Hayden Galloway and Michael Barber, Political Science

I hypothesized that the party leadership in the U.S. House of Representatives converges their roll call voting behavior to the median of their parties after being selected as party leaders. This measured through ideological differences in congressional roll call voting behavior.

The median voter has classically been used to represent the cumulative preferences of a voting body. The preferences of each voter can be mapped on a policy space. Each voter in a policy space has an ideal point where utility is maximized: a preferred outcome. In a policy game where the preferences of the actors are known, candidates seeking to win elections should strategically position themselves at the median of their constituencies to win. This is the median voter theory originally intended for politicians seeking election. Does the median voter theory apply to party caucuses in the U.S. House of Representatives?

Republicans and Democrats in U.S House of Representatives vote to elect leaders to represent their party’s interests. Before elected to leadership, the voting behavior of party leaders is less likely to be observed by their fellow party members. This would leave House members seeking party leadership less incentive to vote congruently with their party before holding office. Therefore, party leaders may change their voting behavior after winning election to leadership. To define the distance party leaders hold from the middle of their parties, the ideology of each House member can be measured through their roll call voting behavior relative to their party’s median voter. This difference must be measured for leaders before and after election to calculate the effect holding office as a party leader has on voting behavior.

Scholars have measured the ideology of Congress through roll call votes and have detected differences between party leaders and the middle of their party. None of these models have considered the possible change in roll call behavior after becoming a leader while also accounting for change in the cumulative ideological movement of Congress.

I used DW-NOMINATE data to measure each house members’ ideology. DW-NOMINATE scores have a range of [-1, 1]. Each legislator’s voting record is a utility maximizing point in multidimensional space. The deviation of each House members’ ideal point from their peers’ ideal points is calculated in their DWNOMINATE score. I apply this data from the 95th Congress (1977) to the 112th Congress (2011).

Once elected to party leadership, we can determine the roll call vote differences of a party leader from their peers. I isolated the net change over time for House leadership roll call voting behavior. Therefore, I used a difference-indifferences model to calculate the difference in leadership roll call behavior relative to the party median: an accurate and consistent measure of the party middle. I calculated the difference between two points in time to test my hypothesis: the session of Congress immediately before elected to party leadership and the session immediately after. This tested the immediate effect that a leadership position has on roll call behavior.

The difference in DW-NOMINATE scores between party leadership and their party caucuses is recorded from one consecutive congressional session to the next. I took the absolute value of every difference, because the direction of the difference from the median is not the focus, but the magnitude of the difference. With these measurements, I calculated the net difference in DWNOMINATE scores, or difference in differences, between party leadership and the median of their party. This net change in differences was calculated for each party leader at select points in their congressional career: the session before and after being selected as a party leader. The cumulative net difference in differences is calculated for Democratic Leader, Republican Leader, Democratic Whip, and Republican Whip in the U.S. House of Representatives.

Rank and file House members do not deviate much from the median. It is evident that the maximum possible deviation is zero or near zero for most House members and it becomes a rarer event as the possible maximum deviation grows. The majority of the observations have a maximum deviation less that 0.1. Party leaders’ role call behavior converges to the median of their party prior to leadership. This is not an immediate change. The term before ascension to party leadership is not statistically different from a leader’s first term in leadership. This proves my theory false: election to party leadership is not the cause – movement to the center takes time. Members aspiring to party leadership could foresee the need to represent the center of their party to obtain a leadership position. Another possibility is that House members may generally converge to the party middle over time. We proved that this movement is rare for House members not aspiring to positions of party leadership. We found that most House members do not change their roll call behavior as do party leadership. This shows that House members that seek party leadership are conscious of the need to represent the median of their party, and are willing to modify their voting behavior to increase their probability of being elected. This change does not occur at the event of assentation to party leadership, yet is a process that occurs prior to election.

By isolating the movement of roll call behavior over time, we find that congressional party leadership responds to the middle preferences of their parties. This movement is characteristic of party leadership. This validates existing assumptions of representation in a highly scrutinized realm of politics. Further research of why this occurs is necessary to understand the process of becoming a party leader in the U.S. House of Representatives.