To estimate the popular-vote totals, I then need to do two things. First, I need to know what share whites and non-whites will each have in the electorate. In most cases the pollster provided that information, but where it was not provided I could derive the totals through interpolation. Whites plus non-whites must equal 100 percent of the sample, and since I know what share of each group each candidate takes, I can use middle-school algebra to determine the shares of whites and non-whites in the electorate. That table is presented below.
You again see remarkable similarity. If you average these together, you get a rough estimate of 74 percent white, 26 percent non-white. This is identical to the 74–26 breakdown in 2008, and much more non-white than the 76–24 breakdown in 2010.
Finally, I need to estimate what share of the vote each candidate will get among whites and non-whites after undecided voters have made up their minds. I assume third-party candidates will receive about 2.5 percent among whites and 2 percent among non-whites, for a total of 2.33 percent overall. I then allocate 60 percent of the remaining undecideds to Romney in each group. That means my calculations are based on the assumption that Romney will beat Obama by 57.7 to 39.8 percent among whites, and that Obama will beat Romney by 77.4 to 20.6 percent among non-whites.
My final prediction is based on applying those percentages to a 74.5–25.5 white/non-white breakdown, a slightly more optimistic forecast for Governor Romney. I did that because most polls’ unweighted samples have produced a much higher white share of the electorate than the average based on self-reported voter enthusiasm and other factors. Many observers have argued that these measures tend to undercount non-white voters as a share of the likely voter sample; I decided to slightly reduce the non-white share of the electorate from my rough estimate to account for the likelihood that white turnout will be slightly higher than the pollsters’ estimate.
As you have probably surmised, this conclusion is extremely sensitive to changes in assumptions regarding white and non-white voters. So I asked my research assistant, Brad Wassink, to create multiple scenarios based on different assumptions. I asked him to run every possible scenario assuming that Romney’s share of the white vote went from 57.5 percent up to 59 percent, and with his share of the non-white vote ranging from 18 to 20 percent. (I did this on Sunday afternoon before the latest polls were released, and those polls showed Romney getting larger shares of the non-white vote.) I also asked him to run each of these scenarios with white voters’ share of the electorate going as low as 73.5 percent and as high as 76 percent.
Here’s the gist of his findings: Romney cannot win the popular vote unless he takes a minimum of 58.66 percent of the white vote and 20 percent of the non-white vote and whites are 74 percent of the electorate. His chances for victory increase the higher the share whites are in the electorate, such that at the 76 percent level, Romney wins the popular vote if he takes 58.66 percent of the white vote, so long as he takes 18 percent of the non-white vote (McCain’s level).
Brad created 150 scenarios; Romney wins the popular vote in only 33 of them. But even this masks how hard it is for Romney to win the presidency. For reasons I explain later, Romney has to win the popular vote by a minimum of 1 percentage point to have a chance of winning the Electoral College. Romney’s margin exceeds 1 percentage point in only six of those 150 scenarios.
It’s not that Romney can’t win; it’s that he’s highly unlikely to do so without getting the highest share of the white vote for a Republican since 1988 and facing the whitest electorate since 2004 unless he has substantially improved his standing with Hispanics relative to McCain’s. But there is precious little evidence that is happening. The 2008 exit poll showed McCain getting 31 percent of the Hispanic vote; a Fox special poll of Hispanic likely voters released last week showed Obama leading by 77–19. In the absence of instant “self-deportation,” Romney is likely to at best replicate McCain’s total among Hispanics and thus is exclusively reliant on the white vote to win.
And Blue All Over
Now, it’s not as if Republicans have never received 60 percent or more of the white vote. GOP candidates have done that three times in the last 40 years — 1972 (68 percent), 1984 (64 percent), and 1988 (60 percent). When they did that, they carried the midwestern battleground states of Pennsylvania, Ohio, Wisconsin, and Michigan. So Romney’s path is clear: Break the 60 percent hurdle and victory — popular and electoral — is assured.
The problem with that is the type of white voter he needs to win to get there. Republican candidates break the 60 percent barrier only when they do well among midwestern and northeastern Catholic and Lutheran blue-collar voters, many of whom belong to a union or come from a union heritage. These white, working-class voters have so far been resistant to Romney’s charm in both the Republican primaries (areas dominated by these voters tended to vote for Santorum) and in all of the midwestern state polls to date. Without them, Romney cannot win the election.
The data from Ohio are particularly poignant. Racial breakdowns are available in five recent polls, and Romney receives between 50 and 53 percent of whites in all of them. Whites will likely comprise 83 to 85 percent of the Buckeye State’s electorate, but in order to win, Romney needs at least 56 percent of their votes. He’s not getting it, and in large measure it’s because he’s not persuading blue-collar whites. Nationally Romney, like most Republicans, runs stronger among whites without a college degree than among those with a college education, but not in Ohio. There, according to the CBS/Quinnipiac poll, he’s running five points worse among the white working class than among college-degreed whites.