18
characteristics examined so far suggest that the answer may differ for candidate issue
positions and for their backgrounds.
As discussed previously, the answers respondents give at T1 and T2 to the
campaign knowledge questions place them into four major groups – retaining knowledge,
learning over time, forgetting, and failing to learn. To predict membership in each of
these groups for each question, I employ multinomial logit models. Those who retain
knowledge serve as the baseline group. Tables 4 and 5 present the results. Interpreting
these results proves difficult without further manipulation. While we can identify which
variables achieve significance, the sign and the direction of the effect may differ and, as
with other, simpler, logit models, we cannot judge the magnitude of the effect solely by
looking at the size of the coefficient. A comparison of the significant variables in Tables
4 and 5 tells us that different processes appear at work for the issue position answers than
the candidate background ones. Only some of the personal characteristics achieve
significance. Candidate ads and news rarely reach significance.
Since this paper focuses primarily on the role of advertising on learning,
forgetting, and remembering, I confine the remaining analysis to the candidate
background questions. To aid in the interpretation of these results I employ CLARIFY
to produce changes in predicted probabilities for variables and quantities of interest. I
focus primarily on the role of advertising and of news, both when holding all other
variables at their means and in the presence and absence of motivation to consume
information as determined by interest and party identification. Table 6 demonstrates the
role of advertising through the changes in predicted probabilities when moving from
5
Michael Tomz, Jason Wittenberg, and Gary King. 2003. CLARIFY: Software for Interpreting and
Presenting Statistical Results. Version 2.1. Stanford University, University of Wisconsin, and Harvard
University. January 5. Available at http://gking.harvard.edu/
Convention