We’re now just 100 days away from the midterms, and the big question on everyone’s mind is whether Democrats will net-gain 23 seats and gain control of the House of Representatives.
A look at five different metrics suggests that Democrats are favored to take back the House, though each of them suggest a different level of confidence.
The national indicators suggest Democrats are in a good position, while a race-by-race examination finds them in an even better one. In all cases though, the margin of error is wide enough so that Republicans have a real chance of maintaining control.
1. The President’s approval rating
President Donald Trump’s approval rating will be the single most important factor in this year’s midterm election. Polls indicate that those who approve of the president’s job are almost uniformly voting Republican, while those disapproving are almost uniformly voting Democratic.
The latest Gallup weekly tracking poll gives Trump just a 42% approval rating. In the post-World War II era, there have been only a few presidents with an approval rating this low 100 days out from a midterm election. They include Richard Nixon in 1974, Jimmy Carter in 1978, Bill Clinton in 1994, George W. Bush in 2006 and Barack Obama in 2014. (Others such as Harry Truman in 1946 and Obama in 2010 had approval ratings of 43% and 45% respectively.) Almost all of these presidents saw their party struggle mightily in the midterms.
Add CNN's 2018 elections calendar
Indeed, a 42% approval rating at this point suggests that the president’s party is likely to lose the House. It points to Democrats winning the House popular vote by high single digits. Now, Democrats probably need to do better than usual in 2018 to win the majority given how the congressional district lines are drawn. Still, a win in the high single digits would be consistent with around a 30-seat gain for the Democrats.
Keep in mind, the margin of error of this estimate is high. Jimmy Carter’s Democratic Party, for example, only lost 15 seats in the 1978 midterm elections, even though he had an approval rating of 39% 100 days before the election.
2. The generic congressional ballot
Democrats have constantly led on this question which asks some variant of “in your congressional district, will you vote for the Democratic or Republican candidate.” Over the last few months, that advantage has held steady at around 8 percentage points. That’s again good news for the Democratic Party.
The generic congressional ballot tends to move in predictable ways. Large leads tend to get smaller and the president’s party tends to lose ground. This year those forces are competing against each other and will probably cancel each other out. A look at all generic ballots compared to the House popular vote results since 1942 suggests Democrats will probably win the House popular vote by around 8 percentage points.
This would be close to the best performance for a minority party since at least 1942. If Democrats win the House popular vote by exactly 8 points, it would probably translate to a seat gain in the high 20s. That would be good enough for a Democratic majority.
The problem for the Democrats is this estimate comes with a wide margin of error. A popular vote win for Democrats in the low single digits would be inline with history, but that would not be enough for them to take back the House.
3. Race ratings
All race raters (CNN, Cook Political Report and Inside Elections) are in agreement that Republicans have a lot more vulnerable House seats than Democrats do. The latest CNN ratings, for example, have 36 Republican seats rated as toss-up or leaning towards the Democrats, while only 3 Democratic seats are rated as toss-up or leaning towards the Republicans. That’s a difference of 33.
This is very bad news for the Republicans per a model inspired by political scientist James Campbell. When one party has a lot more seats at risk even at this early point, they usually lose a lot of seats. Looking at the Cook Political Report ratings since 1984 (with the exceptions of 1986 and 1990 for which we lack data) in the late summer, the 33-seat difference between the number of seats Democrats and Republicans forecasts a seat loss of greater than 50 seats. That seems unlikely to happen, but race raters have historically underestimated waves in the early going.
There’s never been a single year in which the Republicans had so many more at-risk seats than the Democrats at this point in the campaign. The last time the Democrats would flip the House (2006), the Cook Political Report had 217 House seats at least leaning Republican at this point. This year it’s just 205.
The perhaps saving grace for Republicans is that this model is based off of only 15 elections. It’s plausible that Democrats will fall short of a majority, even with this rosy forecast.
4. Partisan district level polling
Once September and October roll around, we’ll be able to check in more frequently on individual districts because we’ll have district-level polling. For now, district level polling is mostly sparse and cannot give us too much additional information about individual races.
But as I pointed out a few weeks ago, district level polling can help in other ways. Much of it is put out by partisan outfits (either left-wing or right-wing). Although left-wing groups tend to release more polling in a neutral environment, when the difference is really great it suggests that one side is in trouble. The reason being that if one side gets a result it likes it’s more likely to publish it. If Democrats are publishing a lot more polls than Republicans, they’re receiving a lot of good news.
This year about 90% of the about 50 partisan House polls in the FiveThirtyEight database have been from left-wing groups. That’s even more lopsided than at this point in the 2006 cycle when left-wing groups were putting out around 85% of all polls.
As with the race ratings, the disparity in partisan district level polls points to a House gain of greater than 40 seats for the Democrats.
Again though, watch the margin of error. With a sample size of election cycles dating back only to 2006, a Democratic net gain of only 10 seats would be within the model’s margin of error.
5. The betting markets
It’s not clear that betting markets are any better at predicting elections than conventional forecasters. They are, however, a good check of the conventional wisdom.
Right now, they are forecasting a Democratic takeover of the House. The website PredictIt suggests that there is a greater than 60% chance of a Democratic takeover with a net seat gain in the high 20s most likely. That’s similar to what the national factors are suggesting, and it gives Republicans a real chance of holding onto power.
What’s so interesting is that betting markets have tended to be conservative when it comes to forecasting waves in years past. In late July 2006, markets gave Republicans around a 55% chance of maintaining control of the House. They would lose 30 seats and control of the House. In late July 2010, markets gave Democrats about 49% of maintaining control of the House. They would lose 63 seats and control of the House.
In other words, the conventional wisdom in 2018, while less bullish than some district level indicators, is more sold on a House flip than it was in either 2006 or 2010.
That could be a sign that a big wave is coming. It could also be the case that forecasters are simply catching on to the political environment earlier.