Throughout the paper, I have used cosponsorship networks as a test of induced legislative collaboration through shared consituency. I believe cosponsorship to be the most appro- priate test for this hypothesis because cosponsorship requires an active decision by both a sponsor and a potential cosponsor. Decisions by a bill’s sponsors about whom to solicit as cosponsors sends a signal to the floor about the legislation being sponsored (Kessler and Krehbiel 1996). Decisions about what to cosponsor send certain kinds of signals about individual legislators to their colleagues (Kirkland 2011). Thus, both actors must wish to send a signal and must agree to send that signal collaboratively. Sponsors can reject cosponsors, and cosponsors can turn down requests from sponsors. Additionally, cosponsorship can occur on any bill sponsored in the chamber. Alternatively, roll call voting is much less a sort of dyadic behavior and is subject to selection mechanisms in ways that make it a less ideal indicator for my analysis. Legislators are expected to take vote on nearly all legislation that comes to roll call, meaning they can be less selective about the type of signal they wish to send with their votes. Additionally, there is much less concern from the sponsor of a bill about who votes for a bill. He or she simply wishes for a majority of legislators to vote in favor of his or her legislation. Crafting the appro- priate coalition is much less important at this stage of legislative deliberation. Finally, roll calls are subject to selection at the committee stage meaning patterns of co-voting at the roll call stage would represent coordinated behavior on only a small subset of the possible opportunities for collaboration.
Nevertheless, there is a lengthy tradition of roll call analysis in legislative studies and if my theory is correct I should be able to detect changes in roll call voting behavior after the change in the North Carolina legislature to single-member districts. Additionally, it is possible that legislators place little weight on cosponsorship making common cospon- sorship an unimportant political activity. Roll call votes are clearly an important veto
point in the chamber and carry a great deal of weight in determining policy outcomes. As such, I have gathered the roll call votes for legislators in the North Carolina House for the 2001-2002 session and the 2003-2004 session. Using Optimal Classification (Poole 2000, 2005), I fit a two-dimensional solution to the roll call votes by North Carolina House members for each session.4 Figure 7.4 plots the coordinates of legislators on the
first dimension for the 2001 and 2003 legislative sessions. Republicans are colored grey and Democrats are colored black. As expected, the first dimension of the solution has a strong partisan structure with very little overlap between the two parties.
As before, I create a vector of differences between actors on the first dimension of the solution. I can then compare the differences between legislators in the same district to dif- ferences between their colleagues from single-member districts. Because the distribution of differences is unknown, I again use the Wilcoxon Rank Test, which assesses whether there is a systematic difference in the ranking of differences between shared district and single-member district legislators. In 2001, the differences between members of the same district but different parties were ranked as statistically smaller than differences between members of different parties and different districts (p-value of 0.021). There was no mea- sureable difference between members of the same party and same district versus members of the same party but different districts. In 2003, the differences in first dimension pref- erences between same districts, cross partisans and different district, cross partisans is no longer statistically significant. Cross-partisans from formerly multi-member districts no longer have distinct behaviors from the cross-partisan single-member colleagues. There were still no measureable difference between members of the same party and same district versus members of the same party but different districts.
Thus, even using the less appropriate but perhaps more politically important roll
●●●● ●●●●●●●● ●●●●●●●●●●●● ●●●●●●●●● ●●●●●●●●● ●●●●●●●● ●●● ●●●●●●● ●● ●● ●●●● ●● ●●●●●●● ●●●●● ●●● ●●●●●● ●●●●● ●●●● ●●●●● ● ●● ●●●●●● ● ● ● ●●● ● ● −0.4 −0.2 0.0 0.2 0.4 0.6 0.8 First Dimension P osition
First Dimension Positions of Legislators in the North Carolina House in 2001
● ● Democrats Republicans ● ● ● ●● ● ●● ●●●●●● ●● ●●● ●● ●●●●●●●● ●●●●● ●●●●●● ●●●● ●●●●●●●●● ●●● ●● ● ●●●●●● ●●●●●●●●●● ●●●●●●●● ●●●● ●●●●●● ●●●●●●● ●●●●●●● ●●●● ●●●●● ●●●● ●● −0.8 −0.6 −0.4 −0.2 0.0 0.2 0.4 First Dimension P osition
First Dimension Positions of Legislators in the North Carolina House in 2003
● Republicans
Figure 7.4: Optimal Classification in One Dimension of Roll Call Voting by NC House Legislators in 2001 and 2003
call voting measures, multi-member districts created increased levels of similarity in vot- ing patterns amongst cross-partisans. These similarities vanished once multi-member districts were eliminated. Recall that these inferences are drawn on the same set of multi-member district legislators before and after their district change, and that their new districts are not particularly different from their prior districts. What has changed is the nature of their elections and the fact that they no longershare a district.