Will the final polls be reliable?

In about a week, we’ll start to see various polling organizations releasing their final polls in key races across the country. How much weight should we give these polls in light of the fact that (a) this isn’t a presidential election year so (b) idenifying likely voters may be particularly problematic, and (c) this looks like a “wave election” for Republicans?
I don’t know the answer, but I thought I’d look at three races from, in essence, the beginning of this election cycle, all of which can be considered “wave” contests. I’m referring of course to McDonnell vs. Deeds for governor of Virginia; Christie vs. Corzine for governor of New Jersey; and Brown vs. Coakley for Massachusetts Senator.
How accurate were the final polls in these races? Quite accurate, it turns out.
In Virginia, the last two polls I found both had McDonnell way ahead. Survey USA had him up by 18 points, PPP by 14 points. McDonnell won by 17 points.
In New Jersey, a race made more complicated by a fairly strong third candidate, the very late polls by well-known pollsters showed Christie ahead. PPP had him up by 6 points; Survey USA had him up by 3. Christie won by 3.5 points. It should be noted, however, that a few very late polls had Corzine ahead. By contrast, a Rasmussen poll taken a short time earlier than the others was on the money, showing Christie ahead by 3 points.
Finally, in Massachusetts, the pollsters caught the wave and then some. The final two polls I saw were ARG (Brown plus 3) and Suffolk (Brown plus 15). The latter poll was an outlier. Brown won by 5 points.
This doesn’t mean that some (or perhaps even many) of the key races on November 2 won’t confound the pollsters. However, I’ll worry about any key race that the polls show to be close and worry a lot about any key race in which the polls show the Democrat with more than a de minimis lead.

Responses