Todd Rogers, Ph.D. student at Harvard, is working on an interesting dissertation on the importance of messages in voter turnout. They randomly assigned voters in California (before the 2006 primary election) and New Jersey (before the 2005 general election) to receive a message that either emphasized low voter turnout (LTO) or high voter turnout (HTO) and saw what influence it had on whether voters actually voted.
They found that the HTO message actually produced higher turnout among those who heard the message and the LTO message reduced voter turnout. This is contrary to the *rational choice* model that would assume that voters who expected lower turnout would vote more, since they would perceive that their vote should matter more (as a percentage of all votes cast).
Interestingly, they found that the LTO vs. HTO message did not much affect the frequent voters who were likely to vote regardless of the message, but the HTO message was more likely to mobilize the infrequent voters. (It should be noted that the message — heard one time by those in the experiement — while it did change the intention to vote statistically significantly, did not produce an enormous effect — the HTO message roughly made voters 3% more likely to turnout and the LTO message surpessed voter intention by a smiliar amount.)
The researchers weren’t constrained by having to deliver truthful messages, but Todd pointed out to me that in any given year, for example with increasing population, you could emphasize high turnout messages such as “more people voted in the last election than ever before”, even if the percentage voting had decreased, and make it more likely that one would achieve the high turnout result desired.
The paper is consistent with a whole body of “social norms” research (summarized in the Rogers paper) that shows that people are more likely to conform to what they believe are social norms: for example, drinking less in college when low rates of alcohol abuse are publicized, stealing petrified wood more from forests when told that others do, reusing towels more in hotel rooms when told that others reuse towels at high rates, etc.
Note: The Rogers and Gerber paper unfortunately could only focus on “intention to vote” as a dependent outcome varaible, not actual vote turnout, so they will need to do further work to make sure that the follow-through on “intention to vote” is actually high on these more marginalized voters.
One wonders whether this applies to other forms of civic participation. Presumably it is helpful only in a tipping point sort of behavior where a fairly large number of people do this already and thus others can be encouraged to do likewise, and presumably the benefit would be greatest when there is the greatest discrepancy between people’s guesses about how often a civic action occurs and how often it really does. If some behavior is relatively infrequent (say going to a political rally in the last year), one runs the risk that disclosing how frequent this is could have the adverse affect (at the margin persuading those who do the behavior currently to quit). But some manipulation of the norms could be used, to for example, emphasize how many millions of people went to rallies in the last year rather than focusing on the fact that it was only, say, 15% of the population.
This research will be forthcoming in “Descriptive Social Norms and Voter Turnout: The Importance of Accentuating the Positive” (The Journal of Politics, forthcoming) with Alan Gerber (of Yale Univ.). Earlier version of this paper available here.