We’re talking Donald Trump — again — in this week’s 2016 Slack chat, but with a thespian twist. As always, the transcript below has been lightly edited.
micah (Micah Cohen, politics editor): OK, we’re going to do some role-playing today. We’ve been very skeptical of Donald Trump’s chances of winning the Republican nomination here at FiveThirtyEight. To really stress-test that position, Nate Silver, our editor in chief, is going to be his normal self (a Trump skeptic); Harry Enten, our senior political writer, is going to pretend he’s a Democratic-leaning political scientist who thinks the Republican Party is in a such a mood that Trump could win (we’ll call him “PhDemocrat”), and I’m going to play the role of a die-hard Trump fan named “Trumpfan1959” (drawing inspiration from the many Tweets and Facebook comments his supporters have sent us).
PhDemocrat: Nate, you always talk about small sample sizes, and there is perhaps no better case of a small size than presidential nomination races. We have had only 12 open primaries in which the “Party Decides” hypothesis — that the party establishment, including elected officials and party operatives, has a lot of control over who wins the nomination — has supposedly held true. So tell me why we should be putting so many chips on a theory with so little data backing it up?
natesilver: Well, PhDemocrat, if you wanted to argue that Trump’s chances are higher than zero, I’d agree with you. But the conventional wisdom isn’t just saying that Trump has a chance. It’s increasingly seeing him as one of the MOST LIKELY nominees. Betting markets — as good a quantification of the conventional wisdom as you’ll find on short notice — have him at 22 percent. Higher than Cruz. About twice where they have Jeb Bush. Trump is four times more likely to be the nominee than Christie is, they’re saying. I think 22 percent is too high.
It’s also significant that not only have Trump-like candidates not won, but they haven’t come particularly close to winning.
PhDemocrat: What is a Trump-like candidate? Part of the reason Trump-like candidates haven’t won is because they haven’t run. In most presidential primary campaigns, it’s only people who have held elected office who have run. The only examples of candidates like Trump running I can think of are Steve Forbes in 1996 and 2000, Herman Cain in 2012, and maybe Ben Fernandez in 1980. (Granted, Fernandez had experience in government.)
Trumpfan1959: Listen, if either of you ever left your East Coast liberal hangouts, you’d realize that there haven’t been any Trump-like candidates. This is someone who built a multi-billionaire dollar empire. He led a hugely successful reality show. He wins. We’ve never had a candidate like that, so the normal rules don’t apply.
natesilver: By a Trump-like candidate, guys, I mean someone who openly defies his party and whose party is openly rooting against him. It’s one thing to say that the party chooses its nominee. That’s not always true. But can the party prevent someone from hijacking its nomination when the consequences for it would be disastrous? My guess is that it can.
Also, we do have some empirical data on how non-politicians perform in campaigns. It’s not uncommon to see them in U.S. Senate races, for instance. And the answer is that they tend to perform poorly. They often fade down the stretch run — look at Meg Whitman in 2010, for example — because of a tendency to commit gaffes and a lack of organization on the ground-game side of things.
Trumpfan1959: Trump has committed plenty of “gaffes” — at least, according to you people — and he still leads every poll.
PhDemocrat: Yes, Nate, but they also win. David Perdue won in Georgia in 2014, for example, and he survived both a competitive primary and a general election campaign. And Perdue also committed some gaffes. What I hear from you is a lot of guessing.
natesilver: You guys are arguing against a straw man. I’m not saying it’s impossible for Trump to win. But it’s unlikely — less likely than betting markets and the conventional wisdom hold.
Also, there’s an important difference between a one-off election like a Senate or gubernatorial primary and a presidential nomination. In a presidential race, voters and the party establishment have time to read and react. You’ve had candidates like Newt Gingrich and Pat Buchanan win individual states before, but they haven’t been able to sustain their momentum.
Will an onslaught of super PAC ads against Trump hurt him or help him? I dunno! But the party has a lot of time to experiment until it finds the right formula.
Trumpfan1959: It seems to me that you have no idea what’s going to happen — maybe there’s zero empirical evidence that candidates like Trump can win, but there’s just as much evidence that candidates like Trump CAN’T win. Trump is dominating the race, media and polls. How is that not the best position to be in?
natesilver: Let me turn this around a bit. You both seem to be placing a lot of emphasis on Trump’s polls. I don’t think they mean very much.
Trumpfan1959: If Marco Rubio were winning, you’d be citing the polls all the time. You just don’t like what they show.
natesilver: Dude, this isn’t complicated. Go back and look at past polling frontrunners at this stage of the campaign. They have a poor track record. By contrast, go back and look at who was leading in general elections in late October. They have a very good track record.
The point of being empirical isn’t that you love polls. It’s that you learn from experience, and our experience tells us that polls aren’t reliable predictors at this stage of the race.
PhDemocrat: 1980: Reagan led in polls and won. 1984: Mondale led in polls and won. 1988: Bush led in polls and won. 1996: Dole led in polls and won. 2000: Bush and Gore led in polls and both won. 2012: Romney was top of polls for most of primary and won. That’s seven out of 12 times the person who led in the polls at this point has won. That’s greater than 50 percent. And it’s far greater than the chance you give to Trump at this point.
natesilver: Most of those candidates had far more of the vote in polls than Trump’s 25-30 percent.
PhDemocrat: Romney didn’t.
natesilver: But Romney had more room to grow because he had establishment support and was not a divisive figure within his party. It’s those qualities — the voters and elites consolidating around a broadly acceptable nominee — that historically allow candidates to zoom up from 25 percent to 50 percent or whatever once they win a few states.
But more importantly, voters aren’t paying a lot of attention. Only 20 percent or so of the voters in Iowa have come to a final decision. Half the voters in New Hampshire won’t decide until the final week of the campaign.
BTW, the same conventional wisdom that is increasingly keen on Trump’s chances also insisted for months that Hillary Clinton’s campaign was in disarray and that Joe Biden would enter the race.
Trumpfan1959: I don’t have much faith in that self-reported “when I decided” data.
PhDemocrat: Yeah, keep in mind that some 30 percent of voters in the 2012 general election claimed they didn’t decide whom to vote for until the last two months of the campaign. Twenty percent said they waited until the final month. But you and I both know that there are VERY few real swing voters deciding that late.
natesilver: It’s interesting to me that more legitimate sorts of Internet polls, which also have trouble drawing a random sample, show Trump doing a lot better than traditional random-digit-dial polls do.
As Harry said to me earlier today, Ann Selzer, who has as good a track record as anyone, doesn’t have Trump doing all that well in her Iowa polls. And she was the one who caught Obama’s surge of new voters in 2008.
You guys are also neglecting something else. Some voters may be coughing up Trump’s name in polls because he’s the only candidate they’ve been hearing about. The media has given his campaign more coverage than literally all the other Republicans combined.
Trumpfan1959: So why aren’t you citing that as a Trump advantage? He knows how to draw attention to himself. You don’t think Rand Paul or Lindsey Graham would kill to have that skill?
PhDemocrat: I’d still like to note that Nate, who prominently featured that exit poll data, should acknowledge that even though exit polls are good, they are not gospel, especially when it comes to people self-reporting when they decided. And I should note that even Ann Selzer gets it wrong once in a while. She had John Kerry carrying Iowa in 2004. That was wrong. She had the establishment candidate, Terry Branstad, winning the 2010 Republican gubernatorial primary by 28 percentage points. He won by only 10 points. She’s good, but she isn’t perfect.
natesilver: Because coverage will probably tend to even out a bit as we get closer to the election. Also, voters will draw more from local news sources in addition to the national ones.
Trumpfan1959: This is all starting to sound very pundit-y.
natesilver: It was interesting that we saw Trump’s numbers wobble a bit after the first couple of debates. Also interesting that he seems to be a bit lower in Iowa, where voters are paying more attention, than everywhere else.
PhDemocrat, no pollster is perfect. Nor are we! But you and TrumpFan1959 are creating a moving target.
You say: WOW! LOOK AT TRUMP’S POLLZ!
I say: Polls don’t mean much at this stage and aren’t very predictive.
So then you say: you can’t prove they’re NOT predictive. Which is a much different standard.
Trumpfan1959:

nuff said.
PhDemocrat: Hold on a second. Hold on a second. No, what we’d say is that there is a strong correlation. Heck, there’s a strong correlation between polls from the first half of the year before the election and primary results. The polls aren’t perfect, but what we’d say is that more often than not the polling front-runner does win. And for you then to say, it’s only a 5 percent chance — that, I think, is wrong.
natesilver: A polling front-runner wins more often than not when the front-runner is at 50 percent in the polls, like Hillary Clinton is now. But Trump’s at 25-30 percent nationally and a bit less than that in Iowa. Also, if you built a simple model with polls and endorsements, it would not have Trump doing all that well.
I guess what it comes down to is that there’s a difference between saying “EVERYTHING LOOKS GREAT FOR TRUMP” and “THERE’S A LOT OF UNCERTAINTY ABOUT THE RACE.”
I emphatically agree with the second statement, not the former.
Trump doesn’t have the profile of a traditional nominee. Both in the sense of his running an anti-establishment campaign, and in terms of his metrics. Polling at 25-30 percent doesn’t mean all that much unless you have a lot of other things going for you too, which will allow you to continue adding support as you move through the process.
HOWEVER, none of the other candidates looks all that great either.
Rubio has a lot of upside potential, but it’s not like he’s sitting there with hundreds of endorsements. Cruz could be a perfect bridge between the establishment and insurgent sides of the party, or get the worst of both worlds — maybe he doesn’t win establishment backing either, but he can’t fire voters up as much as Trump does.
PhDemocrat: Exactly, none of the other candidates looks that good. None of them has raised a TON of money in non-super PAC cash. Trump has a ton of personal money if he needs to spend it. Let’s say you have Trump at 8 percent or whatever, and I have him at more than double that. You say there is uncertainty and so do I. But then you take that uncertainty and still peg Trump at only 8 percent to win. Here’s where I am: There’s a lot of uncertainty, so just because Trump leads now doesn’t mean he’ll win. But he has led in the polls, there is no clear money front-runner, there is no endorsement front-runner, Trump has money, GOP voters are upset, and the betting markets have Trump’s chance of winning north of 20 percent… So you know what? I’m going north of 20 percent too.
natesilver: A completely uninformed model would give Trump a 1 in 14 chance, since there are 14 candidates left. Which is 7 percent. So I’m actually pretty close to that.
But if you want to think about this in a technical way: it’s not just that Trump has no support from his party. It’s that the party is actively looking to stop him because he’d be a catastrophe as their nominee.
Endorsements are just a proxy for party support. Jim Gilmore doesn’t have any party support — or any endorsements to speak of — but he isn’t in the same category as Trump. If you were building a model, it’s as though you’d want to give Trump negative-100 endorsement points.
Read more: John Kasich’s Quiet Campaign To Cut Abortion Access