Back in chapter 2 I offered a simple argument for laissez-faire. If you leave people free to exchange goods on mutually acceptable terms, the result is to move all goods to their highest valued uses, producing the efficient allocation of existing goods. If a good is worth more to a potential consumer than it costs a potential producer to produce it, the latter will find it in his interest to produce the good and sell it to the former, with the result that goods get produced if and only if they are worth producing.
This argument assumes that the only cost to me of selling one more unit of a good is the cost of producing it. But what if one effect of trying to sell more units is to drive down the price? Suppose, for example, that I own the only grocery store in a small town. The cost to me of selling an extra ten gallons a week of milk is not only what it costs me to get and sell the additional milk but also the lost revenue on milk I could have sold at a higher price if I had been content with selling less.
A numerical example may make the argument clearer. I am choosing whether to sell a hundred gallons a week at $2 a gallon or a hundred and ten at $1.90. The cost to me of the extra ten gallons is $1.50/gallon. Expanding output gives me $19 of revenue, ten gallons of milk at $1.90/gallon, at a cost of only $15. That sounds like a good deal until it occurs to me that it also costs me $10 of lost revenue on the hundred gallons that I could have sold at the higher price and must now sell at the lower.
If I considered only the cost and revenue directly associated with the additional gallons, I would keep selling more milk as long as the price was above my cost, since as long as price is above cost I am making money on each additional gallon. That describes the behavior of a seller in a perfectly competitive market—one with so many firms that his sales have no significant effect on price. But a monopoly, more generally a firm with some market power, knows that it can only sell more by selling at a lower price—which costs it revenue on the units that it could have sold at the higher price if it had been content with a lower volume. For me, that lost revenue is a cost. For everyone put together, it is only a transfer; what I lose on milk that my customers were willing to buy at the higher price, they gain.
To see why selling at a price above cost is inefficient, consider a customer to whom an additional gallon of milk is worth $1.80—less than the price I am selling milk for but more than the cost to me of providing it. If I produced the extra gallon and gave it to him, I would lose $1.50, he would gain $1.80, for a net gain of $.30. The net gain would be the same if I produced the milk and sold it to him for $1.50, although the division of the gain between us would be different. As long as I am selling milk for more than its cost to me, there will be customers who value the milk at more than its cost but are not getting it, which is inefficient. This simple point, usually illustrated with a diagram showing the monopoly's cost curves, demand curve, and profit maximizing price, is the standard economic argument for the inefficiency of a monopoly seller.
[math link]
The failure to sell to everyone who values milk at more than it costs me is a problem for me as well as for economic efficiency; I am missing out on potential profits. One solution is to find some way of selling milk at different prices to different customers, charging a high price to those willing to pay it and a lower price to those who will only buy at a low price. In some cases such price discrimination is a practical option, but in many others it is not, either because I cannot tell who will or will not pay a high price or because I cannot prevent people to whom I charge a low price from reselling to people to whom I am trying to charge a high price.
If I somehow solve these problems, and do it well enough so that I can sell to every customer at the highest price he is willing to pay, the standard argument against monopoly vanishes. As long as there is a customer who values milk at more than it costs me, it is in my interest to sell it to him. The result is that with perfect price discrimination it is in my private interest to sell the efficient quantity of milk, the quantity such that everyone who values it at more than its cost of production gets it.
Perfect price discrimination is perfectly efficient, at least so far as quantity is concerned, although that is not true for the imperfect price discrimination that is usually the best a seller can do. Perfect price discrimination also results in all of the gain from the transaction going to the seller and none to the buyer, like a bilateral monopoly bargain where one side is a much better bargainer than the other. But since economic efficiency is concerned with the size of the net gain, not who gets it, that should be irrelevant to efficiency. Or perhaps not. Stay tuned for late breaking updates.
In a competitive industry, above normal profits, profits that more than pay the normal rate of return on capital, make it in the interest of someone to start a new firm, driving output up and prices and profits down. So in a competitive industry in long-run equilibrium, firms sell at a price that just covers all of their costs, including a market return on the stockholders' capital. Economic profit, defined net of the normal cost of capital, including the capital of the stockholders, is zero. In a monopoly industry, on the other hand, there is only room for one firm. The result is at least the possibility of monopoly profit.
The year is 1870. Somewhere west of civilization is a valley of fertile farmland, into which it will some day be worth building a rail line. Whoever builds the first line will have a monopoly; it will never pay to build a second. If the line is built in 1900, the total profit it will eventually produce after paying all costs, including a normal market return on the capital used to build it, will be $20 million. If the railroad is built before 1900, it will lose a million dollars a year until 1900 because there will not be enough people in the valley to support the cost of maintaining the rails.
I, knowing these facts, propose to build the railroad in 1900. I am forestalled by someone who plans to build in 1899; $19 million is better than nothing, which is what he will get if he waits for me to build first. He is forestalled by someone willing to build still earlier. The railroad is built in 1880. The builder receives only the normal return on his capital for building it. The logic of the situation is identical to the logic of inefficiently early homesteading in chapter 10 and inefficient theft in chapter 3.
In such a situation, monopoly profit ends up not as a transfer to the firm from its customers but a net loss. The higher the monopoly profit, the more resources firms burn up competing to be the one that gets it. If so, perfect discriminatory pricing is not the best solution to the problem of monopoly but the worst. Since each customer is buying at the highest price he is willing to pay, all of the gain from the firm's production is transferred to the firm as monopoly profit—and all of the monopoly profit is burned up in the cost of acquiring the monopoly. So the distribution of the gains from trade does matter after all, not as part of the definition of efficiency but, in this situation at least, as an incentive to inefficient behavior.
We now have two arguments for the inefficiency of monopoly—each presented in a simplified scenario but with more general application. One is that a monopoly, in the process of maximizing its profit, sells too low a quantity at too high a price; if it expanded output, its customers would gain more than it would lose. The second is that the opportunity to acquire monopoly profits creates an incentive for inefficient rent seeking, for spending resources making sure that your firm, rather than someone else's, ends up with the monopoly.
If monopoly is inefficient, what can we do about it? Before attempting that question, there is another we must answer first:
The only grocery store in a small town is a natural monopoly. If someone tried to start a second store, one of the two would eventually go out of business because one store selling to everyone has lower average costs, and so can afford to sell at a lower price, than two stores each selling to only some of the customers. The same thing could happen on a larger scale in an industry where a large part of production cost was independent of how much was produced—the cost of designing a product, writing a computer program, tooling up a factory. The more units the fixed cost is divided among, the lower the average cost, so big firms can undersell small ones.
Almost all firms have some fixed costs, so why isn't every industry a natural monopoly? One reason is that economies of scale in production are balanced by administrative diseconomies of scale. The bigger a firm is, the more layers there are between the president and the factory floor and the harder it is for the former to control what is happening on the latter. Once a firm has gotten big enough to take advantage of most of the potential economies in production cost, further growth may cost it more in higher administrative costs and less carefully managed factories than it gains in lower production costs. That is one reason that most industries consist of many firms.
One reason there may be only one firm in an industry is natural monopoly. Another is that if someone tries to compete with him, the monopolist will call the police. The original meaning of "monopoly" was an exclusive right to sell something. Typically such monopolies were either sold by the government as a way of raising money or given to people the government liked, such as relatives of the king's mistresses. Monopolies of this sort are still common. An example is the Post Office: the Private Express Statutes make direct competition illegal.
For a third source of monopoly, and one that brings us closer to legal issues associated with antitrust, consider an industry made up of only five large firms. It occurs to the president of one of them that if they all reduce production, prices will rise; they will gain more from higher prices than they lose from lower sales. The result is a cartel—a group of firms coordinating their behavior to hold output down and prices up as if they were a single monopoly.
One problem for the cartel members is that while each of them is in favor of the others keeping output down and price up, each would like to expand its own output to take advantage of the high price. It can do so by chiseling on the cartel price, selling additional output for a little less to favored customers, defined as customers who can be lured away from a competitor and trusted not to tell anyone about the deal they are getting. Each cartel member gains by chiseling—at the expense of the others. If they do enough of it, they drive the price back down to what it was before the cartel was formed.
One solution is for the members of the cartel to sign a contract agreeing to keep their output down and prevent chiseling by doing all of their selling through a common agent. In much of the world, such contracts are both legal and enforceable. In the United States, they are neither. They have long been unenforceable as contracts in restraint of trade; under current antitrust law they are also illegal.
That is not true of a cartel agreement enforced by the government, such as the airline industry prior to deregulation. The enforcement agency enforcing the domestic airline cartel was the Civil Aeronautic Board, which had veto power over fare changes, including fare cuts. A similar function was served in the international market by IATA, most of whose members were government airlines whose owners could enforce the agreement by denying non-members landing rights in their countries.
Another solution is for the firms to merge into a monopoly. This may raise costs somewhat, since the reason there were originally five firms instead of one was that the firms were already up to the size at which average cost was minimized. But it may also raise profits if the merged firm has enough of the market to be able to restrict output and drive up price. In the United States, such mergers are subject to disapproval by the antitrust division of the justice department.
A friend of mine has suggested an ingenious way in which the antitrust division could distinguish "procompetitive" mergers, mergers designed to join firms that can produce more cheaply by combining their assets, from "anticompetitive" mergers designed to create a monopoly. A procompetitive merger makes things worse for other firms in the industry, since it produces a more efficient competitor. An anti-competitive merger makes things better for other firms in the industry, since they will benefit when it restricts output in order to drive up the price at which it—and they—can sell. All the antitrust division has to do, when a new merger is proposed, is see who objects. If the other firms in the industry object, it approves the merger; if they don't object, it rejects it. Unfortunately, this only works until the other firms catch on and revise their tactics accordingly.
Cartels and anti-competitive mergers both result in both of the sorts of inefficiency described earlier. In either case, one consequence of the monopoly is to push prices above marginal cost, reducing output below its efficient level. In the case of a cartel, rent seeking takes the form of expenditures by the cartel members to enforce, and evade, the cartel restrictions, as well as bargaining costs over creating and maintaining the cartel. In the case of merger, the inefficiency of having a firm too big to minimize average cost is a rent seeking expenditure, paid by the merging firms in the process of getting a monopoly in order to transfer money from their customers to themselves.
Suppose that in some industry economies and diseconomies of scale roughly balance; over a wide range of output, big firms and small firms can produce at about the same cost. It is widely believed that such a situation is likely to lead to an artificial monopoly; the usual example is the Standard Oil Trust under John D. Rockefeller.
I am Rockefeller and have somehow gotten control of 90 percent of the petroleum industry. My firm, Standard Oil, has immense revenues, from which it accumulates great wealth; its resources are far larger than the resources of any smaller oil company or even all of them put together. As long as other firms exist and compete with me, I can earn only the normal market return on my capital.
I decide to engage in predatory pricing, driving out my competitors by cutting my prices below my (and their) average cost. Both I and my competitors lose money; since I have more money to lose, they go under first. I now raise prices to a monopoly level. If any new firm considers entering the market to take advantage of the high prices, I point out what happened to my previous competitors and threaten to repeat the performance if necessary.
This argument is an example of the careless use of verbal analysis. "Both I and my competitors are losing money . . ." sounds as though we are losing the same amount of money. We are not. If I am selling 90 percent of all petroleum, a particular competitor is selling 1 percent, and we both sell at the same price and have the same average cost, I lose $90 for every $1 he loses.
My situation is worse than that. By cutting prices, I have caused the quantity demanded to increase; if I want to keep the price down, I must increase my production—and losses—accordingly. I lose (say) $95 for every $1 my competitor loses. My competitor, who is not trying to hold down the price, may be able to reduce his losses and increase mine by cutting his production, forcing me to sell still more oil at a loss. He can cut his losses by mothballing older refineries, running some plants half time, and failing to replace employees who move or retire. For every $95 I lose, he loses (say) $0.50.
But although I am bigger and richer than he is, I am not infinitely bigger and richer; I am 90 times as big and about 90 times as rich. I am losing money more than 90 times as fast as he is; if I keep trying to drive him out by selling below cost, it is I, not he, who will go bankrupt first. Despite the widespread belief that Rockefeller maintained his position by selling oil below cost in order to drive competitors out of business, a careful study of the record of the antitrust case that led to the breaking up of Standard Oil found no evidence that he had ever done so. The story appears to be the historian's equivalent of an urban myth.
[article link: McGee]
In one incident, a Standard Oil official threatened to cut prices if a smaller firm, Cornplanter Oil, did not stop expanding and cutting into Standard's business. Here is the reply Cornplanter's manager gave, according to his own testimony:
Well, I says, "Mr. Moffett, I am very glad you put it that way, because if it is up to you the only way you can get it is to cut the market, and if you cut the market I will cut you for 200 miles around, and I will make you sell the stuff," and I says, "I don't want a bigger picnic than that; sell it if you want to" and I bid him good day and left. That was the end of that.
—quoted in John S. McGee, "Predatory Price Cutting: The Standard Oil (NJ) Case," Journal of Law and Economics, Vol. 2 (October, 1958), p. 137.
Predatory pricing is not logically impossible; if Rockefeller can convince potential competitors that he is willing to lose an almost unlimited amount of money keeping them out, it is possible that no one will ever call his bluff, in which case it will cost him nothing. But the advantage in such a game seems to lie with the small firm, not the large, and the evidence suggests that the artificial monopoly is primarily a work of fiction. It exists in history books and antitrust law but is and always has been rare in the real world, possibly because most of the tactics it is supposed to use to maintain its monopoly do not work.
Since competition is efficient, one might think that the solution to the inefficiency of monopoly is to break up the monopoly firm. But if a natural monopoly is broken up into smaller firms, average cost will go up—that is why it is a natural monopoly. Since average cost falls as output increases, one of the firms will expand, driving (or buying) out the others. We end up where we started, with a single monopoly firm.
The inefficiency of monopoly is an argument for breaking up artificial monopolies or preventing their formation by laws against predatory pricing, but I have just argued that artificial monopolies created by predatory pricing are for the most part mythical. It is also an argument for breaking up monopolies created by government regulation of naturally competitive industries. But in the case of natural monopoly, perfect competition is simply not an option. We don't want every small town to have ten grocery stores.
The cure that economics textbooks traditionally offered for the efficiency problems of natural monopoly was government regulation or ownership. One problem with this approach is that a regulator, or an official running a government monopoly, has objectives of his own—some combination of private benefit to himself and political gains for the administration that appointed him. A sensible policy for the regulator might be (on the historical evidence often is) to help the monopoly maximize profits in exchange for campaign contributions to the incumbent administration and a well paid future job for the regulator.
Suppose we somehow solve that problem and put a natural monopoly under regulators who have only the best of intentions. After reading the first half of this chapter, they conclude that the solution is to force firms to charge marginal cost, to sell a gallon of milk, or, more realistically, a kilowatt hour of electricity, at exactly what it costs to produce.
This leads to several problems. The first is finding out what the firm's costs are—real monopolies, outside of textbooks, do not come equipped with a diagram showing their cost curves. One approach is to simply watch, see what it costs to produce each unit of output, and set prices accordingly. But relating costs to output is not a simple matter of observation. To determine marginal cost, we have to know not only the cost of the quantity the firm is producing but also what it would cost to produce other quantities.
A second problem is that the regulator observes what the firm does, not what it could do—and the firm knows the regulator is watching. It may occur to the firm's managers that if they arrange to produce the last few units in as expensive a fashion as possible, the regulators will observe a high marginal cost and permit them to charge a high price.
Suppose the regulators see through any such deceits, correctly measure marginal cost, and set price equal to it. A natural monopoly exists because the cost of producing additional units decreases as output increases, giving a larger firm a cost advantage over a smaller firm. But if marginal cost is falling, then average cost, which includes the cost of the earlier and more expensive units, is higher than marginal cost. So if a natural monopoly is forced to sell at marginal cost it will eventually go broke or, if the regulation is anticipated, never come into existence. To prevent that, the regulator must find some way of making up the difference between price and average cost.
One solution might be a subsidy paid for by the taxpayers. While this arguably makes economic sense it is in many cases not a practical option, since regulatory agencies are rarely provided, by Congress or state legislatures, with a blank check on the treasury. The usual alternative is to get the money from the monopoly's customers. Instead of requiring it to charge marginal cost, the regulators require it to charge average cost, a less efficient outcome but still better than the price the monopoly would set for itself.
How does the regulator find out what average cost is? If he simply asks the firm's accountants to calculate how much it spent this year and sets next year's prices accordingly, the management of the firm has no incentive to hold down costs, especially the cost of things that make the life of management easier. Here again, management knows that the regulator is watching and modifies what it does accordingly.
The real-world version of this approach to controlling natural monopolies is called "rate of return" regulation. The idea is to set a price that gives the stockholders of the regulated utility—the most common example of a regulated natural monopoly in the United States at present—a "fair rate of return" on their investment. The cost of inputs other than the stockholder's capital is set at what the regulatory commission thinks it ought to be, based on the experience of past years.
How much do investors have to get to make it worth investing in utilities? The obvious answer is "the market rate of return"—but on how much capital? If regulators measure the size of the investment by how much investors initially put in, investors in new utilities face an unattractive gamble: if they guess wrong the company goes bankrupt and they lose everything, if they guess right they get only the market return on their investment. So a regulator who bases rate of return on historical costs must somehow add in a guesstimate of the risk premium that investors would have required to compensate them for the chance of losing their money.
What about measuring the current value of the investment by the market value of the utility's stock and allowing the utility to set a price that gives a market return on that value? Unfortunately, this ends up as a circular argument. The value of the stock depends on how much money investors think the company will make, which depends on what price they think the regulators will permit it to charge. Whatever amount the regulators allow the utility to make will be the market return on the value of the stock, once the value of the stock has adjusted to the amount the utility is making.
Regulatory commissions exist in the real world, hold hearings, and publish press releases describing what a fine job they are doing in protecting customers from greedy monopolies. What they really do, however, and what effect they really have, are far from clear. In a famous early article on the economics of regulation, George Stigler and Claire Friedland tried to determine the effect of utility regulation empirically, by looking at the returns to utilities in states where regulation came in at different times. So far as they could tell, there was no effect.
[article link]
One issue that antitrust law has paid a good deal of attention to is the possibility of a firm that has a monopoly in one market using it to somehow get a monopoly in another. A prominent recent example is the controversy over charges that Microsoft is trying to use its near monopoly in the market for desktop operating system to get a second monopoly in the market for web browsers. This issue appears in at least three different legal contexts: Vertical integration, retail price maintenance, and tie-in sales. In all three contexts, as we will see, the legal analysis that has been widely accepted by the courts is inconsistent with the relevant economic theory. And in all three cases, the result of showing that is to leave us with a puzzle. Having shown that the court's explanation of these practices is wrong, we have to explain why they nonetheless exist.
Suppose steel production happens to be a natural monopoly and I have it. It occurs to me that making cars requires steel, and I am the only source. I accordingly buy up a car firm, refuse to sell steel to its competitors, and soon have a monopoly in cars as well. I am now collecting monopoly profit on both the steel industry and the auto industry, so I, and my stockholders, are happy.
What is wrong with this strategy is not that it will not work but that it is unnecessary. If I want to drive the price of cars up, I don't need a car company to do it. All I have to do is raise the price at which I sell steel to the existing companies. The car companies will pay the higher price to me, pass the increase on to their customers, and so provide me with my monopoly return without any need for me to get into the car business.
The reason this argument matters to the law is that one of the things antitrust law regulates is vertical mergers, regarded as suspect on the theory that they make it possible for the monopolist to expand his monopoly. The argument so far suggests not only that vertical mergers should not be suspect but that they should not happen, leaving us with the question of why they do.
One reason, of course, is that it is sometimes cheaper for a firm to make its own inputs or sell its own output, with the result that, even where no question of monopoly is involved, we observe quite a lot of vertical integration. A more interesting reason, where a firm does have a monopoly at one stage of the production process, is that vertical merger is a way of reducing the inefficiency due to its monoopoly and, in the process, increasing the firm's profits.
When my steel monopoly pushes up the price of the steel it sells to auto companies, they respond by using less steel and more aluminum and plastic. To the extent that the substitution is driven by my monopoly price, it is inefficient. The car company is using a hundred dollars of aluminum to substitute for steel that costs it a hundred and twenty dollars to buy but costs me only eighty dollars to produce. That represents a net loss of twenty dollars to efficiency and, potentially, to profit.
One solution is for me to buy the car company. I then instruct its managers that in deciding when it is cheaper to use steel they should base their calculations on its real cost of eighty dollars, but that in pricing cars they should do their best to extract as much monopoly profit as possible. I thus eliminate one of the inefficiencies due to my monopoly price on steel, while still selling autos at a monopoly price and collecting the corresponding monopoly profits.
Retail price maintanance is the practice of a producer controlling the price at which retailers are permitted to sell his products. For many years, federal law permitted states to decide whether or not such contracts were permitted and enforced. Under current law, explicit contracts of that sort are illegal everywhere, although in practice that rule is widely evaded—as you can easily check by a little on-line price comparisons of (for example) the latest models of Macintosh computer.
One argument for banning retail price maintenance agreements is that they are a way in which the producer, who has a "monopoly" of selling his own products to retailers, extends that monopoly to the retail market, presumably in exchange for a share of the monopoly profits that doing so produces for the retailers. Here again, the problem with the argument is not that the strategy would not work but that it is unnecessary. A producer is free to charge retailers whatever price they are willing to pay. If he wants to raise the retail price, all he has to do is raise the wholesale price. Without any price maintenance agreement, the retailers will compete down their margin until it just covers their costs. Instead of getting a share of the revenue from the higher retail price, the producer gets all of it.
Having explained why retail price maintenance does not exist, we are left with the puzzle of explaining why it does, why some producers attempt, where it is legal, to make and enforce agreements controlling the price at which their goods may be sold.
I am a retailer of expensive hi-fidelity audio equipment. In order to sell it, I spend a considerable sum maintaining a showroom where potential customers can listen to different producers' equipment, consult with my expert salesmen, and so decide which products to buy.
Judged by the state of my showroom, all is going well; my salesmen hardly have a free moment. Judging by my books, however, something is wrong; lots of people are looking but almost nobody is buying. While trying to solve this puzzle, I decide I am in need of some fresh air, so go out for a stroll. Just around the corner, I find the explanation—a new catalog discount store, with a small office and no showroom, selling the same products I sell at eighty percent of my price. Taped to the door is a map showing the location of my showroom.
This is a problem both for me and for the producers of the audio equipment I sell. Since customers can get my expensive presale services for free and then buy from my lower cost competitor, I stop offering the presale services. I close down the showroom, fire most of my salesmen, and cut prices to match the competition. Customers no longer have the option of trying my goods before they buy. They respond by going to competing retailers selling different brands of equipment, brands whose manufacturers insist on a minimum price for their equipment sufficient to cover the cost of salesmen and showroom. Those retailers provide the presale services, secure in the knowledge that nobody can undercut their prices.
Long ago, when computers required rooms instead of desktops and belonged only to large firms and governments, there was a company called IBM. Floppy disks had not yet been invented. To get information into a computer you punched it into a large deck of paper cards and ran it through a card reading machine. IBM had something close to a monopoly on selling and leasing large computers. One term in their agreement, one eventually declared illegal, was that customers had to use IBM punch cards. Why?
[case link]
Here again, the obvious answer is in order to extend the monopoly from computers to punch cards. Here again, that answer does not work.
Punch cards are not exactly high tech items; lots of firms could produce them and did. IBM could require its customers to use its punch cards but had no control over what cards were used by people using other computers. If IBM took advantage of its monopoly on punch cards used with IBM computers by raising their price, the result would be to make using IBM computers more expensive. But they could have done that much more easily by simply raising their prices. Insisting that their customers use expensive punch cards instead of cheap ones is an indirect way of raising the price of the computer.
It is tempting to reply that IBM can get away with expensive punch cards because their customers have nowhere else to go. But that is wrong. To begin with, their customers have the option of not using a computer at all, an option many firms took. They also have the option of using computers made by other firms—and will take it if IBM gets too expensive.
The more fundamental response is that if IBM can insist on expensive punch cards without losing any customers, that is evidence that they could also have raised the price of their computers without losing any customers, in which case they should have done so. Once they have gotten to the profit maximizing price, the price at which further increases lose them more in sales than they gain in revenue per sale, any further increase, whether per computer or per card, makes profits lower, not higher.
Again, I have explained too much. Having shown that IBM had no reason to insist on a tie-in between cards and computer, I must now explain why they did.
One mundane explanation is that IBM cared about the quality of the punch cards. If something went wrong, they might have to service the machine, and if too many things went wrong, their reputation might suffer. One way of controlling quality was by making the cards. A similar explanation has been offered for an earlier round of antitrust cases involving a giant company, the IBM of its day, making shoe manufacturing machinery.
[case link: United Shoe]
A more interesting explanation is that IBM was engaged in a clever form of discriminatory pricing. The value of the same computer is different to different customers; ideally, IBM would like to charge a high price to a firm that gets a lot of use out of the computer, and is therefore willing to pay a high price, while charging a lower price—but enough to more than cover production cost—to more marginal users.
Customers willing to pay a high price are unlikely to mention that fact to IBM. But, on average, high value customers are also high use customers. High use customers use a lot of punch cards. By requiring all customers to use IBM cards and charging a high price for them, IBM is, in effect, making the same computer more expensive to customers who use it more. Combining expensive cards with somewhat less expensive computers lets it keep the low use users, who are compensated for the high price of one with the low price of the other, while milking the high use users, who are the ones least likely to abandon their computers.
I do not know when this explanation for tie-in sales was first offered by an economist, but I suspect that a lawyer beat us to it. The earliest tie-in case I have come across involved not a computer but a printing press. The tie-in was with the paper the press used. The attorney defending the company's right to require a tie-in offered a simple explanation. If the company covered all of its costs, fixed and variable, in the price of the press, small printers would be unable to afford it. By charging a lower price for the press and a higher price for the paper, the company made the combination affordable for small printers, who didn't use all that much paper, while covering its fixed cost with the extra money it made from big printers, who did. It was precisely the economist's explanation of tie-in sales as a form of discriminatory pricing—presented, as favorably as possible, from the monopolist's point of view. And it correctly pointed out the efficiency advantage produced by price discrimination—a larger quantity of output, due to the ability to cut prices for some customers, in this case indirectly, without cutting them for others.
While these arguments imply that tie-in sales are sometimes efficient, it does not follow that they always are. One cost of requiring customers to buy expensive punch cards is that they will take expensive precautions to avoid using any more of then than necessary. That is inefficient if the cost of the precautions to the user is higher than the cost of the cards saved to IBM. The inefficiency due to overpricing the cards must be balanced against the efficiency gain due to making computers available to the lower use customers who would otherwise be priced out of the market. There are no theoretical grounds on which we can predict what the net effect will be; it might go either way.
This chapter has been devoted mostly to explaining issues rather than analyzing legal alternatives. One reason is that both antitrust theory and antitrust law are complicated areas and I have done little work in either. But it is worth, at this point, at least trying to summarize the subject from a legal rather than a purely economic point of view.
Antitrust law ultimately involves three different approaches to reducing the costs associated with monopoly: Controlling the formation of monopolies, regulating monopolies, and controlling efforts to misuse a legal monopoly.
The formation of monopolies is controlled in three different ways. One is by restrictions on the mergers of large firms, where the antitrust division believes that the merger will have an "anticompetitive" effect. If two firms wish to merge, one of which controls forty percent of the pickle market and one fifty percent, the antitrust division may decide that ninety percent of pickles is too near a monopoly for comfort. In principle, their decision is based not only on the percentage of the market but also on how easy it is, if the merged firm tries to exploit the consumers of pickles with high prices, for other pickle producers to expand or for new firms to enter the market, and on how willing consumers are to substitute other things for pickles if pickles become too expensive. If the conclusion goes against the merger, the firms have the choice of either remaining separate or having one of them spin off its pickle business before merging.
A second way in which formation of monopolies is controlled is by restrictions on behavior believed to create them, in particular on predatory pricing, selling below cost in order to drive competitors out and establish a monopoly. I argued earlier that such restrictions are a cure to an imaginary disease, but the antitrust division may not always agree. Similar arguments apply to controls over tie-in sales and retail price maintenance agreements.
A final, and perhaps most important, way of controlling the formation of monopolies is by making it harder for firms in concentrated industries to cooperate, to form a virtual monopoly, a de facto cartel, by jointly holding quantity down and price up. One way of preventing that is by refusing to enforce cartel agreements, as is done in the United States and (more recently) the U.K. Another is by making such agreements, including secret price fixing agreements, illegal.
Regulated monopolies in the United States are mostly public utilities—electricity, natural gas, water, telephones—and regulation is mostly at the state level. For reasons I have already discussed, it is unclear whether utility commissions can be trusted to try to produce efficient outcomes, whether they can do so if they try, and even whether they have any significant effect on the industries they regulate. The theoretical rule—set price equal to marginal cost, and find the money somewhere to cover the difference between that and average cost—is straightforward. The practical application is not.
Controlling attempts to misuse a legal monopoly gets us into the topics I discussed under the general subject of extending monopolies. My arguments suggest that, while the behaviors in question may be a result of monopoly and may increase the monopoly's profits, it is not clear that they make the rest of us worse off. It is therefore also unclear whether there is any good reason to restrict them.
For about a decade, roughly the eighties, federal antitrust activity was at a relatively low level, in part due to the influence of the sort of arguments I have just offered, arguments which suggest that antitrust activity often does more harm than good. More recently, it has revived again, largely targeted at the computer industry. The current showpiece is the Microsoft antitrust trial.
One reason may be that software provides a particularly striking example of a natural monopoly. Once a computer program is written the cost of producing additional copies is close to zero, so the more copies you sell the lower the average cost. One result is that, at any given time, there is likely to be a single dominant product in each niche—one dominant word processor, one dominant photo editing program. Current champions are Microsoft Word and Adobe PhotoShop.
Stanley Liebowitz, an economist studying these markets, tried an interesting experiment. He graphed market share in each of a variety of niches against average rating in computer magazine reviews. The pattern was striking. At any one time, there was usually a dominant product. It stayed dominant until one of its competitors started getting consistently better reviews, at which point the competitor rapidly took over the market. In an industry where, once a program is written, it costs relatively little to crank out another million copies, market share can change with startling speed.
Just as the American form of marriage has been described as serial polygamy, so the software industry provides a striking example of serial competition. At any given instant there is a dominant product, but which one it is changes over time. During the relatively short history of the personal computer, the dominant spreadsheet on Intel machines has gone from VisiCalc, the original spreadsheet, to Lotus 1,2,3 to Microsoft Excel. The dominant word processor has gone from WordStar to WordPerfect to Microsoft Word.
You may think you see a pattern there. So did the antitrust division. Microsoft has not always been the winner; Adobe, for example, continues to dominate a group of related niches involving graphics and desktop publishing. But Microsoft's share of successful software applications is high and rising. One explanation offered by its competitors is that ownership over the operating system, first MSDos and later Windows, gave Microsoft an unfair advantage in writing software, since they knew more than anybody else about the underlying code with which that software interacts.
While that explanation sounds plausible, it is contradicted by a striking historical pattern. Insofar as Microsoft has such an advantage, it is limited to machines running their operating systems, so Microsoft applications ought to succeed only, or at least mostly, on Intel platforms. But Word and Excel are not only the dominant word processor and spreadsheet under Windows, they are the dominant ones on Macintosh computers as well.
An obvious explanation is that Microsoft used its operating system advantage to obtain a dominant position in the Intel world and then spread from there to the Macintosh, taking advantage of the desire of Macintosh owners to use products compatible with what other people were using. While that sounds plausible, it does not fit the historical facts. In the early years of the Macintosh, the dominant word processor on Intel machines was WordStar. The dominant word processor on Macs was Word. In both the word processor market and the spreadsheet market, Microsoft first obtained a dominant position in the Macintosh market, where it had no more access to the operating system than anyone else and less than Apple (which produced a competing word processor) and then extended that to the Dos/Windows world.
Network Externalities and the Qwerty/Dvorak Myth
The latest version of an economic theory to explain monopoly and justify antitrust action goes by the name of "network externalities." The underlying idea is that there can be economies of scale associated with consumption as well as production. It is convenient for me to use the same word processor as people with whom I want to exchange documents, so the more people are using Word the greater the incentive for me to abandon my trusty WriteNow and go with the crowd. It is convenient for my telephone to be able to reach as many other people as possible, so the larger the size of the telephone network the greater the value it provides to each customer. As with economies of scale in production, the likely result is a natural monopoly.
The classic example offered for the real-world importance of this effect is the Qwerty keyboard, the arrangement of keys on a conventional typewriter. According to the widely accepted story, the Qwerty layout was originally designed to slow typists down, in order to reduce the problem of keys jamming in early typewriters. It achieved success as a result of being used by the world's only touch typist in a crucial early typing contest. Once established, it maintained its position through the power of network externalities despite the existence of a greatly superior alternative, the Dvorak keyboard. In a world dominated by QWERTY machines, practically nobody wanted to learn Dvorak.
Some years ago, Stanley Liebowitz and Stephen Margolis published an article, "The Fable of the Keys," demonstrating that every single fact in the above story was false. QWERTY was designed to prevent key jamming not by slowing typists but by putting pairs of letters that frequently followed each other on opposite sides of the keyboard, thus alternating between the two banks of keys in the early machines; that pattern is still a desirable one, since it means that typists tend to type with alternate hands, which is faster and less tiring. There were many early typing contests, different machines won different contests, and the recorded scores make it clear that there was no single typist with a large speed advantage over everyone else.
Perhaps their most damning result concerned not Qwerty but its competitor. It turned out that the great superiority of Dvorak was demonstrated only in tests run or supervised by August Dvorak, its inventor. The advocates of network externalities, in taking the Dvorak/Qwerty case as evidence for their theory, were treating advertising puffery as scientific data. Tests by independent third parties interested in the possibility of adopting the new layout showed it to be at most a few percent faster than the existing standard.
Liebowitz and Margolis did not claim that the network externality argument was impossible, although they have argued that it is mostly a relabelling of phenomena already familiar in the context of economies of scale and natural monopoly. What they claimed was that, at least in the typewriter case, its effects were unimportant. After all, while some typists need to be able to move from one typewriter to another, many others do not. Many writers preferred to do all their work on a single typewriter, especially back in the days of mechanical typewriters, which varied a good deal more than computer keyboards today. The costs of modifying a typewriter to change the keyboard layout were never terribly high, and became much lower when IBM introduced the Selectric, a model which permitted multiple interchangeable type balls. They became lower still when the world switched from typewriters to word processors, since a computer's keyboard layout can be remapped in software. Every Apple IIc made came with a built-in switch that toggled the keyboard between QWERTY and Dvorak.
If Dvorak had been as much better as its advocates claimed, it should have rapidly established a dominant position among typists who did not have to be able to use other people's machines. Having demonstrated its superiority there, it should have spread. By now, Qwerty should have been relegated to the dustbin of history. It didn't happen that way.
A similar issue arises in the context of computer software, where compatibility is again of significant value. Here again, Liebowitz and Margolis offer evidence that while the fact that other people are using a word processor may increase its value to me, the effect does not seem to be very large. If the main question deciding what word processor I use is what word processor everyone else uses, and similarly with other products, then the first dominant product should also be the last, since no competitor will ever have a chance to compete. It follows that I must be currently running WordStar or MacWrite, and that VisiCalc still owns the world of spreadsheets. It didn't happen that way.
They also offer some less direct evidence. Suppose there are two equally good word processing programs, one with 95% of the market, one with 5%. If network externalities are important, the dominant program should be worth substantially more than its competitor to users—say a hundred dollars more. The rational monopolist should raise his price accordingly, to take advantage of his customers' willingness to pay. He won't raise it by the full hundred dollars, since at that price his competitor might start to expand, but raising it by somewhat less, say fifty dollars, permits him to both maintain his monopoly and exploit it.
It follows that if externalities are important, the dominant product in each niche should cost more than its competitors. Empirically, that does not seem to be the case. Here again, the conclusion is not that network externalities do not exist but that they do not seem to matter very much, at least in this market.
I have said a good deal about the background to recent antitrust actions but very little about the case that is the current high profile example. One reason is that by the time this book is published the Microsoft case will probably be over and something else occupying the headlines. Another reason is that I do not know enough of the detailed allegations in that case, or the evidence for and against them, to want to offer an opinion as to whether Microsoft has or has not been doing the various wicked things that its competitors accuse it of doing.