Font Size: A A A


‘Don’t Look for a Soothsayer’: An Interview with Harry Enten

Matt Seaton
Harry Enten: What happens if the economy rebounds and has a strong month just before the election? The possibility of unforeseen events is greater now than perhaps ever. For people in my business, humility will help a lot.

Phil Penman for The New York Review of Books

Harry Enten, Central Park, New York City, May 23, 2020

A few weeks ago, I was scanning the polling pages on RealClearPolitics essentially looking for bias-confirmation that President Trump’s handling of the federal government’s response to the coronavirus pandemic must be killing his approval ratings. What I found was a disagreeable surprise: the president’s standing and his job approval since the Covid-19 crisis took hold appeared to be, well, healthy. Though still not great by historical standards (and declining in the weeks since), Trump’s negatives were lower than at any time since January 2017.

I knew there was only one person who could explain this to me. I beat a path to Harry Enten’s door.

Enten was still in his twenties when colleagues at Nate Silver’s FiveThirtyEight site dubbed him the “whiz kid” for “his sharp analysis and unique personality,” as an admiring profile in the Columbia Journalism Review reported in October 2016. A few weeks later, of course, some of the shine came off data journalism and opinion polling with Trump’s shock win.

But that bout of public skepticism hasn’t held Harry up much. Today, aged thirty-two, he is a senior writer and polling analyst for CNN Politics, where his colleague the anchor Jake Tapper has called him “the kind of guy who makes us all smarter.”

By way of disclosure, I’ve known Harry since 2012 when we were colleagues at the Guardian US—from where he was hired away by Silver in 2013. More recently, I’d been following Enten’s commentary on CNN’s Forecast Fest podcast (currently in abeyance, but ripe for revamping in the autumn election season)—I feel I always learn something from his blend of astute poll-watching, deep knowledge of election precedent, and incongruously old-time New York-accented delivery.

We arranged a social-distanced dialogue via email: starting with whether we can even trust polling, what follows is an edited transcript of our exchange.

Matt Seaton: There has been a conventional wisdom after Trump won in 2016 that everything the polling pundits were telling us was wrong, opinion polls don’t work anymore, and it’s all bunk. Do you still hear that? And how do you respond?

Harry Enten: I hear it a lot. And here’s what I say.

First, the polling was one of the early indicators of Trump’s strength. He gained the lead in the national primary polls (and state primary polls) less than a month into his bid. Had political commentators in fact listened to the polls instead of dismissing them, we’d have been better-off. That’s why, in this cycle, I wasn’t as dismissive of Biden as many were. Part of it was that he had many more endorsements than Trump, but part of it, too, was a consistent edge in the polls.

Second, the national polling was actually pretty good for the general election. Remember, polls are only snapshots in time. A good analyst’s job is to remind people that the numbers can change and try to explain how they might. With just a few days to go before the 2016 election, I wrote a piece titled “Trump Is Just a Normal Polling Error Behind Clinton.” Turns out he really was.

Third, the states’ general election polling could have been better. Those surveys were off by more than they’d been in any election since the 1980s. Part of that was just bad luck, though part of it, no doubt, is the emerging education gap in American politics. Whites without a college degree (a heavily pro-Trump group) were represented at too low a rate in those polls, and often the voters in that demographic whom the pollsters did get on the phone were disproportionately Democratic-leaning.

More pollsters now weight their surveys by education, and the state polls in the 2020 states most likely to be pivotal were more accurate in 2018 than they were in 2016. But this state polling distortion from 2016 is still what worries me most.

So why, in fact, should we pay attention to polls and polling analysts like yourself now?

Some folks don’t like the horserace; I do. The recent track record of polling—in the midterms and the primaries—actually paints a strong picture. In 2018, many polling analysts, myself included, were predicting a Democratic takeover in the House and for the GOP to hold the Senate—and that’s exactly what happened.

On top of that, we numbers people thought Biden was the frontrunner for almost all of the primary season after he declared his candidacy. Yes, he lost the first few states, but he always placed first or second in the polling.

And what should the informed, discerning consumer of political news be looking for from polling analysis? What is it good for, why does it matter, and what are its limits?


Polling can be used to call an election or merely to describe the populace’s feelings at a given moment. Knowing how Americans feel about stay-at-home orders, for example—and they’ve been solidly in favor of them—is important, and that goes far beyond any election polling.

Election polling can be used to help understand which issues matter to people, which can in turn help us understand why someone gets elected. I personally think the horserace component is fun, and plenty of people seem to agree.

But don’t look for a soothsayer. Look for analysis that knows its limits. Look for someone who explains why they think what they think. Someone who clearly states what their confidence is in any given forecast. Polling itself, and polling analysis, have a margin of error—and a margin of interpretation; anyone who is too certain should not be trusted.

There are a lot of new variables in this election cycle: Will there be a second wave of coronavirus infections starting in the fall; how will most Americans be voting (by postal ballots, etc.); what effect will an election season without the usual conventions, campaigning, and perhaps even TV debates have on the outcome; will November, in fact, turn into a referendum on Trump’s pandemic response, and so on? How difficult do all these new factors make your job?

This is somewhat new territory: there hasn’t been a single election like this one in the polling era, which began, roughly speaking, in the 1940s or 1950s. Fortunately, most events during a political campaign don’t tend to move the polls. Moreover, we’re dealing with a contest that has been incredibly static over time in terms of movement in the polls: Biden’s been ahead by around a little more than five points nationally for nearly a year and a half. Campaigns that are steady in the early going have a tendency to remain that way to the end.

Of course, we have to allow for a lot of uncertainty into that. What happens if we get a vaccine or a great drug for treatment that helps us beat coronavirus late in the campaign? What happens if the economy rebounds and has a strong month just before the election? The possibility of unforeseen events is greater now than perhaps ever.

For people in my business, humility will help a lot.

I recall that one of the first pieces you ever wrote for The Guardian was about how strongly predictive the state of the economy and economic sentiment is on the outcomes of presidential elections. How do you see that theme playing out in this one?

Certainly, the economy is in bad shape right now by almost every single objective measure, and many economic models that try to forecast the election would have Trump losing. Still, we don’t know what economic growth will be like just before the election.

That aside, though, I’m not sure how much it matters. Trump’s approval ratings have not correlated with changes in the economy. His net approval rating on the economy is, and has been, positive throughout much of his term, and yet his overall net approval rating has been negative.

The state of the economy may matter less in this presidential election than any since perhaps 1968, when the Vietnam War and the voter reaction to that war were paramount.

If not the economy, what are the voting sentiment drivers you’re particularly following in this cycle?

My study of past elections in which there was a dominant issue other than the economy indicates that whomever is most trusted on that issue wins the election. Right now, Biden is most trusted on handling the coronavirus. If that holds, he’ll likely win. If Trump overtakes Biden on who is trusted most, Trump will probably prevail.

There’s been a lot of commentary on how Trump’s base seems unshakeable, and he enjoys incredibly solid support from rank-and-file Republican voters even when, at times, the party establishment has seemed more squeamish. But how do you read that support, especially now?

Unless something crazy happens, Trump doesn’t seem to score lower than 80 percent approval from Republicans—and currently, he’s at around 90 percent or better. Even so, Trump’s overall approval rating cannot get above 45 percent. That simply tells you that we live in a hyper-polarized time.

A strong base is enough for Trump to stay competitive, but not enough to win. Moreover, if his approval rating among Republicans drops closer to 80 percent, his overall approval rating will approach 40 percent. It would be very, very tough to win under such a circumstance.

For Trump to emerge victorious in November, he’ll either need to get his ratings up with non-Republicans, or hope that Biden is very disliked and that voters see Trump as the lesser of two evils. That doesn’t seem likely right now.


Puzzling to me—no doubt, in my bubble—has been that Trump is apparently seeing the least negative approvals of his entire presidency these past two months. Specifically, his coronavirus response approval rating was even in positive territory for a while. How do you interpret this?

Phil Penman for The New York Review of Books

Enten, Central Park, May 2020

Look at all the elected officials in this country and those abroad: almost all of them have received boosts in their popularity. It’s called a rally-round-the-flag effect that often occurs in times of national emergency. The fact that Trump got one isn’t too surprising. What’s perhaps more telling is how small his was and how short it lasted. Almost every governor saw a bigger bounce in their ratings than Trump did. His boost was among the smallest for any such effect I could find, and one of the shortest, historically speaking.

So how significant is it electorally if a still-significant part of the population think he’s doing a not-horrible job in this crisis?

His ratings have fallen back somewhat, yet it gives you an indication that Trump can win in November. There is a segment of the population that doesn’t support him now but could back him. The election looks to be close enough at this point that a shift of support in Trump’s direction like that could very much matter. On the other hand, the fact that he got such a small bounce is an indication that if he wins, it will likely be by a small margin.

Purely from a pollster’s match-up perspective, did Democratic primary voters—those who got to choose—make the right call in pivoting quickly and decisively to Joe Biden?

They went with the candidate they thought was most electable. There is evidence (including from some work that I’ve done) that being more center-of-the-road does help pick up voters at the margins who might otherwise have voted for the other party or stayed home. We’re talking 1 to 3 percentage points most likely, but that could be enough to make the difference. Right now, Biden is seen as more moderate than Trump, which was not, in fact, the case with Clinton.

Obviously, there were other moderate choices, such as Amy Klobuchar and Pete Buttigieg, who may have been more electable than Biden. The problem was that they couldn’t win enough non-white Democratic primary voters, which made Biden the only realistic moderate candidate.

We saw in 2016 the electoral college effect of 70,000 or so votes in Wisconsin, Pennsylvania, and Michigan. But which states are you watching especially closely in this cycle?

I think we’re looking at eight states. There are the three you mentioned; then Florida (Clinton only lost it by 1.2 points) and Arizona (a 3.5-point margin for Trump, but the polling actually shows it’s a better pickup opportunity for Biden than Wisconsin). Carrying either Florida and one other state or Arizona, Michigan, and Pennsylvania would be enough.

Outside of those five, you have North Carolina (a 3.7-point Trump win) and Georgia (a 5.1-point Trump win. And then there’s the big one: Texas. Trump won Texas by 9 points in 2016, but Beto O’Rourke lost by less than 3 points in his 2018 Senate bid. If you give Biden Texas and its thirty-eight electoral votes, he doesn’t need to win any other state besides the ones Clinton won.

Whatever happens in the race for the White House, the outcome of November’s election for the Senate will determine a huge amount of what happens after January 2021. What are you seeing in the Senate race?

In a word, it’s tight. To gain control, Democrats need a net gain of three seats (if Biden wins), or four (if Trump wins). I recently gave the Democrats a 60 percent chance of the former result and a 50 percent chance of the latter. They’ve got a good shot in Arizona, Colorado, and Maine, while they’re likely to lose Alabama. North Carolina is tight.

Beyond these states, Democrats have at least a 25 percent chance of winning in Kansas (depending on a Republican primary outcome), as well as in Iowa and Montana. Based on some recent polling, at least one of the two Georgia Senate seats that are up this fall might also be at this level.

The bottom line is that, unlike the last cycle (the 2018 midterms), the political geography of Senate races in this one favors the Democrats: they only have to defend one seat in very Trump-friendly territory and are on offense in a lot of states.

The Democrats really did get a blue wave in the 2018 midterms, at least in the contest for the House, and arguably they built that success on talking about bread-and-butter issues like health care, not talking about Trump that much, and running some moderate candidates in hard-to-win purple districts. From the polling on the issues, would you expect to see the Democrats adopting a rerun of that approach?

You run toward the middle, and you run on the issues on which your side has the most popular positions: that’s what winning campaigns usually do. This year, you probably will see more talk about the coronavirus pandemic and issues connected with that because it’s a top concern, and Trump is weak on it.

And with Trump himself on the ballot, you’ll hear his name mentioned more. How could it be otherwise? But the approach will go beyond his character flaws; it’ll be about his policy failures, as it was with health care in the midterms.

Harry Enten, Central Park, New York City, May 23, 2020

Harry Enten, Central Park, New York City, May 23, 2020

You can follow Harry Enten on Twitter at @ForecasterEnten.

Subscribe and save 50%!

Get immediate access to the current issue and over 25,000 articles from the archives, plus the NYR App.

Already a subscriber? Sign in

© 1963-2024 NYREV, Inc. All rights reserved.