More on that blockbuster CNN poll showing Trump up two

Jazz already wrote on it but it’s the big news of the day so we can squeeze an extra few paragraphs out of it. Ben Shapiro calls the CNN survey the most important poll ever for Trump, for good reason: It’s the first time he’s cracked 41 percent in the four-way race since the conventions. In fact, the 45 percent CNN has for him is the best he’s ever polled in the four-way race. Only once before, during his post-convention bounce, did he reach as high as 44 percent. This isn’t the familiar “Hillary slips while Trump’s numbers stay flat” story that’s explained most of the tightening in the race over the last few weeks. This is Trump actually picking up support. That’s a big deal. And some very bad timing for this new story from Politico about whether Trump might have hit his ceiling.

His favorable rating has improved by a net 12 points since July, making him slightly more popular than Hillary now:


He’s also out to his biggest lead yet on the economy, perennially a key issue for voters:


Of the last four four-way polls taken, three show Trump either tied or in the lead. One is this, one is Rasmussen (which has tended to show better numbers for Trump than the rest of the pack), and the third is that tantalizing IBD poll from Friday. IBD has a good track record and they saw the race dead even at 39. That was part of the “Hillary’s sliding, Trump’s not gaining” trend, but if you’re looking for corroborating evidence that Trump really has pulled even one way or another, that’s it.

Why has the race possibly changed? For starters, the CNN poll was conducted from September 1st to 4th; the FBI released Hillary’s dubious interview answers on the 2nd. There may be a backlash to her — as there was after Comey’s press conference in July — that’s being partially priced into the data here. On top of that, Trump’s somewhat more disciplined campaigning lately may be paying off for him. He did the press conference with Mexico’s president, he’s been making an effort to reach out to black voters — all of that might be softening up some of the reluctant white women and college-educated voters who’ve been holding off on supporting him so far. In fact, his margin with white voters overall in CNN’s data is 19 points, comparable to Romney’s advantage in 2012 and a bare minimum for what he’ll need to have a chance to win.

But. There’s a wrinkle in CNN’s methodology that must be noted:

A total of 1,001 adults were interviewed by telephone nationwide by live interviewers calling both landline and cell phones. Among the entire sample, 28% described themselves as Democrats, 32% described themselves as Republicans, and 40% described themselves as independents or members of another party.

A sample that favors Republicans, let alone by as much as four points, is noteworthy. When CNN last polled this race, in late July, the sample was 28 percent Democrat and 24 percent Republican. HuffPo’s tracker of party identification gives Democrats a seven-point advantage, 35/28. The last time Republicans won the presidency, in 2004, exit polls showed turnout split dead even at 37 each for Republicans and Democrats. A four-point Republican turnout advantage on Election Day would be unprecedented in modern times.

But what about the fact that CNN’s new one moved from registered voters to likely voters? Don’t likely voters tend to be more Republican? Yes indeed — although I’ve blogged multiple polls this summer showing a strange anomaly in which Clinton tends to do better in polls of likely voters, possibly because she’s outperforming Trump among college grads and educated people are more likely to vote. A shift to a likely voter screen isn’t as sure of an explanation for better Republican numbers this year than it has been in the past. This is surprising too in light of other recent polls:


One of the most damaging trends for Trump in recent surveys is that Clinton tends to do a bit better with her own party than he does with his. That killed him, in fact, in CBS’s battleground poll over the weekend: There were simply too many Republicans withholding their votes from him to get him past Clinton in states like Pennsylvania. CNN’s data isn’t finding that at all. Trump is competitive with Clinton in terms of drawing support from his own party and his voters are more enthusiastic than hers. (Which also wasn’t true in the CBS poll). Maybe some #ReluctantTrumpers on the right have finally come home.

…Or maybe this poll is an outlier. To be clear, by noting the sample split above I’m not claiming the poll is “skewed.” (The Twitter pal who brought it to my attention, though, was convinced that CNN deliberately gamed the sample to make this look good for Trump so that it can revert to a more traditional D+4 sample after the debates and launch a “Clinton comeback!” narrative.) It really may be that CNN is picking up a trend towards Trump. But outliers do happen, as a glance at the final Obama/Romney polls in 2012 remind us. I think the CNN poll is the most important one ever for Trump if we get a few more this week showing the four-way race newly tight and Trump busting through his previous ceiling of 41 percent. Until then, I’ll reserve judgment.

Update: Here’s another reason to wonder about CNN’s results:

Hillary led comfortably in Pennsylvania in CBS’s poll this weekend. If Trump is pulling ahead nationally, we should see movement in PA too. Does one poll or the other have a sample problem? A problem with its likely voter screen? Or will the PA polls tighten this week too?