The Washington PostDemocracy Dies in Darkness

Opinion polls can be self-fulfilling prophecies

In the United States, support for gay marriage has increased rapidly in recent years. Is this partially because people are told that more people support gay marriage? (Graphic source: Gallup)
Placeholder while article actions load

Public opinion sometimes changes quickly over relatively short periods of time. For instance, support for same-sex marriage has substantially increased to becoming a majority view over the past few years, while support for the death penalty has been steadily declining. These rapid shifts are the product of people actually changing their views rather than just cohort replacement (younger people replacing older ones in the population).

What causes these opinion cascades? Obviously, new information about policies and position taking by political elites (e.g., President Obama’s support of same-sex marriage) play a role. But people may also change their opinions by observing their fellow citizens. When they learn from public opinion polls that majorities in favor of a public policy are growing, they may jump on the bandwagon and also support these policies. This could be because they want to feel liked or accepted, they could be learning from the wisdom of crowds, or they may want to resolve cognitive dissonance by supporting what will most likely become the policy anyway. In addition to polls, there are numerous other sources for people to learn about collective opinion such as aggregated forecasts (such as The Washington Post’s Election Lab) and prediction markets. These recent phenomena have raised the question of whether the measurement of public opinion can itself influence public opinion.

In an article (open access) recently published in the journal Research & Politics, we conducted a survey experiment using a national sample recruited by YouGov/Polimetrix to explore this issue. Before asking people about their opinions on three contemporary public policy issues — withdrawing troops from Afghanistan, free trade agreements and public financing of elections — we randomly assigned respondents to receive aggregated polling information showing support for the public policies ranging from 20 percent to 80 percent. Averaging across the three issues, we found that moving public support across the range increased support for the policies by over eight percentage points. In other words, people were more likely to support the policy initiatives if they were told that they had more support in the general public.

The effect was not constant across issues. We observed the highest effect for the free trade issue (13.5 points) and smaller effects for the Afghanistan (6.3 points) and public financing (3.5 points) issues.

We offer three, related, suggestive reasons why this was the case:

  1. People are more likely to start out ambivalent on the issue of free trade, making them more susceptible to new information.
  2. People were unsure of existing public opinion on free trade, meaning that the polling information had more of an opportunity to have an effect.
  3. Free trade is not defined as much along partisan lines and people have weaker partisan predispositions on the issue. Hence, learning about majority opinion can cause attitude shifts, but the effects are strongest among people without well-defined prior support, who are previously uniformed on public opinion, and if the policy question is not highly partisan.

This first piece of research has shown that at least a component of opinion cascades is people jumping on the bandwagon after learning about the views of their fellow citizens; but, this research raises several additional questions which we hope to explore in future research. How does this effect differ between public policy questions and elections? At minimum, most elections are partisan, by definition, but they can vary in the extent to which voters are ambivalent and possess information. More generally, what type of information about collective opinion are most impactful and why? What is the mechanism underlying the results: desire to be part of the winning team or the rational processing of signals about the quality of public policies? Which types of people are most susceptible to learning about public opinion through polling information? Given the importance of studying opinion cascades given recent national debates about important public policies, we look forward to scholars addressing these questions.

Neil Malhotra is a professor at the Stanford graduate school of business. David Rothschild is an economist at Microsoft Research.