Government officials and legislators have come to rely on public opinion polls. Almost daily, media outlets and polling firms—not to mention the parties’ own pollsters—ask a variety of Americans for their policy preferences on a broad set of issues. Statistics derived from these polls become the justification for building new programs or sustaining existing ones. Since the New Deal, the growth of government services has been in part the result of politicians seeking to give the people their hearts’ desire.
Satisfying people’s desires is not a bad thing. It is, in fact, the essence of democracy. And polling has become the key source of information on what the public thinks and wants. The problem lies, on the one hand, with how polls are conducted, and on the other hand with how we interpret polls, and the information we draw from them.
Pollsters are professionals who want to do their job right and provide good data and valid statistical analysis. In order to ensure objectivity and neutrality and not be tainted by accusations of partisanship and “leading” questions, the polling industry has developed standard ways of asking questions. And there are good scientific reasons for that: survey respondents’ answers can be swayed by the wording of the question, and even the order in which questions are asked. Studies going back to some famous experiments in the 1980s conducted by Daniel Kahneman and Amos Tversky show that even small word changes that do not alter the message can produce major shifts in public opinion. As a result, polls primarily stick to simple single issue questions.
The question is not simply whether you like the idea of lower taxes or better education. We all do. But we have to consider which programs we are willing to give up or what balance to strike between programs, each of which requires money to effectively run.
And here lies the problem of interpretation. When we ask people about single issues, they tend, in the aggregate, to rate almost everything as important and worthy of government expenditure. Assistance for needy children? Check. Low-cost health insurance? Check. More police on the streets? Check. More funding for education? Check. Improve roads and highways? Check. Benefits for our veterans? Check. Job training? Check. Immigration enforcement? Check. Cleaner air? Check. Safer chemicals? Check. But then comes the kicker: Lower taxes? Check.
This practice of single-issue evaluation creates the illusion of a compartmentalized world where issues are one-dimensional and can be put into individual boxes. This is a problem; we may all want “more guns and more butter” but the reality is that we cannot afford, particularly right now, to pay both for all the guns we think we need for defense and all the butter we think we need to keep people educated and healthy.
The real world of politics is not made up of simple choices; it consists of tradeoffs between multiple public goods. The really hard work of politicians is to decide how finite resources can be best divvied up among many different programs and needs. And the guidance they get from their constituents via polls is no guidance at all.
We see this a lot in our research. Last year, on Election Day, we ran an exit poll in Rhode Island. On that poll we asked people about the importance of various issues from health care to foreign policy on their choice of presidential candidates. About 70 percent of our respondents said that all eight issues we asked about were either important or very important to their vote. People want more of everything, except taxes. In the abstract, without any discussion of financial consequences, we all want lots of the good things that government does, but we don’t want to pay for it.
When it comes to immigration, a policy area that happens to be of interest to the authors of this article, the polls suggest that Americans want it all. According to recent Galluppolls, the vast majority of the public supports tightened border security, including increasing technology and personnel for the Border Patrol (83 percent), requiring employers to verify “that all new hires are living in the U.S. legally” through a government-run database (81 percent), creating a system to track the departures of foreigners who enter the U.S. legally (71 percent), and increasing government spending on enforcement at the U.S. borders (68 percent). At the same time, equally high proportions support a path to citizenship for undocumented immigrants (72 percent) and more visas for immigrants with advanced skills (71 percent).
Some observers assume that this frustrating situation arises because the public lacks political sophistication. Simply put, Americans behave like children who want every piece of candy in the candy store but have no idea how much it is going to cost. There is some element of truth to this as research by political scientist Suzanne Mettler shows that Americans have little understanding of how much government does for them or what it really costs. For example, in a recent poll we conducted, a minority of seniors said they receive benefits from a government program, despite the fact that the vast majority receive both Medicare and Social Security.
While our desire for lots of government programs but low taxes is documented, current polling practices are at least slightly to blame for the fact that politicians feel the need to fulfill all of the public’s “wishes.” What politicians see is that the public supports all of these programs. But if we ask our questions differently, we might get different answers—answers that suggest the public is a lot more intelligent and willing to make trade-offs than politicians seem to think is the case.
How do we know this?
We recently ran a series of experiments where we asked poll respondents to make the tough choices. In these experiments, instead of asking people whether they wanted two expensive programs independently, we first explained that there was a limited pool of resources. Specifically, we explained that the city government had two ways of spending a pre-defined amount of money: it could either hire more police to identify and deport undocumented immigrants or it could spend it on job training. We then allowed people to allocate money so they could decide how much money would go to each program. What we found is that when they have to make a choice people prefer job training to immigration enforcement spending. Fifty-two percent devoted the lion’s share of available funds to job training while only four percent did to immigration enforcement. About one-fourth split the funds equally between the two options. As important as immigration control may be, when the choice is between spending money rounding up undocumented immigrants and spending money to help the unemployed become competitive in the modern workforce, the unemployed win.
While there were differences in the allocation preferences for both Democrats and Republicans, even Republicans preferred job training when they were forced to make a choice. Democrats allocated the vast majority of funds to job training; only 13 percent of people who even loosely identify with the Democratic Party gave a majority of the funds to immigration enforcement. But even among Republicans, 40 percent gave a majority of funds to job training, and another 25 percent split the funds evenly. In total, 65 percent of Republicans chose to spend at least half of the available pot on job training.
Some observers would say that these findings do not reflect the real world any more than single-issue questions do. After all, the choice between “more guns and more butter” is abstract unless embedded in a political, social, and economic context. How is John Q. Public supposed to make such a decision if he does not know whether there is a recession or economic growth, war or peace?
Preferences may be context-specific and people may prefer more spending on enforcement when the economy is weak (to keep competition away), and they may want more spending on job training when the economy is strong (to ensure that more people have good jobs). Our experiments checked for that, too.
Before asking people to make a choice, we gave survey respondents some hypothetical economic projections. This information did not change people’s preferences; they still opted for job training for American workers over police enforcement of immigration law when they had to pick one.
If these experiments reflect the reality of Americans’ policy preferences, what does that tell us about the state of our political affairs? First, that reliance on a naïve interpretation of polling data does not help decision-makers to understand and implement the public will. And second, that citizens need to think about the many dimensions associated with their policy preferences.
The question is not simply whether you like the idea of lower taxes or better education. We all do. But we have to consider which programs we are willing to give up or what balance to strike between programs, each of which requires money to effectively run.
While our politicians, in their interactions with constituents, should emphasize this multidimensionality of policymaking and not simply capitalize on people’s naïve preferences and emotional responses, we should also offer more opportunities for citizens to learn and appreciate the complexity of public policy. A good example is programs in participatory budgeting that some towns are trying out.
In Chicago’s 49th Ward, ordinary citizens can take part in decisions on how to allocate a portion of the budget. Individuals propose new programs, recommend cuts for others, and decide how their community should spend its limited resources. In the process, they learn what it takes to divvy up the pie and how challenging it is to serve the public good. Pollsters, too, should experiment with more complex question structures to better gauge the depth of people’s preferences. Otherwise, public discourse will continue to be inundated with poll analyses that say little about what people would choose if they were offered actual choices based on the reality of the government’s pocketbook.