This survey, though, is a veritable salt mine, offering enough grains of the stuff to keep skeptics' blood pressure high at least through November.
Let's walk through them quickly:
The crossover vote reported by the surveying firm, Mercury Analytics, works both ways. In their estimation, 14 percent of Republicans go for Hillary Clinton.
Sure, the alleged crossover for Trump is bigger, but it's not like Mercury is saying that Trump peeled away a fifth of Democrats and no Republicans went for Clinton.
Which brings us to…
2. What is this survey, anyway?
Here's how Mercury describes its methodology on its own site:
Mercury Analytics, a technology-driven consumer research and political polling firm based in Washington, D.C. conducted a survey and ad “dial-test” of Trump’s first campaign spot among a national sample of n=916 “likely voters”.
So they showed voters Trump's campaign ad and then asked some questions, apparently. It's really not clear how it worked, and there isn't much data available on the site. The questions are referred to obliquely. There are only isolated bits of data. What's more, the survey was conducted online, which we've discussed before.
The kicker, though, is this.
3. That survey still has Clinton winning.
Mercury buries this detail a bit, writing that "Hillary Clinton has the edge, but would be facing a very tough competitor." "Has the edge" means "is leading" (though Mercury doesn't say by how much.) What it suggests, though, is that this 20 percent crossover wouldn't make much difference.
4. Polls from established pollsters show much less crossover.
Two recent surveys from traditional pollsters show partisan crossover — but at much smaller levels. CNN/ORC polled in late December, and Fox News released a poll on Friday. In CNN's survey, 11 percent of Democrats go for Trump and 9 percent of Republicans go for Clinton. In Fox's, 9 and 8 percent switch teams.
These are traditional, live-dial surveys from pollsters that you have heard of before. That's worth keeping in mind.
5. Ignore head-to-head polling!
We have to reiterate that polls pitting a Democrat against a Republican are still way too early to be predictive. Both parties are seeing contested primaries in which partisan loyalty is split. Recent history shows that once the campaign gets down to one Democrat and one Republican, party loyalty kicks in. What's more, there has been no campaign yet! Trying to predict the winner in November right now is like trying to predict a horse race right after seeing two colts be born.
Sure, Trump is a unique force, as has been well established. And maybe he will peel 20 percent of Democrats away. Who knows?
But in a moment with a lot of weird things happening, it's worth spending more time being skeptical, not less. Leaping to conclusions about a nebulous detail from a question about one ad is a lot of fun. Look before you leap.