Romesh Vaitilingam interviews Martin Kocher for Vox

May 2011

Transcription of a VoxEU audio interview [http://www.voxeu.org/index.php?q=node/6466] 

Romesh Vaitilingam:  Welcome to Vox Talks, a series of audio interviews with leading economists from around the world. My name is Romesh Vaitilingam, and today's interview is with Professor Martin Kocher of the University of Munich. Martin and I met at the European Economic Association's annual meetings in Glasgow in August, 2010, where he presented a paper about his experimental research with children and adolescents. The research looks at the attitudes of these young people to such issues as risk aversion, time preference, and social behavior.

Martin Kocher:  We selected a sample of schools and a large sample of school children that are aged from seven to 18 years old, and we tracked or we compared the behavior that is economically relevant over time actually. So we looked at concepts that are important for economics like risk preferences, ambiguity preferences, social preferences, preference towards cooperation, lots of other things, trust, for instance. We conducted experiments with them and conducted them in a way that they're really comparable to real experiments in the laboratory because we incentivized them so they earn money.

Romesh Vaitilingam:  So you're doing similar kinds of experiments as you do with students and with grown up people?

Martin Kocher:  Yep.

Romesh Vaitilingam:  Give us a feel for the kinds of games that the children would play.

Martin Kocher:  Some of them are very easy like individual decision-making tasks for risk preferences, so it just takes decisions between two outcomes and lotteries, and it's very easy to calculate real risk preferences from this choice list of tasks, actually. Some of them are real interaction experiments. So for instance for social preferences, we measure allocation preferences, we measure some ultimatum games—or similar to ultimatum game behavior--trust games, public‑goods games, so very simple economic games where there is a person who has a choice to make and another person that is affected by the choice and can reciprocate or not, so very very simple interactions, that we also use, of course, very often the laboratory. We know very well how grown ups behave on average in these set ups.

Romesh Vaitilingam:  Do children behave in a similar way to adults?

Martin Kocher:  Not really. It really depends a lot on what you analyze. Regarding these individual decision-making tasks, we were quite surprised that there is not a huge difference. So, in risk preference that's a nice example, for instance, we were expecting that there is a huge development. Actually, there is only a small development. So, the way we rank kids, 7, 8 to 10, 12 years they are a bit less risk averse than we observed for older cohorts. But starting from age 12 on average they are almost the same level of risk aversion than grown‑ups, adults, whatever. Of course, there's huge individual originality as always in these data, but there's not as strong development. With regard to social preferences, sharing behavior, there is indeed some kind of development; it seems to be the case that when kids grow up that they become more social. It was a bit surprising for an economist. Our very layman like, psychological thinking and reasoning in the beginning was they grow up, they learn it's difficult to live in this world. They should trust too much so be more careful. Don't give away money. It's the other way around, actually. So people become more social when they become older.

It's very important to also note that this is always within anonymous interactions. So they interact with an anonymous partner. It's not the kind of in-family interaction where you would probably suppose that kids at a very young age are quite social, although you never know. If you have kids, you know that they don't like to share their toys actually.

Romesh Vaitilingam:  Well, I guess I would have guessed that from my experience with my own children and other people's children that they are more likely to be more selfish than adults who have learned social behavior. But I also would have guessed that they would have more risky behavior because they would be less aware of the great risks that there are in life. I would have guessed they would have had a time preference at one things now rather than later because they haven't learned about displacing their activities now for greater benefits in the future.

Martin Kocher:  So, for time preferences we were very surprised that we didn't find it. Actually, we expected the same. We thought that the immediate gratification was more important for younger kids than for older ones. It doesn't seem to be the case on average. Development is very small. It's difficult to say. There are one or two studies that show a bit of a different result. It depends also very strongly on how to incentivize that. Actually, what we do is really incentivize it by money which makes it more serious for the kids, and probably that's the reason why even the young kids think a lot about the decisions because of it. Our experience shows that incentivizing kids with money makes it a real decision for them. If they only get toys or whatever, they get sweets or something like that, it's a bit different for them.

Of course, if you work with younger children, it's almost inevitable to that, but we had positive experiences with using money from an age starting at seven or eight years. It's OK and, of course, we have to add that the kids agreed with it and all the parents agreed, all the principals agreed, teachers agreed. Of course, we had all the permissions. Usually if you run economic experiments, there is no ethical issues involved, almost no ethical issues involved in the laboratory. Once you start running experiments with kids, you have to be taking care of that, of course.

Romesh Vaitilingam:  Are you able to look at differences between kids in terms of their degree of advantage and how well educated their parents are, how much money their parents have, what kind of social class they might belong to?

Martin Kocher:  We have some background data. Surprisingly, many of them don't explain much of the variation in the data. Sometimes, something shows up. What you see, for instance, that's nice as kind of a robustness check; we have some data on how good they are in school actually. And what we see is that those that are not very good at school, supposedly those kids that develop a bit slower than others, they show these slight changes in behavior a little bit later than the others. So, the cognitive development is, to a certain extent, mirrored in the economic behavior. And that is kind of a robustness test that everything is fine.

But all this background behavior, background data on immigration background of the family. There is the religious background of the family. We have data on the profession of the parents, do they grow up alone, whether they have siblings. Sometimes they show up significant, but there's no clear overall pattern, probably because, even though we have a very large sample, it's not large enough to pick up the supposedly small effects that these background variables play.

Romesh Vaitilingam:  Can you give us a feel for the range of differences of behavior in terms of risk aversion, for example? Are some kids very, very risk averse and some kids very keen on taking a risk?

Martin Kocher:  So, what we see, and that's also quite in line with expectations, is that there's a huge variation when they're young, and it becomes small and becomes closer to the variation that we also observe for adults when they grow up, when they become 15 or 16 years old, actually. But for, if you take a look at eight year old kids, there are indeed a few people, or a few kids there that show really risk-loving behaviour. We're pretty sure that they understand what they're doing, because we train them. We exclude observations for kids who were not able to answer some control questions, so it's probably their preferences that they really have.

So they, probably they are, what you probably mentioned before that, they are a bit over-optimistic. They think that good outcome will come true anyway. But you always have, of course…Fairy tales always work that way, so they have more positive feel for life, probably, than older kids.

Romesh Vaitilingam:  Do you think that the behaviors, presuming once you get in to teenage years, that's going to carry over into future adult behavior?

Martin Kocher:  Yeah, for most of these aspects that we studied, there is not a huge development from an age of 12 on, actually. One example where this is the case, but it is still quite a huge development from 12 to age of 18 to 20, is trusting behavior. So, what we observed is actually that kids don't trust at all, almost nothing, and it's actually borne out by the behavior of the trustees, so, they are not very trustworthy either. So that the amount of trust they have, and the amount of trustworthiness increases almost linearly up to the age of 18 or 20. So, this is the only example where it really has a huge development, still, from an age of 12 to 18.

Romesh Vaitilingam:  Presumably, your kind of experiments with these children and assessing their preferences and their likely behavior, it could be an incredibly valuable policy tool. I mean, there's lots of talk, certainly in the UK, which I know of, and in other countries, about interventions that you can make during childhood to have an effect on later life outcomes, the way people might behave in terms of drug taking, or excessive drinking, or proneness to obesity or these kind of health behaviors.

Martin Kocher:  No, absolutely. And I think that's the most important implication. I mean, it's interesting, of course, from a basic research point of view to know about these preferences of kids, but in the end, everything comes down to policy interventions. And what we observe, and that's quite interesting, is that already, at a quite young age, we are able to predict real world behavior by this very easy experimental test. So, risk preferences, measured in the laboratory or with these experiments predict behavior regarding drinking, smoking, sometimes sports choice. Or time preferences predict, to a certain extent, the BMI, the body mass index or the probability where the people learn an instrument and practice long hours.

These predictors are surprisingly good. We didn't expect it to be that good at that young age. We thought, that's, had a lot of noise going on, but we are able to predict quite well already at that age, from this data.

And of course, if you know that, you can decide some policy interventions that will help you. And, for instance, part of the things that was not involved in this study, but it was with the same subject pool. Mathias Sutter of the University of Innsbruck did it with a few coauthors. They analyzed selection into competition. And there's a long discussion that females are less eager to select themselves into competitive environment, and it's very important to know at which age this kicks in—that stronger competitive behavior of boys in comparison to girls. It seems to be the case that it is already there at an age of three years old, which was surprising to see.

It seems to be there at a very early age already, becomes a bit stronger later on, but it's there very early already. And if you know that, you know what kind of interventions might help, and which kind of interventions might not help, actually.

Romesh Vaitilingam:  Final question, Martin, you mentioned the ethics of this research, which was an unusual challenge for an economist to be dealing with. But I think it sounds like, with the positive interventions, just with the subject group that you've used, you might be in a position where you could advise certain ones to pursue, or their parents or their teachers, to intervene in some way in their lives.

Martin Kocher:  No, absolutely. This has to be tested, of course. These are the first results. There are some other studies, but this is a very new field. Actually, psychologists have been doing it for a long time, development psychologists, but they use different methods, in which we don't believe totally sometimes. Sometimes we believe in them as economists. I think there is something out there, and it's important to check the robustness of these results. But then, of course, we are hopefully in a position to see some risks already at a very early age, based on very easy tests that we can administer to the kids.

As you know, it's an experiment. It's just filling out a page and checking his preference, for instance time preferences. In the end, one has to bear in mind, of course, these are always probability statements. We know that the probability increases that somebody, for instance, later on becomes a drug addict. But, I mean, if you can explain some small variations it might be helpful already to notice variation.

Romesh Vaitilingam:  Martin Kocher, thank you very much.

Martin Kocher:  Thanks a lot.