There’s an incredible statistic out there that says that 70% of the population of France’s prisons are Muslims, despite only 8-10% of the population being Muslim. It’s everywhere – in this Telegraph article by Harriet Alexander, in this Washington Post article by Molly Moore, in this New York Review of Books article by Scott Atran and Nafees Hamid, in this report by a French politician (p. 6), and most recently in this Sunday Times article by Niall Ferguson (which he quickly corrected after I pointed out the mistake).

But it’s not true.

Its origin of this stat seems to be a book called Islam in Prisons by Farhad Khosrovkhavar, a French sociologist, though he says he doesn’t use it himself and the figure has been misattributed to him. Prof Khosrovkhavar carried out a survey of four prisons in ‘sensitive’ areas in Paris and the North of France (out of 188 across France).

I emailed Prof Khosrovkhavar, who rejects the 70% figure altogether and says that he reckons a true figure is ‘around half’ – 40%-50%. But (he stressed) these are just estimates, because the French government does not record these things. 

The closest thing to an official figure is the number of French inmates who registered for Ramadan – 18,300 out of a total prison population of 67,700, or 27%, back in 2013 according to Agence France Presse. Prof Khosrovkhavar suggests that this could be an underestimate, because some Muslims will fear being ‘noted’ by the intelligence services. A Brookings Institution report says that “Muslims are greatly overrepresented in prisons and within the eighteen- to twenty-four–year-old age group in particular: they make up only 8.5 percent of that age cohort in France, yet 39.9 percent of all prisoners in the cohort.” Nobody seems to know for sure.

This, obviously, is not to suggest that France doesn’t have a serious problem with integrating Muslim men (in England and Wales, 15% of the prison population is Muslim from a total population of 5%). But the enormous 70% figure is false, and should not be used – no matter how many reputable-seeming outlets have been taken in by it.

This all reminded me of a passage in Quine and Ullian's The Web of Belief (my thanks to my mother for finding the passage for me):

An author of this book remarked after walking about the principality of Monaco, "Just think-only eight square miles!" "I don't see how you even get eight out of it," his brother replied. The map was conclusive: you couldn't. Yet the Encyclopedia Britannica, the World Almanac, Scott's stamp album, various American atlaseses, and the gazetteers in the dictionaries had agreed on eight square miles. Hachette and Larousse turned out to agree rather on 150 hectares, or less than three fifths of a square mile. A subsequent check of the Britannica (eleventh edition) revealed arresting detail: "Area about 8 sq. m., the length being 2 1/4 m. and the width varying from 165 to 1100 yds." Even this arithmetical absurdity had not prevented the producers of all those other reference books from copying the figure of eight square miles, if the Britannica was their source. We are happy to report that the myth broke at last and the "sources" subsequently consulted converged on 0.59 square miles. There is even a new alertness: 0.71 is now reported, because of 76 acres lately reclaimed from the sea. But there is very likely some unwarranted figure on another topic that we are all accepting still, or even newly. 

The policy of seeking safety in numbers by checking multiple sources is an excellent precaution; but, as the above example illustrates, it can fail when the sources are not independent. No one would check a newspaper report by checking more copies of the same newspaper. There is a saying that 4 X 107 Frenchmen can't be wrong, but the contrary is the case if they all believe what one wrong Frenchman tells them. In the foregoing example, admittedly, the Frenchmen were right.