Bringing a Survey to a Gun Fight, Part I
“Pollingism” Has Failed Democrats and Voters. Tuesday Showed the Power of Magnetism.
This piece first appeared, in much longer form, in Michael Podhorzer’s Weekend Readings.
In the wake of Tuesday’s Democratic rout, we see a flurry of takes on what lessons this offers. Indeed, if directives about how Democrats ought to campaign were an energy source, they could power America. But what they deliver in volume, they often lack in depth. Despite the data behind many of these dives, they fail to consider whether the methodology used actually tells us what we need to know to win elections. Let alone enact a governing agenda, which is the actual purpose of politics.
There are two different ideas about what it takes to achieve political victory, but only one gets real airtime. This model views voters as rational actors, making electoral decisions based on their conscious issue preferences. Pollingism, as I’ve coined it, assumes said preferences are static, discernible via testing, and hold ultimate sway over voting behavior. Pollingism views the political task as winning elections and treats the work of governance as something to be hashed out later.
Pollingism proponents believe that data “shows what voters really think, not what people who work in politics wish they thought.” (Nevermind that folks making this claim work in politics.) In their minds, the data have set them free from their biases, including holding fixed stances on right and wrong. The trouble with this is that data aren’t conjured but rather solicited. In other words, you only get answers to the questions you ask. And you only get reactions to the ads you produce. And you only assess impacts in the artificial environments you construct.
An alternative worldview understands voters as social beings, driven by unconscious cues. The refrains of friends, family, and trusted messengers, as well as voters’ identities, hold the greatest sway in their actions. Political persuasion happens through offline interactions, persistent media narratives, and social movements, as well as formal campaigns. This approach, which I call Magnetism, proceeds from the notion that if you want people to come to your cause, you must be attractive. This requires having a cause to which to draw people. And it also means having a polarity that distinguishes you from your opposition. For Magnetism to work, you meet people at the place of their broadly shared values, not their podcast-promoted prejudices.
In short, strategic campaigns begin by asking the question: How do I force the issues and conversations that most benefit my side to the fore?
A Message Nobody Hears Cannot Move Them
At first glance, Pollingism espouses a simple truth: ask people what matters to them and repeat it back. It often gets short-handed as “say popular things” and, by extension, eschew unpopular ones. This dominant approach to campaign message crafting and evaluation involves (1) randomized-controlled trial (RCT) testing of messages and/or ads, and (2) issue-based polling to figure out which topics voters rate most vital and therefore merit airtime. But this fails to consider, let alone optimize for, how we get a message to be heard.
What Happens In-Channel, Usually Stays In-Channel
RCTs and surveys have their utility; I’ve helped conduct them for nearly 15 years. But you cannot wield a thermometer to change a tire. There are inherent limitations to how experimental results apply in the real world of campaigns.
RCTs and sequential surveys occur “in-channel.” People are recruited to respond (usually through online solicitations), made to pay attention (with financial incentives or due to how testing platforms are configured), and register their preferences knowing they’re being observed.
Responses to a single exposure of a message are declared to be the message’s effect. But messages aren’t single-use plastics. Many don’t work until they are repeated. Some are more rebuttable than others. Some could work if spoken by messengers other than those in the materials tested.
Further, there are problems with what Pollingism deems worthy of measuring. While the fixation on vote choice may seem logical — the task come election day is to net more votes — this simplifies how one reaches that goal. Often, what you need to measure is what would make the story that favors your side seem credible, what punctures the key strength of your opponent, or what — only upon unrelenting repetition — reframes the electoral stakes to your advantage.
Humans Are Social Creatures
Pollingism tests content among a general electorate, optimizing for what “moves” engaged respondents after a single view, nearly always without exposure to the opposition messaging on repeat in the real world. If you’re scoring by what the greatest number of people in your sample — including those who will never vote for you, despite what they report in an RCT — find palatable, you’ll come up with something that doesn’t offend in testing but also doesn’t prove memorable once aired.
The choir — not the candidate — carries the most credible tune
For Pollingism, focusing on what they label “the base” is “preaching to the choir,” an activity viewed as not just unnecessary – under the theory that progressive Dem voters have nowhere else to go – but detrimental.
Yet, in the real world, it’s said “choir” that makes joyful noise for the congregation, which goes out and converts new adherents. Further, there is another place that voters who revile the opposition can go: the couch. In every election, there are (at least) three candidates running — yours, theirs, and not voting at all.
Pollingism assumes that what wins elections is persuading the “median voter” that your candidate aligns better with them on their most important issues than the opponent does. For this to hold, voters would have to know candidates’ positions on said issues and/or believe candidates’ claims about them. All of this rests upon the fiction that what people believe about a Democrat running is made out of what that Democrat, or their surrogates and Super PACs, say.
Politics is a shouting match, not a soliloquy — not just for who gets heard, but who gets believed. Especially higher up the ballot, direct communication from a candidate or aligned-organization is a fraction of the information a person receives. Americans mainly get their cues about politics from family, friends, acquaintances, and media they trust. What people believe about politicians comes from what is said about them. Yet, Pollingism proponents advise spending hundreds of millions of dollars on end of cycle ads. And, in this age of rampant disinformation, fail to deploy research methods to even consider what would make campaign claims seem more credible, including which messengers would need to utter them and at what frequency.
People believe what they think people like them believe
Very few campaigns effectively mobilize credible messengers on their behalf, but “yes we can” do so. One-on-one conversations to turn people into leaders who spread the word to others require serious resources and commitment. This is the kind of work that evangelical churches do for the right and unions used to do even more extensively for the left. And it’s part of what made Zohran Mamdani’s campaign and the attendant record-breaking voter turnout in New York City so incredible.
But the dominant Democratic campaign model is to mostly use paid advertising and provide some often non-partisan funds to community groups to do “field,” registering voters and/or getting out the vote (GOTV). This approach uses poll-tested talking points in “off” years and polished content in campaign season, where operatives who also do production and ad buying control each word uttered and image used.
Republicans’ approach has long been completely different. They test for the “red meat.” That is, a message that keeps their base engaged and enraged — ready to repeat a refrain until the muddled middle has it cemented in their minds. This is the precursor to the “manosphere,” in the form of aforementioned evangelical churches, supper clubs, NRA affiliates, and so on (long before podcast was a word).
Trump took this basic formula and injected it with steroids. He throws “eating the dogs and cats” at the wall and sees what sticks. Trump won over “persuadable” voters in 2016 and 2024 not by discerning their issue preferences and moderating his message accordingly, but by motivating and equipping his base to carry his tune. When you are a low-info voter, seeing people you’ve known forever wearing the MAGA hat tells you that this is what people like you believe. And so you do.
Magnetism, in understanding that voters — like all humans — are social creatures who come to judgments on the basis of what their identity group holds as common sense, requires deeper exploration of and attention to people who have previously voted Democratic, regardless of their race or age, educational degrees or origins, or even stated partisan preferences.
I emphasize this because, generally, Pollingism adherents view voters through the lens of their demographic categories, with race and educational attainment at the fore. They accurately note that key groups are shifting away from Democrats and in this find evidence that politicians aren’t saying popular things. But instead of zeroing in on the topics and tactics that would drive the voters who last brought us to victory (and not just those who match their demographics) to the polls, the same Democratic campaign operatives who lament our lack of message reach keep producing content aimed at the whole electorate.
PART II of “Bringing of Survey to a Knife Fight will post Monday, November 10




I love that you have published an Anat Shenker-Osorio piece at the Contrarian. I hope you have her on to speak with Jen sometimes. I always find the things she talks about insightful, clear, and important.
Once again excellent. Because it is true. Any industrial or political organiser can attest to its truth. From you to someone’s ears! 🤞