insignificant in the other three. The measures of change in the issue distances between both Carter
and Ford are significant in the June and October waves, but insignificant in April and August.
Changes in attitudes about Ford—both retrospections about his presidency and assessments of his
character—are significant in the last two waves while the changes perceptions of Carter’s character
are significant in the final three waves. I will turn to the substantive impact of these effects shortly,
but the lesson from these results is that persuasion matters. The changes in voter’s attitudes that occur
during the campaign influence their vote choice.
The fit of the model also improves. Adding the persuasion effects to the model allows it to
accurately predict an additional seven percent of the cases, from 70 percent to 77 percent. This is a
23 percent reduction in the error in the model—a sizable improvement in the ability to predict
individuals eventual vote choice. Clearly, the persuasion that occurs during the campaign is an
important part of voters’ choices.
We have now reached the standard conclusions about campaigns—persuasion, when it
occurs, matters. But this is neither new nor surprising. The central argument put forth here is that
the changes in attitude strength, more specifically changes in a voter’s uncertainty, alters the weights
they apply to the attitudes they hold. The next step is to add the effects of changes in attitude
strength to the persuasion model. To do this I add both the change in the level of attitude strength
and the interaction term between attitude strength and the various attitudes to each of the time
indicators in the model. Table Four presents the results of the model that add the uncertainty terms to
the persuasion model. The joint tests of significance test the null that the changes in uncertainty do
not moderate the effect of issue positions and candidate traits. These tests only account for the
moderating effect of uncertainty and not the direct effect of changes in uncertainty on the probability
that the respondent votes for Carter.
6
Several of these changes in uncertainty terms are significant, but these do not get at the central hypothesis
about the moderating role of attitude strength. Adding them to the test statistic would artificially increase the
statistical significance of the test.
19
Convention