The Case of the Vanishing Moderates: Party Polarization in the Modern Congress *

Transcription

1 The Case of the Vanishing Moderates: Party Polarization in the Modern Congress * Sean M. Theriault Assistant Professor Department of Government The University of Texas at Austin 1 University Station A1800 Austin, TX seant@mail.utexas.edu September 23, 2003 Abstract The two political parties in Congress are as ideologically divergent as they have been at any point in the last three decades. Scholars, analyzing the factors that have propelled this polarization, have come to different conclusions. Several recent studies finds that fundamental shifts in party strategies inside Congress has polarized the parties (Collie and Mason 2000 and Roberts and Smith 2003), while others claims that the shift in the electorate has driven party divergence (Fleisher and Bond 2003 and Stonecash et al. 2003). This article takes a necessary step back to analyze the micro-level changes leading to these enormous macro trends. By analyzing the individual ideological changes, I find evidence for both the institutional and electoral explanations for party polarization. A more complete picture of how Congress has polarized helps explain why Congress has polarized. * The author thanks Scott Adler, R. Lawrence Butler, Roger Davidson, Jen Lawless, David Lewis, Keith Poole, Eric Schickler, John Sides, and Barbara Sinclair for their helpful comments. Also, an earlier presentation of this article in front of a rigorous and helpful Texas A&M audience, especially Jon Bond, immeasurably improved this article.

2 Theriault 2 The Case of the Vanishing Moderates: Party Polarization in the Modern Congress The two political parties in Congress are as ideologically divergent as they have been at any point in the last three decades. Scholars, analyzing the factors that have propelled this polarization, have come to different conclusions. Several recent studies finds that fundamental shifts in party strategies inside Congress has polarized the parties (Collie and Mason 2000 and Roberts and Smith 2003), while others claims that the shift in the electorate has driven party divergence (Fleisher and Bond 2003 and Stonecash et al. 2003). This article takes a necessary step back to analyze the micro-level changes leading to these enormous macro trends. By analyzing the individual ideological changes, I find evidence for both the institutional and electoral explanations for party polarization. A more complete picture of how Congress has polarized helps explain why Congress has polarized. After surveying the positions of the major political parties in 1968, George Wallace, a thirdparty presidential candidate, famously remarked, There ain t a dime s worth of difference between them. In some ways, his quote reflected cutting edge work in political science. Anthony Downs (1957) predicted that as these electoral coalitions raced toward the political middle in hopes of capturing the pivotal median voter, political parties would converge. The ideological positioning of the congressional parties at the end of the twentieth century could not have looked more different than the parties Wallace observed. In fact, Bill Clinton assumed the divergence of the congressional parties when he developed his triangulation strategy. By rising above the infighting of Democrats and Republicans in Congress, Clinton hoped that he would appear to be above politics in promising to stake out a third way. The congressional parties were so polarized in the 106 th Congress ( ) that only two Republicans in the House were more liberal than the most conservative Democrat. 1 In the Senate, the most conservative Democrat, John Breaux, was more liberal than the most liberal Republican, Jim 1 The most conservative Democrat, Virgil Goode, switched parties in the 107 th Congress.

3 The Case of the Vanishing Moderates 3 Jeffords. 2 As the parties polarized in Congress, the political middle vanished. In 1968, when Wallace made his acute observation, over half (235) of all members were in the middle. 3 In 1998, less than one-fifth (84) of all the members were in the middle third of the ideological continuum. During the same time period, the number of moderates in the Senate fell by more than 50 percent from 48 to 23. Polarization, in addition to providing fertile research ground for political scientists, has had important consequences on lawmaking in Congress. Moderate legislators have a privileged position in deciding the final disposition of legislative proposals (see, for example, Krehbiel 1998; Brady and Volden 1998). Long-standing congressional rules like the Senate s filibuster and the constitutional requirements for overriding presidential vetoes force super-majoritarian outcomes. If moderates become sparser, more ideologically extreme members will fill these pivotal legislative roles and, some have argued, leave the vast middle in the electorate under-represented (Poole and Rosenthal 1984). 4 Additionally, congressional polarization also fuels the negative view that most Americans have of Congress. Polarization and its evil twin, partisanship, have been blamed for a decrease in congressional comity (Jamieson and Falk 2000 and Uslaner 1993) and an increase in legislative gamepersonship (Aldrich and Rohde 2000 and Sinclair 2000a, 2000b). The public frequently points to these activities in explaining why it loathes Congress (Hibbing and Theiss-Morse 1995 and 2002). Because of these consequences, congressional scholars have begun to explain the growing party divergence in Congress. Their findings can be broken down into two broad categories. The first explanation rests upon a fundamental change in the internal dynamics of Congress. At least three changes in congressional decision-making over the last 30 years could have caused party divergence. First, the House rules have become so restrictive that it is difficult for moderates to demonstrate their 2 Jeffords became an independent in the beginning of the 107 th Congress. Since this switch, he has voted much more liberally. 3 Throughout this article, I analyze Poole-Rosenthal (1997) DW-NOMINATE scores. Roughly, they exist on a 1 (liberal) to +1 (conservative) scale. Scores in the ideological middle are from 0.3 to I do not mean to suggest from this discussion that legislation is necessarily more difficult to pass. The results from the studies examining the causes of legislative outputs vary (see, for example, Mayhew 1991 and Howell, Adler, Cameron, and Riemann 2000). The point I am making simply suggests that the polarization has a profound effect on how legislation is produced, not necessarily on the amount of legislation produced.

4 Theriault 4 moderate ideologies (Smith 1989 and Sinclair 2000b). Second, the political issues in Congress today are much more ideological than the issues decided during the moderate heyday of the 1960s and 1970s. Third, the party leaderships of both Democrats and Republicans in both chambers have become much more explicitly ideological than they were before (Collie and Mason 2000, Evans and Oleszek 1997, and Roberts and Smith 2003). The second explanation is electorally based. As members districts have changed, so have their voting behaviors. Stonecash et al. (2003, 18) maintains, The emergence of party polarization is because the electoral base of each party has evolved from being fairly diverse to being more uniform. Jacobson (2001), Fleisher and Bond (2003), and Carson, Crespin, Finnocchiaro, and Rohde (2003) argue that this gradual transformation is jolted every 10 years by redistricting. Whether by gradual demographic change or by design, the argument goes, more homogenous constituencies elect more ideologically charged members. These explanations are not necessarily mutually exclusive. Advocates of both explanations borrow from the other in developing their argument. Rohde (1991), one of the earliest observers of this polarizing trend shows how the homogenization of a respective party s districts have emboldened its party leaders to pursue explicitly ideological strategies. Nonetheless, the roots of and the primary basis for the party polarization in Congress remain in doubt. The scholars who study polarization typically have offered their macro descriptions of party divergence before launching into an analysis of their causes. Almost uniformly, these studies skip an integral step analyzing the micro-level foundations of polarization. This article analyzes individual member voting to provide a more solid base for evaluating the explanations of polarization. If member replacement caused party polarization, the underlying culprit is more likely to be electorally based changes whereas polarization caused by the gradual conversion of long-serving members is more likely to be institutionally based. Understanding how Congress has polarized will help explain why Congress has polarized.

5 The Case of the Vanishing Moderates 5 This paper provides three contributions to this literature. First, it brings the Senate squarely into the analysis. Polarization in the Senate almost perfectly reflects House polarization; nonetheless, with few exceptions, polarization studies have focused exclusively on the House. 5 Given the similarities between the chambers ideological migration, polarization explanations cannot be exclusive to one chamber. Second, the micro-level analysis provides evidence for evaluating the electoral and institutional explanations for the parties divergence. At the end of the day, the analysis in this paper provides evidence for both explanations. Third, this article evaluates polarization at a broader context. The parties are 25 percent more polarized now than they were in the 92 nd Congress (1971-2). While there is still room for more divergence, it is unlikely that the future political parties will pull apart as quickly as the current parties have. With apologies to David Mayhew (1974), Morris Fiorina (1977), and the others who have written about the Case of the Vanishing Marginals, this article investigates The Case of the Vanishing Moderates. Section I briefly describes the polarization of the congressional parties. Section II outlines the two underlying mechanisms leading to the polarizing of the U.S. Congress. Section III describes the congress-to-congress changes that undergird the polarization and ties these results to the institutional and electoral explanations. Finally, Section IV speculates about the future of party polarization in Congress. I. The Polarized Congress The polarization between the legislative parties is, perhaps, one of the most obvious and recognizable trends in Congress during the last twenty years. Because this polarization has been described effectively by a number of scholars (see, for example, Aldrich 1995, Coleman 1987, Collie and Mason 2000, Fiorina 1999, Fleisher and Bond 2000, 2003, Jacobson 2000, Roberts and Smith 5 Fleisher and Bond (2003) is one good exception.

6 Theriault , Rohde 1991, Sinclair 2000a, and Stonecash et al. 2003), this section provides only the bare minimum to motivate the casual reader and to remind the forgetful reader. The parties ideological positions in 1968 contained considerable overlap (see panel A of figures 1 and 2). Giving credence to Wallace s statement, the parties were ideologically closer to one another in the late 1960s and early 1970s than they would be at any point thereafter. The change from Wallace s statement (1968) to Clinton s presidency ( ) was cataclysmic. The Democratic party became considerably more liberal while the Republican party became much more conservative. The parties in Congress are now almost completely distinct (see panel B in figures 1 and 2). 6 Insert figure 1 about here. Insert figure 2 about here. Figures 3 (the House) and 4 (the Senate) depict party polarization from a different perspective. The light (dark) grey area represents the spread of the Democratic (Republican) party one standard deviation above and below the mean. The lines running through the grey areas are the respective parties means. 7 Both figures show essentially the same pattern: a drastic separation of the parties from the 1970s to the 1990s. Although the polarization within any one Congress is subject to a whole set of idiosyncrasies, the overall trend is clear Democrats in Congress are becoming more liberal and Republicans are becoming more conservative. Insert figure 3 about here. Insert figure 4 about here. This diverging trend is not a function of the Poole-Rosenthal DW-NOMINATE scores. Different scholars using different methods show the same basic pattern with different roll call summary scores: party votes (Coleman 1997 and Stonecash et al. 2003), party unity scores (Coleman 6 In analyzing the same scores in different congresses, Jacobson (2000, figure 2-1) shows the same trend. 7 Jacobson (2000, figure 2-2) shows that for this time period, the mean and the median are almost identical. Unlike the median, however, the mean s second moment can be calculated.

7 The Case of the Vanishing Moderates , Rohde 1991, and Stonecash et al. 2003), Americans for Democratic Action (ADA) scores (Stonecash et al. 2003), American Conservative Union (ACU) scores (Collie and Mason 2000), and a mixture of ADA and ACU scores (Fleisher and Bond 2000). Shipan and Lowry (2001) even show how the parties have diverged in a particular policy area. The overwhelming evidence suggests that parties throughout the last thirty years have radically diverged. II. The Underlying Dynamics Leading to Polarization The polarization studies, for the most part, ignore the individual member changes in demonstrating the macro-trend of the two parties ideological separation. Prior to the polarization studies, however, political scientists studied these individual member changes to explain the timing of legislative outcomes. Scholars asked, for example, why did the liberals prized open housing fail throughout the 1960s only to pass during the least liberal congress of the decade? Was it member conversion or member replacement? As the parties separated, the macro polarization research swamped the micro-level member congress-to-congress changes. These individual changes, however, can be used to evaluate the explanations for the parties diverge in Congress. The micro-level mechanisms of polarization are introduced through two vignettes. In 1976, Gale McGee, Chairman of the Post Office and Civil Service, sought a fourth term for his Wyoming Senate seat. McGee was legendary in using his committee power to send money back to Wyoming. His opponent, Malcolm Wallop, turned McGee s position into a liability when he asked voters, How efficient do you think the postal service is? With that question as the cornerstone of his campaign, Wallop, a pro-environment, pro-choice Republican defeated McGee. In his first congress, Wallop had an ideological score of 0.35, placing him in the middle third of all Republicans. As his career progressed, however, his very conservative military and foreign policy views prevailed, moving him to the extreme right wing of the Republican party. By the time he

8 Theriault 8 retired, Wallop had a 0.79 ideology score making him the second most conservative member of the Senate. Next door to Wyoming and two years after Wallop s retirement, in 1994, Helen Chenoweth challenged incumbent Larry LaRocco for his western Idaho seat in the House. In two terms, LaRocco had built a fairly moderate voting record, scoring a 0.25 in his second term (the House Democratic average was 0.35). In a highly contested Republican primary, Chenoweth, who demonstrated her conservative credentials by holding an endangered salmon bake fundraiser, came out on top. She went on to defeat LaRocco in the general election by painting him as a liberal lieutenant in Clinton s War on the West. In her first Congress, Chenoweth, with a 0.84 ideological score, was the fifth most conservative representative in the House. The difference between LaRocco and Chenoweth produced the fourth biggest ideological change caused by the historic 1994 elections. Fundamentally, member conversions like Wallop s and / or member replacements like LaRocco-to-Chenoweth s have forced the political parties in Congress apart. No matter the particular underlying reason or reasons, it is the numerous conversions and replacements over the years that decimated the ideological middle in Congress. Member Conversion If Downs (1957) is correct, political parties will converge to the middle in hopes of adding the pivotal median voter to their electoral coalition. Even if two general election candidates start out with relatively extreme ideologies (perhaps, because of a tough primary), they know that the median voter in the electorate is the one who decides the contest. In an attempt to appeal to this median voter, they will moderate their positions. Even though the data on candidate convergence has been less than monolithic (see, for example, Ansolabehere, Snyder, and Stewart 1998), the logic is quite compelling. Knowing that they are most vulnerable in the beginning of their careers (Jacobson 2001), newly elected incumbents will try to maintain a moderate position. Though the threat of a contested

9 The Case of the Vanishing Moderates 9 primary may lure the member away from the middle, first-term members know that defeat in the general election is much more likely. Realizing that a few votes against the party position is a fair trade-off for electoral success, parties are careful not to pressure new members to vote in a way that may damage their future electoral prospects (Fleisher and Bond 2000). As members begin to transition out of their expansionist stage and into their protectionist stage of their careers (Fenno 1977), the electoral pressure to remain moderate decreases and the party pressure to cast a reliable party vote increases. The trust that the member has gained from her constituents (Bianco 1994) and the reciprocal comfort level that both she and her constituents have for one another permit her to exercise a bit more discretion in voting on the floor (Fenno 1977). Asher and Weisberg (1978, 393-4), Brady and Sinclair (1984), Burstein (1978, 1980), and Jones (1974) find small, but pervasive, member conversions. Member Replacement The LaRocco to Chenoweth replacement is one of three categories that comprise the universe of member replacements. First, as in the case with LaRocco, the member can be defeated by the opposite party in the general election (incumbent defeat). Second, member retirement (or defeated in the primary) can cause a switch in the party representing a particular district (cross-party replacement). Third, the member can be replaced by someone from her own party (same-party replacement). An analysis of these categories provides evidence for the polarization explanations. It is the first of these categories that has attracted the bulk of scholarly attention. Writing even before the Watergate Babies (1974 liberal Democrats), Newt s Lieutenants (1994 conservative Republicans), and the Bush Babies (2002 conservative Republicans), Burnham (1970) followed by a long list of others including Brady and Lynn (1973), Ginsberg (1973, 1976), and Brady (1978, 1991) attributed sudden swings in congressional policymaking to massive election turnover. Member replacement need not only happen in volatile elections. According to Clausen (1973) and Kingdon (1989), the changes brought about by member replacement can be more gradual.

10 Theriault 10 A slightly more conservative Republican may replace a more moderate Republican or a more reliable Democrat can replace a maverick Republican. In other words, not all replacements have happened as a result of wholesale electoral shifts like 1974 or Obviously, member conversion and replacement can occur simultaneously. Some members may be becoming more ideological at the same time as more extreme members replace more moderate members. Indeed, simultaneous member conversion and member replacement is perhaps the most compelling and pervasive finding of those studying congress-to-congress changes (Asher and Weissberg 1978; Sinclair 1977, 1982; and Brady and Sinclair 1984). III. Member-level Ideological Congress-to-Congress Changes The unit of analysis throughout this paper are the micro-level ideological changes from one congress to the next. The simplicity of the previous sentence belies the difficulty in making these comparisons. Deaths, resignations, and party switching complicate the analysis. 8 To simplify and to systematize the process, the members scores elected during a particular general election are compared to the members scores elected during the previous general election; in other words, special elections are ignored. 9 In the House, reapportionment and redistricting exacerbate the difficulty of matching members across congresses. Frequently political scientists throw away the elections that happen immediately after reapportionment and redistricting. I develop an algorithm to rescue the 8 Like Poole and Rosenthal, I consider a member who switches partisan identification mid-congressional career to be a completely new member. 9 For example, in October 1995, Mel Reynolds, who had admitted to a sexual relationship with an underage campaign worker, resigned his seat in the House. Three months later, Jesse Jackson Jr. won the right to complete Reynolds s term. In November 1996, Jackson won reelection to the 105 th Congress. The comparison for Jackson s score in the 105 th Congress is Reynolds s score from the 104 th Congress, even though the transition happened midway through the 104 th Congress. All other intra-congress changes are handled in the same manner.

11 The Case of the Vanishing Moderates 11 preponderance of the data from these congresses. 10 In total, I analyze 6385 congress-to-congress ideologies in the House and 1485 in the Senate. 11 In 1997, Poole and Rosenthal presented a procedure to compute member ideological scores that could be compared across congresses. These procedures take into consideration all nonconsensual votes across all of the congresses. 12 They do, however, warn that the scores from one congress should not be compared with a congress from a different party period. To insure that the findings are not dependent upon the particular scores used, I supplement the House first dimension results with analysis based on both dimensions of the DW-NOMINATE scores, 13 Americans for Democratic Action (ADA) scores, and Turbo-ADA scores. 14 Groseclose used ADA scores to 10 First, when incumbents ran for reelection in a 2 year, they were matched up with their 0 year districts except if they lost in an incumbent-versus-incumbent match-up, in which the 2 year comparison is to the 0 year incumbent who won. Second, if after matching all the incumbent reelection efforts, only one district in a particular state from the 2 year and the 0 year remained unmatched, I paired them even if the unmatched 2 year was in a different part of the state than the 0 year. Although this step does not represent a true conversion or replacement for the constituents, it does confine these ideological swings to a particular state. Given that redistricting happens at the state level, it seems that capturing some of these data in this manner is preferable to deleting them. Fourth, if more than one district from both the 0 year and 2 year remained, I matched up the districts by comparing the old lines to the new lines. When no obvious match-up existed, which was rare, I deleted the districts from the analysis. As a last resort, I deleted new seats granted to states with growing populations and old seats taken from states with declining populations. As a consequence of reapportionment and redistricting, relatively few districts went unmatched (12 in 1972, 17 in 1982, and 22 in 1992) and almost all of them included states that either lost or gained seats in reapportionment. Although the states have marginally shifted their districts lines mid-census, California in 1974, was the only state to adopt radically different lines. Tracing the antecedent districts in these mid-census redistricting states is easier because the total number of the states seats does not change. 11 Over the 26 congresses, I deleted 284 House cases (1.02 percent): 51 due to redistricting and 89 because an ideological score could not be computed for one of the members in the comparison. I deleted 15 Senate cases (1.00 percent): 5 because an ideological score for one of the members could not be computed and 10 because they involved third party members. 12 From Keith Poole s webpage, accessed on December 9, Poole and Rosenthal restrict an individual member s score to change linearly over the course of her career. They maintain that higher polynomials in time did not appreciably increase the fit of the ideological score compositions. Undeniably, cross-congress comparisons can be subject to stringent, and perhaps appropriate, criticisms. Differences in membership, differences in party leadership, differences in legislative agenda, and differences in legislative procedures can all cast doubt on the reliability of cross-congress comparisons. Even with their faults, these scores provide the most effective avenue for analyzing across-congress ideological change. 13 With the collapse of the second dimension in the later half of the post-world War II era, it could be that party separation that once existed on two dimensions has collapsed to show more separation on the first dimension (Poole and Rosenthal 1997). Keith Poole and Eric Schickler were very helpful in bringing this potential problem to my attention and helping me resolve it. Lowry and Shipan (2002) also suggest that this might make Poole-Rosenthal scores inadequate for studying this problem. 14 ADA and turbo-ada scores were downloaded from Tim Groseclose s webpage, accessed on April 10, 2003.

12 Theriault 12 compute Turbo-ADA scores so that they, like DW-NOMINATE, could be directly comparable across congresses. To flesh out the construction and interpretation of the polarization scores, consider just the six members from Kentucky in the 104 th Congress (presented in table 1). In the 1994 elections, Kentucky Districts 3, 4, and 5, reelected their incumbents. Jim Bunning s and Scotty Baesler s ideological scores from the 104 th Congress were the same as their ideology scores from the 103 rd Congress, so the polarization score brought about by the 1994 elections, was 0.00 in these two districts. Harold Rogers, however, became more conservative in the 104 th Congress than he was in the 103 rd Congress. Because he was a Republican and a more conservative ideological score polarizes the parties, his polarization score for the 104 th Congress is Each of the comparisons from these three districts is classified as member conversions. Insert table 1 about here. District 3 elected Democrat Mike Ward to the open seat caused by the retirement of Democrat Romano Mazzoli. Because a Democrat replaced a Democrat, the transition is categorized as same-party replacement. Ward s ideological score was more liberal than Mazzoli, so the polarization score for Ward in the 104 th Congress was In district 2, Ron Lewis was elected to the seat that William Natcher held at the beginning of the 103 rd Congress. Because Lewis, a Republican, replaced a Democrat, the transition is coded as cross-party replacement. In district 1, Edward Whitfield defeated incumbent Tom Barlow in the general election. This transition is coded as incumbent defeat. The computation of the polarization scores in these two districts is more complicated because the seats went from one party to the other. As such, you cannot simply subtract the 103 rd score from the 104 th. The members score from both the 103 rd and the 104 th Congress is compared to their party s mean. For example, Barlow s 0.30 is 0.05 more conservative than the mean Democrat s score of Whitfield s 0.37 is 0.10 more liberal than the mean Republican s score In transitioning from Barlow to Whitfield, District 1 s polarization score is 0.05 because

13 The Case of the Vanishing Moderates 13 the new member was less polarized than the old member. In District 2, Natcher s ideological score was 0.01 more conservative i.e., less polarized than the mean Democrat s score and Lewis was 0.04 more conservative i.e., more polarized than the mean Republican s score. So, the polarization score for District 2 is In aggregate, the polarization score for the six Kentucky districts in the 104 th Congress is By summing the polarization scores from the six congressional districts in Kentucky with the polarization scores from all the other districts throughout the United States, we can ascertain the polarization score for the 104 th Congress (14.74). Table 1 shows the polarization process over the last 30 years for both the Senate and the House. According to the Senate scores in column A, only two congresses were more moderate than their predecessors (1970 and 1978). The polarizing scores ring true with conventional wisdom. The three most polarizing elections in the last 30 years were 1994 (the Republican Revolution), 1996 (Clinton s second election), and 1980 (Reagan s first election). Insert table 1 about here. What exactly do the individual congresses polarization scores mean? Like the composition of the Kentucky polarization score, the 1996 polarization score of 2.68 is the sum of all the polarization scores from all 100 seats in the Senate. The scores are on the DW-NOMINATE scale. If every senator went from being completely moderate (0.00) to completely ideological (-1.00 or 1.00), the polarization score for that particular congress would be The 2.68 score from the 105 th Congress implies that each seat became, on average, more polarized (i.e., each Republican senator became more conservative and each Democratic senator became more liberal). On average, a Democrat s scores would increase about the same as Paul Wellstone, who went from 0.77 to 0.80 and a Republican would increase, on average, about the same as Bob Smith, who went from 0.84 to In reality, of course, some scores stayed exactly the same (such as Spenser Abraham s at 0.35 and Patty Murray s at 0.48) while other scores experienced more radical changes such as the 0.25 polarization score when conservative Republican Gordon Smith (0.27) replaced

14 Theriault 14 moderate Republican Mark Hatfield (0.02) and the even larger change resulting from liberal Democrat Ron Wyden s (-0.43) replacement of moderate Republican Bob Packwood (0.06). 15 The individual polarization scores resulting from the 1996 election are not insignificant. Other congresses, however, experienced a fraction of the polarization emanating from the 1996 election. As a result, the average ideological congress-to-congress changes are much more subtle. Each senator became approximately 0.01 more polarized each congress over the 15 congresses. This small change is barely detectable, but when it is aggregated 100 times over 15 congresses, the unquestionable trend present in figure 4 takes shape. Column B of table 2 shows the first dimension DW-NOMINATE scores for the House of Representatives. Only twice over the 15 congresses did the House moderate (the 1970 and 1974 elections). Unsurprisingly, the 1994 and 1984 elections were the most polarizing. Over the 15 congresses, each House seat, like each Senate seat, became 0.01 more polarized with each passing congress. Again, while this change is miniscule, the trend becomes clear (figure 3) when it is aggregated over 435 House seats in 15 congresses. Columns C, D, and E show the polarization scores using different roll call summary scores. Column C takes into consideration both the first and second dimensions of the DW-NOMINATE scores. 16 Column D presents the ADA scores and column E presents the Turbo-ADA scores. The overall results are not dependent upon either the votes used (ADA versus NOMINATE) or the procedure used to estimate them (first dimension versus both dimensions and Turbo-ADA versus 15 The polarization resulting from the Packwood for Wyden swap is more difficult to determine because the respective senators came from different parties. If Packwood s NOMINATE score were the Republican average and Wyden s NOMINATE score the Democratic average, the swap would not have polarized the Senate at all. But, because Packwood s score is 0.31 more moderate than the average Republican and Wyden s score is 0.07 more liberal than the average Democrat, the swap results in a 0.38 polarization score. All changes that result in a party switch whether from retirement or electoral defeat are similarly calculated. 16 To compute the polarization score for both dimensions, I computed the member s two-dimensional distance from the origin (0,0) and subtracted the previous two-dimensional polarization score from it. To reflect the disproportionate importance of the first dimension scores (Poole and Rosenthal 1997), the first dimension is weighted double compared to the second dimension scores. This weighting had only a marginal impact on the individual scores and no substantive impact on the results.

15 The Case of the Vanishing Moderates 15 ADA). The correlations between the polarization scores are high 17 ; the correlations of the disaggregated individual change scores are even higher and more statistically significant. 18 Due to the similarity between columns B-E in Table 2, I present only the data for the first dimension DW- NOMINATE scores throughout the remainder of the article. Conversion versus Replacement Table 3 shows the breakdown in conversion and replacement for both the House and the Senate. Member conversion accounts for 49.7 percent of the polarization in the Senate and 32.6 percent of the polarization in the House. 19 These percentages are in the middle of the previous estimates. Fleisher and Bond (2003) find that around 90 percent of the moderates in Congress disappear as a consequence of replacement. Only 10 percent leave the moderate category by voting differently (i.e., member conversion). Roberts and Smith (2003), on the other hand, maintain that member conversion accounts for between 50 and 82 percent of the polarization of the parties during their respective polarizing phases (the 98 th to 100 th Congresses for Democrats and the 102 nd to 104 th Congresses for Republicans). Insert table 3 about here. The actual individual member conversions are again miniscule in fact, they are a fraction of the average 0.01 micro-level changes. Whereas conversion is responsible for around one-third of the total polarization, it accounts for 84 percent of the congress-to-congress pairings in the House and 87 percent in the Senate. Nonetheless, the successive tiny steps, however small, that members have taken away from the ideological middle have had a profound effect on the parties divergence. 17 The first dimension DW-NOMINATE scores are correlated with both dimensions at 0.44 (p=0.10), with ADA scores at 0.29 (p=0.32), and with Turbo-ADA scores at 0.54 (p=0.05). 18 The first dimension DW-NOMINATE scores are correlated with both dimensions of the DW-NOMINATE at 0.91 (p=0.00), with the ADA scores at 0.78 (p=0.00), and with the Turbo-ADA scores at 0.79 (p=0.00). 19 This percentage is similar to the percentages based on the other scores: 30.0 percent for both dimensions of DW-NOMINATE, 44.8 percent for ADA scores, and 35.9 percent for Turbo-ADA scores.

16 Theriault 16 Member Replacement Member replacement, while only accounting for 15 percent of the congress-to-congress comparisons, is responsible for more than half of the Senate polarization and more than two-thirds of the House polarization. Table 4 shows that the lion s share of the Senate replacement polarization comes from cross-party replacement. Same-party replacement also drives the parties apart. In contrast, when Senate incumbents lose, they are more frequently replaced by a more moderate member, though that trend has become decreasingly less pronounced over the 30 years. Insert table 4 about here. The numbers in table 4 highlight an important trend. The largest ideological and partisan shifts in Congress in the post World War II era have occurred within the South (Bullock 2000, Rohde 1991, and Stonecash et al. 2003). Whereas the Democrats occupied two-thirds of the southern Senate seats in the early 1970s, within three decades, the parties fortunes in the South perfectly reversed. 20 The switch from southern Democrats to conservative Republicans also accounts for more than half (53.3 percent) of the total polarization. In aggregate, the other replacement categories had a net moderating affect on the Senate (-2.75). Table 5 shows the comparable numbers for the House. The disparity between the three replacement categories is not as large: 25.4 percent of the total polarization is brought about by same party replacements, 28.6 percent by cross-party replacements, and 13.3 percent by incumbent defeats. The regional disparity is also great in the House. Whereas Republicans only filled 34 southern seats (31.7 percent) in the 93 rd Congress (1973-4), they held 71 seats (56.8 percent) in the 106 th Congress. As southern Democrats, the bulk of whom were in the middle third of the ideological continuum, died, lost, retired, or otherwise vacated their seat, they have been, for the most part, replaced by 20 Data from table 1.3 (page 10) in Norman J. Ornstein, Thomas E. Mann, and Michael J. Malbin s Vital Statistics on Congress, , Washington, D.C.: AEI Press.

17 The Case of the Vanishing Moderates 17 conservative Republicans. Quite simply, when extremists replaced moderates, the ideological middle disappeared and the parties diverged. 21 Insert table 5 about here. The loss of southern Democrats by both cross-party replacement and incumbent defeat accounts for more than half of the total polarization brought on by replacement. In the House, 57.3 percent of the replacement polarization and 38.6 percent of the total polarization results from the replacement of moderate southern Democrats by conservative Republicans. Ninety-two percent of the polarization brought about by a party switch happened in the South. Evidence for the Institutional and Electoral Explanations Three findings from this paper bear directly on the polarization debate in the congressional literature. First, any complete explanation for the divergence between the parties in Congress must not be specific to either chamber. As shown in figures 3 and 4, both chambers have become similarly polarized since Although the polarization numbers vary between the chambers, they are highly correlated (0.64; p=0.01). Explanations that rely upon characteristics or trends specific to one chamber must be viewed skeptically or, at least, largely incomplete. Second, it is more than just the changes in the House floor procedures that have caused the parties in Congress to diverge, but the changes in the House rules have played a role. The introduction of closed and other special rules cannot alone account for the divergence in the parties. As demonstrated by figure 4, the parties have also diverged in the Senate, which operates under unanimous consent agreements and typically does not debate particular bills under specific rules. But for the House rules, the other institutional characteristics the increasing importance of party leaders (though not to the same degree) and the issues on the political agenda affect both chambers similarly. 21 Unlike replacement, member conversions in the South did not differ in any appreciable way from the rest of the country.

18 Theriault 18 The institutional changes in the legislative process should affect all members similarly irrespective of chamber. The only category of micro-level changes involving the same member over time is member conversion. Due to the increasing importance of party leadership and the changing legislative agenda over time in both chambers, the member conversion scores by congress in the House and the Senate are highly correlated (0.96; p=0.00). One of the major institutional features that the chambers do not share is the use of rules during floor consideration. If the increasing use of closed or modified rules has an impact on party polarization in Congress, its effect should be felt only in the House. Given the high correlation between the member conversion scores, this test will be very demanding. A regression of House member conversion scores on Senate member conversion scores and the percentage point change in House bills debated under a restricted rule 22 will reveal if the growing frequency of closed and modified rules has an effect on party polarization. Table 6 shows the regression results. Both independent variables are statistically significant. Substantively, each percentage point increase in the number of House bills debated under a restricted rule increases the House conversion polarization score by (p=0.11). Given the low number of cases (just 15 congresses) and the high correlation between House and Senate member conversion scores, it is surprising that the House rules variable can explain any of the House member conversion variation. The fact that it almost reaches the conventional level of statistical significance is astounding. Insert table 6 about here. Third, it is more than just redistricting that has caused the parties in Congress to diverge, but redistricting has played a role. Again, the first clue that redistricting is not the only culprit of the polarized Congress is figure 4 showing that the Senate, which does not reapportion or redistrict, has also polarized. Replacement, in contrast to conversion, is more susceptible to electoral factors like 22 Data for restrictive rules were obtained from table 8-2 in Davidson and Oleszek (2004, 243).

19 The Case of the Vanishing Moderates 19 redistricting. Even after reapportionment and redistricting, incumbents will continue to win. The newly elected representatives, however, will be the most affected by the changes in the electorate. But for redistricting, these electoral factors are constant across the chambers. Replacement is also correlated across the chambers (0.47; p=0.08), but not nearly as highly as member conversion. Regressing House replacement polarization on Senate replacement polarization and a measure of redistricting 23 reveals that in the first two elections after redistricting, the polarization in the House is 2.71 polarization points higher than in years without redistricting (see table 7). Insert table 7 about here. VI. Discussion and Conclusion The analysis in this article reconciles the micro-level changes in members ideological scores with the macro-level divergence of the political parties in Congress since the 1970s. Through an examination of member congress-to-congress changes, the analysis provides the fundamental building blocks to understand the mechanisms responsible for party polarization. 24 Regrettably, in politics and most other questions of relevance in today s world, there rarely is a smoking gun that can explain everything. Rather, most right answers have the flavor of a little of this or a bit of that. The bureaucracy, redistricting, a weakening of parties, and members personal empires all had a role in the Case of the Vanishing Marginals. In the case of the vanishing moderates, the answer of both seems right. The parties in Congress have become more polarized as both a consequence of electoral changes and institutional changes. On top of the factors common to both the House and Senate, the 23 The redistricting variable is coded 1 for elections immediately after a redistricting cycle, 0.5 for the second election after redistricting, and otherwise 0. The regression results are robust to different operationalizations of this variable. 24 Various scholars, including Fiorina (1999), Lowry and Shipan (2002), and Poole and Rosenthal (1984) have offered more comprehensive explanations for party polarization; however, their explanations can be broken down pretty easily into the institution-based and districting-based categories.

20 Theriault 20 increasing adoption of restricted rules and the cycles of redistricting have exacerbated the polarization in the House. Although figures 3 and 4 show a systematic trend toward polarization, are they simply showing a magnification of a miniscule congressional trend? After all, an average change of 0.01 in a member s ideology score is negligible. Am I (and the other scholars studying polarization) making a mountain out of a molehill? One way to ascertain the magnitude of the trend is to compare the 106 th Congress ( ) to the 92 nd Congress (1971-2). If every member and senator from the 92 nd Congress had voted perfectly ideologically (i.e., every Democrat voting against every Republican), the polarization score of the 93 rd Congress could have been for the House and for the Senate (instead of 0.30 in the Senate and 1.67 in the House). In the 106 th Congress, if every member and senator had voted perfectly ideologically, the total possible polarization scores would have been and The reduction from the to (77.83 to 58.22) implies that the House (Senate) in the 106 th Congress is 24.1 (25.2) percent more polarized now than the House (Senate) in the 92 nd Congress. Although the congress-to-congress changes are indeed infinitesimal, the total change is substantial. Although the Congress is around 25 percent more polarized now than it was 30 years ago, it is unlikely that the parties will continue to polarize at the rate that they have since the 1970s. Several trends are reaching their natural conclusions. First, the majority of the rules in the House are already restrictive, so closed and modified rules cannot have as large of an effect as they have had in the passed. Second, the number of competitive races in the House continues to shrink. Third, it is mathematically impossible for as many moderate southern white Democrats to leave Congess as left between the 92 nd and 106 th Congress. Whereas 68.2 percent of the southern seats in Congress were held by white Democrats in the 92 nd Congress, they held only 28.8 percent of the seats in the 106 th Congress.

21 The Case of the Vanishing Moderates 21 In short, party polarization in Congress is real and significant; it is not likely that we have seen the end of this trend, though it is likely that we have already seen the most dramatic changes in the trend.

22 Theriault 22 Bibliography Aldrich, John H Why Parties? The Origins and Transformation of Political Parties in America. Chicago: University of Chicago Press. Aldrich, John H., and David W. Rohde The Consequences of Party Organization in the House: The Role of Majority and Minority Parties in Conditional Party Government in Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, D.C.: CQ Press. Ansolabehere, Stephen D., James M. Snyder, Jr., and Charles Stewart III. July 29, Candidates Positioning in U.S. House Elections. Unpublished Manuscript. Asher, Herbert B. and Herbert F. Weisberg Voting Change in Congress: Some Dynamic Perspectives on an Evolutionary Process American Journal of Political Science 22 (May): Bianco, William T Trust: Representatives and Constituents. Ann Arbor: The University of Michigan Press. Brady, David Critical Elections, Congressional Parties and Clusters of Policy Change. British Journal of Political Science 8: Brady, David W Critical Elections and Congressional Policy Making. Stanford, California: Stanford University Press. Brady, David, and Naomi Lynn Switched-Seat Congressional Districts: Their Effect on Party Voting and Public Policy. American Journal of Political Science 17 (August): Brady, David and Barbara Sinclair Building Majorities for Policy Changes in the House of Representatives. Journal of Politics 46 (Nov.): Brady, David W., and Craig Volden Revolving Gridlock: Politics and Policy from Carter to Clinton. Boulder, Colorado: Westview Press. Bullock III, Charles S Partisan Changes in the Southern Congressional Delegation and the Consequences in Continuity and Change in House Elections, eds., David W. Brady, John F. Cogan, and Morris P. Fiorina. Stanford, California: Stanford University Press. Burnham, Walter Critical Elections and the Mainspring of American Politics. New York: W.W. Norton. Burstein, Paul A New Method for Measuring Legislative Content and Change. Sociological Methods and Research 6: Burstein, Paul Attitudinal Demographic and Electoral Components of Legislative Change: Senate Voting on Civil Rights. Sociology and Social Research 64: Carson, Jamie, Michael H. Crespin, Charles J. Finnocchiaro, and David W. Rohde Linking Congressional Districts Across Time: Redistricting and Party Polarization in Congress. Paper presented at the 2003 Annual Meeting of the Midwest Political Science Association. Clausen, Aage R How Congressmen Decide: A Policy Focus. New York: St. Martin s Press. Coleman, John J The Decline and Resurgence of Congressional Party Conflict. Journal of Politics 59 (February): Collie, Melissa P. and John Lyman Mason The Electoral Connection Between Party and Constituency Reconsidered: Evidence from the U.S. House of Representatives,

23 The Case of the Vanishing Moderates 23 in Continuity and Change in House Elections, eds., David W. Brady, John F. Cogan, and Morris P. Fiorina. Stanford, California: Stanford University Press. Davidson, Roger H. and Walter J. Oleszek Congress and Its Members, 9 th Edition. Washington, D.C.: CQ Press. Downs, Anthony An Economic Theory of Democracy. New York: Harper and Row. Evans, C. Lawrence, and Walter J. Oleszek Congress Under Fire: Reform Politics and the Republican Majority. Boston: Houghton Mifflin Company. Fenno, Richard F., Jr U.S. House Members in Their Constituencies: An Exploration. American Political Science Review 71 (September): Fiorina, Morris P The Case of the Vanishing Marginals: The Bureaucracy Did It. American Political Science Review 71 (March): Fiorina, Morris P Whatever Happened to the Median Voter. Paper presented at the Midwest Political Science Association Annual Meeting at Chicago, Illinois, April (updated paper prepared for the MIT Conference on Parties and Congress, Cambridge, MA, October 2, 1999). Fleisher, Richard and Jon R. Bond The Shrinking Middle in Congress. British Journal of Politics Forthcoming. Fleisher, Richard and Jon R. Bond Partisanship and the President s Quest for Votes on the Floor of Congress in Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, D.C.: CQ Press. Ginsberg, Benjamin Critical Elections and the Substance of Party Conflict: Midwest Journal of Political Science 16: Ginsberg, Benjamin Elections and Public Policy. American Political Science Review 70 (March): Hibbing, John R., and Elizabeth Theiss-Morse Congress as Public Enemy: Public Attitudes Toward American Political Institutions. New York: Cambridge University Press. Hibbing, John R., and Elizabeth Theiss-Morse Stealth Democracy: Americans Beliefs about How Government Should Work. New York: Cambridge University Press. Howell, William, E. Scott Adler, Charles Cameron, and Charles Riemann Measuring the Institutional Performance of Congress in the Post-war Era: Surges and Slumps in the Production of Legislation, Legislative Studies Quarterly 25: Jacobson, Gary C Party Polarization in National Politics: The Electoral Connection. Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, D.C.: CQ Press. Jacobson, Gary C The Politics of Congressional Elections, 5 th ed. New York: Longman. Jamieson, Kathleen Hall, and Erica Falk Continuity and Change in Civility in the House. Polarized Politics: Congress and the President in a Partisan Era, eds. Jon R. Bond and Richard Fleisher. Washington, D.C.: CQ Press. Jones, Charles Speculative Augmentation in Federal Air Pollution Policy-Making. Journal of Politics 36 (May): Kingdon, John W Congressmen s Voting Decisions. Ann Arbor: The University of Michigan Press.