As you will likely know, the work of a professor goes far beyond the walls of the lecture hall. While relatively unknown to most of her students, Tilburg University is home to a very passionate department of researchers, each active on the forefronts of their particular field. In this article, we will highlight three of them. First, Tobias Klein will tell us about trying to answer economic questions with data by bringing methods together with questions. Next, Jaap Abbring will tell us about the dynamics of oligopolistic markets and last but definitely not least, Hein Fleuren will inform us about the way he applies operations research and the complications of actually implementing methods and ideas.
After studying in Mannheim, Berkeley and London, Tobias Klein was very happy to land a professorship in Tilburg. “Tilburg University is a very special place,” he says, “Our university is doing extremely well and many people in the region do not realize this; it is a very attractive place to work for. What makes it so special is that it is a research university; most people here do both research and teaching, not just one of them. The idea is that the teaching becomes better if it is done by the people who actually also do research themselves. And at the same time, the research becomes better as a result of your teaching. We are all scholars, you know.”
When asked about what he is currently working on, Tobias notes that all of his current projects involve data in some way. Like progress in society, Tobias believes that progress in research comes from many small steps and many small questions, rather than one big step. Tobias does like theory, but feels that theory without any link to actual data is empty. In his view, theory should be tailored to the question you want to answer. The way it usually works for him is that he or one of the various people he collaborates with comes across some interesting question with an interesting dataset and, perhaps most importantly, something new to measure. For example, he is currently collaborating with Blue Mango – a firm in Eindhoven that Tobias acutely nicknames ‘Google of the Netherlands’ – on a project for the Staatsloterij.
You cannot understand a teacher without also understanding his research
The fact that they have access to a dataset that contains information on Staatsloterij’s radio and television advertisements, as well as online sales data, is what makes this project so special. Now, as you might imagine, these data can be used to measure the short term effects of the advertisements. But the power of data does not stop at that point: they are also used to build a model to think about the long term effects of the advertisements. In particular, the effect of the timing of advertisements is interesting to consider: should Staatsloterij advertise close to the deadline, which is when people tend to buy their lottery tickets, or should they spread advertisements instead, in order to reach more people? It turns out that the former option is in fact more advantageous than the latter.
The paper that Tobias is most proud of has recently been accepted for publication by one of the top five journals in the field. It also involves a great dataset and an interesting question. In this paper, reputation systems are investigated, which have become more and more common on the internet—think of Iens, Tripadvisor, Airbnb and Uber, where people rate one another. The question Tobias has tried to answer in the paper is what these ratings actually do. It turns out that more transparency leads to better behavior of the sellers, which is what a reputation system is all about. Normally, the effects of such a reputation system would be very hard to measure, but due to a clever – Tobias calls it lucky – trick, they were able to measure this nonetheless, by applying various econometric techniques. “I am very proud of this paper because it has a positive message and because we captured something that is normally hard to measure. And on the way, we did econometrics”, he says.
At some point in his career, Tobias would like to do some work into actually measuring health. In particular, the effects of investment in one’s health and the long run effects of behavior that one has early in his/her life. Due to the heavily endogenous nature of treatments, he is not certain whether this question can be answered, but at the same time, he is intrigued by it.
By now, you might be wondering whether this data-enthusiast is involved in the brand-new data science programs. Indeed, he tells us, he will be teaching a course in the bachelor: Data Science Research Methods. This course is intended to teach students to answer questions by applying the wide range of methods to find patterns in data that they will have learned by then. Given Tobias’ own research interests, it seems that this course would fit him perfectly.
We would like to thank Tobias for his time and wish him the best of luck in using data to his advantage to answer interesting questions.
“So then he said ‘sure, you should come visit’ ”, Jaap laughs, talking about none other than Nobel Prize winner James Heckman, who had invited him to Chicago as a visiting assistant professor. After he had returned from Chicago – a great school for economists, he notes – he spent most of his time in Amsterdam and Rotterdam, among others serving as director of education of the Tinbergen Institute for several years. When this came to an end, Jaap realized that his research field had shifted from labor economics – something in which they specialize in Amsterdam – toward industrial organization and insurance markets. This caused him to move his work to Tilburg, which has proven to be an excellent choice.
Currently, Jaap is working on a diverse selection of projects, most of which involve dynamic choice models and dynamic games in some way. For example, in one of his current working papers, he considers a dynamic model of an oligopoly in which firms may enter or exit. These kinds of models may be used in practice to evaluate the existing competition policy, or to determine how competitive the market currently is. However, making predictions based on these models has proven to be rather difficult, since they commonly do not have a unique equilibrium and the existing equilibria are hard to find.
Sometimes simpler models are just more useful
In Jaap’s line of research, he is trying to reduce this difficulty by relieving the models that exist in the literature – which generally incorporate many complexities a market might have – of some of their unnecessary complications. This could allow for the resulting models to be calculated much more quickly, while not significantly compromising the integrity of the solution. Then, because such a model can be solved quickly, it could easily be evaluated many times, to be able to find parameter values for which the model’s predictions match available data. As a direct consequence, these slightly simpler models might be much more useful in practice than the existing, very complex ones. According to Jaap’s research, this approach has appeared to be working well.
An example of such oligopolies would be local cinema markets in the United States. “This sounds very specific, but has been widely studied,” Jaap says, “even in the 1950’s, when television started coming around and people expected cinemas to disappear as a consequence of the competition the new medium would offer.” Because there are relatively few players on each local cinema market – pretty much everyone sticks to the town he lives in to visit a cinema – it can be modeled as an oligopoly. Through deriving the theoretical properties of this model, coming up with the algorithms to numerically find the equilibria and applying econometric methods to find estimators for crucial parameters, Jaap and his collaborators are able to estimate the effect of a new player – say, Netflix – entering in this market. “So, analogously to that,” he tells us, “one could model any market in which one has some form of local competition of retail businesses that serve local customers and analyze them with these methods. There is some methodology there, but also some applications. It is just fun to do some applied work.”
This is also the reason that he is most proud of this particular work. They have done some very serious theoretical work while keeping it simple enough to be actually put to practice. But that is not all: they have also formulated their own practical methods. This breadth is what he likes. “It is just the most satisfying to me when you can see how theory becomes useful,” he concludes.
We would like to thank Jaap for his time and wish him the best of luck in his continued research.
As opposed to Tobias and Jaap, who are working in the field of econometrics, Hein specializes in operations research (OR). His research is commonly practically oriented: more often than not, his research is directly tied to some problem an organization may have. It is on the edge of research and practice where Hein enjoys himself best. “Usually, when research is being done into a certain problem, especially when it is fundamental research, some aspects of the problem that are particularly complex are left out, which is fine: in fundamental research one really tries to get to the essence of the problem. However, when I look at some situation in practice, I often find that these fundamental results cannot be used. So then, what my research is all about is trying to find ways to be able to apply those fundamental results anyway.”
Typically, Hein works with large organizations such as TNT Express or the United Nations World Food Programme. These organizations offer problems that are so complex that through Hein’s research, enormous differences can be made. One of the things that Hein has learned from his projects at large corporations is that unlike some students like to believe, some of the major factors that have to be taken into account are people’s unwillingness to change and simply their lack of understanding of the benefits of optimization. Hein tells us that at the “wonderful playground of optimization problems” that is TNT, they tried to solve those issues by setting up the GO (Global Optimization) Academy: a two-year series of courses for the TNT employees, designed to teach them the basics of optimization in a non-mathematical way. “What we did not expect, and was really great to see, was that through only a little explanation, people started recognizing optimization problems by themselves. All of a sudden, we were no longer the ones trying to convince them but instead, the employees themselves came to us to ask for help.”
Only after we have calculated the optimal solution, the big job actually starts
Next to such projects, Hein also supervises master’s theses and PhD candidates. He often does this in cooperation with a large organization, where they will set up a series of multiple theses for several students. Currently, this organization is the World Food Programme. One of the theses that has been written there, the thesis by Koen Peters, is one that Hein is especially proud of. The WFP provides food to people in areas struck by disaster. The required food comes from all over the world and has to be transported to areas that are hard to reach. And, to complicate matters even further, just serving the people in need with only Brussels sprouts for several months is not going to take care of their nutritious needs too well.
The way this used to be approached by the WFP is to have two separate divisions: one that optimizes the baskets of food to be as nutritious as possible, and one division that optimized the transportation of those baskets. Now, in Koen’s thesis, both divisions are actually joined and the transportation of certain nutrients is optimized, rather than complete baskets. This change by itself has allowed for a million more people in Syria to be fed on the same budget as before, which is of course a great achievement.
When asked about a field of research Hein would like to get into at some point in the future, he tells us about two fields. Firstly, the issue of acceptation of operations research solutions and the psychology involved in this is very interesting to him. For example, we might calculate the optimal solution to some depot-allocation problem and think we are done. In practice, however, this is where the big job actually starts. Depots might have to be built or improved, people might lose their jobs due to ‘our’ cost reduction; these are issues that have to be dealt with. In short, it boils down to the question “why are the beautiful things that we compute implemented so little?” Secondly, Hein is very interested in the humanitarian applications of OR and the complications those bring to the table, which we do not commonly see in ‘western optimization’. For example, the simple lack of a distance matrix we all know and love is only the tip of the iceberg.
We would like to thank Hein for his time and hope he is able to find out how we can get people to implement the beautiful things we have computed.
Text by: Pepijn Wissing