There’s something strange about how Americans see the economy these days – or, at least, how they’re describing it to pollsters. For one thing, there’s a stark partisan divide in responses to even factual economic questions. Take an April CBS News/YouGov poll that found Democrats were not only 41 percentage points likelier than Republicans to say that the condition of the national economy was good overall but also 29 points likelier than Republicans to correctly say that the number of jobs in the US had risen over the past year. There are similar fault lines on concrete questions about people’s own behavior and experiences. In a recent CNN poll, Republicans were 22 points likelier than Democrats to say they’d cut back on driving due to economic conditions, and 25 points likelier to report delays in receiving purchases. In another poll taken this spring, Republicans were more than twice as likely as Democrats to say they’d changed their summer vacation plans due to gas prices, 41% to 16%, dwarfing the divides along other demographic lines. And in a Monmouth University poll released late last year, the share of Republicans saying they found it at least somewhat easy to pay grocery bills plunged 38 points from 2019, while the share of Democrats saying the same rose by 7 points. “There has always been something of a partisan divide on the question of making ends meet based on who controls the White House,” Monmouth polling director Patrick Murray wrote. “But the huge shift in this poll, driven mostly by Republicans, begs the question of whether we are measuring the primacy of partisan identity more than an accurate self-assessment of economic conditions at home.” The same is true of broader economic perceptions. Polling found Democrats’ and Republicans’ relative impressions of the economy’s strength inverting shortly after the 2020 election. And these increasingly visible partisan dynamics pose new difficulties for attempts to measure Americans’ objective views. “Unfortunately, the size of the partisan divide in expectations has completely dominated rational assessments of ongoing economic trends,” wrote Richard Curtin, who leads the University of Michigan’s closely watched consumer sentiment surveys, which have polled Americans on the economy since the 1940s. In a January memo, he noted that the partisan gaps seen on consumer expectations during the Biden and Trump administrations were roughly twice as large as they’d been during the Obama and George W. Bush presidencies. “This situation is likely to encourage poor decisions by consumers and policy makers alike. While there have always been partisan differences in preferred policies, the overwhelming size and persistence of the partisan gap has generated substantial economic uncertainty.” But although the results are obviously influenced by partisanship, it’s less clear what precise form that influence is taking. In one reading, the divides are largely an example of partisan cheerleading – respondents using polls as a convenient opportunity to tout their partisan allegiances, rather than a faithful recounting of what they believe to be the truth. In another reading, they’re a sign that partisan thinking, fueled by polarization and reinforced by partisan divides in news consumption, has fully overtaken the way many Americans perceive their own reality. That ambiguity can make it difficult to figure out the extent to which people mean what they tell pollsters – a challenge that has key implications for understanding public opinion not only on the economy, but also on everything from vaccine refusal to a susceptibility to political misinformation, such as the false belief that the 2020 election was stolen. People don’t always talk literally, and they don’t always answer polls literally, either. For some Americans with strong political views, answering a poll may represent an easy opportunity to vocally demonstrate their loyalties, an effect known by terms such as “partisan responding” or “expressive responding.” “Just as people enjoy rooting for their favorite sports team and arguing that their team’s players are superior, even when they are not,” the authors of one 2013 paper explained, “surveys give citizens an opportunity to cheer for their partisan team.” When the impulse for hyperbole collides with partisan loyalties, it can produce credulity-stretching results on far more than economic assessments. The same tendency may help to inflate estimates from the share of Americans willing to secede from the US to the number of unvaccinated workers willing to quit over a vaccine mandate – both actions that are easy to endorse in a poll but that would be drastic and painful in practice. It may also complicate efforts to measure Americans’ willingness to embrace political violence. That’s a topic that’s sparked some disagreement among political scientists, who’ve argued over just how closely a theoretical willingness to endorse some kind of political violence translates into support for specific violent acts or a willingness to personally commit such acts. And the partisan urge can even lead people to deny evidence that’s plainly visible. In a 2017 study, when Americans were shown side-by-side images of the crowds on the National Mall during the Trump and Obama inaugurations, 15% of Trump voters claimed the sparser photograph from the Trump event showed a larger group of people. “The debate about whether this is partisan cheerleading or genuine partisan-motivated reasoning – where people really, truly believe these things – comes down to what’s really in people’s hearts and minds. It’s ultimately sort of a really difficult thing to know or get a sense of,” said Brian Schaffner, a political scientist at Tufts University who co-authored the study with YouGov’s Samantha Luks. “But the fact that as soon as the presidency changes hands from one party to the other, all of a sudden people evaluate their economic situations very differently suggests that this is not necessarily just genuinely held beliefs. … Some of this, at least, is driven by people who are just using the survey as a way to express their support for their party.” The tendency of some poll respondents to answer expressively isn’t new, and it’s not always motivated strictly by partisanship. Immediately following the 9/11 terrorist attacks in 2001, for instance, the share of Americans who told Gallup the economy was good or excellent spiked from 32% to 46% in what may have been a brief bout of patriotic reluctance to criticize the US in any way. But in an era marked by high partisanship, pollsters and academics have made a variety of efforts to separate out such expressive responding from genuine belief. One experiment, conducted in 2008, found that offering small financial incentives for giving correct answers to factual, partisan-inflected questions (for example, how employment numbers or the federal deficit had changed during the Bush administration) was enough to to diminish the gap between Democrats’ and Republicans’ responses. The authors cited the outcome as an indication that “some portion of partisan polarization in survey responses about facts – perhaps a very large portion – is affective and insincere.” A more recent effort used a technique known as a list experiment in an attempt to measure the share of Republicans who truly believed that the 2020 election was stolen from Donald Trump. Under this format, poll respondents are asked to read a list of statements and then – rather than registering their responses to each individually – to tally up the total number they agree with. This, at least in theory, deprives them of “the reward of telling a pollster something that expresses a feeling rather than a belief.” The share of Republicans who reported believing that Biden had legitimately won was just as low among those who completed the list experiment as among those who answered a more standard question, which, the authors suggest, indicated that “Republicans are reporting a genuine belief that Biden’s election was illegitimate.” Asking follow-up questions can help pollsters refine their understanding of the strength of respondents’ convictions. Back in 2011, a Washington Post poll found that a tenth of Americans subscribed to the false theory that Obama was born outside the US – when pressed, however, most admitted that that was only their “suspicion,” rather than something for which they believed there was “solid evidence.” More recently, CNN polling found that 61% of those Americans who falsely maintain that Biden did not legitimately win the 2020 election say they think there’s solid evidence for their claims, with 39% saying it’s only a suspicion. Data on people’s actual behavior can sometimes validate polling as well. One useful case study for this approach has been America’s Covid-19 vaccination campaign, for which there is both a plethora of polling on Americans’ vaccine uptake and corresponding statistics from the Centers for Disease Control and Prevention. While neither data source is perfect, during the initial US vaccine rollout, the CDC’s numbers generally mirrored overall vaccination rates as reported in polls – and also confirmed that the relatively low uptake among Trump voters measured by surveys was real. That array of evidence suggests that there’s no single answer on the extent to which expressive responding drives survey results. “One of the most consistent things I find across my research is that none of these phenomena are all-or-nothing. They’re all context-specific,” said Matt Graham, a political scientist at George Washington University who has examined different methods for measuring the effect of expressive responding. Nor is there, necessarily, even a clear dividing line between a sincerely held belief and a sufficiently determined willingness to act as though that belief were true. On a practical level, the difference may not always matter that much. When it comes to this year’s midterms, for instance, a voter who honestly believes that unemployment is on the rise, and a voter who’s sufficiently motivated by partisan animus to act as though it is, are both likely to cast the same vote. But it’s a useful reminder that public opinion is complicated – and that the way people express their views about politics isn’t always entirely literal.