Polling Has a 98 Percent Chance of Being Bad for You
It’s not the accuracy of the polls that matters, it’s their ubiquity.
In 2016 and 2020, Americans were hard pressed to escape election polls. From the whiplash of Trump’s obsessive tweets—“the polls are fake,” (Oct. 13) “poll numbers looking very strong,” (Oct. 16)—to news coverage saturated with the latest findings, the ever-shifting numbers were omnipresent. On Election Day, everyone watched in real time to see how predictions were lining up with exit polls.
But on the eve of the 2017 French presidential election, French citizens had to look internationally for their fix of data. A 32-hour legally enforced blanket of silence around the election fell over the country at midnight, just as it has since 1977 when the law was first passed. No campaigning, no statements from campaigns, and, perhaps strangest of all, no polling or publication of polls, on pain of a 75,000-euro fine. Political junkies had to settle for Swiss and Belgian reports, marked on social media with #RadioLondres, a reference to coded messages sent during World War II from a London-based radio station of the French Resistance.
It’s hard to imagine such restrictions in France, where censorship is reserved for offensive speech and religious symbols. (A stricter weeklong blackout was scaled back in 2002 over free speech concerns.) And yet the law endures, alongside similar laws in a host of other countries, among them Italy, Spain, Portugal, Taiwan, and Peru. In Honduras the embargo on polls lasts a whopping 45 days. Common to all the laws is the original motivation for the French law: concern about the effects of exposing the electorate to polling data.
This concern is not just about bad polling data, which is usually the focus of critics. Bad data comes in two varieties. There are deliberately fraudulent polls like those that flooded Mexico before the 2018 election. And there are incompetent or inaccurate polls, a major topic in the wake of Brexit and America’s last two elections. “If public-opinion data are unreliable, we’re all flying blind,” wrote David Graham in a post-mortem of the 2020 “polling catastrophe” in the Atlantic. In response, pollsters and modelers have defended their trade by suggesting polls performed as advertised but statistical simpletons persist in misreading them.
Yet prohibitions on publicizing polling results near an election are not motivated primarily by preventing access to bad data, or bad interpretations of good data. However imperfect, the laws acknowledge an obvious truth that 21st-century information gluttons have forgotten: Sometimes not knowing is a blessing. While legislation may not be the ideal way to secure salutary ignorance, neglecting its existence is a mistake. As French lawmakers argued in 1977, sondomanie or “poll mania”—already an issue back then—can make it difficult to clear one’s mind and evaluate the candidates on their merits. They lamented the resulting “sheeplike movement” and “collective madness.” Better to secure an environment well suited to responsible decision-making, which they described using the terms “serenity, tranquility, calm, and silence.”
Imagine what those lawmakers would think if they saw the explosion of modern polling: thousands of pollsters producing reams of data months in advance of elections, like Christmas decorations on sale in May, and the media and the public consulting the results like an oracle, compulsively gorging themselves on data.
Political science has multiple terms for the potential influence of polling data, among them the bandwagon effect (conformity with perceived majority) and spiral of silence (reluctance to express perceived minority opinions), all of which can influence voter choices and voter turnout. The laws against publishing polling results are meant to insulate against these psychological intrusions, even if they reflect accurate information.
Our myopic focus on bad data and fallacious interpretation results from an assumption shared by pollsters and their critics alike, an assumption so deeply engrained that questioning it is taboo: Accurate information is always good. Yes, misinformation should be fought. So should information illiteracy. But surely an educated public should welcome good data, the more the better? Suggesting otherwise—perhaps flying blind is better under certain circumstances?—reeks of censorship and paternalism. At FiveThirtyEight, the popular polling aggregator, Maggie Koerth makes the case that access to information is a crucial right, regardless of its potential influence. “Polls may well have changed the outcome of an election in France,” she concedes of the 2002 French presidential primaries, when an extreme far-right candidate shocked the nation with a surprise win. “But that was a choice the voters had the right to make.”
But the right to do something does not entail the wisdom of doing it. Ending the discussion with our right to polling data relegates a suite of important questions to the scholarly shadows: How many polls are enough? With what frequency should the public consume them? Even if they are accurate, what exactly is their value to individuals and society? Is it pure entertainment, and if so, does that cheapen what ought to be a serious topic?
Political scientists have long wrung their hands about how polls encourage shallow horse-race coverage, which in turn shapes attitudes toward politics. A growing body of research suggests it contributes to distrust in politicians and media, diminished coverage of third-party candidates, and, according to a 2018 meta-analysis, “inhibits the development of an informed citizenship because the public is mostly familiar with the political rivalries instead of actually knowing what the substantive debate is about.”
We can only consume a limited amount of information, just like we can only consume a limited amount of food. Is it possible that a diet of election coverage dominated by polls might be unhealthy? Is checking FiveThirtyEight on a daily basis like eating a pint of ice cream once a day? (According to Nate Silver, the site’s founder, daily check-ins are at the low end of what’s necessary to “know all that you need to know.”)
In other contexts, we routinely restrict information access to facilitate ideal forms of decision-making. Blind wine tastings are not a safeguard against fraudulent labels or fake prices. They allow participants to reflect on and evaluate the flavor of the wine undistracted by prejudgments. When participants vote on the best tasting wine, the result is “better” because it was produced in accordance with certain criteria of what constitutes objective, unbiased judgment.
Unbiased judgment is an impossible and contested ideal, but it would be foolish to abandon it. Virtually all democracies place limitations on electioneering near polling places, ranging from prohibitions on campaign paraphernalia to bans on political persuasion. These restrictions on free speech serve an obvious and laudable purpose, shared by the French lawmakers in 1977: to create an environment conducive to good decision-making.
Election silence legislation is not the best way to achieve this goal in the United States. General prohibitions can easily backfire as they did in France, ceding the dynamics of political discussion to social media algorithms. Banning legitimate polls in the media will almost certainly create a vacuum filled by illegitimate polls distributed online.
The absence of legislative solutions, however, does not mean the absence of all solutions. Here the food metaphor is helpful. Overconsumption of calories has become an urgent global health crisis. Our global food ecosystem is breaking down, with companies stoking pathological demand and then happily meeting it. Unfortunately, soda taxes and bans on school vending machines will not solve the problem, and more draconian government measures would be dystopian—but the problem is still there. Individuals and communities will have to make a concerted effort to change that ecosystem, instituting new norms around food distribution and consumption, even actively changing their palates.
Likewise, the information ecosystem is in the middle of a calamitous breakdown. Just as technology facilitated an unprecedented increase in ultra-processed foods—cheap, calorie-dense, and highly palatable—it is now facilitating the distribution and consumption of ultra-processed information: ubiquitous, highly palatable pictures of the world chosen not according to their value or their effects, but for the compulsion with which the public consumes them.
It’s easy to blame outside influence or domestic extremists for this, whether it’s Russian interference or QAnon. But even honestly produced, well-intentioned information can be toxic in excess. Election polls are a textbook example. Useful to policymakers that need to understand their constituents, the data has no obvious use for voters, especially months out from the election. Yet it’s consumed voraciously, each new round sending ripples through the media. Our brains flood with the pleasure of prophetic power. “I’m a pollaholic,” confesses David Brooks in the New York Times, and then turns to the food metaphor himself. “It’s hard to figure out how each candidate will handle the so-called budgetary fiscal cliff, or the uncertainties involved with Iran. But the polling numbers are like candy. So clear and digestible!”
Candy bars taste great. We have the right to consume them and companies have the right to produce them in whatever amount the market can bear. None of this changes the fact that in massive quantities they harm our bodies. The recent history of prohibitions on polls reflects a shared intuition that in massive quantities they, too, can harm our body politic. Lack of attention to actual issues, the disappearance of complicated individuals into demographic homogeneity, increased skepticism about every percentage presented by the media, whether election outcomes or vaccine efficacy: Look closely and the signs of chronic illness are there. And if we don’t change our collective information palate, sondomanie and other forms of information overload may become a full-blown epidemic.