Fun With Statistics

The Washington Post is at it again. An atrocious front-page, top-of-the-fold story claimed that Bush has a lead over Kerry in a recent poll. But the 50-47 lead they cite is within their admitted margin of error. For non-stat majors out there, the margin of error is the space within which you can't confidently say anything about the difference in two numbers. In short, a spread less than this margin is statistically insignificant, and absolutely nothing can be said with confidence aside from that the values are equal. The poll shows a tie. Period.

I called the ombudsman (202.334.7582), who promptly rationalized the Post's error by claiming that all news organizations report polls this way. That's right, it's OK for the Post to misrepresent the facts because everybody else does it too.

Wow.

For all the non-wonks asking why this is important: the media have a habit of systematically mischaracterizing poll results, which they then run prominently as fact. Degradation of the discourse aside, these stories can influence people's perceptions of the nature of the presidential race, and a sense of hopelessness can deter voters. Also, the story is an example of the increasing encroachment of analysis over substance in the news.