Q&A: In defense of polling — and a warning for the Democrats
Chris Wilson, CEO of WPAi, a U.S. polling company, who points out the difference between media and partisan polls
In an effort to make some sense of what is happening with the whole ... [gestures to the United States] thing, The Line interviewed Chris Wilson, CEO of WPAi, a U.S. polling company on the Republican side. Wilson has worked with more than 100 current or former members of the U.S. Congress and with the 2016 Ted Cruz presidential campaign. At the time of the interview, ballots were still being counted in many states; the purpose was not to predict what the final result will be, but to analyze the result and respond to criticisms being made of the polling industry as a whole.
The Line: Let's start with the basic question. At the time of this interview, it looks like Joe Biden is going to grind out an electoral college win, and if he catches all the breaks, maybe a sizable one. But we don't know yet, so let's not forecast what's going to happen. Let's just talk about what did happen. What happened in America on Tuesday?
Wilson: Gosh, there's so many things! First, Republicans did better than expected, especially with the minority vote. Far better than we expected. Trump might get as high as 30 per cent among Black and Hispanic voters. That's huge. That's the best showing for any Republican presidential candidate as far back as the ‘50s — I didn't have time yet to look back further. [laughs] So that's big. Phenomenal. Overall, turnout was huge. Both sides brought out millions more voters. We had unprecedented numbers of voters. The "Trump Coalition," which is still something that's changing and defining itself within the GOP, turned out for Trump. That's working-class voters. Blue-collar workers voted for Trump right across the board, in all states. Nevada to Maine.
The Line: Did any of that surprise you?
Wilson: One big surprise, looking at exit polling, is that the narrative of suburban women abandoning Trump was more complicated. Unmarried suburban women did not support Trump, but married suburban women were one of his best demographics. He did better with married women than with married men! So the entire category of "suburban woman" is going to need to be broken down. We might have to start looking at marital status as being a metric on par with educational level, which was huge. In polling you're always fighting the last war. In 2016, education was huge. In 2020, marital status was. As for other surprises, I was only surprised in places where we weren't polling. The media polling is getting a lot of really sharp criticism, and a lot of it is deserved. But the partisan polling was very good. I wasn't surprised in areas we were polling.
The Line: Let me interrupt you here — can you explain to the readers a bit more what you mean by media polling and partisan polling?
Wilson: Sure. A media poll is commissioned by a media company, right? And it's intended for publication. Partisan polling is commissioned by one of the parties, and it's for internal use. It shapes our strategy and our decisions, but it's not intended for public use. Sometimes, sure, you can release a partisan poll strategically, to show that you're up somewhere. But generally, you just use it for your campaigns. And the partisan polling is very, very good. We were polling this election, and we won everywhere we thought we were going to win. I have a lot of friends who are Democrats, and we communicate. And they knew where they were going to win, too. So the partisan polling was good on both sides. But it doesn't get released. The media polling isn't as good, and it gets all the attention. So when people are angry at the pollsters, they're angry at one part of the industry.
The Line: Why is partisan polling more accurate than media polling? Or put another way, what would the media companies have to do to improve their polling?
Wilson: There's two problems with media polling. One is a crime of commission. Some media companies go out there with a poll to set a narrative. That's not what polling is for. It's a big problem. That irritates me the way I imagine you guys would be irritated if you saw a journalist writing partisan drivel to skew a race but insisting it was news. So that's a big problem. But the bigger problem is one of omission. Let me be clear: I don't believe there's a secret plot among media pollsters to help Biden or the Democrats. That's not true. But the media polling companies assume narratives around things like turnout and expected minority group vote share, and they use that to weight their analysis. They looked at early voter turnout and adjusted their models. There needs to be a lot of soul searching among media pollsters about how they allow narratives to affect how they analyze their samples.
But one more thing I'd mention, and this is a crime of commission again, is the weighting of Hispanics. The media pollsters assumed it was going to be like 90 per cent against Trump. I don't mean that literally, but in big picture terms, that was the narrative, right? But like I said, Trump has the best performance among Hispanics since the ‘50s. Look at what happened in Florida. Look at the Texas districts along the Rio Grande, where the media pollsters gave Trump no chance of winning. The media pollsters are getting educated Hispanics and upper-income Hispanics that are easier to get on the phone, and that skews their sample urban and Democratic. They weren't sampling enough in Florida and rural Texas, where you've got a lot of Cubans and Venezuelans who don't like socialism and didn't like what they saw in the Democrats. Every few years the Democrats say, oh, we'll flip Texas or Florida, but they never do. Because Hispanics are voting Republican.
So that's it, really. There's more I could say, but if you want to understand the difference between partisan polling and media polling, it's weighting for demographics and geography and how that impacts analysis. That's a Grand Canyon of difference.
The Line: I think something we should mention is that political parties have money to spend on polling. Media companies don't have nearly as much. They can't afford to do the kind of polling you can do. Would they do better if they spent more, or is there just too much bias built in?
Wilson: The money thing is real. Yeah. But I don't think the problem is bias, no. This is going to sound self-interested, but the media companies need to hire partisan pollsters. [laughs] You're right, there's a lot of cost concerns for the media companies. But it's mostly analysis and weighting. They need more guys like us helping them with their samples. Here, let me get really down in the weeds for a minute. We build our turnout model based on predictive analytics. We score every single voter in the entire country on a zero-to-one scale on whether they're likely to turn out. Anybody above point five we estimate is likely to vote; anybody below a point five, we estimate behaviorally is not likely to vote. Those above point five are the ones that make up our turnout model. We have one media client. It's the Las Vegas Review Journal. We put out a poll in Nevada, showing a dead heat between Donald Trump and Joe Biden. We were the only one that came close. We had Biden ahead! But in single digits. Everyone else had him in double digits. And reporters for other papers were criticizing us, saying we didn't know what we were talking about. That we were trying to push a narrative. Well, look at Nevada. A dead heat. [Note from The Line editor: at the time of the interview, the spread between Trump and Biden in Nevada was less than one point, to Biden's advantage.] Another good example of that is Ann Seltzer in Iowa. She knows the state of Iowa, she has clients there and built her model accordingly. She put a poll out that was different from what everyone else was showing, and took heat for it, and, well, guess what — she was right. Media pollsters don't have the experience that we do, so they need to build a model on assumptions and prior elections — 2016, the 2018 midterms. This is a challenge, and it's more than just a budget factor. Why would they go out and build a predictive model in a state or a nation without having clients to use it? You wouldn't do that just for political polling.
The Line: Well, there is one more way money matters. Guys like you have more money than the media companies. So instead of them hiring guys from the partisan polling world, I suspect it's the reverse: anyone with any talent in media polling gets snapped up by you guys.
Wilson: Yeah, there's a lot of truth in that.
The Line: Let's talk methodology a bit more. With all this anger out there about polling, the media polling companies talk about changes in public behaviour. We were hearing a decade ago about how young voters ditching landlines and just using cellphones was changing everything. Is that still true?
Wilson: Nah, not really. It costs more money to interview somebody on their cellphone, but it's still very doable. Cellphones are different than they were a decade ago. If you had a call on your cellphone 10 years ago, and you didn't know who it was, you're pissed off because they're costing you money and minutes. Today, it's just like a home phone. So it's a bit more expensive but it's not more difficult.
The Line: What do you think of the narrative that polling is struggling?
Wilson: It's not. Media polling is struggling. There hasn't been any major shift in the accuracy of partisan polling in decades. But the public only sees the media polling so their impression is of an industry in crisis. It's not the case for us. There's a real incentive difference here. If a media pollster blows a poll, he shows up at work the next day anyway and explains what went wrong and that's it. If I blow a poll, my clients fire me. [laughs]
The Line: We noted here at The Line a few days ago that media companies have lost just huge percentages of their workforce, and they're using polling to try and make up the difference. Instead of having reporters on the ground in a Midwestern state, they call a thousand people who live there, apply a formula and then report what everyone in that state is thinking. But if the model is slightly off, the reporting is way off.
Wilson: Yup. [laughs] I wanted to think of a more thoughtful response, but yeah. That's exactly right.
The Line: If you're advising the Democrats, what do you tell them was their lesson learned from 2020?
Wilson: Oh, that's simple. A truly moderate Democrat could win 40 states, but their embracement of socialism is going to destroy them as a party. They're going to marginalize themselves and end up in a situation where they are representing such a small segment of the population, that they risk going extinct. Look at Florida and Texas. That's a wakeup call to Democrats that I hope they ignore. [laughs]
The Line: Same question, but for the GOP?
Wilson: That's harder. It's a very different situation for the Republicans. A lot of the focus is going to be on the presidential race, and right now, that looks like it's going to Biden. But overall, the Republicans are doing very well. We're up in the House. We didn't lose the Senate, though that's still up in the air a bit. I'm not a conspiracy theorist but I have some concerns about what's happening in Michigan, I'd like more transparency there — I think John James would be a great senator and I think he should have won that. But overall? I think there are some House races that the Republicans could have won but didn't, and campaigns matter. I think there's a lot of Republicans left over from 1994 who think you win campaigns today the way you did then. You don't. Republicans are going to have to think a lot about Trump and the lessons. He pushed hard for minority votes in a way that Romney didn't, George W. Bush didn't, that McCain didn't. To find that you probably have to go back to Jack Kemp! Trump had so many major endorsements. 50 Cent, Kanye West, Jim Brown, Herschel Walker, and so many others. Criminal justice reform was a big issue that Trump actually delivered on. Was Trump as articulate as a Reagan? No, but he worked hard at this ... and it worked. And I've seen others do this. Ted Cruz does this in Texas, and he gets 30-40 per cent of the Hispanic vote. So does Governor Abbott [of Texas].
Trump showed us that if you do aggressive outreach to minority communities, get good surrogates with huge social media followings, and push hard, it can have a huge impact. Working-class minorities are a growth area for the Republicans.
The Line: And that advice is true for the Democrats too, but reversed. If they lose those voters, they're in serious trouble.
Wilson: Absolutely. Embracing Black Lives Matter but not strongly condemning the violent riots in the cities hurt them, even with Black voters. And stuff like "Latinx?" Hispanics hate that. Keep up stuff like this and they'll drive down their Hispanic vote share into single digits. [laughs] I hope they do!
This interview has been edited for length and clarity.
The Line is Canada’s last, best hope for irreverent commentary. We reject bullshit. We love lively writing. Please consider supporting us by subscribing. Follow us on Twitter @the_lineca. Fight with us on Facebook. Pitch us something: email@example.com