Polling is “90% bullshit.” In fact, all political data is “garbage.” The Democratic consultants who traffic in such numbers are perpetrating a “scam” against their own party and are largely responsible for President Donald Trump’s victories. Instead of trying to gauge public opinion through pseudo-scientific surveys, Democrats should mostly just read history and the classics.
This is the gist of John Ganz’s recent column, “Against Polling” — a widely-shared polemic that actually earned plaudits from some Democratic pollsters.
Both this piece and its reception are puzzling. Ganz is a brilliant writer with many insightful things to say about history and political philosophy. (I recommend subscribing to his newsletter and buying his book.) Yet his diatribe against “data” is unfair and unpersuasive. He patently mischaracterizes the positions he’s arguing against and provides little evidence for his own. He does not acknowledge some obvious objections to his anti-empiricism, let alone rebut them. His piece’s valid assertions are uncontested while its contentious ones are unvalidated.
Nevertheless, it was warmly received, even by some whose vocations it disparaged.
I’m not certain why this is. But I fear that Ganz’s argument is appealing for the very reason it undermines clear-eyed thinking about electoral politics: it offers an elaborate rationalization for dismissing any data one does not like.
Ganz makes some valid points (that no one actually disputes)
Ganz’s critique is aimed at both political data in general, and a specific set of ideas about Democratic electoral strategy: Principally, the notion that Democratic candidates should seek to increase the salience of their popular positions, avoid talking much about their unpopular ones, and give greater deference to public opinion than the party presently does. This basic outlook is often described as “popularism” (a hideous but useful neologism). And it is championed by, among others, the Democratic data scientist David Shor, the commentator Matt Yglesias, and less prominently (and more equivocally) myself.
In prosecuting his case against popularism, Ganz says many things that are inarguably true. For example — after spending the bulk of his column arguing that public opinion data is “garbage” and “90% bullshit” — he retreats to the claim that polling is “part of getting a picture of the world,” just “not the entirety of it.”
Needless to say, the idea that polling shouldn’t be your only tool for discerning reality and the idea that polling is almost entirely fraudulent are pretty different. The first claim is indisputable; the trouble is that no one disputes it.
This is the problem with virtually all of Ganz’s valid assertions. He correctly observes that polling is flawed, that public opinion isn’t fixed, that not all useful knowledge about politics is quantifiable, and that there is more to good campaigning than mirroring the public’s policy preferences. But he does not quote a single Democratic consultant or commentator who rejects these truisms, likely because none do.
Instead of refuting the popularists’ actual ideas, Ganz rebuts an absurd ideology of his own invention. He writes that “the worldview of the data guys is based on a giant mistake” — namely, that “there’s an objective world out there, and it doesn’t change.”
Yet no serious person has ever claimed that public opinion doesn’t change. It obviously does. And this is not lost on the “data guys.” David Shor, to take one example, has argued that Democratic politicians have the power to reshape many of their base voters’ views, that the Dobbs decision made Americans more liberal on abortion, and that “what people care about and trust [the Democrats] on really is responsive to concrete events that happen in the world.”
On this point, the actual dispute between Ganz and the popularists is not about whether public opinion can change, but about how much scope Democratic politicians have to reshape the views of swing voters — which is to say, voters who do not particularly trust Democratic politicians.
Everyone recognizes that this scope is limited. Most progressives would doubtlessly agree that Democrats can’t persuade swing voters to support large new taxes on meat. There may be a strong moral case for making steak more expensive, given the cruelty and ecological harms inherent to large-scale animal agriculture. Were Democrats to campaign on the case for making meat less affordable, however, they would surely do less to change swing voters’ views on factory farming than to poison the Democratic Party’s image. I seriously doubt that Ganz would contest this.
Assuming he doesn’t, then the debate on this point isn’t about whether Democrats must acquiesce to the public’s existing preferences on some subjects. Rather, it is about 1) what those subjects are and 2) how they can be identified.
These are difficult questions. To answer them, one must make not only empirical judgments but normative ones (chiefly, about how Democratic politicians should weigh the risk of alienating voters against the benefits of evangelizing for worthy causes). It is reasonable to argue that Shor, Yglesias, or any other “data guy” gets these questions wrong. But demonstrating that requires engaging with their actual premises, not shredding a caricature of their worldview.
Did Democratic pollsters misjudge the politics of immigration?
Ganz comes closest to addressing the data guys’ actual views in his discussion of immigration.
In 2024 — and during the early months of Trump’s presidency — popularists encouraged Democrats to focus on the electorate’s economic concerns, rather than emphasizing the moral case against Trump’s mass deportation plan. Ganz argues that this was a mistake, one rooted in an overvaluation of polling data and an underestimation of the electorate’s moral judgment.
His argument is as follows: Polling in 2024 indicated that voters agreed with Trump about mass deportation. This led the popularists to discourage Democrats from attacking the cruelty of Trump’s vision. But surveys on this subject were fundamentally misleading: Most voters did not harbor any deep commitment to purging the country of hardworking, law-abiding immigrants. And once people actually saw what Trump’s policy entailed, they recoiled in horror. Now, a majority of Americans disapprove of the president’s handling of immigration.
In Ganz’s view, we should not blame the public for failing to anticipate the consequences of Trump’s agenda, but rather, the “lazy politicians” and “craven advisers” who prevented Democrats from alerting voters to those harms. By refusing to appeal to the electorate’s “faculties of judgement and imagination,” Ganz suggests that the “data guys” abetted Trump’s election.
There are many problems with this argument. But most derive from two unsound assumptions at the core of Ganz’s reasoning:
• If voters soured on Trump’s immigration agenda after witnessing its effects in 2025, then Democrats could have changed their minds about that agenda in 2024, had the party only helped Americans picture those effects.
• If voters disapprove of Trump’s handling of immigration, then the Democratic Party must not have any interest in reducing that subject’s salience relative to economic issues.
On the first point, it simply isn’t true that Democrats failed to warn voters about what mass deportation would entail. To the contrary, on the campaign trail last September, then-Vice President Kamala Harris explicitly appealed to voters’ “faculties of judgement and imagination” on this subject, warning that Trump had “pledged to carry out the largest deportation, a mass deportation, in American history. Imagine what that would look like and what that would be? How’s that going to happen? Massive raids? Massive detention camps?”
At the time of these remarks, Democrats had been making versions of this argument for nearly a decade. At one point during that period, Trump had used presidential power to separate migrant parents from their children, in a notorious scandal. If all this wasn’t enough to persuade swing voters that Trump’s approach to immigration was intolerably cruel, why should we be confident that Harris could have convinced them of as much, if only she’d made the case more forcefully?
Ganz offers evidence that public opinion on immigration is sensitive to shifts in objective conditions and media coverage. But it doesn’t follow that such opinion is highly responsive to Democratic rhetoric.
Yet there is a more basic problem with Ganz’s case: Current polling still indicates that immigration is a source of relative strength for Trump and weakness for the Democratic Party.
In RealClearPolitics’s polling average, immigration remains Trump’s best issue, with voters disapproving of the president’s handling of that subject by only 5 percentage points as of August 15. By contrast, voters disapprove of Trump’s management of inflation by 20.5 points, and of the economy more broadly by 11.6 points as of the same time.
More critically, some recent polling indicates that voters still prefer Trump’s cruel brand of immigration enforcement to the Democratic Party’s perceived laxity on the issue. In a July Wall Street Journal poll, voters said that they trust Republicans more than Democrats to handle immigration by a 17-point margin. On illegal immigration, meanwhile, voters favored the GOP by 24 points.
Therefore, the popularists’ basic proposition on this subject — that Democrats have an interest in emphasizing economic issues over immigration — is as plausible today as it was in 2024.
To be clear, none of this necessarily means that Democrats shouldn’t spotlight the cruelty of Trump’s immigration policies, for non-electoral reasons. In my own view, the popularists can get monomaniacally fixated on political optimization, at the expense of other considerations. Democratic Sen. Chris Van Hollen’s decision to agitate for the release of Kilmar Abrego Garcia — an immigrant whom Trump unlawfully deported to El Salvador — may or may not have marginally increased the Democrats’ odds of winning the 2026 midterms. But doing so brought comfort to a longtime US resident who was suffering in a nightmarish foreign prison. That seems worthwhile.
Yet Ganz does not frame his case against popularism in purely moral terms. Rather, he asserts that the “data guys” are specifically wrong about how to win elections — so wrong that their polling amounts to “garbage” and a “scam.” His only evidence for this extremely strong claim is that public opinion about immigration has changed since 2024.
In other words, Ganz is guilty of precisely what he accuses the “data guys” of doing: drawing sweeping conclusions on the basis of a myopic and incurious reading of cherry-picked polling data.
If empiricism is dead, then everything is permitted
It is easy to enumerate the liabilities of polling and election data analysis. The real difficulty lies in naming alternative methods for ascertaining political reality that are more reliable and less vulnerable to motivated reasoning. And Ganz fails to meet that challenge.
He argues that politics is an art — not a science — and can therefore only be mastered through humanistic methods: There is no alternative to studying “the words and actions of politicians past” and the vast philosophical literature on effective rhetoric, beginning with the ancient Greeks. FDR’s oratory may not offer the precise guidance of polling, message tests, or elaborate statistical analyses of election results. But Ganz argues that such number crunching can obscure more than it reveals by abstracting away essential context, which can only be captured through a qualitative examination of political history and the classics.
It’s surely true that some useful political knowledge can’t be represented mathematically, and that history is an indispensable supplement to political science. Yet as a tool for anticipating how voters will respond to a given agenda or message, the humanistic study of “politicians past” has obvious flaws.
For one thing, as Ganz himself emphasizes, public opinion changes over time. Rhetorical tactics and substantive positions that worked in the past may therefore have less purchase in the present. Polling can offer a portrait of contemporary attitudes; history can’t.
Further, what history and the classics tell us about optimal electoral strategy in 2025 is extremely indeterminate: By focusing on distinct historical examples or emphasizing different pieces of context, one can draw an enormous variety of different conclusions.
Of course, one can apply quantitative tools in biased ways. But scientific methods impose far greater constraints on motivated reasoning than humanistic inquiry does. Rigorous polling can falsify one’s assumptions about public opinion (or at least, cast them into doubt). Analyses of which candidates have outperformed their party in recent elections can validate or undermine certain theories of political best practice. By contrast, no one has ever learned that their policy preferences were unpopular by reading Aristotle.
Ganz’s piece unintentionally illustrates his method’s susceptibility to biased reasoning. To appreciate how, it’s worth quoting his conclusion at length:
The statistical fixation of the early 21st century that’s made so many bad predictions and fathered so many puzzling defeats must be abandoned. We are not in an era of small calculations but of great movements. Politicians with a vision and a strong, clear rhetorical appeal, like Trump, Bernie, AOC, and now Mamdani, are those who excite people. People become disappointed and disillusioned when they lapse into focus-group-tested canting. Say something for a change.
It’s difficult to say exactly what Ganz is asserting here, since his language is a bit vague and aphoristic. If you squint, you could read him as merely asserting a bunch of truisms (it is bad for a politician to speak in canned lines that sound inauthentic; charisma matters; having an energized base is desirable, all else equal).In the most straightforward reading, though, Ganz appears to be making at least three contentious propositions:
• Democrats’ recent defeats were caused by excessive deference to public opinion data.
• Trump’s success demonstrates that paying close heed to polling is less important than offering a vision that excites people and mobilizes a “great movement.”
• For a model of the type of Democratic politics that works in our era, one should look to Sen. Bernie Sanders, Rep. Alexandria Ocasio-Cortez, and New York mayoral candidate Zohran Mamdani.
None of these claims are self-evidently true. On the first point, one could just as easily assert that Hillary Clinton and Kamala Harris lost because they paid too little attention to opinion polling, rather than too much. After all, both declined to embrace many majoritarian positions that contradicted progressive principles. And Harris famously ignored the counsel of data-driven Democratic consultants, who had implored her to focus less on “democracy” and more on the cost of living.
Similarly, it’s not obvious that Trump has been more successful than a less charismatic, but more moderate and disciplined, Republican would have been in his place. It’s worth remembering that Trump 1) lost the popular vote by 2 million ballots in 2016, despite running against a historically disliked Democratic nominee, 2) lost the presidency in 2020, despite running against a senescent man who could not reliably speak in coherent sentences, 3) won narrowly in 2024, even with the tailwind of a global anti-incumbent backlash, and 4) through it all, has had an unusually low approval rating (for a president or party leader).
It is entirely possible that Republicans would have done much worse over the past nine years, had they been led by a straitlaced moderate with poorly attended rallies. But how does Ganz know that?
Finally, it’s difficult to see a firm basis for selecting Bernie Sanders, Alexandria Ocasio-Cortez, and Zohran Mamdani as exemplars of effective Democratic politics in 2025. All are gifted politicians. And I think there are things that the broader Democratic Party can learn from each. Yet they have collectively won zero general elections outside of New York City and Vermont — two of the most Democratic jurisdictions in America. And Mamdani has a negative 9 percent favorability rating in New York state, according to a Siena College poll released this week.
The challenge facing Democrats today is not how to win a New York City mayoral election, but how to win presidential races in Pennsylvania and Senate contests in North Carolina. And Ganz offers no evidence that appealing to majoritarian opinion in those places (as measured through scientific methods) is less important than exciting “great movements,” in the manner that Sanders, AOC, and Mamdani have.
All of which is to say: By dismissing “scientific” methods of evaluating claims about political reality, Ganz enables himself to draw strong conclusions about how Democrats can best disempower the GOP — without providing much in the way of substantiation or even argument. And those conclusions happen to be ideologically convenient for Ganz, whose social democratic politics are well-represented by Sanders, Ocasio-Cortez, and Mamdani.
This is what I find insidious about his whole argument: It serves to insulate progressives’ intuitions about electoral politics from any empirical challenge.
Progressives have strong incentives to engage in wishful thinking about electoral politics
This is a problem, not least because the left’s political intuitions are liable to be biased, unless disciplined by data.
Left-wing activists and public intellectuals have strong incentives to believe that there are minimal tensions between the progressive movement’s factional project (to pull Democrats leftward) and the Democratic Party’s electoral one (to disempower the Republican Party). If there were large trade-offs between these two endeavors, then such progressives might have a moral obligation to counsel some form of moderation. After all, as the Trump administration demonstrates on a near-daily basis, the stakes of keeping the authoritarian right out of power are extremely high.
Yet one cannot advocate for a more ideologically cautious Democratic Party without risking estrangement from other progressives. In principle, there is no reason why it couldn’t be true that social democracy is the most just political system and that — at this particular point in history — the Democratic Party would win many more elections if it moderated on some issues. The latter is an empirical judgment, not a normative one. Yet to articulate this view as a progressive is to jeopardize your sense of belonging and esteem among those who share your moral commitments. Doing so is sure to get you derisively branded as a “centrist.” Some progressive writers may even feel comfortable calling you a “craven” scammer without evidence.
Thus, anyone who finds community and identity in progressive politics — a group that includes a large share of Democratic operatives, staffers, and commentators — is liable to err on the side of underestimating the political utility of moderation. Polling and election data are the only real checks on such a bias. And Ganz’s piece provides a rationalization for dismissing them.
None of this is to say that the popularists are necessarily right about how Democrats can win elections, much less about how the party should balance the dictates of principle against those of political expediency. The latter is an inescapably value-laden question, which data cannot resolve. Yet you can’t formulate a morally serious answer to that dilemma without a plausible conception of political reality. And scientific methods remain our best tools for forming such a conception. For all of polling’s flaws, there is still no better way to find out what voters think than to ask them (or more specifically, a representative sample of them, using maximally neutral question wordings). And for all the methodological disputes among political scientists, there is still no better way to discern which types of candidates most voters favor than to rigorously examine whom they elect.
The post A dangerous new idea about what Democrats are doing wrong appeared first on Vox.
