Liberals Are Far Too Obsessed With Polls

Ryan Cooper

Opinion polls give us important insights to build political strategy from. But the way Democrats have become obsessed with them is completely distorting the party’s strategic thinking. It’s a recipe for incoherence and defeat.

Bill Clinton at a rally at North Carolina State University in Raleigh on October 4, 1992. (Kenneth C. Zirkel / Wikimedia Commons)

Interview by
Luke Savage

From media coverage to electoral strategy, the overbearing influence of opinion polls on American politics is difficult to overstate. For one niche of Democratic strategists in particular, polling offers a simple and elegant route to success: a way of thinking about campaigns and policymaking that has been aptly dubbed “survey liberalism.”

Ryan Cooper is a national correspondent at The Week. In this interview, he speaks to Jacobin’s Luke Savage about the issues raised by his recent essay on the many limitations of survey liberalism — and why left-wing politics should use, but not be bound by, opinion polls.


Your piece is engaging with a phenomenon that’s widely associated with politics, and I think generally disliked — namely the way professional strategists and politicians chase opinion polls. But there’s actually a very specific theory of electoral politics at work here — and one that enjoys a real constituency within the Democratic Party. It’s something Aaron Freedman has dubbed “survey liberalism.” What exactly is survey liberalism?


Basically, figures like data scientist David Shor say you should spend a lot of money in cheap media markets and swing states close to the election, and that you should take positions that are popular while avoiding positions that are unpopular. So, it’s simple enough, in a sense. But I think it has some downsides in the way it’s presented and packaged — specifically within a certain cohort of people.


One reason I was keen to discuss the piece with you is that the idea that Democrats generally chase what’s popular actually seems quite counterintuitive to me. I think there’s a case to be made that America’s political class in general often operates against majority opinion, thanks in part at least to the way it’s captured by organized money. How exactly do you see the picture?


Yes, you could certainly point to a number of positions that Biden and moderate Democrats take that are popular. If you look at the polling on “defund the police,” for instance — which is a big axis of controversy in this crowd — it doesn’t poll very well. The police are generally pretty popular. But you can also point to a lot of stuff that moderate Democrats, especially moderate Democrats in Congress, support that are all really unpopular: bank deregulation, for example. Josh Gottheimer is currently taking the entire Biden agenda hostage to get a state and local tax (SALT) deduction (which was kept in the 2017 Trump tax cuts).

That was basically a finger in the eye of blue states, because it was a de facto subsidy from the federal government to wealthy homeowners — people who owed a ton of property tax could write it off their federal taxes. More than half the benefits of repealing the SALT cap would go to the top 1 percent, while the bottom 80 percent of people get nothing. This type of massive tax cut for the rich that this guy — he’s in New Jersey — is pushing is not particularly popular. And I think it would especially not be popular if the party were to frame it as such. But it’s popular among the people who fund Democratic campaigns in New Jersey, and in New York and San Francisco, because they have a lot of wealthy homeowners.

That really puts the lie — aside from the other problems with survey liberalism — to the idea that the party is dedicated to doing popular things. But I think it also shows you the power of being stubborn and committed to things. You can get a policy through, against the will of majority opinion, if you are determined (and whether or not that redounds to the benefit of the Democrats in 2022 remains to be seen). If you’re courageous in this sense — in this case, if you have the courage to do something terrible — you can accomplish a lot.


In many ways, the real thrust of your argument against survey liberalism is concerned with the moral and ethical limits of, well, merely taking a show of hands on every issue. I want to come back to this, but first: you also run through some of the basic technical and empirical issues with opinion polling (which, after all, is the whole basis for this way of thinking about electoral politics).

Can you discuss some of these? I think it would be fair to say the industry as a whole has rather successfully constructed a mythos of, if not infallibility, then at least steadiness and reliability. Our political or cultural outlook, so this thinking goes, is inseparable from subjectivity and personal bias, but the data doesn’t lie.


For one thing, you have issues and problems related to sampling. Obviously, you can’t go door to door and ask every single person in the country what they think, so you have to take a sample — say, a thousand adults — and extrapolate from that. If your statistics are good, that’s usually a pretty accurate way of doing a poll. But if you look at the history of polling over the last four to eight years, it’s really developed some serious problems. People don’t have landlines anymore (that used to be sort of the bar standard). A lot of people don’t really open their mail that much anymore, because there’s so much junk mail. There are so many spam calls, especially for mobile phones, that a lot of people just don’t answer their phone if the call isn’t coming from someone they know. All of that is leading to serious sampling problems.

If you look at opinion polls asking people who they were going to vote for in 2016 and 2020 . . . everyone knows the polls were off in a few states. They were badly off by nearly 10 points in the case of Wisconsin. I believe nationally they were closer, but it was still not quite there. That led to a lot of panic and soul-searching among the polling firms, so they altered their methods and, in 2020, they missed again: the same way that they did in 2016, namely in underestimating Donald Trump’s support.

So that’s one problem. The other problem has to do with the bias of the polls: i.e., are you sampling the genuine opinion of the electorate, or are you biasing the question? This is something that’s been known basically since George Gallup started his polling firm back in the 1930s.


There was an example related to polling around Social Security in your piece that I thought was particularly illustrative.


Gallup has this whole page of Social Security polls, and a lot of them are basically relics from before 2016, when there had been decades of austerian neoliberalism — starting in the 1980s when they cut Social Security benefits, subsequently trying to get rid of it. That’s been a political project for decades, and so Gallup eventually internalized this ideology and wrote these polls that are self-evidently biased against them. They’re sort of push polls.

In one of them, the prompt is “Next, I’m going to read a list of problems facing the country. How much do you personally worry about the Social Security system?” Another one is: “Which of these statements do you think best describes the Social Security system? It is in a state of crisis; it has major problems; it has minor problems; or it does not have any problems.” Another is: “How long do you think it will be until the costs of the Medicare and Social Security programs create a crisis for the federal government?”

These are all incredibly leading questions, and any pollster would tell you that questions like this are going to drastically overestimate the number of people who say that Social Security is a problem, because of the overall framing. The way the issue is presented is, “How do we fix the problems with Social Security, and which benefits should be cut?” rather than, “Do you like the program?”


There are also very real qualitative issues with structuring a policy agenda or electoral strategy around opinion polls — that is, issues that go beyond their empirical reliability. Something that registers as popular in a particular moment, for example, might become unpopular later or have damaging electoral consequences down the road. What are some major examples of this?


In terms of how polls change, a great example that’s happening right now is the polling on the withdrawal from Afghanistan. Look at polling a week or two ago: an overwhelming majority of people, even a solid majority of Republicans, said we should get out. Now we’ve got this week of coverage, particularly in the mainstream, “objective” media, just killing Joe Biden over pulling the troops out. A lot of Democrats are also refusing to stand up for Biden and are instead attacking him, putting out press releases saying he should have done it better or magically conjured some sort of competent imperial bureaucracy

It was kind of remarkable that Biden, for all of his victim blaming of the Afghan soldiers and stuff, didn’t back down and instead said, “I’m doing it. I don’t care what anybody thinks, there was no way around it.” And opinion has changed really rapidly. I forget the figures, but there was a Morning Consult poll that had support for withdrawal down something like 20 points among Republicans, 15 points among Democrats, and something like that among independents. So it’s basically a split decision now, with a slight majority saying that it was wrong to withdraw the troops.

If you go by survey liberalism, you’d say, “We got our poll. Let’s go ahead and do it. Oh, no, now there’s backlash, let’s retreat. Let’s start the war up again” — i.e., not thinking about, for lack of a better word, political leadership. Instead, this has to be, “I’ve got to analyze the situation on the ground and do what seems like the best combination of what’s reasonable and what seems politically popular.” I think that if Biden and the rest of the party stick up for themselves and say: “Look, sorry, maybe in the best-case scenario we can get some tens or maybe hundreds of thousands of Afghan refugees out of here, but at the end of the day, the Taliban is going to win, and there’s nothing we can do about it, we just have to rip the Band-Aid off.” I think that’s a position that would stand up after a while. The media coverage would fall back down. But Democrats have to stand up for themselves. And you can’t stand up for yourself if you’re just following the weathervane of political opinion, wherever it happens to blow.

One more example of that kind of logic failing can be seen in the consequences of Barack Obama refusing to bite the bullet when his administration was looking at the housing market collapse in 2009 and 2010. I believe Austan Goolsbee, Christina Romer, and other economists told him that they figured there was $750 billion to $1 trillion of negative equity in the housing market at that time (in other words, that was the loss that somebody was going to have to eat). And they didn’t want to do another bailout for homeowners, because it would have been a great big, expensive thing. They didn’t want to push the losses onto the banks, because the banks couldn’t take the hit. That was their argument, which was quite self-serving for the bankers.

So they just let homeowners eat it. They had a slush fund of money from the Troubled Asset Relief Program (TARP) bailout in 2008 that they could’ve used to write down a lot of that bad debt. FDR did this in the 1930s. But they didn’t, because it would’ve put a lot of losses on the banks. Maybe that would have been politically unpopular (Rick Santelli, after all, started up the Tea Party shrieking about the prospect of homeowners getting a bailout). But the ultimate result was basically to create little neutron bombs of economic disaster all over the country from people who got foreclosed on. About 10 million people lost their homes, and that screwed up the construction industry, home building, and made unemployment almost certainly way higher than it would have been on Election Day 2010. If unemployment had been, let’s say, 6 percent instead of 10 percent . . . every point is probably another ten to fifteen seats that the Democrats don’t lose in the House, and maybe another two hundred seats that they don’t lose in state legislatures.

They failed to recognize that a homeowner bailout might be unpopular in the short term and wouldn’t poll that well, but that doing it would have long-term consequences that were going to be positive for the party. And what happened was the exact thing they were trying to avoid. It’s like, “Oh, we can’t do this. We’ll lose our seats,” and they all lost their seats anyway.


The last few paragraphs of your recent piece make the case against survey liberalism with some urgency, and you discuss a few rather pressing issues related to democracy. Why, in this moment especially, is survey liberalism a particularly counterproductive way of thinking about democratic reform?


We’re currently facing an effort to overthrow whatever raggedy democratic institutions exist in this country (obviously the Electoral College and the Senate already make a mockery of democracy). But I think it’s a pretty compelling example of the folly of treating something that is so morally charged and politically central by looking for a show of hands. Shor, to his credit, does suggest that we need to add more states and get rid of the filibuster, and he argues that we possibly need to add more justices to the Supreme Court. None of that is popular, especially packing the court.

You don’t get majority support for any of those things, but if you look at the polling, it’s all very equivocal. Most people in America, for example, don’t even really understand what the filibuster is. And if you got rid of it, they probably wouldn’t notice. People judge you based on how you perform in office, particularly with regard to the economy, and clinging to stuff like the filibuster or not adding new states because there’s a poll that says it’s not particularly popular is a lot less important than doing what’s right to preserve and protect America’s democratic structure such as it is. The attitude inculcated by survey liberalism is one of timid defensiveness: you should always be trimming down your programs and hiding in terror of the electorate.

We need confidence in our leadership class. The elected leadership of this country has to be part of any kind of lefty political project, and just democracy in general. What has been lacking over the last twenty years is any type of vigorous defense, any type of bold leadership. That was what was so striking about Biden pulling out of Afghanistan. Not that it was done in a particularly moral fashion, but that he stood up for himself. That’s something that Obama almost never did. Bill Clinton not only didn’t stand up for himself, he was just actively doing the bad things. I think it’s more critical to cultivate a sense of righteous passion and a vision that’s morally and historically grounded — sometimes doing the unpopular things as necessary. If you’re just looking at polls all the time, you’re going to get the opposite.