Theory: While the link between presidential approval and congressional election outcomes is long established, scholars have generally ignored the role of a member’s own voting record in mediating these effects.  If citizens truly use the congressional vote to express support or opposition toward the president, then they should not punish or reward all of his fellow partisans equally.  Instead, the degree of reward or punishment meted out by voters ought to depend on the member’s level of support for the president’s legislative initiatives.

Hypotheses: Citizen perceptions of a member’s level of presidential support should be substantially grounded in reality—that is, actual presidential support should be a large and significant determinant of perceived presidential support.  Citizens who dislike the president and perceive their member as supportive should rate the member lower on a feeling thermometer, be less likely to approve of the member’s job performance, and be less likely to vote for the member’s reelection.  These same patterns should hold for citizens who like the president and who perceive their member as generally in opposition.  These effects should outweigh simple partisan cues in explaining citizen attitudes toward congressional incumbents.

Methods: All data are drawn from the 1993, 1994, and 1996 National Election Studies.  Perceived presidential support is modeled as a function of actual support, member’s party, strength of partisanship, and projection effects.   Incumbent feeling thermometer ratings, job approval, and congressional vote choice are modeled as a function of the interaction between presidential approval and perceived member presidential support, along with several control variables.  All equations are estimated using OLS, probit, or ordered probit, as appropriate.

Results: Perceived member presidential support is strongly related to actual support, tracking it closely in all three years data among all groups of respondents.  Additionally, the interaction of citizens’ attitudes toward the president and their perceptions of how often their members supported his proposals powerfully influence feeling thermometer ratings, incumbent job approval, and congressional vote choice, dwarfing the effects of simple partisan heuristics.  Thus, citizens appear to be much more sophisticated than is typically assumed in using the congressional vote as a referendum on presidential policy.