Dumb Polls

Yesterday, the LA Times screamed that Obama was leading McCain by 12 percentage points in their opinion polls. My first reaction was, "Big fucking deal? It's June." My second reaction was, "Man, could that be any more misleading?"

I do a lot of statistics as part of my job, so bear with me here. The poll is misleading because it was based on national opinions, but that's not how our presidential elections are scored. A more accurate poll would show which candidate is leading in which states and how that adds up in the electoral vote (see Bush v. Gore, 2000). Also, note that not all polls are equally accurate.

Good pollsters divide their samples into many groups (white men over 40, Hispanic females with children, etc) as they have different voting rates, regardless of what they say is their likelihood of voting. This year, one of the big wild cards could be the African-American vote. Obama's candidacy raises the question as to whether African-Americans will vote in greater numbers (which the polls can right now only guess at) in November than in past elections.

Don't necessarily be thrown by the excitement in the primaries as the turnout between them and the general election is not as highly correlated than you think. Also, Republicans only tend to get a sliver of the African-American vote (Bush got about 9%, and that was after some serious ass-kissing with some black evangelicals) anyway, so it's not as if Obama will get a much higher percentage. The big question is whether more African-Americans will come out and vote. The pollsters who have a bead on that will be the most accurate.

Newspapers and websites LOVE polls because they are a short-cut to discussing the real issues. Just be sure you read a bit more into them before deciding if they are telling you anything meaningful.