Suicide & statistics
Following a series of suicides in the French telephone company France Telecom, there have been attempts at explaining what was perceived as an unusually high rate of suicide within a company: 24 suicides over 18 months, i.e. about 15 per year for a company of 100,000 employees. Without going into these analyses, I want to focus on the recent outcry provoked by the “analysis” of a French statistician, René Padieu, that follows simply reproduces an earlier remark by the New York TImes:
“In statistical terms, the 24 suicides at France Télécom since February 2008 — including eight since the beginning of summer, with the latest confirmed on Monday — are not extraordinary for a company employing 102,0000 people in France.” New York Times, Sept. 29, 2009.
Several other statisticians have already commented on this very crude extrapolation, including Avner Bar-Hen who is the current president of the French Statistical Society (FSDS). (See also the Désintox column in Libération.) It indeed does not make sense to compare the suicide rate of a global population, whose causes are immensely diverse, to a wave of job related suicides within a single company. Especially when considering that a large majority of the suicides was by men over 50… What’s most definitely annoying is that biased use of statistical reasonning from someone signing his tribune as affiliated to the French Statistics Institute (INSEE) and the French Statistical Society (FSDS)—as the president of the deontology commission!. And thus mistakenly involves both institutions...
Ps-There is actually a column in Le Monde of October 26, with the same title as this post and whose argument completely makes my point: by accepting the statistician’s arguments of René Padieu, the columnist argues that statistics has clear limitations for this type of debate and he ends up with Disraeli’s quote about lies and statistics… The disservice to the profession due to such an un-professional attitude is therefore terrible!
October 27, 2009 at 8:34 am
I have not read the initial tribune, so I won’t comment on that. And you know me enough to realise I would not defend (or contradict!) somebody’s argument just because he is from INSEE…
The only thing I am saying is that you can’t discard model 1 (Tadieu comparing rates), without discarding model 0 (trade unions stating an absolute number, without any form of comparison)… Hopefully, someone will come up with a reasonable model 2 some day.
October 24, 2009 at 12:21 pm
I am sorry too (!) because you seem somehow to come to the defense of a INSEE member instead of seeing the overall impact of a radical judgment made by someone whose affiliation to both institutions (INSEE and SFDS) was on top of his tribune (and later reported as such by Le Monde) and thus perceived as such by readers and the public opinion altogether… The original tribune in La Croix was unscientific and unsubstantiated. What Tardieu may know or not besides what he wrote is of no interest to me in this discussion. My whole point is about this: using a title or a (former) affiliation does not make anyone an expert on anything, whether one is a professor in Dauphine, a retired INSEE inspector or you-name-it ! No one is an expert without proper arguments and facts.
October 24, 2009 at 9:43 am
hmm, sorry but
1) I heard Tadieu on the radio yesterday (on France Inter), the journalist who introduced him made very clear that Tadieu was speaking on himself, and was not representing Insee’s opinion in any way.
2) I am sure Tadieu is not an idiot, he knows that one should use a more refined model. What I find much more striking in this story is that journalists has raved for months about the “OMG so many suicides” before someone dared to compute a simple ratio! In the UK, the RSS would have reacted a bit more quickly, I presume, and certainly in a more balanced way than Tadieu, calling for a more refined analysis. That said, what is worse: a crude model (comparing rates), and no model at all (OMG many suicides)?
Whatever one’s political agenda (I’m not saying you have one, but in France so many people have, on both sides), one has to understand that a silly argument may be turned both ways. What if tomorrow a boss says everything is fine in her company because there was zero suicide out of 1000 employees for the last ten years?
As a statistician, the only two things I learned from this are: (a) a suicide rate is a very bad indicator of “bad social health” (whatever that means), and (b) journalists can’t do divisions, which is very amusing in a country where people complain about too heavily math-oriented school curriculums…
Now, I am not saying “work-related psychological pain” (sorry, I am certainly not using the right term) is not worth everyone’s attention, in FT and elsewhere in France, but I think that for a a more balanced statistical analysis can only be beneficial in this respect, possibly with the help on this new survey they’re implementing right now in FT.