Skip to main content
Menu
What Will Virginia And New Jersey Tell Us About Trump?

No offense to the 8.4 million people who live there, but I really don’t care who the governor of Virginia is. I don’t care much about New Jersey’s either.1 I do care a lot about national politics, however, so I’m interested in what next Tuesday’s gubernatorial elections in Virginia and New Jersey will tell us about the national political environment. Should a Republican win in Virginia — where polls have been inconsistent but show only a 3 or 4-percentage-point lead for Democrat Ralph Northam, on average — be cause for Democratic panic? Should Democrats get any credit for a double-digit win in New Jersey (as seems likely based on polls there), or would that just be par for the course?

These questions are tricky. On the one hand, one needs to be careful about finding national lessons in state and local elections (including gubernatorial races). The relationship between local and national politics can be rough: Consider that Massachusetts and Maryland, which are reliably blue in presidential elections, have Republican governors right now, while solidly red Louisiana and Montana have Democratic ones.

On the other hand, while individual governors’ races can be quirky, gubernatorial races on the whole usually do a fairly good job of reflecting the national environment. For example, Republicans turned in excellent performances in gubernatorial races in 2010 and 2014, when they also had great midterm wave years. Democrats did the same in 2006.

In theory, that dynamic should carry over to this year. Because the overall political environment is seemingly good for Democrats, they should expect to perform well in most governors races this year and next, outside of extremely red states. If they don’t, it will be reasonable to ask whether the results in particular states were flukes — or instead, whether the national political climate wasn’t as toxic for President Trump and Republicans as it had seemed.

Let’s look at some data to see how gubernatorial races have behaved historically. I’ve compiled a database of all gubernatorial races since 2001 in which there was no incumbent running (both Virginia and New Jersey are open-seat races this year), excluding races where a third-party candidate finished first or second, or when multiple candidates from the same party appeared on the ballot together.2 This works out to a total of 90 elections.

In those elections, there was a comparatively poor relationship between gubernatorial and presidential voting. In the chart below, I’ve compared the FiveThirtyEight version of a state’s partisan voting index (PVI) — how it voted in the previous two presidential elections, relative to the national average3 — to the margin of victory or defeat for the Republican candidate in open-seat gubernatorial races. A state’s PVI mispredicted the gubernatorial result in 36 of 90 cases, meaning that a state that leans Democratic in presidential elections voted Republican for governor, or vice versa, 40 percent of the time. There certainly is some correlation,4 but it’s rough. The relationship would be much stronger if you’d run the same analysis between presidential races and congressional races instead.

On the other hand, as I mentioned, gubernatorial races usually do a pretty good job of reflecting the national mood when you aggregate enough of them together. In 2006, for example, Democrats won the average open-seat gubernatorial race by almost 10 percentage points, close to their 8-point victory in the total popular vote for the U.S. House that year. Four years later, in 2010, it was Republicans’ turn for a wave. They won the average open-seat gubernatorial race by 8 points, similar to their 7-point margin in the House popular vote.

Gubernatorial races can get swept up in political waves
YEAR NO. OF OPEN SEAT RACES* AVG. PARTISAN MARGIN
2016 7 R +7.7
2014 8 R +6.6
2012 5 D +0.5
2010 21 R +8.0
2006 10 D +9.9
2004 5 D +2.9
2002 20 R +1.2

* Excluding races where an independent candidate finished first or second, or when multiple candidates from the same party appeared on the ballot together.

Source: Dave Leip’s U.S. Election Atlas

So what does that mean for Virginia and New Jersey? Using these past gubernatorial races, I built a regression model to estimate gubernatorial results based on the generic congressional ballot — which currently favors Democrats by 9 percentage points according to our tracker — and a state’s PVI. You can think of this as a “fundamentals”-based prediction: That is, what you’d think “should” happen in each race based only on its partisan lean and the overall political environment, without looking at who’s running or the polls in each state.

That regression model says that Democrats “should” win by 9 points in Virginia, and by 13 points in New Jersey, given that the generic ballot suggests we’re in a strongly Democratic-leaning environment and that both states are bluer than the national average (New Jersey more so than Virginia). In New Jersey, that projection pretty closely matches the polling average, which has Democrat Phil Murphy ahead of Republican Kim Guadagno by 15 to 16 percentage points. Virginia is somewhat tighter than the fundamentals project, however, with Northam leading Republican Ed Gillespie by 3 to 4 points.

But note that this fundamentals-based projection isn’t very precise: Historically, it’s missed the final result in each state by an average of about 10 points.5 To put it another way, there’s a lot of room for error: Republicans would still win under these conditions in Virginia 23 percent of the time, and in New Jersey, 14 percent of the time, despite their seeming disadvantages. Individual gubernatorial races just aren’t all that predictable.

Therefore, the challenge of interpreting Tuesday’s results is that there aren’t many of them. While gubernatorial races can tell you a lot once you have enough examples, only two states are voting on Tuesday — and individual gubernatorial races can be quirky. Still, even accounting for the low predictability of gubernatorial races and the wide spread in the polls in Virginia — which have ranged from showing a 17-point Northam lead to an 8-point Gillespie lead — Democrats probably ought to be winning both races. If Northam loses, it will be hard to know whether that says something about Northam and about Virginia — or instead about the national political environment. But either way, it won’t be a great sign for Democrats.

At the same time, if Northam wins, I don’t think analysts should be going out of their way to parse the meaning of his margin of victory (unless perhaps it’s very close or a massive blowout). Races for governor are a noisy signal, and anything from about a 2-point Northam win to a 15-point win would qualify as a pretty “normal” result under the circumstances.

New Jersey shouldn’t be overlooked, either. Gubernatorial races are rarely “gimmes,” so Democrats probably should get some credit for a win there, provided that it comes by a solid margin like the one polls project. New Jersey isn’t the best data point — you’d rather have a state where the candidates were starting off with more of a blank slate and not one where the outgoing Republican governor, Chris Christie, was so historically unpopular. (Guadagno is Christie’s lieutenant governor.) Nonetheless, political analysts and reporters will surely go out of their way to draw implications from Virginia, despite it being just one data point. Do yourself a favor and double the sample size by incorporating New Jersey into the story. That should put your own Wednesday morning narrative on slightly firmer ground.

Footnotes

  1. My interest in New Jersey is slightly higher, given that I live in neighboring New York.

  2. That means I excluded all Louisiana “jungle primaries,” and the 2003 California gubernatorial recall election; however, I included the Louisiana runoff elections in 2003 and 2015.

  3. In calculating PVI, FiveThirtyEight assigns a weight of 3 to the most recent presidential election, and a weight of one to the next-most recent one.

  4. Specifically, the correlation coefficient is .48

  5. By comparison, polling averages miss the final result in governors races by an average of about 4 points, according to the FiveThirtyEight polling database.

Nate Silver is the founder and editor in chief of FiveThirtyEight.

Comments