archive.md
The Problems with Polls | Samuel Earle | The New York Review of Books
October 17, 2024, issue.
30 - 39 minutes
https://archive.md/Y7AmQ
The twenty-first century was supposed to be a new golden age for political polling. In 2008 Nate Silver, a thirty-year-old sports journalist, became an overnight celebrity after predicting Barack Obama’s election victory with uncanny accuracy, calling forty-nine of fifty states correctly on his personal website, FiveThirtyEight. His method was to aggregate multiple polls, weight them based on various factors, and then subject them to the kind of forensic statistical analysis used to evaluate the performance of baseball players.
In the 2012 presidential election, Silver went from celebrity to sage. He picked the winner in all fifty states while traditional pollsters delivered mixed results. “You know who won the election tonight?” asked Rachel Maddow. “Nate Silver.” According to Marie Davidian, the president of the American Statistical Association, the reason Silver “could predict the election perfectly” was simple: “dispassionate use of the data.” The New Republic declared that it was “1936 all over again”—a reference to the year that launched modern polling, when pollsters like George Gallup and Elmo Roper predicted Franklin D. Roosevelt’s victory, upstaging more old-fashioned election forecasts that he would lose to Alfred Landon. Their innovation was the sample survey—gathering responses from a group of people deemed representative of the entire population, according to characteristics such as age, gender, and race, rather than gathering as many responses as possible through much larger but untargeted opt-in surveys or straw polls. Silver’s innovation was to bring the sample survey into the age of big data.
The excitement of 2012 proved short-lived. In the 2016 election, polls were ubiquitous—by one count, television networks discussed election forecasts around sixteen times a day—but Donald Trump defied almost all their predictions and won the presidency. Worse than that, the polls were accused of enabling his victory by creating a fog of complacency that inadvertently sank Hillary Clinton’s candidacy. In his book A Higher Loyalty (2018), for example, former FBI director James Comey expressed regret for publicizing the bureau’s resumed investigation into her e-mails mere days before the election. “I had assumed from media polling that Hillary Clinton was going to win,” he wrote.
After Trump won, the polling industry joined journalists—many of whom were lulled into similar complacency by misleading polling numbers—in a period of soul-searching. How had their supposedly objective methods underestimated Trump’s support so starkly? Their British colleagues’ failure to foresee the Brexit vote months earlier enhanced the mood of doubt and introspection. Then, in 2020, after concerted efforts by polling companies and their aggregators to correct previous mistakes, the polls ended up being more inaccurate than at any time since 1980. The polling industry plunged into a reputational crisis from which it has yet to recover fully.
In Strength in Numbers: How Polls Work and Why We Need Them, the journalist and data scientist G. Elliott Morris sets out to defend the polling industry against its detractors and restore some self-confidence to his peers. “The rush to declare polling dead is misguided,” he writes. Morris understands the challenges polls face today: plummeting response rates, rising costs, erratic voting behavior, and public suspicion of pollsters (particularly among Republicans). But he argues that the real problem is not so much the polls as the public’s and the press’s misunderstandings of how they work. For Morris, the answer is not fewer polls but more of them, with audiences better educated to interpret and—most importantly—appreciate them. After all, he asks, “would we want to go back to sending out newspaper reporters to trawl the streets for enough willing participants to release straw polls before voting day?”
Morris’s bullishness is typical of the polling industry, a reflex that shields it from facing knottier questions about polling’s political and social usefulness. To many, the point of it seems self-evident: political polls measure public opinion, and every democracy should want its leaders to know more about what the public thinks than the broad results that elections can provide. “Good polls can reveal the will of the people,” Morris writes. “Condemning them as worthless is dangerous to this cause.” But that obscures their greatest achievement and larger influence, which lies not in any particular prediction or service to democracy but in the industry’s complete co-option of our understanding of public opinion, a concept that predates polling but that we can no longer imagine without it. The nature of this conquest now seems so natural, so self-evident, that it passes without remark—even in a book on the achievements of polling.
Public opinion has always been an elusive concept. “How does this vague, fluctuating complex thing we call public opinion—omnipotent yet indeterminate—a sovereign to whose voice everyone listens, yet whose words, because he speaks with as many tongues as the waves of a boisterous sea, it is so hard to catch—how does public opinion express itself in America?” the British jurist, historian, and Liberal politician James Bryce asked in The American Commonwealth (1888). A half-century later Gallup invoked Bryce and announced that he had found the answer: polling with sample surveys. It was as if polls would do for public opinion in the twentieth century what Eadweard Muybridge’s photographs had done for animal motion in the nineteenth century: reveal to a wide audience what was previously imperceptible to the naked eye. But like those of photography, polling’s claims of accuracy—one early pollster called the sample survey a “psychological X-Ray”—veiled an intrinsic deception: it was all too easy to forget how reality is framed and flattened by the medium’s design. (Pollsters have cultivated comparisons with photography, describing their polls as “snapshots”—ironically as a way to prove both their accuracy and their partialness.)
Following their success in 1936, Gallup and his fellow pollsters—the term was coined in 1939—promised that polling would revolutionize not only our understanding of public opinion but democracy itself. No longer would voters need to rely on elections to make their voices heard. No longer would politicians need to gauge public opinion by the size and volume of the crowds that cheered or jeered them or by the ventriloquism of journalists. Thanks to their “new instrument,” Gallup wrote, “the will of the majority of citizens can be ascertained at all times,” realizing a “truer democracy” and ensuring—“with little probability of error”—that dictatorships will “become mere bogey stories to frighten our great-grandchildren.” Such optimism was shared by Roper, who claimed that the public opinion survey represented “the greatest contribution to democracy since the introduction of the secret ballot.”
Gallup and Roper did not invent the sample survey. They imported it from the increasingly professionalized field of market research, where both their careers began. It is hard to determine whether advancing democracy was an honest goal or simply part of their marketing spiel. But it’s clear they thought that political polling would make them rich. “If it works for toothpaste, why not for politics?” Gallup reasoned. “I saw [it] as a veritable gold mine if we could learn fast enough how to use it in all of its ramifications,” Roper said.
These early pollsters preferred to ground the industry’s origin story in the scientific method rather than the profit motive. To this end, journals, institutions, and complex terminology proliferated in the field’s first decades, giving polling the aura of scientific inquiry. Gallup played the role of scientist, comparing his craft to that of a meteorologist. He made sure his name was always prefixed by “Dr.”—he had received his Ph.D. in applied psychology in 1928—and he made a great performance of not voting in elections, which supposedly proved that he was separate from “the new science of public opinion” he studied. “We have not the slightest interest in who wins an election,” Gallup said. “All we want to do is be right.” Roper agreed, describing the field as an “infant science.”
Some of polling’s problems in measuring public opinion are indeed typical of the natural sciences: supposedly “objective” methods were, and still are, suffused with the prejudices of their day, creating blind spots and distortions that only become clear in hindsight. In the early decades of polling, for instance, college-educated white men were widely assumed to be more interested in politics than anyone else, and so survey research drastically underrepresented black people, women, and low-income households in pursuit of accuracy. (Surveyors also preferred spending time in more affluent areas and households, while poorer neighborhoods were sometimes avoided out of fear.) Such problems persist: one explanation for polling’s failure to predict Trump’s win in 2016 is that college graduates, who were more likely to favor Clinton, were overrepresented among respondents.
Other problems with polling are typical of the social sciences: every attempt to study how people think and act has the potential to influence how they think and act, thus changing what is being recorded, either in self-fulfilling or self-negating ways. The results of any poll on a particular issue are liable to change how people think about that issue, just as any poll showing a candidate’s popularity is liable to influence that candidate’s popularity. The effects are unpredictable: some social scientists record a bandwagon effect, when people rally behind a candidate who is ahead in the polls, while other studies point to an underdog effect, when the opposite happens. Add to this respondents’ hypersensitivity toward the wording and ordering of questions—Roper once quipped that “you can ask a question in such a way as to get any answer you want”—and any analogy between opinion polls and “a weather forecast,” which Morris makes at least twice, collapses. (Like “snapshot,” the weather forecast analogy suggests both accuracy and unreliability.)
But the most fundamental problem with polling is that the phenomenon it claims to record—public opinion—has no coherent meaning or existence. The polling industry resolves this conundrum by simply making “public opinion” synonymous with its methods: polls record public opinion; public opinion is what polls record. Skeptics could see this sleight of hand from the start. “Dr. Gallup does not make the public more articulate,” Lindsay Rogers, a political scientist at Columbia University, wrote in an early polemic against polling in 1949. “He only estimates how in replying to certain questions, it would say ‘yes’ or ‘no’ or ‘don’t know.’ Instead of feeling the pulse of democracy, Dr. Gallup listens to its baby talk.”
Polling, in this analysis, was not so much an infant science as an infantilizing one: political matters were reduced to facile either/or stances, with little concern for how lightly or intensely one held an opinion or whether the opinion even existed before the survey. One of the oldest and most ambiguous concepts in the social sciences—a survey of the literature in 1965 quoted almost fifty conflicting attempts at a definition of “public opinion”—was reduced to a simple percentage: “60% think this, 40% think that.”
The conceits of such a percentage—its mirage of an equally informed, equally engaged citizenry, its impression of a country that has spoken—have been criticized by figures as varied as Martin Luther King Jr. and Pierre Bourdieu, for whom public opinion was too amorphous and impressionable to be fixed in the form of a number. Those conceits have also been exposed by many researchers. In an experiment conducted in 1980, people were asked whether they thought “the 1975 Public Affairs Act” should be repealed: a third gave an opinion, even though the act does not exist. In 1995 The Washington Post replicated the study with similar results, but found that another tenth could be goaded into an opinion with a follow-up question. (“Which [stance] comes closest to the way you feel?”) When people were told that either President Clinton or the Republicans wanted to repeal the act, more than half of respondents had a view. More recently, a UK poll found that nearly half of respondents claimed an opinion on a nonexistent politician, who actually proved relatively popular. (Anyone who has knowingly nodded along to a name they’ve never heard, hoping to avoid embarrassment, can relate to this.)
No poll can ever be sure what portion of answers are similarly offered off the cuff or to what extent respondents hold their positions outside the survey setting. The sociologist Leo Bogart said in 1972, “The first question a pollster should ask is: ‘Have you thought about this at all? Do you have an opinion?’” But usually polling companies don’t want to know: adding questions costs time and money, and ideally they want everyone to have an opinion on everything.
Morris has strong opinions about polling and a wealth of experience beyond his years. Born in 1996, he rose to prominence while still an undergraduate at the University of Texas at Austin by accurately predicting that the Democrats would regain the House in the 2018 midterms. After graduating he joined The Economist as a data analyst and journalist. He published Strength in Numbers in July 2022. In May 2023 he was announced as Nate Silver’s successor at FiveThirtyEight. (Silver left amid a round of job cuts at FiveThirtyEight, now under the stewardship of Disney, with about two thirds of its staff reportedly laid off.)
Morris’s book is filled with fighting talk: whatever the doubters say, polls remain “one of the most democratizing forces in American political history”; they can “reveal the will of the people”; they serve “as a pipeline from the governed to the government and as a bulwark against despots”; they are “the key to social knowledge”; they “hand a megaphone to the voice of the people, causing it to reverberate through the halls of government.” In one of his most strident moments, Morris even suggests that critics of polling are enemies of democracy: “In many cases, the denigration of polls is made by elites, elected officials, and ideological activists who have a stake in the public’s voice not being heard”—a claim that would be easier to take seriously if he engaged with the critical scholarship on the polling industry. Rogers’s 1949 polemic and Gallup’s combative response receive a few paragraphs. Susan Herbst and Sarah Igo are referenced in the acknowledgements, but any influence of their important work on how polling hollows out understanding of political participation and on the foundational ties between the polling industry and market research is hard to find in the main text of Morris’s book.
The reasons Morris gives for his fervent faith in polling are underwhelming and overwrought. In the introduction he celebrates how Republicans and Democrats now use polling to determine which presidential primary candidates participate in debates, hailing this as proof that “you don’t have to look far to find concrete examples of polls serving meaningful functions in our electoral, judicial, and governing systems.” He omits the fact that even major polling companies have criticized this use of their findings. (“I just don’t think polling is really up to the task of deciding the field for the headliner debate,” Scott Keeter, then Pew’s director of survey research, said in 2016.) Later we learn how in 1960 John F. Kennedy’s pioneering pollsters “advised…a strategy for his upcoming debate [with Richard Nixon], telling him to come off strong, competent and understandable to the average American”—leaving us to wonder how any candidate could ever have fared without such scintillating guidance.
But Morris also knows that polls are not the “crystal balls” that their most avid cheerleaders sometimes claim them to be, and he vacillates between championing their indispensable place in democracy and admitting their fallibility. Morris the populist revels in nebulous expressions like “the will of the people,” “the power of the people,” and “the voice of the people.” Morris the scientist takes every opportunity to plead for caution and emphasize plurality. Morris the populist prevails: by the book’s conclusion, he is still insisting that “the will of the people is now quantified and easily accessible by any reformer, legislator or interested citizen”—despite beginning the same paragraph with a nod to “what we have learned about the uncertainty in polling and the varying quality of public opinion across issues.”
Caught between the seriousness of its science and the need to market its product, the entire polling industry is trapped in a version of this double act. Gallup was no different. As Igo noted in The Averaged American (2007), he wrote that “the American people are as various as their land” and in the same article repeatedly invoked the mythical “average man” discovered by polls.
Morris concedes that, overall, polling has yet to live up to its lofty promises. But his reasons for why polls don’t work are even less convincing than his reasons for why they do. His main targets for blame are not pollsters or their methods but the public and, above all, the press. As I read his defense of polling, the words of Oscar Wilde came repeatedly to mind: “The play was a great success. The audience was a failure.”
According to Morris, the public has failed to appreciate that every poll comes with a margin of error, so really no poll can be wrong: “Consumers of polling and election models should not trick themselves into mistaking polls and projections for a science they’re not—and will likely never be.” While more polls—particularly in the very close 2024 presidential election—have started to include the margin of error in their results, Morris’s mixed messages will hardly help confused consumers: he advises resisting total faith in polls but also says that “they are scientific” and that “informed readers” should turn to “RealClearPolitics and Pollster to know who’s ahead, and to FiveThirtyEight to know whether they’ll win.”
But Morris saves his harshest words for the media, decrying “the damage done to the polling industry by an overconfident and naïve press.” The polling industry and the media have always had a difficult, if also mutually dependent, relationship. While many journalists initially resisted polls as an encroachment on their craft and authority—“Today, unless you can say ‘According to the Poop-A-Doop survey, Umpty-ump percent of the people chew gum while they read Hot Shot News!’ you fail to make an impression,” one journalist lamented in 1950—it’s also true that from the start, the pollsters’ most important client was the press, and the two quickly established symbiotic ties. The press commissioned polls to generate news stories and bolstered its reporting with persuasive statistics, while polls relied on the press for funding and, crucially, publicity. By the end of the century, most major news organizations had their own in-house polling operations or formal partnerships with polling companies.
This partnership inevitably affected the nature and purpose of polls: newspapers didn’t want to pay for boring findings; they wanted engaging, dramatic stories, tales of conflict and controversy. The polling industry obliged, with varying degrees of reluctance and enthusiasm, and received not just money and publicity but an alibi: the media could now be blamed for its worst traits—exaggerating social conflict, simplifying issues, overstating accuracy.
In the same vein, Morris insists that whereas the media “want attention-grabbing, confident predictions,” pollsters understand “all the nuance and uncertainty that are inherent in their data.” Elsewhere, Morris concedes that “pollsters systematically overestimate their own accuracy,” but the nature and gravity of this contradiction—that pollsters understand and systematically ignore inconvenient truths—elude him.
The fact is that polling companies need engaging, dramatic results, not only because such results keep their patrons in the press happy but also because interesting poll results travel further and faster, spreading the name of the company and thus attracting more clients. While Morris laments how the pollsters and the press both do a “poor job” conveying polling’s limitations, with no account—no mention—of the business side of polling and no sense of how polls need publicity, he misses how pollsters can become invested in their own simplifications and misinterpretations. In pursuit of both accuracy and profit, compromises are made.
It’s hard to believe, given the number of polls being conducted in 2024, but Gallup and Roper were always skeptical of election forecasts. “All of us in the field of public opinion research regard election forecasting as one of our least important contributions,” Gallup said; Roper thought they were “socially useless” and might “do very much more harm than good.” But election forecasts are the only verifiable “theory” that this “science” puts forward: their accuracy is fact-checked by the final ballot in a way that other opinion polls never can be. For polling companies, election campaigns are thus marketing campaigns. The results are twofold: an inordinate number of polling companies participating in the game of predicting elections, on the one hand (in 2020, there were at least 1,572 state-level preelection polls, including 438 in the final two weeks alone, by over 200 different polling companies—all eager, in Morris’s telling, to strengthen American democracy); and on the other, a huge investment in election-forecasting over opinion-measuring methods.
Polls may have once promised to make politics about more than elections, but in practice they have surely done the opposite, with each vote presaged by months, sometimes years, of obsessively dissected forecasts and horse-race coverage. No one embodies this trend more than the politically indifferent, election-obsessed Nate Silver. “With the politics stuff, I just like the elections part,” he told The New Yorker as he was leaving FiveThirtyEight.
Perhaps the polling industry’s standing in society today is most analogous to that of the advertising industry that spawned it: polling organizations are similarly ubiquitous, profitable, and treated cynically by members of the public, who suspect an ulterior motive. Like advertising, political polls are increasingly associated with attempts to manipulate public opinion, tailor messaging in superficial ways, and inform public relations strategies. Politicians of all stripes denigrate polls in public and obsess over them in private. “I don’t have a pollster,” Trump declared on the campaign trail in 2015, before soon hiring one. “No one tells me what to say.” In the months before Joe Biden’s decision to withdraw his candidacy, his advisers frequently attacked the polls in the press for underestimating his support. Despite polls that increasingly indicated he was unlikely to defeat Trump, Biden refused to leave the race until the political fallout from his disastrous debate performance forced his hand.
In the 1990s a new technology replaced polling as the tool destined to transform democracy: the Internet. In a fuller realization of polling’s potential, people would be able to speak up and share their opinions at all times, leading to a better-informed public, more responsive governments, and a truer version of democracy for all. Just as Gallup promised to bring the “town meeting” ideal into the twentieth century—“This time, the whole nation is within the doors,” he wrote—the Internet promised to bring it into the twenty-first. “The function of the Net, in this conception, is to facilitate a running national poll of public opinion, with immediate electronic feedback from citizens to government and vice-versa,” the political scientist Bruce Bimber explained in 1998.
Soon a specific kind of website became the medium for these hopes and dreams: the social media platform, Facebook and Twitter in particular, which launched in 2004 and 2006 respectively. Twitter pitched itself, in a way reminiscent of Gallup’s early polls, as “The Town Hall Meeting… In Your Pocket” and a “real-time measure of public opinion.” It also seems relevant that as a nineteen-year-old sophomore at Harvard, Mark Zuckerberg first made a name for himself by designing an online poll: FaceMash had users choose the more attractive of two female students from their photos from the Internet and built a university-wide ranking. “I almost want to put some of these faces next to pictures of farm animals and have people vote on which is more attractive,” the young Zuck wrote on his blog before launching the short-lived site. Facebook followed a year later, and soon he was celebrated as a champion of democracy.
Like the polling industry, social media platforms’ ties to the advertising industry were either downplayed or ignored: the aim was to give people a voice to enrich democracy (and then use what they said and did to sell them stuff in increasingly sophisticated ways). Social media platforms arrived at the same convenient conclusion as the polling industry: healthy markets and healthy democracies needed the same thing—to know what the public thinks. But surveys were no longer necessary: through social media, users’ thoughts and actions could be tracked at all times. By 2008, advertising gurus excitedly announced, the Internet had already overtaken all other market research methods—“postal, face-to-face and telephone”—to become “the leading global modality for quantitative data collection.” “No longer is recruitment an issue; no longer is the phrasing of the question an issue; no longer is the duration of the interview an issue; and no longer is respondent fatigue an issue,” Finn Raben, director general of ESOMAR, one of Europe’s largest conglomerates of market researchers, enthused in 2010. “If the topic is of interest, then the material is already there…thus is born the ‘Age of Listening’ as opposed to the ‘Age of Questioning.’”
What the advertising industry celebrated as “listening,” however, others saw as something more sinister. The digital economy, premised on the invasion of privacy, was soon denounced as “surveillance capitalism.” Information became its lifeblood, and digital companies developed insatiable appetites for more and more information on users, however trivial. This created a double dynamic: a desire not just to record information but to generate more information.
This is one of social media’s most significant resonances with the polling industry. Just as polls want respondents to have an opinion on everything, cuing views through specific questions and portraying an opinionated public while claiming a neutral detachment, social media platforms like Facebook and Twitter, now X, repeat the same trick on an even greater scale. “You don’t have to have an opinion on everything” has become a refrain online, reflecting how much pressure is applied in the opposite direction. Facebook asks its users, “What’s on your mind?” X prompts its users, “What is happening!?” (The panicked exclamation mark is a new addition, neatly symbolizing both the platform’s neediness toward its users’ information and our disorienting present moment.) Shortly after purchasing the platform in October 2022 for $44 billion, Elon Musk implored users, “If I may beg your indulgence, please add your voice to the public dialogue!”
Social media companies assure the public that their ravenous hunger for opinions simply stems from their deeply felt desire to give people “a voice.” (Speaking at Georgetown University in 2019, Zuckerberg used the word “voice” over thirty times in his thirty-five-minute address.) But what they really want is a reaction: a “like,” a “share,” an emoji, a short comment, or some other form of quantifiable communication that, following Lindsay Rogers, we might call twenty-first-century “baby-talk”—information that can then be packaged, analyzed, and sold. In this monetized vision of the “town square,” more talk means more profit, and users are ideally both perpetual pollsters—always courting reactions to their thoughts and experiences—and obsessive respondents, offering simplified views on a huge number of issues, from politics to brands of toothpaste.
The affinity between social media and polling is perfectly captured by the polling function on many social media sites, which brings the straw poll into the age of big data. Twitter launched one in 2015, and Facebook followed two years later. As Twitter’s new CEO, Musk was initially fond of using the feature. In November 2022 he announced that his decision to reinstate Trump—who was banned from the platform after the storming of the Capitol on January 6—would be determined by a Twitter poll. More than 15 million users voted, and 51.8 percent voted “yes.” “The people have spoken,” Musk tweeted. “Trump will be reinstated. Vox Populi, Vox Dei.” (Trump’s account was restored, but it wasn’t until August of this year that he added his voice to the platform’s public dialogue once again.) In December 2022, facing mounting criticism over his leadership of Twitter, Musk held another Twitter poll on whether he should continue as CEO. The online survey lasted twelve hours and 17.5 million users responded, with 57.5 percent wanting him out. Musk remains in his post—and his penchant for hosting polls seems to have passed. But he continues to defend their integrity. In March, spreading the conspiracy theory that the polling industry uses fake interviews, he posted: “The vast majority of polls are bs. Polls on this platform at least reach some real users.”
In 1921, as the editor of the student newspaper at the University of Iowa, the nineteen-year-old George Gallup wanted to attract new readers. He published a notorious article titled “The Unattractive Women,” which took the form of an ostensibly overheard conversation between two male students and declared that it was women’s “duty to…make themselves as attractive as they can”—a duty that, like Zuckerberg some eight decades later, Gallup seemed to think many women on his campus were failing at. The article led to a spike in circulation and on-campus misogyny. “All of the girls were angry,” Gallup later recalled, but “from that day on, the newspaper was eagerly read.”
Gallup’s interest in getting attention and his desire to discover “what the public wants” were two sides of the same coin. His Ph.D. dissertation sought to pioneer an “objective” way of measuring what parts of a newspaper readers spent time on. Gallup found that they really enjoyed looking at comic strips and pictures, not the hard news they liked to claim in surveys, and he called on “the modern newspaper” to offer more of both “to get itself read” and become more appealing “from an advertising point of view.” In Gallup’s crowd-pleasing quest, polls were doubly useful: they were both a means to discover what people wanted (respondents’ dishonesty notwithstanding) and a product that people wanted—a form of journalism that, like cartoons and pictures, could make politics light and accessible.
Today that product remains overwhelmingly popular: polls saturate election coverage, turn politics into a spectator sport, and provide an illusion of control over complex, unpredictable, and fundamentally fickle social forces. That isn’t to say that polls don’t have uses beyond entertainment: they can be a great asset to campaigns, helping candidates refine their messages and target their resources; they can provide breakdowns of election results that are far more illuminating than the overall vote count; and they can give us a sense—a vague and sometimes misleading sense—of what 300 million people or more think about an issue. But, pace Morris, the time for celebrating polls as a bastion of democracy or as a means of bringing elites closer to voters is surely over. The polling industry continues to boom. Democracy isn’t faring quite so well.
Silicon Valley ultimately peddled the same feel-good story about democracy as the polling industry: that the powerful are unresponsive to the wider public because they cannot hear their voices, and if only they could hear them, then of course they would listen and act. The virtue of this diagnosis is that structural inequalities in wealth and power are left intact—all that matters in democracy is that everyone has a voice, regardless of background. In a very narrow, technical sense, their innovations have made this a reality. But the result is a loud, opinionated, and impotent public sphere, coarsened by social and economic divisions and made all the more disillusioned by the discovery that, in politics, it takes more than a voice to be heard.
—September 18, 2024
