The Forecast: Election night could go into overtime

Senate forecast: Republicans will hold 52 seats (and maintain control of the Senate) next Congress while Democrats will hold just 48. Anything between Republicans holding 49 seats and 56 seats is within the margin of error.

For those of you hoping the 2018 midterms will end on Election Day, I have some bad news: It's possible that we go into overtime.

I'm not just talking recounts. I'm talking about runoff elections in Georgia and Mississippi.

Let's look at Georgia first.

Georgia state law requires that a candidate receives a majority of the vote in order to win. Right now, it's not clear that either Democrat Stacey Abrams or Republican Brian Kemp will reach the magical 50%+ on November 6. That's because Libertarian Ted Metz is also running.

When one averages all the polls taken in October and allocate the undecideds between Abrams and Kemp, Kemp leads with 49.3% to Abrams 48.3% to Metz's 2.3%. In other words, if the polls were right on (and they probably won't be), Kemp and Abrams would face each other in a runoff on December 4.

It would be tempting to assume that Kemp would be a favorite for that runoff. Not only would he have earned the most votes in the general election, he'd be likely to pick up more of Metz's votes than Abrams would.

Additionally, runoffs in Georgia have not been kind to Democrats over the last 26 years. Back in 1992, Democratic Sen. Wyche Fowler won a plurality (not a majority) in the general election. He lost the runoff to Republican Paul Coverdell with turnout a little more than half what it had been in the general election.

In the 2008 Senate race, it was basically the same story. Republican Saxby Chambliss got three points more of the vote than Democrat Jim Martin in the general election but fell short of 50%. Chambliss, however, would beat Martin by 15 points in the runoff. Like in 1992, turnout was significantly lower in the runoff than it had been in the general election.

Low turnout on its own didn't hurt the Democrats in those races. It was the fact black voters didn't turn out in as high numbers relative to white voters in either runoff. Although Georgia's vote isn't divided along racial lines as much as it is in Mississippi, Democrats need high black turnout in order to win in Georgia.

The question, therefore, is whether black voters will turn out for a runoff election in 2018. It's not clear whether 1992 or 2008 are a good guide for what might happen.

Both 1992 and 2008 were presidential years, which usually feature significantly higher turnout than a midterm election. This is a midterm election where turnout is going to be lower. It's taking place in a year when we've already seen high Democratic turnout in special elections off cycle.

Perhaps as importantly, 2008 featured the first black presidential major party nominee in Barack Obama, who was not on the ballot in the runoff.

This year is something different all together. Abrams, who is aiming to become the first black woman governor, will be on the ballot in both the general election and the runoff. It's not difficult to imagine that black voters would come out to vote in the runoff in a way they might not otherwise if the Democratic candidate were white.

Indeed, Louisiana's recent runoffs tell a very different story than Georgia's. Looking to Louisiana isn't a perfect guide, but it gives us some guidance on midterms. Unlike Georgia, Louisiana, another southern state with racially polarized voting patterns, has recently held major statewide runoffs in midterm cycles. Both the 2002 and 2014 Senate races required runoffs.

Instead of seeing the black percentage of the electorate fall in the runoff relative to the November election, it actually rose in both 2002 and 2014. In 2002, it went from 25.9% to 27.1%. In 2014, it climbed from 28.8% to 30.3%. These 1-point to 2 point differences may not seem like they matter. They could.

Because of the state's racial voting patterns, increasing the black percentage of the electorate by even a point could swing a 1 point Kemp lead in the general election to a 1 point Abrams win in the runoff.

The power of the black vote in southern politics shouldn't be underestimated. In 2014, the top vote getter in the Mississippi Republican Senate race changed between the primary and primary runoff. Why? It was thanks to an unusual and seemingly successful attempt by Thad Cochran's campaign to get black voters who hadn't voted in the primary to vote in the runoff.

Speaking of Mississippi, it seems likely at this point that there will be runoff in the state's special Senate election. This Tuesday's election will act as a primary in which all the candidates, regardless of party affiliation, will run against each other.

The major vote getters in the primary look to be Democrat Mike Espy, Republican Chris McDaniel and Republican Sen. Cindy Hyde-Smith. The polling isn't entirely clear, but it seems plausible that Espy will finish first because the Republican candidates will split the Republican vote.

The bad news for Espy is that none of the polls have him close to a majority on Tuesday. In fact, no poll has any of the candidates close to 50% at this time. They all, however, have Espy and Hyde-Smith finishing either first or second with McDaniel in third. If no candidate receives a majority, a runoff in late November will be held between the top two vote getters.

Hyde-Smith would be a heavy favorite in a runoff against Espy. Our forecast has her favored by 15 points.

Were McDaniel to make the runoff with Espy, polls indicate that Espy would have a real chance of winning. That match up seems unlikely at this point.

One final and exhausting scenario is that after Election Day, Democrats control 50 Senate seats, Republicans control 49 Senate seats and Mississippi requires a runoff. If Espy and McDaniel make that runoff, we might not know who controls the Senate until late November. If it's Espy and Hyde-Smith, we won't be able to call Senate control next week, but the answer to the question of who will control will be more of mere formality than anything else.