Race drives the presidential horserace


Of all the many things that might affect the outcome of November’s presidential election, the most powerful could be a single number – the percentage of voters who are nonwhite.

President Obama currently gets the support of 3 out of 4 nonwhite voters, according to Gallup and other surveys. So as the nonwhite share of the electorate rises, so does his standing.

One way to gauge the impact is to look at two polls that have slightly different estimates of the nonwhite share of the vote – the Pew Research Center survey released Thursday and the Gallup tracking poll. Both are carefully executed surveys by organizations with long histories of producing accurate reads on public opinion, but each uses a somewhat different approach to putting together a representative sample of the electorate.


The Pew survey, conducted June 28-July 9, found Obama with a 50%-43% lead over Mitt Romney, the expected Republican presidential nominee. Compare that with Gallup’s tracking poll for the three-week period that overlapped with Pew’s, June 18-July 8. It averaged a closer result – Obama’s 47% to Romney’s 45%.

According to figures released by the two organizations, the voter population in the Pew survey was 28% nonwhite (including 13% black and 8% Latino), while the Gallup sample was 25% nonwhite (including 11% black and 8% Latino).

Obviously, the racial gap doesn’t account for the entire difference between the two surveys – and in any case, no two random surveys are likely to be entirely in line with each other. But Gallup consistently has had a slightly smaller percentage of nonwhite voters in its surveys than Pew and some other polls, and that is one reason it has tended to report lower numbers for Obama.

Which poll is “right”? That’s not a question that can be fully answered in advance of the election.

In 2008, nonwhite voters made up 26% of the electorate, according to exit polls. The nonwhite share of the electorate has grown slightly in each of the past several presidential election cycles, largely reflecting the growth of the Latino voting population. If that pattern holds again this year, then Pew’s 28% nonwhite share will be on the money. That would also be good news for Obama.

On the other hand, job losses in the recession hit minority communities disproportionately hard, forcing a lot of people to move and meaning that many blacks and Latinos will have to re-register in order to vote in November. So it’s possible that the nonwhite share could drop a point, which would correspond to Gallup’s number and would also make Obama’s reelection much harder.

Why do polls differ on something as basic as the percentage of nonwhites in the electorate? Because most people don’t respond to surveys – a problem that has gotten worse in recent years – pollsters have to “weight” their samples somewhat in order to make sure that the people they have polled accurately reflect the voter population. But weighting always involves trade-offs – there’s no perfect way to do it – and each polling organization goes about the work somewhat differently.

With race such a factor in this year’s election, those subtle issues of polling methodology can make a significant difference.