Politicians Don’t Actually Care What Voters Want
We like to think that politicians care about what their constituents want. If voters in a legislative district have certain views about, say, the legality of abortion, we assume that their representative’s decisions will be shaped, or at least influenced, by those views. To a large extent, democracy depends on this assumption: The beliefs of voters should be reflected, however imperfectly, in the leaders they elect.
Illustration by Nicholas Konrad;
photograph by Christopher Lee for The New York Times
But there is reason to question this assumption. It is easy to think of issues, climate change and gun control chief among them, where the consensus of public opinion has provoked little legislative action. How much do legislators really care about the views of their constituents?
Over the past two years, we conducted a study to find out. We provided state legislators in the United States with access to highly detailed public opinion survey data — more detailed than almost all available opinion polls — about their constituents’ attitudes on gun control, infrastructure spending, abortion and many other policy issues. Afterward, we gauged the willingness of representatives to look at the data as well as how the data affected their perceptions of their constituents’ opinions.
What we found should alarm all Americans. An overwhelming majority of legislators were uninterested in learning about their constituents’ views. Perhaps more worrisome, however, was that when the legislators who did view the data were surveyed afterward, they were no better at understanding what their constituents wanted than legislators who had not looked at the data. For most politicians, voters’ views seemed almost irrelevant.
The study worked as follows. First, we took data from the Cooperative Congressional Election Study, an enormous online survey of political attitudes that contains enough participants to allow the data to be informative about the specific political beliefs of constituents within thousands of state legislative districts. Then we hired a design firm to create a visually appealing, easy-to-use website to display constituents’ preferences to legislators. We called the website District Pulse.
We then contacted 2,346 state legislators. At random, we offered about half of them access to District Pulse. We invited the others to view a version of District Pulse that did not offer information about the attitudes of their particular constituents but rather provided information about the attitudes of all Americans, split up into the four broad regions used by the Census Bureau.
By randomizing in this way, we were able to isolate the effects of the district-specific polling information. If some legislators were lazy, for instance, and failed to use the site for that reason, we could be confident that their laziness was not skewing our findings about district-specific polling information, since both groups would be likely to have roughly similar shares of lazy legislators.
We gave each legislator a unique username and password, which allowed us to track exactly who used the site and for how long. Shortly thereafter, we surveyed the legislators who went to the site about their perceptions of their constituents. We then compared the accuracy of those who’d received the district-specific information with those who saw information only about the four census regions. (To unobtrusively measure representatives’ perceptions in the follow-up survey, we asked a colleague of ours to distribute it, with no reference to us or District Pulse.)
When we first did the experiment, in the fall of 2017, only 11 percent of legislators viewed the website to learn about their constituents’ policy preferences. Among those who did, their accuracy in understanding their specific constituents’ preferences was statistically indistinguishable from the accuracy of those legislators who hadn’t seen their constituents’ preferences. In both cases, legislators had little sense of the policies their constituents supported.
To understand how this looked in practice, consider the following. In the average legislative district in our experiment, seven out of 10 constituents support eliminating mandatory minimum sentences for nonviolent drug offenders. But when legislators were asked how many of their constituents they believed supported such a policy, they responded that fewer than two out of 10 constituents did — even if they’d seen their own constituents’ preferences.
Last fall, with national elections looming, we re-sent the polling information and re-administered the survey to legislators who hadn’t used the website previously. This time, we even explicitly reminded a random subset of legislators about the coming election and how polling data would be valuable to their re-election efforts. Yet our results were unchanged: Not many legislators visited the site, and those who did and who saw their district’s preferences were no more accurate for doing so.
No one wants or expects politicians to march in lock step with their voters. Politicians are not supposed to mechanically replace their own views with the views of their constituents. But constituents’ perspectives should carry considerable weight. Our study suggests that for most politicians, voters’ views carry almost no weight at all.
Joshua Kalla is an assistant professor of political science and data science at Yale University. Ethan Porter is an assistant professor of media and public affairs at George Washington University.
Who is online