In recent years scholars have begun to focus on the consequences of individuals’ exposure to civil war, including its severe health and psychological consequences.
…We exploit the presence of thousands of international soccer (football) players with different exposures to civil conflict in the European professional leagues, and find a strong relationship between the extent of civil conflict in a player’s home country and his propensity to behave violently on the soccer field, as measured by yellow and red cards.
…Reinforcing our claim that we isolate the effect of civil war exposure rather than simple rule-breaking or something else entirely, there is no meaningful correlation between our measure of exposure to civil war and soccer performance measures not closely related to violent conduct.
A new paper by Miguel, Saiegh and Satyanath.
8 Responses
There was a discussion at my old blog in 2008 about a similar paper by the same authors: http://elcriadordegorilas.blogspot.com/2008/04/cmo-va-la-causalidad.html
My question still is: couldn´t it be that referees have a racial bias that is not accounted for by simply controlling by region? A referee biased against black players would only be harsh on, for example, black South Americans but not on white South Americans. And if civil wars were more prevalent on South American countries with larger black populations, that would have an effect on the authors´ findings.
Devil’s advocate time:
The authors find a correlation between civil war exposure and aggression in a common setting–maybe the only forum in the world where things are remotely ceterus parabus for all nations. Pithy as the setting is, it approximates (forgive the pun) a level playing field. The findings and conclusions are fair game for criticism. But if you don’t like the result, how do you explain the correlation? One either needs to dispute the statistics or the causal identification or the interpretation, otherwise it stands.
If you are worried about external validity, your case is stronger. But soccer players are not representative now. At the time they were recruited, most were not so rich. Again, the correlation is a puzzle to be explained.
Also, do the authors reach so far in their generalization as your criticisms suggest?
Let me be honest. I have not read the latest paper. But I read the NY parking one. The data is based on transparency international’s corruption perception index. The casual reader of the article and the popular media coverage would not be able to ascertain that the result is based on questionable data. But they will find the result quite intuitive because it fits their prior. I am glad you are defending your thesis supervisor. I love Ted’s work, for example, I think his work on global warming and conflict is brilliant. But as you can guess I am not impressed by this kind of work–the Freakonomics genre. It does not further knowledge and as Ian suggested it may lead to racial profiling. The correlation suggests violent behavior is innate and no amount of exposure to normal time will ameliorate such behavior. Quite depressing, don’t you think?
To be fair to Ted and Fishman, a former student of mine who did an MBA at Columbia (now a reporter for Washington Post) told me that Fishman was quite surprised by the attention the NY parking paper received. Apparently, they were being playful–kind of like Lant Pritchett/Summers infamous “let them eat pollution” memo.
@Ian
I completely agree with you. This is a genre of paper started by Ted and Fishman that looked at the relationship between unpaid parking tickets by UN diplomats and their country’s corruption ranking. Guess what? They found positive correlation–diplomats from corrupt countries behave corruptly in squeaky clean NY. It is just silly, as the Economists pointed out US diplomats have the highest number of unpaid congestion fee in London. More importantly, the authors used TI’s “perception of corruption” index which is doubly problematic. You cannot pass verdict on a country by the misbehavior of its soccer players and UN Diplomats!
Now Ted is spinning the same twine in case of soccer player. Ted has done valuable work. But this kind of work is tantamount to the sensational headlines you see in National Inquirer.
Errr am I the only one thinking that all this study really tells you is about the propensity of rich footballers from poor was afflicted countries living outside their countries of origin to e more violent …..when they are playing football.
These football players are hardly typical of their poor non football playing compatriots, and their behaviour on the pitch isn’t necessarily a reflection of their behaviour off it. So the only practical implication I can see for the research is that refs should watch players from civil way affected countries a little more closely for fouls – but then that might seem a little too much like racial profiling.
Actually, now that I look at their descriptive statistics, the inclusion of Wales (13 civil wars), Scotland (13 civil wars), and England (13 civil wars) is probably going to heavily bias the results. That’s a lot of country-years of civil war relative to the total. If these are dropped, and the effect they find is actually coming from the style of the English game, which is recognized as being much rougher (Go Chelsea!), then the results are suspect. It would be interesting to see the results if these cases are set to zero and the regression is re-run.
Thoughts?
The data is from the UCDP/Prio database. The US is coded in UCDP/Prio as having a number of civil wars with Al Qaeda from 2001-onward. Scotland and Wales are not included in the UCDP/Prio database, so the authors presumably took the civil war data from the Great Britain country-years, where civil war was coded for the problems with the IRA.
There is a problem in using the UCDP/Prio data in general for this reason, and for my own work I’ve thought to recode these cases to “0”, because independent variables linked to civil war usually are related to domestic causes, and so looking at the US’s GDP to predict it’s “civil war” in Afghanistan doesn’t make sense. The sensible course of action is to code civil wars only for the actual territory where these wars occur, or else you’re testing hypotheses about domestic causes of civil war on territories where they didn’t actually. (I’m currently writing an article using UDCP/Prio) I can’t really fault the authors too much for this however.
The paper is interesting but it lacks a definition of civil war. Given that it finds 13 years of civil war in Scotland and four in the US between 1980 and 2005, a qualifier would be appreciated. Whatever they might have had had in Scotland or the US, I doubt it can be put on the same level as what Colombia or Bosnia Herzegovina underwent during the same period.