Home > News > What Polls Don’t Mean
113 views 2 min 0 Comment

What Polls Don’t Mean

- February 27, 2009

Gallup declares “Obama Speech Bolsters Confidence for Many Americans.”

bq. President Barack Obama’s address to Congress Tuesday night appears to have bolstered confidence among many Americans. Four in 10 (41%) say they are now more confident in his plans to improve the economy, including 57% of those who watched or listened to the speech live.

bq. In a one-night Gallup Poll conducted Wednesday night, Americans overall were about evenly split between saying Obama’s speech made them more confident and saying it had no effect on their opinion. But those who reported watching or listening to the speech live were far more likely to say it made them more confident, out numbering by a 2-to-1 margin those who said it had no effect. Fewer than 2 in 10 Americans in either group said the speech made them less confident.

Nothing about this poll is particularly useful. In fact, its value as social science is about nil. Why?

* People do not accurately report whether they have watched, read, or heard some particular media event. Indeed, if we believed surveys that measure attention to, say, network news, the audience would be roughly three times larger than it really is. See this paper by Markus Prior.

* People do not accurately report their internal mental processes. Lee has posted on this before. So when survey respondents tell you that something has made them “more confident” or has had “no effect,” they’re not telling you anything reliable.

Perhaps the only trustworthy thing to emerge from this poll is that Democrats were likely to report that Obama’s speech made them “more confident” than were Republicans. There’s plenty of evidence that Democrats and Republicans perceive political events through their own partisan lens.

Oh, and to top if off, Gallup reports that Obama is “gaining” in the polls (see here), even though his “gain” cannot be outside the margin of error.

Argh.