The Billboard Won't Save You

Part 2 of a series of four posts, see part 1 here

I want to be precise about what I'm arguing here, because the internet has a way of flattening nuance into combat, and that's not the exercise.

Sarah Longwell is good at this. She runs The Bulwark, she does serious focus group research with actual persuadable voters, she has spent years doing the grinding work of building an audience for anti-authoritarian conservatism at considerable personal and professional cost. I have real respect for what she's built and what she believes. The Secret podcast she does with JVL is genuinely one of the better things in political media right now.[1]

She is also, on the question of whether persuasion is the primary engine that can pull us out of this, working from a model that the data has been quietly demolishing for about three years. And because she's smart and her thesis is appealing and her billboards are going up across the country as we speak, it seems worth saying plainly: the persuasion lane is narrower than she thinks, the non-voter lane is not what she's assuming, and the information environment has eaten the mechanism that makes her model work.

Let's do the numbers first, because the numbers are genuinely clarifying.

The 'Thirds' Problem

Here is the actual shape of American electoral politics, expressed as simply as possible.

Roughly 245 million Americans were eligible to vote in 2024. About 155 million of them did. Trump got approximately 77 million votes — call it 31% of the eligible electorate. Harris got approximately 75 million — call it 31%. The remaining ~90 million eligible voters stayed home, roughly 37% of the pool.

So: one third voted MAGA, one third voted Democratic, one third didn't vote. That's the map. That's what we're working with.

Now, the persuasion thesis says: Trump's numbers are dropping (true — he's hitting new second-term lows, net approval around -18 in current aggregates)[2], people are feeling economic pain (true), therefore persuadable soft Trump voters can be moved. And the non-voter pool is a reservoir Democrats can tap. Right?

Here's where it gets uncomfortable.

Pew Research found that in 2024, non-voters split 44% Trump preference, 40% Harris preference. Essentially a coin flip. In 2020, non-voters split 46% Biden, 35% Trump — a clear Democratic lean that has since evaporated. The non-voter pool is not a Democratic reservoir sitting there waiting to be unlocked by the right message or the right candidate. It is a genuinely mixed group, and the portion of it that leans Democratic has either been shrinking or is simply harder to mobilize than the Republican-leaning portion.

That Pew number is the one that should be tattooed on the wall of every Democratic strategy session. The cavalry is not coming from the sidelines. They don't know whose side they're on.

The non-voter pool is not a Democratic reservoir sitting there waiting to be unlocked by the right message. It is a genuinely mixed group, and they don't know whose side they're on, or even if they are playing the game.

The 40% Floor Is Not a Performance Review

The second problem with the persuasion model is what it assumes about the Trump coalition that hasn't moved.

His aggregate approval has been grinding down — 37 to 40% depending on the poll — but that floor has been remarkably sticky considering what's been placed on top of it. Two impeachments, a criminal conviction, January 6th, four years of chaos, a first term of norm-breaking that would have ended any prior presidency twice over, and now tariff-induced economic pain that voters are feeling directly at the pump and the grocery store. None of it has cracked that floor in any durable way.

Sarah's model says: people are feeling the pinch, attach the pinch to Trump, run the "ballroom not your groceries" message, and moveable voters move. It's not a stupid theory. The problem is that the voters still in that 37-41% band aren't doing a performance review. They've incorporated MAGA into their identity. Telling them Trump's ballroom is expensive is not received as information — it's received as an attack on them. The brain doesn't process identity threats as evidence. It processes them as aggression and responds accordingly.

The identity threat research on this is pretty unambiguous. When a significant chunk of the population feels their cultural status is under threat, they consolidate around that identity with a ferocity that overrides other political preferences, including economic ones. We watched this happen in real time after the George Floyd summer. A moment that moved 52% of Americans — briefly — produced a Republican backlash that delivered Virginia in 2021 and moved Latino men toward Trump by 2022. The persuasion ran backward.

The voters who are genuinely moveable by economic pain — who were soft Trump supporters and are now genuinely reconsidering — are real. They exist. They moved his numbers from the mid-40s down to where they are now. But they are not infinite in number, and the ones who haven't moved yet are increasingly the ones for whom identity is doing more load-bearing work than economics. You can't billboard your way through that.

The Echo Chamber Eats the Message

Here is the mechanism problem that Sarah's model doesn't fully account for.

Her persuasion framework is implicitly designed for a shared information environment. You make a compelling argument, it reaches people who haven't hardened against it, some of them update. That's how persuasion has worked in every era of American politics up until roughly fifteen years ago.

The actual media ecosystem doesn't work that way anymore. "Trump's ballroom, not your groceries" is a good line. It'll reach people who already agree with you efficiently and completely. It will reach soft Trump voters inside a Fox/talk radio/MAGA podcast information environment approximately never, or it will reach them pre-rebutted — wrapped in a counter-narrative that explains it as liberal hysteria, fake news, or Democratic projection. The inoculation arrives before the persuasion can take hold, and it arrives faster and louder and with more emotional resonance than a billboard on a highway.

The identity threat literature and the media siloing research point in the same direction: a message designed to shift a voter's economic calculus has to first get past the tribal processing layer that decides whether the message is credible. In a fragmented media environment, the tribal processing layer is effectively impenetrable for a significant portion of the electorate. You're not delivering information to a person. You're trying to slide a note under a door that has been specifically reinforced against notes from your direction.

This is not Sarah's fault. She didn't design the media ecosystem. But it does mean her model is running the right play in the wrong stadium. As they say, it is fighting the last war.

You're not delivering information to a person. You're trying to slide a note under a door that has been specifically reinforced against notes from your direction.

The Timing Problem Nobody Talks About

There's one more structural issue that gets almost no airtime in these discussions.

Persuasion — real persuasion, the kind that shifts deeply held identity-adjacent beliefs — works on long timescales. Repeated exposure, cognitive dissonance building over months, a gradual permission structure that allows someone to update their self-concept without feeling like they've lost a war. That's not a sprint process. That's a marathon process.

But the political calendar runs on sprints. The 2026 midterms are in November. The persuasion window is essentially right now, and whatever doesn't move by September is probably not moving. For that compressed timeline to work, you'd need voters who are already mostly there — already experiencing the cognitive dissonance, already quietly reconsidering, already looking for permission to move. And some of them are. But the question is whether there are enough of them in the right districts to matter, and whether the mobilization of Democratic base voters can happen alongside and not instead of the persuasion play.

The 2018 midterms are Sarah's proof of concept, and it's a legitimate one. Suburban voters moved dramatically, Democrats took the House, the environment was real. But 2018 also had a specific ingredient that's harder to replicate: a relatively unified outrage response to a first-term Trump who was still novel enough to shock. By the second term, the shock has worn off. The people who were going to be shocked are already Democrats. The people still in the coalition have processed four more years of this as normal.

What The Model Gets Right (And Why It Still Matters)

Look, I'm not saying stop making billboards. I'm not saying persuasion is zero. And I'm genuinely not saying Sarah is wrong that Trump's economic numbers are his most vulnerable flank — they are, and hammering them is the correct play compared to the alternatives.

What I'm saying is that the persuasion lane is probably narrow enough that it cannot be the primary theory of the case for Democratic recovery. It can be part of the picture. The soft Trump voter who's genuinely feeling the tariff squeeze and hasn't yet hardened into pure identity politics — that person is real and worth the effort.

But the thing Sarah's model requires — a significant, durable shift of enough voters in enough places to produce governing majorities capable of actual structural reform — that math is very hard to make work through persuasion alone, especially given what the Pew data says about the non-voter pool. You need that pool to be a Democratic reserve. It isn't, not anymore, and the 2024 turnout data on why Harris underperformed Biden by six million votes while Trump gained three million tells a story about mobilization failure that no billboard campaign addresses.

The honest version of this is: persuasion is necessary but nowhere near sufficient. The scale of what's required to actually shift the structural trajectory — not just win a midterm, but generate the kind of governing majorities that could do the ugly, bare-knuckled institutional work we discussed in the last piece — exceeds what any realistic persuasion operation can deliver.

Which is a hard thing to say, and a harder thing to sit with. But you can't build a route out of this if you won't look honestly at the terrain.

The map is the map. The billboard is a billboard.

Next up: How we got here. The slow demolition of pre-political virtue from the robber barons through Milton Friedman to the tribal sorting that made all of this feel, in retrospect, inevitable.

1 - The Secret, and The Triad are why I continue to pay for the sub to The Bulwark

2 - this is the first time Nate Silver's aggregate of Trump's net approval has cracked below 40%. Something I will keep an eye on. It is still far higher than the "Bush" line, and a galaxy away from the "Dick" line.