The correlation between party identification and issue attitudes is known to increase with political sophistication (Delli Carpini & Keeter, 1996). But do these more knowledgeable citizens bring their party identity into line with their issue attitudes, or do they bring their issue attitudes into line with their party identity? Zaller (1992) argued that individuals with higher levels of political sophistication are able to avoid accepting information that runs counter to their party identity—but what happens when acceptance cannot be avoided or when acceptance occurs before one finds out where the parties stand? This is the central question in the debate over the relationship between attitudes and party identification. Building on the work of Taber and Lodge (2006) and Gaines, Kuklinski, Quirk, Peyton & Verkuilen (2007), I expected sophisticated partisans to be more adept at reconciling cognitive inconsistencies in favor of their party. Political sophistication was therefore interacted with disagreement to determine whether it moderated the relationship between disagreement and party identification.
The predicted probabilities in Figure 4.4 show that political sophistication did indeed have a substantial moderating effect on the relationship between disagreement and party identification strength even in the absence of cognitive load. Those at the lowest end of the political sophistication scale dropped from a 68% probability of continuing to report a strong party identity to a 17% probability of reporting a strong party identity when exposed to disagreement. In comparison, there was an increase from 92% to 96% probability of continuing to report strong party identification among the most sophisticated members of the sample. This extremely large effect among the least sophisticated partisans again suggests that partisan updating occurs predominantly among the most vulnerable members of the electorate. However, because the effect emerges without inducing cognitive load, proponents of the revisionist model of partisanship might argue in favor of an alternative explanation. Instead of defending their identity, highly sophisticated partisans may simply not be swayed by disagreement with their party over a single issue when there are many other issues on which they agree.
When subjects were placed under cognitive load, the enormous effect of disagreement on the least sophisticated appears to drop off. Although the coefficient remains negative, the effect of disagreement on the least knowledgeable is no longer statistically distinguishable from zero (see Appendix Tables 4.4 and 4.5). From this result, it might appear as if cognitive load undermined reception of the disagreement stimulus, but this possibility was ruled out in the manipulation check. Therefore, this pattern, although nonsignificant, seems to support the revisionist model of party identification
Figure 4.4 The figure illustrates predicted probabilities obtained from the ordered probit regressions reported in Appendix Tables 4.4 and 4.5. For the purpose of prediction, pretest party identification strength was set to 3. Therefore, results represent the probability of maintaining a strong party identity from pre-treatment to post-treatment. "High sophistication" means that predicted values are calculated with political sophistication set to 7; "low sophistication means" that predictions were calculated with political sophistication set to 0.
rather than the dual motivations theory. It appears that cognitive load may have inhibited the ability of the least sophisticated individuals to adjust their party identification in response to disagreement. It is noteworthy, however, that more sophisticated partisans appear to have moved in the opposite direction, becoming slightly more susceptible to identity change in response to disagreement. Although this slope shift does not reach statistical significance either (p = .19) (see Appendix Table 4.5), this pattern favors the dual motivations theory over the revisionist model (see Figure 4.1, Panels B and C).
The probability of maintaining a strong identity in response to the disagreement treatment is displayed for those highest and lowest in political sophistication in Figure 4.4.
In sum, although cognitive load may inhibit updating among the least sophisticated partisans, it may also increase partisan updating among the most sophisticated partisans. This suggests that cognitive load may at first reduce defensive abilities, but once cognitive resources drop below a certain threshold, partisans are no longer able to defend or update their party identity. This conclusion is, of course, merely speculative.