March 14, 2006 (4:30pm) / 22
control
treatment
5%
25%
45%
65%
85%
All Participants
control
treatment
5%
25%
45%
65%
85%
Democratic Participants
control
treatment
5%
25%
45%
65%
85%
Republican Participants
Figure 2: Perseverance in the John Roberts
Experiment. All panels plot percentages disapproving
of John Roberts. Black bars represent 95% confidence
intervals. In all cases, the treatment consists of reading
the transcript of an advertisement critical of John Roberts
and later learning about the ad’s retraction following
sharp bipartisan criticism of its factual merits.
The top panel presents results from all participants
and suggests that there was a moderate perseverance
effect. But the bottom panels reveal that aggregating over
parties masks substantial differences between partisans.
Among Democratic participants, the perseverance
effect was large. Among Republicans, it was quite weak
(control=11%, treatment post-discrediting=14%, p=.25).
All subjects were asked to report their attitude toward Roberts “as a Supreme Court
Justice” on a seven-point scale ranging from “disapprove strongly” to “neither approve nor
disapprove” to “approve strongly.” Control-group subjects were asked these questions after
reading the introductory text; treatment-group subjects were asked after reading the transcript of
the ad and after the discrediting. Subjects were also asked to report how sure they were of their
position on a five-point scale ranging from “not at all sure” to “extremely sure” and to report their
party identification on a three-category scale as in the previous experiment.
5.2
Results and Discussion
Figure 2 reveals a strong perseverance effect. At the end of the experiment, 33% of control-group
subjects disapproved of the candidate, while 42% of treated subjects did the same (p = .01).
Among Democrats, the effect was profound (control disapproval rate = 56%, treatment = 72%,
p
= .01). Among Republicans, it was weak (control = 11%, treatment=14%, p = .25). Even after
reading of the ad’s withdrawal and a sharp, factual, bipartisan evisceration of its insinuations,
Convention