Patrick Legros is a Professor of Economics at Université libre de Bruxelles (Belgium) and do research at ECARES. His research interests are in theory of contracts, microeconomics, industrial organization, competition policy and regulation. He has taught courses in intermediate and graduate microeconomics, intermediate and graduate industrial organization and antitrust, graduate courses in contract theory. He has been an editor and he is currently the managing editor of one of the top journals that specialize in industrial organization, the Journal of Industrial Economics.
Why did you decide to become an economist?
It was by accident. I was hesitating between law and economics for my undergraduate studies and I decided to study economics because I had heard that mathematical skills were useful there. I really developed a taste and appreciation for economics during my master’s and my first experiences with research in game theory. For my French doctorate, I got a grant to spend a year in the US and I visited the MEDS department at Northwestern University. I had the chance to be there when researchers like Myerson, Holmström, Milgrom, Matthews, were on the faculty and the economics of information was yet developing. That year reinforced my desire to become a researcher. I ended up doing a PhD at Caltech where I worked on mechanism design and contract theory.
What is the feeling for a well-established economist to be located at the heart of Europe where all the important economic decisions about its future are taken? Is it a source of inspiration or a constraint over the potential research projects he can choose (e.g. EU policy agenda)?
It is an opportunity more than a constraint. We have colleagues at ECARES who do purely theoretical work and have relatively little interest in applying it to policy issues. We have other colleagues who are directly interested in policy topics, both because of the agenda and the availability of data sets. For instance, if you work in industrial organization or competition policy there are clearly a lot of things to do, either by helping the commission or by litigating. Policy questions often raise interesting theoretical questions. Being confronted to policy questions is also a good way to re-evaluate your academic work, but reciprocally academic work can feed the policy debate. So it is really an opportunity, not a constraint. Nobody forces you to do something you do not want.
You are the managing editor of a very top IO journal, the Journal of Industrial Economics. Is it a time consuming activity? What is added value for a researcher to hold such a position?
I used to be an editor and I am now the managing editor. As an editor you have a lot of papers to review and this is demanding but also very valuable because you can feel the pulse of the research in industrial organization. You see all different research agendas, you become aware of datasets. You also get an idea about which agendas are on the decline because they become crowded, as well as some potential interesting topics that can be further examined. Hence, you can see many different trends in research and this is very interesting. As a managing editor you have a different role. You manage, you try to give directions. You may push for special issues (like those we did recently on IPR and innovation), sponsoring conferences, and the like. So, while you are less directly involved in editing papers, it is also interesting because you still get the idea of what is going on in our profession and follow up. It is again a nice opportunity.
You have published in many top economic journals and you have evaluated many contributions of others as an editor that eventually have been published. Since our magazine is mainly addressed to our PhD and master students, I would like to ask you what do you think that are the necessary ingredients of a paper to be publishable to a high quality journal such as the Journal of Industrial Economics?
Whether it is theory or econometrics, what is very important is to have a good and well-developed question. Then, it is important to show that it is relevant, meaning that you know the effects are sufficiently significant that it becomes worth it for colleagues to read your paper. This is a key step. Then, style and writing matter; you have to present your work in a way that is both attractive and accessible.
Most students that publish directly after their dissertation tend to spend a lot of time in the details of their model or technical aspects. This is important, but often people spend too much time on this rather than trying to really make as clear as possible the one point that they want the readers to understand, what the intuition behind the main result is. You have to remember that after one or two years working on a model, you know the model, you know the notation, you know the logic you used to prove the results and you expect that by reading it, the reader will understand what you are doing. This often does not happen because the reader has not spent two years working on your paper. Getting results is only the first step in producing a paper; writing may take more time sometimes than deriving results.
In order to make your paper effective, you may need to take the reader by the hand, especially if there are complex technicalities. You also have to convince him that your question is interesting and novel, so you need to know the literature well and you should be able to differentiate sufficiently your paper from the existing literature. You should also give a good sense of how you got from your assumptions to your results in a way that is relatively intuitive. So writing is what makes a paper effective. My advice is to streamline as much as possible the technical stuff, to make sure that what you put in the paper is really necessary for your point. We (the readers) do not need to see the many intermediate steps that are behind your main message, even if it took you months to develop them, and even if you are quite proud of your technical skills.
For econometric papers, whether you do structural or reduced form analysis, causality is really an issue you need to address.
In terms of what can fit in a journal is journal-specific. On a given question, there are many dimensions and the literature resembles a partition, where some boxes are full and others are empty. For the Journal of Industrial Economics, papers that fill “boxes” are not exactly the type of things we are looking for unless there are significant new insights. Writing a paper to fill a box is useful because it makes a contribution, but you should give a sense that there is something new, hence that you potentially shed new light on a phenomenon. Note that for this, you do not necessarily need new models. You can borrow from an existing model, but simply you say “Look: there is something that people get to understand and which becomes very relevant when I look at it in this way”.
Many people say that the field of theoretical industrial organization due to its numerous past contributions is not recommended for young researchers, since it will be very difficult to produce good quality work on new ideas and phenomena. Do you agree with this view?
Not completely. What is true is that if you have a new model that derives a new theorem by changing assumptions in a given existing model, you have little chances to publish it in a good quality journal. This type of theory is certainly past due. But, there are many very important questions in industrial organization that have not been treated so far. We do not understand so much about multiproduct firm competition. We certainly have to learn many things about the interaction between the theory of organizations and industrial economics, topics that I am particularly interested in. We have there a lot of interesting topics, both from theoretical and empirical perspectives. Another important question is how to organize a market for property rights. We have very little understanding on that, and the stakes for policy and growth loom large. So, I think that there are plenty of difficult theoretical problems. If you can make a contribution on such problems, you can publish.
Now, if you want to have a more positive type of paper, it is not sufficient to say that with these assumptions you get these results and this is what will happen. People then will say, “ok, tell me why this is more likely to happen than something else”. And for that, often the most effective papers are those that will use some type of empirical evidence or econometric analysis. So, I would say that if you want to do pure theory you better have something on a topic that is not “well-travelled” yet, or if it is you better have results that will shake received wisdom.
Theory is important even if you do empirical work because it tells you how to look at your data, or which data to look for. Empirical work is about uncovering causality relationships and how can this be done without a theoretical construct? What may be puzzling empirical findings within a given theoretical framework may be easily explained within another theory. Theorists provide new ways to look at the world and at the data, and this role must not disappear, especially when there are policy considerations at play, as in industrial organization.
As you already said, you are interested in the topic of organizations from industrial economics point of view. Your approach deviates from the standard neoclassical cost-minimizing approach. Could you tell us something about it? What are the main insights on how organizations affect the behavior of firms in the markets and what is your future agenda on this topic?
When you do economic theory, you try to find the most convenient model to get to the results. One question is why firms are so simple, so “black-box”, in industrial organization? You get a lot of results from that, a lot of insights. So, why should we bother to deviate from this approach and deal with a richer theory of organization for firms? It is because we get results that would be very difficult to get with a neoclassical firm – or you need very complicated stories in terms of market power and information structure to get these results. Let me give a few illustrations, based on joint research with Andrew Newman, of what organization theory can bring to industrial organization
One very well documented empirical fact is that you have heterogeneity of firms in a market economy in terms of performance. Some firms are very productive while others are not very productive, even if they have access to similar technologies. In a competitive market when firms are not productive they should disappear. So the question is: why do they survive? Using organization theory can help. Think of the standard conflict of interest within an organization between getting monetary benefits that are correlated to variables that you observe and the variables you cannot contract on like private benefits and costs. Organizations where you have low productivity may also have high private benefits and those with high productivity low private benefits. While performance is different, the welfare of the decision makers may be the same, explaining survival of the low performing firms. Heterogeneity creates a loss in output in the economy and the question is how to correct for that, which brings us to the policy dimension of the project, in particular the role of corporate governance, or the control of integration in competition policy.
Heterogeneity and the resulting inefficiencies are affected by demand. It is in fact when elasticity of demand is high that you have the most inefficiency in a competitive setting: this is because the loss in output due to low productivity is magnified. This is in sharp contrast with the usual finding in monopoly theory that the more elastic the demand, the lower the market power of the firm is.
This approach generates a very simple relationship between the level of some market variables like prices and the types of organizations you are likely to observe. When prices are high, monetary benefits are high and therefore the tradeoff between monetary benefits and private benefits changes in favor of increasing productivity. This may lead to more integration in some models. Hence, you have a way to pin down why firms are organized as a function of the level of demand.
This approach may also explain why productivity shocks that will lead to an aggregate productivity gain are not realized in the economy. They are not realized because the productivity shock changes the organizational benefits between integration and non-integration. Some firms may decide to integrate and because they integrate they supply more to the market. This has an adverse effect on the price, which has an adverse effect on firms that are not subject to the shock and may decide to go to less productive form of organizations. So, on average you may have actually a strong dampening effect of the shocks.
The organizational approach leads very naturally to a reverse causality in market settings when you think of the role of integration for foreclosure or predation. The usual idea is that integration may enhance market power and help foreclose the market, then generating higher prices once predation or foreclosure succeeds. So, the causality goes from the use of integration (organization choice) to price levels. But given what I just said before, high prices are conductive to incentives to integrate even in the absence of market power. So, now the causality may go in the other direction, from price levels to integration. The question is how to disentangle the causal effect of market power, organization and prices from the other effect which is when prices are high, demand in high and you have already incentives to integrate without the desire to foreclosure. This opens a host of questions that are important from a policy point of view.
I like this project quite a lot! Actually, it is not so recent. It is based on some older work I have done on matching with Andy, and the current agenda is developed with Thomas Gall and other collaborators. With Andy we have a series of papers on matching, in which surplus of matching is not perfectly transferable, a setting that arose from our research on organizations. In many models of organizations you get a production function where there is no perfect transferability. If I give you a higher wage, you may increase total output and hence my revenue by more or less than the wage. One dollar that I give you may cost me more than one dollar because of moral hazard or other informational asymmetries. So, when you think of organizations, you are forced to look at non-transferabilities. And then when you think about how the economy will be organized you have to think of matching with non-transferability.
The project on education takes the view that the returns from education are realized later in organizations where you have non-transferabilities, and therefore the marginal returns from education may be distorted. For example, when you have your high school diploma, you may go after to the university. What happens at the university will depend on the quality of your teachers, the quality of the infrastructure of your institution, the quality of your peers. All these enhance your own benefit, but, it is not transferable, there is often a missing pricing mechanism. We show that this downstream matching market (university) has a big effect on incentives to match in the upstream market (high school). This may lead in particular to excessive segregation. If it is very difficult to change non-transferabilities, and policies that are targeted to the downstream market may actually help change the matching earlier at the high school level.
One beautiful example of that is the Texas 10% system that was put in place after white students went to court to complain about affirmative action policies by Texas University being discriminatory. So, Texas decided to put a policy where the top 10% of every high school had priority access to the university. They were hoping by doing so to replicate the affirmative action policy, and to have a bigger mix of social classes at the university. It did not happen. It did not bring more Afro-Americans or Hispanic to Texas University. But there was a big effect at the high school level: students who wanted to go to Texas U. realized that it is easier to be top at a not so good school than to be top in the best schools. With work with Thomas Gall and Fernanda Estevan, we show that there was indeed more integration at the high school level following this policy. That is a very nice example of why changing the way you can match in the downstream market (university) may affect matching in the upstream market (high school). You may not see what you would like to see happening downstream but you see many things happening upstream, leading eventually to delayed benefits downstream.
Education policy is indeed complicated. Policy makers tend to focus on one element of the vertical chain (primary- secondary-tertiary-labor). If there is a problem at the high school level, let’s correct the problem there. You could for instance give access to good high schools to unprivileged students. But do these students want to use this option? Do they want to travel half an hour more to go to a better school if they expect that they will get a low return in the downstream market? So, giving opportunities is good only if people take them. And they will take them if they expect that in the future this will be valuable. Sometimes, targeting the policy on downstream segments may be more effective than acting directly on the upstream segment, or at least can be a significant complement. This is true for high schools and universities, but also more generally for education and the labor market.