First, there are some good reasons to be dubious about the results. The authors include a doozy of a to-be-sure paragraph, noting that their situation is a hypothetical, and if it were actually attempted it would produce strong backlash, including from Republicans, crash markets, and probably fail to pass constitutional muster.
But that’s just a start. The researchers note that before asking respondents whether they would support delaying the election, they asked several other questions:
Respondents were asked whether Trump won the popular vote, whether millions of illegal immigrants voted, and how often voter fraud occurs. These questions evoke arguments frequently made by Trump and others about the integrity of the 2016 election.
In other words, the respondents were primed to be thinking about Trump’s (spurious) claims of widespread voter fraud, already shading their impressions before they got to the central question. Pollsters and political scientists long ago showed that the sequence and tone of questions can help determine the way people will answer the question. That’s especially true with the idea of delaying the election, which few voters will have considered since it’s not a topic of open debate. Poll respondents often offer what are called “doorstep opinions”: They’d rather answer a question than admit they haven’t thought about it. As the Harvard government professor Ryan Enos put it on Twitter, “Ask people about something complex they've never considered and tell them somebody they trust supports it and they will also support it.”
Besides, there are lots of dubious partisan views that show up in polls. Respondents consistently say things that are counterfactual or plainly nuts. How many voters really believed that Barack Obama was a Muslim, or born in Kenya? Probably fewer than told pollsters they did. Even the belief that Trump won the popular vote, which has shown up in other polls, could fall under the same umbrella. As Julian Sanchez has argued, such results look like “symbolic beliefs,” offered mostly to affirm fidelity to a party or politician. In the early ’90s, bumperstickers suggested, “Annoy the media: Re-elect Bush.” In today’s even more polarized, media-hating environment, the temptation to annoy the media, liberals, and Republican elites alike by backing outlandish ideas in polls is even stronger.
These weaknesses are not confined to Republican voters. The delayed-election result feeds the prejudices of progressives whose reaction to Trump is the potent cocktail of ridicule (those idiots!) and terror (those idiots are in charge!). But Democrats have sometimes expressed equally worrying views in polls. In the summer of 2016, for example, a pollster found that two-thirds of Democratic voters would trade an unconstitutional third term for Obama if it meant avoiding either Clinton or Trump. Perhaps you think, They must have been joking, and would never have followed through. But that’s just the point: What happens in polling often stays in polling. (That gives us reasons to take polls like this with a grain of salt. But as Brendan Nyhan writes, it doesn’t let people off the hook for unwise statements. Where can one draw the line between “real” and partisan beliefs?)