WikiSummary, the Social Science Summary Database

 

Gerber, Green, and Shachar. 2003. Voting may be habit-forming: Evidence from a randomized field experiment. AJPS 47:540-550.

Main Argument

This is an updated, more persuasive form of the argument in Green and Shachar (2000). The authors use an experimental design to demonstrate that the act of voting in one election makes you more likely to vote in future elections, even with everything else controlled. Three contributions:

1. Demonstrates that randomized experimentation can show whether "habit" plays a causal role (as in this case) or whether it simply masks omitted variables;
2. Shows that, at least in the short run, voting in one election influences turnout in the next. Past voting appears to boost the probability of current voting by 47 percentage points. Contrast this with Wolfinger and Rosenstone's (1980) finding that those with a postgraduate education are 26 points more likely to vote than those with a high school education, or that those in their 80s are around 30 points more likely to vote than those in their 20s, and you see what a large effect habit has.
3. Suggests the importance of considering possible long-term effects (e.g. habit) when studying the behavioral consequences of campaigns.

Methodology

Design and Conditions

In 1998 elections, 25000 registered voters in New Haven were randomly divided into one of several subgroups. Other than the control groups, all subjects received one or more of these treatments:

• Personal contact. Paid assistants, mostly graduate students, would visit you at home and deliver a brief message urging you to vote.
• Mail: You would receive one, two, or three mailings urging you to vote before election day.
• Telephone appeal: Similar to the personal contact. Dropped from the analysis (didn't boost turnout).

Outcome Variables

They then used public records to determine exactly who voted in 1998. Sure enough, personal contact and mailings boosted turnout (by one or two percentage points, for the most part).

Habit

They then used public records to determine exactly who voted in the 1999 mayoral race. Though turnout generally was down (b/c it was a less prominent race), there were still residual effects. Using instrumental (2SLS) regression analysis, those who were contacted in 1998 were still more likely to turn out in 1999. They use several controls at various points: Whether you voted in 1996, demographic and political controls, etc.

Theoretical explanations

Recognizing that their data may not be persuasive, they nonetheless evaluate four possible explanations for these findings. They like the third.

1. Parties and candidates send more mail to people who voted last time than to nonvoters. Problem: This may be true, but the empirics control for how frequently you have been contacted by various groups, so this should be controlled for.
2. Participation boosts political efficacy (see Finkel 1985). This may be true, but the ANES controls for feelings of political efficacy.
3. The ANES does not control for "conative attitudes," a pretentious term that basically means self-confidence in the physical voting process. If you've never voted, you might not know how to register, or request an absentee ballot, or you might be afraid of being embarrassed by your inexperience when you arrive at the polls. But prior experience voting makes you less worried about these things. Thus, these "conative attitudes" might explain "consuetude."
4. Voting reinforces an identity as a good, civic person. As this identity is constructed, you continue to vote. Thus, both voting and abstention influence your self-view, thereby influencing future turnout.