of externalities can induce much more awkward behavior than simply creating an equilibrium with the wrong quantity. Vaccinations against infectious diseases is a case in point. Vaccinations impose a cost on those who have them: they are inconvenient, scary for little kids, and scary to the parents who worry about the rare cases of 170

c a



Percentage of population immunized


FIGURE 5. Externalities associated with vaccinations can lead to instability in vaccination rates.

serious adverse reactions that result from some vaccinations. Vaccinations also incur a private benefit: they make it less likely that the recipient contracts the illness. In addition, there is a public benefit: a vaccinated person is much less likely to transmit the illness to others. An interesting feature of vaccinations is that the private benefit of having a vaccination is lower the greater the fraction of the population already vaccinated. When most of the population is vaccinated, the chances of an unvaccinated person getting the illness is small, so why would she bother with the private cost? Figure 5, which plots the private costs and benefits of vaccinations as a function of the fraction already vaccinated, illustrates the consequences of this type of externality. The private costs of vaccinations are constant, regardless of the fraction vaccinated. The benefits, however, decline as the fraction rises. Imagine we start out at point a, with, say, 30% of the population vaccinated. At this point, the benefits exceed the costs, so every person considering whether or not to get vaccinated chooses to do so. But when everyone makes


this decision, before you know it, 100% of the population is vaccinated, so society moves to point b. At b, however, people see that the benefits are less than the costs, so the next wave of people decide not to have the vaccination. As time passes, everyone rejects the vaccination option, and we move to point a. And from here, of course, b becomes the next outcome. Our simple consideration of the benefits and costs associated with vaccinations suggests that we should see cycles in vaccination rates. Figure 6 provides some data for the case of whooping cough (pertussis) in the United Kingdom. The left panel shows the massive decline in illness rates after the introduction of the vaccine. The inevitable consequence in the 1970s (which was further encouraged by publicity surrounding cases of adverse reactions) was that vaccination rates declined dramatically, from a high of 81% in 1968, to only 30% in 1977. This decline in vaccination rates induced an equally dramatic resurgence in cases of whooping cough, which rose from virtually zero in 1973 to almost 70,000 cases in 1977. With whooping cough again so prevalent, immunization rates rose again, reaching a high of 94% in 1995.


FIGURE 6. Pertussis (whooping cough) vaccination rates and disease incidence in the United Kingdom. Source: Wray and Sommerville (2005).

One way to prevent this cycling of vaccination rates is to impose vaccination requirements on certain segments of the population. To do so one needs a stick with which to enforce vaccination. In the United States, this has been done by linking vaccinations to primary school enrollment. For medical reasons, vaccination against whooping cough must be completed before a child is of age six. Therefore, by requiring that children entering school for the first time be vaccinated, the US has been able to maintain vaccination rates of around 95% for many years. Example 4. Network Externalities. A network externality arises when the benefit of having a product depends positively on the number of other users who have it. There are many examples in modern society. It is not much use having a telephone if no one else does. Microsoft Word dominates the word processor market mainly because it is more difficult to use any other word processor and share your files with colleagues. Figure 7 illustrates the costs and benefits of a product that has network externalities. For simplicity, we assume again that the cost, c, is the same


regardless of the fraction of the population that owns the item. The benefit, however, is increasing in the fraction with the product. If currently no one has the product, then the cost exceeds the benefit (point a). So no one considering the product chooses to buy and we stay at a. In contrast if enough people already have the product, then the benefit exceeds the cost. Then everyone choose to buy the product, and we attain an equilibrium at b. Note the stark contract with Example 3. When the benefit of something declines when the number that already has it increases, we end up with an unstable situation in which we fluctuate between no one having it and everyone having it. But when the benefit increases with the fraction already owning a product, we end up with two very stable equilibria. Either no one has it, or everyone has it. A firm selling a product with a network externality might do well to begin selling at a low price, say at p0. Once enough people have bought the product, it can raise the price, say to p1. The high price allows the firm to recover excess profits which might be more than enough to offset the initial sales at a price below cost. To do this, however, the firm must be able to control the price it must be able to prevent others from coming in at a late stage and selling at or near cost, c. This requires that the firm have some monopoly power, which we will study in Chapter 8. If the firm can not prevent others from entering the market, it will never sell below cost because it knows that it cannot recover its losses from higher prices later on. In such cases, the government must intervene if it wishes to see the product established in the market. It will have to temporarily subsidize the product until a sufficient number have it. Then, it can do away with the subsidy.


b p1
benefit cost

a p0
0 100

Percentage already owning product

FIGURE 7. Network externalities can create two stable outcomes.

Example 5. Lojack versus The Club. The lojack is a transmitter hidden in your car to help in recovery should it be stolen. If your car is stolen, the transmitter is activated, and police radio monitors may detect the signal. The club is a locking device for your steering wheel. It is a highly visible red color, and deters auto theft. Both of these anti-theft devices involve externalities, but one of them should be taxed while the other should be subsidized. Steven Landsburg, writing for Slate magazine, explains with an analogy to house burglar alarms:
When your neighbor installs a burglar alarm, thoughtful burglars are encouraged to choose a different target like your house. Its rather as if your neighbor had hired an exterminator to drive all the vermin next door. On the other hand, if your neighbor installs video cameras that monitor the street in front of your house, he might be doing you a favor. So the spillover effects of self protection can be either good or bad. Landsburg (1997)


The difference between the Lojack and the club is that the former is invisible to thieves while the latter is visible. When the Lojack is used in a city, car thieves do not know which cars have them and which do not. So when you buy a lojack and raise the fraction of cars in the city that have them, all cars receive additional protection. One might expect, because car thieves respond to incentives as well as any other person, that auto thefts decline. This is indeed the case. Studying data from 1986 to 1994, Ayres and Levitt (1998) found that a one percent increase in Lojack sales in a city reduced thefts by about 20 percent (see Figure 8). Thefts fell 50 percent in Boston as a result of the introduction of Lojack, and most of the benefit of the purchase of a Lojack accrues to strangers. Because there is a positive benefit to purchase of a Lojack over and above the benefit that accrues to the buyer, sales of the Lojack should be encouraged, perhaps by use of a subsidy. In contrast, the Club is highly visible. A car thief cruising a neighborhood is less likely to steal a car fitted with the Club, but he is more likely to steal the one next to it. If you use a Club, your neighbors suffer, and so its use should be discouraged, perhaps by use of a tax.

FIGURE 8. Auto theft and the Lojack. Source: Ayres and Levitt (1998).


2. Private Solutions to Externalities: The Coase Theorem

When externalities are present, there is a market failure. Our analysis suggests that the market, if left to itself, will not attain as an equilibrium the quantity of output that maximizes economic welfare, The implication is that we may need some government policy to adjust the equilibrium. But before we go that route, it turns out that the private sector, if left to itself, may have a trick up its sleeve. This trick is explained by a powerful theory developed in 1960 by Nobel laureate Ronald Coase. Its the 19th century, and the West has been won. Big capital has succeeded in laying down railroads from the East Coast to the West Coast. Farmers have established themselves in the Midwest, and corn covers the landscape. And now weve got a problem. As steam trains run from coast to coast, they emit sparks. And the sparks set fire to the corn, which naturally upsets the farmers. The railroads are not required to pay for the burnt corn, so they dont internalize this part of the cost of running trains. So it appears that too many trains are running. Should the government intervene and place a tax on trains? The Coase Theorem says not necessarily. If we can get the farmers and railroad companies together in a room, they might be able to work out an efficient solution for themselves. In particular, farmers could offer to pay the railroads not to run the trains. If the cost of the payment necessary to stop a train from running is less than the damage to the corn field when the train is running, then it is worth making such a deal and the train wont run. If the necessary payment exceeds the cost of the damage, then it is not worth making the deal, and the train will run. It turns out that farmers, making this calculation, will stop just enough trains and not too many -- so that the number of trains left running after the deals have been struck is exactly the number that maximizes economic welfare. Figure 9 illustrates the ideas behind the Coase Theorem. The supply curve, S, reflects the railroads operating costs. To this we must add the average damage 177



D q* q Quantity of trains

FIGURE 9. Maximum compensation farmers are willing to pay under the Coase Theorem.

caused by a train run. Let this amount be d, so that the true social cost of train runs are given by the curve SC, which lies above S by the distance d. As railroads dont have to pay the cost of damage to the corn, they choose an equilibrium number of trains indicated by q. The socially-optimal quantity is q*. For all the trains runs in excess of q*, farmers suffer a loss equal to the shaded area (which equals d x (qq*)). The Coase Theorem states that farmers will be willing to pay this amount no more and no less --as an inducement to the railroads not to run trains in excess of q*, and this payment will in fact achieve an output of q*. To see why this is the case, consider Figure 10. Farmers begin by considering whether to pay the railroads not to run the very last train, so that output will decline from q to q1. Should they do it? The answer is yes. The damage they suffer is equal to the vertical distance, d. The profit earned from the very last train is the vertical distance a. So farmers need only pay a to compensate the railroad, and they will gain in doing so by the amount da. Farmers will continue to make compensation to reduce the number of trains running as long as the profits that the railroads make from running them is no greater


Avoided damages in excess of compensation


S d a
Compensation for lost profit

D q* q q1 Quantity of trains

FIGURE 10. Compensation paid by farmers under the Coase Theorem.

than the damage caused. If output is restricted to q*, the profit from the last train is given by the vertical distance between the private cost curve and the demand curve (because the demand curve will determine the price), and this distance is given by d. Any trains to the right of q* generate less profits than d, so farmers are willing to pay the lost profits to avoid the damage. Any trains to the left of q* generate more profits than d, so farmers will not be willing to pay the lost profits to stop these trains running. Hence, by voluntarily making compensation payments, farmers will induce the efficient level of output, q*. You might think this is all unfair. After all it is the trains that are causing the damage, so why should farmers be forced to pay? There are two answers to this. First, the trains by themselves are not causing any damage. It is the juxtaposition of the railroad and the corn fields that result in damage. Thus the farmers, by having cornfields next to the railroad, are contributing just as much to the damage of the corn as are the trains. Second, if you still dont like the fact that the farmers must pay for the damage, as a policy maker you can change the rules. Government could pass a law stating that farmers have the right not to have their cornfields burned.


Then they could stop all the trains running simply by refusing to give permission. But what would the railroads do? They would offer compensation for damage in order to persuade farmers to allow them to run trains. As long as the profit from running a train exceeds the damage payment, then the railroad would be willing to make the compensation. If the damage exceeds the profit, then no payment will be made and the train wont run. You should be able to work out that, when railroads make their calculations, they will pay compensation to run all trains up to q*, and they will choose not to run trains beyond q*. The total amount of payment made to farmers is given by the shaded area in Figure 11. When the two parties can get together to negotiate payments, the Coase Theorem states that whatever the law is, the outcome is always the same: the socially optimal level q* will be arrived at through bargaining. The only role for the government is to decide who pays. This is an issue of equity. If the railroads are given the right to run their trains, then farmers must pay the shaded amount in Figure 9, and railroads make a lot of profit (what is the total profit earned by railroads in Figure 9?). If farmers are given the right not to have their crops burned, then the railroads pay the farmers the shaded area in Figure 11. Farmers are better off, but railroad profits are much lower (what are railroad profits in Figure 11?).

Limitations of the Coase Theorem. The Coase Theorem states that individual
participants in a market with an externality can attain the socially-optimal level of output by themselves as long as they can get together to bargain. In many instances, however, this will be impossible. For example, there are many farmers along a railroad, and any one of them could refuse to contribute to the compensation that must be paid to railroads. It is unlikely that the railroad would be willing to bargain with each farmer individually, because any one of them could become a holdout. Sometime, local governments can overcome this type of problem. For example, a city government could bargain with a local factory to reduce pollution in the




D q* q Quantity of trains

FIGURE 11. Compensation paid by railroads under the Coase Theorem.

citys river, and it could enforce collection from its citizens to pay for it by means of taxation. But who would the same city government bargain with about acid rain? How would they prove how much damage acid rain is causing? And how would they identify who is causing it? When there are too many actors involved for bargaining to be feasible, when there is too much uncertainty about what the true costs of damage are, or when it is difficult to determine who is causing the damage, bargaining is unlikely to work. In such cases, which may well be more common than cases when the tenets of the Coase Theorem are relevant, we may see continued need for government intervention.

3. The Policy Challenge: Measuring Externalities

We have already seen how the government can correct a market failure due to externalities with a judicious use of quantity restrictions (e.g. quotas), taxes, or subsidies. In each case, the policy challenge is to attain the output that would have been chosen had the producer paid the cost of the externality, or had the consumer reaped the benefit of the externality. But to do this, the government


has to know what the cost or benefit of the externality is. This is hard to do because, by their very nature, the value of externalities cannot be observed in the market. There is no market, and hence no price to observe. As a consequence, it has proved necessary for economists to devise novel schemes to measure the size of externalities. Only when this has been done can government intervene with the right policy. In this section, we briefly review some of the methods that have been developed. The message you will take away from this section is that measuring externalities is in practice difficult. Economists must often be inventive to find a way to measure them, and even then, they probably only come up with rough numbers. Different methodologies and assumptions used by economists, even when made in good faith, will produce different estimates of costs. This means in turn that policies will at best only be approximately optimal, and at worst different interest groups will be able to cite equally scientific-looking studies that support their own position and stump politicians trying to design a reasonable policy. The correct valuation of an externality is how much people are willing to pay to have a benefit or avoid a cost. How much are people willing to pay to avoid the consequences of airport noise? How much do people value a reduction in their risk of death in car accidents? How much do people value an extra acre of Everglades wilderness? If there were markets for these things, we could simply observe the demand curve and work out what the willingness to pay was. But there are no markets. After all, it is the fact that there are no markets for these things that causes the problem in the first place! Nonetheless, sometimes observable behavior does allow us to recover rough estimates of peoples willingness to pay. Consider the following examples: The expenditures people undertake to insulate their houses against noise provides a rough estimate of their willingness to pay to avoid the noise. This estimate is likely to be a lower bound on the true willingness to pay, because everyone that values noise-avoidance more that the cost will pay the cost. It is


likely to be a lower bound on the true willingness to pay also because there is some noise, such as in the back yard, that can be avoided with insulation. Even though the estimate from avoidance expenditures is likely to be on the low side of the willingness to pay, it can still be very useful. Lets say you are considering building a new airport. If the airport is located right next to the city, it will be worth more than if it is located a good distance away. But you know that 5,000 local households will suffer from the noise. You do a study and find that households located near existing airports around the country spend an average of $10,000 to insulate against noise. Thus, you know that it will be worth at least $50 million to avoid the noise. Hence, if you can solve the noise problem by relocating the airport for less than $50 million you know it is worth doing. You are thinking about whether to preserve some wetlands in South Florida. Developers have made the case that draining and developing the wetlands will be worth $10 million. To decide whether you should allow them to do so, you need to know the value of the wetlands in their present state. One way you can get some numbers is to document how much people are willing to spend to visit the wetlands. If collectively visitors spend at least $10 million, you know that the value of the wetlands must be worth more than that. It must be worth more for two reason. First, because people will travel only if the cost is less than what the visit is worth to them. Second, there are many people who do not travel to the wetlands but nonetheless would be willing to pay some money to preserve them. Tobias and Mendelsohn (1991) used a clever variation on the travel cost method to estimate the value of the Monteverde Cloud Forest Reserve in Costa Rica. If it were not a reserve, Monteverde could be used for agriculture. The Costa Rican government has 24 national parks, and it must show that they offer a demonstrable value in excess of their value in an alternative use. Tobias and Mendelsohn collected data from all visitors to the reserve during a given time period. From these visitors, they


collected data about where they were from. For each region in the country, they then calculated the park visitation rate by dividing the number of visitors from that region by the population of the region. When these visitation rates were compared with the cost of traveling from each region to Monteverde, the authors found a negatively sloped demand curve: the higher the travel cost, the lower the visitation rate. Figure 12 illustrates the demand curve they obtained from this exercise. Now consider a nearby region, with travel cost c0 and visitation rate v0. People from this area earned consumer surplus equal to the triangle ac0d. Visitors from a distant region paid a higher travel cost, c1, and had a lower visitation rate, v1. They earned consumer surplus equal to the area ac1b. Adding these consumer surpluses up for all the regions, Tobias and Mendelsohn found that visitors valued the park at $110,000 per year, equivalent to $11 per hectare. This exceeded the annual profits to be had from a hectare of agricultural land. Consequently, Monteverde was found to be more valuable as a reserve than as agricultural land.

Travel Cost a c1 b


D v1 v0 Visitation rate

FIGURE 12. Travel costs and visitation rates to Monteverde.


You are thinking about passing a law to limit smoke emission in a city. This will raise business costs and reduce economic activity. It will be worth it if people value the cleaner air more than they value lost economic activity. One way to get an estimate of peoples willingness to pay for cleaner air is to look at what they are willing to pay for a house that is in an area with clean air. Take two houses with essentially the same characteristics size, features, attractiveness of neighborhoods, crime rates, school quality, etc. but with different air qualities. If you have accounted for all the factors that explain house prices, the difference between the two prices must be attributable to the difference in air quality. Brookshire et al. (1982) conducted just this sort of analysis for house prices in various Los Angeles suburbs. They had data on the sales prices, house characteristics, and local air quality for 634 house sales during 1977. Air quality was measured by NO2 concentration. They found that, after controlling for differences in various features like house size, whether it had a swimming pool, school quality, crime, and so on, a 10 percent increase in pollution resulted in a 2.2 percent decline in the price of the house. Taking this estimate, average house prices, and mortgage rates, the authors concluded that an improvement in air quality from poor to average was worth about $50 per month to a household. Multiply this by, say, half a million households and you find that cleaner air in these neighborhoods is worth $25 million per month. If you can devise regulations and policies that raise air quality for less than $25 million per month, it is worth doing. What do you do if you cannot find any observable behavior with which to estimate peoples willingness to pay? The answer is easy: go ask them. Surveys aimed at extracting information about willingness to pay are known as contingent valuation (CV) surveys. CV surveys ask people about hypothetical situations, but respondents are assumed to indicate how they would behave were they actually faced with the choices being asked about. After extracting


individual responses form the survey, the answers are aggregated up to get an estimate of the value for the community as a whole. This particular example does not have anything obvious to do with externalities, but it is a nice example of how CV works. Whittington et al. (1993) used CV techniques to estimate how much Ghanaian households were willing to pay for improved sewerage systems. In many developing countries, sewerage systems have been built, but households could not afford to connect with them. The systems then go unused. Before cities undertake such enormous expenditure, it is therefore important to know how much people are willing to pay. Whittington et al. collected data from 1,224 households in the city of Kumasi. They asked how much households would be willing to pay for improved latrine systems. Table 2 provides the basic numbers. Column 1 lists the type of system a household already had. Columns 2 and 3 list, for each existing type of system, the average household willingness to pay for an improved latrine system. Two options were on the table. The first was what is known as a Kumasi ventilated pit latrine (KVIP), the second was a modern water closet (WC). The researchers were expecting households to be willing to pay a lot more for the WC than the KVIP. They turned out to be surprised. The valuations of the KVIP and the WC were virtually the same. Because the KVIPs were a lot less expensive to build than WCs, a tentative plan to build WCs was abandoned in favor of KVIPs. Note that the numbers in Table 2 do not refer to the social value of sanitation. Although that is an issue we may care about, Whittington et al. were concerned with the citys ability to recover the cost of building a new sanitation system by having enough people sign up to pay. Thus, what they asked for, and what they got, was the households perception of the private value of sanitation.


TABLE 2. Average willingness to pay for improved sanitation (selected results dollars per month) EXISTING SANITATION Bucket latrine Public latrine Pit latrine KVIP 1.13 1.55 1.23 WC + SEWER 1.24 1.66 1.26

CV surveys are useful ways to get at valuations when there are no better methods available. But one has to be careful. Sometimes, you get silly answers, because people do not understand, or think about, the questions. Ask people how much they would be willing to pay to clean up the Miami River. Some people will have in mind a simple dredging and trash removal operation. Others will imagine the warehouses being replaced with boardwalks, cycle paths, and restaurants. People will make very different statements of their willingness to pay, depending on what they imagine a cleaner river to involve. Sometimes, people get strategic. If they think they will have to pay more than their valuation if a project goes ahead they will choose to understate their true valuation. Those who think they will have to pay less than their true valuation will overstate their true valuation. Sometimes you get non-responses that can bias your results. Ask 100 Republicans how much they would be willing to pay to have Miami International Airport renamed the George W. Bush International Airport, and they might all say $15. Ask 100 Democrats and you might get 50 of them say $0, and the other 50 say bugger-off. If we only count those who give a number, we would conclude that the average willingness to pay is $10 [=(100x$15 + 50x$0)/150], when in fact a more accurate number would be $7.50 [=(100x$15 + 100x$0)/200].


4. Concluding Comments
In this chapter we have studied externalities as a source of market failure. In markets that have costs or benefits not paid or received by producers or consumers, the market will attain an equilibrium in which economic welfare is not maximized. If there is a positive externality, with social benefits not captured by consumers or producers, then the equilibrium quantity is less than the socially-optimal quantity. If there is a negative externality, with social costs not captured by consumers or producers, then the equilibrium quantity is greater than the socially-optimal quantity. This divergence between the equilibrium and optimal output levels provides a justification for policy intervention. In contrast to the examples of policy intervention seen in Chapter 5, an appropriate policy can, in the presence of externalities, raise economic welfare. However, such interventions may not always be necessary. The Coase Theorem states that if people affected by an externality can get together with those causing the externality, they may be able to negotiate a set of payments that allows the market to attain the socially optimal level of output. When such negotiations are feasible, the role of government is essentially to write law to as to define who must make the payments. This is a problem of equity, not efficiency. When the Coase Theorem does not provide a way out (because negotiation is impossible), government needs to design an appropriate intervention such as a Pigouvian tax (for a negative externality) or a Pigouvian subsidy (for a positive externality). To get the tax or subsidy right, the government must be able to measure the size of the externality. This can be quite difficult in practice. We saw a number of examples in which economists have tried to be inventive in measuring such externalities. Economists have many more tricks up their sleeves, but in all cases we must accept that at best we can only obtain rough estimates of the size of an externality, and hence only rough guidelines about the correct size of the tax or subsidy.


1. Demand and supply curves are given by

q d = 80 2 p , q s = 20 + 2 p .
In previous problems, you already calculated the free-market equilibrium price, quantity and welfare for this market. Imagine now that production emits pollution, and your analysts have concluded that every unit of output causes enough pollution to impose $4 of additional social costs on the community. (a) What is the socially optimal output level? (b) How much lower is economic welfare in the free-market equilibrium than if output were at the socially optimal level? (c) To attain the optimal level of output by means of different policies (i) What is the appropriate price cap? (ii) What is the appropriate price floor? (iii) What is the appropriate tax? (d) If it were possible for the polluters and community to negotiate an agreement according to the tenets of the Coase Theorem, (i) If the producers had the legal right to pollute, how much would the community pay the firms to restrict production? (ii) If the community had the right to clean air, how much would the firms pay the community to allow them to produce? 2. Continue to use the demand and supply curve given above. In many cases, the size of the externality is not a constant regardless of the quantity produced. For the following cases, calculate the socially-optimal quantity that should be produced, and the tax or subsidy that would attain this output level. (a) There is a positive externality in consumption. For the first 20 units consumed, there are social benefits beyond the private willingness to pay equal to $2. For each unit beyond 20, the external social benefit is $4.


(b) There is a positive externality in consumption. For the first 20 units consumed, there are social benefits beyond the private willingness to pay equal to $2. For each unit beyond 20, the external social benefit is $0. (c) There is a negative externality in production, equal to 10 percent of the private production cost. (d) There is a negative externality in consumption, equal to 10 percent of the private willingness to pay. That is for each dollar of willingness to pay, society pays a cost equal to 10 cents.

ACE (1998): Straight Talk About College Costs and Prices. Report of The National Commission on the Cost of Higher Education, Washington, DC: American Council on Education. Ayers, Ian, and Steven D. Levitt (1998): Measuring the positive externalities from unobservable victim precaution: an empirical analysis of Lojack. Quarterly Journal of Economics, 113(1):43-77. Brookshire, David S., et al. (1982) "Valuing public goods: a comparison of survey and hedonic approaches." American Economic Review, 72(1):165-177. Coase, Ronald H. (1960): The problem of social cost. Journal of Law and Economics, 3(1):1-44. Kiester, Edwin (1999): A darkness in Donora. Smithsonian Magazine, November. Landsburg, Steven E. (1997): Property Is theft: when protecting your own property is stealing from others. Slate.com , August 3. Smith, Arthur G. (1990): Pittsburgh Then and Now. University of Pittsburgh Press. Tobias, D., and R. Mendelsohn (1991): Valuing ecotourism in a tropical rainforest. Ambio, 20(2):91-93. Wile, Jay L., and Erica A. Sommerville (2005): Vaccines are Incredibly Effective at Preventing Disease. At http://www.apologia.com/vaccines/vac_effective.html. Accessed 25 February, 2005. Whittington, D., et al. (1993): Household demand for improved sanitation services in Kumasi, Ghana: A contingent valuation study. Water Resources Research, 29(6):1539-1560.


Chapter 7 Messed-up Markets II: Public Goods, Common Property Resources and Club Goods
Maria and Ana are roommates, and they are deciding whether to buy a new television. The value of the TV to each of them is $300, but the TV costs $400. Should they buy the TV? Of course, its a no-brainer. Each of them contributes $200, and each of them gains $100 in consumer surplus. Buying the TV makes them better off. But thats not what happens. Ana is an economics major, which makes her smart but selfish. Ana doesnt know for sure, but she suspects that the TV is worth $300 to Maria. Ana then claims that the TV is worth only $101 to her. If Ana is right, she calculates, Maria will agree to pay $299 and gain $1 in consumer surplus. But Maria is also an economics major and, just like Ana, she claims the TV is worth only $101 to her. In trying to increase their consumer surplus, both roommates end up offering to pay $101, and the TV goes unpurchased. After this failure, Maria and Ana come up with a voting scheme. If both vote to buy the TV, they split the cost. If only one votes to buy, then she pays the full cost. Here is Anas thought process. If Maria votes to the buy the TV, I have two choices. I can vote to buy, in which case I pay $200 and get consumer surplus of $100. Alternatively I can vote not to buy, and I pay nothing and get $300 in consumer surplus. If Maria votes not to buy, then I pay $400 and have consumer surplus of -$100 if I vote to buy, or I pay nothing


and have consumer surplus of $0 if I vote not to buy. No matter what Maria votes, I am better off voting No. Of course, Maria goes through exactly the same thought process and she also votes No. The TV is not bought, and Ana and Maria forego the consumer surplus they could have gained with the TV (see Figure 1).

MARIA VOTE NO ANA VOTE NO VOTE YES $0, $0 $100, $300 VOTE YES $300, $100 $100, $100

FIGURE 1. The payoff matrix for Ana and Marias voting problem. The left number in each cell is the consumer surplus payoff to Ana, the right number is the payoff to Maria. Ana and Maria will both vote No and end up in the upper left cell. But they would have been better off if they had both voted Yes.

Maria and Anas failure to buy a TV arises because, for them, the TV is what economists call a public good. If Ana buys the TV, she cannot stop Maria from gaining benefit from consuming it. And if Ana watches TV, it does not stop Maria from watching it as well. Using the buzz words of public goods. Consumption of the TV is said to be non-rival and non-excludable: Non-rival: The consumption of a public good does not diminish the ability of others to consume it as well. Non-excludable: If a subset of the population buys and consumes a good, the remainder of the population also get to consume it, even though they didnt pay anything.


Any good that is non-excludable and non-rival is said to be a public good, and every potential consumer of a public good, just like Ana and Maria, has an incentive to free ride on other peoples purchases. But when everyone free rides, no one actually buys the good, so it is never consumed. The problem of public goods is a special type of externality. My benefit from consuming a public good has spillover benefits to strangers that I dont take into account. In this case, I privately choose to buy too little of the good. It is a special type of externality because the size of the spillover for each free-riding stranger is just as large as my private benefit. This induces me to choose to go so far as to buy none of the good. Viewed as an externality, the only price that would induce me to buy any of the good is a zero price. This requires a 100% subsidy from the government. But, obviously, a 100% subsidy is equivalent to the government buying the good itself, and providing it free to the public. So the standard solution for the public good problem is for the government to provide the good. This is what it does with defense, for example. It decides how much defense to produce, and then uses the coercive power of government to collect revenues through taxation. The challenge for the public policy maker is work out how much of the good to produce. We will come back to this question later.

1. The private provision of public goods

Before the government steps in, it is worth checking whether the private sector has some means to provide a public good. In fact, there are many cases where philanthropists and other altruists provide public goods of one form or another, by donating for museums, donating blood, and so on. Economists have explored the motivations behind philanthropic acts, but we shall not do so here. Instead, we briefly consider two examples in which self-interested individuals have found ways to provide public goods.

The Lighthouse. For many years, economists used the lighthouse as a prime
example of a public good that has to be provided by the government. A


lighthouse warns of a danger or provides an aid to navigation simply by shining light out to sea. Any ship that passes benefits from the lighthouse. Its consumption is non-excludable: a ship passing by in open water can consume it whether or not it pays. It is non-rival: the lighthouses value for any one ship does not diminish because of the presence of another ship nearby. But Ronald Coase (1974) explained that the lighthouse was in fact not such a great example of a good that must be provided by the government. It turns out that in 19th-century Britain, many lighthouses were privately built and maintained. The trick to make this work required that owners of lighthouses directly tackle the problems of non-rivalry and non-excludability. Their solution was quite clever: collect tolls from ships that must have passed by the lighthouse from the ports they enter. The benefit to the port in collecting tolls on behalf of the lighthouse is that they will get more business if the lighthouse makes use of the port safer. The benefit to the lighthouse is that the port can refuse to allow ships that dont pay to dock at the port. Thus, private lighthouses solved the public goods problem by tying the value of consuming the services of the lighthouse to an excludable good.

Broadcast television. Another common example of a public good is broadcast

television. When a TV station emits its programs, it cannot exclude anyone from watching, and it cannot easily collect payments from its audience. In some countries, the government has provided the broadcast television, and collected the cost from the public with taxes. In other countries, broadcasters have found their own solution by tying the broadcast of a TV program to a good that is excludable and rivalrous, namely the ability to broadcast commercials. Advertisers want people to see their commercials without paying to see them. Broadcasters can exploit this because advertising time on a television station is excludable (you dont get to advertise without paying for the slot) and rivalrous (two firms cannot advertise in the same slot). Of course, if the value of advertising declines dramatically, broadcasters will lose their ability to receive payment for sending programs over the airwaves.


Not surprisingly, then, new technology that undermines the value of commercials has the broadcast industry upset. In October 2001, major US television networks sued the manufactures of ReplayTV, a digital video recorder with the feature that it automatically skips commercials. The following year, the CEO of Turner Broadcasting took an even harder line:
Because of the ad skips [TiVo is] theft. Your contract with the network when you get the show is you're going to watch the spots. Otherwise you couldn't get the show on an ad-supported basis. Any time you skip a commercial or watch the button you're actually stealing the programming. Jamie Kellner, CEO of Turner Broadcasting CableWorld, April 2002.27

2. The optimal quantity of a public good

In a regular market one for private goods all consumers pay the same price for the good, but they get to consume different quantities. Public goods are radically different. With a public good, everyone is consuming the same quantity. If the US government builds a missile defense shield, everyone in the US gets to consume exactly one missile defense shield. More precisely, if the government spends $100 billion on a defense shield, everyone consumes exactly $100 billion of defense. Given this difference, how do we work out what the optimal quantity of a public good is? The answer is, fortunately, intuitively simple. Figure 2 illustrates. The supply curve traces out, as usual, the cost of each unit of the good that might be produced. Imagine that there are two people, each of whom have their own willingness to pay schedule. Individual A, for example, is willing to pay $10 per unit for ten units, and $5 per unit for 20 units. Individual B is willing to
I wouldnt want you to think that Mr. Kellner was being mean: hell let you go pee. When asked if he considers people who go to the bathroom during a commercial to be thieves, he responded: "I guess there's a certain amount of tolerance for going to the bathroom.


Price D 30 B 20 15 10 5 A S




FIGURE 2. Deriving the optimal provision of a public good.

pay $20 per unit for 10 units, and $10 per unit for 20 units. If we add these individual willingness to pay curves, we find that society (consisting of these two people) is willing to pay $30 for 10 units, and $15 for 20 units. This social willingness to pay is found by vertically adding the two curves, and it is represented by the curve D. The optimal social quantity is found by equating the social willingness to pay with the demand curve. Thus, the optimal quantity of the public good that should be produced is 20 units.

Inducing truth-telling. In order for the government to work out this optimal
quantity, it needs to be able to measure everyones valuation of the good. This is difficult to do, because almost everyone has an incentive to lie! Imagine that four people live in a community, and they are considering a project to construct a streetlamp. Each person has a different valuation of the streetlamp, their valuations being $50, $30, $20 and $10. The total willingness to pay is $100, so it makes sense to go ahead with the project. But, it will be difficult to find out what peoples valuations are. If it is already decided that everyone contributes an equal share of the cost, then those whose valuation exceeds $25 have an incentive to exaggerate their willingness to pay in order to ensure the 196

project goes ahead, while those with valuations under $25 have an incentive to understate their true willingness to pay to make sure the project does not go ahead. If instead it is agreed that everyone pays an amount proportional to their true willingness to pay, then everyone has an incentive to underreport their valuations. Is there are anything that can be done to extract truthful statements? Economists have dreamed up some dastardly schemes to make it in everyones best interest to tell the truth. The most famous of these is the Groves-Clarke mechanism. It goes like this: The mechanism starts from the premise that over- or underestimation of ones true valuation does not matter if doing so does not change the decision. The only statements that matter are those that are pivotal: they change the outcome from doing the project to not doing it, or vice versa. The GrovesClark mechanism imposes a tax on pivotal people, equal to the cost that their valuation imposes on everyone else. Table 1 illustrates for our streetlamp example. Column 1 is the $25 cost share that each person will be required to pay if the project goes ahead. Column 2 is each persons true willingness to pay, and column 3 is each persons consumer surplus (i.e. willingness to pay minus cost share). The total consumer surplus is positive, so the project should go ahead. But would each person reveal his or her true willingness to pay?


COST SHARE (1) 25 25 25 25 100

TABLE 1. The Groves-Clarke Mechanism WILLINGNESS TO PAY CONSUMER SURPLUS (2) (3) 50 25 30 5 20 5 10 15 110 10

TAX (4) 15 0 0 0


Column 4 provides the tax that the Groves-Clark mechanism would impose on each person. To see where these come from, consider first individual A. If A is taken out of the decision-making process, the sum of the consumer surpluses of the remaining three people is $5$5$15=$15. That is, with A in the process the project goes ahead, with A out of the process the project does not go ahead. A is pivotal and so will be charged the tax equal to everyone elses consumer surpluses. B, C, and D are not pivotal. Drop B and consumer surplus falls to $5, which is still positive. Drop C and consumer surplus falls to $5. Drop D and total consumer surplus rises to $25. In each case the project will go ahead whether we drop one of these three or not. Why does this tax scheme induce truth telling? Consider A. She gets the project she wants, but has to pay a tax of $15. She still gets a surplus of $10 and she still wants the project to go ahead. If she overstates her willingness to pay, it would have no impact the decision is unchanged and so is her tax. If she understates to avoid the tax, she would have to reduce her stated valuation to less than $15 (so that the sum of everyones stated consumer surplus is less than zero with or without A). Doing this avoids the tax, but also makes her lose the consumer surplus of $10 she gets from the project with the tax. What about C? C would prefer the project did not go ahead. But to make this happen he would have to declare a consumer surplus worse than $15, so that the sum of all four stated consumer surpluses becomes negative. But doing this makes him pivotal. He then pays a tax of $15, equal to the sum of the consumer surpluses of the other three people. Thus, the only lies that change the decision enable C to gain $5 from not having the project, but only at the price of paying a tax of $15. You can work out that B and D also have no incentive to lie with the Groves-Clark tax mechanism. Devious scheme indeed. But governments dont use it in practice. First, they are often dealing with projects that affects thousands or even millions of people. To induce truth telling, you would have to ask every single person to


state their valuation. But in practice governments only ask a small sample of people and then use this small sample to estimate total population willingness to pay. Second, only a small number of people are ever going to be pivotal, and they will pay much more than others. In Table 1 for example, everyone except A pays $25 for the lamp, while A pays $40. This likely to be politically infeasible. So, the Groves-Clarke mechanism is not really a practical idea for most government problems. But it is popular in college classroom experiments. And who knows, maybe you can use it to get your roommate to tell the truth for once!

3. Common property resources

Public goods are nonrivalrous and nonexcludable. Some goods are nonexcludable but they are rivalrous. Imagine you are fishing in Biscayne Bay. You have three nice trout, enough for dinner, in your cooler. You catch yourself another one. Perhaps, you think, you dont really need this one. You can release it and let it grow. Next week, youll come back and catch it again, only then it will be bigger. The only problem is that there are another 99 boats out there fishing. If you return the fish, you will have only a one percent chance of benefiting from releasing the fish. Faced with such odds, many people choose to keep the extra fish. Your consumption of the fish is rivalrous. If you consume it, no one else can. But in an open water like Biscayne Bay, the chance to catch fish is nonexcludable because the fishery is a common property resource. A common property resource is one that can be exploited for private gain, but no one actually owns the resource. Anyone can go fishing, but no one actually owns the fish. Thus, there is little benefit to any individual from conserving the resource stock. Too little is conserved, and then everyone suffers. Figure 3 illustrates the overfishing problem using the language of externalities. The private cost of extracting fish is lower than the social cost. The difference is the impact that your fishing has on other peoples success rate. Because of this


Value and cost

Social cost Private cost

Deadweight welfare loss from overfishing

Private benefits Q q Harvest rate

FIGURE 3. Overfishing in a common property resource.

externality, individuals harvest too many fish, and total welfare is reduced by the indicated area, where social cost exceeds the private benefit. When everybody harvests too much the fish stock will be lower. In many instances, overfishing may be so severe that the stocks all but vanish. In 1996 the International Commission for the Conservation of Atlantic Tunas (ICCAT), issued a report on the demise of big-game fish tuna, marlin, sailfish and swordfish. Until 1960, most commercial fishing for these species was carried out by harpoon. A good boat could catch perhaps a dozen fish per day. But in 1961, the US fishing fleet began employing longlines. These are lines of baited hooks, up to 20 miles in length, that yield massive quantities of fish. The stocks of fish began declining almost immediately. As Table 2 summarizes, between 1961 and 1995, populations of major species declined by up to 88 percent. In 2001, the government banned longline fishing off the East Coast of Florida. Although the serious scientific analysis has yet to be done, sport fisherman already report a massive improvement in fish stocks since then. Outright bans are not always needed. Governments often use quotas to restrict the amount of fish caught or the number of boats fishing. Other government 200

have come up with ways to reduce the capacity of boats or raise their costs, for example by only allowing them to fish on certain days.
TABLE 2. Population declines in the Western Atlantic, 1961-1995. Blue marlin White marlin Sailfish 85% 88% 54% Swordfish Bluefin tuna 68% 88%

The decline of the fisheries stock is an example of the tragedy of the commons, a term first coined by Garret Harding (1968):
The [problem] is to be found in a scenario first sketched in a little-known Pamphlet in 1833 by a mathematical amateur named William Forster Lloyd (1794-1852). We may well call it the tragedy of the commons, using the word "tragedy" as the philosopher Whitehead used it: "The essence of dramatic tragedy is not unhappiness. It resides in the solemnity of the remorseless working of things." He then goes on to say, This inevitableness of destiny can only be illustrated in terms of human life by incidents which in fact involve unhappiness. For it is only by them that the futility of escape can be made evident in the drama.

Picture a pasture open to all. It is to be expected that each herdsman will try to keep as many cattle as possible on the commons. Such an arrangement may work reasonably satisfactorily for centuries because tribal wars, poaching, and disease keep the numbers of both man and beast well below the carrying capacity of the land. Finally, however, comes the day of reckoning, that is, the day when the long-desired goal of social stability becomes a reality. At this point, the inherent logic of the commons remorselessly generates tragedy. As a rational being, each herdsman seeks to maximize his gain. Explicitly or implicitly, more or less consciously, he asks, "What is the utility to me of adding one more animal to my herd?" This utility has one negative and one positive component. The positive component is a function of the increment of one animal. Since the herdsman receives all the proceeds from the sale of


the additional animal, the positive utility is nearly + 1. The negative component is a function of the additional overgrazing created by one more animal. Since, however, the effects of overgrazing are shared by all the herdsmen, the negative utility for any particular decision-making herdsman is only a fraction of - 1. Adding together the component partial utilities, the rational herdsman concludes that the only sensible course for him to pursue is to add another animal to his herd. And another.... But this is the conclusion reached by each and every rational herdsman sharing a commons. Therein is the tragedy. Each man is locked into a system that compels him to increase his herd without limit -- in a world that is limited. Ruin is the destination toward which all men rush, each pursuing his own best interest in a society that believes in the freedom of the commons. Freedom in a commons brings ruin to all.

4. Club Goods
Finally, we turn to goods that are excludable but that are, at least up to a point, nonrival. That is, you can charge people for consuming a good, but the cost to the owner of the good (and to other users) of allowing one more person to consume the good is essentially zero. Consider the example of a road. It is perfectly possible to charge drivers for using it this is common on freeways in Florida. But what do you charge drivers? Let us imagine that it costs $10 million to build a road. It costs $10 million to provide the first trip, while for every driver after the first, the cost is zero. Figure 4 illustrates. If the owner charges a price of $10 million dollars per trip, then the demand for trips will be zero. The road is built but no one uses it. If the owner charges a price of zero, then there will be Q trips but the owner has no way of recovering the construction cost. Perhaps the owner settles on a compromise: charge everyone the same price, and make the price just enough to cover the construction cost. Figure 4 plots out the average cost per trip. If there is just one trip, the cost is $10 million per trip, two trips and cost falls to $5 million. At q the cost per trip is $(10 million)/q, and at this 202

$10 million cost for first trip


Zero cost for each additional trip

a Average Cost q Q Number of car trips

FIGURE 4. Pricing road trips (I)

price, drivers demand exactly q trips. This way of setting the price, called average cost pricing, enables the owner of the road to recover its costs while drivers freely choose the amount of trips they want to make at the going price. But average cost pricing is unsatisfactory. There are a number of trips that might be made, namely those between q and Q, for which the benefit of the trip (i.e. the willingness to pay) exceeds the cost of allowing those trips to be made (i.e. zero). Average cost pricing forces us to give up social benefits equal to the area aqQ. The natural solution is to have the government build the road, fund it with taxpayer dollars, and then let everyone use the road free of charge. By this analysis, then, the Florida turnpike system is inefficient.

Congestion goods. But analysis of road problems is not so easy. In reality, each
person who uses the road adds to the cost of other users in the form of congestion and increased accident risk. These additional costs are not taken into account by each person contemplating making a trip, and so a government would want to charge appropriately. Moreover, it seems reasonable to believe that the cost a driver imposes on others is greater the greater is the number of drivers already using the road. This is for two reasons. First, with many users there are more people that a driver can hit (an


External Cost of Trip

p1 p2

Peak Demand

Off-Peak Demand

Number of car trips

FIGURE 5. Pricing road trips (II).

important consideration in Miami). Second, an additional driver is more likely to slow other drivers down when the road is already crowded, and when he does he is going to slow down more of them. How should we price in this case? We already know the principle of the answer: set a charge equal to the costs that a driver imposes on others. Figure 5 illustrates. The external cost of a trip is plotted as a positively-sloped line. It has a positive slope because the external cost is larger the greater is the number of drivers already on the road. The figure also illustrates two demand curves. During peak rush hours, demand is high. In this case, the correct toll to charge is equal to p1. In off-peak hours, the demand is much lower, and consequently the correct toll, at p2 is much lower.28 It may turn out that the costs of the road construction are fully covered by charging these congestion prices. When that happens, we can allow the


And as the Florida turnpike system does not adjust prices by time of day, by this analysis it is still inefficient.


private sector to build and operate a toll road, knowing that it can set prices efficiently. But here is the paradox. Other things being equal, we can assume that the more you spend on building a road, the lower the congestion will be. But the lower the congestion, the lower the efficient congestion prices, and less revenue is generated. Congestion pricing has been applied in many settings. Golf clubs charge less for a round of golf at odd hours of the day. My father, for example, often gets up to play golf in England at 4:00am (cheapskate that he is). Restaurants offer early-bird specials to anyone willing to have dinner at 5:00pm. Telephone charges are often lower during weekends and evenings. You will no doubt be able to think of many other examples of congestion pricing.

5. Concluding comments
In this chapter we looked at several different types of goods: Public goods are nonexcludable and nonrival. A market left to itself usually will not produce any of the good. In some cases, it is possible to tie consumption of a public good to consumption of an ordinary (private) good, and when this is the case private producers can be induced to supply the good. Common property resources are nonexcludable but rival. They are not owned by anyone but they can be exploited for private profit. The usual outcome when a market is left to itself is for each person to exploit too much. Policy is usually required to restrict the rate of exploitation of the resource. Club goods are excludable but nonrival. When a private firm tries to provide a club good, it usually charges too high a price (in comparison with the social optimum) in order to cover its costs. Too little is consumed.


Congestion goods are excludable and partially rival. In off-peak periods they may be nonrival, while in peak periods they become much more rival. Such goods require different pricing schemes at different times of the day.

1. In the main text, we summarized Ana and Marias public good voting problem with a payoff matrix,
MARIA VOTE NO ANA VOTE NO VOTE YES $0, $0 $100, $300 VOTE YES $300, $100 $100, $100

where the left hand number in each cell was Anas payoff and the right hand number was Marias payoff. We found that Ana and Maria would both choose to vote No. One way of thinking about why this is the case, is to realize that when both vote No, neither individual has an incentive to change her decision. Doing so will make her worse off. When this is true, economists say we have found the Nash equilibrium.29 Using this notion, identify the Nash equilibria in the following payoff matrices. These are not necessarily public goods problems, and there may be more than one or less than one equilibrium

MARIA VOTE NO ANA VOTE NO VOTE YES $0, $0 $5, $5 VOTE YES $5, $5 $0, $0


John Nash, he of book and movie A Beautiful Mind, developed the concept.


MARIA VOTE NO ANA VOTE NO VOTE YES $0, $0 $5, $5 VOTE YES $5, $5 $0, $0

We will return to the concept of Nash equilibria in Chapter 9. 2. Here are two individuals demand curves for a public good: Person A: q d = 20 2 p , Person B: q d = 40 p , and here is the supply curve:

q s = 20 + 2 p .
Calculate the optimal quantity of the public good that should be produced. What is each persons willingness to pay at this quantity? 3. Given the following numbers, calculate the Grove-Clarke tax for each person.


COST SHARE 100 100 100 100

WILLINGNESS TO PAY 95 120 80 110

Coase, Ronald H. (1974): The lighthouse in economics. Journal of Law and Economics, 17(2):357-376. Hardin, Garret (1968): The tragedy of the commons. Science. 162:1243-48.


Chapter 8 Messed-up Markets III: Monopoly Power

Although we have looked at a several types of market failures, we have continued to assume that there are in each case many potential suppliers of the good. This is a central feature of competitive markets. When there are many potential suppliers, no one firm can exert an influence on price. Each firm therefore simply compares the market price with its production costs and decides how much, if any, to produce. But many markets dont have this feature at all. Miami has only one cable TV company. Water and electricity are similarly provided by just a single company. If you want to buy an ipod, you dont have a large choice of potential suppliers there is only one. In other cases, there are just a few suppliers: automobiles manufacturers, airline companies, and so on. In yet others, there are many supplier of similar products, but somehow they all seem to charge different prices. For example, there are many companies selling fast food, but each of them sells a somewhat different range of items, and each of them has its own pricing scheme. In each of these cases, firms have the power to adjust the price at which they sell their products without suffering extreme changes in quantity demanded. If a firm in a competitive industry raises its price, it will lose all its sales to other firms. But firms in which there is only one, or there are only a few, sellers can usually raise their price without losing all their customers. If they can do so and raise profits, then we expect that such firms will charge higher prices than 208

they would if they had to compete in a market with many sellers of an identical good. It is no surprise that people tend to feel exploited most by companies that dominate their market. Markets that have only one supplier are called monopolies. Markets with a small number of buyers are referred to as oligopolies. And markets in which there are many suppliers of related goods, but in which the goods are sufficiently different for each firm to have some power over the price they sell at, are known as monopolistically competitive markets. In this chapter and the next, we will look at each type of market. We shall take a selective look at some features of markets in which firms have power to set the price they sell at. In this chapter we focus on the optimal pricing decisions for a firm. In the next chapter, we focus on the problem of strategic interaction. When there a just few sellers in a market, decisions made by any one firm affect the profits of other firms and the decisions they should make.

1. Maximizing Monopoly Profits

How does a monopolist decide how much to produce and at what price to sell a good? These are, of course, not independent questions. Once a monopolist has decided what price it will sell at, the demand curve will tell us what quantity will be sold. Similarly, once the monopolist has chosen its quantity, the demand curve will indicate what price is necessary to allow that quantity to be sold. So the monopolist still has only one choice to make, just as in a competitive market. But when a firm has monopoly power, we will see that the firm almost invariably chooses to sell a quantity less than would have been sold in a competitive market where producers have the same costs of production. Restricting quantity below the competitive equilibrium level causes price to be higher. What the monopolist loses from the reduced quantity is more than offset by what it gains from the higher price. The monopolists challenge is to find the quantity that maximizes its profits.


Price Demand

A pm B
(Monopoly profits)

C Cost





FIGURE 1. The monopolist creates profit by restricting supply

Figure 1 illustrates what is going on. We assume that the monopolist can produce each unit of output at a constant cost, as indicated by the horizontal line. The demand curve slopes downwards, as usual. In a competitive market with many producers, we know that the equilibrium quantity would be where the cost curve intersects the demand curve (at q*) and the market price would be p*. Because all firms have exactly the same costs, the competitive market supply curve is horizontal and no firms are marking any profits. Consumer surplus is given in the competitive market by the sum of the areas A, B, and C. We know from Chapter 4 that this equilibrium maximizes welfare. Every unit of the good for which consumer willingness to pay exceeds the production costs is being produced, and no good for which costs exceed the willingness to pay are being produced. The monopolist can generate some profits for itself by restricting output. If it reduces output to qm, the market price will rise to pm. The monopolist therefore earns profits equal to area B. Consumer surplus falls to area A, and economics welfare declines as area C is lost. Note that our analysis of the monopolists decision problem looks exactly like our analysis of a production quota (Chapter 5). In a competitive market, no one firm has the ability to raise 210

the price, but the government can create profits for producers by stepping in and imposing a quota. A monopolist does have the power to restrict quantity and raise price it simply sets a quota on itself knowing that its monopoly power prevents other firms from entering the market and undermining its manipulation of the market price. What a monopolist wants to know, then, is what quota it should impose on itself in order to make area B as large as possible. We analyze this problem first with a numerical example. Consider a monopolist that can produce as many units of a good as necessary at a constant cost per unit of $1. The demand curve is given by

q = 10 2 p .
Table 1 records for each possible output level the corresponding price such that consumers will demand exactly this quantity, total revenues, total costs, and profits.
TABLE 1. The Monopoly Profit-Maximization Problem Quantity 0 1 2 3 4 5 6 7 8 9 10 Price 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 Revenue 0.0 4.5 8.0 10.5 12.0 12.5 12.0 10.5 8.0 4.5 0.0 Total Cost 0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 Profit 0.0 3.5 6.0 7.5 8.0 7.5 6.0 3.5 0.0 4.5 10.0

As the table shows, the firms revenues are maximized when it restricts output to 5 units. However, its profits are maximized when it restricts output to 4


15 10 5

Total Costs

Total Revenues
0 -5 -10 0 2 4 6 8 10



FIGURE 2. The monopolists profit maximization problem.

units. Figure 2 illustrates this numerical problem graphically. The figure plots the data from the revenue, total cost and profit columns in Table 1. Profit is equal to the vertical distance between the revenue and total cost lines and is maximized when output equals 4 units. Is there a more convenient way than tabulating all possible profit levels to solve the monopolists problem? There is, if we think about marginal changes in profit. We call the change in revenues associated with a one unit increase in output the firms marginal revenue, and the change in cost associated with a one unit increase in output its marginal cost. For example, as the firm increases output from 4 to 5 units, its revenues increase by $0.50 that is, its marginal revenue is $0.50. However, its marginal cost is $1, As revenues increase by $0.50 while costs increase by $1, profits must decline. More generally, if marginal revenue is greater than marginal cost, the firm can raise profits by increasing output. If marginal cost is greater than marginal revenue, the firm can increase profit by reducing output. And yes, youve guessed it, profit is maximized when marginal revenue equals marginal cost. Figure 3 illustrates by plotting marginal cost (MC) and marginal revenue (MR) along with the demand curve (D) for this numerical example. Marginal


Price 6
5 4 3 2 1


0 0 2 4 6 8 10


FIGURE 3. The monopolists profit maximization problem, again.

revenue equals marginal cost when output is equal to 4 units. With output equal to 4 units, the market price is $3. Profit, indicated by the shaded area B, is equal to $8. Compared with the competitive equilibrium (output equal to 8 units and price equal to $1), the monopolist produces too little and charges too high a price from the perspective of maximizing economic welfare. Just as in the case of a production quota, the monopolist gains profits at the expense of consumers. The loss to consumers exceeds the gain to the monopolist so there is a net decline in total welfare. This is the market failure that monopoly brings about. Of course, we had asked previously whether there was a more convenient way to calculate monopoly output without tabulating the numbers for all possible output levels, and we have still not answered this question. What we need, and this should be clear from Figure 3, is an equation for the marginal revenue curve. Let us now derive that. Figure 3 shows two features for the marginal revenue curve. First, when quantity is zero, the demand curve and the marginal revenue curve intersect 213

the y axis at exactly the same point. Second, the marginal revenue curve intersects the x axis at the midpoints between zero and the point at which the demand curve intersects the x axis. The usual form for our demand curve is q = a bp , for some numbers a and b. and from this we can calculate that the demand curve intersects the y axis at a price p=a/b, and it intersects the x axis at a quantity q=a. So if the marginal revenue curve has the form MR=cdq, we know two points this curve passes through: (i) MR=a/b, q=0, and (ii) MR=0, q=a/2. We can use these facts to solve for c and d (I leave this as an exercise):


a , b


For example, in our numerical example, we have a=10 and b=2. Thus, the equation for marginal revenue is MR = 5 q . In our numerical example, marginal cost is 1. Hence, we need to find the output level such that 1 = 5 q , which is q=4. When q=4, we can now use the demand curve to find that the price must satisfy 4 = 10 2 p , and hence that p=3. Of course, this way to obtain the equation of MR involves little more than asserting the relationship it has to the demand curve. Let us now be a little more formal. The change in revenue resulting from an increase in output consists of two terms. The first is the effect of a decline in price on the q units already sold, which is caused by the increase in output. Denote this change in price by p, so the resulting change in revenue because of this effect is pq. The second term is the additional revenue from the increase in output, q; multiplying this by price gives the second term, p q. That is,
change in revenue = p iq + p iq .

The first of these two terms is negative because p < 0, while the second is positive because q > 0. Thus revenue may increase or decrease when the monopolist increases output. But of course we already knew this from Figure 2. Marginal revenue is the change in revenue resulting form a one unit increase in output. So set q = 1, to obtain 214

MR = p iq + p .
From the demand curve, if q increases by one unit, it is easy to verify that p falls by an amount 1/b (you should check why this is the case). Thus, p = 1/b. We also know from rearranging the demand curve that p = a b 1b q . Plugging these into the marginal revenue equation gives

1 a 1 MR = q + q

a 2 q, b b

which is what we arrived at before.

2. What Makes a Monopoly?

So a monopoly is a market in which a firm can prevent others from entering the industry. How does it get to have that power? There are a number of ways. We focus on two here. 2.1 Dominant technology There may be many firms capable of producing a good, but one of them has sufficiently better technology that its cost of production is much lower than all other firms. Figure 4 illustrates. The dominant firm both has a large capacity and a low production costs. If it acts as a monopolist, and sets marginal revenue equal to marginal cost, it will establish a price of pm. As long as the monopoly price is lower than the unit costs of other firms, no other firms will be able to enter profitably at this price. Thus the firm is effectively a monopolist even though there exist other potential producers.


A firm may also be the sole producer but be unable to charge the monopoly price. Figure 5 illustrates. Being the sole producers, the firm would like to


Lowest price at which potential competitors could begin to enter

Costs of potential competitors


Capacity of low-cost firm

MR qm
Marginal cost of low-cost firm

D Quantity

FIGURE 4. A monopolist with potential entrants.

Highest price that can be charged and prevent potential competitors from entering

pm pa

Costs of potential competitors

Capacity of low-cost firm

MR qm
Marginal cost of low-cost firm

D Quantity

FIGURE 5. A monopolist with potential entrants, again.


charge the monopoly price pm. But if it did so, then many other firms could profitably enter and take market share away from the low-cost firm. What should the firm do? The answer is straightforward, charge the highest price it can consistent with no other firms entering. This requires that it sets a price just below the unit production cost of its potential competitors, indicated by pa. In this case, even though the firm is the sole producer in the industry, it cannot act exactly like a monopolist because of the threat of entry by other firms. Markets in which the price charged by the low-cost producer is limited by the threat of entry are known as contestable markets. Note that if the cost advantage of the low-cost producer is small, it will end up setting a price very close to the competitive market price. In such markets, the welfare loss of monopoly is also going to be very small.

Patents. If having a better technology is going to allow a firm to be a

monopolist, it must be able to prevent other firms from copying its technology. This is the purpose of patent protection. In the United States, firms that develop a new technology can secure a patent that gives them the right to be the sole owner of the technology for up to 17 years. During this period, the firm can effectively act as a monopoly if the technology gives them sufficient advantage over potential competitors. The monopoly protection of course induces a welfare loss for the duration of the patent, so why would government do this? The answer of course is that developing new technologies costs money, and firms need to be able to recover the cost of development by earning profits. The firm cannot earn profits if other firms are free to copy its technology as soon as it is invented.30 So patents involve a tradeoff. On the one hand, the government wants to provide an incentive for firms to develop new technologies, because lower


If it were possible for every firm to immediately copy an invention, we would have a public good problem.


production costs raise economic welfare (you have repeatedly analyzed the welfare effects of a decline in production costs). On the other hand, the government would prefer not to suffer the welfare loss associated with monopoly power. It needs to balance these competing concerns, and in the US it has settled on allowing monopoly profits for up to 17 years for pretty much everything. If the patent system is a compromise, it is probably not the best way to stimulate new technologies. Michael Kremer of Harvard University (1998) has suggested an alternative mechanism that provides an incentive to invent but does away with the social cost of monopoly power in the market. He suggests that the government buys all the patent rights and donates them to the public domain. The purchases would be financed out of our taxes. How would the government work out how much to pay for a patented invention? It has an incentive to pay as little as possible but, knowing this firms would not invest enough in developing new technologies. Kremer suggests allowing other firms to bid for the patent rights, say using a Vickrey auction. The price arrived at in the auction is the price the government pays. We know in a Vickrey auction that the final bid price will be close to the valuation of having a patent as long as bidders bid their true valuations. But this approach begs another question. If the government is going to buy an invention and give it away, what incentive do private firms have to bid their true valuation? Kremer proposes his own compromise. Let the winning bidder have a chance of keeping the patent. Every patent invention is auctioned, and the winning bid and bidder identified. Then the government rolls a die. With, say a 10 percent probability the winning bidder gets the patent right to have a monopoly, and with 90 percent probability the government buys the patent. Thus, there will still be some monopolies, but at least there will only be 10 percent as many as in the current system.

John Harrisons Clock. Sailors navigating by the stars were easily able to
calculate their latitude, but longitude was a different matter. One possible way 218

of measuring longitude was to observe the time the sun rose or set. As a vessel sailed west, the sun would rise later and later, and by observing the time it would be possible to calculate just how far west the ship had sailed. The challenge was that there was no clock sufficiently accurate to do the job. Accurate pendulum clocks had been available since the early 17th century, but the rocking motion of a ship made them unusable for measuring longitude. In 1714, the British government decided to tackle this problem by offering a prize of $20,000 to the person who could develop a sufficiently accurate clock to be able to provide a ships longitude to within 40 miles or so after a voyage from London to the West Indies. To do so, the clock would have to maintain time accurate to within two minutes over the two-month duration of the voyage. Why would the government offer a prize rather than a patent? There are two possible answers. First, the social welfare that would be gained by having such a watch probably far outweighed the profits that could be made, so a patent system would not provide sufficient incentive to develop such a clock. Second, if the government owned the technology, it could more easily prevent other countries from gaining access to it, and this would help the British government dominate the oceans. The challenge was successfully met by John Harrison. He began work on his first model in 1730, and eventually produced a watch that met the governments accuracy requirements in 1753 (see Figure 6 remarkably it was a pocket watch). To test the clock Harrison's son set sail for the West Indies on 18 November, 1761. He arrived in Jamaica 62 days later, and the watch turned

FIGURE 6. Harrison watch, 1753.

National Maritime Museum, London.


out to have lost only 5.1 seconds. This would be the end of the story, except for one thing. The government committee responsible for awarding the prize, the Board of Longitude, refused to pay up. In fact, the Board only paid up twenty years later, after the king had ordered it to do so, and even then it paid only 8,750. The governments credibility in offering prizes to purchase inventions had been irretrievably damaged.

Malaria vaccines. Malaria kills 1.1 million people per year. Five hundred
million more suffer infection each year. Yet, until just a few years ago, not a single firm was attempting to develop a vaccine. The patent system was clearly not providing a sufficiently strong incentive. The reason is that people who die from and suffer from malaria are extremely poor. Although they stand to gain an enormous amount from having a vaccine, their income levels do not permit them to pay for it. Pharmaceutical companies, seeing little prospect of recovering their development costs, naturally choose not to pursue a malaria vaccine. The World Bank and some individual governments have stepped in to provide the necessary incentives. The World Bank has promised to spend billions on the purchase of a vaccine at market price. In November 2004, the British government announced an intention to buy 200 million to 300 million doses. Pharmaceutical companies must believe these promises are more credible than the Board of Longitude: by late 2004, about 80 candidate vaccines were being tested around the world.31 2.2 Natural monopoly A second reason that monopolies arise is because production of some goods requires a very large layout before any output can be produced, but then the cost of producing each additional unit beyond the first is quite modest. The first cost is known as a fixed cost. Industries with high fixed costs and low


This remarkable initiative was designed and promoted by Michael Kremer.


marginal costs are called natural monopolies. Examples include local utility services, where it is very expensive to layout pipes and cables but very cheap to provide increased services once the cables and pipes have been laid. In natural monopolies the unit cost of production is very much higher if more than one company pays the fixed cost of production. Imagine if a dozen water utilities each laid its pipes throughout Miami and then competed to provide services! Figure 7 illustrates. For each company there is a large fixed cost and a low, constant marginal cost. The average cost of production, which equals the sum of fixed and all marginal costs divided by the level of output is high when output per firm is small, and it declines when output is increases. The average cost of producing, say, output q is lower if one firm produces q rather than having two firms each produce q/2. The figure illustrates a case in which, at the equilibrium price that corresponds to an industry output level of q, two firms each producing q/2 make losses equal to area A each. In contrast, a single firm producing q makes a profit equal to area B. City governments understand this and so they usually license just a single firm

Price D
Average cost with two firms each producing q/2

Average cost with one firm producing q

A B Average cost MC q Quantity


FIGURE 7. Mergers are profitable in a natural monopoly.


to provide utilities. They use their regulatory powers to decide which firm gets to be the monopolist. Some city governments even decide to take ownership of the firm themselves. If an industry is a natural monopoly but it is not regulated, then one would expect to see mergers. At any given market price, the value of a single firm producing q exceeds the sum of the values of two firms each producing q/2. One firm will buy out the other and we would end up with a monopoly anyway. While it makes technological sense to have single provider when a market is a natural monopoly, there is still a problem for the city government. The monopolist is now free to charge a price such that marginal revenue equals marginal cost (this maximizes profits even when there is a fixed cost of production). But, as we already know, monopoly pricing involves a welfare loss, because output has been restricted below the socially optimal level. Figure 8 illustrates. The monopolist sets quantity equal to qm and a high price of pm results. But additional units could be produced at a marginal cost less than the consumers willingness to pay. The socially-optimal level of output is q*, with price set equal to marginal cost. The problem here is that price is now

D pm MR pR p* qm qR q* AC MC Quantity

FIGURE 8. The socially-optimal, regulated and monopoly price.


less than average cost, so the monopolist loses money on each unit sold. Before you know it the monopolist is bankrupt.

Subsidize the bastards. What can be done? There is a policy that allows the
regulator to attain the socially-optimal level of output. It involves providing a subsidy to the monopoly. This lowers its costs and allows it to produce even more. In the Figure drawn above, however, a subsidy alone would not be enough. Even if a subsidy of 100% were granted, so that the firm produces at the point where marginal revenue is zero, the resulting output level would still be far less than the socially-optimal output.32 The government can get more deeply involved in regulation by offering a subsidy and imposing a minimum output level. That is the government could require the monopolist to produce q* output, but then subsidize it to cover its losses. How big would the necessary subsidy be in Figure 8?

Average cost pricing. As you might imagine a subsidy to a monopolist is a

politically difficult policy to introduce. Most governments in charge of regulating natural monopolies are not willing to pay subsidies, so instead they make a compromise. They impose a price cap on the firm, and attempt to set this price cap equal to the point where average cost intersects the demand curve. At this price, indicated by pR in Figure 8, the firm produces qR, and it just covers its costs of production. The regulator has got as close as it can to the socially-optimal output level without inducing financial losses for the firm.


The problem set at the end of the chapter ask you to draw a case where a subsidy of less than 100 percent would succeed in inducing an output level equal to the social optimum.



Shaded area equals revenue to be obtained from fixed part of the pricing

AC Price per unit q* MC Quantity

FIGURE 9. Socially-optimal two-part pricing

Two-part pricing. Another possibility for the regulator is to allow two-part

pricing. Two-part pricing entails customers paying a fixed fee if they consume anything at all, and an additional fee per unit consumed. This is a very common form of pricing for utilities, such as electricity and water. The fixed fee covers the fixed cost of producing the service, while the per unit cost should be close to (ideally, equal to) the marginal cost. Figure 9 illustrates socially-optimal two part-pricing.

Decreasing block rates. Figure 10 illustrates the demand curve for an individual. Decreasing block rate pricing involves charging a high price, p1, for the first q1 units, and lower prices for subsequent units. The closer the cheapest block price is to marginal cost (if it kicks in before quantity q*), the
closer consumption gets to the socially-optimal level. The higher prices charged for the first units consumed cover the fixed costs of production. Consumer surplus under decreasing block rates is given by the sum of areas A, B, and C. One way to think about decreasing block rates is as a generalization of two-part pricing.



A p1 p2 p3 q1 q2 B C MC q3 q* Quantity

FIGURE 10. Decreasing block rates

3. Price Discrimination
Just when you thought monopolies could earn enough profit by virtue of being the only producer, it turns out that they can often do even better. When you travel by air, the chances that you paid the same price for your ticket as the person sitting next to you is pretty close to zero. Airlines charge different prices depending upon when you book your flight (see Table 2), how you book your flight, whether you insist on a direct flight, what time you want to go, whether you stay over on a Saturday night, and so on. You dont pay less for staying over on a Saturday night because it costs airlines less to transport you out on a Thursday and back on a Tuesday than it does to take you out on Tuesday and return you on Thursday. You pay less because your willingness to stay over a Saturday night identifies you as a cheapskate. It is not cheaper for American Airlines to send you to New York via New Orleans, but if you let them they will charge you less than for a direct flight. Your willingness to


book an indirect flight has pegged you as a customer with a low opportunity cost of time.

TABLE 2. Miami to New York, For travel . . Price Same day $724 Next day $464 In 48 hours $234 In 14 days $179 In 21 days $104 Cheapest available one-way flight offered by American Airlines, according to prices quotes given on 27 March, 2005 by Expedia.com.

On the other side of the same coin, people who book a flight for the very next day are usually those who have some urgent reason to travel. Travelers only willing to take a direct flight probably value their time more than those who are willing to put up with a connecting flight. Such travelers usually have a high willingness to pay, so airlines charge them more. Charging different types of consumers different prices for essentially the same service is called price discrimination. Doing so enables a firm to increase its profits. The challenge for monopolists is to identify these different types of consumers.33 But because price discrimination can be very valuable, companies have engaged in a variety of inventive schemes to identify customers with higher willingness to pay.


American Airlines does not have a monopoly in flying from Miami to New York, but it does have a monopoly over the particular flights that it makes. Other firms are capacity constrained so they cannot take all of AAs customers. Thus, AA has the ability to adjust its price.



pT pT+F p F MC


FIGURE 11. Price discrimination by Disney World (I)

Disney World. Friendly though he looks, Mickey Mouse is a rip-off merchant.

Florida residents get a 10 percent discount on a regular day pass. Why is this a rip-off? Because Disney World does it to increase profits. Figure 11 illustrates. There are two types of customers: people who are from out of state on vacation and locals. Given all the other costs of taking a vacation, and the fact that there is not much to do in Orlando if you dont visit the attractions, tourists have a much higher willingness to pay than Florida residents. The demand curve and marginal revenue curve for tourists are indicated by DT and MRT respectively. The demand and marginal revenue curves of Florida residents are indicated by DF and MRF respectively. Treating these as separate demands, Disney charges prices of PT to tourists and PF to Florida residents. If Disney could not distinguish between these groups it would have to charge a common price. This is found by horizontally summing the two demand and marginal revenue curves to get the aggregate demand and marginal curves, indicated in bold. The common price Disney World would charge lies between the two prices already found. Profits are higher when Disney can charge separate prices. How do I know this? Consider Figure 12, which plots profits earned from Florida residents and


Profits b




pT Residents

Price charged

FIGURE 12. Price discrimination by Disney World (II)

tourists at each possible price. Charging tourists PT maximizes profits that can be extracted from them (it is the price at which the marginal revenue form tourists equals the marginal cost), and this yields profits equal to the height of b. Charging residents PF maximizes profits that can be extracted from them, and this yields profits equal to the height of a. Any different price yields a lower level of profits, and this is captured by the inverted u-shaped profit curves. When a common price is charged, it is higher than PF and lower than PT. Florida residents are being charged too much to maximize profits from them, and tourist are charged too little. Profits at this common price are given by the sum of c and d. Clearly, c+d is less than a+b.

Offer discounts, dont charge extra! Disney price discrimination only works if
it can recognize the different groups. If Disney announced a price for residents, and then said that it would raise prices by 10 percent if tourists show their out of state driver license. Obviously tourists would affect a Florida accent (if there were such a thing) and get the lower price. But by offering a discount to Florida residents, it induces residents to reveal themselves. Non-


residents by virtue of not being able to prove they are residents have thereby revealed themselves as tourists who merit getting ripped off.

Make things worse! You are a manufacturer of laser printers. You can produce,
say, 1,000 a month, and you are selling at $500. But at this price you are only selling 500 machines per month. Your warehouse is filling up with unsold laser printers. How can you sell more? Lower the price, of course. But the trouble with doing that is that some people are buying at the current price. If you lower the price you will gain from the extra quantity sold, but you will lose from the price reduction. What you need to do is find a way to separate buyers willing to pay $500 or more for a printer from those only willing to pay less than this. You cant ask people, because they wont tell you. But what you can surmise is that some people have a high valuation because they print a lot, and hence wont be satisfied with a slower machine such as an ink jet. So heres what you do. Mess with some of your laser printers and make them slower. Customers with a high willingness to pay will still buy the laser printer at the high price. Customers with low willingness to pay will pay for the slower machine, which you offer at a reduced price. This is not a fictitious example. The IBM Laser Printer E of 1990 was exactly the same as the IBM Laser Printer, except that engineers had gone in and added a chip to slow the machine down. The E printer cost more to produce by virtue of having the extra chip, but it printed at half the speed and sold at a substantial discount. Selling an unnecessarily inferior commodity to enable price discrimination is an old practice. Jules Dupuit wrote of railroad pricing in the 19th century:
It is not because of the few thousand francs which would have to be spent to put a roof over the third-class carriages or to upholster the third-class seats that some company or other has open carriages with wooden benches. What the company is trying to do is to prevent the passengers who can pay the second class fare from traveling third class; it hits the poor, not because it


wants to hurt them, but to frighten the rich. And it is again for the same reason that the companies, having proved almost cruel to the third-class passengers and mean to the second-class ones, become lavish in dealing with first-class passengers. Having refused the poor what is necessary, they give the rich what is superfluous. Quoted in Ekelund (1970).

The ubiquity of price discrimination. Despite its long history, Odlyzko (2003)
has argued that the internet is encouraging more and more price discrimination as firms are increasingly able to identify different types of consumers. Dell charges different prices for exactly the same computer if you are dumb enough to declare yourself as a small businessperson rather than a health-care worker. In 2000, Amazon.com was caught red-handed charging different consumers different prices for exactly the same DVD.34 Priceline.com makes its money by persuading people to declare their willingness to pay, and then discriminating against them for doing so. To see this, think about how Priceline works. A supplier tells Priceline it has a certain amount of a good to sell, and what the minimum price is that it will accept. Anyone that bids above this minimum price gets the good, by paying what they bid. Anyone who bids below gets rejected. Here is a bricks-and-mortar analogy: Imagine you were to go in to the store and see a nice jacket (for the rare ones among you that wear jackets) for $49. As you are at the checkout with a $50 bill in your grubby paw, you mention to the store clerk that you really like the jacket and you would have been willing to pay $75 for it. The clerk then charges you $75. Youd be pissed, no? But if you play Pricelines game, this is exactly what you are doing every time.35


Companies practicing price discrimination on the internet have come up with a nice euphemism: they call it dynamic pricing.


Priceline tactics are discussed by Carnahan (2000). Carnahan quotes a Consumer Reports observation about even more interesting price discrimination: We successfully bid the


Canadian Drugs. In recent years, there has been a vociferous debate about
importing drugs into the United States from Canada. Most drugs sell for a markedly lower price in Canada, and proponents of free importation argue that it will allow our elderly to considerably reduce their expenditure on drugs. Opponents argue disingenuously that drugs imported from Canada are unsafe. The whole debate is pointless. Because it is currently illegal to import all but the smallest quantity of drugs from Canada, pharmaceutical firms can treat the US and Canadian markets as entirely separate, and they can price discriminate. The profit maximizing behavior depends on demands in the two countries, but it is clearly the case that it has proved optimal to charge a lower price in Canada. If free importation is allowed, prices in Canada and the US must converge. But US demand is more than ten times larger than Canadian demand. So the common price that will be charged will be dominated by US demand conditions, not Canadian demand conditions. At best, US consumers could see very small reductions in prices, while Canadians will see massive increases. So what do you think the Canadian government will do if the US government allows imports? Ban exports, of course.

Higher costs or price discrimination? If two people are being charged different
prices for what appears to be the same item, is it always price discrimination at work? Perhaps not. Perhaps it simply costs more to sell the good or service to one customer than it does to another, and as outside observers we are having a

minimum for everything on our list, saving more than 40% overall. . . Four days later, when our reporter again bid the minimum for the same items, the offers were uniformly rejected. At the same time another staffer, registering as a first-time user, got the entire market basket for the minimums. The implication is the Priceline believes that returning customers have a higher willingness to pay than new customers. I can think of a reason why Priceline would believe this. Can you?


Taken to the Cleaners?

Nobody can explain why laundries charge less for men's shirts than for women's. Steven E. Landsburg Slate.com. Posted Friday, July 3, 1998.T dry cleaner charges $1.65 to clean and press a man's shirt and $5.25 for a woman's blouse. What's going on here? The laws of arithmetic allow only two possibilities. Women's clothing must be associated either with higher costs or with higher profit margins for the dry cleaner. Unfortunately, neither theory seems terribly plausible. Let's start with the "higher cost" theory. In its most naive form, this theory predicts that if I move the buttons on my dress shirts from the right side to the left, the cost of laundering them will more than triple. That one's not going to fly. So, to give the theory a fair chance, we have to look for more significant differences between men's and women's clothing. Well, like what? You could argue that women's clothing is typically made of more delicate fabrics than men's. But if that's the relevant factor, why don't dry cleaners just quote different prices for different fabrics? (For some materials, such as silk, they typically do quote separate prices. The question is why this practice does not completely displace that of distinguishing between men's clothes and women's.) An alternative version of the theory is that women's clothes are costlier to process because women demand higher quality work. I can't disprove that version, but I have no real evidence to support it, either. So, in a search for better alternatives, I called three different dry cleaners and asked for their explanations. The first said that men's shirts are machine pressed, while women's are hand pressed. That left me wondering why they don't simply quote different prices for different kinds of pressing. The second said that women's shirts require specialized treatment because they are typically doused with perfume. That left me wondering why men who use after-shave are not chronically dissatisfied with their dry cleaners. The third said that this was their pricing


policy, and if I didn't like it, I was free to shop elsewhere. In the absence of a clear, convincing story about gender-specific costs, let's see what kind of story we can tell about gender-specific profit margins. In other words, let's ask whether my dry cleaner is exploiting female customers through higher markups. To make sense of that theory, you have to ask why dry cleaners would want to discriminate specifically against women, as opposed to, say, men. That strategy makes sense only if men are more price-sensitive than women and hence more likely to walk away in the face of a high markup. But why should men be more pricesensitive? You could argue that men are less diligent about cleanliness and so more likely to respond to high prices by wearing unlaundered shirts. But as long as we're dealing in stereotypes, you could argue equally well that women are more willing to do their own laundry--in which case women would be more likely to walk away from a high price, and it would make more sense to discriminate against men. So it isn't clear which gender is the more natural candidate for getting soaked at the cleaners. But there's a more fundamental reason to doubt that either gender can be victimized by price discrimination, and here it is: There are over half a dozen dry cleaners within easy walking distance of my house. If they're all earning higher profits on women's blouses than on men's shirts, why hasn't any of them decided to specialize in women's blouses? Let me make that more concrete. Suppose the going prices are $1.65 for a man's shirt and $5.25 for a woman's blouse, even though (under the theory we're currently entertaining) they are equally expensive for the cleaner to handle. Then if I were a dry cleaner, I would announce a uniform price of $5 for all shirts and blouses


--thereby attracting all the women's business and none of the men's. Because nobody has adopted that obvious strategy, we should suspect that despite appearances, the profit margin on women's clothing can't be much higher than on men's. In fact, the process wouldn't stop there. As soon as I announced a uniform price of $5, my neighbor would announce a price of $4.75. Ongoing competition for the (temporarily) more lucrative women's business would quickly eliminate any profit differential. That argument rests on the fact that dry cleaners are highly competitive. If Microsoft ran the entire dry cleaning industry, it might very well choose to discriminate against women (or men, depending on market conditions). But in the world we live in--or at least in the neighborhood I live in--there are so many interchangeable dry cleaners that none of them should be able to get away with exploiting anyone. One of my colleagues' wives insists I've got this wrong--she says she's so loyal to her own dry cleaner that no discounter can lure away her business. If most customers are as devoted as she is, then each dry cleaner is like a mini-Microsoft, with its own captive customer base. In that case, price discrimination can survive. But I am instinctively skeptical that many customers are as fanatically loyal as my colleague's wife. The theory that only a monopolist can price discriminate is standard textbook fare, and it's borne out by a lot of observations. Movie theaters have a certain amount of monopoly power (on a given night, a given moviegoer is likely to have a strong preference for a particular movie at a particular theater), and they price discriminate by offering discounts to senior citizens (which is equivalent to discriminating against everybody under the age of 65). Airlines have even more monopoly power--once you know where and when you want to fly, you are likely to have an extremely

limited choice of airlines--and they heavily discriminate against business travelers by charging more for midweek flights than for weekend flights (when most travel is for leisure). By contrast, in the most competitive industries, there is no price discrimination. As I am fond of pointing out to my students, you've never heard of a wheat farmer who offers senior citizen discounts. Likewise for gas stations, which are ubiquitous and sell to everyone at a single price. Well, at least that's what I used to tell my students. But I might have to make a small change in my lesson plan. The gas station nearest our campus has just announced a policy of senior citizen discounts on Wednesday afternoons. Is this price discrimination in favor of seniors, or does it reflect a genuinely lower cost of serving them? If you push me hard enough, I can probably concoct some kind of story about lower costs. Maybe seniors tend to drive cars with bigger gas tanks, so they buy 20 gallons at a time instead of 10, thereby saving on the cost of processing credit cards. (A significant part of that cost is the time spent waiting for the card to be approved, during which the pump is unavailable.) But if this cost saving is significant, why has only one local gas station recognized it? And why is it significant only on Wednesdays? I have suggested to my colleagues that none of us should be permitted to present ourselves to the world as economists until we figure out what this gas station is up to. Nobody has risen to the challenge. A few have suggested that perhaps the gas station owner is just a little quirky. Maybe that's right. But it would be far harder to believe that the entire dry cleaning industry is just a little quirky. Either there is enough monopoly power to sustain price discrimination, or there is some reason why women's clothes are incredibly expensive to clean and press. But I have no idea which.


hard time identifying what this extra cost is. In the reading on the previous page, Steven Landsburg puzzles over dry cleaning charges. He notes, correctly, that price discrimination requires some degree of monopoly power, so the most likely explanation for charging so much more for womens blouses than for mens shirts is that the former cost more to clean. But Landsburg is skeptical that costs can differ so much. Miami-Dade County has decided to sit on the fence with dry cleaners. MiamiDade Countys Gender Pricing Ordinance and its Dry Cleaning and Laundering Ordinance prohibit businesses from charging different prices for goods and services based solely on the customers gender. So far so good (for women, or the men that wear their blouses). However, if you want to exercise your rights, youd better be a darn good accountant. The ordinance states that a business is permitted . . to charge a different price if the goods or service involve more time difficulty or cost. In other words, consideration must be given to the quality and complexity of the goods and services to determine whether or not you have been discriminated against. The county allows you to collect damages only if you can prove discrimination. Apparently MiamiDade County expects you to have the expertise to do this.

1. The production of good X creates an externality. The following questions are based on the graph, which shows the marginal revenue, demand, unit private cost and unit social cost associated with production of good X.
13 12 social

(i) Is the externality negative? Explain.



7 4 Q1 Q2 Q3

private unit cost

(ii) Suppose the good is produced by a monopolist. Should output be taxed or subsidized? What is the per-unit dollar value of this tax or subsidy?


(iii) Suppose the good is produced by a perfectly competitive firms, Should output be taxed or subsidized? What is the per-unit dollar value of this tax or subsidy? 2. A monopolist faces demand q = 200 bp and its unit cost is $20. What price does the monopolist charge? 3. A monopolist has a marginal cost of $22, and faces a demand curve of qd=2807p. (i) Draw the monopolists profit maximization problem and solve it algebraically. (ii) What subsidy is necessary to induce the monopolist to produce the socially optimal level of output? 4. A monopolist faces a demand curve given by q d = 80 2 p and a constant marginal cost of $30. (i) What is the monopolists profit-maximizing level of output and the monopoly price? What are profits and consumer surplus at this equilibrium? (ii) What percentage subsidy is needed to induce the monopolist to produce the socially optimal level of output? When this subsidy is given, what are monopoly profits, consumer surplus and government expenditure? (iii) How much larger is total economic welfare in case (ii) compared with case (i)? 5. A monopolist faces a marginal revenue curve of MR = 50 4q . If the monopolists unit cost of production is $10, what is the monopoly price? 6. A regulated natural monopoly has a fixed cost of $100 and a marginal cost of $10. It faces demand q d = 80 2 p . If the regulator chooses to set a price cap equal to average cost, what should this price cap be? (Hint: you will need to make use of the quadratic formula for this one). 7. A monopolist has a unit production cost of $10. The firm has identified two distinct groups of customers. Group A has a demand curve of q A = 100 4 pA , while group B has a demand curve of q B = 60 3pB . (a) Which group pays the higher price if the monopolist price discriminates? (b) If the monopolist is not able to price discriminate, what common price does it charge?


8. A pharmaceutical company can sell its drug separately to customers in the United States and in Canada, because trade in drugs in prohibited. Canadian demand is q c = 100 2 p while US demand is qus = 2000 20 p . The marginal cost for the drug is constant at $10 per dose. (i) What prices are charged in each country? What quantities are sold? (ii) Assume now that the US allows trade in the drug. What common price will be charged for the unified market? (iii) What is the percentage increase in economic welfare increase in the US (you may assume that the pharmaceutical company is American, so that profits from sales in both countries count as a contribution to US welfare)? (iv) What is the percentage decline in Canadian economic welfare (which consists only of Canadian consumer surplus)? 9. For each of the following decide if the pricing scheme is price discrimination, if there are differences in costs, or if there is some other reason to for the scheme. In some cases you should be able to provide a case for price discrimination and a case for differential costs. (i) Senior citizen discounts at gas stations. (ii) Early bird specials at restaurants. (iii) Ladies free on Mondays (at a South Beach nightclub). (iv) No cover for students (at a local bar).

Carnahan, Ira (2000): The economics of Priceline. Slate. http://slate.msn/id/82827. Ekelund, R.B. (1970): Price discrimination and product differentiation in economic theory: an early analysis. Quarterly Journal of Economics, 84:268278. Kremer, Michael (1998): Patent buy-outs: a mechanism for encouraging innovation. Quarterly Journal of Economics, 112:1137-1167. Odlyzko, Andrew (2003): Privacy, economics, and price discrimination on the internet. Manuscript: Digital Technology Center, University of Minnesota.


Chapter 9 Messed-up Markets IV: Oligopoly and Games of Strategy

In Chapters 4 through 7 we assumed there are many firms in each market. From the perspective of each firm, it is as though all the other firms have already collectively arrived at the market price in conjunction with the forces of demand. Given this market price, the firm simply has to decide whether to produce and, if so, how much to produce. In Chapter 8 we went to the other extreme: the industry consists of just one firm. The market demand curve is also the firms demand curve. The monopolist sets the price, and usually (unless it is a contestable market) need not worry about other firms at all. Time to do something different. And guess what? Well consider markets in which there are more than one, and less than many, producers. How many? Lets say two, or three. Maybe four. Five? Possibly. Markets in which there are just a few competing firms are called oligopolies. The key feature of oligopolies is that the decisions of any one firm have a significant impact on the profits and the optimal choices of the other firms. In competitive markets each firm is individually too small for its decisions to matter for the other firms in the industry. In monopoly there are no other firms to affect. If what I do may induce you to do something different, then I must take your potential reactions to my decisions into account (as long as it is also true that what you choose to do affects me). This is a much harder problem than what a firm in a competitive market or a monopolist must deal with. My decision may induce you to change what you do, and when you change what you do it alters


what my decision should be. This in turn alters what you do, and this alters what I do, and this . . . . You get the idea. Solving the oligopolists problem involves an infinite chain of hypothetical reasoning. Fortunately, we have a short cut that might enable us to think about how oligopolists will behave. This short cut is the concept of the Nash equilibrium, which was introduced briefly in the problem set to Chapter 7. We revisit the concept and apply it to problems of oligopoly. Unfortunately, the concept only gets us so far. It tells us what is not going to be a market equilibrium, but it leaves open too many possibilities about what could be an equilibrium. The state of affairs in economic modeling of oligopoly is consequently somewhat unsatisfactory. Unlike our models of monopoly and competition, oligopoly models do not make unique predictions.

1. Games, Payoff Matrices and Nash Equilibrium

An oligopoly consists of a small number of firms, but we are going to limit our attention to markets with exactly two firms, which are known as duopolies. Doing so allows us to keep the math as simple as possible. The issues that arise with duopolies are representative of the issues that arise when there are more than two firms. Therefore, imagine there are two companies in an industry, and let us call them Airbus and Boeing. Imagine they produce airplanes. As we are thinking about an entirely fictitious industry, let us further imagine they can choose to produce either 50 or 75 airplanes per year and, if they do, their profits are as given in the payoff matrix in Figure 1. If both firms produce 50 planes each, they each make $100 million in profit. Total supply is restricted, and the price is consequently high. If each firm produces 75 planes, market price will be driven down drastically because of the high industry supply. Profit for each is just $75 million. The best that each firm can do is to produce 75 planes when the other firm produces only 50 planes. The larger producer will then make $125 million in profit, while the smaller producer makes only $50 million.


But it cannot be an equilibrium for one firm to produce 75 and the other to produce 50. To see this, recall from Chapter 7 the concept of Nash equilibrium, which is defined as a pair of choices such that no firm has an incentive to deviate unilaterally from its choice. Consider first the lower left cell, where Boeing is producing 75 planes. Boeing calculates that if it changes its strategy and produces 50 planes instead, it reduces its profit from $125 million to $100 million. However, Airbus calculates that if it raises its production from 50 planes to 75 (given that Boeing is producing 75 planes already), it increases its profit to $75 million. Thus, Airbus changes strategy and the lower left cell cannot be a Nash equilibrium. Similarly, if we begin in the upper right cell, with Airbus producing 75 planes, Boeing will want to increase production. Only when both are producing 75 planes does neither firm want to change strategy. Hence, the Nash equilibrium in this model has both firms producing 75 planes.36 If both firms could have cooperated, their profits would have been much higher. The cooperative outcome involves each firm agreeing to restrict output to 50 planes, and thereby earning profit of $100 million. A game with payoffs such that the Nash equilibrium leads to lower payoffs for both firms than the cooperative outcome is known as a prisoners dilemma. The name comes from a story, given in the inset box on the next page, that economists like to tell to motivate these sorts of games. In a prisoners dilemma, there are a number of possible outcomes. First the firms could each choose to produce 75 planes, which is a Nash equilibrium. Second, they could collude and produce 50 planes each. But there are two main problems with collusion. One is that collusion among oligopolists in the United States is illegal. The other is that even if they could manage to secretly collude, both firms have an incentive to cheat on the agreement. Collusion is


You should additionally verify that both firms producing 50 planes is not a Nash equilibrium.


then risky and difficult to maintain. Hence a third possible outcome is that the firms attempt to collude, but each ends up cheating or in jail.

AIRBUS 50 PLANES BOEING 50 PLANES 75 PLANES $100m, $100m $125m, $50m 75 PLANES $50m, $125m $75m, $75m

FIGURE 1. The payoff matrix for Boeing and Airbus. The left number in each cell is profit to Boeing, the right number is profit to Airbus.

The Prisoners Dilemma

The district attorney has arrested Ana and Maria on suspicion of stealing a television set. The DA puts them in two different cells and does not allow them to talk to each other. He makes Ana the following offer. If Ana confesses and Maria does not, Ana gets to go free. If Ana does not confess but Maria does, Ana gets 10 years in jail. If both confess, they get 5 years each, and if neither confess, the DA has enough to put them away for one year. The payoff matrix looks like this: ANAS CHOICE CONFESS DONT CONFESS MARIAS CHOICE CONFESS DONT CONFESS 5,5 0,10 10,0 1,1

You can see that the DA has carefully constructed a payoff matrix to induce both of them to confess. Both confessing is the only Nash equilibrium, although both would have been better off if they had been able to coordinate with each other and keep quiet. But by keeping them in separate rooms, the DA prevented this coordination.

But oligopoly can lead to yet more outcomes. One of the main difficulties with analyzing oligopoly is that if the technology and market demand were


different (so that the entries in the payoff matrix are different), there could be yet more substantively different outcomes. In Figure 2, we have changed the profits that are earned when both firms produce 75 planes, so that now each firm makes only $10 million in profit. What is the Nash equilibrium in this case?

AIRBUS 50 PLANES BOEING 50 PLANES 75 PLANES $100m, $100m $125m, $50m 75 PLANES $50m, $125m $10m, $10m

FIGURE 2. An alternative payoff matrix.

Consider first the lower left cell, where Boeing is producing 75 planes. Boeing calculates that if it changes its strategy and produces 50 planes instead, it reduces its profits from $125 million to $100 million. Airbus calculates that if it raises its production from 50 planes to 75 (given that Boeing is producing 75 planes already), it also loses profits. Hence the lower left cell is a Nash equilibrium. But so is the upper right cell in which Airbus produces 75 planes while Boeing produces only 50. Thus, the upper right cell is also a Nash equilibrium. The lower right cell and the upper left cell are not Nash equilibria. If both firms produce 50 planes, each one has an incentive to raise its output. If both firms are producing 75 planes, each firm has an incentive to reduce its output. Thus, in this game there are two equilibria, in both of which one firm produces 50 planes and the other produces 75 planes. Can we deduce which firm will produce the greater number of planes? Not in this game. Both


outcomes are possible, and the concept of Nash equilibrium does not help us decide between them. But let us again change the game slightly. Imagine that Boeing entered the industry first. It could then build a factory that is capable of producing 75 planes a year. One can think of this factory as a means of committing to produce 75 planes. Then, when Airbus enters it asks how many it should build. It will choose to build 50 planes. In this game, Boeing has a first-mover advantage. By choosing its strategy first, it gets to choose which Nash equilibrium the industry selects, and it gets to keep the lions share of the profits for itself.

2. Reaction Functions
A major limitation of using payoff matrices is that they restrict us to a small number of choices. But in reality, Airbus and Boeing could not only choose to produce 50 or 75 airplanes. They could also choose 55, 60, 72, 99, and so on. In principle we could just a create a payoff matrix that incorporates these extra choices. For example, if there were three choices, the payoff matrix would have three rows and columns. But, as you can imagine, the matrix would rapidly become unwieldy as the number of choices got large. An alternative way of thinking about the equilibrium is to develop the idea of reaction functions. A reaction function is nothing more than an equation (or, equivalently, a line on a graph) that tells us, for each company, what the optimal choice is given each possible choice the other firm has made. Let us illustrate the idea graphically first so that we can see how it related to the concept of Nash equilibrium, and then we can think about how we might construct the equations for a Nash equilibrium. Figure 3 illustrates a pair of reaction functions for the Boeing-Airbus duopoly. Line A plots the optimal output choice for Airbus for each possible output choice of Boeing. Given an output level for Boeing, Airbus must choose the corresponding output level plotted out by line A in order to maximize its


Boeing Output A

Nash equilibrium



Airbus Output

FIGURE 3. Reaction functions for Airbus (A) and Boeing (B). The Nash equilibrium is at the point of intersection of the two lines.

profit. Line B plots the optimal output choice for Boeing for each possible output choice of Airbus. Given an output level for Airbus, Boeing must choose the corresponding output level plotted out by line B in order to maximize its profit. The reaction functions are downward-sloping for an intuitive reason: if Airbus produces more, the industry price will go down, so the optimal quantity of output for Boeing must decline. If Boeing produces more, it is similarly optimal for Airbus to cut back. The Nash equilibrium must be where the two lines intersect. Only at this point is it true that (i) the output level for Airbus is optimal given Boeings output level, so Airbus has no incentive to change strategy, and (ii) the output level for Boeing is optimal given Airbuss output level, so Boeing has no incentive to change strategy.


Let us now think about the equations that generate the reactions functions and the Nash equilibrium. Industry demand is given by

q = a bp ,
where, as usual, a and b are some numbers. It is useful first to remind ourselves of the monopoly problem from Chapter 8. We discovered there that with this demand the marginal revenue curve is given by

MR =

a 2 q, b b

and the optimal output level is found where marginal revenue equals marginal cost. In our duopoly, industry output is given by the sum of the outputs of Boeing and Airbus:

q = qA + qB ,
and we can write industry demand as

q A + q B = a bp .
Let us now think of the profit maximization problem for Airbus. The demand for Airbus, given that Boeing produces qB output is given by

q A = a q B bp ,
which looks just like the monopoly problem except that the coefficient a from the demand curve is replaced by aqB. Thus, if we replace a with aqB. in the marginal revenue equation, we get Airbuss marginal revenue:

MR =

a qB 2 qA . b b a qB 2 qA , b b

Airbus then sets this equal to its marginal cost, MCA, to maximize profits:

MC A =

which we can rearrange to solve for qA:


qA =

1 b MC A q B . 2 2 2


Boeing faces the same problem, and its optimal output is given by

qB =

1 b MC B q A . 2 2 2


So, imagine that the industry demand curve is given by

q = 80 2 p ,


and both firms marginal cost is $10 (in millions, if you like, but I am trying to keep the math simple!). In the demand curve, a is 80 while b is 2, so Airbuss reaction function is

q A = 30 q B
and Boeings reaction function is

1 2


q B = 30 q A .

1 2


Figure 4 plots equations (4) and (5). We see that they intersect at qA=qB=20. These are the Nash equilibrium output levels. You should also verify this by solving (4) and (5) for qA and qB. Total industry output in the duopoly is 40 airplanes per year. Compare this with the monopoly output level and the competitive output level. By now, you should have no problem calculating that a monopolist facing the demand curve given by equation (3) and a marginal cost of $10 would choose an output level of 30 planes, while the competitive market would arrive at an output level of 60. This should not be surprising. Oligopoly, being somewhere between competition and monopoly, selects an output level between these extremes. If one firm chooses for some reason to produce nothing, then the other firm has the entire market to itself. Not surprisingly, it then chooses to produce the monopoly level of output.


60 50 Boeing Output 40 30 20 10 0 0 10 20 30 Airbus Output 40 50 60 Boeing Airbus

FIGURE 4. Reaction functions for Airbus and Boeing given marginal costs of $10 each and industry demand q=802p.

Changing costs of production. Imagine that Boeings costs were to increase,

from $10 per plane to $20. What would happen to output? The first thing to notice is that nothing happens to Airbuss reaction curve, so we need only worry about what happens to Boeings reaction curve. But this is easy to do. All we need to do is put the new marginal cost into equation (2):

qB =

80 1 1 20 q A = 20 q A . 2 2 2

Plotting this new reaction curve (Figure 5), we see that an increase in Boeings costs shifts its reaction curve inwards. It produces less, while Airbus produces more. The new Nash equilibrium has Boeing producing 6.67 planes (down from 20) and Airbus producing 26.67 planes (up from 20). Total industry output has declined by 6.67. Cournot versus Bertrand competition. In the previous example, we assumed that each firm chooses how much quantity to produce as a function of its competitors quantity. We then found the Nash equilibrium, where Boeing


60 50 Boeing Output 40 30 20 10 Boeing 0 0 10 20 30 Airbus Output 40 50 60 Airbus

FIGURE 5. An increase in Boeings costs shifts its reaction function in and alters the Nash equilibrium.

was choosing its optimal quantity given the quantity produced by Airbus, and Airbuss quantity was in turn optimal given Boeings quantity. But it is equally reasonable to assume that Boeing chooses its optimal price given the price charged by Airbus, and that Airbus does the same. So what? you may ask. After all, each firm faces a demand curve, so that when it chooses how much to produce the price it must charge to be able to sell this quantity is immediately defined. How could it possibly matter whether a firm (a) chooses quantity, and then the price it must charge to sell this quantity is defined, or (b) the firm chooses price, and then the quantity it must sell to attain this price is defined? Unfortunately, it matters dramatically. When oligopolistic firms compete in quantities, as we have already modeled, we call the market a Cournot


oligopoly. When firms compete in prices, we call it a Bertrand oligopoly.37 And, surprising as it may seem, the predictions of the two types of oligopoly model could not be more different. When both firms have the same unit cost of production, in Bertrand oligopoly the equilibrium price is immediately pushed down to the same price as would be charged in a perfectly competitive market. The reasoning is as follows. Imagine, as before that, both firms face a marginal cost of $10 and one of the firms, say Airbus, decides to charge $20. Boeing will immediately calculate that it can capture the entire market if it charged $19.99. But, then Airbus can capture the entire market back by charging $19.98. None of the these prices can be a Nash equilibrium. As long as the price exceeds the unit production cost, one firm can capture the entire market by charging $0.01 less than its competitor. Thus at any price above $10, at least one firm has an incentive to lower its price slightly. The only possible Nash equilibrium is when price has been driven down to the unit production cost. Then no firm can make profits by lowering the price yes, they would capture the entire market but they would incur a loss on every sale. The Nash equilibrium in a Bertrand oligopoly can be illustrated by plotting two reaction functions where prices in instead of quantities are on the axes (Figure 6). The 45o line indicates pairs of equal prices. So consider first Boeings reaction function, indicated by the line abc. For any price that Airbus charges above the unit cost of $10, Boeing wants to charge a price that is just a little lower. This is shown by having Boeings reaction function lie just a little below the 45o line. However, Boeing will never charge a price below $10, so once Boeings price falls to $10, its reaction function becomes horizontal.


Named for the first economists to study these models: Augustin Cournot (1801-1877) and Joseph Bertrand (1822-1900). Cournot had described the quantity game in 1838. In 1883, Bertrand reviewed Cournots work, and developed the price game.


Boeings Price

Airbuss unit cost Airbuss reaction function f 45o c

Boeings reaction function


Boeings unit cost

d 10

Airbuss Price

FIGURE 6. Reaction functions for Bertrand oligopoly.

Airbus faces an identical problem. As long as price exceeds $10, Airbus would like to charge a price just below Boeings. This is indicated by having Airbuss reaction function, def, lie just to the left of the 45o line. And again, Airbus will never charge less than $10, so its reaction function becomes vertical at the unit cost. The Nash equilibrium is as usual where the two reaction functions intersect, and as we see this is where both firms charge $10. How much does each firm produce in the Bertrand oligopoly Nash equilibrium? The sad answer is that we do not know. Both firms charge the same price, and buyers are indifferent between them. We can only presume that somehow buyers turn up at one or the other firms door according to some process that we do not understand. In the Cournot oligopoly, prices are much greater than the unit cost. Recall that when units costs are $10 and firms choose quantities, we found that each firm produces 20 units. We can easily verify what the resulting market price would be. The industry demand curve is q = 80 2 p . Plug an industry quantity of 40 into the demand curve, and we can solve for p=$20.


The Bertrand and Cournot oligopoly models have different predictions about what the equilibrium looks like when both firms have the same unit cost of production. It is therefore no surprise that they will also have very different predictions about what the equilibrium looks like when the two firms have different costs of production. We have already seen that in the Cournot equilibrium, when one firms unit cost increases both firms continue to produce. The firm whose costs increase reduces its market share, while the other firm increases its market share. In the Bertrand equilibrium, the effects of a change in production costs for one firm are much more extreme. If, for example, Boeings production costs rise from $10 per unit to $15 per unit, Airbus will choose simply to sell at $14.99. At this price, Boeing drops out of the market, and Airbus becomes the sole producer.38

3. Collusion, the Prisoners Dilemma, and Repeated Games

When oligopolies compete in prices, competition will drive profit all the way down to zero. Even when firms compete in quantities, competition reduces profits. In the Cournot model from the previous section, we found that each firms charged $20 and sold 20 units. Given a unit cost of $10, this means that each firm earns $200 profit. A monopolist would have produced 30 units of output, at a monopoly price of $25, thereby earning profit of $450. Boeing and Airbus could both do better than engage in competition. They could agree to fix the price at $25, sell 30 planes in total, and split the profit. They would accomplish this by each producing 15 airplanes. Doing so would earn them $225 each. This is a little better than the $200 each earned in Cournot competition, and a lot better than the zero profit earned in Bertrand competition. Agreeing in this way to raise prices and split the profits is


In the language of Chapter 8, Airbus will be a monopolist in a contestable market: it will not be able to charge the full monopoly price unless its costs are much lower than Boeings.


referred to as collusion. We say that a group of firms that gets together to restrict output and raise prices in this way has formed a cartel. Oligopolists like the idea of collusion, especially when profits have all but vanished. In the early 1990s, American Airlines and Braniff were engaged in a price war as they competed for passengers on flights in and out of Dallas. Robert Crandell, the CEO of AA, called Braniffs chair, Howard Putnam, in 1992. Part of their conversation went like this:
Crandall: I think its dumb as hell, for Christs sake, all right, to sit here and pound the shit out of each other and neither one of us making a fing dime. Putnam: Well Crandall: I mean, you know, goddam, what the fis the point of it? . . . Crandall: I have a suggestion for you. Raise your goddam fares twenty percent. Ill raise mine the next morning. Putnam: Robert, we Crandall: Youll make more money, and I will too. Putnam: We cant talk about pricing. Crandall: Oh, bullshit, Howard. We can talk about any goddam thing we want to talk about. Transcript of conversation, as received by the Department of Justice.

The problem was that it was in fact illegal for Crandell and Putnam to talk about prices. In fact it is viewed as serious crime. When Sothebys and Christies conspired to fix the commissions they charged at their auctions, some of them went to jail (see inset box on the next page). The reason is straightforward. When firms collude and set the monopoly price and output level, and then split the profits, they raise their profit by hurting consumers and by reducing overall economic welfare. In the United States, it has long been decided that considerations of welfare justify making collusion illegal. Of course, making something illegal doesnt mean people dont do it. Especially in corporate America.


Collusion and cheating. Figure 7 illustrates both the benefits and challenges of
colluding to fix prices when firms compete in quantities. As we have seen, if there is no collusion, both firms charge a price of $20 and produce 20 planes each, making a payoff of $200 each. This is indicated in the lower right box of Figure 7. We have also seen that if they collude to split the monopoly profits, they each produce 15 planes and make $225, as indicated in the upper left box. What happens if they agree to collude and produce 15 planes, but then one of them decides to cheat? Imagine that Airbus keeps to the agreement and produces 15 planes, but then Boeing cheats. Boeing calculates what its best option is when Airbus produces 15 planes. We can find out what this is by looking at Boeings reaction function, which we found was

q B = 30 q A .
Thus, when Airbus produces 15 planes, Boeing can maximize its profits by producing 22.5 planes. Total supply is then q=15+22.5=37.5 planes, and from the demand curve, q = 80 2 p , we see that price will be $21.25. Now Airbus is producing 15 planes at this price, and makes $168.75 profit. Boeing is producing 22.5 planes at this price and makes $253.13 profit. These payoffs are indicated in the lower left box of Figure 7. A similar outcome arises when Boeing agrees to collude and Airbus cheats.
AIRBUS COLLUDE BOEING COLLUDE DO NOT COLLUDE $225, $225 $253.13, $168.75 DO NOT COLLUDE $168.75, $253.13 $200, $200

1 2

FIGURE 7. Collusion versus Cournot Competition.

The Nash equilibrium in Figure 7 is easy to spot: it is where both cheat and earn profit of $200. Although both firms would be better off by colluding, they


Taubman Sentenced To One Year--Plus A Day

Dan Ackman, April 22, 2002.

NEW YORK - A. Alfred Taubman, one of the highest-ranking American businessmen ever to be convicted of a serious crime, was sentenced to a year and a day in federal prison today for his role in a price-fixing scheme between his Sotheby's auction house and archrival Christie's Taubman, the former chairman of both Sotheby's and Taubman Centers, a real estate trust, was convicted of the charges Dec. 5, 2001. His lawyer, Robert Fiske, argued today that he should receive no jail time in light of his age--Taubman is listed as 78, though there is some dispute--and his failing health and history of charitable good works, which the lawyer called "legendary." A U.S. Probation Department report recommended probation only. Federal antitrust prosecutor John Greene argued that Taubman should be sentenced closer to the three-year maximum. U.S. District Judge George Daniels, citing Taubman's "lack of contrition" and his "arrogance and greed" in heading up the secretive scheme, said Taubman must go to jail for his crime to show that "no one is above the law." But Daniels did ratchet down the sentence because of the factors Fiske cited. The judge tacked on a $7.5 million fine, the amount calculated as 5% of the volume of commerce affected by the price-fixing scheme. Taubman will also have to pay for the cost of his incarceration. Fiske argued that a three-year sentence would in fact be a life sentence because Taubman has just a 3.8-year life expectancy-assuming he gets careful medical care, the kind not available in prison. He called Taubman "a Horatio Alger story," noting that his client started work at age 9 and made his fortune by building shopping malls and in related ventures, Fiske argued that Taubman's philanthropy was unique even for a man with a net worth of more than $700 million, which ranks him at No. 340 on Forbes' 400 Richest Americans list. Speaking for 50 minutes, Fiske recited a long history of Taubman's philanthropy, including funding the revitalization of downtown Detroit, the organization of a program to improve public school teaching in Michigan, scholarships for Jordanian students studying in the U.S. and the creation of a center at Harvard University for the study of state and local government. Taubman gave of his time as well as his money, Fiske said. The Justice Department's Greene argued that deterrence demanded a jail term. Sotheby's has already paid a $45 million fine and has been found liable for hundreds of millions of dollars in civil damages to clients and shareholders. Taubman agreed to pay $156 million of the settlements personally. Daniels acknowledged Taubman's philanthropy, but added, "The law does not countenance a Robin Hood." The victims of Taubman's crimes may have been rich, the judge said, but they were victims nonetheless, and they had lost $43.8 million over six years.


FIGURE 8. Price of crude oil, 1930-2003, adjusted for inflation.

both have an incentive to cheat on any agreement to fix prices. That is, collusion to fix prices is a prisoners dilemma. Figure 7 illustrates why collusion so often appears to be unstable. If two or more firms ever manage to restrict output and split the higher profits, each member of the cartel has an incentive to cheat. At some point, one of them is going to cheat. When they do so, prices will collapse, and there will be a period of intense competition. Perhaps at a later stage, seeing how little profit they are making, the firms get together again and raise prices once more. OPEC is a cartel that makes money for its member countries by restricted oil supply and raising prices. Not all oil-producing countries are members of OPEC. Some of them ignore OPECs collusion, while others go along with it and join the collusive agreement. However, OPEC controls enough of the worlds oil supply to be able to collude. In 1973 it exercised this power, cutting 254

back supply enough to cause oil prices to rise more than 200 percent (Figure 8). Flushed with this success, OPEC struck again in 1980-81. However, as Figure 8 shows, oil prices declined steadily throughout the 1980s and remained quite low during most of the 1990s. Why? There are two reasons. First, oil producing countries both in and not in OPEC succumbed to the incentives they had to expand production, thereby undermining the OPEC strategy. Second (and this came as quite a surprise to OPEC), consuming countries turned out to be surprisingly successful in developing their ability to substitute alternative energy sources for oil and in becoming more energy efficient. In so doing, consuming countries made the demand for oil more price elastic, which as we know from Chapter 8 reduces any producers ability to raise price and reap extra profit.

Repeated games as a way to enforce collusion. If every member of a cartel has

an incentive to cheat, how can a price-fixing agreement ever be sustained? The answer is that members must have some way to punish each other for cheating. In some cases, one can do this with a large supply of missiles. In other cases, a less violent method is needed. When a strategic game is repeated many times, a mechanism for enforcing collusion often emerges naturally. For example, Boeing and Airbus play their oligopoly game with each other year after year. If neither cheats on a collusive agreement one year, they always have an opportunity to cheat next year. Similarly, if one of them cheats this year, the other player in the game has an opportunity to punish next year, and the year after if need be. Imagine, therefore, that Boeing and Airbus play the Cournot oligopoly game for ever. Each year that they collude, they earn $225. If one of them cheats, it earns $253.13, while if both play the Nash equilibrium strategy, they earn $200. Imagine that Boeing makes the following threat to Airbus: If you cheat today, then next year I will choose the Nash equilibrium. Airbus then calculates that if it cheats today, it will gain $253.13 - $225 = $28.13 this year. The penalty it pays is that next year it will earn only $200, a loss of $25 from


the collusive agreement. That is, Airbus calculates that it gains $28.13 this year and loses $25 next year, so it goes ahead and cheats. Boeings threat was not enough to stop Airbus from cheating. Knowing this, Boeing makes a tougher threat: Cheat today and I will choose the Nash equilibrium for the next two years. Airbus can now calculate that it gains $28.13 this year, and then loses $50 over the next two. Seeing this, it does not cheat. Airbus makes exactly the same threat to Boeing, and Boeing does not cheat. Whether such threats work depends on a number of factors. First, the threats have to be credible. Neither Boeing nor Airbus can threaten to do something that, when the time comes, it would not be willing to do. One useful way to ensure a threat is credible is to restrict actions that are threatened to be Nash equilibria, because a firm will always be willing to do something that is a Nash equilibrium (see inset box below). Second, the firm being threatened must care about the future enough. If firms really dont care too much about what happens next year, then there is not much one can do as a threat. Fortunately,

Time Inconsistent Threats

Threats are not credible if the person making the threat would not want to follow through on them when the time comes to act. Such threats are time inconsistent. Heres an example: Charles has just graduated high school and has the summer to enjoy before going to college. He plans to spend the summer on vacation with his friends. Charles parents have a different idea. They want Charles to learn some responsibility before he goes to college, and they want Charles to get a summer job. Charles parents make the threat that if he doesnt get a job they wont pay for college. Unfortunately for his parents, Charles is not dumb. He understands that his parents threat is time inconsistent. Even if he is still irresponsible come August Charles knows that his parents would still prefer that he goes to college than not, and so they wont want to carry out their threat. Consequently, Charles takes his vacation.


most firms do care about their future for the simple reason that a firms stock market value depends not only on this years profits, but all expected future profits.

4. Concluding Comments
In this chapter we studied issues that arise when markets consist of a small number of firms. These markets are characterized by strategic interactions. When the actions of one firm has a significant impact on the profits and the optimal actions of other firms, working out what firms should do involves an infinite regress: my action depends on your which depends on mine which. . . . To get around this problem, we exploited the concept of the Nash equilibrium. We first analyzed this for a duopoly that competed by choosing production quantities (Cournot). We then analyzed the Nash equilibrium for a duopoly in which firms competed by choosing prices (Bertrand). We found that the two models of duopoly gave very different predictions. This is, unfortunately, an unavoidable problem in the analysis of oligopoly.

1. Consider a duopoly where total industry demand is given by q = 100 5 p . Firm A has unit production costs of $10, while firm B has unit production costs of $5. (a) Derive the Cournot (quantity) reaction functions for both firms. Plot them. What is the Nash equilibrium? (b) Plot the Bertrand (price) reaction functions for both firms. What is the Nash equilibrium? 2. Consider a duopoly with the same demand curve as in question 1, but now assume that both firms have unit production costs of $10. (a) Derive the Cournot (quantity) reaction functions for both firms. Plot them. What is the Nash equilibrium?


(b) How much is produced when the two firms collude to maximize profits? Assume they split profits so derived evenly. (c) If this oligopoly game is repeated year after year, for how many years must each firm threaten to play the Nash equilibrium strategy in order to induce the other firm not to cheat on the collusion agreement?


Answers and Hints to Selected Problems

Chapter 1. Auctions
2. $1.5 Million or $1.6 million, depending on the order of bidding. Buyer C wins in both cases. 3. a) Who knows? Certainly less than $1.7 million. b) $1.5 million 4. a i) $190 or $200. The final bidder indifferent between bidding $200 and winning, or not bidding $200 and losing. ii) $200. b i) $100 or $110. (ii) $100 c i) $90 or $100. ii) $100 5. i) ii)
1 4

( 1 (190) + 1 (200)) + 1 ( 1 (100) + 1 (110)) + 1 ( 1 (90) + 1 (100)) . 2 2 2 2 2 4 2 2

1 4

(200) + 1 (100) + 1 (100) . 2 4

6. i) Add the ten highest valuations and subtract the total amount paid. The price is the eleventh highest-bid. ii) $500. iii) 8 are sold. The net welfare loss is $150. To see why, note that two sales are lost, to the two individuals with valuations of $130 and $120. Instead, the two reels they would have bought remain with Manual, who only values them at $50 each. iv) All 10 are sold. There is not net welfare loss. 7. i) Collins wins and pays either $80 or $90. ii) Not necessarily: A Vickrey auction would yield a price of $82. 8. Auction 1: A and B. Auction 2: No one. Auction 3: No one.

Chapter 2. Three Principles

1. These are straightforward but tedious calculations. 2. Draw two budget lines. Indicate one point on the old budget line and one point on the new budget line that satisfies the changes in quantities indicated in the question. Now draw indifference curves tangent to the budget lines at these points.


3. i) Indifference curves have a positive slope. ii) Indifference curves slope down, but moving into towards the origin gives higher utility. iii) Indifference curves slope down until 120 units of soda, after which point they become positively sloped. 4. The budget lines has a king at (x=10, y=10). If good y is on the vertical axis, the slope is 1 to the left of x=10, and 1/2 to the right. 5. The change in price causes the budget line to rotate through the point (10 bushels of corn, 10 pounds of beef) 6. The budget lines has a kink at (10 corn, 10 beef). 7. Note: Study time is a bad. If you put study time on the axis, the indifference curves will have a positive slope. If you put leisure (= 24 study time) on the axis, the indifference curves have a negative slope. i) If study time is on the horizontal axis, the budget line has a positive slope up to 8 hours of study, and is horizontal thereafter. If leisure time is on the horizontal axis, the budget line is horizontal up to 16 hours of leisure, and has a negative slope thereafter. has a positive slope up to 8 hours of study, and is horizontal thereafter. ii) The budget line has a maximum at 5 hours of study. Studying more than five hours lowers the grade.

3. Efficiency and Equity

1. Construct a utility possibility frontier by calculating the utility of each person at each allocation that uses all 8 units of the good. Plot the frontier. (a) Are the listed allocations inside, outside, or on the frontier? (b) Which allocation results in the highest utility for the person receiving lower utility? (c) Which allocation results in the highest total utility? 2. a) No. b) Yes. No matter which policy you begin with 2 groups out of three prefer another.

4. Supply and Demand

1. a) p=25, q=30, consumer surplus = 225, producer surplus = 225. b) p=8, q=16, consumer surplus = 16, producer surplus = 32


2. a) qd=50p, qs=10/3+1/3p. (Hint: For each write a general equation of the form q=a+bp, where a and b are unknown numbers. The graphs show two points through which each line passes. You can used this information to solve for a and b). 3. a) The demand curve shifts up by $5 but maintains the same slope. Find the new equation of the demand curve and solve for the new equilibrium. The new demand curve is qd=902p, The new price is 27.50 and the new quantity is 35. b) The supply curve shifts down by $1, while maintaining the same slope. 4. (a) As you have already done question 1, you know how to do this. (b) The new supply curve is given by qs=80+3p. Use this to find the new equilibrium. 5. If prices and quantities move in the same direction, the dominant shift must have been in the demand curve. If they move in the opposite direction, the dominant shift must have been in the supply curve. 6. Calculate the PS. Or simply draw the two supply curves on the same diagram and eyeball the difference. 7. (a) Price falls by $1 (b) CS rises by $10,000 ; PS does not change. 8. All the low cost producers are active. No high cost producers are active. Price is $150; Profit is $50,000. 9. You are given enough information to conclude that the supply curve is qs=10+p. At the world price, the domestic producers sell 2 units; domestic consumers consume 40 units, so 38 units must be imported.

5. Messing with Markets

1. The free-market quantities are p=$25, q=30, CS=$225, PS=$225, GR=$0, EW=$450. a) p=$30, q=20, CS=$100, PS=$300, GR=$0, EW=$400. b) p=$20, q=20, CS=$300, PS=$100, GR=$0, EW=$400. c) Price cap is not binding. Same as free market equilibrium. d) Same outcome as in (a). e) p=$22.50 (producers), $27.50 (consumers), q=25, CS=$156.25, PS=$156.25, GR=$125, EW=$437.50.


f) p=$27.50 (producers), $22.50 (consumers), q=35, CS=$306.25, PS=$306.25, GR=$175, EW=$437.50.

3. Calculate PS and CS with and without the policy. The cost of the tax borne by producers is the decline in PS as a percentage of the decline in totals economic welfare. 4. (a) Price cap. With policy: p=10, q=20, CS=$700, PS=$50, GR=$0, TEW=$750. (b) A subsidy of $5. With policy: Pp=21.67, Pc=16.67, q=66.67, CS=$1,111.06, PS=$555.69, GR=$333.33, TEW=$1,333.45. (c) Quota of 50. With policy: p=25, q=50, CS=$625, PS=$687.50, GR=$0, TEW=$1,312.50. (d) A tax of $10. With policy: Pp=16.67, Pc=26.67, q=46.67, CS=$544.29, PS=$272.26, GR=$466.60, TEW=$1,283.15. 5. The supply curve has a steeper slope, so they suffer the greater burden. 6. (a) $3.67. (b) $4.40. 7. Price must rise by at least $22.50. 8. No effect on welfare (quantity does not change). 9. The supply curve is horizontal. (a) $0. (b) Only the quota; the other two have no effect on profits. 6. Externalities
1. The free-market quantities are p=$25, q=30, CS=$225, PS=$225. a) q=26. (Note: the socially optimal supply curve is qs=28+2p). b) In the free market, welfare is CS+PSPollution cost = 450 (4 x 30) = 330. In the social optimal, CS+PS Pollution = 442 (26 x 4) = 338. Hence free-market welfare is $8 less than the social optimum. c) (i) $23. (ii) $27. (iii) $4.


d) (i) The maximum amount is the pollution cost avoided by going from the free market to the social optimum. This is $16. The minimum amount is the lost welfare to producers, which is $4. If consumers could also demand compensation, the minimum would be $8. (ii) The minimum amount is the pollution cost from 26 units of output. This is $104. 2. You need to draw each of these cases. a) Hint: The social benefit curve lies $2 above the demand curve up to output equal to 20, and then jumps up to be $4 above. At the intersection of supply with private demand, only the $4 difference is relevant. The optimal output level is 34, requiring a subsidy of $4. b) The free market output level is optimal. c) Hint: The optimal tax is a 10 percent tax. d) Hint: The optimal tax is a 10 percent tax.

7. Public Goods
1. First matrix: {No, No}. Second matrix: no Nash equilibrium. 2. q=20. As wtp is $0, Bs is $20. 3. With all four included, total CS is +5, so the project goes ahead. Hence, only those with positive CS could be pivotal. Both B and D are pivotal. For B the tax is $15, for D the tax is $5.

8. Monopoly
1. (i) Negative. (ii) Subsidy; $3. (iii) Tax; $5. 2. p = 100/b+10. 5. From the MR equation, a/b=50 and 2/b=4. This allows you to solve for a and b, and therefore to obtain the equation of the demand curve. 6. $11.77. Find the equation of average cost: AC=100/q+10. Set p in the demand curve equal to this AC and solve for q. You will get a quadratic equation with two solutions. From the diagram, which of these two mathematical solutions is the economically relevant one? Once you have q, you can get p easily.


7. For part (a) simply solve the monopoly problem twice. For part (b) solve the monopoly problem for the aggregate demand, q=1607p.

9. Oligopoly
1. (a) qA=8.33; qB=33.33; p=11.67. (b) Firm B captures the entire market at a price of $9.99. 2. (a) qA=25qB/2; qB=25qA/2; Nash equilibrium is qA=16.67, qB=16.67. (b) Each produces 12.5. (c) You need to calculate the gain from cheating. If one firm cheats, it gains $7.81 in profits relative to sticking to the collusive agreement. The following year, the firm would lose $6.93 by having the Nash equilibrium forced on it, instead of the collusive agreement. This is not enough to stop the cheating, so the threat must be made to punish cheating for two years.