Lies, damn lies, and statistics

When I took basic statistics and and statistical analysis, Disraeli's quote about statistics (that lies, damn lies, etc.) was always in the forefront of the discussions. There are ways to keep an eye on such things, on the distortions possible via our love of numbers and statistics, but you seldom see them used, especially in the popular media.

So it was with some interest that I saw in yesterday's Best of the Web a reference to a piece of statistical analysis done by the Israel-based International Policy Institute for Counter-Terrorism, called "An Engineered Tragedy". In it, the authors assert that the "scorecard" typically used in the media for comparing the number of Palestinians killed versus the number of Israelis lends a certain distortion to the reporting. Thus, they write that their "database shows a total of 561 Israelis killed, compared to around 1499 Palestinians, up to 30 June 2002." These numbers are essentially what the media often reports, seeming to say, "See, lots more Palestinians die, so all these suicide bombers can't be that bad."

But what ICT's statistical analysis says is:

[S]uch numbers distort the true picture: They lump combatants in with noncombatants, suicide bombers with innocent civilians, and report Palestinian “collaborators” murdered by their own compatriots as if they had been killed by Israel. Correcting for such distortions, we can arrive at a figure of 579 Palestinian noncombatants killed by Israel, compared to 433 Israeli noncombatants killed by Palestinians (see Graph 1.2). While Israelis account for a little more than 27 percent of the total fatalities as generally reported, they represent some 43 percent of these noncombatant victims. There are a number of valid ways of arriving at such corrected figures to compare the extent to which each side has been responsible for the killing of noncombatants; they all show a much less lopsided picture of the conflict than the raw totals do.

The ICT piece goes on to describe methodology, and presents figures and graphs to bolsters it conclusions. All in all, a completely different picture is painted than the one you get with the typical media scorecard.

But then, giving out full and correct numbers would require, well, effort. The boiler plate is sooooo much easier.