Sunday, May 29, 2005

Bad Uses of Statistics



John Tierney gives us examples to ponder over in his latest NYT column (the one that should be Katha Pollitt's if there was any justice on this earth). The column is about how peaceful the world really is today, compared to past centuries:

The only antidote [to feeling that wars are prevalent] is to look at long-term trends instead of daily horrors. For a really long-term trend, consider that of 59 skeletons found in a Stone Age graveyard, at least 24 died from violence. Or that a quarter of the male population died fighting in some pre-agricultural societies.

In the 20th century, despite two world wars, humans had less than a 2 percent chance of dying in war or a mass killing, according to John Mueller, a political scientist at Ohio State. Today the risk is lower still - about a quarter the chance of dying in a car accident.

I mention these numbers not to minimize today's tragedies. I plan to be at a parade on Monday honoring the soldiers who have fallen, especially the more than 1,600 in Iraq. But I will also be thinking about the Progress Paradox and the origin of Memorial Day.

It started after the Civil War as Decoration Day, an occasion for widows wearing red poppies to decorate graves and memorials in virtually every town. If a war of that scale happened now, there would be nearly five million graves to tend. Sixteen-hundred is still too many, but if the trend continues, Memorial Day may eventually become a memory itself.

Can you spot the statistical mistakes here? There are at least two: first, it's incorrect to calculate the probability of dying in a war by including in the base all the people who lived in areas with no wars, and that's what Tierney's two percent figure uses. He then compares this to the findings from one archeological dig from one place. There's no way of knowing how representative that dig is of the general time and place.

The second one is the whole last paragraph which compares oranges to sausages in so many ways that I'm exhausted in just trying to list them. So I won't.

Tierney's column is correct in one sense, though. The world has become safer for many individuals over time, and one of the main reasons for that is something Tierney doesn't mention: the effect liberal and progressive ideas have had on social justice, education and opportunities for all.