Nate always gets it right. I'll post it:
Twelve years ago, in August 2003, Joe Lieberman led in
most polls of the Democratic primary. Eight years ago, in August 2007, Rudy Giuliani
maintained a clear lead in polls of Republicans, while Hillary Clinton
led in polls of the Democratic nomination contest. Four years ago, in August 2011, Mitt Romney began with the lead in polls of Republican voters, but he would be
surpassed by the end of the month by Rick Perry, the first of four Republican rivals who would at some point overtake Romney in national polling averages.
Lieberman, Clinton, Giuliani and Perry, as you’ve probably gathered, are not the faces atop Mount Rushmore. Only Clinton came close to winning the nomination.
But the problem isn’t just that the national polls at this stage in the race lack empirical power to predict the nomination; it’s also that they describe a fiction. I don’t mean to suggest that Donald Trump’s support in the polls is “fake.” I have no doubt that some people really love him or that he’d be the favorite if you held a national, winner-take-all Republican primary tomorrow. However, the “election” these polls describe is hypothetical in at least five ways:
- They contemplate a vote today, but we’re currently 174 days from the Iowa caucuses.
- They contemplate a national primary, but states vote one at a time or in small groups.
- They contemplate a race with 17 candidates, but several candidates will drop out before Iowa and several more will drop out before the other states vote.
- They contemplate1 a winner-take-all vote, but most states are not winner-take-all.
- They contemplate a vote among all Republican-leaning registered voters or adults, but in fact only a small fraction of them will turn out for primaries and caucuses.
This is why it’s exasperating that the mainstream media has become obsessed with how Trump is performing in these polls.
So you should ignore those national polls entirely? In a literal sense, they do
have some correlation with election outcomes: Even this far out, a candidate near the top of the polls is a somewhat better bet to win the nomination than one near the bottom. But that’s like projecting a major league pitcher’s numbers from high school stats: Sure, you’d rather draft a random 17-year-old with a 2.14 ERA than another one with a 3.31 ERA if that’s all the information you have to go by. But that data doesn’t reveal very much, and its predictive power tends to be swamped by other indicators (everything from the pitcher’s strikeout-to-walk ratio to his scouting reports).
In the case of presidential primaries,
indicators such as
endorsements and support from party elites tend to be more reliable indicators of eventual success. To the
extent that you’re looking at polls, you should probably adjust for
name recognition and the
amount of media attention a candidate is receiving. And you should account for favorability numbers and second-choice preferences, since all but a few candidates will eventually drop out of the running.
It’s possible — pretty easy, in fact — for a candidate to improve his standing in the polls while he simultaneously lowers his chance to become the nominee. Currently, the average GOP voter
has a favorable view of seven Republican candidates; being agreeable won’t help you stand out in the near term, even though the nomination is a
consensus-building process in the long term.
What about being a jerk? If you can make yourself the center of attention — and no candidate in modern memory has been more
skilled at that than Trump — you can potentially turn the polls into a referendum on your candidacy. It’s possible that many GOP voters are thinking about the race in just that way now. First, they ask themselves whether they would vote for Trump; if not, they then choose among the 16 other candidates. The neat thing about this is that you can overwhelmingly lose the majority in the referendum — 75 percent of Republicans are
not voting for Trump — and yet still hold the plurality so long as the “no” vote is divided among a sufficient number of alternatives.
Another trade-off comes from entrenching your appeal with a narrow segment of the electorate at the expense of broadening your coalition. I’ve seen a lot written about how Trump’s candidacy heralds a new type of populism. If it does, this type of populism isn’t actually very popular. Trump’s overall favorability ratings
2 are
miserable, about 30 percent favorable and 60 percent unfavorable, and they haven’t improved (whatever gains he’s made among Republicans have been offset by his declines among independents and Democrats). To some extent, the 30 percent may like Trump
precisely because they know the 60 percent don’t like him. More power to the 30 percent: I have plenty of my own issues with the political establishment. But running a campaign that caters to (for lack of a better term) contrarians is exactly how you ensure that you’ll never reach a majority.
3
At FiveThirtyEight, however, we’re fairly agnostic about what will happen to Trump’s polling in the near term. It’s possible that he’s already peaked — or that he’ll hold his support all the way through Iowa and New Hampshire, possibly even winning one or two early states, as similar candidates like Pat Buchanan and Newt Gingrich have in the past.
4 Our
emphatic prediction is simply that Trump will not win the nomination. It’s not even clear that he’s trying to do so.
