Skip to contentSkip to navigation

An Interview with Paul M. Romer

Stanford university economics professor Paul Romer explains how the knowledge economy is creating monopoly power and changing the nature of competition. This stems from the ability of knowledge-based industries to generate increasing returns by capturing as much market share as possible. Unlike traditional monopolies, however, information age giants will face being superseded by new entrants and little danger exists of permanent monopolies -- even in an industry like software.

(originally published by Booz & Company)

It is hard to label an economist who just turned 40 as venerable, but that is what Paul M. Romer's peers call the Stanford University economics professor. Professor Romer, who also teaches at the University of California at Berkeley while conducting research at the Hoover Institution at Stanford and the Canadian Institute for Advanced Research in Toronto, is best known for his theories about the dynamics of growth. In Professor Romer's view, knowledge is the unsung hero of the growth game. While most classical economists -- not to mention Marx -- focused on production, labor and capital, Professor Romer added knowledge and technology to the mix.

The classical approach operates well in the physical economy of resource extraction and commodity production, according to the professor. That economy is characterized by diminishing returns, since each additional ton of copper or barrel of oil is harder to find than the previous one and is -- by definition -- scarcer and therefore more expensive to extract from the earth.

The information economy is different, however. Large upfront costs are incurred to write a complicated piece of software or to discover specific gene sequences. But after the initial work is done, the cost of each additional "unit" is minimal or sometimes even nil. Software can be copied onto disks at very low prices or sent out over the Internet. Each gram of protein produced by genetically altered bacteria only adds to the world's supply. As a result, while returns diminish in the physical economy, they increase in the knowledge economy -- a cause for hope, according to Professor Romer.

There is another important tenet in the professor's world view. It is that the more we discover new things, the better we get at the process of discovery itself. Knowledge builds on itself. As a result, the capacity to create wealth and value increases over time, surely another reason for optimism.

Professor Romer's work is serious and academic. But while most economists have their cadre of on-the-one-hand, on-the-other-hand adherents and detractors, Professor Romer's work has so far attracted far more support than dissent among his peers, prompting speculation that he may someday be a recipient of the Nobel Prize for economics. Economist Paul Krugman and management theorist Peter F. Drucker are among Professor Romer's biggest and most public boosters.

Professor Romer recently spoke with Strategy & Business at his Stanford office, in Palo Alto, Calif. What follows are excerpts from that conversation.

S&B: For the most part, knowledge, like technology, has been taken for granted by economists who look at the factors that propel growth. But in your work, you assert that knowledge can raise returns on investments and that it is a factor of production, like capital, labor and raw materials. Does the information economy work differently from the classical economy?

Paul Romer: Let's back up. My work on growth can be traced back to an attempt to isolate the differences between the information or knowledge-based economy and what came before it. My belief is that those differences are important for our understanding of growth. Those distinctions matter to people running firms and they should matter to policymakers. They are issues that show up at a number of points in the economy. And since those issues are topical right now, it makes my work in this area a lot of fun.

S&B: What differences are you referring to?

Paul Romer: Let me articulate it this way. One feature of knowledge can be summarized by Isaac Newton's statement that he could see far because he could stand on the shoulders of giants. In other words, his notion was that knowledge builds on itself, which means that as we learn more, we get better and better at discovering new things. It also means that there's no limit to the amount of things we can discover.

This is a very important fact for understanding the broad sweep of human history and it is very different from what we are used to thinking about in terms of physical objects, where scarcity is the overwhelming fact with which we have to deal.

S&B: You say we are getting better at finding new things. What exactly do you mean by that?

Paul Romer: There are only a finite quantity of things with which we can work -- basically, the matter in the earth's crust. We've had essentially the identical amount of physical stuff for millions of years. Now a lot in the language of economics -- and in the popular language, as well -- makes it appear as if we are wealthier today because we have more of this physical stuff. It makes it sound as if we have actually produced more physical things. But in truth, that is not right. If you think about it the way a physicist does, the law of the conservation of matter and energy states we have essentially the same quantity of things we have always had.

S&B: So what have we done?

Paul Romer: We have taken the fixed quantity of matter available to us and rearranged it. We have changed things from a form that is less valuable into a form that is more valuable. Value creation and wealth creation in their most basic senses have to do with taking physical objects and rearranging them.

Now, where do ideas come in? Quite simply, ideas are the recipes we use to rearrange things to create more value and wealth. For example, we have ideas about ways to make steel by combining iron with carbon and a few other elements. We have ideas about how to take silicon -- an abundant element that was almost worthless to us until recently -- and make it into semiconductor chips.

So we have physical materials to work with -- raw ingredients -- which are finite and scarce, and we have ideas or knowledge, which tell us how to use those raw materials. When I say there are always more things to discover, what I mean is that there are always more recipes that we can find to combine raw materials in ways that make them more valuable to us.

S&B: What you are saying sounds straightforward enough. What does it mean for the economy and for growth?

Paul Romer: The claim I made a moment ago about standing on the shoulders of giants tells us something important. It says that we can take advantage of a form of increasing returns in the process of discovery. Now it could have been -- I suppose -- that with each new discovery, it got harder and harder to make additional discoveries. In that case, we would -- at some point -- simply give up. Progress would be slowing down and eventually would come to a halt. Of course, this is not what we see when we look at history. From one century to the next, the rate of technological change and the rate of growth of income per capita has been speeding up.

S&B: This is contrary to discovery in the physical world, isn't it?

Paul Romer: Yes. The physical world is characterized by diminishing returns. Diminishing returns are a result of the scarcity of physical objects. One of the most important differences between objects and ideas, the kinds of differences that I alluded to before, is that ideas are not scarce and the process of discovery in the realm of ideas does not suffer from diminishing returns.

S&B: One of the differences? Are there other differences as well?

Paul Romer: Yes. Let's take a particular piece of knowledge, such as a piece of software, and think about the costs of production that a supplier faces. If we do that, we see that this piece of knowledge is very unusual from a traditional economic point of view.

To make the first copy of Windows NT, for example, Microsoft invested hundreds of millions of dollars in research, development, testing and so forth. But once Microsoft got the underlying bit string right, it could produce the second copy of Windows NT for about 50 cents -- the cost of copying the program coded in this string of bits onto a floppy disk. Since then, all subsequent copies of the program have the same cost or even lower costs. For example, if Windows NT is distributed over the Internet, the cost of every additional copy is basically nil. So the first copy costs you hundreds of millions of dollars, but all other copies are free, no matter how many you produce.

This kind of falling cost per unit is a different manifestation of increasing returns. This is very different from the physical economy, where an important part of the cost of each good comes from the process of making an additional copy of that good.

S&B: Paul, is this kind of increasing return the same as or different from the standing-on-the-shoulders-of-giants effect that you mentioned before?

Paul Romer: These two kinds of increasing returns are logically distinct. For example, we could have falling costs of production for copies of a specific piece of software even if the standing-on-shoulders effect were not operating.

Suppose, for a moment, that Windows NT were the last piece of useful software that could ever be written. Once it is finished, it will be impossible to develop any new pieces of software. I know this sounds crazy, but bear with me. This would be an extreme case that is the opposite of the shoulders-of-giants effect.

But even if this were true, the production costs for NT would still have the unusual character of huge costs up front and very, very low costs for each subsequent copy. That feature is very important for understanding what we call the industrial organization of the economy, because it means that you are going to see a great deal of monopoly power in the new economy.

S&B: How does knowledge lead to monopolies?

Paul Romer: Traditionally, in economics, you would assume that there would be a couple of leaders in a field, that their success would make it attractive for other people to come in, that there would be rivalry among all the firms and that the industry segment as a whole would grow and benefit as competition intensified.

For example, in the physical economy, suppose you've got a firm that owns a big ore deposit somewhere in the world, and there are other firms that own other ore deposits in other parts of the world. Each of these firms will operate under conditions of increasing costs and diminishing returns. One of those firms may be able to produce 100,000 tons of refined ore per year. But it can't easily scale up to 200,000 or 300,000 tons because of the nature of its physical resources. There's only one way to mine the ore and it gets more expensive as you do it faster. The physical structure of your assets means that there's a limit on how fast you can exploit it.

So you might ask: Will one company take over the whole world's supply of ore and become a monopolist in world copper?

The answer is no, because the firm that wants to do that would produce up to a point where it would become more and more expensive. Other firms would be drawn into the market and might produce and sell their ore more cheaply. When firms face increasing costs and diminishing returns, no single firm can supply the whole worldwide market. If it tries, it faces increasing cost disadvantages relative to its competitors.

S&B: So there are limits to monopoly power in the physical world?

Paul Romer: Yes. In the physical economy, there is a natural equilibrating process. If one firm tries to take over the whole market, other firms would gain an advantage and will enter. When this whole process settles down, there will be many firms and a competitive market.

S&B: How does this differ in the knowledge economy?

Paul Romer: Let's think about operating systems and the world of computers for a moment. Let's say you have one firm, like Microsoft, which is going to produce Windows NT. Once it produces its first copy, it then faces no cost disadvantage. It can produce millions or billions of copies of NT at little additional cost. It truly can supply the entire worldwide market for operating systems. If anything, it probably gets even easier for Microsoft because the larger its market size gets, the more attractive it is to adopt its software because of what we call bandwagon effects. If everybody else is using a particular piece of software, it usually makes it advantageous for you to use it, too.

Now this difference between diminishing-returns industries and increasing-returns industries completely changes the dynamic of competition. Under conditions of increasing returns, competition is driven by various firms trying to capture as much market share as possible as quickly as possible and by being the first to develop a product and to flood the market with it, even -- at least initially -- at a loss. To see how different this is, think again about mining. If you discover a large deposit of copper ore, do you think you would want to give lots of it away for free to increase your market share? Of course not.

S&B: What does this mean in a business sense?

Paul Romer: First of all, it means that the kind of strategies you see in the knowledge economy are those like Netscape's. In that strategy, a new company comes into a new market and gives away its browser and tries to move very rapidly to supply a large part of the market. Netscape did this and it ended up with a large percentage of the total market for browsers. It is now trying to leverage that advantage into a larger advantage in the more lucrative market for server software. That kind of competition is driven by falling costs and increasing returns in the information economy. That is the way competition unfolds in the information economy.

So we are talking about two extremes: One is software, the other is resource extraction. A lot of traditional business fits somewhere in between.

S&B: How are monopolies sustained in the knowledge economy? Don't new companies come along with new products that replace the old ones?

Paul Romer: Yes, that is true and it brings us back to the two notions of increasing returns that we were discussing a minute ago. Recall the thought experiment where we shut down the increasing returns associated with the shoulders-of-giants effect but assumed that the increasing returns associated with cheap copies were still present. This leads to a rather pessimistic implication. A firm like Microsoft would become the monopoly supplier of operating systems software. It would remain a monopolist forever because no entrant could compete with it with a similar product and, according to our assumptions, there was nothing better that was left for someone else to discover.

Fortunately, we don't get this pessimistic result because the two kinds of increasing returns go together. This means that even though one firm may control a large fraction of the market at any point in time, there is another way for new firms to compete. Instead of trying to enter the market by providing exactly the same product as the incumbents at a lower price, they can enter and compete by selling something that is new and better. This is precisely what Microsoft and Intel did when they displaced I.B.M. as the dominant force in the computer market. It is what firms like Netscape, Sun and Oracle are trying to do now by leveraging the emerging power of the Internet. Economists call this monopolistic competition. It's a form of competition between different firms, each of which sells a different kind of product and can behave like a monopolist at least temporarily.

Of course, it takes various kinds of institutional infrastructure to make this system work. For example, the Government has to grant property rights over intangible assets like ideas.

S&B: The property rights you are referring to are patents, copyrights and so on?

Paul Romer: Yes, and these differ from the property rights we are familiar with for physical objects. You can give dispersed ownership to something like timber or ore. Many different people can own different pieces of the same kind of asset, so they can all compete to supply ore. But if you consider, for example, a pharmaceutical product, you see that once the Government gives a firm a patent on a new drug, this firm becomes a monopolist with regard to that piece of knowledge, that recipe for how to make that drug. We give this company a legal right to keep everyone else from using its knowledge, and this means that the company can gain the dominant position in the market. So it is the combination of low replication costs to the producer and the protection of intellectual property rights that creates the monopoly position.

S&B: This means that competition tends to take place via the introduction of new goods rather than by competing on price in existing goods?

Paul Romer: Yes, that is right. It is a very different vision of how competition works and why markets are so successful in generating high standards of living. You see, the traditional intellectual justification for laissez-faire comes from dealing with a hypothetical world filled only with scarce objects. In effect, you're just trying to figure out which mine should produce how much copper and how much total copper should be produced. In such a world, the classical notion of perfect competition is feasible and it works well. Competing firms will end up charging the right price for copper in a precise sense. The market price will determine the total amount of copper that should be produced. It will also determine how much of the total each mine will produce. Classical economists were able to show that under competition, these production and allocation decisions are efficient.

In this kind of world, it is important for policymakers and the Government to step in and break up any monopoly or cartel. By keeping the cartels at bay, the price will adjust to the efficient level and the right quantities will be produced and sold.

Now let's go to this new world, where the key challenge is to produce and distribute knowledge or ideas. The first thing to notice is that you lose the classical notion of the right price. For example, with the concept of a vaccine, the best thing to do might be to give any firm the right to use this concept for free and let everybody in the world use it to make serums so that people in general can benefit. To do this, you want the price for using the concept of a vaccine to be zero after it has been discovered. Thus, for example, we could have granted a perpetual patent on the basic concept of a vaccine to Edward Jenner, who discovered the concept in the 1700's. In fact, he and his heirs have no intellectual property rights over this idea. They collect no income when researchers and doctors all over the world use vaccines and the concept of immunity to protect us from disease. After the fact, this is efficient.

But let's back up for a moment. If you want to get people to discover extremely valuable concepts like vaccines, you would like to offer them really high prices for their discovery. This creates an incentive to do the work necessary to make the discovery. So, in this world of knowledge and recipes, there is always an unavoidable tension between wanting to have low prices after the fact, so a knowledge-good can be distributed widely, and wanting to promise strong property rights and monopoly protections in advance, as an incentive and a motivation for discovery.

S&B: How do you resolve the problem, Paul?

Paul Romer: What we have to do is work out a balance between tolerating some monopolies and monopoly profits -- since that's how we motivate people to discover new recipes -- with competition to keep prices low and distribute products widely. We encourage this competition by granting property rights that are partial or incomplete. In practice, what this means is that while in the physical economy, with diminishing returns, there are perfect prices, in the knowledge economy, with its increasing returns, there are no perfect prices.

S&B: Are you suggesting that we must simply tolerate monopoly prices?

Paul Romer: Yes. I'm also suggesting that we have to establish property rights that are incomplete. In the world of objects, you don't have to make these kinds of compromises. We use Government policy to limit monopoly. We also establish very strong property rights. For example, it would be very wasteful for property rights on land to expire after a certain period of time and then for everyone to be able to use it freely. But this is precisely what we do with patents. One key difference is that in the world of ideas, you cannot have both strong property rights and competition. The other is that you don't get congestion over the use of ideas. When we look back on systems where a group of people did have free access to a physical resource like land, we speak of a "tragedy of the commons." But there is no similar tragedy of the intellectual commons.

Now, as I suggested above, if there were not all these additional things to discover, we would be pessimistic about this new world of ideas because of all the monopoly power we are creating with patents and copyrights. In fact, if there were nothing left to discover, we would treat all knowledge the same way we treat the knowledge behind the principle of vaccination. It would be common property and everybody would be free to use it.

But fortunately, in the knowledge economy, there are always new things to discover. We grant property rights over knowledge, and this leads to a leapfrogging process whereby the potential of future monopoly profits induces new discoveries. As a result, a new entity will emerge and come into a market at some point and leapfrog all the existing entities. When that happens, you'll get a big jump in terms of productivity and economic value, and the old monopolists will typically be displaced. Remember, this is a different dynamic from the one we described in the classical economy, where all competition was to produce existing goods at lower prices.

So over in this idea part of the economy, we worry much less about having the Government actively intervene to strike down any hint of monopoly power and keep all firms small. We rely much more on the process of what Schumpeter called "creative destruction," where you have a sequence of temporary monopolies that are superseded by new monopolists selling new products and services.

S&B: What are the policy implications of the knowledge economy?

Paul Romer: There are big implications stemming from the recognition of the difference between the classical world of objects and the idea-based world. With objects, we have a good idea of what the right institutional arrangements are. We just establish infinite-lived property rights, make sure that there is no monopoly control and then let the market operate. This will lead to efficient allocations of physical objects. That is the lesson of laissez-faire, and it is still a very powerful lesson, one that too many politicians still do not understand.

S&B: And the aim of laissez-faire in that respect is greater economic efficiency?

Paul Romer: Yes. But more importantly, I would say the aim in both of these worlds is to generate or create more value. The kinds of problems you've got to solve in the classical realm are relatively simple. You just have to allocate scarce commodities among alternative users and between alternative producers.

S&B: It would seem that the aim of regulation in the commodity-producer world is to foster efficient production, while the aim in the knowledge world is to bolster future discovery. These are different roles for the Government, aren't they?

Paul Romer: Yes, they are. There is something inherently static about the classical vision -- that if there were nothing changing, nothing new being discovered, the price systems would be a good way to decide which mine should produce the next unit of copper. It also works well to decide which power plant should generate the extra megawatts of electricity and so on.

But over here, in the knowledge economy, there are these other imperatives. We have to worry not just about getting efficient usage from what we've got right now, but to figure out ways to discover all the new things we might need or be able to use. That is not a set of static problems. It is very dynamic.

S&B: Do the right institutions exist to make those policy choices?

Paul Romer: What's interesting about the knowledge economy is that we haven't figured out what the optimal institutions are. That's still a wide-open question. What is the best way to structure our economic world? This is true not only at the level of policy formation for a nation, but also for firms. Firms are really struggling with how they should organize themselves internally.

S&B: What are some of the policy issues within firms?

Paul Romer: The traditional logic within companies was like that in government. In the Government, people might look at medical discovery, for example, and say, "Gee, this is a market failure, we can't make the market work perfectly here." So the Government would then step in. It would collect taxes, use the revenue to pay for research taking place at universities and use that research to develop discoveries, like the principles behind vaccines. It would then give this knowledge away for free. We invented non-market institutions like universities to aid us in the production and distribution of ideas.

Firms, to a certain extent, have mimicked that solution. They say, "O.K., we need to generate new knowledge. We have different divisions that we can treat as profit centers. So, what we will do is tax those profit centers to create something that looks like a mini-internal university -- the R.&D. department." They then gave the funds to the R.&D. department and told it to do good work, just like the national Government does with the universities.

Over time, when the R.&D. department produces something, it gives that something away for free -- gives this knowledge away -- to any of the operating divisions that needs it.

S&B: That's the classical, pre-breakup AT&T-Bell Labs model.

Paul Romer: Yes, but you also saw it at I.B.M. and at G.M. and on smaller scales in many firms. A number of firms mimicked the institutions and economic policies of government when they realized that knowledge-creation and discovery was central to their growth.

S&B: What came next?

Paul Romer: The firms found that these institutions did not work so well, so now they are trying to come up with modified institutional arrangements. They are recognizing that the simple solution that recreates the Government's command-and-control and tax-and-subsidy mechanisms is not the perfect solution. What they are finding is that while there are a billion haystacks in which there will be some very valuable needles, it is an enormously expensive proposition to go about looking underneath every one. So what they are asking themselves is, how do you allocate the resources that you devote to research most effectively? They are realizing that they cannot just give scientists lots of money and let them follow their curiosity. If they do, this form of tax-and-subsidy system runs the risk of dissipating efforts by looking in too many different directions that don't necessarily lead to the highest returns for shareholders. One way to think about the problem here is that a tax-and-subsidy system does not have the built-in checks on wasteful activities that are present in a market system.

As a result, firms are beginning to create market-like mechanisms that impose market tests on things like R.&D. departments. In some instances, this goes all the way to the extreme of the profit-center model. In these instances, they have set up R.&D. units as profit centers and they charge different divisions for any of the results they produce that other divisions use.

S&B: Does that work?

Paul Romer: Well, it gets back to one of the problems with knowledge, which is, if you charge high prices for knowledge, you don't end up with the efficient distribution of that knowledge.

Let me provide you with an anecdote. Someone at a company I know told me that he had taken over the market research division. The company used to be operated on the tax-and-subsidy model. The corporation used to give the marketing division lots of money and it went out and worked on anything that it wanted to. That system meant that the department was not really working on the right problems. So this manager took the market research budget and allocated it to all the different business units and told the units that they could buy whatever they wanted from the researchers. This of course made the market research folks focus more on the actual marketing problems that the different divisions were facing.

But then he ran into a problem. One division would pay the market researchers to work on problem X. Another division in the corporation would then come in and want the group to do a study on a very similar problem. It was inefficient for the market researchers to go out and redo the study, but the first division had property rights over the existing study. The market researchers could not just give it to the second division without getting permission from the first, and complicated negotiations would sometimes ensue over what the second division should pay the first. This sometimes led to restrictions on flows of information that would have helped other divisions, and this is clearly bad from the point of view of shareholders.

The lesson is that as soon as you start to price knowledge, you get into awkward situations where your knowledge is not being as widely used as it could be. This is just a fact of life.

S&B: So how do you price knowledge? What is the model?

Paul Romer: There is nothing so far that is perfectly neat and clean. My guess, however, is that the answer lies between the two extremes of the market system and the tax-and-subsidy system. My guess is that we're going to start to find much richer institutional arrangements to control the flow of information.

For example, firms are doing things like making people in the R.&D. units go out and spend time with the divisions, just so that they communicate more closely with people in the divisions. Firms are even making R.&D. people go out on sales calls to talk to the customers, to see what they are
really interested in.

This is one way to get people in the R.&D. unit focused on their markets, their market opportunities and their final customers. It's a weak, but apparently effective, way to impose a kind of market discipline on what the researchers do.

But it is an institutional arrangement, not an explicit price system. Firms could eventually augment an institutional arrangement like this with things like promotion and compensation systems, which reward being attentive to customers, as opposed to more traditional systems, which just give rewards for the number of scholarly publications, or the number of patents, whether they are useful or not. So what I expect is that there will be a lot more experimentation with arrangements that are neither pure market systems nor pure tax-and-subsidy systems.

S&B: Another model is to view the R.&D. department as the center of the firm and the rest of the company as the channel for getting the discoveries of the R.&D. department to market.

Paul Romer: Imbedded in your question is what people call the linear model of science and discovery. It was one in which the people in the research lab said, "Look, the whole game is to create knowledge. That's where value comes from. We're the center of that process. We're going to create the knowledge. Once we've created that knowledge, we'll give it to you, so you can market it and make a profit for us."

In practice, that turned out to be a very bad way to structure corporations and a very bad way to think about economic activity. The problem is that it is not well guided. There are many different kinds of knowledge that researchers can produce. Some of these will be more valuable than others, and researchers may not be in the best position to tell which ones are the most valuable ones. So what's happened is that we've kind of turned this process around.

The new perspective says, "Look, the most valuable things we could do will be things we learn about from the problems we are facing in the field." In terms of information flows, that means it goes from the customer to the operating people to the R.&D. people.

The point I am making is that the problem-solving agenda should not be determined by the R.&D. unit, but by somebody else. So even though we now recognize that knowledge creation is really central to the business process, we don't think of it as being a kind of pyramid model, where the R.&D. unit is at the top and everybody else is kind of subservient to it.

It is very important to realize that the immense potential for new discovery brings with it an even more immense potential for wasted effort. Discovery is therefore associated with really hard choices. For example, what's the set of all possible software programs that could be written? If you just do a simple mathematical calculation about how many bit strings could fit on a CD-ROM, it is an unbelievably large number, far larger than the total number of particles in the universe. It's larger than any physical quantity that we can understand. Most of these possible bit strings would be useless junk. A few of them will be the "killer apps" of the future.

What's the significance of this? At the same time that we know there are good things out there to discover -- new pieces of software -- it would be very easy for an organization to produce a lot of software that isn't very useful. As a result, we need market-like mechanisms to focus the research efforts.

Reprint No. 97110

Joel Kurtzman,

Joel Kurtzman is editor-in-chief of Strategy+Business.

Get s+b's award-winning newsletter delivered to your inbox. Sign up No, thanks
Illustration of flying birds delivering information
Get the newsletter

Sign up now to get our top insights on business strategy and management trends, delivered straight to your inbox twice a week.