Tuesday, May 06, 2008
Herrmann et al. Cooperation and Antisocial Punishment Today's comment (probably the last one for a while depending on time and studying) is about the (2008) paper by Benedikt Herrmann, Christian Thöni and Simon Gächter in Science.
The main thrust of the paper is to do with what the authors label as 'antisocial punishment', i.e. punishing those people who are cooperating in a public goods game with you. This is contrary to most of the literature on the subject in which punishment normally ensures that players in public goods games end up at a Pareto superior solution of cooperate (most of the time).
The authors undertook 16 sets of experiments in 15 different countries (you can see the countries in the graphs). The authors then look at some specific reasons as to why the results could indicate such strange results for the different groups in the different countries. They argue that social norms (social preferences?) could exist which would promote specific patterns of behaviour rather than others. Thus they try to use an instrument for civic cooperation in society, which might be indicative of the necessity to punish free riders in order to display that such behaviour is unacceptable. However the converse, that low civic cooperation necessarily leads to antisocial punishment is not necessarily true, but they try to test the thesis regardless. To test this hypothesis they use the World Values Survey, The second hypothesis is that strong rule of law could lead to a situation in which personal revenge punishing, or antisocial punishment would be unacceptable – 'effective, fair, impartial' legal institutions could erase the need to personal revenge punishment (this is supported by an interesting, if slightly more anecdotal article by Jared Diamond in the recent New Yorker). To test this hypothesis the authors use an law indicator provided by the World Bank. Norms of civic cooperation were important in determining punishment in ensuring cooperation, but rule of law was not significant. With respect to antisocial punishment, better norms of civic cooperation and improved rule of law would decrease the likelihood of antisocial punishment. Stating the converse – countries with poor rule of law and low norms of civic cooperation were more likely to have antisocial punishment. Anyway, taking note of these conclusions, I think it is quite cool to have a look at the graphs. The first graph is the graph indicating whether cooperation occurs when the punishment option is not allowed, i.e. this forms the control group against which the punishment group is compared.
As is normal, we see that the groups start off cooperating, but then basically move downwards. However, what is astonishing is the disparate starting positions for the different cities and the different end positions. Normally, one would expect 'first world countries' (high civic cooperation, high legal institution quality) to have decent cooperation in general. This is not the case, for which Melbourne, Australia is the most interesting exemplar. Istanbul, Riyadh and Athens are however very near the bottom (and stay there in the second graph). Copenhagen has very high norms of cooperation, with cooperation being high in this graph and also in the second. In the second graph (where the experiment is run with the option to punish) we observe a massive change in behaviour for certain developed countries, again Melbourne is interesting as their behaviour has come into line with most of the other developed countries, but is not as high as Copenhagen or Boston.
Something which is fascinating to see when assessing the graphs side by side is that the behaviour in Minsk, Athens and Riyadh (and to a lesser extent Muscat) is almost exactly the same in the control as it is in the punishment treatment. This is quite astounding when considering the evidence that is normally offered in this literature. I think that this warrants substantial additional research. What makes the second graph particularly interesting is the variation in the final outcomes. In the first graph there was some variation (difference max-min = 7.6 units), but the shape of the curves was fairly similar for each of the societies. In the second graph there is dramatic variation in the outcomes for the different societies (difference max – min = 12.3). These are very rough, I admit - I don't have the data set, but I encourage you to read the paper for some more direct and applicable analysis of the dataset. I would also intuit that if you looked at the averages for the two groups Chengdu up to Boston against Minsk down to Athens you would get a significant difference between these groups, whether these groups deserve aggregation is something worth checking in terms of the civic cooperation and the legal institution index.
One additional comment I'd like to make is that I think that this kind of research would be especially worthwhile in both South America and Africa. The sample at hand covers groups from the US, Europe, Asia and Austalasia. It would be cool to see a couple more from the Australasia/Oceania region too, but I think there should be some experiments done in South America and Africa, as already stated. These would face additional and particular challenges, but are warranted nevertheless.