9
1.
8
If, however, none of the overall increases occur among those with prior knowledge, then we can conclude
that the treatments fail to cause priming (or the other effects in the first column), with a few caveats that I
discuss below. Finally, if the increases instead occur primarily among those who learned the parties’
positions, then we can attribute them to issue-driven or vote-driven campaign effects, as presented in the
second column of Table 1.
This approach is not ideal. At best, it may allow us to conclude that the increase in issue-vote
consistency arises either because of campaign effects that require prior knowledge or because of campaign
effects that result from learning. It can differentiate between the columns in Table 1, but not between the
rows. Given the nature of the data, however, this simple approach may allow for strong conclusions while
avoiding the problems with experimental or IV approaches.
I apply this research to cases where previous researchers have reported that messages on an issue
increased dramatically between waves of a panel and produced increases in issue-vote consistency. I
searched the literature for all such cases and selected those in which the survey asks about perceptions of the
parties’ positions on that issue in both waves. Unfortunately, a number of the priming studies that examine
position issues with panel data lack such questions (Berelson et al. 1954; Krosnick and Kinder 1990;
Mendelberg 2001). Cases that do meet this requirement include European integration in the 1997 British
election, Social Security policy in the 2000 U.S. election, and defense spending in the 1980 U.S. election.
Further search revealed an additional case where campaign attention to an issue increased and priming
appears to occur, at least according to the conventional test: Public Works projects in the 1976 U.S.
presidential election.
Before proceeding, a few methodological comments are in order. As dependent variables (DVs), I
use vote intent or vote choice when available because of its greater political importance compared to
alternatives such as feeling thermometers. The cost of this choice, however, is fewer cases. I have replicated
the results below with more continuous DVs and found very similar results. As with previous panel studies
on priming, I compare the strength of the relationship between vote choice and issue attitudes only among
those individuals who responded to these questions in both waves of the panels. Consequently, I do not test
whether campaign or media messages on an issue cause citizens who were previously undecided to vote for
the party that shares their position on the issue. As in all analyses with public opinion measures as
explanatory variables, measurement error in the issue attitudes and in measures of prior knowledge and
8
Increases among those with prior knowledge could arise for reasons other than salience. For
instance, by conveying information that fosters greater certainty about individuals’ perceptions of the parties’
positions, the treatments may cause individuals to change their vote choice to match their issue position. As
people become more confident about their correct perceptions, they may become more likely to vote for the
party that shares their position. Alvarez (1997) presents evidence for such effects in U.S. presidential
elections.
Convention