Mike Espy Needs An Unprecedented Swing To Win The Mississippi Runoff

We don’t often see a runoff in a general election, but if Sen. Cindy Hyde-Smith were to lose Mississippi’s Senate runoff on Tuesday, after the two Republican candidates combined to win a sizable majority of the initial vote, that would be even more unusual.

In the first round, Republicans Hyde-Smith and Chris McDaniel combined for a bit less than 58 percent of the vote, while Democrat Mike Espy and one other candidate from his party together won a little more than 42 percent. For Espy to win, the runoff vote has to swing more than 15 points more Democratic than the initial vote margin. But if we look at the five Senate elections since 1990 where an initial round of voting was held on the national Election Day and two candidates advanced to a runoff,1 no challenger has ever come close to outperforming the previous round of voting by the kind of margin Espy would need to win. What’s more, no runoff has ever shifted that much in either party’s direction.

Democrats need a record-setting swing to win Mississippi

Shifts in vote share margin between first election and runoff, in Senate races since 1990 where the first election took place on the national November Election Day in a midterm or presidential cycle

vote share margin
Year State First Election Runoff Swing
1992 GA D+1.6 R+1.3 R+2.9
2002 LA R+2.9 D+3.4 D+6.3
2008 GA R+2.9 R+14.9 R+12.0
2014 LA R+12.3 R+11.9 D+0.4
2016 LA R+25.4 R+21.3 D+4.1
2018 MS R+15.3

In cases where the first election included multiple candidates from the same party (all Louisiana and Mississippi races in the table), the margin is based on the difference in the total vote share between all Democratic and Republican candidates. In some cases, parties had unequal numbers of candidates running, which may have exaggerated the differences in vote share. Some data may not add up due to rounding.

Sources: ABC News, Georgia Secretary of State, Louisiana Secretary of State, UVA Center for Politics, Federal Election Commission

In fact, the biggest swing recorded — 12 points in the 2008 Georgia Senate runoff — favored the incumbent and not the challenger. Of the five previous Senate elections that went to a runoff, four featured incumbents, and in those contests, the swing in party vote share moved in the challenger’s favor only once. In Georgia’s 1992 Senate election, Democratic Sen. Wyche Fowler missed a majority by 1 percentage point in the general election, and in the runoff, challenger Republican Paul Coverdell squeaked out a narrow victory as the margin shifted 3 points in the GOP’s direction.

As for that big swing in Georgia’s 2008 Senate election, it came about in part because of the unprecedented circumstances surrounding the general election. The initial vote took place during Barack Obama’s first presidential bid, which saw African-American voters turn out at much higher rates than usual, both nationally and in Georgia. That turnout did not repeat itself in the runoff, however, and Republican Sen. Saxby Chambliss won re-election by 15 points.

The next-largest shift (and the largest shift in a midterm election) also benefited an incumbent, although it was much smaller. In 2002, four GOP candidates combined for a narrow edge over two Democratic candidates in the initial vote, but Democratic Sen. Mary Landrieu narrowly won the Louisiana Senate runoff by shifting the margin 6 points in a Democratic direction. So a Democrat has moved the needle enough to win a runoff in the Deep South in the recent past, although the politics of Louisiana (and the rest of the region) have shifted further toward the GOP since the early 2000s. Moreover, Landrieu didn’t need nearly as large a swing as Espy does in 2018.

From this admittedly small sample of Senate runoffs, an Espy win looks like a long shot: These races rarely swing by nearly as much as he needs to win, and when they swing at all, it’s more often toward the incumbent than the challenger. Furthermore, the Nov. 6 result in Mississippi looks similar to the 2014 and 2016 contests in Louisiana, which also bodes poorly for Espy. In those races, the GOP candidates combined for a double-digit margin in the first election, and in the runoff, the Democrat only gained a little ground. In that 2014 race, Landrieu was voted out of office after nearly 20 years in the Senate (no incumbent ran in 2016). As Nathaniel Rakich wrote in our preview of the Mississippi runoff, Espy needs today’s electorate to be substantially different from the voters who turned out earlier this month; Espy likely needs a strong turnout from black voters, who lean very Democratic, combined with some apathy among white voters, who are strongly Republican in Mississippi.

Something Looks Weird In Broward County. Here’s What We Know About A Possible Florida Recount.

The Florida U.S. Senate race is still too close to call. According to unofficial results on the Florida Department of State website at 11:45 a.m. Eastern on Friday, Nov. 9, Republican Gov. Rick Scott led Democratic Sen. Bill Nelson by 15,046 votes — or 0.18 percentage points. We’re watching that margin closely because if it stays about that small, it will trigger a recount. It’s already narrowed since election night, when Scott initially declared victory with a 56,000-vote lead.

The changing margin is due to continued vote-counting in Broward and Palm Beach counties, two of Florida’s largest and more Democratic-leaning counties. On Thursday evening, the supervisors of elections in the two counties told the South Florida Sun Sentinel that vote counting there was mostly complete. Under Florida law, counties have to report unofficial election results to the secretary of state by Saturday at noon, but Nelson’s campaign is suing to extend that deadline. Scott’s campaign and the National Republican Senatorial Committee are also suing both counties for not disclosing more information about the ongoing count, and Scott called on the Florida Department of Law Enforcement to investigate Broward’s handling of ballots.

Unusually, the votes tabulated in Broward County so far exhibit a high rate of something called “undervoting,” or not voting in all the races on the ballot. Countywide, 26,060 fewer votes were cast in the U.S. Senate race than in the governor race.1 Put another way, turnout in the Senate race was 3.7 percent lower than in the gubernatorial race.

Broward County’s undervote rate is way out of line with every other county in Florida, which exhibited, at most, a 0.8-percent difference. (There is one outlier — the sparsely populated Liberty County — where votes cast in the Senate race were 1 percent higher than in the governor race, but there we’re talking about a difference of 26 votes, not more than 26,000, as is the case in Broward.)

To put in perspective what an eye-popping number of undervotes that is, more Broward County residents voted for the down-ballot constitutional offices of chief financial officer and state agriculture commissioner than U.S. Senate — an extremely high-profile election in which $181 million was spent. Generally, the higher the elected office, the less likely voters are to skip it on their ballots. Something sure does seem off in Broward County; we just don’t know what yet.

One possible reason for the discrepancy is poor ballot design. Broward County ballots listed the U.S. Senate race first, right after the ballot instructions. But that pushed the U.S. Senate race to the far bottom left of the ballot, where voters may have skimmed over it, while the governor’s race appears at the top of the ballot’s center column, immediately to the right of the instructions.

Sun Sentinel reporters talked with a ballot expert, who said that some voters may not have noticed the Senate race (perhaps thinking it was just part of the ballot instructions) and started filling out their ballot with the governor race instead. That theory is supported by a data consultant who’s worked for several political campaigns in Florida, who found that the parts of Broward County that fall in the 24th Congressional District did see higher levels of undervoting than other parts of the county. That might be because the 24th District was uncontested, which according to Florida law means that the congressional race did not appear on the ballot at all. As you can see in the sample ballot above, the congressional race would also appear in the lower-left corner on many ballots, along with the Senate race. In districts where there was no congressional race on the ballot, however, that corner would have looked even emptier, perhaps making it easier for voters to inadvertently skip over the Senate race.

An alternative explanation is that an error with the vote-tabulating machines in Broward County caused them to sometimes not read people’s votes for U.S. Senate. If that’s true, we would probably only find out if there is a manual recount. According to Florida law, any election that’s within half a percentage point (as this one currently is) triggers a machine recount; then, after the machine recount, if the race is within a quarter of a percentage point, it goes to a much more complex manual recount — a.k.a. each ballot is recounted by hand. As long as the machine recount doesn’t change the Senate results too much (barring a surprise in the remaining ballots in Broward and Palm Beach), it looks like that’s where we’re headed. In addition, Republican former Rep. Ron DeSantis and Democratic Tallahassee Mayor Andrew Gillum are separated by just 0.44 points in the governor’s race, so that could go to a machine recount, too.

But recounts rarely change the outcomes of elections. A FairVote analysis found that the average recount from 2000 to 2015 shifted the election margin by an average of just 0.02 percentage points. The largest margin swing was 1,247 votes — coincidentally also coming in Florida, in the 2000 presidential race. If Nelson is going to stage a comeback in the Sunshine State, he’ll almost certainly have to close the gap between him and Scott even more in the next couple of days.

Why The House And Senate Are Moving In Opposite Directions

At first, I was a little skeptical of the narrative that Justice Brett Kavanaugh’s confirmation process could send the House and Senate moving in opposite directions. Usually in politics, a rising tide lifts all boats — so whichever party benefited from the Supreme Court nominee’s confirmation would expect to see its fortunes improve in both its best states and districts and its worst ones.

But a House-Senate split is exactly what we’re seeing in the FiveThirtyEight forecast. Democratic prospects in the Senate are increasingly dire, having fallen to about 1 in 5. Indeed, it’s been hard to find any good news for Democrats in Senate polling lately. In the House, by contrast, their opportunity is holding up relatively well. In fact, Democrats’ chance of taking the House has ticked back upward to about 4 in 5, having improved slightly from around 3 in 4 immediately after Kavanaugh was confirmed. And while district-by-district House polling has been all over the place lately, Democrats’ position has improved slightly on the generic congressional ballot.

On the surface, you might reason that House and Senate battlegrounds aren’t that different from one another. Yes, the most competitive Senate races this year are in really, really red states. Specifically, the average competitive Senate race, weighted by its likelihood of being the decisive state in determining the majority according to FiveThirtyEight’s tipping-point index, is 16 percentage points more Republican than the country overall.1 But the average competitive House district is also pretty red: 8 points more Republican than the country overall, weighed by its tipping-point probability.

The more time you spend looking at the battlegrounds in each chamber, however, the more you’ll come to two important conclusions:

  1. The House and Senate battlegrounds really aren’t that much alike. In several important respects, in fact, they’re almost opposite from one another. For example, House battlegrounds are more educated than the country overall, while Senate ones are less so.
  2. The Democrats’ map in the House is fairly robust, because they aren’t overly reliant on any one type of district. (This stands in contrast to the Senate, where most of the battlegrounds fit into a certain typology: red and rural). While House battlegrounds are somewhat whiter, more suburban and more educated than the country overall, there are quite a few exceptions — enough so that Democrats could underperform in certain types of districts but still have reasonably good chances to win the House. This differs from Hillary Clinton’s position in the Electoral College in 2016, in which underperformance among just one group of voters in one region — white working-class voters in the Midwest — was enough to cost her the election.

So let’s look in more detail at the characteristics of the House and Senate battlegrounds, starting with some basics: their geographic region (as according to the U.S. Census Bureau) and whether they’re incumbent-held or open-seat races. In the series of charts that follow, I’ll show what these characteristics look like in an unweighted average of all 435 congressional districts, and compare that to what’s happening in battlegrounds. (Rather than handpick the “battleground” contests, I’m weighting all races by their tipping-point probabilities in the House and the Senate; contests more likely to prove decisive have outsized sway in the calculation.)

How House and Senate battlegrounds compare based on incumbency and geography

Chance that a race will have a certain characteristic

Characteristic All congressional districts House tipping-point districts* Senate tipping-point states*
Democratic incumbent 40% 6% 46%
Republican incumbent 46 71 29
No incumbent 14 25 25
Characteristic All congressional districts House tipping-point districts Senate tipping-point states
Northeast 18% 19% 2%
Midwest 22 27 32
South 37 27 41
West 23 27 25

*Weighted by tipping-point probability, which is the likelihood that a particular state or district is decisive in determining majority control. Tipping-point determinations are as of 12 p.m. on Oct. 13.

One obvious but overlooked difference between the House and the Senate is that Democratic incumbents have very little exposure in the House but a ton of it in the Senate. For instance, there’s a 46 percent chance that the tipping-point race in the Senate will be one featuring a Democratic incumbent, but only a 6 percent chance for the House. Just the reverse is true for Republicans; they have tons of vulnerable incumbents in the House but few in the Senate. What this means is that, other factors held equal, things that reduce the incumbency advantage will tend to hurt Democrats in the Senate but help them in the House. If, for instance, the Kavanaugh hearings turned the public further against incumbent politicians in both parties, that could contribute to the increasing divergence we’re seeing between the House and Senate forecasts.

Geographically, the competitive House races are almost evenly divided between the four major Census Bureau regions. This is part of what makes Democrats’ map robust: Even if they completely flopped in one of these regions, they could still narrowly win the House by performing to the model’s expectations in the other three. Note, however, that relative to population, the South is slightly underrepresented among the competitive House districts — it represents 37 percent of congressional districts overall but 27 percent of the competitive ones. That’s probably a good thing for Democrats because their coalitions in the South tend to rely on racial minorities and on younger voters — groups that don’t turn out as reliably in midterms as they do in presidential elections.

By contrast, there’s a 41 percent chance that the tipping-point state in the Senate will be in the South. And whereas the Northeast is fairly important to the House map, with a number of competitive races in New York, New Jersey and Pennsylvania, there’s almost nothing interesting happening in the Northeast on the Senate side of the ledger.

Next, here are the political characteristics of competitive districts, based on FiveThirtyEight’s partisan lean index and on whom they voted for in 2012 and 2016. (To make for an apples-to-apples comparison between the House and the Senate, the statistics for Senate races in these next two charts are based on congressional districts within states with competitive Senate races rather than the states overall. For instance, if Tennessee has an 11 percent chance of being the tipping-point state in the Senate, that 11 percent chance is divided between the nine congressional districts in Tennessee to estimate how important individual congressional districts are toward determining Senate control.)

How House and Senate battlegrounds compare based on recent voting patterns

Chance that a race will have a certain characteristic

Characteristic All congressional districts House tipping-point districts* Senate tipping-point districts*
Obama-Clinton 39% 15% 27%
Obama-Trump 5 17 4
Romney-Clinton 3 17 2
Romney-Trump 52 51 67
Characteristic All congressional districts House tipping-point districts* Senate tipping-point districts*
Very Democratic† 32% <1% 16%
Competitive 29 82 21
Very Republican† 39 18 63

* Weighted by tipping-point probability, which is the likelihood that a particular state or district is decisive in determining majority control. For the Senate, statistics are compiled on a district-by-district basis but weighted based on the probability that the state will be decisive to Senate control. Tipping-point determinations are as of 12 p.m. on Oct. 13.

† At least 10 points more Democratic or Republican than the country overall.

It’s true that House battlegrounds are Republican-leaning — but for the most part, they’re Republican-leaning and not much more than that. Only 18 percent of competitive House races have a partisan lean of R+10 or better for the GOP, for instance. By contrast, weighted by their importance to determining the outcome of the Senate, 63 percent of competitive districts are at least R+10. Romney-Clinton and Obama-Trump districts — that is, districts that split their vote between the past two presidential elections — are quite important in the House but not really a factor in the Senate.

Finally, here’s what competitive districts look like based on three characteristics that are increasingly predictive of voting behavior: race, education and urbanization.

How House and Senate battlegrounds compare based on key demographics and urbanization

Chance that a race will have a certain characteristic

Characteristic All congressional districts House tipping-point districts* Senate tipping-point districts*
High education (>=35% bachelor’s degrees) 28% 37% 13%
Medium education 35 41 43
Low education (<=25% bachelor’s degrees) 37 22 45
Characteristic All congressional districts House tipping-point districts* Senate tipping-point districts*
<50% non-Hispianic white 29% 18% 24%
50-80% non-Hispanic white 48 51 38
>80% non-Hispanic white 23 30 38
Characteristic All congressional districts House tipping-point districts* Senate tipping-point districts*
Urban (>2500 people per square mile) 21% 8% 14%
Suburban (500-2500 people per square mile) 23 29 14
Exurban (100-500 people per square mile) 31 39 26
Rural (<100 people per square mile) 25 24 46

*Weighted by tipping-point probability, which is the likelihood that a particular state or district is decisive in determining majority control. For the Senate, statistics are compiled on a district-by-district basis but weighted based on the probability that the state will be decisive to Senate control. Tipping-point determinations are as of 12 p.m. on Oct. 13.

I’ve defined high-education districts as those where at least 35 percent of the adult population has a bachelor’s degree. By that measure, 28 percent of congressional districts in the country overall are highly-educated. But 37 percent of the House battleground districts are highly-educated. By contrast, only 13 percent of the districts that are most important to Senate control have high education levels.

Racially, both House and Senate battlegrounds are whiter than the country as a whole, but the skew is worse in the Senate: 38 percent of districts most important to Senate control are at least 80 percent non-Hispanic white, compared to just 23 percent of districts in the country overall.

Lastly, competitive House districts are concentrated in the suburbs and exurbs (defined based on population density), which make up about half of congressional districts overall but closer to two-thirds of competitive ones in the House. Suburban and exurban districts are quite unimportant to the Senate, however, where the key races are disproportionately based on rural states or in states such as Nevada with a mix of urban and rural areas with little in between.

I’m not going to go on for too much longer because I hope those numbers really speak for themselves. I suppose I had some intuition about how different House and Senate battlegrounds are different from one another — but that intuition didn’t match how sharp the differences are in the data.