There are a lot of mea culpas floating around from people who thought it would snow in July — in Miami — before Donald Trump became the Republican nominee for president. I, to take one example, was wrong on Trump. (Although, strictly speaking, I never mentioned snow in Miami; no, I said “Trump has a better chance of cameoing in another “Home Alone” movie with Macaulay Culkin — or playing in the NBA Finals — than winning the Republican nomination.”) I wrote a pre-emptive mea culpa back in December, when it was already clear that my initial skepticism of Trump was overconfident. But anytime something so unexpected happens, it’s worth stepping back and thinking about what we should learn. Here are four initial thoughts:
1. Don’t rule out the ahistorical when there’s little history.
My skepticism of Trump boiled down to this: No party had ever nominated someone like him. In the modern era, parties have tended to nominate ideologically reliable candidates who are also electable — for Republicans, electable conservatives. Trump seemed neither electable nor all that conservative.
Indeed, Trump’s résumé didn’t resemble those of recent nominees of either party. Since nominees began to be selected mostly by caucuses and primaries in 1972, no major-party candidate without elected office experience had won. Not since Wendell Willkie in 1940 has a party nominated a candidate who wasn’t either a politician or a war hero.
The problem, of course, is that this data set is incredibly small. There have been only 14 party primaries without an incumbent president running for re-election since 1972. Any pattern that appears in such a small data set may have occurred only by chance. At the very least, the relationship may not be as strong as it appears to be.
2. Take a nuanced view of the polls.
I focused on a lot of data in the summer, fall and winter of 2015 — before I really started to think Trump had a chance. We had data on Trump’s ideology and his lack of party support. We had data suggesting he’d face long odds in a general election. But I focused too little on polls that clearly showed Trump with a good chance of winning the primary. I should have given them more weight, especially as the Iowa caucuses neared.
Trump led in the vast majority of polls. I went back and looked at the 549 polls in our national primary polling database taken after Trump entered the race June 16. He led in 500, or 91 percent.1 More than that, Trump jumped into the lead very quickly. He led in 75 percent of the polls taken in July, and it only climbed from there. As Natalie Jackson, Ariel Edwards-Levy and Janie Velencia wrote Thursday at The Huffington Post, “A Trump nomination shouldn’t be a surprise based on polls.”
The same thing goes with state-level polls. Although Trump struggled in Iowa and ultimately lost there, he led in 83 of the 86 New Hampshire surveys conducted after he declared his candidacy. Trump was ahead in 43 of the 45 polls taken in South Carolina after he jumped into the race. Even in Jeb Bush and Marco Rubio’s home state, Florida, which Trump used to exorcise Rubio from the race, Trump was ahead in 49 of the 50 polls with field dates entirely after his entry.
Why did I ignore these polls? Early polls, even a month before a primary election, haven’t been very predictive. The candidates typically have varying levels of name recognition; most voters don’t make up their mind until right before they cast a ballot. We all remember President Howard Dean, right? Yet, there were plenty of other candidates who led in the polling throughout and won, such as George W. Bush in 2000. I still think it was right to be highly skeptical of the polls in the summer and fall of 2015. But as Trump maintained and even expanded his lead into 2016, I should have been quicker to give them more weight than I gave to the lack of a precedent for Trump. (I did do this eventually; it just took too long.)
3. Maybe favorability ratings aren’t as hard to change as we thought.
I didn’t ignore the polls completely. Instead, I focused on Trump’s low favorability ratings. That wasn’t a mistake. The error was in not recognizing that favorability numbers can change as much as horse race numbers — even for a well-known politician — and that a candidate in a large field may need only a base of strong supporters to win.
Trump had very negative net favorability ratings before he entered the race, scoring a -35 percentage point net favorability rating among Republicans in a Monmouth University survey in June. Usually, that would be a death sentence for a candidate. Trump, though, was able to swing that quickly and scored a +17 percentage point net favorability rating in an ABC News/Washington Post poll the next month.
Trump’s strength came from a core of supporters who viewed him “very favorably.” Consider that ABC News/Washington Post poll. Trump had a net favorability rating 16 percentage points lower than Jeb Bush’s. But Trump also had a “strongly favorable” rating of 30 percent, higher than Bush’s 19 percent. Trump’s supporters were a solid bloc, while opinions were far softer for the other candidates. Trump’s strongly favorable rating really didn’t move over the course of the campaign. Even by September, when his net favorability rose, his strongly favorable rating was still 31 percent in the ABC News/Washington Post poll. When his net favorability dropped by early April, his strongly favorable rating barely moved.
4. Don’t assume the party knows what it’s doing.
Political parties have been fairly powerful institutions. And I thought Republican elites — party operatives, elected officials, grass-roots activists — would work tirelessly (and efficiently) to stop Trump. You could see Trump’s lack of party support in the endorsement race. Trump received very few endorsements from local and national political figures, and most of the ones he got came well after the primaries started. Historically, that hasn’t been characteristic of winning candidates. In fact, Trump is the first candidate since at least 1980 to win the nomination of either party while having fewer endorsements than another candidate on the day he became the presumptive nominee.
The problem was that the party didn’t or was unable to coordinate on whom to support as an alternative to Trump. Even when there seemed to be some rallying to Marco Rubio in the late winter, most major party officials were staying on the sideline. We saw this most clearly with Ted Cruz — very few major officials rallied to his cause. Maybe in the year of the “anti-establishment,” this didn’t matter. After all, state legislators were more likely to endorse Cruz than national officials. Either way, voters went in a different direction.
Still, we don’t truly know what the 2016 Republican results mean for the idea that party elites can influence presidential primaries (or as it’s shorthanded, the theory that “the party decides”). Hillary Clinton had an early, commanding lead in endorsements on the Democratic side, and she seems to be on a glide path to the nomination. Again, there have simply been too few primary elections to make any confident statements one way or the other.
Listen to the latest episode of the FiveThirtyEight politics podcast.
Bonus: Trump!
Let’s give some credit to Trump himself! No, I don’t think that Trump is a strategic and tactical mastermind who planned every move he made, or even that every move was successful. On the whole, though, more of what he did worked than didn’t work. Trump generated a ton of free media coverage; that helped him. He was willing to challenge Republican orthodoxy; that, at the very least, didn’t hurt him. I don’t know whether he’s built a new political coalition or the Trump phenomenon is sui generis, but whatever the guy did, it worked.
Whatever happens with Trump in the general election, it was a humbling experience for the guy writing these words to you now. The lesson I take from it is you shouldn’t dismiss polling data, even when it doesn’t line up with your priors. That’s especially the case when your priors are informed by a small number of elections.