A NYT Reporter on What Happened to the Needle on Election Night

(And also what happened on election night.)

Voters waiting in line, some on their phones.
Arizona voters wait in line at a polling place in Phoenix during the midterm elections on Nov. 6. Ralph Freso/Getty Images

On this week’s episode of my podcast, I Have to Ask, I spoke to Nate Cohn, who covers elections for the Upshot at the New York Times and is also my good friend. Ahead of Tuesday’s vote, he also helped oversee a massive project in which the Times and Siena College polled a huge number of House and Senate races. Below is an edited excerpt from the show. In it, we discuss how optimistic Democrats should be about the Midwest, what Tuesday’s results suggest about Trump’s odds in 2020, and what happened to the Needle on election night.

You can find links to every episode here; the entire audio interview is below.

Isaac Chotiner: What’s your big takeaway from what we saw on Tuesday night?

Nate Cohn: That it was a good night for the Democrats. It wasn’t the night of their dreams necessarily, especially in the Senate. But I thought that their performance in the House was really, really impressive given the formidable structural disadvantages they faced in the chamber. I think in the end they’re probably going to win something like 39 seats.

Clearly, the Democrats have to be disappointed by their performance from high-profile contests in the Sun Belt. In Arizona, where they may yet win when all the votes are counted, it was not a clear victory for Kyrsten Sinema that a lot of people expected. And in Florida, Bill Nelson still has an outside shot to win, but it looks like they will narrowly lose both the governor’s race and the Senate race.

But overall I don’t see how you can conclude it was anything other than a great night for the Democrats. They’ll probably win the national popular vote by 7 points, which is better than what the Republicans got in 2010 and 1994. I think that by most of the standards that you use to assess these sorts of things, this is about as good as an election that you can get in American politics right now.

Was there any big surprise to you?

The turnout was astonishing. I think we came in with high expectations. But it’s very easy to look at highly competitive congressional districts and find places where the turnout matched or exceeded 2016. And there is no precedent for that in contemporary American politics. I think that when all the votes are in and counted, we’ll get up to something near 115 or maybe even 120 million votes cast nationwide. In 2014, it was 82 million. In the president’s election, it was 137.

Aside from maybe suburban white women, who we’ve heard a lot about for a very long time, was there any group’s turnout that particularly surprised you?

Across the board it was a little weaker in white working-class areas. It was weaker among Hispanic voters. But those are common patterns in midterm elections. I’m not sure the Democratic turnout will ultimately be assessed to have materially exceeded Republican turnout, if it did at all. It was just a very high turnout election across the board.

Were you surprised that Republicans managed to win as many Senate seats as they did?

I am not surprised by the red-state parts. I’m not surprised that the Democrats lost big in Indiana, Missouri. And I’d add Tennessee to that list. In our polling, the Democrats just never really were doing exceptionally well in deeply Republican areas. And my understanding is that the private polling showed something similar. And I don’t fully understand why there were a lot of state public polls that at the end of the race showed Democrats faring very well in places like Missouri and Indiana.

I am surprised by the result in Arizona and Florida to some extent. I don’t think that it’s a huge polling error. But it’s a 3-point polling error in two states that were polled a lot—or, rather, were polled a lot by a diverse set of pollsters using diverse methodologies. And in general those polls were high-quality. Those are races that on paper the Democrats ought to win in a wave election. But one useful thing to do is compare it to a House race. In the House races where a Republican retired, and Trump won by 3 points, you would expect that the Democrats should have won that seat comfortably. I think if you went down the list, you would find the Democrats won every seat that was comparable to the Arizona Senate race, and yet the Democrats didn’t win that. [Editor’s note: As of this publication, the Arizona Senate vote is still being tallied. The outcome is not yet known.]

Polls before the 2016 election understated the amount of lesser-educated white voters. Do you have some sense of what happened this time?

I don’t. I think it’s troubling that we had another wave of final polls in Ohio, Iowa, Indiana, Michigan, and Missouri that all seemed to overstate the strength of the Democrats.

Were you generally surprised by the results in Florida? Not just the Senate race but the governor’s race, where Gillum’s support was also overstated?

Yes. I was. And I’ve only glanced at the results by county. But what’s really striking is that it really looks to me like Gillum and Nelson did the things they were supposed to do. They won Pinellas County, which is St. Petersburg, by a 4- or 5-point margin. They won Duval County, which is Jacksonville, which Democrats basically never win. I can’t tell you the last time a Democrat won Jacksonville in a high-profile state election. Obama didn’t win it. [Gillum] won Seminole County, which is sort of east, which is suburbs north of Orlando. Obama didn’t win those counties either, even as he carried the state.

It seems to me that just at a glance, the results in Miami-Dade County and in the Orlando area, where there is a large Hispanic vote, looked fairly disappointing for the Democrats, but I don’t think that covers a full 3-point error in the polls.

There had been a lot of talk after 2016 that after Democrats got slaughtered in the Midwest, at some point they might have to turn to a different map that could include Arizona, Nevada, maybe Florida, Georgia, or North Carolina. But looking at Tuesday night, it seems Democrats did better in the Midwest. They have the governorships in Wisconsin, Minnesota, Pennsylvania, and Michigan. Trump’s approval rating is below 50 percent, even somewhat significantly below 50 in some of these states. What did the results there tell you about Democratic strength in the Midwest? And what did it make you think about 2020, if anything?

Well, first let me say, and you know this because we talk all the time, that I have always felt that the Democratic path is in the Midwest. There are more swing voters there. The Sun Belt states I think offer relatively limited upside for Democrats. And the final thing is that if the Democrats don’t have a strategy intended to stem the bleeding on white working-class voters, it could get worse for them. There is no reason to suppose that 2016 was the floor among that group, and there are additional electoral votes for the Democrats to lose in a place like Minnesota or Maine.

That said, I thought the election results were broadly consistent with the view that the Democrats could win those states back. But I don’t know that I thought that it was a very impressive performance, Isaac. I thought that Debbie Stabenow’s performance in Michigan was pretty disappointing. Given that it was a wave election, where the Democrats won the national popular vote by 7 points in the end probably, an incumbent Democrat winning Michigan by 7 points or so does not impress me all that much. The Democrats fell short of reclaiming the governorship in Iowa. Scott Walker did lose in Wisconsin, and that is important. But if the Democrats want to win through the Midwest, they need all of Wisconsin, Michigan, and Pennsylvania. While I think they showed strength there, I’m not sure that they showed enough strength to indicate the Democrats were gonna do better there than they would in the national popular vote, which had been the case before Trump was the president.

At the same time, I don’t think that their performance in the Sun Belt should leave them very optimistic about their ability to break through there, either. I mean, they got a very strong turnout from black voters in Georgia, and didn’t quite get over the top in Arizona and Florida. [Editor’s note: Again, these races are still being counted.] I think they can be extremely proud of how they did in Texas. But Texas is the state they were furthest from winning. I continue to think that there is danger for the Democrats in sort of getting caught in between—where they are doing way better than they did in the Sun Belt, but not quite good enough to win, and they are doing worse than they used to do in the Midwest, and maybe not quite good enough to cobble together 270 using those states alone. Because you would have to run the table in Minnesota, Michigan, Pennsylvania, and Wisconsin to get over the top without any Sun Belt states that were carried by the president.

Can you imagine a situation where Ohio is 2020’s deciding state?

I don’t see Ohio as the deciding state. Maybe if Sherrod Brown was the Democratic nominee.

And when we talk about the Midwest, are we talking about Pennsylvania?

I am not. So, there’s a big debate. But to me, it’s not in the Midwest. It does share some demographic characteristics of the Midwest. Like, they’re both relatively white. But I don’t think that, historically, they move in unison.

If you go through Pennsylvania on one of your fancy Acela trains, I think that it doesn’t count as the Midwest.

There you go. I get to see Pennsylvania out of the Acela all the time.

So then, how did you think the Democrats did in Pennsylvania, and what do you think of that state in 2020?

I think that Pennsylvania has to be a decent state for the Democrats in 2020. I thought they did OK there. I didn’t think they did great, though. I mean, they only picked up three House seats despite a new map that was drawn in their favor. They weren’t able to get over the top in the 1st Congressional District, which was based in Bucks County. And admittedly, the Democrats had maybe their worst nominee of the cycle in that district. But, you know, it wasn’t an exceptional performance, either. They did do very well in the governor’s race and the Senate race, though. Better than they did in equivalent races in Michigan and Ohio. So, if I were ranking the states right now, based strictly on the midterm result, I would feel better, if I were a Democrat, about Pennsylvania than I would feel about Michigan.

Your friend Harry Enten had a Twitter thread this week saying Trump is likely to lose re-election. I’m curious if you saw that thread—well, you should have seen it, because I sent it to you.

I’m reluctant to read into the state of a presidency. Or rather, to read into a president’s presidential election chances based on their standing at the midterms. You know, I remember back at this time in 2010, there were a lot of people that thought that Barack Obama was going to be in a lot of trouble in the Midwest because his approval rating was under 50, and because the Democrats lost a whole bunch of governor’s races and Senate races in many of the same states we’re talking about today. So I do think that it would be a mistake to just assume that because the president is where he is today, that that’s where he’ll be in two years. That said, you know, the president’s approval rating has been really static and stable. And you know, Trump won the election in an unusual way, which is that he won it with 46 percent of the vote, and didn’t get close to 50 percent in any of the states that we’re talking about.

I think one plausible interpretation of all this is that the sort of voters who decided either not to vote because they didn’t like both candidates, or the voters that elected to write in a candidate or support a minor-party candidate, continue to feel as negatively towards the president as they did at the time of the 2016 election, except that now they would be more likely to support a Democratic candidate. Either because now politics are more defined by the president himself, or because the Democrats are likelier to find a more palatable nominee. I’m not saying that interpretation is right, by the way, but I think that that interpretation has always had some merit, and I think it is at least consistent with the results. So if I were a Democrat looking to be optimistic, I would focus more on that possibility than the assumption that if the president’s approval rating is at 46 today, that he will be in trouble in 2020.

Does this election give you any kind of insight into the type of candidate you think Democrats should run in 2020?

I would point out two things about what we see in the results so far: One is that just being a progressive superstar is not enough to fundamentally transform an electorate and win a race. There were a lot of progressive candidates who won primaries this cycle on some sort of argument that if we mobilize the base, we can transform the electorate and win places where we don’t usually win. It didn’t happen. The Democrats that I saw who outperformed the most were people who were relatively moderate. I mean, they weren’t necessarily centrist or something, but they weren’t running as progressive firebrands. And yet something about their biography still made them really compelling. Both to the progressive base and to moderate voters. A lot of them had military backgrounds. A lot of them were just compelling candidates, really talented candidates who came forward in a year when Democrats needed them to. And so if I were a Democrat looking at 2020, I would look to the people who did best in this year, and I would say that they are young, and that they still manage to excite people without listing off every policy dream of the left. That’s not to say, by the way, that you can’t win doing that. It’s just that I don’t think there’s necessarily all that much upside if you can excite people by other means.

To me, it’s not all that different from what Obama did. I mean, one of Obama’s great strengths was that he managed to sort of be something for everybody. You know, if you were a centrist, you could see Obama as a centrist. If you were a progressive, you could see him as a progressive. And I think that Donald Trump is a similar candidate in his own respect. There are conservatives that see him as a conservative, but there are a lot of white moderate voters in the Midwest who voted for Barack Obama who don’t see Trump as a conservative extremist at all. They see him as someone who is fighting for working people in much the same way that Democrats have traditionally been thought to fight for working people. So if I were a Democrat, I’d be looking for someone who has that combination of appeal—someone who has the ability to reach out to moderates on pocketbook issues, who has a compelling biography. I’d probably suggest that they were a young person. And I wouldn’t make it an ideological referendum.

We’re entering an era of American politics where the Senate will be very tough for Democrats, because of the way the Senate functions and the urban-rural divisions. If the parties keep going in these directions, what can Democrats do?

Well, they need the parties not to be going in these directions. I don’t think that there’s an answer to your question conditional on the first clause of it. I mean, I think that if you continue to polarize the country along racial and educational lines, Democrats will keep doing better in urban states that are diverse and well-educated, with large populations. And that will not be rewarded in the Senate. I mean, look at the places where the Democrats had their best nights: Texas, Georgia. I mean, those are states where the Democrats just are not going to be rewarded for it in the Senate.

If you can’t keep open Democratic appeal among white working-class voters, then there’s nothing you can do about it, from a Democratic standpoint.

And what do you think it would take to appeal to more of those white working-class voters?

I think that as long as American politics is defined by immigration and by issues that polarize the electorate along racial and geographic lines, that we’ll probably see a continued trend towards polarization along racial and geographic lines. So I don’t think that there’s all that much the Democrats can do to stop this, as long as the basic breakdown of these coalitions continues, and the Democrats continue to fight over these same issues. Now, that said, sometimes the issues change. In 2012, we were not talking very much about immigration. In 2012, we talked about gay marriage and abortion a lot. In 2012, Planned Parenthood was an issue. In 2012, the Democrats were the ones running on trade and outsourcing and Bain Capital. So it’s possible that the Democrats can at some point return to a set of issues that are a little bit more favorable to them with white working-class voters. But I don’t think that if you keep relitigating the issues of the 2016 election, that the Democrats are going to get a different result in terms of the overall geographic breakdown of the electorate.

So what are those issues that you’re talking about?

Immigration, being on the pro-trade side of trade. I think that combination of issues is really tough for Democrats in a lot of places. Although I don’t like the term “identity politics,” I mean, I think that as long as that’s a major force in the culture, that that’s tough for Democrats too, in a lot of these places.

Right. The caravan was a huge issue that the White House, Fox News, and conservative media were talking about every day. And there were debates in mainstream media and among liberals about whether Democrats needed to discuss it more. And naturally, now Fox has completely dropped it. The White House isn’t talking about it anymore, which suggests that maybe they didn’t think it was so important in the first place. How do Democrats deal with something like that, especially if Democrats feel that the issue is being ginned up for political purposes, and it’s not actually about addressing some real problem in society?

I think the Democrats have a real dilemma on immigration. And I think the basis of the dilemma is that they think it’s a moral issue. The political reality, though, is that because of the way our electoral system is configured, that the sort of people who disapprove of the Democratic view on immigration have a lot of sway. I don’t like comparing everything to 1992 and Bill Clinton, but there were a lot of issues where the Democrats were sort of outpacing, you know—they were a step ahead of the electorate on a number of issues at that time. Bill Clinton didn’t go to being conservative on all of those issues, but he at least softened the Democratic position. I think that if the Democrats could do something like that on immigration, it would probably be in their interest to do so. And if they did soften their stance on immigration, I think they would have more credibility to put up a fight with the president on the president’s most outlandish views.

So take the caravan, for instance. I don’t think the Democrats have a ton of credibility with the voters they care about most on those issues so long as they are talking about abolishing ICE. And I think that the Democrats would probably do well to take a step back on those sort of issues—if they can, and feel morally like that’s something they can do.

It’s a problem with Trump, because he’s so good at commanding attention. Also, when you’re not in power, you can’t set the agenda in the same way. So sometimes it feels like the agenda’s being set for them, but they have to respond. And it’s tricky to figure out how exactly to do that.

Yeah, that’s true. I think that from a messaging standpoint, if you’re a presidential candidate, we’re not very far removed from when Barack Obama talked—he supported deportations, supported more border security in the way that he framed his stance on immigration. And I feel like that would be a problem for a Democrat in the 2020 primary, to basically rehash Obama’s message on the issue.

I think that when the Democrats move far enough to the left on an issue, they lose a little bit of credibility to push back on the most extreme stuff on the right. And again, I’m not trying to make an argument about what the parties should do necessarily, because political parties always have to make—always have to balance electoral considerations with the issues that define the party, and immigration is increasingly an issue that defines the party. So if they’re saddled with the issue, then they just have to try and make the argument as best they can. But I do think that the 2008- and to a lesser extent the 2012-era Democrats’ messaging on immigration would be more effective for them than the one they have now.

The New York Times has this thing called the Needle, which is a kind of live look at the odds that Democrats will win the House or Republicans will win the Senate, as the results are coming in. You guys had a bit of trouble with the Needle on Tuesday night. Can you explain exactly what happened?

The first thing I’ll say is live forecasting is really hard from a technical standpoint. You don’t really get an opportunity to test it. The election begins, and if it doesn’t work, then you have to try and debug it then. We had issues in what I can casually describe as the data pipeline getting data to us and to the model. That meant that we were not able to publish the Needle until later in the night than we had hoped.

In particular, we actually had ambitions for our forecast to be much more powerful than it had been in the past. We were going to rely on precinct data from Georgia, Virginia, Florida, Minnesota, California, maybe some other places, to really supercharge our estimates, particularly early in the night when there isn’t much hard data yet and you’re only looking at early votes. We thought that by 7:45, we would have an extremely granular understanding of the race in a way that no one else would, and literally none of the precinct data materialized and never did. We had to flip to a model that depended loosely on counties, like our 2016 one. Once we did, things looked pretty normal and right, and we published it.

At Nate Silver’s FiveThirtyEight, his House model also got a little funny there for a while. It went up to about 92 percent Dems take the House, and then very quickly went down to about 38 percent, and then boomeranged back up to about 57 percent, where it stayed for a while, and then it went up again. The kind of objection you see to this sort of live modeling on election night is that it drives people crazy.

I don’t know what’s going on with the FiveThirtyEight model. I can just say that when we turned the model back on and backfitted it, we never had Democratic chances drop beneath 85 percent.

I think of election night forecasting as—it’s almost self-evident that you would do it. We are always trying to figure out what’s happening in the world as soon as we can. That’s what journalists are always trying to do. Waiting until every last vote is counted is usually entirely unnecessary, and if we can tell you something about what’s happening in the world before, then we should.

Do you think that given that FiveThirtyEight and your model both had trouble, even though it was a different variety of trouble—I guess what I’m asking is: This is such a fraught thing in that people are so on edge, and it’s so hard to get right because it’s so complex. Is that in itself a reason to not trot things out on Election Day?

No. I think that technical issues in a live setting are—they’re going to happen. We didn’t publish anything. Frankly, our forecast was really good. We would not have called a single race wrong if we had used it. We would have said throughout the entire night, even when people were freaking out about Democratic chances, that they were on track to win the House. Although the precincts didn’t work, if we had, we would have shown Democrats on track to win in Virginia-2 and Virginia-7 very quickly, which would have led the night to have a totally different feeling for most viewers, I think.

I am sad that we were unable to publish it as quickly as we had hoped, with all the data we had hoped. But my only regret is that it didn’t work more quickly, not that it shouldn’t be done.

What time did you go to bed on Tuesday night or Wednesday morning?

7:30 a.m.

And when did you wake up again?

Maybe noon.

And did you sleep the next night?

I got to bed maybe by 2:00 a.m. I’m trending back.

I should note by the way, as you know, I’ve been staying up until 5:00 a.m. for a while now. I’ve been doing all of that to build these precinct projections in the key states that ended up being completely useless to us, so—

That’s a very sad story.

It is sad.