When you see surveys showing fairly high numbers of Obama-Trump voters, remember that such surveys are unreliable because too many people claim retroactively to have voted for the winning candidate. Once you control for this by re-interviewing voters whose votes were recorded contemporaneously, you find that 1)there weren’t all that many and 2)most of them were Republicans whose votes for Obama were aberrational, and would have been overwhelmingly likely to vote for the Republican nominee irrespective of who either party nominated:
Democracy Fund found a fairly ordinary crossover vote in 2016: 9.2 percent of Obama voters supported Trump and 5.4 percent of Mitt Romney voters supported Clinton. That was a “typical” and unsurprising degree of partisan loyalty. “The 2016 election did not create more instability, in the aggregate, than others,” it reported.
And those Obama voters who did cross to Trump look a lot like Republicans. The AFL-CIO’s Podhorzer analyzed raw data from the Cooperative Congressional Election Study , out in the spring, and found that Obama-Trump voters voted for Republican congressional candidates by a 31-point margin, Republican Senate candidates by a 15-point margin and Republican gubernatorial candidates by a 27-point margin. Their views on immigration and Obamacare also put them solidly in the GOP camp.
“Democratic analysts who are looking to solve the party’s problem by appealing to this small group of Obama-Trump voters are pointing themselves to a group that by and large is a Republican group now,” Podhorzer told me. “The bulk of Obama-Trump voters are not fed-up Democratic voters; they are Republican voters who chose Obama in 2012. As such, few are available in 2018 or 2020.” Democrats should instead appeal broadly to working-class voters, he said.
In 2008, a larger-than-usual number of Republican voters went with Obama during an extraordinary time, when the economy was in free fall and an incumbent Republican president was deeply unpopular. ANES polling found that 17 percent of Obama voters in 2008 had been for George W. Bush in 2004, compared with the 13 percent of Trump voters, the same survey found, who supported Obama at least once. These people aren’t Obama-Trump voters as much as they were Bush-Obama voters.
Trump’s popularity will be a far more important variable in the 2020 election than the identity of the Democratic candidate. The key for the Democrats in 2008 was not Obama per se but a very unpopular incumbent in a bad economy, which created a lot of crossover voters. It’s not surprising that Bush-Obama voters moved back towards Romney and even more moved back for Trump with an incumbent no longer on the ballot. And as they try to get votes at the margin, Democrats are better off trying to mobilize turnout from their own base than trying to attract disaffected Republicans. If Trump is extremely unpopular you’ll get some no matter what and if Trump is reasonably popular they won’t be gettable anyway.