Mobilization Effects Using Mail
| Author | Richard E. Matland,Gregg R. Murray |
| Date | 01 June 2014 |
| DOI | 10.1177/1065912913499234 |
| Published date | 01 June 2014 |
| Subject Matter | Articles |
research-article2013
Article
Political Research Quarterly
2014, Vol. 67(2) 304 –319
Mobilization Effects Using Mail: Social
© 2013 University of Utah
Reprints and permissions:
sagepub.com/journalsPermissions.nav
Pressure, Descriptive Norms, and Timing DOI: 10.1177/1065912913499234
prq.sagepub.com
Gregg R. Murray1 and Richard E. Matland2
Abstract
We use field experiments in Texas and Wisconsin to address voter mobilization and turnout by evaluating nonpartisan
get-out-the-vote (GOTV) messages delivered via mail during 2010 gubernatorial campaigns. We manipulate three
factors in the messages: social pressure, descriptive- and injunctive-voting norm consistency, and message timing.
The results present an initial field-based confirmation that norm-consistent messages increase turnout; demonstrate
significant message timing effects, which are mediated by state election rules; and indicate social pressure’s effectiveness
varies significantly more than previously found. These diverse findings suggest researchers place a greater emphasis on
context when evaluating experiments and the effects of mobilization messages.
Keywords
voter mobilization, voter turnout, social pressure, social norms, field experiment
We use a field experiment to test the effects of nonparti-
decision to vote. We found previously that voters’
san get-out-the-vote (GOTV) mailings to contribute to
response to GOTV messaging varies by voters’ propen-
two separate literatures. First, we add to the mobilization
sity to vote. We retest the model in a new context.
literature on the effectiveness of GOTV techniques. In
Furthermore, motivated by expectations derived from
recent years, there has been a surge in field experiments
the RAS model, we provide a first test of whether timing
testing various mobilization methods (Abrajano and
impacts mailing’s effectiveness.
Panagopoulos 2011; Garcia Bedolla and Michelson 2012;
We start by discussing previous scholars’ results,
Gerber and Green 2000; Gerber, Green, and Larimer
pointing to a number of unresolved questions. Next, we
2008; Green and Gerber 2008; Green, Gerber, and
present our design and empirical results. We conclude
Nickerson 2003; Michelson 2003, 2006; Michelson,
with a discussion of implications for voter mobilization
Garcia Bedolla, and McConnell 2009; Nickerson 2008;
and turnout as well as with an admonishment that the
Panagopoulos 2010, 2011, forthcoming). While these
context in which a mobilization message is being used
studies provide evidence that GOTV tools can increase
can dramatically influence its effectiveness.
turnout, our knowledge remains incomplete concerning
under which conditions these mechanisms work. We pro-
Mobilization Experiments Testing
vide the first field test of Gerber and Rogers’ (2009) argu-
ment that GOTV messages using consistent injunctive
Social Pressure and Norm
and descriptive norms increase turnout more than mes-
Consistency
sages where the two norms are inconsistent. Furthermore,
In a recent review, Green, McGrath, and Aronow (2013)
we consider the effectiveness of social pressure in raising
find direct mail effects on turnout are generally very
turnout. Previously published field experiments have
small, but effects are larger for mailings that invoke social
been conducted primarily during low-salience elections
pressure. Their meta-analysis shows the seventy-nine
and in a limited geographical area. We test for effects in
“conventional” nonpartisan mailings surveyed produced
the context of competitive gubernatorial elections with
substantially higher base turnout. We test for these effects
1
in two very different settings with distinct politics and
Texas Tech University, Lubbock, TX, USA
2Loyola University Chicago, IL, USA
political cultures: Texas and Wisconsin.
The second literature we contribute to is the voter
Corresponding Author:
turnout literature. We retest our model of voter turnout
Richard E. Matland, Department of Political Science, Loyola University
Chicago, 337 Coffey Hall, 1032 W. Sheridan Rd. Chicago, IL 60626,
(Matland and Murray 2012), which adapts Zaller’s
USA.
(1992) Receive-Accept-Sample (RAS) model to the
Email: rmatlan@luc.edu
Murray and Matland
305
an average increase in turnout of 0.19 percentage points
self-mailer tested in a mayoral race in a small California
(confidence interval [CI] = 0.11–0.28), while the twenty-
municipality. With a control sample turnout of 10.6 per-
nine social pressure treatments generated an average
cent, they found self-mailers increased turnout by 1.4,
increase of 2.85 percentage points (CI = 2.69–3.01). The
2.0, and 3.1 percentage points. Finally, Green, Larimer,
impact of social pressure mailings varied depending on
and Paris (2010) report an unpublished experiment con-
the sample used, the political context, and the form of
ducted just prior to the 2008 presidential election where
social pressure exerted. On the high end, Gerber, Green,
individuals contacted for the 2006 study were sent a new
and Larimer (2008) used a single mailing that empha-
mailer once again showing their voting history. In a very
sized the civic duty to vote, provided citizens with their
high-salience election environment,3 this second applica-
own voting history and those of their neighbors, and
tion of social pressure to the same population failed to
promised to mail updated voting records for the neigh-
have an effect.
borhood after the election. This mailer led to an impres-
Although these results show that social pressure mail-
sive 8.1 percentage-point jump in turnout in the August
ings have an impact, there are still unanswered questions.
2006 statewide primary in Michigan; control turnout was
First, these studies have been conducted primarily in low-
29.7 percent. They also tested a “self” mailer, in which
salience elections, particularly special elections, prima-
individuals were urged to vote, made aware they were
ries, or local elections. These messages may get habitual
being monitored by including copies of the voting history
voters to the polls for a low-salience election, but they
of household members, and promised an updated voting
may be less effective with citizens who vote less regu-
record after the election, but with no explicit threat of
larly. Of the published work, only the Mann (2010) study,
public shaming. This “self” message raised turnout by 4.9
which used a sample of low-propensity voters, was con-
percentage points. In a follow-up, Gerber, Green, and
ducted during a statewide general election. In addition,
Larimer (2010) identified a sample of voters for local
many of the studies have been fielded in Michigan. The
elections in Michigan in November 2007 who had voted
robustness of social pressure effects is still up for debate.
in only one of two previous elections. They randomly
These findings should be replicated in different electoral
treated half of the sample with mailers describing the
contexts and state political cultures. To do precisely that,
individual as a voter and the other half with mailers
we test the following hypotheses:
describing the individual as a nonvoter. Messages indi-
cating an individual had voted in a recent election
Hypothesis 1A (H1A): Subjects receiving a social
increased turnout by 4.1 percentage points, while mes-
pressure mobilization message will vote at higher
sages indicating an individual had failed to vote increased
rates than subjects receiving no mobilization
turnout by 6.4 percentage points. Sinclair, McConnell,
message.
and Green (2012) replicated the “self” mailer in a
Congressional special election held in Chicago in 2009
Hypothesis 1B (H1B): Subjects receiving a social
and got an estimated effect of 4.2 percentage points.
pressure mobilization message will vote at higher
Some studies show more modest impacts. Mann
rates than subjects receiving a general civic duty
(2010) tested three versions of the self-mailer varying the
mobilization message.
amount of social pressure and without a promise to send
a follow-up mailer with updated voting records. He pro-
We also desire to test norm consistency in GOTV
duced effects of 3.1, 2.6, and 2.5 percentage points in the
mailers. Gerber and Rogers (2009) distinguish between
2007 Kentucky governor’s race. His sample, however,
two types of social norms: injunctive norms (what people
was exclusively single females who were historically less
ought to do) and descriptive norms (what people actually
likely to vote.1 Abrajano and Panagopoulos (2011) tested
do). An injunctive norm is that people should not walk
the impact on Latino voters of a postcard that noted the
against the light at a crosswalk. A descriptive norm, at
voter failed to vote in the previous municipal election and
least in most college towns, is that almost everyone walks
urged them to vote in a special election to fill a New York
against the light. Virtually all GOTV messages empha-
City council district seat. Interestingly, they find an
size the injunctive norm that one should vote, but implicit
English-language mailer had a stronger impact than a
in many GOTV messages is a descriptive norm bemoan-
Spanish-language mailer despite the target audience
ing the fact that many do not vote. Gerber and Rogers
being Latino. The impacts were modest with 3.9 percent
suggest that this inconsistency...
To continue reading
Request your trial