One of the first field experimental projects I worked on (around 2005) sought to figure out why ethnic heterogeneity is so often associated with lower levels of public goods production. We tried to answer this question by looking at how people respond to social dilemmas differently when they are playing with people from their "own" group and people from other groups. Based on play in across a whole series of experimental games we implemented in Kampala, we argued that behavior that might look like it reflects deep seated in group preferences, really reflected a set of social norms. The key evidence was that people were not particularly likely to favor coethnics when playing they were acting anonymously; but they did start favoring coethnics when their actions were observable to their coethnics. The book does a lot more than the article, exploring mechanisms in more detail and trying to assess external validity in various ways.
Looking back I still like the analytical characterization of the the mechanisms through which ethnicity might operate. I also like the care we took not to assume the existence of groups or objective ethnic categories, relying instead on various notions of subjective demography. I am less happy now with the use of games, finding them hard to interpret and very sensitive to details of implementation. I also worry that some of the results come from the analysis of a set of subgroups that we did not specify in advance. I think our results might have looked different if we had written up a preanalysis plan in advance. Finally I regret that the piece is so often cited to support the argument that diversity creates collective action dilemmas. Our interest though was always in figuring out why this might be the case and not whether it is the case; in fact we found that overall ethnicity was not a very powerful predictor of behavior in the games we examined.