For decades, political scientists have puzzled over problems of collective action that arise when large numbers of people are asked to contribute to a public good. When any one person’s contribution to the collective cause is negligible and the outcome can be enjoyed even by people who do not contribute, collective action fizzles because no individual has an incentive to sacrifice for a collective cause. This grim analytic framework is often applied to voting, where individuals are asked to expend time and effort, yet have little chance of casting a pivotal vote (Downs 1957).

The fact that large numbers of people do in fact vote has led scholars to theorize about the “selective incentives” (Olson 1965) that induce people to participate in elections. One hypothesis is that people derive intrinsic satisfaction from casting their ballots. They either enjoy the act of voting per se or feel good about themselves for advancing a partisan cause or honoring a civic obligation. A second hypothesis, and the focus of this symposium, posits that people receive side-payments when they contribute to a collective cause. In electoral systems where bribes and other material inducements are rare, incentives are thought to be social in nature: voters are rewarded by the approbation of others, while nonvoters are criticized or shunned. In other words, people are rewarded or punished according to whether they comply with social norms (Cialdini and Trost 1985), such as the expectation that citizens ought to participate in elections.

When authors in this volume use the term “social pressure,” they have in mind communications that play upon a basic human drive to win praise and avoid chastisement. Social pressure may be exerted by praising those who uphold norms or scorning those who violate them. Social pressure increases with the amount of praise or scorn, which is why social pressure is thought to be amplified when a person’s compliance with social norms is disclosed to others. For this reason, social pressure communications typically involve three ingredients: they admonish the receiver to adhere to a social norm, indicate that the receiver’s compliance will be monitored, and suggest that the monitored behavior will be publicized.

In practice, social pressure inducements to vote vary from draconian to discreet. At one extreme are policies and programs designed to shame those who fail to vote. At various times and places, governments and newspapers have made a point of chastising nonvoters (see Gerber et al. 2008, p. 35 and Panagopoulos in this volume for examples). At the other end of the spectrum are gentle reminders to do one’s civic duty by participating in elections. Somewhere in between are messages that subtly inform recipients that whether they vote is public information by thanking them for participating in a recent election.

How much social pressure is needed in order to induce people to vote? What kinds of social pressure messages are most effective in generating turnout, and why? Although these questions are closely connected to venerable literatures in psychology on compliance with social norms, the creative spark that attracted interest in this subject came from a researcher outside the academy. A campaign consultant named Mark Grebner, in search of a cost effective way to increase turnout, developed a form of direct mail that presented voters with information about whether they and their neighbors voted in recent elections. Grebner’s intuition was that nonvoters are ashamed of abstaining but believe they can get away with it because no one knows whether they in fact voted.

While conducting GOTV campaigns, I never met a voter who genuinely needed reminding, since none of them were surprised to be told it was Election Day. Nor did I meet anyone who resisted the idea that voting was good, normal, important, expected, a duty. But some of them—I later found by checking records—still failed to vote. My attention was drawn to the remarkably large fraction of people who falsely claimed to have voted, as reported by surveys following each November election. Ten days didn’t seem like enough time to have genuinely forgotten; it seemed more like intentional lying. This led me to suspect that a substantial number of potential voters believed that voting is essentially a private act, rather than a public one, and it was safe to routinely misreport voting not only to researchers but to friends and family. That would explain why reminding was ineffective; voters weren’t unaware of the message, but deliberately evading it. I decided to try to test that model, by seeing if the threat of public exposure would force at least some of them to abandon their pose, and actually vote, by making them think they couldn’t continue to get away with it. (Personal communication, 2010)

His self-funded early experiments showed promising results: disclosing vote histories increased the effectiveness of direct mail by an order of magnitude.

After Grebner brought his in-house experiments to our attention, the authors collaborated with him and Christopher Larimer on a five-group experiment designed to isolate the active ingredients in Grebner’s social pressure mailings. One ingredient is scolding: our first experimental treatment was a mailing that enjoined recipients to do their civic duty. Another ingredient was surveillance: in addition to telling people to do their civic duty, we announced that they were part of an academic study, and that their participation in the upcoming election would be monitored. A further ingredient is disclosure of norm compliance: a third treatment reported whether each voter in the household had voted in recent elections and promised to send an updated mailing to indicate whether ballots were cast in the upcoming election. A final treatment amplified the level of disclosure by reporting not only whether members of the household voted but also whether others on the block voted. Turnout in each experimental group was compared to turnout in a large control group, which received no mail.

These mailings, which Grebner designed and distributed to 80,000 households, demonstrated that turnout increases as social pressure increases. When maximal social pressure is applied, a single mailing increased the probability of voting by more than 8% points, which is an astonishingly large effect that rivals even the effect of face-to-face contact with a canvasser. Even the mailing that disclosed the voting record of one’s household had an effect of between 4 and 5% points. Bear in mind that a single piece of direct mail, partisan or nonpartisan, rarely increases turnout by more than 1% point (Green and Gerber 2008).

Like the early airplane, Grebner’s invention was both intriguing and hazardous. From a theoretical standpoint, the newfound ability to generate large treatment effects meant that social scientists could now introduce exogenous changes in turnout rates in order to study how campaign messages diffuse within and across households (McConnell et al. 2010). The magnitude of the effect also facilitated the rigorous investigation of heterogeneous treatment effects across individuals (Feller and Holmes 2009; Imai and Strauss 2009) and electoral contexts (Larimer 2009). The capacity to generate large treatment effects using social pressure laid the groundwork for methodological inquiry into efficient experimental designs (Gerber et al. 2010a) and the accuracy of estimates generated by regression discontinuity designs (Green et al. 2009). As a practical matter, however, the use of social pressure messages presents risks for actual political campaigns, which worry that heavy-handed tactics will alienate potential supporters.

This volume presents four essays that contribute theoretically to the understanding of social pressure while at the same time offering insights that potentially inform campaign practice. The first essay, by Christopher Mann, was conducted in collaboration with an actual voter mobilization campaign. The research design sought to maximize effectiveness while minimizing backlash against the organization sending the mail. All four of the experimental mailings revealed the past voting records of members of a household. The first condition replicated the Gerber et al. (2008) Self treatment, but the remaining three experimental conditions attempted to disguise the social pressure. In the second experimental condition, monitoring of past voting was said to be the reason the mailing provided assistance in finding information about candidates and polling places locations. In the third treatment, the recipients were told they were part of research on voting behavior, and the final treatment characterized the mailing as part of research and included a survey of recipients’ opinions. The question is whether softening the social pressure message with offers of assistance, a survey, or a research cover story lessens its effectiveness. The answer is no—in comparison to a control group that received no mail, each of the treatment mailings produces sizeable increases in turnout of similar magnitude. The implication is that surveillance drives the effect. The mailers work because they each make apparent the fact that the recipient’s compliance with the norm of voting is a matter of public record.

Tiffany Davenport’s study of social pressure reinforces this conclusion using a very different mode of communication, door-to-door canvassing. Working with a nonpartisan political group that focuses on the concerns of public housing communities, Davenport randomly assigned tenants in public housing developments to receive one of two kinds of canvassing appeals. The first treatment group received a standard get-out-the-vote message. The second treatment group received, in addition, an official-looking document listing their past participation in national, state, and local elections. Canvassers transmitted this information without admonishing recipients about their failure to vote in the past. Davenport finds significant effects for both kinds of canvassing, but the effects are especially strong when canvassers distribute voter histories. The implication is that monitoring of voting behavior—without scolding nonvoters—is sufficient to generate substantial increases in turnout.

The essay by Costas Panagopoulos shows how the social pressure effect depends on whether one promises to reveal compliance or noncompliance with voting norms. In three related experiments conducted under similar conditions, Panagopoulos randomly distributed mailings in which people were either notified that voters would be honored by having their names listed in a local newspaper or warned that nonvoters would have their names revealed in a local newspaper. Although both mailings are effective, the evidence suggests that the latter, the shaming approach appears to produce a much larger increase in turnout. Whereas the shaming mailing increases turnout among all segments of the electorate, the pride treatment only increases turnout among those with a high propensity to vote.

Whereas the Panagopoulos study looks at the promise of future praise or shaming, the Gerber et al. (2010b) study in this volume randomly varies the information people receive about their past voting record. Focusing on voters who had voted in just one of the past two elections, the authors sent a social pressure mailing that reported voter turnout in one randomly selected election. Turnout rises sharply in both experimental conditions, but the effect is significantly stronger when people were shown a past election from which they abstained. This finding may be interpreted to mean that social pressure messages are especially influential when they call attention to observable noncompliance.

Taken together, these four essays illuminate several important aspects of social pressure as applied to voter participation.

  1. 1.

    Monitoring of compliance with social norms appears to be crucial. Admonitions to vote without monitoring (or credible indications that monitoring occurs) are ineffective.

  2. 2.

    The content of the feedback also seems to matter. Calling attention to past nonvoting is particularly effective, as is the threat of shaming non-compliers.

  3. 3.

    Consistent with the hypothesis that voters are chronically aware that voting is normative, it is unnecessary to hector recipients about the importance of fulfilling their civic duty so long as one demonstrates that compliance will be monitored.

The final essay in this collection provides a brief synopsis of the long-term effects of social pressure interventions. Due to their large size, the three studies using direct mail provide an especially clear indication that effects persist from one election to the next. The Gerber et al. experiments conducted in 2006 and 2007 continued to have significant effects on voter turnout through the presidential primaries of 2008 but show no detectable effects in the November 2008 general election. Mann’s interventions in 2007 significantly increased turnout in both the 2008 primary and the general election. The Panagopoulos findings fall in between: significant effects in the November 2008 are found in one of the two sites that showed significant effects in 2007. These enduring effects are open to different interpretations. One hypothesis is that social pressure mailings have special resonance, and recipients remember them long after the election has passed. Another hypothesis is that by increasing turnout in one election, the mailings create voting habits. By demonstrating the persistent effects of social pressure, these experiments set the stage for more refined experimental designs that will help sort out these competing explanations.

The rapidly growing experimental literature on social pressure is not only of immense significance in political science, where problems of collective participation abound, it also speaks to cognate literatures in social psychology and behavioral economics. Recent years have seen a surge of interest in compliance with pro-social requests. For example, field experiments have gauged the extent to which different kinds of feedback encourage consumers to conserve resources (Goldstein et al. 2008; Ayres et al. 2009) or make charitable donations (Karlan and McConnell 2009). Many of these interventions, particularly those indicating that compliance will be monitored and disclosed to others, are similar in structure to the voting experiments described above. Given the pace, scale, and nuance of social pressure research across social science disciplines, we should soon expect a new empirically grounded understanding of the conditions under which people comply with norms.