3---voters distinguish between state and federal action---it controls voting patterns

An inspection of Table 4 shows modest support for the federalist voting hypothesis. Respondents with accessible attitudes on unemployment are more likely to link vote for (against) their incumbent senator when they attribute functional responsibility to the national government and credit (blame) government for policy outcomes in connection with this issue. Respondents with accessible attitudes on unemployment but who attributed functional responsibility to the state government showed a similar significant and positive relationship between causal responsibility attribution for unemployment and their vote choice for governor. This pattern did not extend to mayoral voting behavior.

Accessible causal responsibility attributions regarding education and traffic congestion linked to mayoral vote choice when functional responsibility for these issues coincided with the city. In contrast, accessible causal responsibility attributions for these issues did not affect senate or gubernatorial vote choice when functional responsibility matched those levels of government. Moreover, if accessibility for these attitudes is lowered, at a certain point all of these coefficients shrink and become statistically insignificant. To illustrate, Figure 1 displays the combined effect coefficients for unemployment (and their 95% confidence intervals) calculated at different levels of attitude accessibility in the gubernatorial vote choice model. Note that as causal responsibility attributions regarding unemployment become less accessible (i.e., the latency score increases), the combined effect shrinks and the confidence interval crosses zero. Consequently, the federalist voting hypothesis may only apply, if at all, to a subset of voters for whom a particular issue is salient and readily accessible.

The pattern of statistical significance for the coefficients in Table 4 is also intriguing. Unemployment only has an effect on national and state-level voting behavior, while education and traffic congestion only have an effect on city-level voting behavior (as long as functional responsibility is also attributed to those levels). In some sense, this pattern matches the actual assignment of functional responsibility for these issues. The national government has more to do with developing policies that affect unemployment, while local governments are more involved with setting education policy and handling traffic congestion. While only suggestive, these data lead to a hypothesis worthy of further study: Perceived functional responsibility attributions affect voting behavior to the extent they match the actual assignment of functional responsibilities.

The cause of this match between perceived and actual functional responsibilities is far from clear. It may reflect knowledge on the part of the respondent. After all, these results only work for respondents with highly accessible attitudes on the subject. Perhaps these individuals care a great deal about these issues and are motivated to pin credit or blame on the appropriate office. On the other hand, it may reflect the set of issues discussed in the campaign. There are at least two reasons why campaigns may be the main culprit for explaining this pattern. First, campaigns at different levels of government tend to emphasize the issues over which officials are functionally responsible (Atkeson & Partin, 2001; Tidmarch, Hyman, & Sorkin, 1984). Second, when campaigns focus on an issue, considerations about that issue, such as causal responsibility attributions, are more likely to be accessible (Zaller, 1992).

The findings with regard to unemployment may provide some evidence for the campaign-effects claim. In 2002, unemployment was an issue in both senatorial and gubernatorial campaigns. The relationship at the state level is likely further enhanced by the fact that all three of the incumbent governors in the state analyzed here were Republican, the same party as the sitting president. President Bush and his policies became an issue in each of these gubernatorial campaigns, which may be responsible for some degree of spillover from the national to the state level.

Conclusion

Citizens do appear capable of making distinctions in terms of what different levels of government do, and these distinctions structure where they attribute credit or blame for policy outcomes. However, these distinctions only affect voting behavior at different levels within the federal system under specific conditions. The data analyzed here suggest that the issue upon which voters are attributing responsibility must be highly accessible and the level of government perceived to be functionally responsible for that issue should coincide with the level of government that is actually responsible. As already discussed, future study should hone in on the explanation for this nuanced finding. Does it reflect campaignspecific cues, knowledge on the part of the voter, or both?