October 12, 2007

Statistics on BBC radio

I thought Mark's discussion of the almost wilful ignorance of basic statistical concepts in our culture was not just fascinating but very important. And here is a footnote to it. A few minutes ago I heard on BBC's Radio 4 morning news program "Today" a reporter explaining that blue tongue disease in sheep could cause "up to 70 percent mortality in some animals."

I suppose there have been days when I felt 70 percent dead, but that's not what she meant. She just wandered from talking about a statistical fact concerning populations to talking about effects on individual sheep. She didn't mean "in some species of animal", if that's what you're thinking, because she had already limited the claim to how devastatingly serious blue tongue disease is in sheep.

Doubtless just an on-the-fly slip; but in a national news broadcast, it seemed to me a significant indication of quantitative carelessness nonetheless.

As if to emphasize the point, within about a few minutes on the same program another story, about a UN report on childbirth mortality, mentioned that half a million women die each year in childbirth and that the worst thing was that this figure "has remained unchanged" for a number of years. If that claim were true, it would of course be great news: since the population of the world has been increasing and the number of births is linearly related to the total population, this would mean that deaths in childbirth, per thousand births, were constantly declining from year to year. But again, that's not what was meant, since the hook for the story was a new report concerning the failure of the situation to improve. I think it was just one more case of pulling out a constant number when only a percentage would tell us what we needed to know. And that, along with its converse (giving us a percentage where only an absolute quantity would make sense) must surely be the most prevalent of all the conceptual slips in talking about quantities that Mark addressed.

Update: Jonathan Weinberg points out to me that in fact "the (very careful) UN report goes on to place the 529,000 figure between a lower uncertainty bound of 277,000, and an upper uncertainty bound of 814,000, maternal deaths." In other words, they do not really know whether the figure is even roughly constant — not to within one to three hundred thousand maternal deaths. That was not mentioned in the radio stories about the UN report, and makes the error of asserting that things are getting worse (or not getting better) much worse.

Posted by Geoffrey K. Pullum at October 12, 2007 02:00 AM