Skip to main content
Menu
Unskewed Polls, Early Voting Edition

Does the early vote suggest the polls are wrong? Democrats may hope so. The argument goes like this: Sure, the polls show the Republican winning in Colorado, Georgia and Iowa, but the early voting data in those states suggests Democrats have an advantage. Between the polls and actual votes, we should trust the votes (we don’t know who early voters voted for, but we know important information about them, such as party registration or race).

There’s two problems with this line of thinking. First, early voters and Election Day voters aren’t each drawn randomly from the electorate. Research shows certain groups are more likely to vote early. Second, early voting isn’t a secret; pollsters account for it.

The public polls in Colorado, Georgia and Iowa look right on target. They may even be slightly underestimating Republican strength.

Let’s start in Colorado. Democrats have argued that the public polls there, which show an average 2 percentage point lead for Republican Cory Gardner, are incorrect. They say their own polls show the race closer to a tie.

Democratic pollsters who have shown the race close to a tie have modeled an electorate in which registered Republicans make up 5 percentage points more of the electorate than registered Democrats. The two public pollsters, Marist College and Monmouth University, who released data on party registration gave Republicans a 7 percentage point registration advantage on average.

Because the state has adopted all-mail voting, there has been a substantial portion of the vote already reported (90 percent of the number of votes cast in the last midterm). Through Monday night, Republicans are holding a 7.5 percentage point lead in party registration among those who voted — slightly more Republican-leaning than the average in the public polls. Moreover, the Republican lead has held fairly steady through Monday. Democrats would need an incredible surge in the final day of voting for the actual registration of voters to look like what the internal Democratic polls had. That’s possible, but it looks unlikely right now.

Georgia, like Colorado, shows no signs in the early vote that the public pollsters are off. The key in Georgia is race, a good predictor of voting behavior in a state where the overwhelming majority of African-American voters favor Democrats and the overwhelming majority of white voters favor Republicans. Among seven pollsters (Landmark twice, Marist, Rasmussen, SurveyUSA twice and YouGov) who have released surveys in the last 10 days with a racial crosstab, the median poll had African-Americans making up 30 percent of voters. Perdue led by a median of 3 percentage points in these polls.

In the early vote so far, African-Americans made up 32.8 percent of early voters. That may make the polls seem wrong, but we probably will see the African-American percentage of voters drop to right around 30 percent given that African-Americans have always made up a larger percentage of the early vote than the total vote. In 2012, they made up 33 percent of early voters and 29.9 percent of all voters who casted a ballot.

Iowa doesn’t look much better for Democrats either. In party registration, Democrats lead 41 percent to 39 percent among early voters. Another 20 percent of early voters have been unaffilated, and some Democrats have claimed a 33 percentage point edge among that group. Let’s give them that. If both candidates hold onto 90 percent of their base, and unaffiliated early voters tip Democratic by a 2-to-1 margin, then Democrat Bruce Braley leads in the early vote by about 8 percentage points.

Six polls (the Des Moines Register, Fox News, Loras College, Marist and Quinnipiac University two times) over the last two weeks have released crosstabs on Iowa’s early voters. The median result of those crosstabs shows Braley up 9 percentage points among early voters.

Braley’s problem is that Republican Joni Ernst leads by 2 percentage points in the median of these same six surveys. In other words, polls are picking up Braley’s early vote edge, and they still show him losing — Election Day voters push Ernst over the finish line.

Now, none of this proves Republicans are going to win these seats. It’s important to not make too much of the early voting data. There are still plenty of votes to be cast in these three states. But Democrats claiming the polls are wrong based on the early votes should take a closer look at the data.

Harry Enten is a senior political writer and analyst for FiveThirtyEight.

Comments