New Yorkers Support Free Buses, but They Don’t Think It Should Happen

If voters “support” a policy proposal in a poll, does that mean they think the government should enact that policy?

It sounds like a reasonable assumption. At the very least, it’s an assumption that’s been made by many public opinion researchers and political scientists. I’ve made it myself.

But “support” for a policy may not mean what it seems, based on this week’s New York Times/Siena poll of New York City. Many voters who tell a pollster they “support” a policy might not believe that the government “should” enact that very policy.

In four of five cases, fewer people said the city “should” enact the proposal than said they “support” the same policy. In the most striking case, 60 percent of New Yorkers said they supported free city buses, but only 44 percent said the city “should” make buses free. A majority, 55 percent, said the city should not.

While it’s typical for different question wordings to yield different poll results, this particular difference highlights a potentially serious issue. It raises the possibility that one of the most common measures in polling isn’t necessarily measuring what pollsters and poll consumers use it to measure. And while it is only one small experiment, it adds to lingering doubts about whether pollsters are succeeding in their highest mission: to divine the will of the people in a democratic society.

Over the last decade, polls asking whether voters “support or oppose” policies have played a modest but meaningful role in American political debate. On issue after issue, survey results have given the impression that a majority of Americans support progressive policy priorities, such as near universal support for firearm background checks, or widespread backing for Medicare for all, or Joe Biden’s legislative agenda.

In the real world, Democrats have struggled to translate apparent support in the polls into similar levels of success at the ballot box or in Congress. Republicans control the federal government and a majority of state legislatures. Mr. Biden’s “popular” agenda did little to improve his approval ratings. Referendums on background checks and single-payer health care failed or badly underperformed in blue and purple states.

The disparity between the polls and political outcomes has contributed to the sense that American democracy is failing to reflect the will of the people. Judged against the polls, academic researchers have found that elected officials greatly overestimate the conservatism of their constituents. Just two weeks ago, two excellent political scientists found that congressional decisions align with public opinion only 55 percent of the time, counting a “representation failure” as when Congress fails to pass a bill that voters say they “support” in polls.

These theories have merit, but there’s another possibility: that the polls overstate how much voters actually want policies they say they support. Voter initiatives on both the left and the right have tended to underperform positive poll numbers, raising the possibility that the polls miss a lowercase “c” conservative tendency to err toward the status quo. Pollsters reasonably counter that initiatives shouldn’t be expected to follow the findings of general issue polls, as the referendums follow protracted campaigns and often include arcane language. But the Times/Siena experiment doesn’t follow a campaign and uses the ordinary language of a poll. It offers new reason to think that the typical framing of poll questions might miss an important conservative dimension of public opinion.

The terms “support” or “oppose” can be nuanced. In many contexts, to give “support” is to express positive encouragement. There are many cases when someone might want to express a positive sentiment about something with which they have real reservations. In everyday life, one might offer support to a friend who quit a job, even if it was probably a mistake. And if you want a roof deck, it would be hard to say “I oppose building a roof deck” simply because you can’t afford it and won’t do it for the foreseeable future.

The “should do or should not do” question, in contrast, is more squarely about whether to undertake the proposed course of action. It prompts respondents to consider the consequences of acting on an idea, not just their feelings about it — though many may still want to express what they think “should” be done in an ideal world. It also comes closer to testing the respondent’s lowercase “c” conservative aversion to change, although no poll question can capture the gravity of acting in the real world.

On balance, support for the five policy proposals declined in our experiment when voters were asked whether the government “should” do the policy. The decline was greatest for free busing: Voters supported the idea by a wide 60-39 margin, but only a minority said it “should” be done, 44-55. The decline in support was less for other issues, with no decline in support for raising taxes on the rich.

The relative resilience of raising taxes on the rich, together with the decline in support for free busing, suggests that the “should or should not” question does more than account for resistance to change. It turns respondents into hypothetical decision makers, and their answers may be likelier to be shaped by analysis of the cost-versus-benefits, feasibility, opportunity costs and what’s necessary as opposed to what’s simply desirable.

Across these dimensions, free busing faces more serious questions than taxing the rich. In general, free services are easy to “support” — who doesn’t want free things? But voters may see them as costly, challenging to implement and ultimately not worth it. Higher taxes, on the other hand, are generally undesirable but potentially necessary (and far easier to swallow if they don’t affect you). They’re also easy to implement.

And though we can’t know for sure, it’s possible that some voters who say they support free buses might not believe it is a good idea in any meaningful sense. That may seem counterintuitive, but it is possible that some voters see a free public good as something like a slice of cake: They love cake, but know — and feel — they shouldn’t eat it.

Does this mean the “support or oppose” question is bad, and ought to be replaced by “should or should not?” Not at all. Sometimes, all that matters is whether the respondent would like to express something positive or negative about an idea. In this particular case, the “support or oppose” question may suggest that Zohran Mamdani is on relatively safe political ground in pushing free busing as a candidate. Ultimately, the two questions measure different attitudes, and under some circumstances one question may be more to the point than the other.

Public pollsters and political scientists, however, have generally been clear about polling’s purpose: to figure out what the public wants so that its representatives can better account for public opinion, and so that representatives can be held to account. By that standard, it is not at all clear whether the “support or oppose” question is up to the task. Worse, it isn’t clear how one even could determine whether it’s up to the task.

Unlike with election polling, there are no “results” for comparison when it comes to issue polling. That means there’s no benchmark for determining whether any particular question yields an accurate or even useful measure of public opinion. As a consequence, issue polling can never be proved right — or wrong.

For that same reason, this experiment doesn’t prove that the “support or oppose” question is wrong. Maybe it suggests that New York City voters really do want free busing, but they have doubts about its practicality, which Mr. Mamdani might overcome if he could implement the program effectively in office. If so, the story might resemble what happened with the Affordable Care Act, which voters said they opposed before it was passed, but ultimately supported after it was enacted and under threat of repeal.

But let’s say for argument’s sake that New Yorkers do not think the city should enact free buses. Then the result of the “support or oppose” question misrepresented what the public wants its elected officials to do by a wide margin — a margin that would dwarf any recent error in presidential election polling.