March 14, 2006 (4:30pm) / 23
210
203
Number of observations
–331.32
–353.50
Log likelihood
.17
.13
.39
3.14
6
.18
–.53
.28
2.33
5
.19
–1.01
.22
1.51
4
.24
–2.04
.18
.29
3
.27
–2.45
.30
–.56
2
.40
–3.45
.22
–1.43
1
.35
–.53
*
.25
–.07
.25
–.60
*
Treatment
Difference
Republicans
Democrats
Table 2: Belief Perseverance in the John Roberts Experiment. Cell entries in
the first two columns are parameter estimates and standard errors from ordered
logistic regressions. The dependent variable is approval of the John Roberts as a
Supreme Court Justice, a seven-category variable ranging from “disapprove” to
“neither disapprove nor approve” to “approve.” The treatment, described in the text,
consists of reading the transcript of a television advertisement critical of John Roberts
and later learning about the ad’s retraction following sharp bipartisan criticism of its
factual merits. Substantively interesting coefficients statistically significant at 90% using
a one-sided test for H
A
>0 are denoted by
*
.
The results suggest substantial belief perseverance among Democrats: treated
Democrats disapproved of Roberts more than their control-group counterparts, even
though the treatment left them no informational basis for doing so. The difference
between the treatment effects for Democrats and Republicans is also substantial,
indicating that partisanship moderated the extent of perseverance.
treated subjects were less approving of Roberts than their control-group counterparts—but as in
the previous experiment, the effect was sharply moderated by partisanship.
And as in the previous experiment, the moderating effect of partisanship is affirmed by
ordered logistic regression models in which approval ratings are regressed on the treatment. The
difference between the Republican and Democratic treatment effects is significant (p = .06).
The magnitude of the difference is indicated by the bottom panels of Figure 2, in which the gap
between Democratic and Republican disapproval rates grows from 45% in the control condition
to 58% in the treatment condition.
Convention