Affect, Social Pressure and Prosocial Motivation: Field Experimental Evidence of the Mobilizing Effects of Pride, Shame, and Publicizing Voting Behavior

This is a review of Affect, Social Pressure and Prosocial Motivation: Field Experimental Evidence of the Mobilizing Effects of Pride, Shame, and Publicizing Voting Behavior (2010) by Costas Panagopoulos. Political Behavior 32 (September): 369-386. You can find the original in Google Scholar.

Most places give out “I voted” stickers to honor voters. Perhaps a scarlet letter on non-voters would work better.

Suppose a local newspaper planned to honor those who vote by listing their names in a post-election issue. Would you be more likely to vote?

Now, suppose a local newspaper planned to shame those who stayed home by listing their names instead. Would you be more likely to vote?

That’s the question Costas Panagopoulos asks in recent study in Political Behavior, part of a special issue (read some background) dealing with the effects of social pressure on voting. Voting is a norm. Pangopoulous divided a bunch of voters into a control group, an “honor” treatment, and a “shame” treatment. The two treatment groups received a postcard. The first received a card announcing that voters would have their names listed in the paper. The second received a card announcing that non-voters would have their names listed. All these postcards went out in November 2007, a low-turnout municipal election year.

It appears that shame motivates voting better than honor does. Both tactics work, but shame works better. Promising to honor voters boosts turnout by 0.8-1.5 percentage points1 (in Holland MI) or by 4.5-4.7 percentage points (in Monticello, IA). Threatening to shame non-voters boosts turnout by 6.3-6.9 percentage points (in Ely, IA).

Most places give out “I voted” stickers to honor voters. Perhaps a scarlet letter on non-voters would work better. Maybe we should require them to sew a big, scarlet “N” (for “nonvoter”) to their clothing and keep it there until the next election.

Comment and criticism

One problem with the research design is that the two cards went to different cities. The “honor” treatment went to voters in Monticello, IA, and Holland, MI, while the shame treatment went to Ely, IA. Pangopoulos made this decision for a valid reason: He was afraid that if both cards were sent to people in the same city that people who received the “honor” treatment might talk to folks who received the “shame” treatment. If people concluded that the paper was going to publish both lists, that could have distorted the effects. So I fully understand why Panagopoulous didn’t send both treatments out to people within the same city.

Still, this creates a difficulty. The two cities that received the “honor” treatment had a base turnout rate in the control group of 24.4% (in Holland, MI) and 30.9% (in Monticello, IA). But the city that received the “shame” treatment had a base turnout rate in the control group of 15.6% (in Ely, IA).Take a look at the table below:

City Turnout in 2004
(control group)
Turnout in 2007
(control group)
Percent of voters
that are “mobilizable”
Treatment Treatment effect
(ITT)
Holland, MI 48.4% 24.4% 24.0% Honor +0.8
Monticello, IA 69.8% 30.9% 39.1% Honor +4.5
Ely, IA 77.2% 15.6% 61.6% Shame +6.3

Think about what the table shows. In the absence of treatment, we would expect turnout to be higher in 2004 (presidential election year) than 2007 (municipal election year). And looking at the control group’s turnout in each year, we see that such was the case. By subtracting 2007 turnout from 2004 turnout, we can estimate how many voters in each city were “mobilizable”–that is, what percentage of voters had shown an interest in voting in an exciting year (2004) and might be persuaded to come out in an uninteresting year (2004). This number is far higher in Ely than in either other city.

What do we conclude? Even if the same treatment had been sent to all three cities, we would expect the effect to be strongest in Ely.

I’m not sure, then, that we can confidently conclude that shame works better than praise. In my gut, I tend to think that Pangopoulos is right. But I’m not 100% convinced that the question has been answered decisively just yet. It’s possible that whether we honor or shame people makes no difference–maybe the only thing that matters is that we’re publicizing norm compliance.

As a minor side point, why do political scientists give their articles incomprehensible, jargon-laden titles? This is fascinating research that would surely interest political journalists, but a title like “Affect, Social Pressure, and Prosocial Motivation” is far from inviting. Why not something simple like, “A Scarlet Letter for Non-Voters? Shame Mobilizes Voters Better than Praise.”

Possibly related posts: