Polls are an easy target these days. I know from every time I answer the question ‘what do you do?’ and I’m not even someone who has the word ‘poll’ in my job title. Democrats and Republicans, olds and young’uns — everyone likes to hate on polls. There are now polls that show just how much people distrust polls which, no doubt, many people also distrust.
In fact, you should approach many contemporary polls with a healthy dose of skepticism with all that’s going on — declining response rates, scammy online panels with slightly-less-than-human respondents, the difficulty of fixing differential nonresponse and selection bias, and, most egregiously, the inability (or unwillingness) of many pollsters to attach a basic explanation of their methodology to their polls. As we approach one of the most high-stakes elections of our lifetimes after the previous two most high-stakes elections of our lifetimes where polling was deemed ‘a failure’, it’s tempting to want to burn it all down.
But this would be a huge mistake. And that’s because public opinion polls1 have laid (and continue to lay) the very foundation for the modern American state.
Let me explain.
There’s a view in historical political economy — argued by prominent scholars like Theda Skocpol and Stephen Skowronek — that the American state did not really exist until the New Deal. Before then, the federal government was a loose federation of “courts and parties,” with limited centralized capacity. The argument goes that Franklin Roosevelt’s New Deal, the first of its kind to provide a comprehensive federal program of social services, effectively brought together the American state, putting America on equal footing to other western nations with robust welfare states.
From the ruts of rural and urban poverty alike, an American middle class emerged with the help of programs such as the Works Progress Administration (WPA), the Tennessee Valley Authority (TVA), the Civilian Conservation Corps (CCC), and, of course, Social Security.
The New Deal transformation of America may not have been possible without public opinion polling.
As social scientist Jean Converse points out in an illustrious history of American polling, Survey Research in the United States, the WPA funded more than 5,000 survey research projects in an effort to enumerate, and more importantly represent, the conditions of ordinary Americans who were eligible for these relief programs. Represent, in the democratic sense, is the correct term since the cross-tabulations from these surveys were presented to Congress in order to justify the expansion of these programs.
The monographs from these research programs (take a look for yourself) captured, for instance, estimates of WPA benefits received in every county in the United States:
… how this relief fluctuated from month to month, across different geographies:
…and, important for persuading Congress, that the assistance had desirable ‘pull-yourself-by-the-bootstraps’ effects: higher crop yields, increased earnings, and employment outcomes (albeit, in ways that would drive most modern economists mad):
Public opinion surveys allowed the U.S. government to do data science nearly a hundred years before the term existed: measuring the New Deal's massive demand and pinpointing exactly who wanted what from its groundbreaking programs.
Beyond the New Deal, public opinion polls have become a vital tool for the federal government in nearly every imaginable area of governance.
During America’s two great 20th century wars, polling was used to inform how to ration everything from fuel to food. Public opinion surveys helped ensure that resources were being distributed in a way that the public could tolerate—or at least in a way that wouldn’t lead to riots.
The unemployment statistics you see on the nightly news come from polls, and the methods for creating them were pioneered in the era of those New Deal research programs.
Public health is another area where polling plays a huge role. If the COVID-19 pandemic taught us anything, it’s that we don’t just need accurate health statistics but that we need them fast. My colleagues and I at NORC have been fortunate to be at the forefront of this effort, helping the CDC build the NCHS Rapid Survey System, a polling tool to track things like the prevalence and experience of long COVID symptoms in the population. Other colleagues have done amazing work in mapping bilateral hearing loss (the third most common chronic health condition in the United States!) across the American landscape.
This is all good news for liberals, who generally believe in the informed expansion of government to provide these sorts of social services to the people. If you don’t share that view, I have good news for you too: polls also helped shape American capitalism (and made it, too, more responsive to the people). The same techniques used for those New Deal programs were quickly adapted by businesses in the ‘30s and ‘40s, to understand consumer preferences — what we now call market research. This research helped fuel the rise of consumer capitalism, informing everything from the design of household products to the launch of nearly every iconic brand you’ve ever interacted with.
When most people—your weird Thanksgiving uncle and the Brooklyn dudes I keep having to explain myself to—hear the word “poll,” they likely think of horse race polls: those snapshots of who's up or down in the upcoming election. It’s worth remembering that public opinion polling has a deep history that predates election polling by nearly a century. Of course, polling had its problems back then too—the sampling methods, while impressive, were not always representative in the strict statistical sense and the questions themselves, while important, were not always properly standardized.
Nevertheless, polling gave powerful people the best possible answers to the most important questions — and we (your weird uncle, my Brooklyn haters) are all better for it.
At least colloquially, there is a distinction between polls and surveys and it has to do with the scope and length of questioning.
Polls tend to be shorter, asking about salient topics in the moment. There is no such expectation of temporal relevance or expediency with surveys. Take this for what it’s worth, but here are the five words that ChatGPT most associates with the word poll — quick, focused, snapshot, opinion-based, short — and the same for survey — comprehensive, detailed, analytical, long-form, multidimensional (the words themselves are weirdly longer for survey).
To the layperson, my sense is that this format distinction is academic. To everyone else, i.e. the people bothering to even read this long footnote, consider this post to be about public opinion research writ large, encompassing both formats.