The race is tightening! The sky is falling!

The horse race meme is still alive.

You probably saw that the AP published a poll yesterday that showed Barack Obama up by only one point over John McCain. Well, I've got news for you--and for Sen. McCain: The poll is bogus. Or, as the Car Talk Guys would put it, BO-OH-GUS!

One key is in the poll's sample of "likely voters," which is the figure that was reported. (You can see the poll itself here.) The problem lies in figuring out who are the people likely to vote. From all respondents, Obama has a 10 point lead. But among "likely voters," as interpreted by the poll, that margin comes down to a single point. Why? We don't know, because we don't know how "likely voters" were culled from all of those who responded. But we do know that the results are substantially different from other polls.

The factor most often used by pollsters to measure "likely" voters is past voting behavior. But as we know, millions of new voters have registered. Not all of them will vote. But many of them have registered because of their enthusiasm over the race this year. And the lion's share of those voters are going to show up and pull the lever (or, more likely, mark the little circle or fill in the blank) for Barack Obama. So polls of "likely" voters are suspect to begin with.

There is also a significant error on page 24 of the poll, where it is related that 45 percent of respondents said they were evangelical Christians. To begin with, only 55 percent of respondents identified themselves as Christian (54 if you are one of those who doesn't count Mormons as Christians). Even if the percentage of evangelicals is 45 percent of 55 percent, that is 24.5% of the sample, more than there are in the nation (some polls said that 23 percent of voters in 2004 were evangelicals, but in fact evangelicals comprise only about 16 percent of the population, and do not vote at a rate 1 1/2 times that of other groups).

Then, too, the percentage of Jews in the sample is only 1 percent; Jews constitute approximately 3 percent of the population. The percentage of Muslims in the sample is too small to be noted; while Muslims are not a large percentage of voters, they are a measurable fraction.

The AP poll is, at best, an outlier. Very likely it was influenced, albeit unconsciously, by the desire to keep the horse race story plausible.