Trump is not inevitable in Iowa

He's the clear favorite. But Iowa polls aren't very reliable at this stage and the other candidates are only a normal-sized polling error behind.

I promised myself that I wasn’t going to do much writing for this newsletter until I filed my next book chapter. (It’s almost done, Virginia!¹) But I’m going to give myself an out because most of what I want to say can be expressed in a single table. In a poll released yesterday of the Iowa caucuses, Donald Trump led Ron DeSantis 42-19, with Tim Scott in third place at 9 percent. The poll was sponsored by NBC News and the Des Moines Register and conducted by J. Ann Selzer, who is a great pollster. It’s a nice lead for the former president — showing a race a good deal tighter than in national polls, but roughly in line with other recent polls of Iowa.

But here’s the thing: anybody who has studied the Iowa caucuses can tell you that a polling lead doesn’t mean all that much at this point — not even a 23-point lead.

Here’s that chart I teased. It compares the late summer² Des Moines Register Iowa caucus poll against the actual results for every contested³ caucus since 1980.

The poll leader at this point won seven caucuses and lost eight, so they’re batting a little under .500. And a lot of the misses are recent. In fact, no candidate since 2004 who led the late-summer Des Moines Register poll actually went on to win the Iowa caucuses — except for Hillary Clinton in 2016, who just barely did.

That’s not necessarily the best way to look this, since Trump’s lead is larger than most frontrunners. Nonetheless, the average difference between the candidate’s lead in the late-summer DMR poll and his or her margin of victory or defeat in the actual caucuses was 17 points. In other words, Trump, with his 23-point lead, is not much more than a normal-sized polling error ahead of the competition.

Moreover, almost all of those errors went in the wrong direction from Trump’s perspective, with the frontrunner losing ground. Ted Kennedy, for instance, led Jimmy Carter in August 1979 by the same 23-point margin that Trump holds against DeSantis now — but he wound up getting crushed in the caucuses by 28 points.

So it felt a bit off to see headlines touting Trump’s lead as commanding or dominant. Characterizing a poll benefits from historical context. Historically speaking, a 23-point lead on the night before a general election is commanding. The candidate is essentially always going to win. But a 23-point lead five months before the Iowa caucuses is not especially safe at all.

Look, I think Trump is the favorite to win the GOP nomination; in fact, his price at prediction markets (68 percent) seems probably too low.⁴ As many recent candidates have shown, including Trump in 2016 and Joe Biden in 2020, you absolutely do not need to win the Iowa caucuses to win your party’s nomination.

But the Iowa polls — and to some degree the New Hampshire ones — are probably the least persuasive part of the case for buying Trump stock. Those states are famously volatile, with winners like Rick Santorum, John Kerry, George H.W. Bush and Pete Buttigieg⁵ having come from far behind, often quite late in the race. It’s not great for Trump that highly politically engaged voters in Iowa who have taken a longer look at the race have shown more curiosity about the alternatives. Those voters could leave just a crack of daylight open for DeSantis or Scott — both of whom had better net favorables than Trump in the DMR poll — perhaps along with other candidates.

1. Virginia is my editor.

2. I used the poll with a final field date as close to Aug. 16 as possible.

3. In 1992, other Democratic candidates essentially conceded the race to Iowa Sen. Tom Harkin. So that year is excluded, as are years when an incumbent president faced only token opposition.

4. Prediction markets seem to have become a less accurate proxy for “the conventional wisdom”. If you read media coverage of the race, it treats Trump’s nomination as all but inevitable — maybe the equivalent of a 90 or 95 percent probability. I at once think that prediction markets are too bearish on Trump’s nomination chances and that the Indigo Blob is too overconfident about them.

5. If you count him as the winner; Bernie Sanders won the popular vote that year.