Getting To the Truth with Voters

Could studying subconscious responses from voters fix our broken polling?

Election ballot boxes with question mark and data poll graph lines for poll tracker concept background.

After two presidential elections and one midterm election where the polls were off—with Republican/Trump support underestimated in 2016 and 2020 and then overestimated in 2022—a common and understandable response among voters has been, “I don’t trust the polls!”

Who answers polls anyway? And are voters actually telling pollsters how they really intend to vote, at least in any reliable way? What about shy Trump voters, or the silent numbers of Haley Republicans for that matter?

I really appreciate it when researchers start to think outside of conventional norms and approach questions from a new standpoint. And that’s what researchers Shane Snow and Joe Lazer, who are also two of my favorite people and among the brightest guys I know, did recently. They set out to find a way to explore what is going on subconsciously with voters, all in an effort to find a more accurate way to assess their ultimate behavior. 

I found their work and the results fascinating, so I wanted to share our discussion with you. — Jay

LEFT: Joe Lazer. RIGHT: Shane Snow.

Q: Your approach on this poll is fundamentally different than typical polling, trying to get at what people think rather than what they say they think. How do you get at that notion in the poll, specifically explaining what the “Harvard Implicit Association Test” is and how it works? 

Joe: Well to start, I wouldn’t call it a poll — it’s more of a traditional research study, since our goal was to go beyond the polls and understand subconscious associations between the presidential candidates and certain attributes that would be revealing. I personally wanted to take this approach because I’ve been riding the “pollercoaster” every day and it’s making me nauseous. Are the polls wrong?! Of course they are. They’re wrong by 3 points on average, and in the US today, 3 points is everything. 

Also, humans are complicated! We don’t always say what we really think. I could ask you how many drinks you have per week, and you could tell me 5 when it’s really 15 because a) you’re embarrassed about how much you drink, or b) you really think you only have 5 drinks per week because you don’t keep close track. I’m pretty sure I lied to my doctor on this question during every physical in my 20s. The same goes for voting. One theory is that Trump overperforms the polls because there are secret Trump voters who don’t know they’ll vote for Trump until they get into the booth. But I think this is really more interesting when it comes to messaging and issues — people are really bad at identifying what actually motivates them to vote for someone. 

Our test aimed to understand the strength of people’s associations between different attributes. For instance, a strong association between “My Vote” and Donald Trump indicates you’re very likely to vote for Trump. A strong association between “Gets Me” and Kamala Harris indicates you feel a connection with Kamala Harris. The technology we used is called nPLICIT, developed by a neuromarketing research firm called Neuro-Insight based on the Harvard Implicit Association Test. The test itself takes about 10-15 minutes. Participants are asked to sort words and images during an online gaming task, with the speed of their responses indicating the strength of their associations. It uses nanosecond timers and only the responses that hit a Goldilocks zone where people are making subconscious associations count. Shane had studied this test before, and we thought it’d be a good way to get results at scale. 

Shane: We were motivated to do this study because four years ago we did an SST brainwave-scanning study with Neuro-Insight during the Democratic primaries, wherein we showed that the then-trailing Joe Biden was actually voters’ favorite even though he was far behind in the polls at the moment. When we measured liberals’ subconscious brain responses, people who said out loud that they preferred Elizabeth Warren or Bernie Sanders actually felt more warmth and association with Biden when they saw his advertisements. We published our study results on Medium after The New York Times and a bunch of other outlets declined to publish them, and after Biden ended up winning the primary various editors told us that they should have taken our study more seriously.

This time we used a different type of test to get a bit more granular than we did in 2020. Though it hasn’t been applied much to politics (to our knowledge), tools like nPLICIT have been used for years to predict consumer buying trends and to surface unconscious social biases, e.g., this test will surface the fact that if you grow up around certain kinds of people, you will subconsciously hesitate to trust people that aren’t like those you grew up with—even if you consciously try not to discriminate. And it’s shown how exposure to more people outside of your historic “in-group” slowly erases that subconscious hesitation or bias and leads you to actually not discriminate unfairly. In the case of the current election, we were interested in applying this study methodology to see if people who want to believe themselves when they say, “I’m voting for Harris because that’s the right choice” actually subconsciously thought of her as “radical” or “weak,” and/or if they had similar dissonance between how they talk about Trump vs how they perceive him.

Q: You decided to study both swing state voters and specifically a large group of young men of color. Can you explain why you chose these groups? What did you discover that was most surprising or not being picked up by standard surveys?

Joe: Well, as the political junkies who read your newsletter probably already know, one of Trump’s primary targets this cycle is young Black and hispanic men ages 18-35. He’s been blitzing the comedy podcast circuit to reach them — which is smart — and the polls indicate that he’s been making some ground, but there’s a lot of debate over how solid that support is. That was the first group we studied. The second group we studied was just voters from the seven big swing states because, well, they’re going to decide the election.  

Q: One thing that struck me was that the idea of who “Gets Me” is one of the indicia in your study. I’ve never thought of asking respondents whether a candidate likes them, versus the other way around. Can you explain this a bit further, and then perhaps give some examples of why this is important? What did you find?

Joe: So the inspiration for this is the Affinity Voting Theory. If you listen to Ezra Klein—which I do because he’s my Taylor Swift—Ezra talks about this a lot. It suggests that people are motivated not by whether they like a candidate or their specific policies but by whether they think the candidate would like them. They want to vote for candidates that get them, share their values, and can relate to their social, cultural, and economic experiences. This explains why Hillary’s “basket of deplorables” comment was such a shitshow — it made her seem like she didn’t like a good portion of America. 

Shane: The big thing for me here is when you look at the psychology of trust—something I’ve written about extensively over the years—it turns out that humans only use yardsticks like “does this person seem competent?” or “does this person seem capable of protecting me” when selecting their leaders if they either (a) have no other information about the person, or (b) believe that this strong, capable person is going to treat them with “benevolence.” Without benevolence, the more capable a leader is, the worse they could be for you. I call this The Terminator Theory. The Terminator (scary, murdery robot from the future) is extremely capable, always does what he says he’ll do, but you’re not going to trust him when he comes to kill you. However, when The Terminator comes back in the second movie and he’s now there to protect you, then you’re extremely happy trusting him. That’s the key thing when it comes to selecting leaders, and I think we see it play out very obviously with Trump in particular. People who believe that Trump’s got their back are very happy when he shows these displays of strength, even when he bullies people and shows that he’s not a nice guy. And of course, when you’re the one being bullied, the stronger the bully, the less you trust him. This is literally the appeal of a “strongman,” and the reason why people can see a strongman so differently.

The upshot is if a potential leader can show their constituents that they “get them” and will therefore likely treat them with benevolence, they’re going to earn those constituents’ trust. Telling people that you’re strong and powerful and smart and all that just shows them that you could outsmart or overpower them if your interests diverge. And we think that’s going to be a key factor in this year’s nailbiter of an election.

Q: Your conclusions sound a bit skeptical of the Trump campaign’s efforts to win over young men of color. What drives your skepticism, and what is the data telling you?

Shane: The data show that there’s a weaker association with “My Vote” and Trump for young men of color than for other groups. At the same time, there’s a strong association with “Gets Me” and Harris for this group. Which on the face of it is a bit surprising. But in the context of trust, it makes sense. These voters are essentially saying, “I like what this guy’s promising, but I dunno if he actually has my back.” 

The strong association with “Gets Me” and Harris means there’s a likelihood that these voters could come back around to Harris. Now what this study doesn’t show is whether there are other factors that we didn’t measure which could be more important than “Gets Me” to these voters. And it also doesn’t necessarily indicate whether these subconscious associations will lead to the effort to turn out and vote, or depress the vote among those who are subconsciously conflicted. My theory is that it’s more likely for young men of color who think they prefer Trump but who connect with Harris better to simply not turn out than to cast a vote one way or another. The upshot for the Harris campaign is they ought to lean into messaging and events that help her show authentically that she understands what young men of color go through and she has their back.

Q: Kamala Harris has seen her favorables rocket to positive levels in a short amount of time, even while Trump remains 11 points underwater on net favorables in some surveys. Does your data provide any insight into why this might be the case? 

Joe: This is an area where we’re getting more speculative, but my main theory is pretty simple. Even though she’s been Veep for 3.5 years, most of America didn’t really know Kamala. Or at least the voters who matter didn’t. You were only following the primaries in 2020 pre-Iowa if you were a locked-in partisan. She was in the background for most of the Biden presidency. Then Biden drops out, she gets elevated and BOOM — Kamala Brat Summer. Everywhere you look, there’s media coverage of her underdog rise. 

On social, you’re getting coconut-pilled to oblivion, you’re watching her get down and dish life lessons from her mom, and it all makes her seem really relatable yet strong. And in her interviews, speeches, and in the debate, she’s done a really good job of hammering home the story of her middle-class upbringing. All of that is contributing to her strong favorables and the strong connection she has with voters that we saw in the study. I’d love to see her hammer home her middle-class story even more, and tie it to a few more simple, easy-to-understand economic policies. She’s done this with housing, but let’s be honest, The Opportunity Economy probably sounds like BS to a lot of voters. 

Shane: The world-view for Trump is fairly locked in. I have family who still think Trump is a self-made billionaire, and for whom no amount of damning evidence of crimes or his disparaging of immigrants will make them believe anything different about him. In the CNN pre-debate poll, only 9% of voters said they wanted to know more about Trump, while 26% said they wanted to know more about Harris. And I suspect that a larger percentage of people who think they don’t want to know more about her simply don’t know her very well and are gaining a favorable opinion of her simply through the side-by-side comparison with Trump. That’s a large reason Biden won in 2020, in my estimation. People saw Biden as more warm and thoughtful next to Trump, who is not warm and not thoughtful. Harris is coming off as warm and thoughtful, if perhaps a bit calculating as Joe says. Which, if there’s anything to be concerned about on the trust front, coming across as phony is a landmine the Harris campaign doesn’t want to step on. 

This is something I’ve been thinking about lately with Tim Walz, by the way. When we see a politician kiss a baby or pretend to be interested in some guy welding at a factory, we roll our eyes and say “yeah, ok they’re pretending to care.” When you see Walz interact with voters, you see a guy who genuinely looks enthusiastic and who seems like he really does care. That’s showing benevolence, and that’s building trust. Not many people knew Walz before he was tapped to be Harris’s VP, but his favorability is high, and I think it’s because he’s seen as warm and genuine.

Q: Do you see pollsters taking up the idea of trying to get at subconscious rather than conscious behavior in voting? With many voters especially in the swing states remaining undecided, could this help the campaigns better understand how to sway them?

Joe: I’d love to see that. Amongst the external pollsters, I doubt it because it’s so much more expensive and time-intensive. It really doesn’t lend itself to up-to-the-second horse-race polling, which the media is addicted to. I could see campaigns using this more, though, since it works so well for understanding messaging and vulnerabilities. I’d also love to see them conduct brain-tracking studies on campaign ads, like the one we conducted in early 2020 that foreshadowed Biden’s come-from-behind primary win.

Shane: I definitely think we’ll see campaigns start to use this kind of study in the future in order to shape their messaging. Ad agencies and consumer brands use neuro-tracking to design Super Bowl ads; media companies use tools like Immersion Neuroscience to decide how to edit TV shows for greater engagement. Why wouldn’t political candidates pay attention to what people’s brains are saying? Everyone pretends that focus groups are accurate, while it’s widely shown that focus groups can be useful but are not usually accurate or representative of what people really think. What your brain actually says… is accurate.

Q: In 2016, Cambridge Analytica used psychological profiles to target people whom it believed were “persuadable.” That came under a lot of criticism. Do you have any concerns about how subconscious biases, once identified, might be exploited by technologies or bad actors to throw elections?

Joe: I think we’re comparing apples with soulless tech vampires here. People were mad about Cambridge Analytica because they used a personality quiz and a Facebook backdoor to steal the personal data of all your Facebook friends. The data from the quiz was relatively insignificant — what mattered more was all the data they got from illegally stalking you and your friends online. 

To me, this is more of an alternative to polling and focus groups, which we already use to create psychological profiles and figure out what messages will work to persuade people. That all being said, you could imagine bad actors using this technology to identify racial or identity biases to exploit, and that’s certainly a little concerning, as we know they’re doing this through other means already. 

Shane: Every campaign (whether political or in advertising) develops “personas” so it can tailor messaging to people. So whereas I think one could attempt to smear a competitor by saying “They’re tracking your brains!”, the easy reply would be, “We got volunteers to let us study them so we could understand what the people really want, and we’re learning from you.” And like with any technology, if one side gets an advantage from it, the other side is going to adopt the tech too.

Q: This election is going to be won or lost on enthusiasm, meaning turnout before and on Election Day. Does anything in your methodology tell you about how likely it is voters will head to the polls?

Joe: In the swing state voter group, candidates on both sides seem locked into their candidates with strong conviction. I’m tempted to say that it indicates we’re headed toward a high-turnout election, but we don’t have the baselines from past elections to compare it to. The most optimistic area for Harris is amongst young men of color, where Democratic and Democratic-leaning voters are much more locked into her than Republicans are to Trump. 

Q: Do you find the results of your study interesting enough to continue pursuing this approach? What’s next for this project?

Shane: Certainly for the next big election, we’d like to do an even bigger version of this. I’d personally love to combine the SST (brainwave tracking) tools that we used in 2020 with the nPLICIT (implicit association tracking) methodology we used this time to get an even more robust picture of voters’ subconscious biases toward candidates earlier in the campaigns. And I’d like to repeat these studies at regular intervals in order to show movement and draw conclusions about how campaign messaging and events are influencing how voters think at the subconscious level. It’s expensive, so we couldn’t continue doing this on our own without a backer of some sort. But I suspect there are folks out there who will be highly motivated to learn the kinds of things we’re showing are possible to learn with these tools.

Joe Lazer is the best-selling co-author of The Storytelling Edge, a contributor to Fast Company, and writes a popular newsletter on the art and science of storytelling with over 150,000 subscribers. Shane Snow is a prolific journalist, Tony-winning producer, and bestselling author whose work has appeared in GQ, Wired, The New Yorker, and more. He is currently the CEO of the film technology company SHOWRUNNER and is a keynote speaker on innovation, storytelling, and leadership.