Making Congressional Studies Dynamic: Comparing Across Time

Timothy P. Nokken
University of Houston

Ironically, much of what we as legislative scholars investigate is inherently static. Despite the existence of long time series of important data such as election returns, parties' seat shares, and roll call votes, for example, rarely do congressional scholars construct research designs that call explicitly for time series analyses. There are good reasons to explain why most legislative scholars are not time series experts. First, it is not always obvious how to properly address the time dependent elements for much of the congressional data, despite the long time series. Furthermore, many of the phenomena we study are characterized by the fact that they remain stable over time. In the remainder of this note, I explain the point of these seemingly anachronistic statements more fully. I then go on to explain how utilizing the historical record can help to overcome some of the inherent stability we confront in congressional data.

Devising explicit time series research designs to study congressional phenomenon is not an easy task, even with data sets spanning several years. For example, while we have a lot of data on congressional election returns, we do not necessarily have enough to utilize high-powered time series techniques. I would argue that too many factors have changed to make meaningful comparisons of election returns over wide swaths of history - redistricting, changes in candidates, rise and fall of different parties and party systems - make time series comparisons of aggregate, state, and district level election returns over long time spans somewhat difficult. 

Additionally, the most prevalent form of congressional data, the log of recorded roll call votes, is not "time series" ready for three important reasons. First, series for individual congressional seats are frequently interrupted by the replacement of sitting members. The average tenure in office for most MCs falls far short of the length needed for high-powered time series analyses on individual voting records. Second, it is not clear that there is an explicit, easily identified time element to roll call voting behavior. For example, roll call voting would not necessarily exhibit time dependent statistical characteristics like seasonality. Finally, the most basic issue is that the vast majority of members of Congress do not alter their voting behavior over the course of their careers. When we do subject the roll call voting of individual members to some sort of statistical analysis, we find very little variation in the dependent variable. Poole and Rosenthal (1997) succinctly characterize the situation by saying that "members die in their ideological boots" (8). Herein lies the rub. Much of what legislative scholars believe influences roll call voting behavior remains highly stable over time: institutional structures, party affiliation, constituencies (sans drastic redistricting), and individual ideology. The way to overcome the lack of "time series friendly" data is to turn to the historical record and to devise clever ways to exploit the copious amount of data we can access.

In my research, I try to determine how a variety of factors that tend to remain stable over time - e.g., party affiliation, constituency, institutional structures, individual ideology - influence individual level roll call voting behavior. While such factors tend toward stability over time, they are not completely devoid of variation. I exploit natural experiments to "dynamicize" the study of legislative behavior in the United States, searching for change over time. These natural experiments generally involve comparing an individual's behavior prior to and following some sort of event. If one observes a significant behavioral change, one can attribute that behavioral change to the change in the factor of interest, "treatment." One such natural experimental design meant to capture the affect of party affiliation on voting behavior analyzes the roll call voting behavior of those members of Congress who switched parties during their congressional career (Nokken, 2000). I find that those MCs who switched parties exhibited dramatically different patterns of voting behavior following their defection. The party defectors start to compile a voting record that matches the members of their new party. A control group consisting of members who compile similar voting records to the switchers, but who do not switch parties, shows no statistically meaningful changes in their voting behavior. While that particular piece is probably best not characterized as an historical approach since it covers a fifty year time span from 1947 to 1997, it illustrates how one can add a dynamic element to the study of Congress by assessing the importance of variation in a factor normally conceptualized as a constant. The expectations surrounding the behavioral consequences of party defection are not necessarily constant over time. Different party systems or multidimensional policy environments, for instance, may lead to different behavioral consequences in different congressional eras, and have led others to investigate party defection across broader periods of time (Oppenheimer, 2000; Nokken and Poole, 2001).

Historical data also allows one to determine how constituent preferences and electoral concerns influence an individual member's voting behavior. One such question asks whether a MC's pre- and post-election voting behavior remains consistent. In particular, I want to know whether individuals change their votes on specific pieces of legislation at different points in time. The most logical place to look for such changes is among lame duck members of Congress, those MCs who cast roll call votes but will not return to the next Congress due either to retirement or electoral defeat. Lame duck sessions of Congress - held after the November elections, but prior to the seating of new members - provide an extremely rich setting with which to study such effects. Analyses of roll call votes in lame duck sessions held in recent Congresses find some evidence that members no longer subjected to electoral constraints changed their votes on specific roll calls. McArthur and Marks (1988) found that lame ducks had an important and positive influence in passing the Domestic Content bill in 1982.¹ Nokken (2000) analyzes the series of votes pertaining to the General Agreement on Tariffs and Trade in the 103^rd Congress (1993-94), final passage of which was secured following the momentous 1994 elections, and finds some evidence that defeated Democratic House members were more likely to shift from the opposition to support final passage of GATT. 

In modern congresses, roll call votes held in lame duck sessions are rare events. The ratification of the Twentieth Amendment in 1932 changed the date new Congresses convened and thereby eliminated the formal lame duck session of Congress. Those lame duck sessions convened in the December following congressional elections, and met until new members were sworn in the following spring. Furthermore, the lame duck sessions addressed substantive legislative business that resulted in numerous recorded roll call votes. These lame duck sessions serve as a natural laboratory with which to study how the electoral connection affects roll call voting behavior, primarily participation rates and party cohesion. Previous work on lame duck effects essentially conclude members do not vote differently, they simply vote less often (Loomis, 1995; Poole and Rosenthal, 1997). 

In a recent paper, Craig Goodman and I compare participation rates and party cohesion scores in non-lame duck and lame duck sessions from 1870 to 1933 (Goodman and Nokken, 2001). We hypothesize that significant behavioral affects do result from lame duck status, but that the effects would take a subtler form than the aggregate effects Loomis (1995) sought to find. We compare lame duck and non-lame duck voting across important subsets of members, and find that certain groups of members exhibit significantly different voting behavior in lame duck sessions. Generally, we find party cohesion rates in lame duck session vary significantly across congresses, oftentimes exceeding the mean levels of cohesion in a number of non-lame duck sessions. In short, lame duck sessions do differ from "regular" sessions of Congress, but the differences are not uniform across time.

The questions we are most interested in answering in the lame duck research have to do with partisan-based incentive structures. More specifically, are party leaders able to extract higher levels of party cohesion from MCs in return for some sort of partisan-based benefit? We find weak evidence that supports a partisan benefits story. For example, lame duck sessions preceding changes in partisan control of the House are marked by statistically significant increases in party cohesion among returning members of the party poised to take majority control in the next session of Congress. We characterize such lame duck sessions as an audition of sorts for members to try to impress party leaders in an effort to win seats on desirable committees. Departing members also had strong incentives to maintain their partisan loyalty. Our results show that departing MCs who took patronage positions showed statistically significant increases in party cohesion in the lame duck session of their last term in office. The political parties still controlled the distribution of patronage positions during the time period we analyzed, even after the enactment of the Pendleton Act in 1883. Those lame duck members who maintained or even increased their support for party initiatives in lame duck sessions may certainly have done so to receive side payments, most notably a patronage appointment

The findings described here do not represent the final word on the significance of lame duck Congresses. Rather they are preliminary findings that suggest many avenues for future research addressing issue-specific consistency in roll call voting behavior (along the lines of pieces by Asher and Weisberg 1978; Nokken 2000); scheduling strategies of legislative party leadership, and on research on vote buying or side payments that induce party loyalty on important roll call votes. My use of historical data stems not from a simple interest in history per se, but from an interest in analyzing and evaluating important changes in the set of factors that influence roll call voting behavior. Limiting studies of roll call behavior only to modern congresses ignores a number of useful events with which to extend and clarify our findings from the analyses of modern congresses. While such research focuses on congresses past, it certainly provides a wealth of useful information to understand the dynamics of the contemporary congress.

Notes

1. The Domestic Content bill would have required a minimum percentage of labor and parts in order to be categorized as a domestic automobile or else to be subjected to import restrictions.

References

Asher, Herbert B. and Herbert F. Weisberg. 1978. "Voting Changes in Congress: Some Dynamic Perspectives on an Evolutionary Process." American Journal of Political Science 22:391-425.

Goodman, Craig and Timothy P. Nokken. 2001. "Lame Ducks and Roll Call Behavior in the U.S. House of Representatives, 1869-1933." Paper Presented at the Annual Meetings of the Midwest Political Science Association, Chicago, IL.

Loomis, Michael. (1995). An Analysis of Legislator's Utility Function for Voting in the United States Congress. Ph.D. Dissertation. Carnegie Mellon University.

McArthur, John and Stephen V. Marks. (1988). "Constituent Interest vs. Legislator Ideology: The Role of Political Opportunity Cost." Economic Inquiry 26(3): 461-470.

Nokken, Timothy P. (2000). "Roll Call Votes as Exercises in Position Taking: Congressional Reactions to Normal Trade Relation Status for China, 1999-2000." Paper Presented at the Annual Meetings of the Southern Political Association, Atlanta, GA.

Nokken, Timothy P. 2000. "Dynamics of Congressional Loyalty: Party Defection and Roll Call Behavior, 1947-1997." Legislative Studies Quarterly 25:417-444.

Nokken, Timothy P. and Keith T. Poole. 2001. "Congressional Party Defection in American History." Paper to be presented at the Annual Meetings of the American Political Science Association, San Francisco.

Oppenheimer, Bruce. 2000. "Testing Ideological and Partisan Theories in the Structuring of Congressional Decision Making: The Roll Call Behavior of Members Who Switch Parties." Paper presented at the Annual Meetings of the Midwest Political Science Association, Chicago.

Poole, Keith T. and Howard Rosenthal. Congress: A Political-Economic History of Roll Call Voting. New York: Oxford University Press.

Timothy Nokken is Assistant Professor of Political Science at the University of Houston. He received his Ph.D. from the University of Illinois at Urbana-Champaign in 1999. His work has been published in Legislative Studies Quarterly and the Journal of Theoretical Politics. His email address is tnokken@mail.uh.edu