Public Polling Deep-Dive in Georgia and Ohio

It’s always interesting to see where late-deciders break in an election or who else shows up. Looking at Georgia and Ohio, two states we polled ten days before the elections, something stood out to me. Both states had extremely close governor’s races in our polls (2.2 points in Georgia and .2 points in Ohio), and there were third-party candidates who got a not-insignificant percentage of the vote in our final poll. (3.8% in GA and 4.7% in OH).

As we now know, Georgia ended up as a 1.5-point race and Ohio ended up with a 4.3-point difference. What stood out to me was first that the third-party candidates all got a lower than projected percentage of the vote. Second, and most interesting, is that those voters broke one way in Georgia, and another in Ohio.

In Georgia, by a 1:65:1 margin, Abrams – the Democrat – got those votes. In Ohio, by a 2.3:1 margin, DeWine – the Republican – got them. And it’s not as simple as saying those who wanted the right-leaning third-party candidate went to the Republican, and vice versa.

In Georgia, there was only one third-party candidate and he’s a Libertarian. In Ohio, the left-leaning candidate (Green Party) actually lost a larger share of her votes from our poll to the actual election than the right-leaning (Libertarian) candidate.

So how do you explain this?

Obviously it was Oprah knocking on doors for Abrams in Georgia.

Kidding! (sort of…).

Based on our polls, here are some things it wasn’t:

  • Image: Kemp and Cordray had the better respective images by far.
  • Age: Kemp led with voters under 50 and over 50, while DeWine trailed in both age groupings.
  • Trump Impact: Georgia voters said they were more likely to vote for republicans because of or despite their view of the president, and less likely to vote against republicans to send a message to the president.

Based on our polls, here’s what it was:

Voter Propensity: This is our proprietary measurement of an individual voter’s likelihood to vote based on their overall voting history, both in similar elections and others. Voters as categorized in one of three voter-propensity categories: high, medium, and low. Take a look at who voters in each category said they favored in these states:

Ohio Georgia
DeWine Cordray Kemp Abrams
High-Propensity 46.1% 44.7% 53.2% 42.9%
Medium-Propensity 41.4% 41.3% 48.8% 47.6%
Low-Propensity 44.1% 36.9% 41.9% 53.7%


It’s rare that we see the Republican favored with low-propensity voters, but that’s what our poll showed in Ohio. Georgia was much more typical, with Abrams leading by nearly 12 points with low-propensity voters. As a result, with extremely high turnout in each state relative to similar past elections, DeWine’s lead increased and Kemp’s decreased as those low-propensity voters showed up and chose one of the two traditional party candidates.

So what does this all mean?

First of all, it means our numbers were right.

Second, it means that if you only look at the toplines of a poll, you’re missing the point. You have to look at the crosstabs to get the full story – or perhaps more accurately, the full range of possible stories.

A poll isn’t all about getting the final margin of victory correct – although we’re pretty darn good at that, too. It’s about uncovering the clues of what could happen so you can influence what ultimately does happen