For Dems who may be worrying excessively about Trump’s very modest uptick in his approval ratings and the 3.9 percent unemployment rate for April, The Economist magazine has unveiled a wonky new “prediction model for America’s mid-term elections,” with some encouraging data for Democrats. The methodology article is replete with jazzy graphs and is subtitled “Our model estimates both the national political climate and the nuances of each district.” Credit The Economist with putting a lot of thought and effort into their model. In one paragraph, the author(s) explain,
With just 38 House elections in the dataset, there are not enough historical examples available to tease out the individual impact of each of these variables with all the others held constant—particularly since many of them tend to point in the same direction in any given year. However, for the purposes of prediction, we did not need to know exactly how much each type of information matters when taken in isolation. Instead, we can be satisfied simply by determining which composite blend of all the ingredients yielded the most accurate forecasts when presented with data about an election that had not been used to train our model.
Got that? The methodology article gets even more impenetrable for lowly laypersons who prefer cutting to the chase over nuanced statistical analysis. But The Economist is also providing daily updates based on the model, and it will be interesting to see how they perform in the closing days of the 2018 elections. You’ll have to read the article to see the lovely charts, But yesterday, The Economist gave Dems an “around 2 in 3” chance of winning a House of Reps majority. “On average, we expect the Democrats to win a 9-seat majority,” a ballpark 222-213 advantage. That’s better than Sabato’s Crystal Ball analyst Kyle Kondik, who sees “a coin flip battle for the House” in his latest update.
As for the number of House seats in play:
Although every seat is up for election, the real battle for control of the House is fought in a much smaller array of seats. We rate 278 seats, about two-thirds of them, as “safe”—where one party has a better than 99% chance to win. Another 87 are rated as “solid” (with a 90-99% chance to win). That leaves the remaining 70 seats to determine which party will control the House.
Not that it tells us much about the House lineup after the elections, but the model sees Democrats getting 54.1 percent of midterm votes, compared to 45.9 percent for Republicans, who get a hefty helping hand from gerrymandering. The Economist notes that “In 2014, for example, the Democrats won a House seat for every 189,000 votes they received. The Republicans, by contrast, won a seat for every 162,000.”
Further, “In order to be favoured to win the House our model thinks the Democrats must win the two-party national popular vote by about 6.9 points. If as many people vote in 2018 as did in 2014, that would mean the Democrats need to win 5.2 million more votes than the Republicans.” The daily update notes in closing that Democrats lead in “the best measure of how the election is playing out across the country,” the generic ballot by 43-37.
No doubt there are plenty of factors that don’t fit so well into such predictive models, especially unexpected events that impact national security, like a terrorist attack, or a sudden market crash or natural disaster. And no model will change the political reality that Democrats have to improve their recruitment, training and support of candidates, which is a huge challenge, especially with an unprecedented number of candidates running nationwide in 2018.