Using Referenda to Examine the Impact of Voters on Legislative Partisanship.

Overview:One determinant of the parties' ability to resist the median voter may be constituency size. We test this proposition with a measure of legislator ideology that is directly comparable to voter ideology, using referenda votes as bridging observations.Abstract:At the state and local level throughout the United States, there is considerable variation in the power of parties to control nominations, structure legislative voting, and compel non-median behavior from elected officials. Some of this variation stems from the range of institutional arrangements that govern parties. For example, more stringent rules on primary participation may result in the nomination of more extreme candidates (Fiorina 1974; Gerber and Morton 1998). Long periods out of power may also cause parties to unify while nominating more centrist candidates (Bawn et al 2006). Lee and Oppenheimer (1999) suggest a role for constituency size in determining partisanship. They find that senators from larger states tended to be policy activists, while those from smaller states tended toward more centrist behavior.In this paper, we investigate the role of constituency size in inducing party strength. Masket's (2007) study on this phenomenon was hampered by an important methodological barrier; any study that tries to determine whether legislators are more or less partisan than constituents must make very strong assumptions about the ideological scales of those two populations. Even if we accurately measure preferences with legislative votes and with electoral returns, the scales from both measures are arbitrary, and not the same.This is a common problem in the literature, to which we offer a possible solution. We propose to bridge the two populations using referenda, on which the legislature and the electorate both vote. They thus serve as bridging observations between the two distributions (e.g. Bailey 2007). We explore several critical features of this strategy. First, legislators can only vote yea or nay, but electoral districts can pass or fail a measure by any margin. We develop strategies for using that additional information without compromising the goal of comparability. Second, any further application of this strategy will suffer, since referenda are scarce. So we attempt to quantify the additional leverage gained by adding more bridging observations, and the dangers of having too few. This paper would be appropriate for panels on methodology, political parties, state politics, representation and nominations. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]