In this article, I will be discussing my predictions for the 2018 gubernatorial elections, the reason behind them, and noticeable flaws in my forecast.
So, what am I predicting will happen? With tossups, no party has a clear advantage. Democrats are favored to flip many seats, but Republicans are also expected to win in some battleground states and Democratic states. If every tossup went the way they were leaning (which is almost certain not to happen), the Democrats would control 25 state governorships, Republicans would control 24, and there would be 1 Independent Governor from Alaska who caucuses with the Democrats. In short, control over Governorships would be very close, possibly closer than control over the United States Senate.
Map with Tossups:
Map without Tossups:
- Democratic lean in state’s previous Governor election – 50% weight
- Democratic lean in the 2016 Presidential election – 15% weight
- Democratic lean in the generic ballot – 25% weight
- Disapproval lean in Trump’s approval rating – 10% weight
- Senators’ parties – 2.5 points towards each Senator’s party, giving either one party 5 points or canceling out split Senators
- Incumbency – 5 points in the direction of the incumbent
I wanted to be sure to factor in split voting, meaning that a state who votes for a Democrat in the Presidential race might vote heavily for a popular Republican governor. That’s why I gave the last Governor election a 50% weighting, while only giving the 2016 Presidential election a 15% weighting. Also, I wanted to encompass the solid views of a state’s population over time, which is why I gave points for Senators. Using only one or two elections can be dangerous, because a flawed candidate or extremely popular candidate may sway results. However, there are two Senators, each elected at a different time, ensuring the stability and accuracy of voter preference.
Once I had my partisan lean, I used the following system to give ratings:
- Within 5% – Tossup
- Within 10% – Lean
- Within 17% – Likely
- Outside 17% – Safe
I noticed a few flaws and abnormalities under my system that I, while wanting to fix them, had to leave to assure standard practice across all states and reduce my personal bias. I will be disclosing those flaws here:
- I have no true measure for Governor favorability. Some of my reasoning in giving past Governor elections 50% was so that it could include favorability in candidates. However, for states with popular candidates that aren’t running for reelection, this can sway partisan lean. For example, in Nevada, Brian Sandoval won with 70.6%, well above the Republican average for such a swing state like Nevada; but, Sandoval isn’t running again in 2018, so while he election margin plays a 50% role in my prediction, it has virtually zero impact in the 2018 elections because a new candidate will be on the ballot. I’m going to try to fix this, so stay posted for an updated predictions article.
- I have no measure for ideological lean. Many Republican governors in blue states are liberal, and many Democratic governors in red states are conservative. I have no measure for this and thus must assume that the Governor’s ideology is congruent with their party’s ideology. For example, Republican Phil Scott of Vermont is probably going to win reelection. However, Vermont’s partisan results come out as lean Democrat. This is because Scott didn’t get a large enough majority in 2016 to overcome the state’s Presidential election results, the generic ballot, Trump’s approval, and the state’s two Democratic Senators. If another Republican were running, my rating would be more accurate. However, with a liberal Republican like Phil Scott, my Vermont rating just doesn’t seem fit. Again, stay posted for an update to my predictions.
For the most part, the flaws counteracted themselves. Nevada, a swing state if not leaning blue, was safe Republican, while Vermont, a likely red state in this scenario, was leaning Democrat. My predictions are still accurate, but just remember that leaning one way or being “likely” one candidate is not a sure thing.