The presidential campaign begins in earnest this week as 20 Democrats debate each other in Miami over who should take on President Donald Trump in next year’s election.
Is it possible to predict the 2020 election before a single primary vote is even cast?
Disregarding Yogi Berra’s sage advice to “never make predictions, especially about the future,” here’s how a political scientist would do it.
First, I would explain that contrary to popular belief the polls in 2016 were pretty accurate. Nine of the ten major tracking polls predicted Clinton would win the popular vote, which she did by 3 million votes. Eight predicted her vote share within their margins of error.
The polls also accurately predicted the Electoral College in all but one state, Wisconsin. In six states, polls were too close to call, so they were “toss-ups.” Five of these ended up voting Trump. But the election turned on just three states — Wisconsin, Michigan and Pennsylvania — where the combined margin for Trump was about 86,000 votes, less than half of 1%.
The lesson from 2016 isn’t to doubt the polls. It’s when polls tell you it’s too close to call, you should believe them!
So, what about 2020? Political scientists have developed a variety of econometric models for predicting presidential elections without knowing anything about the candidates. One of the best is Alan Abramowitz’s from Emory University. Using the incumbent president’s net approval/disapproval rating and GDP figures the summer before the election, he has accurately predicted every election back to 1948.
Using this model, if Trump’s net approval/disapproval remains where it is today (about -10 points), Trump should be reelected if GDP growth next summer is above 2%. If GDP growth falls below that, or Trump’s disapproval rating gets worse, he is in jeopardy.
So, this looks good for Trump.
There are, however, plenty of good reasons to ignore statistical models in 2020. First, Trump is not like other incumbent presidents. He lost the popular vote by a large margin in 2016, and his approval rating has never been above water during his presidency. At the midterm his approval stood at 37%, substantially lower than Obama’s (50%), Bush II (58%), Clinton (47%), Bush I (64%) and Reagan (60%). Moreover, Trump’s poor approval is despite a strong economy and low unemployment. Who knows what will happen if the economy falters or as more scandals unfold?
A better approach to 2020 is to consider whether the U.S. is going through a political realignment. That’s where existing party coalitions grow unstable, and voters shift their allegiances from one party to another. It’s clear the Reagan coalition is gone — Trump’s Republican Party is nothing like Reagan’s. The electoral fortunes of today’s GOP increasingly depend on a shrinking base of older, whiter, male voters.
These trends have been building for some time, but they accelerated under Trump. According to exit polls from the midterm election, voters under the age of 45 were twice more likely to vote for Democratic candidates than Republicans, while those over 45 split their vote evenly between the parties. Men chose Republican candidates by a 4-point margin, but women voted Democratic by a 19-point margin. And while white voters favored Republicans by a 10-point margin, nonwhite voters overwhelmingly supported Democrats — African Americans by a 90-point margin, Hispanics by 40 and Asians by 54.
The most important predictor of future voting — partisan identification — is also shifting. According to the Pew Research Center, 50% of Americans today identify themselves as Democrat or lean-Democrat, while 42% are Republican or lean-Republican. That 8-point Democratic advantage is wider than it’s been in a decade, and twice as large as in 2016. The gap is greatest among groups with growing demographics, such as millennials, 59% to 32%; Hispanics, 63% to 28%; Asians, 65% to 27%. Among African Americans it’s 84% to 8%, and among women it’s 56% to 37%.
What does this means for 2020? Regardless of who the Democrats nominate, turnout will probably determine who wins the election.
Republicans do well in low-turnout elections because their older, whiter base votes more dependably and because our electoral institutions — especially the Electoral College — favor rural states. Democrats do well in high-turnout elections because, well, more Americans are Democrats, and that percentage is growing.
Democrats have won the popular vote in six of the last seven presidential elections (when turnout is higher) since 1992 but lost seats in all but two midterm elections (when turnout is lower). Democrats did well in 2018 because turnout was the highest for a midterm election in more than a century, 50.3%.
This bodes well for Democrats in 2020. The electorate appears motivated and eager to vote. Of course, there could be a repeat of 2016, when the Democratic candidate wins the popular vote but loses the Electoral College. But that is statistically unlikely to happen again.
Many things can influence turnout — the quality of candidates, campaign strategies, even the weather. But voters have a choice and turnout is key. If you choose not to vote, know that others will.