Saturday, October 24, 2009

Bad math at Fox News

The other day I was working out and flipping channels, and I saw Fox News folks commenting about a recent poll that showed support for the public option in the health care debate. What struck me was that they were latching on to the fact that more of the respondents to the poll identified themselves as Democrats than as Republicans, and they claimed that this made the poll flawed. In fact, they specifically said that the poll should have sampled an equal number of Democrats and Republicans.

Nice sentiment, but bad math. The point of a poll is to figure out the popularity of an issue among the general population (or a specific subset of that population such as voting adults, or seniors, or such if that is the specific goal of the poll). If you believe that the poll should equally sample Democrats and Republicans, that is basically assuming that the percentages of Democrats and Republicans are equal, which they are not. Rather, the ratio in the poll of Democrats and Republicans, if done correctly, should approximately match the ratio in the target population, and this almost certainly is not 1:1, as the Fox commentators seemed to think.

I do not know whether the ratio reported in the poll does match the broader population; if it was off by a meaningful amount, then that would have been a sign of a potential flaw in the poll. But alas, Fox did not report this information, so I find myself strangely uninformed by their broadcast.

No comments: