Ariel Pakes is the Thomas Professor of Economics at Harvard, where he teaches courses in Industrial Organization and Econometrics. His research has focused on developing methods for empirically analyzing market responses to environmental and policy changes, and has done work for a number of consultancies, government agencies, and large firms. Much of his methodological work has been incorporated into the way government agencies evaluate the likely impact of policy changes. Ariel has mentored over fifty doctoral students, many of whom are now leading researchers at prestigious institutions. He received the Frisch Medal of the Econometric Society in 1986, was elected the Distinguished Fellow of the Industrial Organization in 2007 and in 2017 received the Jean-Jacques Laffont prize, after which he was kind enough to spare the time for this interview.
What future do you see for empirical industrial organisation? What new fields could it be applied to?
Industrial Organisation (IO) is getting richer all the time, it can explain many things. Most of what is happening now is that certain subfields are picking up on IO and using the tools that we have developed to analyse different policies. For example, in the U.S. the big issue is health insurance, Obamacare. There has been a lot of work on the implications of the entitlement program and its costs, and the incentive schemes that could make the cost of U.S. healthcare go down at no cost to quality.
A lot of IO tools are also being used for environmental policies – carbon taxes, standards on cars for emissions, coal plants, and public finance. For instance, for the study of the implementation of taxes and other government policies, demand systems developed in IO are used to evaluate welfare.
For people like me, other than when I do methodology rather than actual applications, there are several issues open, such as moving to a more careful analysis of dynamics which will require more theory as well as more econometrics, more data.
Dynamics in our field means the development of products, R&D, telecommunication. We do not quite have our models ready to analyse that, but we’re getting closer! You don’t stop doing research because you are not at a place where you can give answers yet, or you will never get to a place where you have answers.
In your talk you said that one of the goals of IO is to have more productive industries producing more output, and studying the impact of a given policy on productivity. Do you think patents actually encourage innovation, with the current state of intellectual property protection and the theft of property rights ?
It is a field where it is very hard to do research; part of the reason is that the outcome of the activities are truly random, and they are correlated across individuals. A good example would be that if you could cure the common cold you would make a fortune, so there’s been a lot of money put into that. But science just doesn’t have an answer so it doesn’t matter that a lot of money is being put into it and that different people are doing it. It’s a harder area to do research than most.
People don’t understand enough the implications of policy. In the US now, there’s a lot of cry for a policy to keep prices of drugs down. I mean, we should all be aware of the fact that the marginal cost of most pharmaceuticals is near zero. But if you tied the price of pharmaceuticals to marginal cost, all the drug companies would shut down tomorrow, because all the money is put into the R&D, they do research on many things before they find one thing that’s good, so the typical cost of a drug is a billion dollars just to produce it, on average.
I think you have to be very careful not to decrease the incentives of the drug industry, which has probably been the biggest source of productivity gain over the last 20 years. If you ask which industry produced more consumer surplus than any other industry it’s probably pharmaceuticals, so the last thing you want to do is kill pharmaceuticals.
I think one of the big issues in the US now is intellectual property protection and there are difficult questions involved. There are patent trolls who pick up many patents and then try and sue people who are using their patents: it mostly goes to litigation, it’s expensive for all sides but I’m not sure we know how to get around it and the research around this has been inconclusive.
When you’re dealing with intellectual property you have to be very careful with the incentives for innovation. That’s why we have patent laws, that’s why we have R&D subsidies and it’s been terrifically successful for the country, for the world as a whole. You don’t want to disincentivize it. I’m not sure we know enough to go very far beyond that.
I’ve been working on the drug industry quite a bit, and something to remember is that the US produces and prices drugs and then sells them to Europe, including France. The French government authorities then actually determine a price for the drug, which is always much lower than the price in the US. There’s a sense and a feeling in the US that we’re funding your drugs and therefore the benefits of drugs for the European continent and people are quite unhappy with it.
In 2014 you were kind enough to have an interview with our magazine and you said that researchers were likely to start worrying about different equilibrium assumptions in the future, bringing in significant complexity that would require finding a way to simplify the analysis. Do you think that progress is being made ?
There has been progress!
The thing you have to remember about IO is that most of us understand that we’ll never get it exactly right. A market is just too complex for that, there are too many issues involved and that’s not our goal. It is just to make our models better than the next model. Somebody’s going to make a decision along the way both in the firms and the regulatory authorities, and the only question is if we can make the decision better because we’ve dug into it deeper.
One of the big issues in IO has been bargaining between firms. There are many different couples bargaining and if I bargain with you, the outcome of that affects the profitability of people bargaining with other people because we’re competing in the market. The basic assumption is what happens if I deviate, if I don’t contract with you, what is my other option? Because that is my bargaining game, what I can gain from bargaining with you is that minus what I would get without bargaining with you.
So I have to find out what I would get without bargaining with you, as that’s what we’re bargaining over! There has been quite a bit of progress on weakening the assumptions in the bargaining model.
In the dynamic models, there’s been a bit of work on learning: the problem is that firms need to be able to learn enough to be able to reach our predicted equilibrium. If I’m at the same point many times, I see what my competitors are doing at that point all the time and I can figure out what to do for me, as the outcome of my actions depends on what my competitors do.
However, if the world is changing all the time, the competitors are also changing all the time, and it’s not clear that I can figure out what my competitors will do in any given situation because that situation doesn’t happen many times. We’ve started to develop learning models that show how firms adjust to new information and I think that will end up being quite helpful. It’s not finished yet– we’re not ready to go out with policy recommendations on Google tomorrow–but we’ve made some progress, there’s a lot of young people in our field doing very good work!
How do you think antitrust authorities can respond to the tech firms such as Google ?
Rather than answer the Google question, I will talk about a case about breaking up Microsoft: they phoned me and asked me to be on the government’s side about how to do that. I thought about for a while and I said “I don’t know the answer and neither do you, and you won’t know the answer in the next three years.” Somebody has to make a decision and in cases like that, people who are closer to the industry are likely to know more than me, if I don’t have the tools to investigate it in the appropriate way.
A similar question came to me when we worked with General Motors. They gave us their data, took our models and started using them for their own stuff. Their head of research Mustafa Mohatarem, very smart guy, he came to me one day and said “What about dynamics, what should we do dynamically? What new products do we produce?”, and I said “At this stage you know better than me!” [laughs] There are things for which we are not really able to help very much yet.
On collusion and cartels, we are much more able–we’ve had quite a bit of theory and a bit of empirical work. The policy issues are often related to the courts who find it very difficult to bring a case unless there’s what’s called a “smoking gun”, for instance, when somebody has taped somebody about doing something like setting prices in a meeting.
In the ADM case [Lysine antitrust case], the reason that they were able to get a conviction is that they actually sent someone in with a tape recorder to one of the meetings. We know of mechanisms that can support cartel-like behaviour, so for an IO economist, a cartel-like behaviour is behaviour that is above our equilibrium prices.
The equilibrium we usually think about is: I set my price to do the best I can given what everybody else is doing,they set their prices the same way, and then nobody has an incentive to deviate.
A cartel will set a price above the equilibrium price, so you have to ask “Why do other people follow?”, because if it’s above the equilibrium price, it means that the other people are not in equilibrium, and they have incentives to price lower and make more money. The reason they don’t is that there is some punishment scheme in the future: it becomes a dynamic problem. They know that if they decrease their price, then tomorrow I will kill them, price much lower and you will have a price war.
We know a lot about how to handle cartels. One of the problems is that the courts have typically needed some smoking gun, they must be able to tell that you have done something illegal.
If this is the case, in the U.S. it’s a “per se violation”, which means that it doesn’t matter if it has harmed society or not, you’re guilty. In Europe, that’s not the case, it’s the ”rule of reason” which says that maybe it was productive for society to have a cartel. Prices are higher so consumers pay more–that’s bad. On the other hand, as prices are higher you may develop new products and now there is an incentive to build new products.
If you look at the dynamics, it’s not always the case that collusion is a bad thing, hence “the rule of reason” rather than “per se violation”. The problem with doing that more generally is that you will get a whole load of court cases, because it’s very hard to distinguish between what’s in society’s interest and what’s just raising price to make more money! [laughs]
We know about analysing the cartel issue, but the issue is that you have to make judgments on whether this tradeoff between more products or more investment versus prices going up is good or bad for consumers. Often it will be good for some consumers but bad for others. It’s a very complicated trade off but we know how to make it; whether the courts or the government listen to us is different! [laughs].
What are your thoughts on current antitrust policy, especially in the light of increasing corporate consolidation?
I know a bit about antitrust policy in Europe, but less than I know about US antitrust policy. The courts have become more and more able to listen to economists and economists’ models. I’m actually quite surprised ! 15 years ago, nobody used our tools! You would go in, one side would argue one thing, the other side would argue the other thing and there would be no model and no data.
Now the courts listen quite frequently and attempt to use our tools. Most of the time they don’t have the time to do the research that we would consider for an article, but they attempt to use our tools and they listen very carefully to the logic of our argument. I imagine it will get even better. I think the story of the courts’ relationship to economics is that it takes them a number of years to catch up to research, maybe ten, but eventually if the research is good and the methodology is correct, it gets into the court system and they listen to us.
There was a court decision on Amazon and Ebay on books, and I was asked to read the decision, I was surprised with the judge–it was just like I was giving a class on the topic [laughs], I must admit I was rather impressed.. More and more lawyers are learning economics and you have to able to explain it in words that aren’t equations. If it’s a good model, you should be able to do that.
Is there any particular methodology that you would like to see being used in antitrust, bearing in mind the trade off between time and precision, with antitrust authorities obviously having limited time to come up with their findings?
I think that’s right. That’s the basic issue for antitrust authorities: they would use our tools all the time if they could. There are shortcuts that have been developed, like something called the Diversion Ratio. If you increase your price a little bit, you get an extra dollar from people who stay, and you lose price minus marginal cost from the people that leave due to the price increase. When you do a merger, these people who leave might actually go to the product of the other merging firm. A big factor, if you do the demand system appropriately, will be the fraction of people who leave and go to the other product. The higher that proportion, the higher the price increase, according to the theory.
They make serious attempts, when they don’t have time to do a whole demand system, just to estimate the fraction of people who would leave and go to the other products. They do a survey asking if we took this product out of the market, what other products would you go to? That’s an attempt to mimic the demand system, and you can do that relatively quickly, especially if it’s a business firm, where there aren’t that many consumers, just other businesses. It makes sense from the point of view of the model, at least for some things. You have to realise that once these guys increase their price, so will everybody else, and that part of it is not taken into account directly. But it’s an attempt to move toward our direction and thinking about it.
There are mergers of content providers with cable television networks, and the authority that had to sign off on these mergers started using very complicated economic models to figure out what would happen after the merger.
The reason I say that is that, while they don’t bring these models to court, it can allow them decide whether or not it’s worthwhile going forward with the case. The fact that the models aren’t brought out in court doesn’t mean they haven’t been used in the process of deciding what goes to court.
You are very well known for your research, but also because of the amount of distinguished students you have taught and mentored (you mentioned Thomas Wollman and Katie Ho in your presentation yesterday). What would you say is the key to getting the best out of your students ? What advice do you have for the young TSE graduates ?
The most important advice for young PhD graduates is that you have to be serious about your research and honest about the results. We can’t do everything but when you’re doing your stuff and there’s something wrong you tell the reader that something is wrong.
I have something like 60 PhD students and they are mostly at the top universities in the US now as professors. It’s just been fun, they have brought up questions that ended up changing my research, they became friends of mine. It’s a big part of my life–you end up working with people who hopefully are good and that you get along with.
It’s a two way game, maybe I give something to the students but the students give quite a bit to me. I think that my biggest job is advising on my students’ theses.
I teach some of the graduate courses and then some of the students are attracted to writing a thesis, and sometimes it’s about helping them with methodology, but a lot of it is shooting down bad ideas. Sometimes you may be starting a thesis and I know it won’t work and I know why it won’t work and I don’t want you to go there [laughs]. Students will typically come in with 4 or 5 ideas before there’s one that makes sense. Sometimes I make mistakes, by the way and my biggest mistake is letting people do things that are too hard, because students come in and they’re going to change the world, and some things just can’t be changed so fast. It’s good that a student comes in and wants to change the world [laughs], but I have a better sense of what can be done than they do.
In truth, this is what has been keeping me. If you ask me why I’m at Harvard, it’s mostly for the students.
With thanks to Bruno Jullien, Mathias Reynaert, Ana Gazmuri and Marc Ivaldi for their question suggestions.
By Fernando Stipanicic, Sai Bravo and Tristan Salmon