Bill McInturff

In a memo released to reporters, McCain's pollster makes a variety of arguments about how the race is "functionally-tied." Some of this memo is reasonable. But ...

... I direct your attention to his sixth point. Here, McInturff argues that Obama's support among African-Americans is totally factored into the public polls as "most polls" show Obama winning 97-1 or by a similar margin among African-Americans. So, Obama can't expect any more support there and McCain is likely to gain support as the remaining voter pool (non-African-American) is more likely to support McCain. This is utter nonsense for several reasons:

1. McInturff says, “in most polls,” Obama is winning 97-1 among African-American voters. That just isn’t true. Ben Smith provides evidence to the contrary here. More importantly, the state polls that release internal demographic data very frequently show McCain winning far more of the African-American vote than he is likely to win as McInturff suggests happens in his polling in past campaigns. For instance, look at SurveyUSA’s poll from Monday in Virginia. This poll has Obama ahead by 9 points overall in the state and winning 86-13 among African-Americans. Does McInturff concede that Obama is underperforming in this poll among African-Americans? If so, Obama’s lead should be slightly larger.
2. Let’s assume away the problems I identify in point #1 and assume McInturff is right that the African-American support Obama is going to get is already factored into the polling results. McInturff says this means that “it will be very difficult for Senator Obama to perform much above his percentage of the vote in a state” because “the only undecided/refuse to respond voters are white and Latino.” Huh? Barack Obama is winning by 2 to 1 or more among Latino voters in most places and, while he is not winning among white voters, he is not being blown out either. In fact, it is entirely possible that if the only remaining undecided voters are white and Latino voters, Obama could earn a split or even win among these voters. It frankly depends WHICH Latino and white voters are undecided but to assume the vast majority or even the majority of these voters will go to McCain is not an assumption supported by any real evidence.
3. Finally, McInturff discusses turnout elsewhere in the memo but fails to mention that many of the state-level polls are probably underestimating what is likely to be a record high turnout among African-American voters. So, even if these polls are already accurately factoring in the proportion of African-American voters Obama is likely to win (many of them are not), they may not factor in the higher proportion of the electorate African-American voters are likely to constitute. For instance, in the same Virginia poll by SurveyUSA I mentioned above, African-Americans make up 18% of the electorate in their sample. But African-American voters made up 21% of the electorate in Virginia in 2004 according to exit polls. Similarly, in North Carolina, a SurveyUSA poll a week ago estimated African-Americans would make up 20% of the electorate in the state. But, again, according to exit polls in 2004, African-Americans were 26% of the electorate in the state. Even if the African-American portion of the electorate is simply as high as 2004, Obama still has room to grow in these polls just among African-American voters.

McInturff is a smart and good pollster. But this particular point he is making is simply not supported by the data. Beware of the campaign that sends the internal pollster out to explain why things are not as the public polls seem. They usually are as the public polls say.