W
EAVER
9
The Color of the Campaign
D
ATA
& M
ETHODS
Study Design
This study will examine how voters’ evaluations of political candidates are influenced by
the race and skin color of the candidate. To this end, a survey experiment was designed in which
respondents are exposed to visual stimuli (in the form of campaign advertisements) that alters the
race and skin color of the opponent and then asked about their attitudes toward the contenders.
First, subjects are shown the campaign literature of two opposing candidates and
informed that the candidates are running for an open Senate in a neighboring state. The literature
approximates real campaign placards with the candidate pictured in a static pose with a flag in
the background. The candidates are all males of approximately the same age.
To control for
visual candidate differences, this study used a morphing technique using photo editing software,
which digitally averages several faces together to produce distinct candidates, equating all
relevant characteristics, while altering the race and skin color of the target candidate.
Each
candidate is generated from a set of images that share at least one person in common, therefore
minimizing the effective difference between opposing candidates.
This technique allows
researchers to have a comparison of two candidates while carefully controlled visual stimuli can
elicit subject responses to the condition in question – race and skin color. Previous studies have
18
The following people aided in the design of this study: Jim Alt, Mahzarin Banaji, Don Green, Jim Glaser, Shanto
Iyengar, Steve Ansolabehere, Sid Verba, and Taeku Lee. This draft also benefited from extensive comments given
at the American Politics Research Workshop, the Political Behavior and Psychology Workshop, and the National
Conference of Black Political Scientists.
19
The author does not mean to imply that race and skin color only affect male candidates. Given the limited number
of treatment groups and conditions, the gender of the candidates must be held constant.
20
See Appendix II for a description of how images were produced.
21
White candidate 1 = generated from images A, B, C
White candidate 2 = generated from images A, B, D
Ligh-skinned black candidate = generated from images A, E, F
Dark-skinned black candidate = generated from images E, F, G
Convention