Objectives. Are legislators' party affiliations or is district partisanship the greatest predictor of legislative support of the president? Do members of the U.S. House emphasize different policy positions when casting roll calls than when communicating their positions to constituents? We theorize that party is less important in legislators' district-oriented behavior than in roll-call voting. When casting roll calls, legislators are agents facing multiple principals, namely, political party leaders and their district constituencies. When engaging in district-oriented behavior, the only key principal is the legislator's constituency.
Methods. We analyze legislators' positions on roll calls and in platforms. Platforms are examined with a unique data set of franked mass mailings sent by House members. Linear and limited dependent variable models are employed.
Results. Our findings show that constituency preferences are a more consistent predictor of legislative support for the president when analyzing legislators' platforms, and that political party has a relatively limited effect. When analyzing roll-call votes, party is the key predictor.
Conclusions. Political parties may be interested in what legislators do as opposed to what they say. The platform findings are in contrast to most recent empirical work examining position taking, though consistent with the canonical works of Mayhew and Fenno. This has implications for theories of parties in Congress that tie party behavior in the legislature to partisanship in the electorate.