Economics and Finance Symposium - Economic Policy Challenges: Microeconomics and Regulation

ROSE: I'd like to welcome everyone to the final panel of today. If you spend much time around MIT, you all know that physics has an uncertainty principle, and as you've discovered today, so does economics. And indeed, in this panel, the uncertainty principle we're grappling with is will our final panelist make it off the plane that is supposed to be coming from Washington, DC landing in Logan in time to join us at the very end of our panel discussion? We'll wait and see.

So we started this morning with a discussion of contributions that MIT faculty and alumni have made to advance economic modeling methodologies and science, particularly in the microeconomic realm, but broadly. But MIT economists have seldom been content with only these objectives as laudable and as invaluable to academic progress as those may be. Indeed, a hallmark of the MIT economics department has been the firm insistence on integrating economic science with applications.

Economics faculty and alumni have at least one foot on the ground at all times. Sometimes, as you've heard from Esther Duflo, both feet on the ground in entire teams in countries around the globe. So we've asked our current panel to switch gears from a focus on the macroeconomy, as uplifting as the last panel's discussion of that was, and shift attention to discuss what microeconomists have contributed to important policy questions and operations of government policy, regulatory policy, and corporations.

And this panel could, of course, as the others, be an entire day of discussion in and of itself, but Jim Poterba has told me that we do have a firm stop, because people might want to be able to eat at the end of the day. And so we will try to limit our time to give you a bit of a teaser to many of the important issues that we could discuss but give you a sense for some of the contributions.

So I'd like to just briefly introduce our panel, and then I'll let them weigh forth. Our first panelist is Dennis Carlton. Dennis has an MIT Ph.D. In economics and is the Katherine Dusak Miller professor of economics at the Booth School of Business at the University of Chicago. Dennis is going to share his perspectives on economic contributions to competition policy broadly-- that'll be a theme of his-- and is well-positioned to do that, both through his experiences, as been mentioned earlier, as deputy assistant attorney general for economic analysis at the Department of Justice, also as a member of the Antitrust Modernization Commission. Policy roles, which greatly complemented Dennis's research expertise in this area, as well as other broad areas of industrial organization, anti-trust, and regulatory economics.

Our second panelist, Dick Schmalensee, is not quite an MIT lifer, despite the dual bachelor and Ph.D. Degrees from MIT and his role as the Howard Johnson professor of economics and management, director of the MIT Center for Energy and Environmental Policy Research, and dean emeritus at the Sloan School. Now why do I say he's not quite a lifer? Well, Dick spent a few years away from us at the University of California, San Diego, which today looks particularly attractive to him.

Dick has had a career that has spanned a broad range of research areas, ranging from industrial organization, management, strategy to antitrust, regulatory, and environmental policies. And I believe Dick is going to weigh in, in particular, to share with us some of the contributions of microeconomics to solving some of the questions that Bob Gordon raised in the last panel, namely how do we think about intelligent energy policy and environmental policy given the substantial challenges that we face not just as a country, but as a world?

Hal Varian, who is an MIT undergraduate and has a Ph.D. in economics from the University of California, Berkeley, gives us a little bit of a change of pace. So how-- well, he has in his past life been a professor at Berkeley in three departments there-- business, economics, and information management. Currently, he's chief economist at Google. And as you know, economics undergraduates and even a healthy fraction of economics Ph.Ds often choose to ply their trade in the private sector.

It's a little less common for faculty to make that transition from academics to the business world, although MIT alumni have done that consistently over time, and some of our most distinguished alumni have done that in very successful careers in both academics and the corporate world, as well as sometimes the policy world.

I think of people like George Schultz, who moved from the University of Chicago to Bechtel or Gary Loveman who leapt from Harvard Business School to become CEO of Harrah's. And today Hal is going to share with us some of the contributions that microeconomics have to understanding the operation of firms in the information economy, the challenges that are posed by some of the really dramatic technological innovations that we're grappling with, and how we might make sense of that.

And then finally, if US Air is cooperative, and Logan snow crews are doing their job, Mark McClellan, who has a Ph.D. from MIT as well as an M.D. and MPA from Harvard, is going to talk with us about health policy. And Mark, again, as you're catching this theme, has not only had a very distinguished research career in this area, but has been a very important policymaker in a number of health-related policy positions.

So without further ado, I'm going to ask Dennis to kick off this discussion.

CARLTON: Thank you. I'm very pleased and honored to be here to celebrate with MIT, and I want to really thank MIT for the fine education they gave me in economics and in operations research, and also thank them for how nice all my professors were and the MIT administration when I was a student.

My topic today is regulation with a special focus on antitrust, and I'm going to talk about three things. I'm going to talk about the enormous impact that economics, specifically industrial organization, has had on antitrust, and the very important effect that antitrust has had on industrial organization. I'm then going to identify a few key hot topic areas that are emerging in antitrust, and finally, I'll conclude by drawing some lessons about how economics can best be used to influence policy here and around the world.

Before about 40 years ago, antitrust was abysmal in its use of economics, both at the government agencies, and in court decisions with rare exceptions. But industrial organization in microeconomics provided a very simple and powerful way to understand competitive behavior amongst firms. And it really produced a revolution in antitrust thinking, and this revolution is often traced to Aaron Director, who taught at the University of Chicago Law School and co-taught, actually, a course in antitrust with Edward Levy, who was not only president of the University of Chicago, but also an attorney general of the United States.

By the 1960s, we heard this morning, and certainly by the 1970s, there was a growing economics group at the Department of Justice. And in the 1980s, the chief economist position was elevated to deputy assistant attorney general, which put it on a par with the other legal deputies. When I was there at the Department of Justice about two years ago, I had about 60 Ph.D. economists under my direction, and they were doing very important research, and moreover, they were listened to.

On most cases, the economists are asked to produce a memo that goes to the front office, together with the legal staff producing their own separate memo. And each are weighed in terms of making the decision. The same is true about the importance of economics at the other government and antitrust agency, the Federal Trade Commission.

Let me talk briefly now about some of the interactions between antitrust and industrial organization and vice versa. Let's start out with one of the key areas in industrial organization and antitrust-- cartels. It's a pretty simple theory. A bunch of firms get together. They collude to raise price. That's bad. So what does the Department of Justice want to do? They want to stop that.

So what I'm going to illustrate will be an example of how some very simple ideas matter a tremendous amount. What we know is from just very basic economics that your incentive to engage in a cartel is going to depend upon the penalties. So the Justice Department looked and noticed that the fines were pretty low. So the fines were raised from 10 million to 100 million. And then they looked about jail time, and they noticed that they weren't sending very many people to jail for very long. So they raised the jail penalty to 10 years.

And then they did something very clever. They introduced what's called a leniency program, and they said, if you're currently in a cartel, and you come to us and you squeal on everybody else, we'll give you leniency. You won't go to jail. We'll reduce your fine, and when you get sued civilly by a private plaintiff, and you're guilty, because you're telling us you're guilty, you'll be subject to damages. Under the antitrust laws, civil damages are trebled.

But if you come to us, and we give you leniency, we're not going to allow the court to multiply your damages by three. So wow. Pretty powerful incentives. And by and large, there weren't a lot of cartels that were being investigated. It was hard to find them. They put in place these three programs that I just discussed. All of a sudden, people are lining up around the block to say, I'm in a cartel, I'm in a cartel, and I think economists were pretty surprised by how many there were. Very simple idea, very effective.

Secondary are government policy mergers. In the early 1980s, Bill Baxter, a former Stanford professor, a law professor, was the Assistant Attorney General, and he produced the merger guidelines. Reissued merger guidelines would be more accurate. But these merger guidelines were based explicitly on economic reasoning, heavily based on George Stigler's theory of oligopoly, and it gave economic structure for the first time to merger review.

And today, the latest econometric techniques and industrial organization theories are being used by the staff of economists I referred to to analyze mergers. And they do very, very good work. While I was at the Department of Justice, I noticed something. I actually wrote a paper about it, and I said, boy, you guys are doing all this work about using fancy econometric models, merger simulations to predict what's going to happen if a merger occurs, and you're using that to decide whether or not to challenge a merger.

Why don't you keep following things? that way we can test subsequently for those mergers that go through. Which of our techniques, which of our theories actually do a good job at predicting, say, the pricing behavior? Unfortunately, we don't do enough of that at the agencies.

Let me turn to the last sort of important element in the antitrust cases through our cartels, through our mergers. The last area is what I'll call section two cases or bad conduct cases. These are often the cases that make big headlines. So there have been a number over the last several years-- the breakup of AT&T, the government attack on IBM, a whole series of Microsoft cases, and a whole series of litigation involving credit cards.

I've been involved in various aspects of all these cases, and my point here isn't to discuss whether these cases were decided rightly or wrongly. But what my point is to illustrate that each of these cases unleashed an enormous amount of research among industrial organization economists about practices that maybe they were vaguely aware of but didn't really appreciate their importance.

So for example, there's now a huge literature on what I'll call dynamic strategic behavior in which a firm that has some market power is able to engage in various types of strategies against its rivals in order to either preserve its market power or to swing its market power from the product over which it has the market power legitimately into areas where it wouldn't otherwise have market power but for its acts.

What are some of the acts that have been attacked? Bundling. You buy my product, this product. Then you have to buy this other product from me. You know, tie-in sales. Issues about how I deal with my rivals who may depend upon me for certain critical inputs and the like. Another area that developed, a booming area, had to do with networks. Network effects occur when, for example, my use of a word processor program is influenced by how many other people are using it, because I want to communicate with them. Huge area that developed.

Another area is this recent area of what are called two-sided markets. What's an example of a two-sided market? Credit cards. How does that work? My usefulness as a customer of having a credit card depends upon how many firms accept that credit card, so there are two mechanisms of payment by which a credit card company at least could get paid either by the customer or by the firm, and there are various types of revenue flows.

The fact that antitrust cases provide practical examples to industrial organization economists is what is motivating, I think, a lot of very good research into uncovering better how current industry practices work. What are some common features of these cases? Let me just mention two. Many of these industries are undergoing rapid technological change. Now we know from a lot of the panels and a lot of the work of economists in this room how important technological change is.

What does that mean for these cases? That means when you're intervening in a case, you want to make sure you're not preventing technical change by what you're doing, and you're not impeding it. That's a pretty hard problem to figure out how firms compete when it's technological competition rather than price competition that is emerging. We have much better models of price competition than we do of technological competition.

And finally, the second common element of a lot of these cases is a vertical aspect. In many of these cases, the firm that's being attacked under the antitrust laws or what I'll call the dominant firm is providing an essential input to its rivals who then compete against that dominant firm in other aspects. And the question is, what are the requirements one firm has with another about duties to deal and on what terms?

Let me just turn now to what looked like the emerging hot topics in the field of both antitrust and industrial organization. And I'll mention just three. First one, I've already alluded to. That's bundling and tie-ins. When a firm sells a product of which it may have some unique power, market power, it can bundle that product and say, you want to buy that product from me? You also have to buy these following other products from me. Or it might package the products together as one product.

Question is when is that an antitrust violation versus when is it, for example, if I'm bundling products together in one product, when is it desirable technological change? How do you strike that balance? That's been a problem, a difficult problem to figure out in some past antitrust cases, and I predict that's going to continue to be a difficult problem that will dominate a lot of research that IO economists are doing.

Second area, important area, has to do with intellectual property. I'll speak about patents. The patent system has been widely studied, and a lot of people have come to the conclusion that it's broken. Why is it broken? A lot of what are called weak patents are given. What's a weak patent? You get something that really isn't that important, an idea. You get it patented at the patent office, and then, perhaps, if you go to litigation, the court says, eh, it's not so important. It's not valid.

How did this problem arise? An interesting way. In 1982, there was a restructuring of the court system, and the normal appellate court system was replaced by a specialized patent court. And people thought that would be good. We'll get a lot of people who have specialized knowledge, and patents will be more accurately given when they're deserved. But what happened was a little different. Unexpectedly, there was an enormous increase in the number of patents that were granted, and this led to the problem that I just alluded to of weak patents.

And what kind of problems did weak patents create? Well, if you're an inventor, and you come up with a new product, all of a sudden you might get sued by 1,000 different people. And this is especially a problem in high tech industries, where design of a particular product can trigger literally hundreds or even thousands of patents. In the antitrust area, this problem has led to what's called the patent ambush problem, and this is a serious problem. In fact, I'm giving a paper on this in two weeks. There's a conference in Europe specifically with one third of the day devoted to the patent ambush problem in antitrust.

What is that problem? That's a problem where a patent holder, in a sense, misleads, misuses the patent system in order to induce firms to make a technological investment that's irreversible. Once they've made it, the patent holder shows up and says, surprise. I have a patent. You owe me a lot of money as a royalty. Creates ex post market power where none might otherwise exist.

The last hot topic I'll talk about has to do with information, and this is a question as to who owns information. So think about the internet. I buy a lot of stuff on the internet. Do I own that information? Does the seller own the information? Does my credit card company, if I pay by credit card, own the information? Does a search engine own the information? Who can use it and how?

When people start using information, when there's a lot of information available, intermediaries are going to develop, intermediaries that bunch together buyers who can negotiate against sellers. Is that an antitrust violation? How about intermediaries that put together a whole bunch of sellers who then can gang up on one buyer? Is that an antitrust violation? When you start thinking about putting together intermediaries or running auctions, you're getting very close to starting to study or restudy how markets form. Economists have not really paid as much attention as they should to the fact that a market is a product that can be viewed as a product that firms produce.

What are the rules that govern a market? Now if you look at something like the Chicago Mercantile Exchange, they now own the Chicago Board of Trade. There have been futures exchanges around for a long, long, long time. They are in the business of creating markets. We haven't done enough study of markets. We haven't done enough study of how the rules get created to make a market work to generate liquidity. And this is going to get close to not just antitrust, but SEC and CFTC regulation of financial markets.

And the final issue that's related to information has to do with privacy concerns. This is probably less antitrust, but I know it's something that, for example, the Federal Trade Commission is very focused on. Well, let me just conclude by asking what lessons, if any, about regulation generally do I draw? Well, I think there are really three points I want to make. The first is that simple ideas have high value added. And that's the thing where I would stress regulatory policy could enjoy large payoffs. The example of the leniency policy was a huge-- the leniency policy was an excellent example of showing how very simple ideas about economic incentives can change things a lot. Second, a lot of regulation, including antitrust, has an international dimension. And we have to cooperate internationally. The difficulty this raises is that many international organizations, foreign organizations, don't quite have the same staff of economists or the appreciation of economics that a lot of US agencies have.

So how do you convince them what to do? Well, I think the best way to do it-- I mean MIT does it by training Ph.D.s who then go to those countries-- but there's another way of doing it, and that is the agencies themselves can run training programs to bring people to the United States and teach them economics. And I actually started doing that at the Department of Justice, and I think that was a very effective way of teaching the staff, not the political appointees, but the staff of international agencies how they should think.

And the final thing I'll mention may seem like a detail, and I used to think it was a detail, but I've now come to realize it's really not. And that is if you look domestically, the institutional structure of the regulatory agency matters a lot. What do I mean by that? There are some government agencies that have what I'll call a public interest standard in which their goal is to advance the public interest. There are other agencies-- for example, the Department of Justice-- where their role is to enforce the antitrust laws and make sure that there is no harm to competition.

So imagine a telecommunications merger. It turns out that both agencies review the telecommunications merger. The Justice Department puts conditions on to make sure there's no harm to competition. The FCC can go much further. They can put on whatever conditions they feel are important in order to promote public interest. My own experience suggests that the more carefully delineated the objectives are of the regulatory agency, the more likely you are to avoid what looked like arbitrary imposition of conditions that might be used to promote sort of the pet idea of one of the agency regulators. So thank you very much.

[APPLAUSE]

SCHMALENSEE: We are fully staffed. As a near lifer, I'm delighted to be here today, and I must say what a privilege it has been to be here these many years as part of this extraordinary, extraordinary community. As I listened to the previous panel, I was struck by the notion that the difference between policymaking in the macro sphere and in the micro sphere is a little like the difference between drivers and mechanics on a racing team.

Everybody knows what drivers do. Everybody knows what macro people do. They do inflation and unemployment, and their stuff is on the front page of the paper. The mechanic is kind of in the garage late at night, dirt under the fingernails, skinned knuckles, nobody watches, and nobody has a clue what they're doing until the car doesn't go.

It wasn't macro policy that got us into this mess, I think one can argue. And, of course, from a micro point of view, the calls for more regulation as a response, the undifferentiated calls for more regulation that we heard in the aftermath of the initial stages of the crisis are a lot like telling a mechanic to use more wrenches. One does need to get closer to the details.

So my focus is to be on energy and environmental policy, but I do want to say a few words about micro or structural policy more broadly that, to some extent, will echo what Dennis said. It's a lot less fun to teach antitrust and regulation than it was when I started, because there are fewer really dumb policies. There are fewer of the antitrust cases that came down in the '50s and '60s that were really great fun to teach to an economics class, because you could pick them apart. And you could say this decision is pursuing-- whatever it's pursuing, it's not pursuing any single objective. It makes no sense. The economics is bad and so forth.

That has been transformed under the influence of economics, and you get decisions you disagree with, but you don't get decisions and arguments in cases that make no economic sense. They're pretty rare and insignificant cases. And we don't regulate airlines, and we don't regulate interstate trucking, and we don't do a lot of things that were fun to talk about in class, because you could say, why are we doing this, and get students to understand.

There are still some policies that fit the bonehead description, and that's because, I think, like the mechanics, we tend to work not on the front pages of the paper, but in the third page of the business section. And the policies are usually covered-- the debates are covered, if they're covered at all-- by reporters who don't actually understand the issues.

The other point to make, and I'll come back to it briefly, is that there are still plenty of analytically difficult problems, policy problems. There may be political problems and administrative problems, but there are also difficult technical problems. Dennis mentioned two-sided markets and credit cards as a classic example.

Well, the Federal Reserve was taxed by the Dodd-Frank bill with regulating an aspect-- let me not get into it-- of the pricing of debit cards. And while there is a literature on this, which I've contributed, Dennis has contributed, others have contributed going back at least a decade, about the only consensus in that literature is that what the Fed did was wrong.

We don't have the right answer. We don't have a prescription that says, marginal cost pricing. Well, we know that's wrong in this case. We don't have that answer.

So let me tell some good news before I tell some bad news. So there has been a lot of progress beyond antitrust in the general area of energy and environmental regulation. We used to regulate the wellhead price of natural gas. That was another one of those great policies to talk about in class, because it served no useful purpose and had a very high cost.

We moved away in large parts of the country in the 1990s from the traditional model of regulating electric utilities. We found ways, following models in Chile and England, of bringing competition into the system, of reducing the scope of regulation, and increasing the scope of competition. Economists played a significant role.

You even now see some regulators thinking about incentives, about the incentives given to regulated firms. One would not read that in the '40s and '50s too often. Since the Reagan administration, cost-benefit analysis has been required in the process of making federal environmental rules, as well as other rules. Sometimes this is even an aid to decision rather than a defense of a decision already made.

But this is now a bipartisan thing. The guidelines for doing cost-benefit analysis don't get torn up every time an administration changes. It begins to look a little bit like antitrust. It's something you do as part of good government, and that is economics. And finally, at least until the recent demonization of cap and trade, policymakers had come to appreciate the ability of market-based mechanisms to achieve environmental goals at reduced cost. The high watermark, at least so far, is the European Union's emission trading system for reducing carbon dioxide emissions-- ironically, a system sold to them by the first Bush administration and then the Clinton administration based on our experience with the acid rain program enacted in 1990.

There are other elements of good news, other elements where other areas, where economics has had an impact on the way decisions are made and on the policies that have resulted. But frankly, it is amazing when you look across energy and environmental policy, the extent to which policy is still inefficiently pursuing goals that don't make much sense. And I think the hard problems really are political, not technical, and really do come from the fact that this stuff tends to be on the second page of the business section.

Let me give you an example that infuriates me, but I'll try to control myself. It's outside this area. It's Massachusetts auto insurance. Now we're the only state in the union that regulates the prices of auto insurance, the only state in the union. You will occasionally see third page discussions of the business section about, could competition work? This is not an open question. But in Massachusetts, it is.

So why doesn't it just get answered correctly? A, there isn't really intelligent public debate, because it's kind of an obscure question, and not that many people really realize that in the rest of the country, you could have the gecko if you want the gecko, but not here. And they don't quite realize there is an agency that has an interest in continued regulation, and there are a set of companies that don't operate outside Massachusetts that have an interest in the continuation of this regime.

That kind of-- it happens in a small room not on the floor of Congress, not in front of the press. It happens in a small room with a few people taking minutes that decisions are made is a reason why bad policies persist. People don't watch the mechanic. People don't understand the mechanic much of the time. So let me give you a few examples of bad policies that have persisted for awhile, and then just to give us a depressing conclusion that tries to echo the macro panel, I'll talk a little bit about renewable energy.

So since 1970, the Environmental Protection Agency has been required to regulate local air pollution under the demonstrably false assumptions that perfect safety is possible and that damage functions are discontinuous. It is sometimes required by law to disregard cost. Second example-- despite the partial success of the 1990s, much of the US electric power sector-- and I should give Paul Joskow credit for pushing a lot of the change that did occur, and Paul being frustrated by the fact that it stalled. And a large fraction of the country didn't move to restructure. About a third of the country didn't move.

Unlike telecommunications, unlike electric power in the European Union, there is no overall effective national policy. There is, in a continent spanning system, a dominance by state regulators. There is an extraordinary heterogeneity in legal regimes, organizational forms, even though the electrons whiz around. It is a system that has been in place more or less since the 1930s and is dead stuck. Huge barriers to reform-- not technical. Not technical. Lots of great economic work has been done in bringing market forces to bear in electric power, but politically, there are awesome obstacles.

Finally, since the 1970s, with all due respect to Bob Gordon, I think discussion of energy independence has driven out much intelligent debate about energy policy. Except for oil, which we don't use to generate electricity, we're self-sufficient. And despite the subsequent discovery of huge oil fields in Alaska, great advances in technology, and ongoing subsidies, US oil production peaked in 1970. Full stop. Peaked in 1970. Has declined since. We will not drill our way to independence. If you listened to the last presidential debates, you would never get that impression. You would get the impression, if we just drilled a little more-- well, that is technically nuts.

So it is frustrating when one works in these areas that level of debate on important questions about energy policy. So finally, let me talk about the whole issue of renewable energy and climate change. I'm glad Bob Gordon mentioned a climate tax. I've worked the climate change issue in one fashion or another since I was at CEA, and it is, to my mind, the most complex public policy problem I've ever seen, in part because of its international dimension, in part because of its inner temporal dimension, in part because of all the uncertainties involved, in part because of all the mechanism design problems involved, but many of the incoming members of this Congress don't believe there's a problem.

That's step one. And step two, we're not going to deal with it. The state of public debate is we're not dealing with climate. I'm sorry. That was yesterday's question. Well it goes on. The substitute seems to have been, we will do renewable energy. I love the sun, and I love the wind. Just let me be very clear on this. How could you be opposed to the sun and the wind? But let's just for a moment get serious. 29 states and the District of Columbia have requirements that certain fractions of electricity be generated from renewable energy. The president of the State of the Union talked about 80T of electricity generated by clean sources by 2035. However, it has been made clear that almost everything is clean, so we seem to be OK there.

But let's be clear what all of this focus on renewable energy is. It is not a responsible reaction to climate change. It deals only with one sector, provides no incentives in transportation, no incentives for energy efficiency in heating and cooling and industrial processes, transportation, provides no direct incentive to back off coal, which is the main way you reduce carbon dioxide emissions. It's not a good way to create jobs. I mean Detroit learned that having a big domestic market doesn't necessarily give you a healthy domestic industry, and switching from one capital intensive way of generating electricity to another capital intensive way to generate electricity is probably not the recipe to cure mass unemployment.

It is industrial policy. It is industrial policy. It is the government picking a winner and saying the Chinese think solar is a big deal, so we're going to think solar is a big deal. Now if you work in the energy and environmental policy field, and you see industrial policy coming up again, despite the governments-- and that's a plural governments-- despite governments' terrible record at picking winners, and even worse record at developing national champions that can compete internationally, it's a little bit depressing.

This has enormous political momentum. The level of debate is very low. Those people who know the issue and have thought about these things are drowned-- their voices are drowned out by the voices of special interests operating like the mechanics in the garage where it's quiet, not on Sunday afternoon in front of the audience like the macro people. We all have envy of the macro people. They get the attention. But I think the fact that they get the attention means that the debates are at a higher level. The fact that it's economy-wide issues means there's less scope for special interest.

A lot of micro policy is frustrated. Even though there's good economics and progress to be made, it's frustrated by the fact that it is very hard to get past special interests when there isn't widespread attention. So I'm trying to match the earlier panels that dealt with the macro economy for being negative. This is about as well as I can do it. There has been progress. There's been a lot of work done here. Antitrust in particular has been revolutionized by economics, but if you look in energy and environmental policy or broadly in areas of regulation, there is much to be depressed about. There are plenty of areas where economists are unanimous and unable to move the needle. But there's always tomorrow. Thank you.

[APPLAUSE]

VARIAN: So I grew up on a small farm in Ohio, and when I was in junior high, a momentous event took place in my life. The Russians shot off Sputnik. That was a satellite. And Congress got very upset about falling behind in the space race and all of this stuff. They passed something called the National Defense Education Act, and it was because of that act and the scholarships that they provided that I was subsequently able to attend MIT. And that's had a huge intellectual impact on my life, so thanks to Sputnik and thanks to MIT.

Now I'm going to be more optimistic. I think I'll be the most optimistic person that you've heard today, because after all, I do come from California, where it was 63 degrees when I left, and I'm counting the minutes until I get back. But we live in a period of great innovation, and I refer to this as a time of combinatorial innovation, and by that I mean, every now and then you get a set of component parts become available that innovators can combine and recombine to create new inventions.

Now this is an old idea. Marty Weitzman, also an MIT grad, has written about it. Joseph Schumpeter has written about it. He wasn't an MIT grad, but hey, nobody's perfect. Think of something like in the 19th century, there were standardized mechanical parts became available. Beginning in the 20th century, you electrification, motors, electric motors, gasoline engines. In World War II, we saw radar, electronics. In the '20s and after, microelectronics, integrated circuits, software, and now the web.

So the nice thing is you've got these components. You can create all sorts of innovations by combining them in novel ways. And I think exploring this idea might help address what Bob Solow talked about this morning of trying to make that connection between investment and technology and how it actually turns into products. So you get a very rapid pace of innovation.

I'll just give you a recent example. Back in 2005, Google released Google Maps, where you could display maps in a convenient way, and a computer engineer named Paul Rademacher looked at this, and he dumped the JavaScript from the web browser that we used to call Google Maps, and he hacked it for awhile, and he was able to combine Craigslist house rentals with the Google Maps and put these two pieces of software together.

So what did Google do when we saw this egregious violation of our intellectual property? We did the obvious thing. We hired the guy. We put him to work building an API so that anybody could do this. And so Google Maps then a few months later, under Paul's guidance, released this API, and there was a huge growth of innovation as people found all sorts of ways to combine this particular component, a geocomponent, a map with all sorts of interesting capabilities.

Now this kind of combinatorial innovation is a global phenomenon. It's particularly global, because the costs of communication have really collapsed in the last 10 or 15 years. So even a very, very small company, 10 or 15 people, can now be born global and has access to communications and a computation infrastructure that even the largest multinationals could not afford just 10 or 15 years ago.

So small firms can use Skype and chat and wikis, docs, source code repositories, email, video, et cetera to work around the globe and around the clock, because you don't have to worry about everybody going to bed at the same time. You're able to continue production in other parts of the world, and there's a whole host of these what I would call micro multinationals that are springing up in various places, mostly connected to universities, where you can see this kind of innovation take place in a very vibrant environment.

And not only do you have this efficiency in development, you can also host your software products at data centers run by Amazon, Google, Microsoft, and other vendors of cloud computing so that they've essentially changed what was a fixed cost and a barrier to entry into variable cost, so you can get firms entering at much smaller scale, and so you can get much more innovation going on. I think this is also leading to a big improvement in business processes. So I think there's an area that we haven't really gotten into much. That's what you might call the nanoeconomics of the firm. I know we're talking about microeconomics, which is taking the firm as given, but withinside the firm, there are business processes and communication patterns that go on. And the fact that we have these highly networked, highly computerized organizations makes life much more efficient.

As a little example, Yan Chan, who's a professor at University of Michigan, did a study where she took students and gave them questions from Google. One team of students went off to answer the questions using the library, and the other team used a search engine, and you could compare the time to answer these different questions.

Turns out it's about seven minutes using a search engine and 21 minutes using the library. So you're saving about 15 minutes a day on a question, and if you take some average numbers for how often you ask a question on Google, that adds up to about $200 billion of savings and lost time.

Now there are some subtleties here, because it used to be asking questions was expensive, or maybe I should say getting answers was expensive. So we didn't ask very many questions, or we only asked really important questions. But now getting questions answered is very cheap, so we ask a lot of them, so there's a bit of a consumer surplus calculation that has to be gone through to do this right. But no matter how you cut it, you get a pretty big increase in productivity in knowledge work.

Now think about the way we do knowledge work. So everybody in this audience is a knowledge worker, I'm quite sure, and one of the things you do is you produce documents. And in most cases, those documents are reviewed by other people. So you get input into what you do. And traditionally, we had technological devices like typewriters and carbon paper, and you would shorthand. People would transcribe your document. It was sent around for commentary, and people scribbled notes on it, and it was all combined into a new document. It went around again.

Then we had great innovations like white out-- that was a great innovation-- and post-it notes, and that made it much easier to do this kind of process of revision. When we went to email, we mirrored that same process of creating and circulating a document and getting comments on it just using a software for.

So it's very inefficient when you stop and think about it, because you could do so much better by having a master copy up in the cloud and having the computer do all the revision tracking and keeping track of who did what, and be able to restore the document to its stage at any point. And you end up with a huge productivity improvement, and hopefully, you get not only better documents, but better ideas and better decisions by removing a lot of the transactions cost from this process.

And by the way, I think it is tricky, because a lot of those transactions costs don't show up in GDP, so it's a little hard to talk about the productivity increase. I think there's definitely a productivity increase, but there's a challenge in terms of measuring it. Now along with this improvement in computation and communication, the cost of experimentation and exploration has dropped rather dramatically. So I talked about combinatorial innovation, where you have lots of innovators trying different things, but any single innovator or any single firm can also try lots of different things. That's called experimentation, and we heard Esther Duflo talk about that this morning.

We saw experimentation in the age of agriculture with the huge productivity improvements that came from agricultural research stations. You look at the age of manufacturing. David Hounshell has written a wonderful book called From an American System a Manufacturer to Mass Production, and you can see these technological improvements that went place because of tinkering of trying to work on improvement in the production process through assembly lines and time and motion studies. Remember in the '80s, the Japanese system of manufacture, Kaizen, which was the idea of continuous improvement in the production process. And so all of these things dramatically improved productivity by experimenting and trying to find better ways to do things.

Now in the internet age, the fantastic thing is that experimentation and continuous improvement is just dramatically cheaper, because you can design software, so it allows all the parameters to be variable, and then you can implement a system, try out different configurations in real time, and really get a continuous improvement in the processing of your system.

So for example, at Google, we are firmly committed to experimentation. Last year, we ran 5,000 experiments in the search engine side of the business. That resulted in 400 improvements in search quality, and on the ad side, we ran about the same number, 4,000 to 5,000, with, again, several improvements, and at any one time, there's about 500 experiments running on both the ad side and the search side, so 1,000 all together. And anytime you use Google, you're probably in a dozen experiments or a half dozen experiments at least.

So this is a system that is continuously improving, because we're continuously running experiments and trying to improve the system to make it function better. And it's far, far easier to do that in the web environment than it is to do it in Africa for sure, but we know the value of experimentation really pays off. I'll give you an example there.

Back in 2006, Google wanted to get into voice recognition. Now we had no data and no expertise, so the first thing we did is we hired some expertise. We got some of the world experts in this area, and then we created a system called GOOG-411, which was an automated information system, where you could call GOOG-411 and say, pizza, Mountain View, and it would come back and say, do you mean Joe's Pizza on 123 Main Street? And you could say yes, or you could say no. And that allowed you to find what you wanted.

Now the great thing about it is it was a light client on the phone. In fact, it was virtually no client. And all the calculation took place in the cloud. It took place on the Google servers. So this system could learn in real time. When it said something and thought it recognized what you said, it said yes or no, and you answered yes or no. And that was a data point for it to learn from.

And so in the space of a year, we had an absolutely fantastic state of the art data system, and even now, if any of you are using Android phones, it works exactly the same way. When you talk to your Android phone, it is learning from your response whether it's got the right voice recognition, and this system is continuously improving to recognize a variety of accents, a variety of languages, and you can do this in a real time. And by the way, you can do the same thing with optical character recognition, so you can get a much, much better system there by doing the same thing.