Undecided VoterDan Hopkins at The Monkey Cage takes the idea of polling bias for independents more seriously than I do: Polling Biases and Their Potential Impacts. This is in reference to Jay Cost's claim that the independents are going to break hard for Romney.

Frankly, I don't see the problem. Over half of all independents are just Democrats (24%) or Republicans (32%). And my experience is that Republicans who call themselves "independent" normally do so because they are more conservative than their party. At this point, all of these people know how they are going to vote. "Independent" does not mean "undecided."

Let's do a little calculation, shall we? Let's look at Ohio where Obama is leading in the poll average by 48.5% to Romney's 45.8%. I'm going to assume that the makeup of the undecideds is the same as the make up of the independents. This is the best case scenario for Romney. That leaves 5.7% of the electorate undecided. Let's assume that the Democrats will vote for Obama and the Republicans will vote for Romney. That will add 1.4 and 1.8 percentage points to the two candidates. This makes the race 49.9% (Obama) to 47.6% (Romney).

Let's further assume that the remaining (true independents) will break as the partisans did: 51.2% for Obama and 48.8% for Romney. This gives an additional 1.3 and 1.2 percentage points. Thus, Obama should win Ohio 51.2% to 48.8%. (Note: the numbers are almost identical to the numbers before we added the true independents because they split so evenly.) This gives Romney an extra 0.3 percentage points over the total he gets from the poll average.

In order to win 50.0% of Ohio, Romney would have to win 74% of the undecideds. That isn't reasonable given what we know about the electorate.

I did this calculation to show that Romney is not secretly winning in the polls. The kinds of tricks proposed by Jay Cost and his colleagues could also be used to make Obama look better. We could claim that the polls are under-representing who will actually vote. In the end, the polls only give us a decent idea of who will actually vote and how they will vote. At this point, I'm going to believe the polls. The Weekly Standard folk can use the next week to complain. And then we'll see who was right.