Monday, September 15, 2008
Whites, Blacks and Altruism in UCT Students
An Economics Research Southern Africa (ERSA) working paper by Wilhelm Gerhard van der Merwe and Justine Burns uses the dictator game to test altruistic motives in the students. As I have already sent the authors an email with some of my concerns, I am not worried at all about discussing some points about the paper, in fact I hope that more discussion about the paper will improve it prior to its (hopeful) publication. In the spirit of full disclosure: Justine Burns was my 'supervisor' for my MComm in Economics, her husband Malcolm Keswell was my 'boss' when I worked for SALDRU; they are now my 'friends'.
Anyway, back to altruism in UCT students. The authors are very careful in annunciating that the sample that they study should not be seen as generalizable. This is meritorious indeed as far too many people don't make this kind of point sufficiently strongly in their research (including Ernst Fehr on whose recent work I commented a couple of days ago).
The construction of their experiment is as follows. It uses a classic dictator game with samples of students who are randomly paired. They run a control dictator game and then introduce a treatment that reveals the surname of their opponent in order to convey their racial/ethnic group. Their results indicate that the introduction of this information, in general, increases, i.e. simply knowing something about the partner increases the payoffs in general. White proposers make significantly higher offers than Black proposers. White proposers also seem to favour insiders with the median offer in a white-white pairing being double the offer in a white-black pairing. There does not seem to be insider bias for black individuals.
I have one or two concerns, which, as I said, I mentioned to the authors. The main one is about the inclusion of 'altruism' in their title and in the text. I don't believe that the patterns that are observed should be discussed as necessarily altruistic, but could, instead, be indicative of a host of differing theories see Bardsley, 2008; List, 2007 and Levitt and List, 2007 on this. The authors don't place too great a weight on the term altruism, so I don't think that it will be a problem for them to negotiate this problem.
Otherwise, I don't have any major methodological issues with the study and I believe that it is informative and should be replicated, eventually with a more representative sample of the population (this is something I'm particularly interested in seeing, or even doing at a later date). There are one or two issues in how things are presented, but they will be ironed out as the paper goes through iterations of review I am sure.
Please read the paper it is quite short and I think that Justine and Wilhelm would appreciate any feedback I could give them from anyone who reads this blog. It is available free of charge from the link in the first line of the post.
Any idea when the experiment actually happened? Hugh and I participated in a very similar experiment in our second year, ie, 2003. We were both dictators and, naturally, we were both bastards and kept all the extra money... :-)
An additional reason why the results may not be generalizable: http://www.gnu.org/philosophy/economics_frank/frank.html
Gosh Mike, your adherence homo economicus warms my heart. Um... maybe not. Yes, Robert Frank does have good insights, as always. I still think it's a worthwhile arena for research though, don't you?
:-)...
Of course I do - my point was that the kind of research I pointed to was another reason to worry about generalizablity, not a reason for not actually conducting the research.
Yes, my point though was that if you manage to get a nationally representative sample, which some experimentalists have managed to do, then you should be able to make more general statements about the tendency of individuals within that economy. If we are then able to observe such nationally representative samples of experiments from an internationally representative sample of countries, then we would have something that better approximates a 'general' idea.
I think that the above is a bit of a pipe dream, but I'd rather dream it than not even consider its possibility.
Sure, that would be awesome. University students - especially when there are economists involved - are generally far from representative. (Related: http://bps-research-digest.blogspot.com/2008/01/just-how-representative-are-people-who.html).
I was mostly worried about the experiment Hugh and I were involved in. As far as I remember, subject recruiting was done mainly among economics students. And if it's true that economics training significantly affects behavior in games, then we have serious reasons to worry about representativity.
Of course, all these problems can be resolved by getting nationally representative samples. Good luck finding funding to do that though...