For the full results of the interactive regressions, see Table D.1 in Appendix D.

25

Following Benesh (2006, 704), I also controlled for a number of institutional factors such as level of

26

professionalism, dissent rates, and the rate of the exercise of judicial review (see her footnote 8, p. 700).

The addition of these variables to the interactive equation has no effect whatsoever, largely because, as in

Benesh’s analysis, these variables have no relationship with the dependent variable of this analysis.

I have re-run all of this analysis with a more restricted indicator of type of selection/retention system:

27

partisan and non-partisan elections versus all other systems. The results are extremely similar to those

reported in Table D.1 and discussed in the text above, with no modifications whatsoever required in the

substantive conclusions.

-18-

institutional legitimacy, and whether any of the sets of interaction terms achieve significance (see Cohen

et al. 2003).

The analysis reveals that only two of the interaction terms are statistically significant. Both have

to do with the policy talk variables. When these two interaction terms are added to the basic equation,

R

2

changes by 1.8 %, which is statistically significant at *p* < .05. The interaction terms are quite

25

interesting: In states with elected judges, the coefficient for the condition under which the candidate

expresses his policy views is a trivial .04. However, in states without elections, the coefficient is -.24,

indicating that policy talk *detracts* from legitimacy. Similarly, while the coefficient for the policy

promises condition is +.11 in states with elections, it is -.16 in states without elections.

26

From this analysis, an important amendment to the findings reported in Table 3 (above) emerges:

The general finding that policy statements by judicial candidates have no deleterious effects on

institutional legitimacy is confined to states employing elections to select and/or retain their judges. In

states with no such processes, citizens (presumably) have little direct experience with judicial

campaigning and they therefore find it objectionable when they are exposed to such activity in the

experiment.

27

Care must be taken, however, with this finding, for two reasons. First, the sample was, of course,

not designed to be representative of individual states, and, second, this interactive effect is *not* confirmed

on the issues of campaign contributions and attack ads. Perhaps this indicates that campaign