The Washington PostDemocracy Dies in Darkness

What ticket-splitting in the midterms looked like

Voting stickers are ready to be handed out Tuesday at a church gym in Atlanta. (Michael M. Santiago/Getty Images)

If Georgia voters had voted for the same party straight down the ballot, there would be no runoff in the state’s Senate race. Herschel Walker, the Republican Senate candidate, would have beaten Sen. Raphael G. Warnock (D) by nearly 8 points — the margin by which Georgia Gov. Brian Kemp (R) won reelection.

But that’s not how they voted. Walker received 200,000 fewer votes than Kemp. Control of that seat, and potentially the Senate overall, will be settled early next month.

Of course, if voters in Wisconsin had similarly voted on party lines, control of the Senate would probably already be determined by the time of the runoff. There, Sen. Ron Johnson (R-Wis.) narrowly beat Lt. Gov. Mandela Barnes (D). Had Barnes received as many votes as Democratic Gov. Tony Evers, Democrats would have picked up that seat.

It’s not uncommon for voters to either split their votes between parties or vote for a candidate in one race but not another. This year, though, with control of the Senate at stake and many close races making the difference, the distinction between gubernatorial and Senate votes is particularly interesting.

Sign up for How To Read This Chart, a weekly data newsletter from Philip Bump

The situation in Georgia was predictable months ago. Polling repeatedly showed that a chunk of the electorate planned to back both Kemp and Warnock — splitting their vote between the Republican and Democratic incumbents. Polling also showed that voters in Georgia were less excited about voting for Walker than Kemp, suggesting that some might vote for the top of the ticket but not further down.

With (incomplete, in some cases) results in hand, we can now visualize the gap between Senate and gubernatorial candidates in a number of closely watched states. What’s more, we can see how the gap varied within the state.

Here, for example, is a comparison of the gap between the Republican candidates with education levels by county. (The vertical axis on these charts is the two-party margin in the governor’s race minus the two-party margin in the Senate race. Since Johnson outperformed Republican gubernatorial candidate Tim Michels in Wisconsin, these figures are negative numbers for Wisconsin.) You can see a clear correlation in Georgia: as the percentage of residents in a county that has a college degree increases (from left to right) Kemp’s margin was increasingly larger than Walker’s.

In other states, it’s not as clearly correlated. It’s also worth noting that this isn’t as strongly correlated to the county’s 2020 presidential vote, meaning that this isn’t simply capturing big Democratic cities.

We can see a similar pattern when we look at income, which itself correlates to education. Counties with higher median incomes gave bigger margins to Kemp than Walker.

Income also often correlates to race, but we don’t see a strong correlation between the percentage of residents in a Georgia who are White with an advantage for Kemp. We do see a correlation in Arizona, where gubernatorial candidate Kari Lake outperformed Senate candidate Blake Masters by a much wider margin in counties with a lower density of White residents. That there are not many counties in Arizona, though, offers an important caveat.

The charts for Pennsylvania and Wisconsin show something different: Neither of those states has many counties with high percentages of non-White residents.

The divide on education in Georgia is interesting. We saw similar divides in the Republican primaries, where places with more college-educated voters were less likely to support candidates viewed as aligning with former president Donald Trump. That Kemp (Trump target) would fare better than Walker (Trump endorsee) isn’t as surprising in that context. But it also suggests that there weren’t similar divides between the two candidates in the other states.

Another central takeaway here is that there isn’t necessarily a clear explanation beyond that some candidates were more broadly preferable to voters than others. That, say, Pennsylvania gubernatorial candidate Doug Mastriano was far worse at campaigning and faced a much tougher opponent that the Republican Senate candidate in that state, Mehmet Oz.

Or, in that case, that Trump’s endorsement didn’t do either of them many favors.

Lenny Bronner contributed to this article.