This item is only available as the following downloads:
1 RESOLVING DIFFERENCES: BICAMERAL DISAGREEMENT AND RECONCILIATION IN THE POSTREFORM CONGRESSES By JORDAN MICHAEL RAGUSA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT O F THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2011
2 2011 Jordan Michael Ragusa
3 To my parents, for it is in your footsteps I follow a nd to Christine, for her unwavering support of my academic endeavors
4 ACKNOWLEDGMENTS This dissertation is dedicated to my parents my mother, Marilyn Grinnell my father, Don Ragusa and my step mother, Elizabeth Yarris I would not be w riting a dissertation acknowledgment were it not for your intelligence guidance and encouragement Most of all I want to a cknowledge the inspiration and wisdom of my father, a n associate professor emeritus of psychology, who died during my graduate train ing. This dissertation is also dedicated to my fia nce, Christine. I am fortunate to have had your unwavering support while pursuing my Ph D : a challenging time for both of us I am forever indebted to my dissertation advisor and mentor, Larry Dodd, who taught me to c ontinuously with theoretical puzzles. T hough I was afforded significant fr eedom over this project, inevitably voluminous books and articles on Congress informed this dissertation. Moreover, t his project was born in his le gislative polit ics seminar and in our many lengthy but engaging conversations. students routinely cite two traits that make him particularly effective as a mentor : his kindness and unrivaled knowledge of American politics congressional developmen t and empirical theory I hope to imitate his example as I begin my own career a sincere form of flattery indeed Major elements of this project were fo stered, clarified and revised with the guidance and encouragement of my dissertation committee. I am deeply indebted to each of these individuals. Michael Martinez provided insightful methodological suggestions and offered constructive criticism on ways to improve my argumentation He was a constant presence at department seminars where most of these ch apters were unveiled I am indebted to Rich Conley for sharing his expertise on party politics
5 checks and balances and the presidency Rich, along with Larry, had the biggest effect on the substance of this project. Michael Heaney encouraged me to thin k about the theoretical mechanisms driving my empirical results. Marcus Hendershot provided his methodological expertise, particularly in the area of time series analysis an d offered helpful comments on my writing style And last but not least, I am ind ebted to Elizabeth Dale for helping me clarify my argumentation about the Framers and their views on bicameralism I would also like to thank my fellow American politics doctoral students at the University of Florida. Josh Huder, Paulina Rippere, and Wi ll Hicks are close friends w ho supported me during my Ph.D. training both personally and professionally. All three p rovided guidance comments or concerns on key elements of this dissertation I am thankful to have them as both friends and trusted collea gues. I would also like to acknowledge those individuals who commented on chapters of this manuscript at conferences. I am particularly grateful to David Rohde and Bruce Oppenheimer for inviting me to participate in the by the Political Institutions and Public Choice Program at Duke University. Thanks go out to Gregory Koger and Burde tt Loomis for providing feedback at tha t conference. I am also grateful for comments and suggestions received at regional conferences. D avid Rohde provided his thoughts o n Chapter 3 at the 2011 Southern Political Science Association meeting. James Rogers provided helpful criticisms of Chapter 5 at the 2010 Midwest Political Science Association meeting. And Ryan Va nder Wielen commented on Chapter 4 at the 2009 Midwest Political Science Association meeting. I am also thankful to Michael C. Brady who shared data on conference committees and
6 offered substantive advice on this project Each of these individuals has contributed to this disse rtation
7 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ .......... 10 LIST OF FIGURES ................................ ................................ ................................ ........ 12 ABSTRACT ................................ ................................ ................................ ................... 14 CHAPTER 1 INTRODUCTION: BICAMERAL DISAGREEMENT AND RECONCILIATION IN THE POSTREFORM CONGRESSES ................................ ................................ .... 16 The Marriage of the House and Senate ................................ ................................ .. 16 Constitutional Beginnings ................................ ................................ ....................... 17 What We Know About Bicameral Disagreement and Reconciliation ...................... 22 Bicameral Disagreement ................................ ................................ .................. 22 Resolving Differences ................................ ................................ ...................... 26 Amendment trading ................................ ................................ .................... 27 Conference committees ................................ ................................ ............. 28 What Do We Know About Resolving Differences? ................................ ........... 29 Scope and Method: Why Study Bicameral Disagreement and Reconciliation in the Postreform Era? ................................ ................................ ............................ 35 ................................ ................................ ....... 35 Scope ................................ ................................ ................................ ............... 37 Method ................................ ................................ ................................ ............. 41 Theoretical Structure and Empirical Expectations ................................ .................. 42 Bicameral Disagreement ................................ ................................ .................. 43 Resolving Differences ................................ ................................ ...................... 44 Content of the Dissertation ................................ ................................ ..................... 46 2 ESTIMATING THE BICAMERAL HURDLE ................................ ............................ 52 Political Parties and Inter Branch Bargaining ................................ .......................... 54 Political Parties and I nter Chamber Bargaining ................................ ...................... 57 Theorizing About Bicameral Compositional Differences ................................ ......... 58 The Link Between Bicameral Distance and Gridlock ................................ ........ 58 The House and Senate Medians in the Postreform Era ................................ ... 60 Polarization Asymmetry ................................ ................................ .................... 62 Two Additional Bicameral Consequences of Polarization Asymmetry .............. 70 Intra party Bicameral Distance ................................ ................................ ......... 70 House Filibuster Pivot Di stance ................................ ................................ ....... 72 Methodology: Estimating the Bicameral Hurdle ................................ ...................... 74 Bicameral Conflict in the Postreform Era ................................ .......................... 76
8 Independent Variables ................................ ................................ ..................... 81 Bicameral Gridlock ................................ ................................ ................................ .. 85 Trends in Bicameral Gridlock ................................ ................................ ........... 85 Main Bicameral Gridlock Estimates House Initiated Bills ............................... 88 Main Bicameral Gridlock Estimates Senate Initiated Bills .............................. 91 Bicameral Disagreement ................................ ................................ ......................... 93 Trends in Bicameral Disagreement ................................ ................................ .. 93 Main Bicameral Disagreement Estimates ................................ ......................... 94 Discussion ................................ ................................ ................................ .............. 97 3 THE DIMENSIONALITY OF RESOLVING DIFFERENCES ................................ 117 A Unified Typology o f Resolving Differences ................................ ........................ 119 Reconciliation ................................ ................................ ................................ 122 Partisan Conflict ................................ ................................ ............................. 124 Bica meral Conflict ................................ ................................ .......................... 125 ................................ ................................ ..... 126 Theoretical Expectations for Conference Outcomes ................................ ............. 12 7 ................................ ................................ ....... 133 Principal Components Analysis ................................ ................................ ............ 139 The Spatial Dynamics of Resol ving Differences ................................ ................... 143 Cases in Each Dimension ................................ ................................ ..................... 145 Reconciliation ................................ ................................ ................................ 145 Parti san Conflict ................................ ................................ ............................. 148 Bicameral Conflict ................................ ................................ .......................... 150 The Explanatory Power of Each Dimension ................................ .......................... 151 Quasi Divided Government ................................ ................................ ................... 157 Discussion ................................ ................................ ................................ ............ 160 4 RESOLVING DIFFERENCES IN TIME AND SPACE: MODELING BILL LEVEL CONFERENCE OUTCO MES ................................ ................................ ............... 168 Reestimating the Resolving Differences Policy Space by Era .............................. 169 The Winners and Losers: Mapping Conference Outcomes ................................ .. 174 Determinants of Conference Outcomes ................................ ................................ 179 Modeling the First Dimension: Reconciliation ................................ ................. 180 Modeling the Second Dimension: Partisan Conflict ................................ ........ 185 Modeling the Third Dimension: Bicameral Conflict ................................ ......... 193 Discussion ................................ ................................ ................................ ............ 196 5 RESOLVING DIFFERENCES: FACILITATOR OF, OR IMPEDIMENT TO, MAJORITY PARTY AGENDA SETTING? ................................ ............................ 214 Theoretical Foundations for Agenda Control ................................ ........................ 216 The Bicameral Sequence ................................ ................................ ...................... 218 Bicameral Sequence and Agenda Control ................................ ............................ 221 C ................................ ................. 221
9 Amendment Trading: Agenda Control in the House and Senate .................... 222 How Bicameral Sequence Aids the Majority Party ................................ ......... 225 Formal Models of Agenda Control ................................ ................................ ........ 226 Data and Methods ................................ ................................ ................................ 227 The 9/11 Commission Bills: An Example ................................ .............................. 234 Findings ................................ ................................ ................................ ................ 235 Comparing Conferencing and Amendment Trading ................................ .............. 238 Discussion ................................ ................................ ................................ ............ 241 6 CONCLUSIONS ................................ ................................ ................................ ... 250 The Marriage of the House and Senate in Historical Perspective ......................... 250 The Honeymoon Period ................................ ................................ ................. 250 The Emergence of an Activist Senate ................................ ............................ 251 ............................. 253 The Textbook and Reform Eras ................................ ................................ ..... 256 The Postreform Period ................................ ................................ ................... 258 How the House and Senate Coevolve ................................ ............................ 260 Overarching Conclusions ................................ ................................ ...................... 262 Specific Conclusions ................................ ................................ ............................. 263 Additional Empirics ................................ ................................ ............................... 266 Implications ................................ ................................ ................................ ........... 272 Strong Parties or Weak Parties? ................................ ................................ .... 272 The Constitution, Political Parties and Bicameralism ................................ ..... 283 The Nature of Resolving Differences ................................ ................................ .... 292 LIST OF REFERENCES ................................ ................................ ............................. 300 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 315
10 LIST OF TABLES Table page 2 1 .......................... 110 2 2 Temporal trends in the bicameral gridlock ................................ ........................ 111 2 3 Determinants of bicameral gridlock ................................ ................................ .. 112 2 4 Determinants of bicameral gridlock with interaction effects .............................. 113 2 5 Determinants of bicamer al gridlock ................................ ................................ .. 114 2 6 Temporal trends in bicameral disagreement ................................ ..................... 115 2 7 Determinants of bicameral disagreement ................................ ......................... 116 3 1 A typology of resolving differences ................................ ................................ ... 165 3 2 Component loadings (untransformed data, 95 th to 110 th ) ................................ 165 3 3 Component loadings (transformed data, 95 th to 110 th ) ................................ ..... 166 3 4 Explained variance by dimension (untransformed data, 95 th to 110 th ) .............. 166 3 5 Explained variance by dimension (transformed data, 95 th to 110 th ) .................. 166 3 6 Component loadings (untransformed data, 97 th to 99 th and 107 th ) .................... 167 3 7 Explained variance by dimension (untransformed data, 97 th to 99 th and 107 th ) 167 4 1 Component loadings (untransformed Data, 95 th to 96 th and 100 th to 103 rd ) ...... 207 4 2 Component loadings (untransformed data, 104 th to 106 th and 108 th to 110 th ) ... 207 4 3 Component loadings (transformed data, 95 th to 96 th and 100 th to 103 rd ) ........... 207 4 4 Component loadings (transformed data, 104 th to 106 th and 108 th to 110 th ) ....... 208 4 5 Explained variance by dimension (untransformed Data, 95 th to 96 th and 100 th to 103 rd ) ................................ ................................ ................................ ............ 208 4 6 Explained variance by dimension (untransformed data, 104 th to 106 th and 108 th to 110 th ) ................................ ................................ ................................ ... 208 4 7 Explained variance by dimension (transformed Data, 95 th to 96 th and 100 th to 103 rd ) ................................ ................................ ................................ ................ 209
11 4 8 Explained variance by dimension (transformed data, 104 th to 106 th and 108 th to 110 th ) ................................ ................................ ................................ ............ 209 4 9 Determinants of reconciliation outcomes ................................ .......................... 210 4 10 Determinants of partisan outcomes ................................ ................................ .. 211 4 11 Determinants of partisan conference outcomes ................................ ............... 212 4 12 Determinants of bicameral conflict conference outcomes ................................ 213 5 1 All House final passage votes, 94th 110th Congresses ................................ 243 5 2 All House final passage votes, 104th 110th Congresses ............................... 243 5 3 All House conference report votes, 94th 110th Congresses .......................... 244 5 4 All House shuttle final passage votes, 94th 110th Congresses ..................... 244 5 5 All House conference report final passage votes, 104th 110th Congresses .. 245 5 6 All House shuttle final passage votes, 104th 110th Congresses ................... 245 5 7 All Senate conference report votes, 94th 110th Congresses ......................... 246 5 8 All Senate conference report votes, 104th 110th Congresses ....................... 246 5 9 Conference and shuttle interaction effects ................................ ....................... 247 6 1 Conference committee failure rate ................................ ................................ .... 298
12 LIST OF FIGURES Figure page 1 1 Percentage of Hou se bills cleared by the House ................................ ................ 51 1 2 Percentage of Sen ate bills cleared by the Senate ................................ .............. 5 1 2 1 A spatial representation of the 99 th Co ngress ................................ .................. 101 2 2 A spatial representation of the 100 th Congress ................................ ................ 101 2 3 Bicameral distance ................................ ................................ ........................... 102 2 4 Adjusted bicameral distance ................................ ................................ ............ 102 2 5 Polarizat ion in the House and Senate ................................ .............................. 103 2 6 Polarization asymmetry ................................ ................................ .................... 103 2 7 I ntra party bicameral distance ................................ ................................ .......... 104 2 8 Filibuster House distance ................................ ................................ ................. 104 2 9 Bicameral gridlock ................................ ................................ ............................ 105 2 10 Bicameral disagreement ................................ ................................ ................... 105 2 11 Adjusted bicameral disagreement ................................ ................................ .... 106 2 12 P artisan random effect (pre Revolution) ................................ .......................... 107 2 13 Partisan random ef fect (post Revolution) ................................ ......................... 108 2 14 Predicted probability of Senate passage by Intra party*partisan (model 7). ..... 109 2 15 Predicted probabil ity of Senate passage by Bicameral*partisan (model 9). ..... 109 3 1 Scree plot for the untransformed data. ................................ ............................. 163 3 2 Scree plot for the untra nsformed data. ................................ ............................. 163 3 3 3 D spatial map of resolving differences ................................ ........................... 164 4 1 3 D spatial map of resolving differences in the pre Revolution era ................... 200 4 2 3 D spatial map of resolving differenc es in the post Revolution era ................. 201 4 3 2 D spatial map of resolving differences by e ra ................................ ................ 202
13 4 4 Rec onciliation median coordinate ................................ ................................ ..... 203 4 5 Reconcilia tion median absolute coordinate ................................ ...................... 203 4 6 Part isan conflict median coordinate ................................ ................................ .. 204 4 7 Bicam eral conflict median coordinate ................................ ............................... 204 4 8 Distribution of partisan pre conference bills ................................ ...................... 205 4 9 Estimated effect of identical House and Senate partisan bills on pro majority conference outcomes ................................ ................................ ....................... 206 5 1 Formal models of agenda control ................................ ................................ ..... 248 5 2 Conference and chamb er median interaction effects ................................ ....... 249 6 1 Average public law length for bills referred to conference ................................ 299
14 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy RESOLVING DIFFERENCES: BICAMERAL DISAGREEMENT AND RECONCILIATION IN THE POSTREFORM CONGRESSES By Jordan Michael Ragusa August 2011 Chair: Lawrence C. Dodd Major: Political Science This dissertation is about the marriage of the House and Senate a m e tapho r juxtaposing two features of the U.S. Constitution : the distinct personalities imposed upon the two chambers and the requirement that the House and Senate consent to identical versions of legislation This is defined like any relationship b y the frequency and severity of conflict and disagreement The main finding s of this dissertation are as follows. Chapter 2 shows that House and Senate have experienced more frequent and more severe bicameral disagreements over the past thirty years I attribute these trends to growing compositional asymmetrie s between the Hous e and Senate Moreover I find that pro majority House passed legislation has become especially prone to bicameral gridlock over this period Chapter 3 d escribes the process of resolving House and Senate policy disputes via conference committee U sing m ultivariate spatial modeling, t he results show that resolving differences is multidimensional in nature and that the greatest amount of variation in conference roll call patterns is a proces s of compromise and concessio n Chapt er 4 explores conference outcomes at the bill level I find that over the past thirty years there has been: (1) an increase in the variability of conference
15 outcomes, (2) an increase in pro minority conference outcomes, and (3) an increase in c ompromise and concession in conference. Using the multivariate spatial coordinates estimated in Chapter 3 a final section of Chapter 4 finds that conference committees operate in a maj oritarian fashion Chapter 5 concludes by examining resolving dif ferences through formal models of agenda s etting. The results show that partisan models of agenda control perform best when we examine all final passage votes but when we examine bills that went to a conference committee or were shuttled between the chambers a non partisan model outperforms all rivals. The overarching narrative is this : while it is true that parties in Con gress have enjoyed enhanced organizational capacities over the postreform period, the majority party has simultaneously faced growing bicameral constraints has manifested in mo re frequent and more severe House and Senate policy disputes. These developments are paradoxical given the conventional wisdom about parties in Congress
16 CHAPTER 1 INTRODUCTION: BICAMERAL DISAGREEME NT AND RECONCILIATIO N IN THE POSTREFORM CONGRESSE S The Marriage of the House and Senate This dissertation is about the marriage of the House and the Senate C odified in Article 1, Section 7 of the Constitution which requires t he concurrence of both houses on i dentical versions of legislation the union of the House and Senate has been surprisingly resilient. This stands in cont rast to E dmund Randolph Constitutional Convention that could never long co exist (quoted in Wood, 1998, 556 ). Though history has thus far proven Randolph wrong it is true that, as with any marriage, the Hou se and Senate frequently disagree when attempting to enact legislation. Often these disagreements are easily resolved, especially when the policy diff erences are minor. But in some cases the two chambers prefer widely different policies and quick, amicab le solutions are difficult to reach. And in the most dramatic cases, the House and Senate are unable to reach a consensus preserving a status quo unfavorable to a majority (or supermajority) in each chamber To understand the marriage of the House and Senate t his project is the first to explore two critical stages in the legislative process bicameral disagreement and reconciliation jointly. T he topic of this dissertation is animated by two overarching premises First, our understanding of how the Hou se and Senate interact in the policy to day operation Second, the relationship of the House and Senate is variable and thus amenable to systematic study, rather than constant These t wo premises stand in contrast to the tendency in congressional research to examine the operation a single chamber. While single chamber studies are important because they help researchers isol ate key
17 organizational features, institutions, outcomes, they a re limited inherently in their ability to explain certain historical developments and aggregate patterns. And at th e same time, many single chamber studies discriminate between competing organizational theories the distributive, partisan, informational an d majoritarian theories while ignoring the central feature of congressional organization the effects of its bicameral structure Some of the central questions addressed in this dissertation are: What role do political parties play in our system of consti tutional checks and balances? How have the increases in polarization and the organizational strength of parties altered the functioning of legislative checks and balances ? How do competing parties and chambers resolve bicameral disagreements? How do con ference committees affect policy outcomes and how have these patters changed over the postreform period ? Constitutional Beginnings The marriage of the House and Senate was forged in the adoption of a bicameral congress. This decision stands in sharp contr ast to the unicameral legislature of sovereign states established by the Articles of Confederation. But while many items generated considerable controversy and discussion during the Constitutional Convention, once a confederated form of government was aba ndoned by the delegates 1 a federal congress with two chambers (Wood 1998, 553). Thus, during the Founding era t he salient debates regarding 1 That is to say, once the Antifederalists had lost the debate over preserving a confederated system of government and debate turned toward the Federalist plan of fashioning a new national government (see Wood 1998, 547 562).
18 bicameralism concerned two issues: how to ap portion each chamber and how to selec t members. Out of the conflict between states of varying population sizes emerged the well known delegation who proposed the solution) This constitutional compromise established an upper house, the Senate, with two representatives from each state and a lower chamber, the House, with representatives in proportion Thus, unlike some bicamera l legislatures where the interaction of both chambers is the Framers of the U.S. Constitution fashioned a congress with two dissimilar chambers. 2 This decision was not a foregone conclusion or without controversy. bicameral legislat ure with each house apportioned according to population 3 Moreover a handful of delegates, most famou sly Edmund Randolph, thought the establishment of two distinct chambers would create a fatal tension between the two bodies 4 Ultimately however, the F a House and Senate with inherent tensions due to the compromise creating differences in apportionment, thus establishing bicameral cooperation and conflict as a c entral aspect of Congressional operation Though t he House and Senate organization reflected a compromise, the relationship of the two chambers reflected the purported virtues of legislative checks 2 (Tse belis and Money 1997). 3 Of course, the Virginia plan called for members of the House to be elected by the general population and members of the Senate to be elected by the state legislatures. 4 claim that co
19 and balances. 5 This principle is most clearly explicated in Federalist no. 51 where Madison discusses solutions to the problem one solution is divide the legislature into different branches; and to render them, by different modes of election and different principles of action, as little connected with each other as the nature of their common func tions and their common dependence on the society will 6 James Wilson echoed some one or more steps of its progress, undergo the keenest scrutiny. Its relations, whether near or more remote, to the principles of freedom, jurisprudence, and the innovations, crude projects, and partial contrivances will be stifled in the attempt to bring 7 Co ntemporaries ha ve endorsed these claims, noting that the virtues of a and increases the stability of enacted policies ( Riker 1992; Tsebelis and Money 1997). The principle of l eg islative c hecks and balances was implemented in a few ways. First, the new congress was made to represent diverse interests. In part icular, out of the Connecticut C ompromise emerged an inherent tension between state interests (vested in the Senate) and n ational interests (vested i n the House), a principle known as in Federalist no. 62 can now be passed without the concurrence first, of a majority of the people, and then 5 each chamber. 6 All references to the Federalist Papers in this manuscript are in referenc (1961, ed.) The Federalist Papers 7 The Works of James Wilson Edited by Robert Green McCloskey. 2 vols. Cambridge: Belknap Press of Harvard University Press, 1967.
20 of a majority of the states Though this feature reflected a compromise, rather than spirit of legislative checks and balances nonetheless. Second, there was an explicit attempt to focus th e energies of each chamber on qualitatively different policy domains T his principle was codified in the varying t enure of lawmakers. With senators serving six year terms and representatives two year terms, the upper chamber would focus its energies on l ong 8 At t he same time, with longer terms senators were insulated from short term fluctuations in public opinion or was thought to be more responsive to exogenou s changes. Third, many of the Framers be lieved in the normatively beneficial effects of bicameralism on public policy arguing that a second chamber would enhance proper legislative decision making. In fact i t was argued during the F ounding that the Sen ate, because of greater policy expertise, longer terms, and historical precedent, would legislation passed by the House and delay the passage of defective policy ( Federalist nos. 62 and 63 ). 9 Or as Madison famously put it, the Senate would prote 10 8 In Federalist no. 63 this defect must be an additional body in the legislative department, which, having sufficient permanency to provide for such objects as require a continued attention, and a train of measures, may be justly and effectually answerable for the attainment of 9 In Federalist no. 62 impulse of 10
21 constitutional decisions were based on instrumental rationality alone 11 we can speak about the presumed effects of the legislative institutions they conferred us irrespective of their rational or materialistic intentions (Jillson and Euba nks 1984) In this respect, a belief in normatively beneficial effects of inter chamber cooperation and conflict remained an overriding organizational principle. 12 Thus bicameral disagreement and in our system of government, mere byproducts of ou r constitutional framework, they are the manifestations of one of the central features of legislative organization What do we kno w about bicameral disagreement on the one hand, and the process and patte rns of resolving those disagr eements on the other hand? 11 The problem is one of backward induction, a common historical institutionalist critique of For example, during the Constitutional Convention the Federalists argued that the Senate would represent the states and th e House the people only after the Connecticut compromise had been established (Wood 1998, 558). to population rred plans failed to make their way into the and Oppenheimer (1999) note that the Framers favored a bicameral legislature before discussing the issue of the Se Rather, these Constitutional decisions occurred within a particular historical context. Or, as (quoted in Jillson and Eubanks 1984). But as I point out, we can use concerning what they perceived as the ultimate effects of the institutional structures they gave us irrespective of their rational or irrational intentions in order to consider the functioning of our current Congress. 12 And though one constitutional difference between the two chambers was undone with the passage of the 17 th amendment (Bernhard and Sala 2006; Crook and Hibbing 1997), core constitutional differences remain in place.
22 What We Know About Bicameral Disagreement and Reconciliation Bicameral Disagreement Because inter chamber bargaining is an explicit and central feature of congressional organization, it is surprising that a systematic examination of the frequency and severity of bicameral disagreement is absent from the congressional literature Some may say this lacuna represents major gap in our current understa nding W ith few exceptions, the extant discussion o f disagreement between the H ouse and Senate focuses on one or a few cases of major conflict (usually as part of a more ecumenical study) From the Contract with America the Bush era tax cuts and the Comprehensive Immigration Reform Act of 2006 to NAFTA and House and Senate has structured numerous landmark policy outcomes. Among the cases of bicameral disagreement cited in the congressional literature the Contract with America has reviewing the first session of 104 th Congress, Ornstein and Schenkenberg (1995) point at or moderated the majority of those policies Indeed, studies of this landmark period contend that Republicans faced not only the usual challenge of a rival president, but the additional challenges created by ideological divisions within their party a c ross the House and Senate ( Connelly and Pitney 1997; Koopman 1996; Ornstein and Schenkenberg 1995). Additional discussion of bicameral disagreement in the 104 th Congress can be found in landmark studies by Tsebelis and Money (1997, 54), Brady and Volden ( 2006, 142 159) and Binder (1999, 524). The failure to enact immigration reform in the 109 th Congress is another landmark case of
23 bicameral disagreement cited in the congressional literature. For example, Baker (2008) notes that while most issues in the m odern Congress divide lawmakers along ideological and/or partisan lines, the d ebate over immigration reform in the 109 th 21 3). In the end, the reform effort died in a conference committee as negotiators appointed by both chambers were unable to find a workable solution. Conflict between the House and Senate over the North American Free Trade Agreement (NAFTA) is yet another case of consequential bicameral disagreement. In one analysis, Wirls (1998) shows us that the Senate apportionment scheme created critical tensi ons between the two chambers on NAFTA. He concludes that even in a Congress with similar preferences in both chambers, the distribution of those preferences can have major effects on the success or failure of legislation. T he aforementioned studies detailing instances of major House and Senate policy disagreement have three general strands. First, as a matte r of scope and method, none of these studies concern bic a meral disagreement exclusively. Rather, House and Senate policy con flict is a component of a b roader or more ecumenical study Second, each study suggests that disagreement between the House and Se nate is structural stemming from institutional characteristics rather than idiosync ratic. Though this is a simple observation on its face it suggests that we should l ook at the institutional causes of bicameral disagreement rather than looking exclusivel y at exogenous factors such as whether one party controls both chambers. And third, most of the aforementioned studies suggest ( quite explicitly) that ideological differences across the House and Senate arising from within the parties are powerful determi nants
24 of House Senate conflict But while we know intuitively that Democrats (Republicans) in the House often disagree with Democrats (Republicans) in the Senate this observat ion has yet to be incorporated into a systematic theory or subject to empirical examination. This latter point will be developed more fully in a subsequent section. In addition to the cases of bicameral disagreement cited by congressional scholars, related research offers us some clear expectations. Indeed, t he topic of bi cameral conflict is subsumed by the literature on policy production (Mayhew 1974; Binder 1999, 2003, 2008; Krehbiel 1996, 1998; Brady and Volden 1998 2006; Chiou and Rothenberg 2008a, 2008b) and policy durability ( Hammond and Miller 1987; Maltzman and Shipan 2008 ; Ragusa 2010 ; Riker 1992 ). As will be e vident, f or the present project the most theoretically relevant study of policy productivity War gridl ock (Binder 1999, 2003, 2008). throu ghout this manuscript as aggregate gridlock. That is, her grid lock measure combines agenda items that failed on the House and Senate floor, fail ed because of bicameral disagreement and failed because of a presidential veto Nonetheless disagreement be tween the House and Senate receives some explicit discussion in her 2003 book Stalemate For example, Binder estimates that 46.4% of her agenda items died in the second acting chamber after being passed by the first acting chamber. Further, she reports t hat 49.0% of her agenda items were passed by the House but died in the Senate while 43.7% were passed by the Senate but died in the House. Thus, bicameral disagreement is quite common and consequential. This alone makes the systematic study of bicameral disagreement a worthy endeavor
25 (50). This, she says bers of Congress, we cannot simply assume that the two chambers hold similar sets of views But b in the aggregate, the aforementioned items are the only bits of information spe cific to bicameral disagreement. Thus, though research has informed the topic of this dissertation in important ways the congressional literature still lacks a systematic study of the determinants of bicameral disagreement or the causes of the a pparent increase in bicameral roadblocks in the 1990s. It is my hope that this dissertation builds on 13 disagreement is her conclusion that that the id eological distance between the House and Senate medians is a powerful determinant of aggregate gridlock in the post WW II era. In fact, Binder finds that bicameral distance has the most substantively important effect on the overall levels of gridlock (g reater in magnitude, for example, than divided party control). As she notes, this finding helps explain instances of stalemate during unified government in the 103 rd finding has been confirmed elsewhere. For example, Maltzman and Shipan (2008) look at differences between the chambers on the durability of legis lation, finding that enacted laws are less likely to be amended when there are major differences between House and Senate (Ragusa 2010). 13 I would like to thank Sarah Binder for providing helpful correspondence on components of this project.
26 However, in two separate papers, Chiou and Rothenberg (2008a and 2008b) challenge (la rgely along methodological lines). In Chapter 2 I develop a theory and set of hypotheses that hopefully reconciles these divergent findings. In sum the literature on bicameral disagreement is underdeveloped. This is surprising given that the formal i nteraction of the House and Senate is a critical feature of legislative organization and the day to day operation of the U.S. Congress T he relevant literature on this topic focuses on either one or a few cases of major policy conflict (rather than a syst ematic examination) or speaks to the topic of bicameral disagreement tangentially through l arger or more ecumenical research (policy productivity or policy durability ). Given the frequency of bicameral stalemate and the landmark legislation which has expe rienced bicameral disagreement an exclusive examination of House Senate policy conflict is long overdue. But d espite the somewhat limited state of the literature, we can piece together the following expectations : (1) bicameral disagreement is st ructural, due to systematic variation in bicameral design, rather than purely exogenous, a funct ion of party control ; (2) the distribution of preferences within each chamber ( particularly within party House and Senate cleavages ) is a primary determinant of bicamera l conflict; and (3 ) disagreement between the chambers increased around the 1990s. This dissertation will return to these important issues in Chapter 2 Resolving Differences Though the Constitution is silent on how the House and Senate should resolve thei r ( inevitable ) policy disputes, two extra constitutional methods guide the two chambers in merging competing versions of public policy one a formal institution, the
27 other a procedural mechanism. These methods are the convening of a conference committee an d amendment trading (respectively). Amendment t rading The first form of bicameral reconciliation is a mechanism known as amendment characterized by the two chambe rs agreeing to identical versions of legislation after messaging the original bill back and forth with amendments. In modern practice, the second chamber to act will typically substitute its preferred statutory language (an amendment in the nature of a su certain provisions. After the first amendment to th chamber which has the options amending the bill furth ( thereby preserving the exogenous status quo ) On rules (Palmer and Bach 2003). That is, each chamber has one opportunity to amend the first amendment of the other chamber. A third degree amendment is allowed by unanimous consent in the Ho use and Senate or by suspension or special rule in the House. Of course, further degrees of amending are allowed without objection (Oleszek 2004, 259). Thus, the process of resolving differences by amendment trading is analogous to what game theorists ca ll sequential bargaining. With the exception of Chapter 5 this dissertation examines conference committees as the means of resolving bicame ral differences. On the one hand this substantive focus reflects methodological necessity. Conference committees, unlike
28 amendment trading, provide easily comparable and simultaneous roll call votes (where pre conference House and Senate votes can be directly compared to post conference House and Senate votes). On the other hand, conference committees typically recon cile disagreements on the most salient and consequenti al bicameral disputes. Thus, conference committees are in my view, more worthy of systematic study Conference c ommittees If at any time during the legislative process one chamber insists on its pro posal initiated the formation of an ad hoc joint conference committee (for an excellen t review see Longley and Oleszek 1989). Though conference committees are not mentioned in the Constitution, they have been a distinct feature of Congressional organization since the very first session of the first Congress (Longley and Oleszek 1989). T he earliest lawmakers realized soon after the founding that conference committees are efficient solutions to bicameral stalemate. Compared to amendment trading, c onference committees are usually convened when the policy differences of the two chambers are l arge or the bill is controversial. O nce conferees have completed negotiations, a simple a document detailing the proposed compromise. The conference report is then sent back to both chambers who vote on the compromise bill ( Longley and Oleszek 1989; Olesz ek 2007). If both approve forwarded to the e
29 possible, including amendments in true 14 or technical 15 disagreement. However these procedures seldom arise (Oleszek 2007). What Do We Know About R esolving D ifferences? The limited research on bicameral disagreement is matched by the lack of research on r esolving differences. In their seminal work on this topic, Longley and Oleszek (1989) remarked that post Krehbiel (1991, 194) n otes behavior are not well polished as two authors bluntly put it, In fact in the last twenty years only two books have been dedicated to conference committees (Longley and Oleszek 1989; Van Beek 1995), a few journal articles (Nagler 1989; Lazarus and Monroe 2007; Vander Wielen 2010) and couple of book chapters (Smith 1989; Krehbiel 1991). This is reg rettable not only because of the pivotal role of resolving differences for policy outcomes but also because of the significant attention scholars have paid related topics such as standing committees and bicameralism Resolving differences is a critical sta ge in the legislative process one worthy of scholarly attention By my count, of the non commemorative public laws enacted from 1977 to 2008 34% were originally passed by both chambers in disagreement. Of these 14 True disagreement occurs when conferees are unable to resolve certain bicameral differences. In these cases the chambers may opt to resolve these matters through the normal shuttling process. The n on controversial items are usually approved in normal fashion via a partial conference report. 15 If conferees exceed this a point of order.
30 laws, 16% were reconciled in a conference committee while 18% were reconciled by amendment trading. 16 And the frequency of bicameral reconciliation only rises when we consider landmark legislation (Smith 1989; Longley and Oleszek 1989). Indeed, resolving differences has determined the fate of ma ny laws of national significance Thus, if we are to understand topics such as legislative checks and balances congressional organization policy productivity and policy outcomes, we need to examine the critical stage of resolving differences. During the 1960s and 1970s, the pervasive question posed in research on conference committees was: Which chamber dominates the conference process? Or answered) by Steiner (1951) the House. Fifteen years later with the publication of The Power of the Purse (1966) a different answer was offered the Senate. More than a stylized fact, Fenno proposed a theoretical explanation for this pattern. First, he noted a type regularly asked for the higher of the two appropriations figures making it easier for the House to modify its position without sacrificing its preferred proposals. More importantly still, Fenno identi conclusion and theoretical propositions (Manley 1970 and Vogler 1970). Most significant in this line was additional explanation of Senate dominance. They maintained that chamber dominance 16 database. The data on all co nference committees and amendment trading in this period was compiled by the author.
31 is not a function of Senate leverage per se, rather the chamber to act second in the conference process (which is us A shift in the questions addressed by scholars writing on conference committee outcomes occurred under the auspices of neo institutional theory Two pioneers in this new direction Shepsle and Weingast (1987), reshaped our understanding of standing committee power. They successfully shifted the debate over the ubiquitous observation that congressional committees dominate the legislative process beyond the oft repeated vor a rational choice account. In doing this they placed conference committees at the fore of the strong committees thesis. Their bicameral sequences Specifically, since standing committee members are almost always named as managers in conference committee they possess the ability to either modify the policy back to its original committee approved position or simply defeat the proposal (the latter denoting the ex post vet proved fruitful during a legislative era marked by high committee autonomy and relatively weak congressional parties, subsequent researchers questioned the accuracy of their predictions under modern (a decentralized committee structure and strong parties) legislative conditions. Highlighting the departure from the strong committees thesis, Steven Smith (1989) explored challenges to conference committee autonomy by the parent chamber through devices such as the exc hanging of amendments in disagreement and the rej ection of conference agreements. Smith committee autono 232). Keith Krehbiel (1991)
32 proposed an additional rival to the strong committees thesis. He sought to identify whether the main features of U.S. legislative organization are primarily distributive where the policy process is marked by competition among legislators for a limited set of political goods or informational where legislator s pursuit of policy advance the book length research, Krehbiel dedicates a chapter to examining conference committee slation goes to conference? Who goes to conference? Does a well worn path exist on which legislation traverses after its 94). The answer to the first question is that specialized committee bills, not highly distributive bills, tend to go to conference. In regards to the third question, Krehbiel asserts, much like Smith (1989), that the ex post veto/committee dominance theory incorrectly assumes an equilibrium (routine) path through the conference stage. The answer to who goes to conference is particularly relevant for this dissertation. Krehbiel notes that, at first blush, the (216). That is, taking a di an informational lens and consider the role of expertise in conference committee ajoritarianism lurks (perhaps deeply) beneath the surface the 99 th Congress, Krehbiel demonstrates that after controlling for expertise the
33 untenable regarding conferee selection. Thus, the Speaker restrains his or her partisan impulses as managers for the House are not typically ideologues but policy experts likely to produce ma joritarian outcomes (244). With the rise of political polarization and partisanship, new theories have shifted the conventional wisdom about conference committee outcomes Collectively dubbed the 3, 2005; Dion and Huber 1996; Aldrich 1995; Aldrich and Rohde 2001; Rohde 1991), these perspectives posit th at the majority uses institutional mechanisms and procedures to skew legislative outcomes away from the median member of the chamber toward the majo preferred policy position Though primarily formulated for standing committees and floor activity, these theories are ubiquitous in political science scholarship and have recently been applied to con ference committee politics. The first attempt to identify the strategic calculus involved in selecting conferees theoretic approach. Using a revised version of Shepsle and strategic selection of conferees, the outcome of conference committees. In a related study, Carson and Vander Wielen (2002) find that as party polarization increases, se niority is less predictive of conference committee assignment while ideology increases in significance. Thus, they argue that by the mid 1990s seniority and ideology were equally important determinants of conference committee selection. Both studies sug gest that the selection of conferees has a strong ideological component.
34 Lazarus and Monroe (2007) take the next logical step. They propose that when committee members will produce a party damaging conference report, the Speaker can use his appointment power to select a prefera 4). Thus, lawmaking Using data fro m the 97 th through 106 th Congresses they demonstrate that under the conditions identified by their theory 17 the Speaker will name partisans as conferees in addition to committee members. This strategic maneuver, dubbed 18 maintains the norm of naming conferees that are familiar with the bill while presumably allowing the Speaker to put the party stamp on legislation. A recent, but as of yet unpublished, study by Vander Wielen and Smith (n.d.) finds that conference delegations are s imilarly biased in a pro majority direction in the Senate as well. However, while link between a biased conference delegation and partisan conference outcomes is certainly logical, no study has empirically verified it 19 C hapter 4 will test this linkage. 17 The first condition identified by Lazarus and Monroe is when there is a significant difference in the position of the jurisdictional committee and the Speaker. The seco nd condition is when 18 on 19 Nagler (1989) offers some descriptive evidence.
35 Scope and Method : Why Study Bicameral Disagreement and Reconciliation in the Postreform Era? Post Reform E Critical developments in rules, procedures and org anizational structures often delineate one era in Congressional his tory from another. Within the congressional literature, researchers identify the early to mid 1970s as a historical cut point demarcating two period s the pre 1968) and the ntemporary era (1975 present). C ongressional reforms straddling the textbook and post reform eras the culmination of efforts spanning more than two decades (Dodd and Oppenheimer 1997), undermined the seniority system and weakened the power of committee leaders enhanced the power of sub committees strengthened the parties leadership structures (particularly the House Speaker powers ), made party leaders more responsive to the rank and file, and expanded the Zelizer 2006 ; Dodd 1979; Cooper 2005; Deering an d Smith 1997; Stewart 2001). The most consequential effect of the reforms of the mid 1970s was the redistribution of organizational power away from c ommitt ees and committee chairmen and the centralization of power with in the parties and party leaders. Th us, following the reform period Congressional researchers identify an increase s capacity to manipulate and control the policy proces s (leading to increasingly pro majority policy outcomes). Indeed, the majority in the contemporary Congress controls consequential organizational decisions such as the naming of com mittee and subcommittee chairmen the appointment of committee members, co ntrol over the
36 legislative agenda, and (in the House) the ability to elect a Speaker who, in turn, names members to the powerful Rules Committee. With these developments in mind, one popular theory posits that political parties e cartel That is, parties derive their institutional authority by also Cox and McCubbins 1993 and 2007). In particular, the majority controls the consideration of poli cy known as negative agenda control through powers such as scheduling and amending. In this way the cartel model posits that the majority advances legislation to final passage when the proposal is preferred by its members to the status quo. This allows l awmakers within the majority to simultaneously vote their sincere preferences while fostering an electorally beneficial party record. Though sharing a number of empirical predictions, the theory of conditional party government (Aldrich 1995; Aldrich and Rohde 1997 ; Rohde 1991) emphasizes a fundamentally different causal mechanism underlying the parties organization al power in the modern Congress : intra party preference homogeneity and inter party preference divergence. 20 These two factors are, in fact, t and file delegates greater authority to the leadership which, in turn, exploits institutional rules and powers to ensure members of the party act in a w ay consistent with their collective goals. As With growing ideological polarization 20 Sinclair (1995) argues that the majority leadership frequently employs both strategies.
37 make the majority very unhappy [and in this circumstance] members of the majority would have a lot more incentive to empower their leaders to prevent a minority victory on legislation than in the former In short, the power of parties to produce non centrist policy outcomes varies according to the distribution preferences within the chambers. Scope T he postreform era is a theoretically interesting period with respect to bicameral disagreement and reconciliation for a few reasons. One is the simple fact that as pr eviously reviewed, these twin topics are severely understudied individually and have never been studied jointly. In a recent discussion of areas for needed resea rch, Rohde (2002, 34 7) is true of bicameral disagreement as well But b eing under studied does not make a topic worthy of systematic examination (though it certainly helps). Thus, the central reason why studying bicameral disagreement and reconciliation in the postreform period is theoretically interesting is that the curren t literature presents contradictory expectations with respect to the functioning of legislative checks and balances in this period. On the one hand, political parties been cited by political scientists for over a century as a solution to the policymaking inefficiencies built into the Constitution. From Woodrow Wilson (1885), E.E. Schattschneider (1942) and V.O. Key Jr. (1942) to more recent political scientists like Cutler (1988), Kernell (1991) and Sundquist (1988), the conventional view is that strong p arties bridge the gaps created by the Constitution and facilitate policy production and active government. In fact, in 1950, the American Political Science
38 o party system. Because parties were seen as ideological platform (thus encouraging active government and electorally responsible parties). M uch like conditional party government, one of the central themes of these studies was that polarization would spur strong parties. But while recent researchers have found empirical support for the claim that parties bridge legislative executive checks and balance s (Binder 1999, 2003; Conley 2002; Edwards and Barrett 2000; Cameron 2000; Alt and Lowry 1994; Cutler 1988; McCubbins 1991; Kelly 1993 ; Edwards Barrett and Peak 1997), no work has explored this process vis vis legislative chec ks and balances : at least n ot directly. Theoretically, however, we would expect the same effects. That is, as the parties have polarized and become more internally homogenous, and as legislative reforms over the postreform period have delegated greater control over congressional o rganization to the majority party, we would expect the majority to increasingly circumvent the hurdles established by Framers between the House and Senate and coordinate legislation across chambers As V.O. Key Jr. (1942) aptly put it: "the obstructions o f the governmental structure must be overcome, and it is the party, through extra constitutional expedients, that accomplishes However, a clear paradox exists, revealing the limits of our current understanding. A handful of researchers have cited too much polarization as the cause of stalemate and gridlock (Binder 2003; Dodd and Schraufnagel 2009). This is in contrast to the vocates who predicted the opposite. It is worth
39 quoti ng Binder (2003, 26) at length on this issue, as she is one of the few authors to identify this contradiction: The alternative hypothesis, if true, suggest s that polarization might be counterproductive to securing policy responsiveness of political parties turning the argument of the responsible party school on its head. Given the institutional structure of Congress and the separation of powers that distributes vetoes across the system, party polarization might ironically make major policy change less, ra ther than more, likely (26). Research by Dodd and Schraufnagel (2009) and Binder (2003) are the only attempts to reconcile this issue. Dodd and Schraufnagel (2009) propose that the relationship between polarization and policy activity is curvilinear. At the extremes high and low polarization governing institutions are increasingly stalemated whereas during eras of moderate polarization lawmakers are actively engaged in policy creation and change. The theoretical basis for this relationship is a belief t hat moderate conflict enhances policy negotiation and deliberation: both fundamental to p assing major policies ( Dahl 1967). Binder (2003) proposes a different mechanism and relationship. Citing work by Fiorina (2001), she maintains that the negative effe ct of polarization on policy activity is due to a decline in the number of moderates in Congress. With the disapp earance of moderates, the population of lawmakers who find it politically expedient to compromise have disappeared as well, making the task of legislating more difficult ( Binder 2003, 80). While I find both arguments convincing, ultimately this dissertation argues that a different causal mechanism drives the relationship between polarization and stalemate. The key insight is this dissertation that polarization has occurred asymmetrically across the House and Senate Indeed, researchers typically discuss legislative polarization in the aggregate without
40 considering inter chamber differences (or the consequences thereof ) T here are logical reasons to expect the cause and effect relationship between polarization and policy activity to be disjointed with in each chamber. In the House, with greater polarization and greater intra party homogenei ty, there should be less stalemate over the postreform era, not more significant advantages to a ideologically coherent majority. Th is hypothesized effect onal party government thesis and is consistent with the earlier view s advocated by the r esponsible party theorists (Committee on Political Parties 1950, E.E. Schattschneider 1942; V.O. Key Jr. 1942). I find some clear empirical support for this proposition. Figure 1 1 presents the percentage of House bills that were successfully cleared by the House from the 95 th to the 110 th Congress ( calculated as House bills passed / House bills introduced) These observations are publically available on the Congressional Bills Project webpage. 21 We can see an obvious positive trend over the postreform period (in fact, the percentage of House bills passing the House has doubled over this time). In the Senate, however, we would expect that polarization has caused either gr eater stalemate or gr e ater policy moderation. Because sixty votes are increasingly needed to break a filibuster in period of high polarization, the votes of moderates (Binder 2003) and/or legislative deliberation (Dodd and Schraufnagel 2009) are critical to securing passage of Senate bills. Figure 1 2 presents the percentage of Senate bill s that were successfully cleared by the Senate for the same time per iod ( calculated as Senate bills passed / Senate bills in troduced). We can 21 The author is grateful to E. Scott Adler and John Wilkerson for maki ng their database publically available. Any errors are my own. These data are available at: http://www.congressionalbills.org/
41 see no clear, statisticall y meaningful trend. 22 This is in contrast to the notion that polarization is a direct cause of stalemate. The prior empirics suggest that the link between stalemate and polarization is not due to developments that exist at the passage stage but historica l developments that exist at the post passage stage due to greater bicameral disagreement, greater challenges resolving differences or both. Polarization may not cause stalemate because of a lack of moderates or a lack of deliberation within each chamber p er se (though these are certainly indirect causes), rather, polarization creates an increase in bicameral disagreement which naturally increases stalemate. studying bicameral d isagreement and reconciliation over the postreform period (explicitly) is a worthwhile endeavor because we lack a full understanding of an essential question: How have increased polarization and the strengthening of parties over the last 30 years altered t he patterns of bicameral gridlock and policy conflict ? The literature established paradoxical relationships and expectations I propose that exploring bicameral disagreement and reconciliation exclusively (rather than looking at gridlock in the aggregate) provides an av enue to better understand this matter Method Finally, from a methodological perspective, though the reforms compromising the to mid 1970s, the 95 th Congress was selected as the starting point for t his research because of an important con gressional reform which began in full at the opening of the 95 th In January of 1975 (the 94 th 22 Though there is a negative trend shown in Figure 1 2, that trend is far from significant. The estimated linear trend coefficient is .0006 with an associated standard error of .001 (p=.64).
42 Congress) a change in House rules was proposed that would open conference committee proceedings to the public. 23 The rul e was subsequently adopted. Prior to this period, conference committees met in secret, without members of the media or lobbyists present. After some debate, the Senate adopted a similar rule in November of that year (Longley and Oleszek 1989, 53). Thus, the 95 th Congress was the first in which all conferences were open to the public. This reform was consistent with the so c period which were designed to open the legislative process to the p ublic (for example, an additional development was the creation of C SPAN). Authors have noted that this reform altered a crucial aspect of how the House and Senate resolve differences: their secrecy. Indeed, Longley and Oleszek (1989) suggest that conference committees following the ope n conference reform appear to be marked by greater electoral politics including more grandstanding, contentiou sness, and less efficiency ( Longley and Oleszek 1989, 50 61). In response, conferees have increasingly bargained informally before a conference c ommittee officially convenes or meet formally in a manner that discourage s outsiders from attending (Longley and Oleszek 1989). These critical developments in the operation of conference committees and their potential effects on the nature of resolving di fferences are reason to begin the analysis with the 95 th Congress Theoretical Structure and Empirical Expectations This manuscript can be thought of as containing two sections. Chapter 2 explores bicameral disagreement while Chapter 3, Chapter 4 and Chap ter 5 examine the process of resolving differences. Given this structur e, I find it beneficial to present the 23 See House Rule 28 (Longley and Oleszek 1989, 53).
43 theoretical discussion and empirical expectations in the relevant chapters. Here I provide a limited overview of the overarching theoretical ten ets and core empirical expectations guiding this project Bicameral Disagreement Since 1885, political scientists have claimed that parties bridge the constitutional gaps created by the Constitution. Thus, as Congress has experienced greater polariz atio n and as the parties have become organizationally stronger over the postreform period, we would logically expect a decrease in bicameral disagreement ( as the majority carefully control s the passage of legislation ) But contrary to these expectations, ther e have been a number of high profile bicameral disagreements since 1994 (even during unified government) Two recent examples include the Patient Protection an d Affordable Care Act, which was heavily moderated be cause of bicameral disputes and the Compre hensive Immigration Reform Act of 2006, which died in a House Senate conference. These cases among others provide the impetus for reconsidering the link between strong, polarized parties and bicameral conflict Certainly there are good institutio nal reas ons to suspect that the two chambers have polarized at different rates over postreform period and that this has caused a parallel increase in the frequency and severity of bicameral conflict The theoretical connection between Ho use and Senate differences in polarization and bicameral conflict draws on the logic of spatial models ubiquitous in studies of Congress, s eparation of powers and gridlock propose that the distribution of preferences within a political institution, typically arranged on a left right policy space, are critical to determining the performance of that
44 organization. In particular, I argue in Chap ter 2 that polarization asymmetry has created greater tension between pivotal House and Senate actors in three areas: (1) greater intra party bicameral differences, and (3) a growing gap between the House median and Senate filibuster pivot. Thus, contrary to popular exp ectations, the theory I outline suggests that the House and Senate have become increasingly dissimilar over postreform period in terms of the ideological location of pivotal memb ers leading to greater bicameral conflict. This theory and its expectations present a plausible solution to the paradoxical relationships between polarization, the strength of parties and policy outputs Resolving Differences The theor etical discussion and empirics presented in Chapter 2 regarding bicameral disagreement partly inform the remaining chapters on resolving differences. T he two stages are after all, inexorably linked. On the one hand bicameral disagreement has direct effects on the capaci ty of House and Senate negotiators to successfully merge competing versions of legislation. If the pre conference disagreements are severe, for example, there is a lower probability that negotiators will be able to find a workable solution. On the other hand the nature of bicameral disagreements can affect the precise outcomes reached by negotiators. As just one example, if the disagreements are major in scope and /or salience, irrespective of the specific political or institutional cleavages animating th e conflict, there may be a tendency for conferees to make greater concessions and compromises. On the latter point as the House and Senate have become increasingly dissimilar over the postreform period in terms of the spatial location of pivotal lawmaker s, it is plausible that resolving differences has become increasingly strained Rather than a smooth process
45 of resolving bicameral disputes, where minor policy solutions are able to satisfy competing actors, over the postreform period major changes are i ncreasingly needed to reconcile complex policy stalemates. I hypothesize that t his pattern most constrains the actor that wields the greatest unilateral control over the pre conference legislative process the House majority party In Chapter 3 I also prese nt a series of conceptual arguments regarding the process and pa tterns of re solving differences. The central question is best summarized One of the overarching premises is that resolving differen ces is a multidimensional process. By disagreements are intrinsically co mplex in terms of both the divisions among political actors and the policy issues involved. The na tural effect of complex, multidimensional disagreements is that it creates challenges to successfully mending competing versions of public policy. In addition to the challenges facing conferees due to multidimensionality, I argue that resolving difference s is a process marked by great uncertainty and significant t ransaction costs which, ultimately create an incentive for risk averse legislative decisions. In short, I argue that seeking compromise with political rivals at the conference stage is a utility maximizing choice (where at earlier stages in the policy process legislative failures are much less costly or politically damaging). Ultimately these various conceptual arguments regarding conference committees work hand in hand and make a single overarc hing prediction that resolving differences is, first and foremost, majoritarian in nature. That is, I argue that the primary principle structuring the operation of conference committees, and by extension the key
46 determinant of conference outcomes, is comp romise and concession (as opposed to non centrist, partisan outcomes or inter chamber conflict). C ontent of the Dissertation T he empirical chapters have a natural progression Chapter 2 explores the fic attention to the frequency and severity o f bicameral disagreement over the post reform period. Logically, the subsequent chapters examine attempt to resolve disagreements. Chapter 3 explores the ma cro level patterns in this p rocess while Chapter 4 examines the bill level patterns. The final empirical chapter Chapter 5 addresses a potential counterclaim to the argumentation of the prior three chapters providing further evidence in support of this d issertation s claims Chapter 2 is motivated by the overarching question: How have growing ideological polarization and the strengthening of parties in Congress affected inter chamber conflict ? Relying on an original dataset that records both the frequenc y and severity of House Senate disagreements Chapter 2 reveals that over the postreform period the two chambers have come into greater conflict when trying to pass legislation. In other er the last thirty years. Furthermore, the results show that pro majority or has become especially prone to bicameral gridlock. I find evidence that this effect is due to increasing distance s between the medians of eac h chamber as well as growing intra party disagreements that manifest across the House and Senate The congressional literature has been largely silent on these developments. Noting that polarization has occurred at a faster rate in the House than in the Senate, I attribute t he
47 increase in bicameral conflict to growing bicameral compositional asymmetries. These median member, (2) increasing intra party bicameral differences, an d (3) a growing gap between the House median and Senate filibuster pivot. C hapter 3 explores (at the macro level) how the House and Senate resolve differences when they arise referred to passage stage I begin this chapter by developing a unified typology of how the House and Senate might resolve differences, one that hopefully contributes to our conceptual understanding of this process. Using roll call data from all conference committees convened from the 95 th to the 110 th Congresses, I operationalize the typology using multidimensional spatial modeling. There are two primary findings in this section. First, I uncover evidence of multidimensionality in the process of resolving differences. In particular, I find that three qualitatively distinct dimensions reconciliation, partisan conflict and bicameral conflict explain over 80% of the variation in the conference committee roll call patterns. Second, I find that the first dimension of resolving differences (i.e. the one that explains th e greatest amount of variation) is a process of reconciliation, defined here as contrary to the direction of the literature over the past few decades, I do not find that partisanship is the leading determinant of post passage bargaining. This finding is particularly consequential when juxtaposed with findings abo ut the growing bicameral hurdle facing the majority party Where Chapter 3 examine s the aggregate patterns in post passage bargaining, Chapter 4 takes a look at individual conference committee ou tcomes. Following
48 Chapter 3 the first section estimates separate spatial models for the pre Republican Revolution and post Republic Revolution periods. The s e separate models show that the first and second dimensions reconciliation and partisan conflict respectively exhibit stability over time, indicating that even in the contemporary Congresses (where partisan roll call patterns are more pronounced) the proc ess of resolving differences remains a multidimensional process governed by consensus and compromise first and foremost. The second section explores the individual dimensions at the bill leve l using spatial mappings I find that over the postrefor m perio d there has been: (1) an increase in the variability of conference outcomes, (2) an increase in pro minority conference outcomes, an d (3) an increase in the extent of compromise and concession in conference. Overall, these findings suggest that over the p ostreform period the so when competing parties and chambers attempt to resolve differences and that this hurdle has increasingly constrained the majority party. Finally, t he third section of Chapte r 4 uses the multivariate spatial coordinates to examine the factors structuring conference outcomes. The main findings are as follows. First, greater pre conference disagreement necessitates greater reconci liation Second, partisan legislation is moder ated in a pro minority direction in conference. And at the same time, there is no evidence that the majority outcome. T hird, the more widely a bill passes one chamber relative to the other chamber the more the conference outcome shifts in the direction of the rival chamber. Taken as a whole, the findings from this section suggest that conference committees operate in a majoritarian rather than distributive fashion.
49 Chapter 5 concludes the empirical chapters by examining the process of resolv ing differences through formal models of agenda setting. Critics of the prior chapters may argue that the majority introduces and passes legislation in each chamber beyond their median ideal point s o that, in negotiations with the other chamber, they are able to concede a few minor provisions and ultimately emerge from conference with The central question of Chapter 5 is thus: How does resolving differences affect the capacity of parties to control the legislative agenda? I that captures bill le vel variation in agenda control Armed with th is data set competing formal models of agenda control are compare d. The results show that partisan models of agenda control perform best when we examine all final passage votes. However, when we restrict our analysis to bills that went to a conference committee or were shuttled between the chambers in disagreement, a non partisan, majoritarian model outperforms all rivals. This highlights how parties manage, successfully or unsuccessfully, varying institutional hurdles and shows that the majority party typically concedes some of its preferred policies when legislation goes to conference or is resolved via amendment trading. Chapter 6 concludes the dissertation. In the first part I findings in historical context. I note that while the instituti onal relationship has been rem arkably resilient throughout history despite some fears during the Constitutional Convention, it has evolved in consequential ways since 1787 Thus, thinking legislative checks and balances its historical variation and effects on the day to day operation of Congress is absolutely critical. In the second part of the conclusion
50 The narrative I present is that, while it is true that parties in Congress have enjoyed enhanced organ izational capacit ies over the postreform period the majority has simultaneously faced growing bicameral constraints because of an increase in House and Senate compositional differences. Third, I discuss the main implications of this research. One implic ation is that this dissertation offers a parsimonious solution to the paradoxical relationship between polarization, strong parties and policy productivity. Thus, as a practical matter, the present work helps us understand a fairly intuitive observation: while the majority is strong within the House and Senate, occasionally the across the chambers (even during unified government). At the same time, the findings force us to reconsider our conceptualization of and acknowledge the existence of a secondary process that has come to rival or limit the organizati onal strength of the majority. Finally, I discuss the normative undercurrents of this research The limited government and strong legislative checks and balances. The findings are troubled slug gish response to social and economic problems (due, in part, to bicameral constraints)
51 Figure 1 1. Percentage of H ouse bills cleared by the House Percentage calculated as total bills passed by the House divided by House bills introduced. Sourc e: Congressional Bills Project. Figure 1 2. Percentage of Senate bills cleared by the Senate Percentage calculated as total bills passed by the Senate divided by Senate bills introduced. Source: Congressional Bills Project.
52 CHAPTER 2 ESTIMATIN G THE BICAMERAL HURDLE Republicans in the 104 th Congress "have redefined success as getting legislation through one house." Barney Frank (D MA) 1 In January of 1995 the Republican Party assumed control of the House and S enate for the first time in fort y years. 2 permanent lican lawmakers faced an unfamiliar challenge: the task of governing Among the challenges was the fact that, despite their unity in the electoral arena, their victory in 1994 came w ith more salient intra party divisions Indeed, as Chapter 1 argues, one of the biggest hurdles facing Republicans at the opening of the 104 th Congress was how to manage their caucus across two constitutionally separate and distinct chambers. Of the 1994 midterm election, one congressional observer noted that while Gingrich (Hager 1994, 3226). In the waning days of 1994 Speaker elect Gingrich promis ed votes within the first 100 days on each item comprising the Republican Contract with America t he most ambitiou s policy agenda since t he Great Society. House Republicans followed through on Gingrich ten Contract ite ms before the end of April (the exception being a constitutional amendment imposing congressional term limits). But at the close of the fi rst 100 days, only two Contract bills had been enacted into law, with a surprising number having failed in the Republican contr olled Senate. Though inter branch 1 Quoted in Ornstein and Schenkenberg (1995, 203). 2 The 83 rd Congress (1953 1984) had R epublican majorities in both chambers.
53 lenses detailing conflict between Clinton and congressional Republicans few discuss the critical role of inter chamber bargaining 3 The mixed succe ss of the Contract with America raise s a series of theoretical questions. For example: What role do political parties play in our constitutional system of checks and balances ? This is an enduring question But in particular the limited success of the C ontract with America raises the question : How have the increases in ideological polarization and the organizational strength of parties the most notable Congressional developments in the last thirty years affected the functioning of legislative checks and balances? This chapter argues that the Contract with America and its mixed success can be explained by larger developments that occurred over the postrefor m per iod. The central thrust of this chapter is the finding that the House and Senate have become increasingly dissimilar over the postreform period in term s of their internal composition s (in particular, the ideological location of pivotal lawmakers in each chamber) I note three areas of increased bicameral dissimilarity: (1) greater distance betwee s median member, (2) increased intra party bicameral distance and (3) a greater gap between the House median and Senate filibuster pivot. The effect s of these developments ha ve been greater bicameral gridlock and more severe policy disagre ements between the two chambers. Moreover, I show that bicameral gridlock has become especially pronounced for pro House initiated legislation. P ara doxically, the findings show that Congressional polarization and the 3 Two notable exceptions are Binder (1999, 2003, 2005) and Riley (1995).
54 strengthening of parties have had a homeostatic effect on the capacity of the majority to pass its preferred policies as polarization has lead to a higher bicameral hurdle. Political Parties and Inter Branch Bargaining Political parties have been at the center of inter branch bargaining research since at least the 1940s. The conventional wisdom holds that parties bridge the space erected be tween Congress and the presidency attenuating the natural obstacles created by our constitutional frame work The academic constitutio nal structure can be traced to Woodrow Wilson ( Ran ney 1951 ). In Congressional Government (1885), Wilson lamented what he saw as the inefficiencies of the Constitution in particular its limitation on legislative powers favoring instead a Westminster form of government. According to Wilson, political parties possess a redeeming q uality in our flawed system as they help the executive and legislative branches bargain in c oncert. I n 1942, E. E. Schattschneider echoed this sentiment famously noting That same year, V.O. Key Jr. remarked that "the obstructions of the governmental structure must be overcome, and it is the party, through extra constitutional expedients, party system. Parties would be rendered he sense that, with clear inter party ideologica l differences, homogenous intra party preferences and unified governing control, new majorities would act decisively and with a popular mandate giving voters a clear visi on of the
55 failures In sum in the 1940s and 1950s there was surge in and stalemate Unfortunately for th ese writers sometime s dubbed divided party control would become a reoccurring condition. With persistent split party control, the scholarly discussion of parties and inter branch bargaining shifted. Four dec ades removed from the beginning of the responsib le parties movement, Sundquist (1988, 625) noted that the traditional party government thesis advocated by Schattschneider, V.O. Key Jr. and others of both branches o f government. Rereading the literature of the midcentury, one is In fact the swearing in of the 84 th Congress in 1955 marked the beginning of what one noted political scientist has called ided governmen American politics, political scientists and observers alike have lamented divided government as the cause of legislative gridlock ( Cutler 1988; Kernell 1991; Sundquist 1988). But then in 19 91, David Mayhew turned the conventional wisdom about the negative effects of divided government on its head. legislation over the post WW II era, Mayhew found that important legislation passes in approxima tely the same proportion during times of divided and unified government. example, Jones (2001) demonstrated that party polarization and party seat division increase the likelihoo d of gridlock independent of divided government (which has null
56 effects). At the same time, the Pivotal Politics model elaborated by Krehbiel (1996, The central question, ac cording to these authors, is: Why does gridlock occur just as often during unified party control? Th e answer according to Krehbiel and Brady and Volden is that models which examine party control through the lens of simple majoritarianism significantly und erestimate institutional factors that affect the size of legislative coalitions at final passage namely the supermajoritarian veto pivot and preference based (without an explicit role for legislative parties) Of course, consensus on the effects of divided government is nonexistent. If anything, the balance of evidence lies on the side of those who find that divided government is associated with a decrease in policy p roduction (though recent work by Ragusa ( 2010 ) finds symmetrical effects on the overall supp ly of legislation). Edwards and Barrett (2000) examine presidential initiatives and demonstrate that the president is able to initiate a larger percentage of his p referred policies during unified party control while Conley (200 2) shows us that the president is more likely to veto significant legislation during periods of spl it party control ( Cameron 2000). Others have found evidence that divided government leads to higher budget deficit s (Alt and Lowry 1994; Cutler, 1988; McCubbins, 1991). Kelly ( 1993 ), using data on legislation deemed landmark by observers at the time of enactment and years after the fact found that divided control limits the supply of landmark le gislation in direct contrast t finding. Edwards Barrett and Peak (1997) examine the consequences of divided government from the analysis of significant legislation that was
57 proposed but did not pass, they concluded that Mayhew conventional wisdom And Binder (1999, 2003), using a measure of gridlock that taps the total number of failed legislative policies divided by the size o f the policy agenda, finds that a larger percentage of her agenda items fail during periods of split party control. In summary, despite disagreement in the literature, the contemporary wisdom holds that political parties bridge our system of legislative executive checks and balances easing the bargaining process between Congress and the president and positively affect ing the overall supply of legislation. Political Parties and Inter C hamber Bargaining Little research exists on political parties and inter chamber bargaining Instead the vast majority of research on parties and legislative productivity has looked at constitutional tensions between Congress and the presidency. But as Binder (1999, House Senate differences, not simply legislati ve executive conflicts, have A leading study on parties and legislative checks and balances studies, Binder finds that the ideological distance between the House and Senate medians is a powerful determ inant of legislative gridlock over the post WW II era. In fact, in both her 1999 APSR article and her 2003 book Stalemate Binder reports that bicameral d istance has the most statistically meaningful effect in her gridlock model (greater in magnitude, for example, than divided party control). As she notes, this finding helps explain instances
58 of gridlock during unified government ( Krehbiel 1996, 1998 and B rady and Volden 1998). Of course Binder is not without critics. In two separate papers, Chiou and Rothenberg (2008a and 2008b) challenge cy productivity. In their initial paper (2008a), Chiou and Rothenberg note that Binder uses W NOMINATE scores to compare the ideological location of the House and Senate medians These scores, Chiou and Rothenberg note, are not comparable across cores which scale the House and observations Chiou and Rothenberg (2008a) find no evidence that bicameral distance positively affect s gridlock. Though Binder later (2008 ) con curs with Chiou and conference report votes confirms her ori ginal findings ( 2003, Appendix D). However, in a second paper Ch s decision to treat voice votes on conference reports as evidence of bicameral unanimity is seriously flawed. Theorizing About Bicameral Compositional Differences The Link Between Bicameral Distance and Gridlock It is necessary to begin our discussion by noting a conceptual disconnect between disconnect between these two factors is due to fact that the distance between the House and Senate medians leads to one of two mutually e xclusive outcomes: (1) a decrease in the likelihood that the two chambers will pass similar agenda items
59 bicameral gridlock is directly comparable to aggregate gridlock, in cases where the two chambers disagree over the content of legislation but subsequently engage in bicameral bargaining there can be a decrease in the likelihood of aggregat e gridlock. Failing to account for this distinction may have contributed to the mixed results observed by Binder and Chiou and Rothenberg. The following inf ormal spatial models illustrate these two effects using the 99 th and 110 th Congresses as an exampl e Assume we have a game played by three actors the House (H), Senate (S) and president (P) who have single peaked Euclidian ideal points. Further assume for simplicity that the policy space is unidimensional and that the two chambers have majoritarian voting rules. Figure 2 1 and 2 2 present the estimated spatial location of the House, Senate and president in the 99 th and 100 th Congresses (respectively) using th Congress, because rival parties c ontrolled the House and Senate, the distance between the two chambers in comparison to the 100 th Congress. Thus, in the first stage of the game when one chamber passes a bill and submits it to the second cha mber for consideration, there is a greater likelihood that that legislation in the 99 th Congress will experience bicameral gridlock compared to the 100 th Congress. That is to say in the 99 th Congress there is a greater chance that the Senate will reject b ills passed by the House (and vice versa). In the second stage of the game, if the chambers attempt to resolve differences and formally bargain over the content of the
60 ) in the 99 th Congress as compared to the 100 th Congress. If we assume that this bargaining process yields a bill located at the midpoint of the House and Senate positions (denoted M for ), then th e likelihood of a presidential veto is given by the distance between the president and that th Congress though the likelihood of bicameral gridlock is high compared to the 100 th the likeliho od of a veto is dramatically larger after bicameral bargaining compared to the 100 th Congress. The previous discussion highlights the following. On the one hand increases in bicameral distance will cause bicameral gridlock if the ideological distance be tween the chambers creates irreconcilable differences. However, if the two chambers engage in inter chamber bargaining, a moderated bill might decrease the likelihood of a presidential veto Regarding the link between overall gridlock and bicameral dista nce, the connection between these factors capacity for resolving disagreements. Thus, the true effect of bicameral distance manifests in two mutually exclusive ways: (1) a decrease in the likelihood that the chambers will pass ( an increase in ) and (2) an increase in policy disagreement between each chamber ( an increase in ) In short, it is important to examine these two causal paths. T he House and Senate Media ns in the Postreform Era Despite her excellent work on the topic, Binder provides only a limited discussion of historical patterns in the distance between the House and Senate. This is not a primary focus was on the factors that cause gridlock (bicameral distance was simply one of many factor s ). At the same time, her analysis ended with
61 the 106 th Congress in 2001. The eight years elps reveal the forthcoming trends So the question is: Are there any meani ngful trends in the distance between the House and Senate medians over the postreform era? As it turns out, there are Figure 2 3 presents the distance between the House and Senate medians from the 95 th to the 110 th Congresses common space scores 4 From the Figure it is clear that when rival parties control the House and Senate ( the 97 th 98 th 99 th and 107 th Congresses) the distance between the two chambers is rather large. This is to be expected divided govern ment However, despite the fact that thr ee out of the four episodes of quasi divided government occurred in the first half of the postreform period, there is a visible positive trend in bicameral distance. If we remove these periods from the Figure for a moment given that we can safely account for the spikes in the time the temporal trend is stark ( Figure 2 4) The linear fit in Figure 2 4 is given by: | HM SM | =0.73+.0074(Cong). The slope is statistically significant at .05 level with only twelve ob servations. Based on these estimates, we can conclude that over the post reform period the distance between the medians of the two chambers has increased by .12 in the first dimension. To put this estimate into perspective, the mean bicameral distance sc ore for the pre 104 th Congresses (before the Republican Revolution in 1994) was only .013 while in the post Republican Revolution era ( the 104 th to 110 th ) the mean bicameral di stance score is .075 (a nearly six fold increase). To add some context to this development a difference of about .07 in the first dimension is equivalent to the 4 ideological distance in the first dimension between the House and Senate medians using
62 ideological distance that separates Senators Joe Lieberman (D CT) and Harry Reid (D NV). Thus, and perhaps surprisingly, over the postreform period I find that as the parti es have polarized and become more internally homogenous within their respective chambers, the two chambers have become increasingly d issimilar Polarization Asymmetry Before analyzing the effects of House and Senate ideological distance on bicameral gridl ock and disagreement, the pertinent question to ask is: Why has the distance between the House and Senate medians increased so dramatically over the postreform period? A clear development such as this must have an equally apparent causal mechanism. In th is section I will argue the positive trend in bicameral distance can be connected to two major developments over the postreform period which have operated in concert : (1) electoral and geographic polarization and (2) Congressional reforms. I argue that bo th developments have spawned greater polarization i n the House than the Senate, leading to greater House and Senate compositional differences. controversial, their belief that the House and Senate would function differently cannot be argued For sion to allot each state bicameral legislature before addressing the question of apportionment ( Wood 1998, 553). Nonetheless, for the present purposes at least, we can safel y discuss the
63 thoughts on the effects of bicameral design (irrespective of debates about their rational intentions or overarching goals). For instance, Madison argued in Federalist No. 62 genius of the to which Madison alluded was thought to derive from House and Senate representational and deliberative differences. Contemporaries have echoed this sent iment, noting that one of the merits of a bicameral legislature is that it limits what the F Money 1997). Regarding representational differences, the tensions between equal state represent ation and popular representation (i.e. federal bicameralism) leaves each chamber beholden to fundamentally different constituencies. Though no issue pits per se (Lee and Oppenheimer 1997), tensions between the chambers can arise when factors that correlate with state size like geography enhance the scope of policy conflict. Immigration is one perennial example. 5 further differences in represent ation between the chambers with the House focu sed on short term, or popular issues and the Senate alert to long (cf. Federalist nos. 62, 63 and 51). For example, Madison in Federalist no. 51 noted divide the legislature into different branches; and to 5 For example, in the 98 th Congress conferees met 10 times attempting to reconcile between immigration reform bills S 529 and HR 1510. The parallel bills represented the second attempt in as many congressional sessions to enact sweeping immigration reform. Ultimately, the main factor which killed the bill in conference was a disagreement about how to reimburse border states for the increasing financial obligations created by the surge in new citizens under the legalization provision. Conference committee members from states with large illegal alien populations believed the billion dollar block grant to offset social service costs stemming from legalization was too low ( Cohodas 1984)
64 render them, by different modes of election and different principles of action, as little connected with each other as the nature of their common functions and their common dependence on the And finally, regarding deliberative differences between the House and Senate, it was believed during the founding that the Senate, because of greater policy expertise, longer terms and historical precedent, would review legislation pa ssed by the House on the merits, delaying the passage of defective policy (cf. Federalist nos. 62 63 ;). 6 and compositional tensions between the House and Senate. This is not a nov el claim However, if we juxtapose this rather simple observation with what is the do minant and most dramatic development of the postreform period greater ideological polariza tion in Congress we arrive at a powerful hypothesis Based on what we know abou t House and Senate representation we would logically expect the House and Senate to polarize very differently over time The difference is not qualitative but of degree; we would expect the House to polarize at a greater rate than the Senate. This is du e to the simple fact that senators typically represent more diverse constituencies and are constitutionally insulated from exogenous changes while representatives typically represent more homogenous constituencies and are more responsive to exogenous facto rs. It is helpful to briefly review the sources of congressional polarization. 7 6 In Fede ralist no. 62 7 Because of space constraints a few sources of polarization are not reviewed here. These includ e lingering political and policy disputes (Dodd 1991; Uslaner 2000) and the recasting of political issues (Sundquist 1983; Adams 1997; Leege et al. 2002; Green, Palmquist and Schickler 1996; McCarty, Poole and Rosenthal 2006; Carmines and Stimson 1989).
65 The most frequently cited explanation of polarization in the House and Senate is electoral and geographic realignment ( Stonecash, Brewer and Mariani 2003). For example, it ha s been argued that changing migration patterns aided southern Republicans electorally while the resulting political realignment splintered the southern Democratic Party (Jacobson 2000; Polsby 2004; Rohde 1991). These regional characterized by a more homogenously liberal Democratic party and reduced electoral competition for southern Republicans. More recent work has linked polarization throughout the United States with more general migratory patterns. Scholars have begun to advocate that population mobility the increased ability of Americans to relocate geographically causes individuals with similar ideologies to cluster referred to sometimes as partisan or ideolo ( Bishop and Cushing 2008 ; Gimpel and Schuknecht 2003). According to this view, preferences for particular geographic locations and places correlate with political, economic and religious preferences. S ome authors cite gerrymandering as an additional ele ctoral cause of greater polarizati on in Congress (b ut there are significant disagreements in the literature ) Abramowitz et al. (2006) find no significant effect of gerrymandering in period s following redistricting while Oppenheimer (2005) notes that para llel increases in polarization have occurred in the U.S. Senate. Ho wever, a recent study by Carson et al. (2007) reports that when districts undergo significant changes to their boundaries, there is a corresponding increase in polarization. These authors are careful to note, however, that this modest effect is but one of multiple causes of increased congressional polarization.
66 Among these electoral and geographic developments there are clear reasons to suspect that House districts have polarized more th an states. Since population mobility almost certainly declines as geographic distance increases, intra state mobility will create homogenous House districts while having absolutely no effects on the overall state composition. Logically, the only way elec toral polarization can occur equally in the House and Senate is if migrators are just as likely to relocate across states lines as they are within state lines. At the same time, to the extent that gerrymandering af fects polarization the effects can only m anifest in the House as state lines are fixed while district lines are not. Overall, then, to the extent that electoral and geographic trends affect polarization which many authors cite as the leading cause s that effect has almost certainly polarized Hou se districts more than the states A number of prominent studies add institutional reforms to th e determinants of postreform er a Congressional polarization. Ultimately, though these reforms are not a function of constitutional design per se 8 the effects of these reforms have contributed to polarization asymmetry in the same manner as the electoral geographic causes. Legislative reforms in the 1970s, for example, simultaneously decentralized Congress weakening the power of committee chairmen and enhanced the formal power of the majority. These reforms, however, were uneven across the House and Senate. For example, the House reforms included the revival of the Steering and Policy Committee (including the Speaker as its chair), the expansion of the whip sy stem and the empowerment of the Speaker to name members to the influential Rules Committee ( Davidson 1981 ; Deering and Smith 1997; Dodd 1979; Dodd and Schott 1979; Rohde 8 This is so because the U.S. Constitution allows the two chambers to fashion their own rules.
67 1991; Sinclair 1988, 1989; Shepsle 1989; Zelizer 2006). Though the Senate made major changes to its committee system in this period in particular Senate reformers cut the number of committees and sub committees and reorganized committee jurisdictional borders no conseq party or camber procedures (Da vidson 1981; Deering and Smith 1997). A similar narrative is applicable to the reforms implemented after the 1994 congressional mid term. Though the Senate reforms in this period made the upper chamber more like the House (particularly with the adoption of a secret ballot to name committee chairs), the implemented reforms were less consequential in terms of strengthening the two parties ( Davidson 1981; Deering and Smith 1997; Zelizer 2006). Term limits, designed to enhance the responsiveness of committe es and leaders to the rank and file, were less stringent in the Senate compared to the House. 9 While the House cut the number of committee staff by one third, the Senate made no corresponding cuts. And finally, while the House leadership was given enhanc ed power over the sel ection of committee chairmen no such power s were afford the Senate leadership. Ultimately, of the Senate reforms implemented during the 104 th Congress, elimi Figures 2 5 and 2 6 explore whether the previous hypothesis polarization asymmetry is supported empirically Figure 2 5 charts polari zation in the Senate and House (separately) over the postreform period using the distance between the median 9 For example, the six year term limit for committee chairmen in both chambers included subcommittee chairmen in the House. Also, where the house included an e ight year limit for the Speaker, no formal limits were adopted for the Senate majority leader.
68 Democrat and median R epublican in each chamber Higher value s indicate greater polarization For these estimate s I am again using scores because they are comparable a cross chambers Parallel trends in this Figure are evidence against polarization asymmetry. Figure 2 6 reports the difference in each er values indicate the House is more polarized while lower values indicate th e is more polarized A flat trend in this Figure is evidence against polarization asymmetry. The trends in both figures are unambiguous and support the earlier discussion : since the 95 th Congress the House has polariz ed more rapidly than the Senate Fo r example, at the start of the series in the 95 th Congress we can see that the Senate was slightly more polarized than the House. Th e difference was small however (about .05 in the first dimension). In the 96 th Congress, the House and Senate were about e qually polarized according to both figures But by the 110 th Congress, we can see that the House had become much more polarized than the Senate. In the House the average distance between the median Democrat and median Republican was .84 in the first dime nsion while the same distance in the Senate was .73. Thus, in the most recent Congresses the House is about 15% more polarized when compared to the Senate The next question we must address is : What does it matter if the House and Senate have polarized at different rates ? I hypothesize that bicameral variation in polarization and the resulting compositional differences between the House and Senate has increased the frequency (bicameral gridlock) and severity (bicameral disagreemen t) of bicameral policy di sputes. The theoretical connection between differences in polarization and bicameral conflict draws on the logic of spatial theory and the model of
69 P ivotal Such models ubiquito us in studies of Congress, separation of powers and gridlock propose that the distribution of preferences within a political institution, typically arranged on a left right policy space, are critical to determining the performance of that organization. Sp ecifically, each player has a fixed, Euclidian preference in a unidimensional policy space where the spatial location of each player determines organizationa l outputs. Without the support of a few organizational outputs exist within a gridlock interval where preservation of the status Of course, the spatial location s of pivotal actors are not static: quite the contrary. Thus, if polarization asymmetry describes patterns in the distribution of preferences across the House and Senate, we would logically expect this effect to increase the distance between pivotal actors. For example, s ince the chamber median in a two cont inuum, as the two parties polarize the location of the median will shift toward the extremes as well Of course, our conceptualization of s represent an abstraction. Lawmakers rarely consult the chamber median and draft legislation to his or her liking. Thus, there may be pivotal actors that we should consider in addition to the House and Senate medians As I argue in the following section, differences in the rate of polarization across the House and Senate can be theoretically lin ked to two additional types of bicameral compositional tensions And t h ough the theoretical discussion borrows from the general intuition of pivotal politics, unlike many gridlock models my argumentation is pu rely empirical rather than mathematical But like the formal theories
70 on whic h this research is founded, the theory begins with some basic assumptions and uses logical deduction to generate a set of testable hypotheses. Two Additional Bicameral Consequences of Polarization Asymmetry Intra party Bic ameral Distance For the same reasons that polarization asymmetry affects the distance between the medians across chambers, we would also expect an increase in the bicameral distance within the two parties. istance But though academics and observers regularly note that Democrats in the House often disagree with Democrats in the Senate (and vice versa for Republicans), no work has built this simple observation into a single metric. However, as the discussion in the introductory section illustrates, landmark policies such as those items comprising the Contract With America can be greatly affected by exactly this source of bicameral tension. It is worthwhile, therefore, to systematic ally examine this phenomenon. Figure 2 7 presents this factor ( Intra party Bicameral Distance ) over the post median member in the House and their median member in the Senate. Higher values indicate greater within party bicameral distance. From Figure 2 7, in the pre Republican Revolution era (95 th to 103 rd Congresses) the mean median Senator and their median Representative is .03 where in the post Revolution era Congresses (104 th to 110 th ) the mean intra party bicameral distance is .10 (a more than threefold increase). Recall from earlier that a difference of about .07 in the first dimension is about the distance between Joe Lieberman (D CT) and Harry Reid (D NV);
71 so the change in within party bicameral distance over the postreform period is greater than that. It is worth pointing out two notable qualit ies of this measure. First, o ne of the most con tentious debates in the congressional literature surrounds the conceptual distinct ion between partisan effects and preference effects. The two sides rarely see eye to eye. th fellow party members in spite of their disagreement about the policy in question, or do they vote with fellow party members because of their agreement about the policy in mphasis in original). Both bicameral distance (Binder 1999, 200 3) and the House filibuster pivot distance (forthcoming) are non partisan in nature, focusing instead on the role of raw preferences. In contrast, intra party bicameral distance melds the effects of parties and preferences into one measure. Indeed, thoug h the measure is opera tionalized using preferences, it incorporates the pivotal role that parties play in the contemporary C ongress. In fact, we can look to partisan theories of lawmaking to understand why differences in polarization might cause House Sen ate policy conflict. According to the theory of conditional party government, polarization caused the growth in al capacity to produce non centrist outcomes. It follows that if one chamber exhibits greater power in this regard (t he House), it is likely that the majority leaderships of both chambers will advocate and advance different proposals. This possibility is often overlooked. As Smith and Gamm standing differences between the nature o f party
72 A second notable quality of this measure is that it offers a within parties explanation for the formalize given the history and importance of intra party divisions For example, in 1910 the organizational power of the against Joe Cannon, where progressive Republicans joined Democrats in enacting a series of reforms that sever (Baker 1973; Rohde 1991; Riley 1995). Intra party divisions were also critical to the passage of the 17 th amendment. Wirls (1999) demonstrates that contrary to the conventional wisdom, regional intra party divisions were the key factor structuring roll call voting on the direct election of senators. As a third example, n umerous authors have noted that Democratic intra party divisions during the 1960s stymied efforts to enact civil rights reform. James MacGregor Burns (19 63), for example, famously argued that deadlock was due to systematic differences between northern and southern Democrats as well as bet ween Northeastern and Midwestern Republicans. And f inally, intra party divisions also structured in part policy conflic t over the Contract With America. Despite the narrative of Republican unity and decisive action Ornstein and Schenkenberg (1995) point out the inherent tension between Republicans who wanted to balance the budget and those who wanted to dismantle governm On the surface, the two themes are mutually reinforcing. In reality, they are not. I ndeed, there is enough tension between them that they may be close to mutually exclusive (Ornstein and Schenkenberg 1995, 198). House Filibuster Pi vot Distance On the one hand, both bicameral distance and intra party bicameral distance are majoritarian in nature. That is, they fail to take into consideration differences in the
73 institutional rules governing the House and Senate. Where the House is a majoritarian institution because a simple majority is necessary and sufficient to pass legislation the Senate is (often) a supermajoritarian institution where, because of unlimited debate, 60 votes are frequently needed to pass legislation. A comprehens ive examination of bicameral differences must account for the distinction in the rules governing the House and Senate. Accordingly, a systematic examination of bicameral compositional differences must account for the pivotal role exercised by the supermaj oritarian Senate (Krehbiel 1996, 1998; Brady and Volden 1998 2006). In particular, we would expect polarization asymmetry to have increased the consequences of the filibuster pivot on House and Senate negotiations. As B rady and Volden (2006, 38) show s trong parties actually stretch the gridlock interval as legislation located at the party median in the Senate is more likely to face a filibuster from the minority. However, where most researchers model the filibuster pivot as an intra cameral measure usi ng the distance between the Senate median and the filibuster pivot here I am interested in the distance between the House median and the Senate filibuster pivot. Indeed, because a simple majority on the House floor is need to concur with, amend or reject legislation passed by the Senate examining variation in the location of the filibuster pivot vis vis the House median helps us assess the effect on bicameral bargaining Figure 2 8 presents this data over the postreform period using the absolute distance between the House median and Senate filibuster pivot (the 40 th or 60 th Senator, depending on party control, whose vote is pivotal for enacting cloture) using ilibuster pivot is almost always on the
74 some leverage. However, the contribution of conceptualizing the filibuster pivot in this manner is that, as we can see Figure 2 8, as the two parties have polarized the size of the interval between the House median and the filibuster pivot has grown. This is because both the filibuster pivot and the House median have moved toward the extremes over the postreform period. Certainly, the minority is filibustering to a greater extent today than in years past. As with the previous two measures of bicameral differences, as the parties have become more homogenous and distinct presumably a source of greater party power and pro majority policy outcomes they have had to contend with a greater bicameral hurdle that might constrain their power. In this particular case, though the House majority can easily pass policies at its discretion, those policies are increasingly likely to encounter bicamer al gridlock and disagreement given the increasing distance of the filibuster pivot. Methodology: Estimating the Bicameral Hurdle The previous discussion proposed two qualitatively different, but related, forms of bill that dies in the other chamber. Whether the bill passed by the first chamber was chamber floor, the proposal was not preferable to the s tatus quo in the second chamber. On the other chamber counters by passing a different bill. Whether the counterproposal was an entirely different bill or a slightly modifi proposal was not preferable to the status quo in the second chamber without
75 some modification(s). Though both measures tap bicameral policy conflict, bicameral gridlock is a latent measure where we know conflict exists though that we cannot systematically observe its nature while bicameral disagreement is a formal measure where, because each chamber made a tangible proposal, we can systematically observe differences in the policy content of each bil l. To assess bicameral gridlock I coded data on every bill passed by the two chambers over the post reform period (95 th Congress to the 110 th Congress) The primary observation is whether the second chamber in the policy sequence passed the initiating ch 0 if fail). Fortunately, these observations are publically available on the Congressional Bills Project webpage and required only minimal additional coding 10 To assess bicam eral disagreement, I created an original database that identifies every attempt by the House and Senate to resolve differences. every attempt to reconcile House and Senate policy disputes by convening a conference committee or engaging i n amendment trading. 11 final passage vote, I record ed the midpoint of both proposals dimension common space scores. The midpoint of a roll call vote is the position in the liberal to conservative dimension that separat estimating the differences in the 10 The author is grateful to E. Scott Adler and John Wilkerson for making their database publically available. Any errors are my own. These data are available a t: http://www.congressionalbills.org/ 11 Thomas.com website. The data on conference committees was checked against data provided by Michael C. Brady. All errors are my own.
76 ideological differences in the policy content of the two bills. Though mutually exclusive, t he two measures are rel ated. In fact, breaking bicameral gridlock is a precondition for observing bicameral disagreement. This creates a selection effect by definition This effect affects the observation of bicameral disagreement in two ways (bicameral gridlock is unaffected) Firs t, it may be that a majority in the Senate agrees with a proposal passed by the House but, because a minority of senators are ab le to sustain a filibuster, the Senate is unable to pass the House bill. This effect would artificially increase the amount of disagr eement and cause the forthcoming models to overestimate the s alience of the filibuster pivot. But at the same time, a second selection effect removes the most salient disagreements from the analysis This occurs when one chamber chooses inaction as a mea ns of killing the ( a common occurrence ) That is, rather than waste time drafting and passing legislation destined to fail the second chamber ignores the bill or issue This eff ect would decrease the actual amount of disagreement. Bicameral Conflict in the Postreform Era As a first look at bicameral conflict I present descriptive statistics of bicameral gridlock and bicameral disagreement over the postreform period. Figure 2 9 c harts bicameral gridlock for each Congress during this period calculated as the inverse percentage of House and Senate passed bills that cleared the second acting chamber Figures 2 10 and 2 11 by comparison, chart the mean bicameral disagreement score
77 per Congress calculated as average first dimension midpoint distance between bills passed by House and Senate in disagreement For bicameral gridlock, t he raw data reveal that 53.4% of bills passed by the initiating chamber were successfully passed by t he other chamber. Thus, 46.6% of bills passed by the initiating chamber died in the second chamber without passage (the postreform rate of bicameral gridlock) In the aggregate, this estimate is surprisingly rall gridlock estimate ( Table 3 2). For example, passage. Moreover, she reports that 49.0% of her agenda items were passed by the House but died in the Senate while 43.7% were passed by the Senate but died in the House. According to the bicameral gridlock data reported in Figure 2 9 the comparable statistics are 46.1% (died in the Senate) and 47.7% (died in the House). As Binder notes, this small difference in the passage rate for Hou se and Senate agenda items ts the Binder 2008, 15 17). Indeed, both her agenda measure and my data show that the Senate is no more to blame for policy inac tion than the House. The more notable trend in Figure 2 9 is the linear increase in bicameral gridlock over the postreform period (representing a decrease in the percentage of House and Senate initiated bills that passed the other chamber ) The estimate d linear trend line is given by: Pr(B i C j Gridlock ) = 0.57 .010(Cong). Thus, from the 95 th to 110 th Congresses the percentage of House or Senate passed bill s gridlocked by the second chamber increase d 1% per Congress (15% over the entire time series). If we use the average number of bills passed per Congress as a benchmark a decline of 15% over the
78 postreform period accounts for 95 fewer House passed bills passing in the Senate per Congress and 47 fewer Senate passed bills passing in the House per Congre ss. This is certainly a consequential development Another interesting feature visible in the time series is the relatively minor (or perhaps non existent) decrease in the probability of bill passage during periods of congressional split party control. A divided with a growth in bicameral gridlock. But, on the other hand formal models of inter chamber bargaining suggest that during congressional split party control there is greater strat (1988) game theoretic model shows us that during periods of quasi divided government state houses introduce less enacted legislation relative to unified party control. And form al model shows us that party leaders strategically initiate prior work and the results here, it seems that party leaders in rival chambers use their agenda control powers to limit the introduction and passage of bills that have little chance of securing a majority in the other house. Figure s 2 10 and 2 11 present data on the average per Congress House and Senate ideological disagreement. Figure 2 10 presents the entire time serie s As we can see from Figure 2 10, there is clearly greater bicameral disagreement during periods where different pa rties occupied the two chambers. As we would expect, w hen rival parties control the House and Senate they tend to pass ideologic ally different proposals. However, consistent with the results for bicameral gridlock, Figure 2 11 shows a linear increase over time in bicameral disagreement once we ( crudely ) control
79 for periods of quasi divided government (by removing them from the Fig ure ) The estimated linear trend line in Figure 2 11 is given by: | HB i Mid SB i Mid | =.1813 .0048(Cong). Thus, over the postreform period the average ideological disagreement between the two chambers has increased by .072 in the first dimension. As disc ussed previously, a difference of .07 in the first dimension is about the distance between Joe Lieberman (D CT) and Harry Reid (D NV). The increases over the postreform period in both bicameral gridlock and bicameral disagreement are trends that have gon e largely unreported in the congressional literature This is surprising given that these are fairly clear trends and given the potential consequences of these developments. T he only reference to similar trends which I am aware is discussion by Binder of a parallel pattern in her gridlock data. Table 2 1 reported here reproduces for the 95 th to 110 th Table 3 3 cannot simply assume that the two chambers hold similar sets of views on major policy data for bicamer al gridlock and bic ameral disagreement support descriptive statistics while extending the time series beyond the 106 th Congress. It is worth repeating, however, that the present research focuses on these trends in greater theoretical and analytic al depth. Ultimately, my results shou ld be interpreted as extending Finally, comparing patterns in both figures for 106 th Congress might raise a few questions. Given the general positive relationship between bicameral gr id lock and
80 bicameral disagreement, it is perhaps surprising that in the 106 th Cong ress both trend lines spike in opposite directions. Curiously, this indicates that during the 106 th Congress there was simultaneously a higher second chamber passage rate (l ower bicameral gridlock) and greater ideological disagreements between the two chambers. This pattern highlights the fact that bot h constructs though similar are different. On the one hand, Republicans were only moderately successful in terms of clear ing legislation during the 106 th Congress, though they enacted a number of important bills (Mayhew 2005). The successes included codification of permanent normal trade relations with China (HR 4444), major banking and financial reform (S 900), enactment o f the Ed Flex Program (HR 800), legislation closing the loophole on 527 PACs (HR 4762), enactment of the Y2K Dispute Resolution bill (HR 775). But, on the other hand, there were a nu mber o f important failures While a number of these failures stem from Clinton vetoes 12 there were also a number of high profile failures due to inter chamber disputes. For example, along strict party lines, the House passed the Patients' Bill of Rights Plus Ac though also partisan was much more moderate. Trying to merge the different bills was described by ). Ultimately, Republican s were unable to find a workable solution within their party and the two bills died in conference. Cases such as these contribute to bicameral disagreement but not bicameral gridlock. On the flip side, bills passed by both 12 1122)
81 chambers and enacted into law c ontained significant initial disagreements. The most historically important law enacted by the 106 th Congress is a clear example: the Leach Steagall (Parks 1999) ct party lines, the House bill (HR 10) passed with wide, bipartisan support. 13 The substantive differences between the two bills are reflected in midpoints for each final passage vote. The midpoint for the much more moderate House bill was .504 while the midpoint for the partisan Senate bill was .067. It turns out that this difference (.437) is near the mean for the 106 th Congress (.468). The overall point is that during the 106 th Congress there were many policy failures and disagreement that occurred after each chamber passed a bill. And at the same time a number of bills enacted into law contained contentious inter chamber differences prior to bicameral reconciliation This combi nation can lead to lower bicameral gridlock and higher bicameral disag reement. Thus, it is important to distinguish between bicameral gridlock and bicameral disagreement as well as take account of the myriad ways in which bicameral differences can affect policy outcomes. Independent Variables Having found some consequential developments in bicameral conflict over the postreform period and argued for a causal link between the growth in bicameral conflict and increasing House and Senate compositional asymmetries I now turn to modeling the determinants of bicameral gridlock a nd bicameral disagreement. Given the proximity of this project to prior work on policy productivity I adopt 13 S 900 passed 5/6/1999 by a mar gin of 54 44 with only one Democrat joining every Republican in support (roll call 105). HR 10 passed 7/1/1999 by a margin of 343 86 with 138 Democrats joining 205 Republicans in support (roll call 276).
82 gridlock model as a reference (but see also Brady and Volden 1998, 2006; Chiou and Rothenberg 2003; Krehbiel 1996, 1998; Mayhew 19 991). But because the present research differs from a traditional gridlock model, not all factors are the same or have the same expectations. The variable Divided Government Senate were controlled by the presiden three bodies was unified. On the one hand, because this dissertation examines the passage and post passage stages it is expected that the is somewhat limi ted (at least in comparison to traditio nal models of gridlock ) Nonetheless, because lawmakers are strategic actors, pre empt a presidential veto and moderate their policy proposal s prior to passage. If this form of strategic behavior is pervasive we would expect a decrease in both bicameral gridlock and bicameral disagreement during divided government. The assumption is that moderate policy is more likely to secure passage and is less ideologically divisive Budget measures the size of the fede dget surplus as a percentage of federal government outlays. 14 The expectation is that budgetary surpluses will decrease the prevalence of both bicameral gridlo ck and bicameral disagreement as lawmakers have an easier time passing new laws with fewer financ ial constraints 14 These data are available on the Office of Managem http://www.whitehouse.gov/omb/budget/Historicals
83 The variable Party Mandate records the number of prior Congresses a new majority was in the minority averaged for both chambers. 15 When a new majority comes into power with a popular mandate, the expectation is that they will act decisiv ely and break the conditions of gridlock The Republican Revolu tion is a classic example. In the present application the expectation is that party mandates are associated will less bicameral gridlock and less bicameral disagreement. The variable Mood Lag c mood data lagged one Congress or conservative policies. Higher values indicate a more liberal public and a preference for active government while lower values indicate a more conservative public and a preference for less active government T he expectation is that periods when the public prefers a more active federal government are associated with less bicameral gridlock. During conservative eras, by contrast, the expectation is that there will be an increase in bicameral gridlock. I also expect t hat periods where the public prefers an active federal government will be associated with less bicameral disagreement. model, some important differences must be accounted for given the uniqueness of the forthcoming models The first, as and operationalize Bicameral Distance (1998) common space scores. Moderation variable from the forthcoming models Since 15 For example, the value for the 104 th Congress is 12 (Republicans were in the minority in the House for 20 previous C ongresses and were in the minority in the Senate for 4 previous helps the new majority coordinate policies across chambers, it makes sense to average the House an d Senate figures rather than model them separately.
84 moderation taps polarization in Congress, it is highly collinear with the primary independent variables measuring bicameral differences And b ecause the prior discussion cited polarization asymmetry as the cause of greater bicameral differences, I believe the three measures of bicameral differences are the more proximate factors affecting House Senate conflict Third, u odel I include a variable Omnibus as an estimate of the number of omnibus bills introduced per Congress. 16 This control variable accounts for the possibility that the declining likelihood of second chamber passage is a function greater bill size. 17 Fourth, I include a variable that records the number of bills passed in a given Congress (labeled House/Senate Pass ). As a matter of legislative work load, it makes sense that when one chamber passes a greater volume of legislation there is a natural decline in t he organizational capacity of the second chamber to respond. Both the omnibus and workload controls are specific to the bicameral gridlock models. Finally, an important variable concerns partisan divisions at the passage stage. The variable capturing this effect ( Partisan ) is the difference in the percentage of the majority to minority voting to pass a bill added for each chamber. A value of 0 indicates the two bills passed along perfectly nonpartisan lines (such as a unanimous vote) while a value of 2 in dicates that every member of the majority in the House and Senate voted 16 As a technical matter, this variable only records the number of omnibus bills mentioned by CQ Almanac per Congress. Thus, there may be a few omnibus bills per Congress not captured by this variable. 17 The omnibus data were coded from the Policy Agendas Project Congressional Quarterly Almanac dataset: http://www.policyagendas.org/ The data were originally collected by Frank R. Baumgartner and Bryan D. Jones, w ith the support of National Science Foundation grant number SBR 9320922, and were distributed through the Department of Government at the University of Texas at Austin and/or the Department of Political Science at Penn State University. Neither NSF nor the original collectors of the data bear any responsibility for the analysis reported here.
85 against every member of the minority. The roll call data used to construct this measure 18 This factor is expected to have mixed effects. In the bicameral gridlock models, the expectation is that partisan House passed legislation Senate passed legislation, the oppos ite logic and expectations exist ( i.e. less bicameral gridlock). In the bicameral disagreement models, it is expected that the partisan variable will decrease the amount of bicameral disagreement. The logic is that w hen both chambers pass partisan legisl ation there is greater ideological coherence in those proposals. By comparison, w hen one chamber passes a partisan bill and the other passes a moderate bill (a situation represented by lower values on this variable) there is less ideological coherence bet ween the two chamber s proposal s Bicameral Gridlock Trends in Bicameral Gridlock As an initial modeling strategy I examine d in greater detail the increase in bicameral gridlock over the postreform period. On its face, this trend is evidence of an import contrast to the efficient policymaking the chambers that strong parties are thought to foster. But party theorists might counter this trend is to be expected: as the two parties have polarized and become more internally homogenous (e.g. Aldrich 1995; Rohde 1991) they increasingly use powers such as negative agenda control (Cox and McCubbins 1993, 1999, 2007) to block bipartisan legislation (allowing 18 http://www.voteview.com/
86 pr o majority legislation to pass unimpeded). Thus, the increase in bicameral gridlock over the postreform period is actually evidence in favor of the strong parties thesis. Model 1 and Model 3 explore this matter for the bicameral gridlock data (respective ly) using multilevel random coefficient models ( Rabe Hesketh and Skrondal 2008) The models are given by: E ach model includes a random intercept for each Congress and a random coefficient for the variable Partisan T he benefit of a multilevel random in tercept model in this context is its flexibility. With logit and probit models we must impose a particular functional form on the temporal components (treating them as fixed effects). Multilevel models, by contrast, allow these effects to vary in accorda nce with the data. However, as a robustness check I report analogous probit models using a linear time trend ( Congress ) and an interaction between time and the partisan passage variable ( Partisan Pass*Cong ). These models are reported as Model 2 and Mode l 4 for the House and Senate bicameral gridlock data (respectively). The multilevel models presented in Table 2 2 include three random effects estimates. The first ( ) reports the amount of between Congress variation in the likelihood of secon d chamber passage. In both the House and Senate models this variation is statistically significant. 19 The analogous logit results show that the overall 19 This statement requires a qualification as the results reported above do not show significant betwe en Congress variation in the Senate data. As I report in the text, for the Senate model the analysis rejects the multilevel random coefficient model. However, a subsequent analysis rejects a standard logit model in favor of a multilevel random intercept model (Rabe Hesketh and Skrondal 2008). In the latter model the between Congress variation is statistically
87 effect is negative over time. That is, the negative and statistically significant fixed effects on the variable Congress in Model 2 and Model 4 reveal that, over the postreform period, legislation passed by the two chambers has become less likely to pass the second chamber. This confirms as statistically meaningful the trend reported earlier. The two ad ditional random effects estimates ( and ) in Models 1 and 3 explore whether the effect of Partisan varies over the postreform era as well This directly tests if partisan legislation has become increasingly more or less likely to pass the second chamber over time. Likelihood ratio tests reject the null hypothesis that these additional estimates are jointly zero in the House multilevel model but do not reject the null hypothesis in the Senate multilevel model. For the House model Figure 2 12 plots the estimate effects of Partisan for the pre 104 th Congres ses while Figure 2 13 plots the same estimates for the post 103 rd Congresses (with the Republican Revolution demarcating the two periods). We can see that in the second half of th e postreform period (Figure 2 13 ) the effect of Partisan is increasingly negative (as given by the large negative slopes). By contrast, in the pre 104 th era (Figure 2 12 ) a handful of the effects of Partisan are nearly parallel (indicating no change ove r time) or have slightly positive slopes. The conclusion is that partisan House passed legislation has become less likely to pass the Senate in the latter half of the postreform period. This effect is confirmed by Model 2 as the interaction effect ( Parti san Pass*Cong ) is negative and statistically significant. significant. Further evidence of significant between Congress variation is confirmed by the Senate logit model.
88 Overall, these results indicate something substantively important for our discussion about the strength of parties and inter chamber bargaining : over the postreform period, partisan House passed l egislation has become more likely to end in bicameral gridlock (less likelihood of securing passage) because of bicameral disagreement with the Senate while bipartisan House passed legislation has become less likely to end in bicameral gridlock (greater li kelihood of securing passage) This is in direct contrast to the purported benefits of str ong parties: namely their capacit y in the postreform period to foster active government However, I find that this effect does not apply to Senate passed legislatio n. With these preliminary results established, we are ready to turn toward the main estimates. Main Bicameral Gridlock Estimates House Initiated Bills This section takes a comprehensive look at the factors which cause bicameral gridlock. It is importan t to make two points at the outset. First, once we include the full range of covariates in particular those capturing House and Senate compositional differences the between Congress variation ( ) estimated previously becomes statistically insi gnificant On the one hand this suggests that bicameral differences explain the negative trends in the likelihood of passage reported previously. At the same time, this means we are justified in estimating fixed effects models ( Rabe Hesketh and Skrondal 2008) Second, it is important to note that the variables Filibuster House Distance and Intra party Bicameral Distance are highly collinear. Thou gh they represent different constructs in terms of how the House and Senate might differ in their composition I speculated that both factors are derived from the same
89 causal mechanism ( polarization asymmetry ). Rather than exclude one factor or crudely combine them into a single index I estimate separate models with each variable. The results listed in Table 2 3 contain the House estimates. Model 5 includes the intra party bicameral distance variable while Model 6 includes the House filibuster pivot variable. In both models a number of covariates are statistically significant and correctly signed. I find that an increase in the distance between the two chamber medians ( Bicameral Distance ) decreases the likelihood that the Senate will pass legislation different dependent var iable. 20 If we compare the pre Republican Revolution mean bicameral distance score with the post Republican Revolution mean bicameral distance score (excluding cases of quasi divided government), bicameral distance has decreased the likelihood of Senate pa ssage by between 3% (Model 5) and 6% (Model 6). I also find in both models that partisan House passed legislation is less likely to pass the Senate. Indeed, the coefficient on Partisan is statistically significant and negative. This indicates that, irre spective of growing compositional differences, partisan House passed legislation is prone to Senate gridlock. Finally, the results in both models support the two additional sources of bicameral differences as significant contributors to bicameral gridlock in the postreform era. In Model 5 the negative effect on Intra party Bicameral Distance is such that an increase in d istance between the parties across chamber s of .06 in magnitude (the real amount of change from the 95 th to the 110 th Congresses) is esti mated to have decreased the likelihood of Senate passage by 10%. And in Model 6 the negative effect of House Filibuster Distance is such that an increase 20 It is not surprising that we f ind this link as bicameral gridlock is most closely linked to overall levels of gridlock (as the prior theoretical discussion points out).
90 of .34 in magnitude (the real amount of change from the 95 th to the 110 th Congresses) is estimated to have decreased the likelihood of Senate passage by 15%. Together, these results confirm the general hypothesis that House and Senate compositional differences are a main determinant of gridlock and that the historical increase in these factors has foster ed greater bicameral conflict. Overall, these results suggest that differences between the two chambers have a more complex and substantively powerful effect on policy productivity than Binder (1999, 2003) reported The results listed in Table 2 4 include interaction effects between our three measures of House and Senate compositional tensions and the partisan passage variable. These interaction effects allow us to assess whether variation in these factors explain the earlier finding that partisan House p assed legislation has become especially gridlocked in the Senate. Looking at the results in Table 2 4 the models confirm this hypothesis for the Bicameral Distance and Intra party Bicameral Distance interaction effects as given by the negative and statis tically significant coefficients on the interaction terms. However, there is no evidence of such an effect corresponding to the filibuster distance Figure 2 14 and 2 15 present the predicted probability of Senate passage based on interaction effects for Intra party Bicameral Distance and Bicameral Distance (respectively) for t he 95 th Congress and 110 th Congress. For these estimates I vary Partisan Passage from 0 (a perfectly bipartisan bill) to 1 (a perfectly pro majority bill) along the X axis. I comp are the 95 th Congress, where both bicameral distance measures are low, to the 110 th Congress, where both bicameral distance measures are high. We can see in both figure s that the predicted probabilities of passage diverge as Partisan increases. For examp le, according to Model 7 in Figure 2 14 a perfectly
91 bipartisan bill is 5% less likely to pass in the 110 th Congress compared to the 95 th Congress. However, a perfectly partisan bill is 31% less likely to pass in 110 th Congress compared to the 95 th Congre ss. And according the Model 9 in Figure 2 15 a perfectly bipartisan bill is 13% less likely to pass in the 110 th Congress compared to the 95 th Congress while a perfectly partisan bill is 28% less likely to pass in 110 th Congress compared to the 95 th Cong ress. Thus, with these models we have an explanation for the historical increase in bicameral gridlock for partisan House passed legislation. That source stems from increasing distance s between the medians of each chamber as well as intra party disagreem ents across the House and Senate. Main Bicameral Gridlock Estimates Senate Initiated Bills Model 11 and Model 12 contained in Table 2 5 are probit estimates of the likelihood that Senate passed legislation will pass the House. Because Model 3 and Model 4 in the earlier estimates revealed no evidence that the effect of Partisan varies over time I do not report models interacting the bicameral differences measures with Partisan (analogous to Models 7 10). However, I can confirm that those effects are all insignificant. Generally speaking, the Senate models do not perform very well. In the results, there is mixed evi dence that bicameral distance reduce s the likelihood that Senate passed bills will pass the House. Though the coefficient is negative in Mode l 8 it is statistically insignificant. In Model 9, however, the effect is negative and statistically significant, indicating that the greater the distance between the House and Senate medians the less likely a Senate passed bill will pass the Senate. Thi s is consistent with the House models. The positive effect on Party Mandate in both models indicates
92 that the longer a new majority was in the minority, the greater the likelihood Senate passed legislation will pass the House. Finally, though both Intra party Bicameral Distance and House Filibuster Distance are negative in both models as expected, they are statistically insignificant. The null result for the filibuster pivot is unsurprising however: w e would expect the filibuster pivot to have a more p ronounced effect on House passed legislation as it arrives into the Senate ( and the House models confirm this effect ). Overall the evidence reported in Table 2 5 suggests that compositional differences have little to no effect on the success or failure of Senate passed bills This is in stark contrast to the robust finding s for the House initiated bills in this regard Though the effect of bicameral distance is significant and negative in one specification, this is mixed evidenc e at best. On the one han d these null results may be due to the low sample size. With 418 total observations, there is an average of just twenty six observations per Congress. But at the same there are reasons to suspect that these null results reflect reality Given that there is less variation over time in the likelihood of Senate passage, it may be that the upper chamber has experienced li ttle decline in the success of its proposals in the House This may be due, in part, t o the fact that the Senate tends to be the more mode rate chamber Of the sixteen Congresses in our sample, the Senate median was more extreme in absolute value compared to the House in only five (31%) of those periods (95 th 97 th 101 st 102 nd 106 th ). As the more moderate body with a supermajoritarian vo ting requirement, the Senate has greater barg aining leverage over the House.
93 reverse of Washingt does not hold that the House saucer cools Senate legislation Bicameral Disagreement Trends in Bicameral Disagreement Our attention now turns to bicameral disagreement: w here the House and Senate pass ideologically different bills in disagreemen t. Consistent with the previous section, as an initial step I begin by modeling the postreform trends in bicameral disagreement. But w here the dependent variable in the previous section was dichotomous, here the dependent variable is continuous. Because the data for bicameral disagreement follow a non normal distribution 21 I used Generalized Linear Models (a generalization of the linear model ) in each specification After estimating a series GLMs provided by a negative binomial model with a power (1) link function. 22 In the first part for the multilevel models I used the Generalized Linear Latent and Mixed Models (GLLAMM) routine in Stata 10 ( Rabe Hesketh, Skrondal, and Pickles 2004, 2005) Unfortunately, the negative binomial family is not currently available in GLLAMM. Thus, for the multilevel estimates I used a poisson family and log link function which provided the second best model fit Those estimates are given by: The multilevel results repo rted in Table 2 6 reveal mixed fin dings concerning trends in bicameral disagreement (though the best performing model provides evidence 21 decline exponentially in th e positive direction thereafter. 22 Model fit was determined by the model with the lowest scaled deviance and Bayesian information criterion.
94 of a significant positive trend) On the one hand, both the multilevel random coefficient model (Model 13) and the multilevel random intercept model (Mod el 15) are rejected by likelihood ratio tests in favor of a fixed effects specification (Rabe Hesketh and Skrondal 2008). That is to say, the three random effects parameters ( , and ) indicate statistically negligib le variation. Indeed, the additional generalized linear model interacting Partisan and a linear time trend (Model 14) reveals three coefficients that are statistically insignificant. However, when we estimate a generalized linear model without the (insign ificant) interaction effect (Model 16), both Partisan and the linear time trend are stati sti cally significant and correctly signed The linear trend ( Congress) is in the same direction as the descriptive statistics, suggesting that the House and Senate ha ve experienced more severe policy disagreements over the postreform period The model also shows that partisan legislation typically contains f ewer bicameral disagreements compared to bipartisan legislation. This is the exact opposite of bicameral gridlo ck, which is consistent with expectations The finding suggests that when the majority coordinates the passage of partisan legislation across chambers assuming it can get the legislation through both house the result is ideological coherence among the t wo bills Main Bicameral Disagreement Estimates Table 2 7 reports the results for the full specif ication, including all covariates. B ecause the previous results showed no evidence of an interaction between Partisan and time I do not estimate interactio n effects in the full models At the same time, the variables Budget Omnibus and Total Pass are not included among the covariates as these factors pertain to gridlock rather than disagreement. That is, poor budgetary
95 situations, the increased use o f om nibus bills and an increase in the number of total bills passed are all logically connected to the frequency with which the House and Senate pass parallel proposals b ut there are no compelling reasons to think that, once the two chambers successfully pass parallel proposals, t hese factors would affect ideological disagreement on those bills As with earlier, t he results presented in Model 11 and Model 12 include the effect of intra party bicameral distance and the House filibuster distance (respectively). In both models the coefficient on Partisan is statistically significant and negative, indicating that when the two chambers pass highly partisan legislation in concert, the differences between those bills in terms of their policy content is limited. 23 In Model 1 1 only the effect of intra party bicameral distance is statistically significant alongside the partisan passage variable. The positive coefficient indicates that as the distance between the median senator and median r epresentative within each part y increases the divergence policy proposal likewise increases. To put this effect into context, a change in intra party bicameral distance from .029 to .091 (the real change from the 95 th to 110 th Congress) is estimated by the mode l to increase the differences in the policy content of by .11 in the first dimension Recall that earlier I cited the fact that a difference of .07 is equivalent to the distance between Joe Lieberman (D CT) and Harry Reid (D NV). Thus within party bicameral divisions represent a significant bicameral hurdle for the House and Senate as intra party 23 Because the passage models were disaggregated for the House and Senate, I was able to estimate how partisan the b ill was in each chamber. Here, I combine those estimates into a single additive index. The variable now ranges from 2, where in both the House and Senate every member of the majority voted in favor of the bill while every member of the minority voted aga inst the bill, to 0, a perfectively bipartisan bill passed by both chambers.
96 divisions simultaneously cause greater bicameral gridlock (reported previously) and greater bicameral disagreement (reported here) In Mod el 12 both the House filibuster distance and bicameral distance are statistically significant and positive. However, these coefficients are each on ly significant at the .10 level, so these effects should be considered marginally significant The positiv e coefficients indicate that as the distance between the House median and filibuster pivot has increased, and as th e distance between the chamber medians increases, the ideological distance between ea similarity increase s To p ut the filibuster effect into context, a change in the distance from the House median to the filibuster pivot from .12 to .46 (the change from the 95 th to the 110 th bills by .20. At the same time, a change in bicameral distance from .02 to .20 (the change from the 95 th to the 110 th ) is estimated by the model to increase the divergence Overall the less robust results in the models for bicameral d isagreement ( compared to the models of bicameral gridlock ) are understandable As discussed earlier bicameral gridlock has a selection effect that likely tempers the amount of disagreement between the two chambers. When the true level of disag reement is large, the likelihood of bicameral gridlock increases and, when one chamber kills the initiating chambers bill those disagr eements are not included in the present measure. On the other hand, formal work on bicameral bargaining notes that the chambers st rategically and moderate their proposals accordingly Thus,
97 in cases of major bicameral differences there may be greater moderation before passage. Despite these processes which mute the true level of bicameral disa greement there is some evidence that bicameral differences affect the level of bicameral disagreement. distance measure is limited. This is consistent with the initial theoretical discussi on. Indeed, I postulated that the differences between bicameral disagreement and bicameral gridlock might explain the conflicting results uncovered by model and Chiou and Rothenberg (2008a, 2008b). In fact t he robust results for bicameral gridlock and the mixed results for bicameral disagreement may contribute to ambiguous effects in aggregate gridlock models. Distinguishing between these two outcomes is therefore important. At the same time, I find that th e distance between the House median and Senate filibuster pivot is positive and significant, but only .10 level. However, the results do confirm the effect of within party bicameral distance as a leading contributor to bicameral disagreement As was the case with the Contract With America, the two chambers can pass similar bills where major disagreements arise because internal party differences (particularly the majority). Ultimately, the fate of these bills i s determined by how well the chambers can res olve those disagreements (the subject of the forthcoming chapters) Discussion I posed a few overarching questions at the outset of this chapter Broadly speaking, this chapter speaks to the enduring question: What role do parties play in our system of co nstitutionally limited powers? But in particular this chapter answers : How
98 have the increases in ideological polarization and the strength ening of parties in Congress the most notable developments over the last thirty years affected the functioning of le gislative checks and balances? Relying on data of postreform House and Senate policy conflict this chapter presented a series of findings that speak to the prior questions. I began by reassessing the link between overall gridlock, as defined by Binder (1999, 2003) and others, and two forms of House and Senate conflict bicameral gridlock and bicameral disagreement. Descriptive statistics and multilevel modeling showed us that both forms of bicameral conflict have increased over the postreform period. T his is an important development though it has gone largely unnoticed Next, I introduced the concept of polarization asymmetry a term used in reference to House and Senate differences in the rate of polarization over time. I then argued that polarizat i on asymmetry has caused a growth in bicameral compositional differences over the postreform period In particular, I identified three areas where the House and Senate have become increasing dissimilar over the postreform period in terms of their internal makeup : (1) greater distance intra party bicameral differences, and (3) a growing gap between the House median and Senate filibuster pivot The remainder of the chapter explored the consequences of Hou se and Senate compositional dif ferences on bicame ral disagreement and gridlock. One of the main findings from these sections is the finding that partisan House passed legislation has become increasingly gridlocked in the Senate. Surprisingly, I found tha t this effect is due to a growing gap between the medians of each chamber as well as growing intra
99 party ideological differences across the House and Senate And in a final section I modeled the severity of bicameral disagreement and sought to explain its increase over the postreform period. The evidence suggest s that this trend can be explained by the increasing distance between the party medians across chambers with limited evidence that the House filibuster pivot distance and raw bicameral distance pl ay a role as well. Overall these findings help make sense of two paradoxical and seemingly contradictory findings in the literature over the past 60 years. On the one hand, prior t o postreform period scholars were citing the lack of polarization as a r eason for decreased legislativ e productivity ( Burns 1963; Dodd and Oppenheimer 1977; Schattschneider 1942; Key 1942). But today, a number of authors cite greater polarization as a powerful determinant of increased gridlock and less legislative productivit y (e.g. Binder 1999, 2003; Bond and Fleisher 1990; McCarty, Poole and Rosenthal 2006). This disparity has yet to be fully resolved. The only attempt to reconcile these disagreements is recent work by Dodd and Schraufnagel (2009). Dodd and Schraufnagel p osit a curvilinear relationship between polarization and gridlock. At the extremes high and low polarization governing institutions are increasingly stalemated while during eras of moderate polarization lawmakers are actively engaged in policy creation an d change. The theoretical basis for this relationship is a belief that moderate conflict enhances policy ne gotiation and deliberation ( Dahl 1967). Or as Dodd and Schraufnagel (2009, 396) explain, moderate polarization (1) allows parties to adopt ideologi cal positions, thereby engendering meaningful policy debates and (2) affords enough ideological heterogeneity for successful coalition formation at final passage. This present work adds an additional ex planation for the paradoxical effect of polarization on legislative productivity In contrast to the work of Dodd and
100 Schraufnagel (2009 ), I identify an institutional mechanism one the puts legislative checks and balances at the center. The novelty of this work is that it is capable of operating from with in the strong parties thesis. I show that as the parties have polarized and become more internally homogenous, they have exacerbated the natural differences between the House and Senate. As I have shown, a major component of this process has been growing intra party bicameral differences. Additional discussion of these matters can be found in Chapter 6 But d espite the present chapter, we have only a partial look at the overall policy process ( and in particular how the House and Senate interact when t ryi ng to pass legislation ) In this chapter we discussed c onflict between the chambers: how conflict has changed over the postreform period and the factors that drive patterns of conflict However, we have said nothing about how the House and Senate resolve those conflicts when they arise. The next three chapters explore this matter
101 VD BD H M S P Figure 2 1. A spatial representation of the 99 th Congress Spatial locations based on Space scores. BD is calculated as the distance between the House and Senate medians (H and S). M is the midpoint of a House and Senate post (Reagan). VD represents the likelihood of a presidential veto after a post passage bargain. VD BD H M S P Figure 2 2. A spatial representation of the 100 th Congress Spatial locations based on BD is calculated as the distance between the House and Senate medians (H and S). M is the midpoint of a House and Senate post (Reagan). VD represents the likelihood of a presidential veto after a post passage bargain.
102 Figure 2 3. Bicameral distance Bicameral distance is calculated as the distance between the House and Senate medians. Figure 2 4. Adjusted b icameral distance Adjusted bicameral distance is calculat ed as the distance between the House and Senate medians with periods of quasi divided government excluded. Space scores.
103 Figure 2 5. P olarization in the House and Senate. Polarization is calculated as the distance bet ween the median Democrat and median Republican in each chamber. Source: Figure 2 6. P olarization asymmetry Polarization asymmetry is calculated as the absolute difference between the House on scores. Common Space scores.
104 Figure 2 7. Intra party bicameral distance Intra party bicameral distance is calculated as the average distance between Common Space scores. F igure 2 8. F ilibuster House distance Filibuster House distance is calculated as the distance between the filibuster pivot and the House median. Space scores.
105 Figure 2 9. Bicameral gridlock. Bicameral gridlock is calculated as the inverse percentage of House and Senate passed bills that cleared the second acting chamber. Source: Congressional Bills Project. Figure 2 10. Bicameral disagreement Bicameral disagreement is calculated as a verage first dimension midpoint distance between House and Senate in disagreement. Source: Space scores.
106 Figure 2 11. Adjusted b icameral disagreement Adjusted bicameral disagreement is calculat ed as average first dimension midpoint distance between House and Senate in disagreement with periods of quasi divided government removed. Source: Congressional Bills Project and
107 Figure 2 12. P artisan random effe ct (pre Revolution) The slopes represent the estimated effect of Partisan (x axis) on the log odds of second chamber passage (y axis) for each Congress.
108 Figure 2 13 Partisan random effect (post Revolution) The slopes represent the estimated eff ect of Partisan (x axis) on the log odds of second chamber passage (y axis) for each Congress.
109 Figure 2 14. Predicted probability of Senate passage by Intra party *partisan (model 7) Figure 2 15. Predicted probability of Senate passage by Bicam eral*partisan (model 9)
110 Table 2 1. Legislative roadblocks after initial chamber passage Congress (years) Percent killed by bicameral disagreement 95 th (1977 78) 2.94 96 th (1979 80) 3.23 97 th (1981 82) 0 98 th (1983 84) 3.45 99 th (1985 86) 0 100 th (1987 88) 0 101 st (1989 90) 10.53 102 nd (1991 92) 4.35 103 rd (1993 94) 12.5 104 th (1995 96) 0 105 th (1997 98) 26.67 106 th (1999 2000) 21.43 Notes: [Reprinted with permission from Sarah Binder (2003) Stalemate: Causes and Consequences o f Legislative Gridlock (Page 49, Figure 3 3 ). Brookings Institution Press ]
111 Table 2 2. Temporal trends in the bicameral gridlock House passed bills Senate passed bills Model 1 (MLM) Model 2 (Probit) Model 3 (MLM) Model 4 (Probit) Fixed effects Partisa n 0.67*** 0.21 9.59*** 2.82 0.42 0.47 6.56 9.32 Congress 0.03*** 0.01 0.08*** 0.02 Partisan *cong 0.10*** 0.03 0.07 0.09 Alpha 0.24*** 0.08 3.90*** 0.85 Random effects 0.06 0.03 0.23 0.17 0.46 0.27 0.78 1.55 0.10 0.06 0.23 0.40 N level 1 (bills) 3325 3325 418 418 N level 2 (cong) 16 16 LR test Chi2(3)=95.19*** Chi2(3)=7.03* Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The response is coded suc h that 1=second chamber bill passage and 0=bill fail. Negative coefficients therefore correspond to greater bicameral gridlock. Standard errors are robust.
112 Table 2 3. Determinants of bicameral gridlock House passed bills Model 5 (Probit) Model 6 (Probit) Bicameral distance 1.24** 0.64 2.54*** 0.46 Divided 0.01 0.07 0.06 0.08 Mood lag 0.01 0.01 0.01 0.01 Party mandate 0.01 0.01 0.02* 0.01 Budget 0.38 0.63 0.28 0.38 Omnibus < 0.01 <0.01 <0.01 <0.01 Total pass < 0.01** <0.01 < 0.01 <0.01 Partisan 0.50*** 0.07 0.48*** 0.07 Intra party bicameral 3.96** 1.67 Filibuster House distance 1.17*** 0.28 Alpha 0.47 0.64 0.12 0.57 N 3325 3325 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The respon se is coded such that 1=second chamber bill passage and 0=bill fail. Negative coefficients therefore correspond to greater bicameral gridlock. Standard errors are robust.
113 Table 2 4. Determinants of bicameral gridlock with interaction effects Hous e passed bills Model 7 (Probit) Model 8 (Probit) Model 9 (Probit) Model 10 (Probit) Bicameral distance 1.49** 0.75 2.58** 0.46 0.64 0.67 2.01*** 0.50 Divided 0.01 0.08 0.06 0.08 0.01 0.08 0.05 0.08 Mood lag <0.01 0.01 0.01 0.01 0.01 .01 0.01 0.0 1 Party mandate 0.02 0.01 0.02* 0.01 0.02 .01 0.01* 0.01 Budget 0.23 0.63 0.31 0.38 0.49 0.63 0.22 0.38 Omnibus < 0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 Total pass < 0.01 <0.01 < 0.01 <0.01 < 0.01* <0.01 < 0.01 < 0.01 Partisan 0.01 0.17 0. 25 0.17 0.28*** .11 0.27** 0.11 Intra party bicameral 2.28 1.77 4.18** 1.69 Filibuster House distance 1.01*** 0.30 1.18*** 0.28 Intra party *partisan 6.35*** 2.07 Filibuster*partisan 0.79 0.53 Bicameral*partisan 2.71*** 0.98 2.66*** 0.98 Alpha 0.59 0.64 0.30 0.62 0.40 0.64 0.10 0.57 N 3325 3325 3325 3325 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The response is coded such that 1=second chamber bill passage and 0=bill fai l. Negative coefficients therefore correspond to greater bicameral gridlock. Standard errors are robust.
114 Table 2 5. Determinants of bicameral gridlock Senate passed bills Model 11 (Probit) Model 12 (Probit) Bicameral distance 1.58 1.87 3.57* ** 1.37 Divided 0.33* 0.19 0.22 0.18 Mood lag 0.05 0.04 0.03 0.03 Party mandate 0.08** 0.03 0.07** 0.03 Budget 1.17 2.04 1.17 1.06 Omnibus <0.01 <0.01 0.01 <0.01 Total pass <0.01 <0.01 <0.01 <0.01 Partisan 0.29 0.25 0.30 0.25 Intra party bicamer al 8.55 5.57 Filibuster House distance 0.78 0.82 Alpha 2.55 1.92 2.03 1.90 N 418 418 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The response is coded such that 1=second chamber bill passage and 0=bill fa il. Negative coefficients therefore correspond to greater bicameral gridlock. Standard errors are robust.
115 Table 2 6. Temporal trends in bicameral disagreement Model 13 (RC GLLAMM) Model 14 (GLM) Model 15 (RI GLLAMM) Model 16 (GLM) Fixed effects Partisan pass 0.53** 0.21 0.27 0.67 0.53** 0.21 0.19*** 0.03 Congress 0.01 0.01 0.006** 0.003 Partisan *cong 0.01 0.01 Alpha 0.90*** 0.13 0.59 0.72 0.90*** 0.13 0.26 0.32 Random effects <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 N level 1 (bills) 346 346 346 346 N level 2 (cong) 16 16 LR test Chi2<.01 Chi2<.01 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The multilevel models were estimated with a poisson family and log link function. A negative binomial family is not currently available in the GLLAMM package. The GLM models, however, use negative binomial. The response is coded such that higher values represent greater ideological distan ce in the first dimension between House and Senate passed bills. Standard errors are robust.
116 Table 2 7. Determinants of bicameral disagreement Model 17 (GLM) Model 18 (GLM) Bicameral distance 0.20 0.23 0.45* 0.27 Divided 0.02 0.03 0.02 0.03 Mood lag 0.01 0.01 0.01 0.01 Party mandate 0.01 <.01 < 0.01 <0.01 Partisan 0.18*** 0.03 0.18*** 0.03 Intra party bicameral 1.50** 0.60 Filibuster House distance 0.40* 0.22 Alpha 0.73*** 0.27 0.66 0.31 N Level 1 (b ills) 346 346 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The response is coded such that higher values represent greater ideological distance in the first dimension between House and Senate passed bills. Standard errors are robust.
117 CHAPTER 3 THE DIMENSIONALITY OF RESOLVING DIFFERENCE S At the conclusion of Chapter 2 I summarized the main findings in the following manner : (1) over the postreform period the House and Senate have come into more fr equent and mo re severe conflict when trying to enact legislation ; (2) the increase in bicameral conflict is a function of growing compositional asymmetries and greater ideological tension between pivotal actors in each chamber ; (3) part isan House passed legislation has become especially vulnerable to bicameral gridlock over the postreform period I argued that these findings help make sense of the paradoxical claims scholars have made regarding the link between polarization, strong parties and policy productivity In the final paragraph I noted that despite findings we have only a partial look at h ow the House and Senate operate when trying to enact legislation. Though it is important to understand conflic t between the House and Senate and in particular h ow these p atterns have changed over time it is equally important to understand how the two chambers resolve those conflict The present chapter explores this process at a high level of abstraction by examining the dimensionality of resolving differences The limited research on bicameral disagreement (e.g. Binder 1999, 2003, 2008; Chiou and Rothen berg 2008a, 2008b; Riker 1992) is matched by the limited work on how the House and Senate resolve differences Furthermore, the body of existing research on reso lving differences is of limited conceptual scope (though see Longley and Oleszek 1989). What I mean is that the body of existing work examines resolving differences as a means of discriminating between competing theories of legislative organization. Duri ng the 1960s and 1970s the pervasive question was: Which chamber
118 dominates the reconciliation process? Studies by Steiner (1951), Fenno (1966), Manley (1970), Vogler (1970), Ferejohn (1975) and Strom and Rundquist (1977) provided competing answers to this question. And in the 1990s and 2000s, the leading studies have explored whether the primary features of congressional organization were majoritarian or partisan. Krehbiel (1991) argued that inter chamber bargaining foste rs informational advantages, favo ring lawmakers with p olicy expertise and fosters majoritarian policy outcomes. On the other side, Nagler (1989), Carson and Vander Wi e len (2002), Lazarus and Monroe (2007) and Vander Wielen and Smith (n.d.) argue that post passage politics advances the m over the composition of conference committees. Taken as a whole, the body of existing work on resolving differences has p osited the existence of three distinct dimensions of resolving differences one bicameral one majoritarian and one partisan. But given the ecumenical focus of these prior studies, the se dimensions have been explored in isolation. In addition existing re search has ignored what I argue d is the central feature of inter chamber bargaining consensus and compromise ( Ferejohn 1975). Or a and take, comp romise, horse claims victory, because most everyone does win In this chapter I develop a unified typology of how the House and Senate res olve differences; it is my hope that this typology contributes to our conceptual understanding of this process. In the first section, I ask what resolving differences
119 Using roll call data corresponding to all conference committees conv ened from the 95 th to the 110 th Congresses, I operationalize the typology using multivariate spatial modeling. This method identifies the policy space in which resolving differences occurs. There are two primary findings in this section. First, I uncove r evidence of multidimensionality. In particular, I find that three qualitatively distinct dimensions reconciliation, partisan conflict and bicameral conflict explain over 80% of the variation in the conference roll call patterns. Furthermore, the result s reveal th at no single dimension captures a majority of the roll call variation In providing these results concerning, this finding supports albeit perhaps only indirectly, some emerging research which finds evidence of greater dimensionality in the ro ll call record ( Crespin and Rohde 2010; Roberts, Smith and Haptonstahl 2009; Talbert and Potoski 2002). Second, I find that the primary dimension of resolving differences (i.e. the one that explains the greatest amount of variation) is a process of reconc iliation defined here as contrary to the direction of the literature over the past few decades (Nagler 1989; Carson and Vander Wielen 2002; Vander Wielen and Smith n.d.; Lazarus and Monroe 2007), I do not find that partisanship or ideology is the leading determinant of post passage bargaining. This second result is particularly consequential when juxtaposed with findings. A Unified Typology of Resolving Dif ferences The process by which the House and Senate resolve policy disputes is poorly understood. This is true from a theoretical standpoint but also in a conceptual sense. Though recent work has begun to fill the void, mostly on the theoretical side of t he ledger, passage
120 ection is to sketch a view of th e policy space in which the House and Senate resolve differences. In other words: What does resolving differences The present research on resolving bicameral differences is concerned with two primary analytical units: parties and chambers. Thus, the conceptual framework has four primary actors (1) the majority party in the House of Representatives, (2) the majority party in the Senate, (3) the minority party in the House of Representatives and the (4) minority party in the Senate. In a gam e theoretic sense, resolving differences can yield for each of these four actors one of three mutually exclusive outcomes. In the compared the initial chamber pas secured favorable concessions during inter chamber bargaining. In the second compared to the initial chamber passe inter chamber bargaining where the actor either voluntarily conceded or lost desirable policy provisions during inter chamber bargaining And in the third outcome, resolving differences yields a final po the third outcome occurs when the final policy proposal is preferred equally to the initial chamber passed bill; either no consequential changes were made or the policy gains were equivalent to the policy losses.
121 As a first step in mapping the reconciliation policy space I invoke one assumption our four actors are linked by their partisan and chamber no des. That is, I assume intra chamber and intra party preference similarity The sug gestion of intra party convergence is clear and non controversial. Indeed, roll call votes frequently pit Democrats and Republicans across chambers against one another (P oole and Rosenthal 1997; Poole and Rosenthal 2006) But the notion of intra chamber preference similarity requires some additional discussion and clarification. What I suggest is that, when we think about bargaining between the House and Senate, there ar e often policies pack aged within legislation that many members within the chamber support independent of other considerations (like partisanship or ideology). This pattern manifests most frequently when competing voting blocs reach a compromise on the flo or t amendment in Patient Protection and Affordable Care Act of 2010. Though amendment prohibited federal funding for abortion, the policy was defended by many lib eral Democrats, including Speaker Pelosi, because it was vital to secure passage of the larger bill. As New York Times columnists David Herszenhorn and Jackie Calmes (2009) aptly put it To save the health care bill she The argument here is simply that this is a systematic feature of bicameral bargaining. Once we invoke the prior assumptions, the process of resolving differences reduces to fifteen mathematically possible reconcil iation outcomes. It is fundamental to keep in mind that this typology represents the patterns of resolving differences not the
122 aggregate policy outcome 1 A quote by S trom and Rundquist (1977) is a fitting illustration of this distinction. They maintain battles to be won and more like the peace talks that occur after major battles have been we uncover three qualitatively distinct for ms of resolving differences (with sub divisions for zero sum and positive sum conflict). Reconciliation In this chapter, reconciliation is defined as the process of resolving House and Senate policy disputes by compromise and/or concession. In the current usage, reconciliation occurs when all four actors have similar preferences for the bicameral agreement as compared to their initial chamber passed bill. I n the first form, denoted by four positive signs, reconciliation yields a final policy closer to ea ed by each chamber are non overlapping (giving negotiators the option of avoiding difficult cuts). Research into the effects of bicameralism on policy outcomes supports the existence of this form of inter chamber bargaining. For example, in a comparati ve 1 For example, consider a case where the majority party in each chamber passed bills at or more extreme than their ideal points (i.e to the left or right or the party median). An d during reconciliation, a few of the relatively minor proposals were removed from the final bill. As a r. The overall process described here indicates partisan conflict during reconciliation as the final outcome shifted toward the minority. be closer to th
123 analysis of bicameral legislatures, Heller (1997) found that gridlock is typically overcome by higher government spending. In the second form o f reconciliation, denoted by four negative signs, bicameral bargaining yields a final proposal further from the ise). Support for this form of reconciliation can in the literature on bicameralism as well. For example, a number of studies posit that bicameralism leads to more stable policies because of greater policy moderation (Lijphart 1984; Riker 1992; Tsebelis a nd Money 1997). 2 This form of resolving differences is most likely to manifest when the policies advocated by each ch amber are overlapping (where difficult cuts are simply unavoidable). And finally, in the third for m reconciliation, denoted by four zero signs, bicameral bargaining yields a final proposal that is preferred by each actor equally to their conferees located the conference agreement in a spatial location where the gains and concessions were equivalent across our four actors. Typically, this occurs when only minor changes were necessary to create a workable solution. House and Senat e disagreement(s) were resolved in a way where the concessions accommodated all sides equally in a qualitative sense. No one actor or team of actors 2 Of course, this view is predicated on the notion that each chamber has dissimilar preferences. When preferences are convergent, like in some bicameral system, this argument breaks down (Heller 1997; Lijp hart 1984)
124 was able (or willing) to impose its will on the others and extract a biased conference agreement Partis an Conflict Partisan conflict is the process of resolving House and Senate disputes according to inter party policy disagreement. In this way, the node linking the majority and minority across chambers drives the pro cess of resolving differences such that the status quo shifts preferences that are systematically different across the parties, where bargaining will make one party better off. Conceptually, partisan conflict can take two very different forms: eit her one party gets what it prefers at the expense of the other party or one party gets what it prefers without a corresponding shift for the other party. If the form of partisan c onflict is the former where utility maximization is cou ntervailing this is a zero sum form of reconciliation. That is, when the majority (minority) gains concessions in bicameral latter form of partisan conflict where u tility maximization for one party is associated with no change in utility maximization for the rival party represents a positive sum form of reconciliation. Th at is, one party gains or losses policy concessions in bicameral negotiations where the concessi ons are not offset by changes for the other party. The logic of partisan conflict is prevalent in the literature on the U.S. Congress (e.g. Aldrich and Rohde 1997 ; Rohde 1991; Cox and McCubbins 1993, 2005). Moreover, this view is prominent in the study o f post passage bargaining. Studies by Lazarus and Monroe (2007) and Vander Wielen and Smith (n.d.), for example, propose that in both the House and the Senate, the authority granted the leadership over the membership of
125 bicameral committees confers signif icant control over the range of policies considered as solutions to inter chamber disagreements. A ccording to these studies, partisan conflict should be the primary dimension governing resolving differences and we should observe a preponderance of pro maj ority shift s Bicameral Conflict Bicameral conflict is the process of resolving House and Senate disputes according to inter chamber policy disagreement. In this way, the node linking the majority and minority within each chamber drives reconciliation. I n other words, the status quo shifts in conference the existence of preferences that are systematically different across the two houses, where bargaining will make one house better off. As with partisan conflict, this form of reconciliation can take two different forms: either one chamber gets more of what it wants at the expense of the other chamber (zero sum) or one chamber has a sizable ideal point shift (positive or negative) without a corresponding shift for the other chamber (positive sum). The literature on post passage bargaining supports the existence of such a dimension. In fact, the question of a conference committee original ans wer to this question the House subsequent work has proclaimed the Senate the winner on average. Fenno (196 6) argued that the Senate tends to ask for the higher appropriations Figure making it harder for the House to take away monies, and that the Senate bargained as a more homogenous collectivity. Subsequent work posited an additional expla nation of Senate dominance. They maintained that chamber
126 dominance is not a function of Senate leverage per se, rather the chamber to act is that a third dimension likely exists in the process of resolving House and Senate disputes one marked by conflict between the two chambers. 15 16 ) formal work on inter chamber bargaining reveals qu alitatively similar di mensions U sing cooperative and non cooperative game where the final outcome makes both chambers better off. Or, in other words, the status outcome is legislation that is of higher quality and less likely to be reversed in future policy contests (wher the citizenry). This dimension is similar to my chamber dimension is similar to my bicameral conflict typology. As Tsebelis and Money note, research focuses almost exclusively on either the political or efficient dimension. 3 Though there are similarities between the typology presented here and that proposed 3 bicameralism, which tends to emphasize either the political or the efficient dimension of
127 by Tsebelis and Money, the key difference is that the present work adds a second (partisan) po litical dimension to the analysis. Theoretical Expectations for Conference Outcomes One of the o verarching premises guiding my expectations for resolving differences is that both the process and issue space are inherently multidimensional. By I mean that the methods of resolving House and Senate disagreements are inherently complex in two ways : (1) the divisions among political actors competing for favorable outcomes and (2) the policy issues involved. On the process side, I believe t here is significant complexity involved in satisfying competing pivotal actors during bicameral disputes. Longley and Oleszek (1989, 193, emphasis in original) offer perhaps the best articulation of committees can best be viewed as multilateral rather than bilateral when the House and Senate pass policies in disagreement, there are typically significant tensions between one or multiple pivotal actors. Though polic y differences are a staple of legislative politics those tensions are minimized when the House and Senate pass legislation without disagreement. By virtue of reaching the process of resolving differences only exacerba tes and exposes these latent tensions. Potential cleavages include those between the two chambers, and leadership across chambers and key committee leaders in each chamber. These t ensions are derived, in part, from basic constitutiona l differences between the House and Senate S pecific to conference negotiations, we know that House members typically play the part of negotiations given their s ervice on only a few committees. Senators, by comparison, are s aid to be have in conference as
128 der the larger implications of the legislation (Carmines and Dodd 1985; Longley and Oleszek 1989, 99 100). Resolving difference is also characterized by significant issue space complexity As with th e political cleavages I believe this is an inherent feature of resolving differences When policy problems occupy the political agenda, compelling a Congressional response, often the two chambers draft and pass very different solutions to those problems (no matter how uniform the external signal may appear). Though many times the differences are similar in nature, often bills in disagreement contain For example, the House and Senate may pa ss legislation to curb global warming, falling within the domain of environmental policy but the Ho use bill contain s provision s pertaining to strategic petroleum reserve while the Senate bill contains provisions pertaining to nuclear power What began as two clean, uni dimensional b ills where the issue space for each bill was easy to comprehend each chamber suddenly require s a multidimensional solution Simply including or excluding all non overlapping policy proposals is an impracticable solution as it risks upsetting coalit ions in each chamber. Complexity among pivotal actors and in the issue space is exemplified by the 9/11 th Congress. Though both chambers passed similar bills under conditions of unified government, there wer e key differences between the House proposal and the Senate proposal. Specifically, the House bill focused on immigration reform as a national security issue (such as enhanced border security) while the Senate bill contained no such provisions. There was also non p artisan House and Senate conflict over how to allocate funds to the states (Baker
129 2008). In this case, as is true for many bicameral disagreements, conferees had to work within multiple issue spaces immigration and national security simultaneou sly, making successful resolution challenging. Moreover there was real tension between the relevant committee chairmen within and across the chambers who wanted to project their jurisdictional terrain, and party leaders responsive k and file (Fessenden 2004 ). After the conference committee drafted a compromise the Speaker o f the House Dennis Hastert delayed brining the conference report to the House floor for a vote, fearful that his caucus would reject the compromise ( Kady 2004). The natural effect of complex, multidimensional disagreements is that it creates challenges to successfully mending competing versions of public policy. These challenges, I propose, are likely to bias conference outcomes in a certain direction and cr compromise, concession and policy moderation. Simply put, majoritarian outcomes are ne cessary in conference committee in order to prevent stalemate. No one single politica l actor is able to manipulate the process of resolving differences precisely to their liking under such circumstances. Instead, outcomes must be carefully located toward the center of the policy space, nearest to all pivotal actors. Both formal and empir ical research supports this contention. For example, Heller (1997) has shown that the greater the number of actors who can affect the content of legislation, the greater the number of compromises that must be made in order to pass budgetary legislation ; t he result is higher budget deficits. Researchers have noted a similar effect with respect to legislative executive checks and balances (Alt and Lowr y 1994; Cutler 1988; McCubbins 1991).
130 In addition to the challenges facing conferees due to multidimensio nality and the difficulty satisfying competing pivotal actors, resolving differences is a process marked by uncertainty and significant transaction costs which, untimely, create an incentive for risk averse legislative decisions. Regarding uncertainty, fr equently there are doubts among lawmakers before a conference committee convenes concerning the ability of bicameral solution Indeed, a handful of conference committees fail each year. As one example, landmark immigr ation reform died in conference during the 109 th Congress. There is also uncertainty with respect to the preferences of the pivotal lawmakers in each chamber. This is revealed by the fact that prior to any conference committee, the initiating chamber mad e a formal policy proposal and that proposal was rejected by the second chamber Moreover, pivotal actors in each chamber have an incentive to misrepresent their true position. Conferees will typically argue that that certain provisions will never pass t heir chamber, germaneness requirements As Longley and Oleszek (1989, 2004) note: procedures, and psychology are among the devices that conferees use to feel each o ther out in an attempt to gain In short, because of uncertainty successful reconciliation is a challenging endeavor and is not a foregone conclusion. I also believe there are significant transaction costs involved in resolving d the passage and postpassage stages (for example, those associated with drafting, marking up, and passing legislation in each chamber as well as those associated wit h studying the policies in disagreement, considering rival solutions, identifying the position
131 of relevant actors and, most importantly, debating and drafting a workable solution). The decision to incur these costs is an important legislative decision. A t any given point there may be multiple bills in conference committees and countless other bills waiting to receive consideration in standing committees or on the chamber floor. Because time is finite, there is an incentive ceteris paribus to adopt an eff icient solution a low cost compromise that quickly satisfies competing actors while maintaining provisions in the original proposal. In short, because of significant uncertainty and the costly nature of bicameral bargaining, there is an incentive for con ferees, committee chairmen and party leaders to adopt a risk averse bargaining strategy. Once legislation reaches the conference stage the majority of the work is complete. Suffice it to say that any member who voted for the initial bill has invested the ir limited legislative resources (time, money, bargaining power, vote trading, hearing time, campaign promises, staff expertise, etc.) in the legislation waiting in conference. If conferees, party leaders or committee chairmen try to push a political outc ome at this late stage it may upset fragile voting coalitions and/or spur the minority to engage in dilatory tactics (a forthcoming section reviews these possibilities). Simply put, if a conference bill dies, or the agreement causes significant delays in adoption of a conference report, the cost to lawmakers who voted to pass the bill or invested resources in a conference committee m ay outweigh the costs of advocating a simple compromise from the outset. Thus, seeking compromise at the conference stage ca n be a utility maximizing choice (where at earlier stages of the policy process legislative failures are much less costly or politically damaging).
132 The expectation that conference committees are characterized by compromise, concession and majoritarian outc He views delegation in all legislative institutions, not just conference committees as a decision driven by informational gains posits that legislative institution s are organized in order to collect, analyze and transmit policy information back to the chamber. Based on principal agent theory where the principal is the chamber median and the agents are the committees the basic logic is that information is costly to obtain and evaluate (a problem committees solve). U sing data on the preferences of conference committee delegations, he finds that, consisten t with his informational theory, conferees represent the preferences of their parent chamber rather than their com mittee or party Thus, a ccording to Krehbiel, despite the rules which give party leaders the power to name partisan conferees W e would expect conference outcomes to be majoritarian rather than partisan for procedural reasons as well First, conference agreements are debatable in the Senate, giving the minority the option of filibustering (Oleszek 2007; Palmer and Bach 2003). Since the minorit y almost always possesses the votes to sustain a filibuster manipulating conference committees for partisan gain risks killing an entire bill which the majority has invested significant costs The minority in the Senate can also block a conference commit tee from convening. This was common occurrence in the 108 th Congress where Senate Democrats, upset about being blocked from conference committees on a pair of earlier bills (HR 1, the Medicare drug bill, and HR 6, an energy bill) blocked the opening of nu merous subseq uent conference committees ( HR 1904,
133 HR 7, HR 3108). Such a maneuver forces the majority into a more costly reconciliation process (using informal negotiation, amendment trading or passing a new bill). Thirdly, the range of policy solutions (Longley and Oleszek 1989; Lazarus and Monroe 2007; McCown 1927). This rule states that conference agreements must be within the interval connecting each y not be added to a conference report raised against the conference report in either chamber. Thus, the majority is constrained from revising legislation in conference howev er they see fit. Finally, the rules of each chamber governing voting on conference reports give wide latitude to conferees. One a conference agreement is struck, it returns to each chamber privileged ( meaning it cannot be blocked from receiving a vote) a nd under a closed rule (a rule preventing further modification). Thus, if conferees propose a genuine compromise take it or leave it form. Though the leadership can adva nce a motion to recommit the bill to a conference with instructions, not only are those motions rare (Tsebelis and Money 1997, 201) but the instructions to conferees are non binding. In sum, there are significant procedural constraints on the majority whe n they attempt to use a conference to seek partisan outcomes. Observing Resolving Differences To assess the proces s of resolving bicameral disputes we need the revealed preferences of representatives and senators for proposals that were, first, passed b y the two chambers in disagreement and, second, successfully reconciled. Armed with this
134 post passage stage. To obtain this information in a systematic fashion over th e entire post reform era I rely on roll call votes. The strength of such an approach is its conceptual clarity. Though bicameral disagreement and reconciliation are complex and often veiled process es we can straightforwardly explore the contours of inte r chamber b argaining as the two votes (pre and post conference ) lie in the same policy space. In other words, the contours and severity of pre conference disagreement are easily compared to the contours and severity of the post conference agreement for o ur four actors But roll call votes contain some apparent limitations in this application. One possible limitation of using roll call votes for this type of analysis is that it limits our generalizability to the so called formal bicameral disagreements This is in contrast to informal bicameral disagreement and reconciliation. For example, we know that par ty leaders and committee chairme n frequently coordinate the introduction and passage of legislation with their counterparts in the other chamber. T hough this form of provides some level of generalizability to these cases. Fi rst, it is important to keep in mind that Article I sec. 7 of the Constitution (the every attempt to pass legislation. And though they may not bargain formally, the institutional constrain ts ategic behavior ( Sin and Lupia 2004; Taylor 2008; Rogers 1998, 2005). Thus, the absence of formal bicameral bargaining does not indicate the absence of the bicameral hurdle and its effects. Second, and more importantly, it is important to understand how, exactly, the so icameral
135 reconciliation works. During a series of personal interviews I conducted as part of this project, House and Senate legislative and committee staff described to me a mu lti tier, hierarchical conference negotiation process that typically occurs before a formal conference committee is convened. On the lowest tier of the hierarchy are the routine, non controversial provisions that are resolved during informal meetings amon g junior by demarcate on a provision by provision basis the statutory differences between the the middle tier, policy disagreements are either more consequential or those that a tier the most contentious issues. These issues are negotiated with direct conferee involvement. For the most consequential and divisive issues, negotiations are members) or t committees). The previous discussion has two points relevant to generalizability of the present methodology. First, the vast majority of bicameral bargaining associated with a formal conference actually takes place behind the screens. By one account, 90% of the disagreements are resolved by staff at the lowest two tiers. But second, even the most contentious issues are sometimes resolved before a formal conference is convene d. In my interviews, legislative staff routinely reminded me that conference committees are
136 4 once recalled on rees debate the issues in private. Nearly all of the toughest decision come down to private negotiations disagreements still speaks to the larger process of inter chamber bargaining as the formal and informal processes are interrelated. Thus, to observe the process of resolving differences in a systematic fashion over the postreform period, four observations were recorded for every conference committee where the disagree ments were successfully resolved the chamber yeas and nays for the majority and minority on approving the original chamber passed bill (t 1) and the chamber yeas and nays for the majority and minority on the final conference report (t). This data was used parties (P) in each chamber (C). These shift parameters are constructed as the difference in the roll call from the post conference vote to the pre conference vote. Mathematically, a positive shift parameter ind icates actor P i C j increased its vote share from the original chamber passed bill to the conference report while negative shift parameters indicate the actor P i C j decreased its vote share between the two periods. A value of zero indicates no change in vote share. Conceptually, the shift parameters reveal how the process of resolving differences affected legislative outcomes from the perspective of competing parties and chambers and allows us to assess the policy space. Shift Parameter P i C j = (P i C j Aye t / P i C j Voting t ) (P i C j Aye t 1 / P i C j Voting t 1 ) 4 Senator Grassley (IA). American Jobs Creation Act of 2004 Congressional Record (October 9, 2004) p. S22971 Available from: LexisNexis Congressional; Accessed: 11/8/2010.
137 It is important to point out that in this application usable votes are the exception rather than the rule. This is because the methodology requires four roll call votes on a single proposal one for the origi nal House passed bill, one for the original Senate passed bill, one for the House approving the conference agreement and one for the Senate approving the conference agreement. Unfortunately, unanimous consent and voice votes are common on at least one of these votes (especially in the Senate and on conference report votes). Overall, though the data collected for this study contains more than a thousand conference committees convened during the post reform period, only about one in five (n=206) have four u sable roll call votes. It is also important to point out that the missing observations differ from the non missing observations in at least one important way they are the least partisan issues. This intuition is born out in the data as statistically meani ngful. 5 6 Though prior work has simply assumed that voice votes on a conference report indicate unanimity (Binder 2003), I concur with Chiou and Rothenberg (2008a and 2008b) who argue that such an assumption underestimates the true extent of bicameral dis 2008 reply). In fact, the problem of assuming unanimity is only magnified when 5 Using the roll call data I calculated the difference between the proportion of each pa rty voting differences in the roll call in the form of percentage (where 1 is a perfectly partisan bill and 0 is a perfectly bipartisan bill). In the House, the means a re .27 and .36 for the missing and non missing observations (respectively). In the Senate the means are .13 and .23 for the missing and non missing observations (respectively). Both differences are statistically meaningful (t= 4.06 for the House and t= 5.56 for the Senate), indicating that the missing cases are less partisan than the cases included in my sample. 6 successful conferences. Though these are ultimate ly partisan issues, the low likelihood of failure at this stage indicates that larger point about a bias in the direction of finding a partisan dimension still holds.
138 constructing the shift parameters. 7 In the end because the missing values tend to be the least partisan bicameral disagreements in the population, the shift parameters used in the forthcoming principal components analysis are likely biased in a direction that overestimates the partisan dimension: a bias that favors the conventional wisdom (Carson and Vander Wielen 2002; Lazarus and Monroe 2007; Nagl er 1989; Vander Wielen and Smith n.d.). At the same time, because we have a bias in favor of the most contentious issues, the sample used in the forthcoming analyses contains the most salient and important policy disputes throughout the postreform period. Overall, I regard and value It is also important to state very clearly what the shift parameters tell us. It may be appealing to regard the conference shift parameters as the utility of our four actors for t he final outcome. This is not the case. Rather, the conference shift parameters reveal the utility of our four actors for the final outcome relative to the bill at final passage. So, for example, we may find that a conference committee modified the origi nal bill in a direction unfavorable to the Senate minority. This does not mean the Senate minority wanted on the main details of the policy). Thus, the shift parameters help us assess the patterns of conference operation their aggregate nature, determinants and historical patterns 7 This occurs because invoking the unanimity assumption (Binder 2003) often results in ve ry large, positive shift parameters, creating a massive skew in the direction favoring reconciliation. Only rarely in the data containing complete roll call votes do we observe total agreement on the conference report.
139 Principal Components Analysis The typology outlined theoretical discussion outlined ea rlier suggests that the the process of resolving House and Senate disagreements is : (1) intrinsically complex and (2) that no single deductive logic or overarching the oretical perspective adequately explains the overall process. Though prior work on the topic of inter chamber bargaining does not dispute this point, the slate of existing studies have yet to address the matter theoretically or empirically. As such, this section advances our understanding of resolving differences by discriminating between competing theoretical traditions and, u ltimately, providing a broader conceptual l ook at the patterns of inter chamber bargaining. Given the hypothesis of multidimensi onality, an analysis of the shi ft parameters using univariate statistic s means, medians, frequen cies, etc. or bivariate statistics correlations, regressions, etc. would result in a loss of information and limit our ability to understand the data generatin g processes. Indeed, conference outcomes can fall along multiple dimensions Consider the following case In July of 1999 Clinton signed a law preempting lawsuits arising from the so (106 th Congress, HR 775) At the passag e stage 95.6 % of House Republicans supported .4 % of House Democrats In the Senate, by comparison, 92.5 % of Republicans supported their version of the legislation compared to only 26 .1 % of Senate Democrats. Though these aggregate figures are comparable, t he House and Senate bills contained notable differences, with the House bill characterized as a more partisan bill compared to the Senate proposal (Ota 1999). After a conference between House and Senate negotiator s, at the poastpassage stage
140 98.6% of House Republicans supported the conference report compared to 89.8% of House Democrats while 96.3% of Senate Republicans supported the conference report compared to only 64.4% of Senate Democrats. What we see quite in tuitively in these roll call patterns is that the conference agreement was preferable to all four actors compared to the initial House and Senate bills Clearly this is a case of reconciliation. But at the same time we can see that the conference agreem ent was especially preferable for Dem ocrats. Clearly, the conference agreement falls along two dimensions simultaneously a compromise and a pro minority outcome. A quick review of some contemporaneous accounts of the negotiations supports this claim 8 O f the agreement, Republican Orrin Hatch stated on the Senate floor that the final bill reflects the spirit of compromise. But I must admit that I believe the original Judiciary and Commerce Committee bills -along with the House bill -wou ld have been far m ore 9 Univariate statistics would obscure these nuanced patterns; t hus, I tu rned to multivariate analysis. A branch of statistics concerned with the interdependence of multiple variables, multivariate analysi s allows the researcher to describe meaningful patterns in complex processes (where data of high dimension are difficult to interpret graphically or through simple correlation plots). In particular, the multivariate results presented in this chapter 8 In conference, Clinton and congres sional Democrats accepted a cap on punitive damages for small business, a provision to waive civil penalties for small business and a scaled back version of an original provision that would have limited the ability of financial institutions to collect debt payments from individuals negatively affected by the Y2K glitch. These concessions were favored by Republicans. Democrats, on the other hand, gained a provision in conference requiring lawsuits seeking over $10 million in damages to be filed in federal court, keeping many possible class action lawsuits in state court, as well as a provision that increased the liability of defendants who cause economic damages. For a an overview of these issues see Ota (1999). 9 Senator Hatch (UT). Y2K Act Conference Re Congressional Record (July 1, 1999) p. S8032 Available from: LexisNexis Congressional; Accessed: 6/2/2011.
141 were d erived from principal components analysis. Principal components analysis 10 (or PCA) is a statistical technique used to simplify multivariate data by transforming a series of input variables into a low dimensional data configuration (in political science, s ee Heckman and Snyder 1997). PCA achieves this by producing a series of linear combinations of the original input variables that explain variance in the original variables. The first principal component contains maximal overall variance, the second princi pal component contains maximal remaining variance and so on, until all the variation contained within the input variables is represented via the estimated components. And because the components are orthogonal, they represent qualitatively different bits o f information. Inferential methods using principal components analysis is based on the assumption of multivariate normality ( Rabe Hesketh and Everitt 2007) 11 As it turns out, this assumption is violated by all four shift parameters. 12 In particular, the data are leptokurtotic, meaning there are many observations around the mean, resulting in a hand, this is a natural feature of bicameral reconciliation. Only rarely are major changes 10 All analyses were conducted in STATA 10. All techniques employed in this chapter followed two STATA reference manuals ( Rabe Hesketh and Ev eritt 2007; StataCorp 2009). In addition, I consulted Jackson (2003). 11 It is important to recognize that the descriptive aspects of principal components analysis are unaffected by this assumption only inference based the estimated standard errors (StataC orp 2009). 12 I used the omninorm routine in STATA 10. This routine produced a test statistic of 780 (p<.001).
142 non overlapping, major changes are simply unavoidable. Despite the leptokurtosis we can attempt to normalize the data using a variety of transformations. 13 These transformations result in the normalization of three out of the four shift parameters (the shift parameter for the Senate majority is still kurtotic). Thoug h the data have been dramatically improved in terms of satisfying the assumption of multivariate normality, ultimately this assumption is still violated. 14 Since the descriptive aspects of the principal components analysis are unaffected by the violation o f multivariate normality (StataCorp 2009), those results are discussed substantively. The inferential discussion, where multivariate normality becomes potentially problematic, makes use of the transformed data as well. A decision has to be made at the ou tset concerning the appropriate number of components to retain from the principal components analysis. The literature contains two recommendations ( Jackson 2003 or Borg and Groenen 2005). The first is to keep any component with an eigenvalue larger than one. The second is to retrain any present scree plots with confidence intervals for the untransformed and transformed data. Since there is no clear elbow in either plot, I r etrained the first three components. Though the eigenvalue for the third component is estimated at less than one, it remains within acceptable bounds based on the confidence intervals. 13 First, kurtosis was corrected using an arctan transformation (arctan((X i skewness was corrected using the lnskew0 routine in STATA 10. 14 With the transformed data, the test of multivariate normality yields a test statistic of 39. Though this is still a significant departure from multivariate normality, the over 9000%.
143 The Spatial Dynamics of Resolving Differences The results from the principal component analysis scaling the entire time series (95 th to 110 th Congre sses) can be found in T ables 3 2 to 3 5 Tables 3 2 (untransformed data) and 3 3 (transformed data) present the component loadings for each shift parameter. The component lo adings represent the correlation between the shift parameter and the revealed dimension. Thus, t he greater the influence of each shift magnitude and, most importantly, the directi o n of these loadings allows us to interpret each dimension in relation to the theoretical forms of resolving differences (see the row The componen t loadings in T ables 3 2 and 3 3 reveal some clear patterns regarding what resolving positive and approximately equally sized loadings for our four shift parameters on the first component reflect a process of reconciliation Recall that reconciliation is defined as of resolving bicameral disputes by the equally signed loadings indicate that all four shift parameters fluctuate in the same direction (either positive or negative). When all four shift parameters are positive, each actor P i C j increased its utility after a conference committee met to resolve bicameral differences. Conceptually, this outcome reflects a form of bargaining analogous to logrolling or vote trading. On the other hand, when all four shift parameters are ne gative in the first dimension, each actor P i C j decreased its utility after a conference committee met to resolve differences. This outcome has been described colloquially as plit the where moderation of the o riginal bill is preferred to the status quo than no bill at all.
144 The opposit ely signed loadings by party on the second component reveals a process of inter chamber bargaining marked by partisan conflict. These loadings indicate that as one party increases its utility after conferencing the other party decreases its utility Thus, partisan conflict is the second leading form of resolving differences. It is important to note that the oppositely signed loadings of relatively large and equal magnitudes sugge st that this form of reconciliation is zero sum partisan conflict where in some cases the majority party gains favorable provisions at the expense of the minority (a pro majority shift), and in the other cases the minority party gains favorable provisions at the expense of the majority (a pro minority shift). I find no evidence of positive sum partisan conflict in the results. This comports with classic conceptions of an ideological and/or partisan policy space (that the parties have Euclidian single pea ked preferences that anchor opposite ends of linear dimension). However, though the results show that partisan conflict is an important component of resolving differences they also show that partisan conflict is not the leading determinant of this proces s (contrary to the conventional wisdom). And finally, the oppositely signed loadings by chamber on the third component indicate a process of bicameral conflict. 15 The loadings reveal that as the majority and minority in one chamber increase their utility a s the result of a conference the majority and minority in the rival chamber decrease their utility As with the second dimension, this is a zero 15 The flipped signs for the component loadings in dimension three across the transformed and untransformed datasets are not consequential. This difference merely reflects how the algorithm arranged the linear combinati on of the original shift parameters for this dimension. What is important are the magnitude of the loadings and their signs within each dimension relative to one another.
145 equal in magnitude to the other chamb wisdom suggesting that our institutional design occasionally pits distinct interests in the chambers against one another. However, because it is the third of three components, the implication is that bicam eral conflict is less salient a form of resolving differences than reconciliation or partisan conflict. Cases in Each Dimension Reconciliation Recall that the scaling procedure revealed that the first dimension of resolving differences is reconciliation. Also recall that within this dimension, positive coordinates reflect a reconciliation process where all four actors increased their utility or vote share after a conference committee (a process of logrolling or vote trading). On the other side of the dime nsion, negative coordinates reflect a reconciliation process where all four actors decreased their utility or a compromise). The legislative history of the Telecommunications Act of 1996 exemplifies t he former while the 2001 transportation appropriations bill exemplifies the latter. Once enacted into law, the Telecommunications Act of 1996 made sweeping changes to communications policy by increasing long distance telephone competition, deregulating t he cable television industry and reversing limits on media ownership. In short, this law followed the deregulatory trends of the 1980s and 1990s. Though both the Senate bill (S. 652) and House bill (H.R. 1555) passed by comfortable margins (81 18 and 414 16 ), the overall complexity of both proposals created a number of contentious inter chamber disputes (Carney 1995) The two main areas of contention concerned the regulation of cable rates and the concentration of media ownership
146 (Carney 1995) Though t here were overlapping provisions in each bill, Republicans believ ed Senate Republicans sacrificed too many provisions to the Democratic minority while Senator Ro bert Dole (the majority leader) complained about the lucr ative giveaways to broadcasters contained in the House bill (Carney 1995). Comparing the debates in both chambers on the conference agreement reveals the nature of the compromise. Of the agreement, Edward Markey (D,MA) said 16 the conference report is a much improved piece of legislation. It scales back or removes many of the problematic provisions of H.R. 1555 while retaining pro compe titive, pro consumer ohn Linder (R,GA) explained 17 that long months of negotiation. I believe that the conferees have wo rked in good faith to create a balanced bill .. this improved piece of roll call patterns used in the principal components analysis. pre conference proposal, 92% o f Republicans and 50% of Democrats voted to pass the bill while 100% of Republicans and 92% of Democrats pre conference proposal, 94% of Republicans and 65% of Democrats voted to pass the bill while 98% of 16 Representative Markey (MA). Conference Report on S. 652, Telecommunications Act o f Congressional Record (February 1, 1996) p. H1169 Available from: LexisNexis Congressional; Accessed: 11/13/2010. 17 Representative Linder (GA). Conference Report on S. 652, Telecommunications Act of Congressional Record (February 1, 1996) p. H1145 Available from: LexisNexis Congressional; Accessed: 11/13/2010.
147 Republicans and 91% of Democrats voted to pass the conference agreement. Clearly, the agreement struck between House and Senate negotiators was more palatable to all sides as compared to the original chamber passed bills; thus, this case refl ects reconciliation in the positive direction On the other side of the reconciliation dimension where the scaling procedure produces negative first dimension coordinates competing actors sacrificed preferred compromise). An example in the dataset is the 2001 transportation appropriations bill. The House initiated the spending proposal, a $55.2 billion measure, by a vote of 395 15 S even Republicans and six Democrats voted against the House bill, which did little more than establish authorization levels for transportation agencies and programs. The Senate acted second, passing an amended version of the House bill by a vote of 99 0. With a price tag of $54.8 billion, on it s face the Senate bill resembled the House proposal But unlike the House bill, the Senate bill contained a number of narrow policy provisions such as the establishment of national drunk driving standards, new regul ations on the amount of rest required of truck and bus drivers and language to block recently enacted rollover standards Under the compromise forged in conference, the drunk driving standards and rest regulations w ere given delayed implementation deadlines while the language striking the rollover standards was removed from the final agreement. To make these difficult agreements more palatable, $3 billion in last minute earmarks were included in the conference agree ment. Because there were a number of nonoverlap and cuts were unavoidable, a number of
148 lawmakers in both chambers withdrew their support for the final legislation. In the end, the House passed the conference agreement 34 4 50, with 32 Republicans joining 5 in the Senate, the conference reported passed 78 10, with 7 Republicans joined by 3 struck between House and Senate negotiators made a number of tough changes to the original chamber pass ed bills, though no one chamber disproportionately benefited or sacrificed. Partisan Conflict Recall that the sc aling procedure revealed that the second dimension of resolving differences is partisan conflict. Also recall that negative coordinates reflect partisan conflict where the minority party secured favorable concessions at the expense of the Intelligence Reform and Terrorism Prevention Act of 2004 (the 9/11 Commission Recommendations bill discussed earlier ) is one example of this form of resolving differences. The bills pass ed by the House and Senate implementing the 9/11 Commission recommendations sought to overhaul encies as well as centralize power in a new agency with expanded budgetary and legal powers. The Senate acted first, passing a bipartisan bill (S 2845) that hewed closel y to the 9/11 ( "2004 Key Votes: A Mix of Hits and Misses For The House, on the other hand, passed its own version of reform (HR 10) along strict party lines (only eight Republicans voted against the pr oposal compared to 125 Democrats). The major differences between the two proposals included the amount of power afforded the new national intelligence director
149 and a series of immigration restrictions and law enforcement provisions added to the House bill Because of these differences as well as major intra party bicameral disputes among Republicans (Baker 2008) Democrats held a strategic advantage. By siding with Senate Republicans, Democrats in both chambers were able force concessions in conference from House Republicans ( "2004 Key Votes: A Mix of Hits Ultimately, 67 House Republicans voted against the compromise proposal compared to only 8 on the initial House passed bill. House Democrats, on the other hand, voted 183 8 in favor of the conference report compared to only 69 125 in favor on the original bill. Though the Senate roll call patterns are much less dramatic (because the bicameral compromise tracked the can senators were openly dismayed at James Inhofe (R OK), voted against the agreement even though it was nearly identical to the bill he supported just a few months earlie r. On the Senate floor 18 Inhofe lamented items in the House bill that were struck form the conference agreement namely the strict immigration provisions. 19 liked. I have a very short list of things t hat were taken out of the House bill in 18 Senator Inhofe (OK). "Intelligence Reform and Terrorism Prevention Act of 2004 Conference Congressional Record (December 8, 2004) S 11963 Available From: LexisNexis Congr essional; Accessed: 11/13/2010. 19 The immigration provisions excluded from the final agreement included a provision requiring proof of lawful presence in the United States, enhanced licensing requirements, restrictions on ign documents and language to close a gap on the U.S. Mexico border near Sand Diego.
150 improved their utility after the conference while the majority ( Republicans ) sacrificed a number of their preferred provisions ( namely on immigration). Bicameral Conflict Recall that the scaling procedure revealed that the third dimension of resolving differences is bicameral conflict. Also recall that within this dimension positive coordinates reflect a pro House outcome. This r econc iliation outcome is embodied in the American Jobs Creation Act of 2004 (H.R 4520). Though the bill enacted into law made some of the most significant changes to the corporate tax structure in nearly 20 years, the bill began in the House as an attemp t to repeal U.S. export subsidies that the World Trade Organization had deemed illegal. Despite bipartisan agreement on replacing the subsidies, significant disagreements e xisted over what to enact in their place The House bill favored reducing taxes on U.S. multinational corporations and included a buyout for tobacco farmers. Because the tobacco provision was in direct nate took the rare step of adding a provision to its bill after passage one that matched the tobacco buyout with greater FDA tobacco regulations. A separate Senate provision also required that the law not add to the deficit rporate In conference, negotiators accepted the Senate proposal that limited tax shelters, reducing the cost of the law. However, not only did the tobacco buyout remain in the final compromise, but the FDA regulations were stripped from t he conference agreement dra matic losses in conference is that House Republicans, aided by a handful of House Democrats, w ere able to vote in complete unity (Mullins 2004) As t he close of the 108 t h
151 Congress drew near, the status quo (raising WTO sanctions) was worse than letting the bill die. Though Senate Democrats were especially unhappy with the outcome, a number of Senate Republicans were as well. On the Senate floor, Mike DeWine (R,OH), who voted for the initial Senate bill explained his decision to reverse and vote against the conference agreement, saying 20 reform, sending i t on to the House, and then having it stripped out of this conference passed bill to vote and voted in favor the conference agreement. The House was the clear winner in this case. The Exp lanatory Power of Each Dimension The prior results and discussion contribute to our broader understanding of how the House and Senate resolve disagreements. Such an analysis of post passage reconciliation has been conspicuously absent from the congression al literature. But in additional to classifying conference outcomes, w e can also explore the relative explanatory power of the three dimensions. For example: How much variation in the patterns of resolving differences does the three dimensional typology e xplain? To what extent does reconciliation the first dimension explain the roll call patterns vis vis 20 Representative Linder (GA). American Jobs Creation Act of 2004 Conference Report Congressional Record (October 9, 2004) S10936 Available From: LexisNexis C ongressional; Accessed: 12/2/2010.
152 partisan and bicameral conflict? Answering these questions moves us beyond the conceptual contribution and allows us to discriminate between competing theoretical accounts of congressional organization and legislative politics. Table 3 4 (u ntransformed data) and Table 3 5 (transformed data) present the amount explained variation captured by each dimension. The first column contains the eigenvalue for e ach component while the second column contains the percentage of explained variation. For the percentage of explained variation I also estimate confidence intervals, allowing us to compare the relative importance of the various dimensions. Tables 3 4 (untransformed data) and 3 5 (transformed data) reveal that the first three components explain 84% of the variation in the shift parameters. Though this is quite strong and validates the three dimensional typology, it also suggests that 16% of the variati on in conference outcomes is driven by alternative forms (perhaps a fourth 21 dimension). For example, it could be that the Senate minority, aided by its ability to exploit unlimited debate, is able to extract concessions that it favors relative to the othe r three actors. Though such a dimension is mathematically capable of man ifesting in the results, in the end this dimension does not capture enough variation in the original shift parameters. The implication, then, is that any fourth form of post passage bargaining and conflict is rare compared to the three main dimensions. Looking more closely at Tables 3 4 and 3 5 we can see that the first component (reconciliation) accounts for between 34% and 38% of the explained variation in the 21 Though I followed the recommendation in the literature to not retrain any component with an eigenvalue less than one, I explored the fourth dimension regardless. I was unable to discern a meaningful logic in the shift parameter loadings. This suggests that the fourth dimension is weak and probably captures idiosyncratic instances of reconciliation.
1 53 shift parameters d epending on whether we analyze the untransformed or transformed data (respectively). The results also reveal that the second component (partisan conflict) accounts for between 24% and 29% of the vari ation in the shift parameters. And finally, the third c omponent (bicameral conflict) accounts for between 21% and 22% of the variation in the shift parameters. For both the partisan conflict and bicameral conflict the loadings revealed that these outcomes were zero sum. The fact that no single dimension cap tures a majority of the variation in the process of resolving differences is further evidence for the multidimensionality hypothesis. However, closer inspection of the confidence intervals shows that the explained variation captured by reconciliation is s tatistically larger than the explained variation for partisan conflict In the untransformed data, the p value of a difference of means test is .056 (reported in the Table notes) In the transformed data, were inferential statistics should be more valid because of the data transformations, the p value of this test is less than .001. Thus, where the vast majority of studies find that roll call voting in the House and Senate is driven almost entirely by an ideological dimension separating liberals (Democra ts) and conservatives (Republicans) in the postreform period, these results suggest that resolving House and Senate policy disagreements is driven, first and foremost, by a qualitatively different process one governed by compromise and/or concession. This finding stands in direct contrast to the conventional wis dom about conference committees: that they are manipulated by the Speaker and Majority leader in order to produce a pro majority biased outcome. Though certainly true in some cases, I find that par tisan conflict this is not the over arching function of resolving differences Moreover, though partisan conflict is the
154 second leading component, additional tests reveal inconclusive results concerning the significance of the meaningfulness of the differe nce between dimensions two and three. In the untransformed data, the results appear significant at .01 level according to a differences of means test while in the transformed data the differences are not significant (p=.20). Thus, it is theoretically and substantively interesting that bicameral conflict cannot be distinguished between partisan conflict (at least when we scale the entire time series) As one additional piece of evidence supporting the finding of multidimensionality I estimated the spatial ( Rabe Hesketh and Everitt, 2007 ). These scores identify a point in the three dimensional resolving differences policy space corresponding to the dimension or dimensions that explain the ind ividual conference outcome. This allows us to create a visual map of how resolving differences affected the final policy and quite literarily negative values indicate tha t our four actors lost vote share after a conference committee while positive values indicate that our four actors increased vote share. In majority while positive valu Figures 3 3 maps these outcomes in three dimensions 22 The values in each of these figures were scaled in three dimensions so that the origin (the point at X=0, Y=0, 22 For this I used MATLAB.
155 Z=0) has a real interpretation it is the point in the space where all shift parameters equal zero, indicati ng that the changes made during confe rence negotiations had no effect on the pre conference House and Senate roll call patterns. Such a pattern typically arises when the changes made in conference were relatively minor relative to the full bill. You can notice in Figure 3 3 that most of the observations are located around the origin of the three dimensional coordinate system. This indicates that, overall, major changes in the pre to post conference roll call patterns are rare Previously I cited researc h by Strom and Rundquist (1977) who noted that conference after major b 450). The clustering of observations around the origin of Figure 3 3 lend s credence this characterization of conference committee outcomes Further, notice that the variation away from the origin is seemingly random. That is, conference modifications are not predominantly located on one or two dimensions but distributed rela tively evenly, across all three. This visual evidence confirms the prior argument about the multidimensional character of conference outcomes. Though the prior findings are unique in that they apply to specific aspect of the legislative process, and hel p us understand the functioning of a conference committee, they also addresses a major debate in the congressional literature. A majority of studies in the congressional literature both empirical and formal follow Poole and 2007) lead i n conceptualizing the policy space and roll call voting as uni dimensional (particularly during the post reform period). Of course, Poole and
156 did not appreciably improve roll call predictions. To be fair, the approach used here is different in important ways compared to the traditional roll call study. Namely, the present work examines a process of resolving disagreements which is dynamic and thus qualitatively different than a one shot roll call vote at the passage stage At the macro level, what reaches the roll call stage often pits Democrats a gainst Republicans, liberals against c onservatives Still, the finding of multidimensionality and to a lesser extent the finding that party conflict is not the leading component is imp ortant in the context of emerging work which has b egun to challenge the universality of a unidimensional Cong ress. For example, Roberts, Smith an d Haptonstahl (2009, n.d. ) find that the predictive power of uni dimensionality is a function of aggregating roll call votes at the two year, Congress level. At the level of individual bills, they find that multidimen sionality is the norm for most major legislation in the House and Senate. As they note in the abstract, floor) a l ow dimensional space exists. Crespin and Rohde (2010) report evidence of NOMINATE scaling procedure for a subset of votes on appropriations legislation. Though their methodologically was very similar that the process of aggregating roll call votes at th e Congress level masks evidence of high dimensionality. In short, though the present analysis is
157 methodologically different than all three previous studies, the results support those albeit perhaps only indirectly, who find evidence of greater dimensiona lity at the bill level. 23 Simply put, there is a great deal of complexity in bicameral bargaining and reconciliation. Quasi Divided Government In a series of separate analyses I explore the conference patterns for instances of quasi divided government when the House and Senate are occupied by rival parties. In the post reform era, there have been four instances of quasi divided government (the 97 th 98 th 99 th and 107 th Congresses). I aggregated these periods into a separate sample given their theoretical distinctiveness from the entire time series Ultimately, because of the low sample size (n=35), there are some challenges to identifying each dimension and making clear inferences. Nonetheless, we can glean some meaningful results. Ultimately the substa ntive interpretation of our three dimensions remains the same even during periods of quasi divided government. However, the interpretation of the partisan dimension requires some additional discussion. As with earlier n egative values in the partisan dim ension indicate that the minority in both chambers gained favorable concession in conference. Given split party control, this indicates greater bill 23 I leave it up to the reader to decide the extent to which these findings confirm the earlier work. At worst, the evidence I report only tangentially supports these earlier authors. On the one hand, the earlier work looked at roll call votes on the actual policy content of the legislation. My work, by comparison, looks at dimensionality in the reconciliation domain (in particular policy shifts in conference). Though these are different legislative actions, they concern the same goal: implementing public policy for the social and/or economic betterment of society or electoral gain. Further, not hing in my methodology constrains the PCA results to be multidimensional. If conference outcomes were mostly ideological that result could, mathematically speaking, manifest in the results.
158 moderation, as it did previously. H owever, positive values in the partisan dimension indicate that the ma jority in both chambers gained favora ble concessions in conference. Ultimately, as we shall see outcomes loading onto the second are driven by pro minority shifts (negative values) that represent greater bill moderation and centrist outcomes Positive v alues are very rare. The results from the quasi divided government principal components analysis for the untransformed series are available in Tables 3 6 and 3 7 24 The component loadings in Table 3 6 reveal the presence of the same three main dimensions o f bicameral reconciliation reported previously. However, the interesting finding is that during quasi divided government partisan conflict is the first principal component. The oppositely signed loadings by party in the first component indicate that as o ne party increases its vote totals after conferencing, the other party decreases its vote totals. Moreover, the explained variation for this component (Table 3 6) is 47%. This is the highest proportion of explained variation among the results. The secon d principal component represents reconciliation (the first principal component over the entire time series). This dimension of resolving differences explains 30% of the variation in the shift parameters. However, it should be noted that two of the shift parameters in the second dimension have low loadings (.0349 and .0292 for the House majority and Senate minority respectively). This is likely a function of the low sample size (n=35) And finally, the third estimated component represents bicameral confl ict. The positive loadings for the House and negative loadings for the Senate indicate that, in this dimension, the vote share of each chamber moves in opposite directions after 24 The results from the transformed series are not substantively i nterpretable due to instability in the data from the low sample size and data transformation. Those results are not reported.
159 conferencing, irrespective of party differences. However, the loading for th e Senate minority is low ( .0261), probably as a result of the low sample size. This final dimension explains 18% of the variation in the shift parameters. It is important to keep in mind what, exactly, drives the increased salience of partisan conflict during quasi divided government. Though Chapter 4 takes a more comprehensive look at each dimension in isolation, some discussion is necessary here. As a technical matter, the increase in the explanatory power of this d imension represents a combination of greater pro majority shifts and greater pro minority shifts. But in this particular case, there are good reasons to suspect that the results stem from greater pro minority shifts. Because rival parties control each chamber, a pro minority shift indica tes a majoritarian (moderated) outcome. This outcome makes sense as each party has what amounts to a legislative veto over the policy content of any bill the 99 th Congress as an example. Assume that we have a game played by our four actors and that they have single peaked Euclidian ideal points that exist in a unidimensional policy space (liberal to conservative). Further assume (for simplicity) that the two chambers have majori tarian voting rules. Figure 3 4 presents an informal spatial model The House acts first passing a bill at its chamber median ( .122 denoted HB ). The Senate acts second, making a counteroffer located at its chamber median (.067 denoted SB ). In the initial round of the game, the House proposal is .468 away from the H ouse minority Republicans ( .122 .346 denoted HMin ) while the Senate proposal is .408 away from the Senate minority Democrats (.067 ( .341) denoted SMin ). Assume in the second round of the game a conference committee meets and
160 House and Senate bills ( .0275). The agreement after reconciliation is now .374 away from the House minority Republicans and .314 away from the Senate minority Democrats. Thus, under these condi tions we would expect the minority in each chamber, despite the fact that they are of the opposite party, to gain favorable concessions in co nference. As empirical verification of this spatial model and the prior argumentation the shift parameters for o ur four actors during periods of quasi divided governmen t where the principal components load 25 onto the partisan dimension are .065 and .02 for the House and Senate majority (respectively) and .15 and .16 for the House and Senate minority (respectively). This indicates that partisan conflict in conference during periods of quasi divided government is driven by greater pro minority shifts. Discussion Little work has explored how, exactly, the House and Senate resolve disagreements when they arise. Furth ermore, the body of existing research on this topic is of limited scope (though see Longley and Oleszek 1989). Indeed, existing work on resolving differences explores this stage in the legislative process as a means of discriminating between competing the oretical perspectives of legislative organization looking almost exclusively at the tension between competing chambers or the explanatory power of distributive, majoritarian or information al theories. Thus, the conceptual dimensions of resolving differenc es have been explored in isolation. More problematic is that existing research has overlooked what is the central feature of inter 25 I used any bill that loaded onto the second dimension at .5 or higher.
161 chamber bargaining consensus and compromise (but see also Ferejohn 1975; Manley 1970). In this chapter I developed a unif ied typology of how the House and Senate resolve differences; one that I hope contributes to our broader understanding of the process. Using roll call data from all conference committees convened from the 95 th to the 110 th Congresses, I operation a lized th e typology using multidimensional spatial modeling. There are two primary findings. First, I found evidence of multidimensionality in the process of resolving differences. In particular, I find that three qualitatively distinct dimensions reconciliation partisan conflict and bicameral conflict explain over 80% of the variation in the conference committee roll call patterns. Furthermore, the results revealed that no single dimension dominates the process. This provides some much needed conceptual clari ty concerning how the House and Senate resolve differences. However, in providing these descriptive results the findings sup port an emerging view that roll call voting in both chambers contain s greater dimensionality tha n most authors acknowledge ( Crespin and Rohde 2010; Roberts, Smith and Haptonstahl 2009; Talbert and Potoski 2002). Second, I find that the first dimension of resolving differences (i.e. the one that explains the greatest amount of variation) is a process of reconciliation defined here as (Nagler 1989; Carson and Vander Wielen 2002; Vander Wielen and Smith n.d.; Lazarus and Monroe 2007) I do not find that partisanship is the leading determinant of post passage bargaining. This finding is particularly consequential when juxtaposed with
162 findings concerning the growing salience of bicameral hurdle faced by the majority party Though this chapter has described the patterns of resolving differences, and applied those descriptive results to theoretical debates about congressional organization and roll call voting, we have not explored the patterns of resolving differences with e ach dimension in isolation For example, what explains the ability of conference negotiators to engage in successful reconciliation? Is the second dimension driven by pro majority shift s or pro expense of t he other, as initial researchers of conference committees wondered? And finally, how have the patterns of resolving differences cha nged over the postreform period? Chapter 4 explores these questions in detail
163 Figure 3 1. Scree plot for the untransfo rmed data Figure 3 2. Scree plot for the untransformed data
164 Figure 3 3. 3 D spatial map of resolving differences The dots represent how conferees modified the original House and Senate passed bills according to the princ ipal components analysis reported in Table 3 2 the initial roll call patterns. HB CA SB HMaj SM in SMaj HMin Figure 3 4. Spatial represe ntation of the 99 th Congress Spatial locations based on (Democrats), Hmin is the median of the House minority party (Republicans), SMaj is the median of the Senate maj ori ty party (Republicans) and SM in is the median of the Senate minority party (Democrats). HB represents a bill passed at the House median and SB represents a bill passed at the Senate median. CA represents a bill reconciled at the midpoint of the House and Senate bills.
165 Table 3 1. A typology of resolving difference s Shift Parameter P i C j House Majority House Minority Senate Majority Senate Minority Typology + + + + Reconciliation 0 0 0 0 + + Partisan Conflict (Zero Su m) + + + 0 + 0 Partisan Conflict (Positive Sum) 0 + 0 + 0 0 0 0 + + Bicameral Conflict (Zero Sum) + + + + 0 0 Bicameral Conflict (Positive Sum) 0 0 + + 0 0 0 0 Table 3 2. Component loadings (untr ansformed data, 95 th to 110 th ) Shift Parameter Dim ension 1 Dim ension 2 Dim ension 3 House Majority .4453 .6078 .3688 House Minority .4848 .4439 .6662 Senate Majority .5828 .3571 .4779 Senate Minority .4765 .5532 .4379 Interpretation: Bicamera l Reconciliation Partisan Conflict Bicameral Conflict Notes: N=206 The loadings were estimated via principal components analysis. The component loadings represent the correlation between the shift parameter and the dimension The magnitude and di rection of the loadings help us interpret the dimension.
166 Table 3 3 Component loadings (transformed data, 95 th to 110 th ) Shift Parameter Dimension 1 Dimension 2 Dimension 3 House Majority .4778 .6959 .0811 House Minority .5459 .0532 .6531 Senate Majority .5042 .0613 .7522 Senate Minority .4686 .7136 .0342 Interpretation : Bicameral Reconciliation Partisan Conflict Bicameral Conflict Notes: N=206 The loadings were estimated via principal components analysis. The component loadings r epresent the correlation between the shift parameter and the dimension. The magnitude and direction of the loadings help us interpret the dimension. Table 3 4 Explained variance by dimension (untransformed data, 95 th to 110 th ) Dimension Eigenvalue Ex plained Variance SE 95% Confidence Interval Dimension 1 Reconciliation 1.37 .34 .03 .29 .39 Dimension 2 Partisan Conflict 1.14 .29 .02 .24 .33 Dimension 3 Bicameral Conflict 0 .84 .21 .02 .17 .25 Notes: N=206. The explained variances were estima ted via principal components analysis. T test Dim1 to Dim2 = 1.586 (p=.056); T test Dim2 to Dim3 = 2.4307 (p<.01); T test Dim1 to Dim3 = 4.113 (p<.001) Table 3 5. Explained variance by dimension (transformed data, 95 th to 110 t h ) Dimension Eigenva lue Explained Variance SE 95% Confidence Interval Dimension 1 Reconciliation 1.53 .38 .03 .33 .44 Dimension 2 Partisan Conflict 0 .99 .24 .02 .20 .29 Dimension 3 Bicameral Conflict 0 .89 .22 .02 .18 .26 Notes: N=206. The explained variances were estimated vi a principal components analysis. T test Dim1 to Dim2 = 3.9550 (p<.001); T test Dim2 to Dim3 = 0.8378 (p=.20); T test Dim1 to Dim3 = 4.7947 (p<.001) All significance tests are two tailed.
167 Table 3 6. Component loadings (untransformed data 97 th to 99 th and 107 th ) Shift Parameter Dimension 1 Dimension 2 Dimension 3 House Majority .6549 .0349 .3993 House Minority .2152 .7143 .6311 Senate Majority .2163 .6983 .6645 Senate Minority .6913 .0292 .0261 Interpretation : Partisan Conf lict Bicameral Reconciliation Bicameral Conflict Notes: N=35 The loadings were estimated via principal components analysis. The component loadings represent the correlation between the shift parameter and the dimension. The magnitude and directio n of the loadings help us interpret the dimension. Table 3 7. Explained variance by dimension (untransformed data, 97 th to 99 th and 107 th ) Dimension Eigenvalue Explained Variance SE 95% Confidence Interval Dimension 1 Partisan Conflict 1.86 .47 .07 .32 .61 Dimension 2 Reconciliation 1.21 .30 .06 .18 .43 Dimension 3 Bicameral Conflict 0 .71 .18 .04 .09 .27 Notes: N=35. The loadings were estimated via principal components analysis. T test Dim1 to Dim2 = 1.7230 (p<.05); T test Dim2 to Dim3 = 1.6 547 (p<.10); T test Dim1 to Dim3 = 3.4649 (p<.001) All significance tests are two tailed.
168 CHAPTER 4 RESOLVING DIFFERENCE S IN TIME AND SPACE: MODELING BILL LEVEL CONFERENCE OUTCOMES Chapter 2 examine d the frequency and severity of House and Senate c onflict over the postreform era. One of the more consequential findings of Chapter 2 is that the House and Senate have experienced more fr equent gridlock and more severe ideological disagreements when trying to enact policy in the latter half of the postr eform period The results linked this trend to growing compositional asymmetries between the two chambers. These compositional asymmetries include greater distance between intra party bicameral differences and a g rowing gap between the House median and Senate filibuster pivot. The evidence also showed that bicameral gridlock has become increasingly prono unced in the contemporary era for partisan House initiated legislation and that the primary determinant of this effect has been growing intra party bicameral cleavages T he overarching narrative given these findings is that the enhanced organizational capacities of the House and Senate majority have coincided, oddly, with a growing that dispropor tionally constrains the majority Chapter 3 then explored the next logical stage in the policy process: How the House and Senate resolve disagreements when they arise. Using multivariate spatial modeling I identified the conceptual features of resolving differences using pre and post conference roll call votes. There were two primary findings. First, the process of resolving differences is multidimensional. That is, three qualitatively distinct dimensions reconciliation, partisan conflict and bicamer al conflict explain over 80% of the variation in conference committee outcomes with no single dimension capturing a majority of the explained variation On the one hand these findings provide some
169 conceptual clarity concerning how, exactly, the House and Senate resolve differences. Second, the results in Chapter 3 showed that the first dimension of resolving differences (i.e. the one that explains the greatest amount of variation) is a process of compromise and concession. Thus, contrary to the directio n of the literature, I do not find that partisanship is the leading force structuring conference committee outcomes (though it certainly plays an important role) But despite the research and findings reported in Chapter 3, we are yet to explore the pat terns of resolving differences in great detail. Indeed, Chapter 3 examined the macro level patterns of resolving differences. Though these aggregate patterns are useful, especially given the dearth of research in this area, it is w orthwhile to examine co nference outcomes at the bill level The present chapter takes up this issue by : (1) examining postreform changes in policy space governing re solving differences ; (2) presenting bill level conference outcomes using multidimensional spatial coordinates; an d (3) modeling the determinants of bill level conference outcomes. Reestimating the Resolving Differences Policy Space by Era This section examines historical patterns in the process by which the House and Senate resolve differences in conference. As note d at various points in this dissertation, little research exists on bicameral disagreement and reconciliation. This is especially true when we consider temporal trends in how the House and Senate meld competing policy proposals In a recent chapter discu ssing areas for needed research,
170 The multivariate spatial models estimated in Chapter 3 revealed that resolving differences is a multidimensional process. That is, the policy space in which parties and chambers resolve bicameral disputes is marked by three qualitatively distinct dimensions consensus and compromise, partisan conflict and bicameral conflict (in that order, from most to least e xplanatory power). To examine if and how this process has changed over time I re estimated the dimensions for the pre and post Republican Revolution pe riods. On the one hand, the Republican Revolution was selected as the cut point because it marks a period of heightened partisanship ( Bond and Fleisher (Critchlow 2004; Dodd and Oppenheimer 2005). But more important for the present bargaining in particular. During a series of interviews conducted as part of this project senior committee sta ff described an informal norm developed during the 104 th Congress 1 According to appoint ed conferees. Rather as is commonly thought, Gingrich was able to influence conference outcomes by appointing smaller, more predictable sets of House managers Thus, to identify historical developments in resolving differences I fol lowed methodology with the sole difference being that, this time, I estimate d separate policy spaces for the pre Republican Revolution (95 th to 103 rd Congresses) 1 It was suggested that this procedure is attributable t o Dick Armey (R TX).
171 a nd the post Republican Revolution (104 th to 110 th Congresses) era s The interest ed rea der should refer to Chapter 3 for the methodological details The component loadings for the untransformed pre Revolution data are contained in Table 4 1 while the component loadings for the untransformed post Revolution data are contained in Table 4 2. The component loadings for the transformed pre Revolution data are contained in Table 4 3 while the component loadings for the transformed post Revolution data are contained in Table 4 4. The corresponding estimates of explained variation by compon ent are reported in Tables 4 5, 4 6, 4 7 and 4 8. For comparability across time, periods of quasi divided government were not included in these analyses 2 Unfortunately, because of instab ility across the transformed and untransformed estimates in the pre Revolution period we are unable to make definitive conclusions about the explanatory power of the dimensions across time 3 Nonetheless, the descriptive statistics are unbiased (StataCorp 2009) and shed light on some temporal patterns. 4 Further, we can s afely discuss the amount of explained variation in the post Revolution sample and make statistical comparisons 2 We would get a biased look at the two period with these Congresses included in the samples as three periods of quasi divided government occur in the pre Revolution sample compared to only one in the post Revolution sample. 3 Indeed, recall that the virtue of reporting the results of the transformed data series is that it produces less biased results. 4 Recall from the previous chapter that descriptive principal components results are unaffected by violations of multivariate normalit y. But because the inferential results are potentially biased (namely the confidence intervals around the estimates of explained variation), I include a second set of estimates from a transformed dataset. Comparing the estimates for these two series reve als some contradictory results. While reconciliation is identified as the first dimension in both datasets across time, partisan conflict and bicameral conflict switch places as the second component in the pre Revolution transformed and untransformed samp les. In both cases, the third dimension is uninterpretable. Since these estimates are meant as a descriptive look at the process of resolving differences, I rely on the untransformed series. Given the
172 According to the untransformed data, reconciliation is the leading dimension of resolving differences in both the pre Revolution and post Revolu tion periods while partisan conflict is the second leading dimension in both pe riods as well ( Tables 4 1 and 4 2 ). These results are identical to those reported in Chapter 3 where the policy space was scaled over the entire postreform period. Thus, even in the contemporary Congresses where partisan roll call patterns are more pronounced (Poole and Rosenthal 1997, 2006) it is telling that resolving differences remains a process of consensus an d compromise first and foremost. This an important result in t erms of our ability to characterize the overall process of resolving bicameral disputes as well as the claims about how conference outcomes have changed over time. It is important to point out that there is instability a cross the untransformed and transformed data sets in the pre Revolution period. 5 Thus, we are unable to draw definitive conclusions about the explanatory power of each dimension across time. However, in the post Revolution sample we can examine the amoun t of explained variation in the transformed dataset. In Table 4 8 we can see that though the amount of explained variation by the reconciliation dimension is larger than the partisan dimension, that difference is only significant at the .10 level. Given that this difference was significant at the .05 level over the entire time period (repo rted in Chapter 3 ), this result suggests that the partisan dimension has increased in salience over time (though instability in the results we are unable to make def initive conclusions about the explanatory power of the dimensions in relation to one another or over time. 5 Though reconciliation remains the leading component in both series, the second dimension switches between bicameral conflict and partisan conflict (and in transformed pre reform series the third dimension is not interpretable).
173 perhaps only modestly). The subsequent section will tak e up this issue in greater detail as this can be evidence of growing pro majority outcomes, pro minority outcomes or both One interesting postreform development is evident if we compare the loadings for the third dimension in our two time periods. When we modeled the policy space for the entire time period in Chapter 3 the third component was identified as bicameral conflict. This was revealed by the oppositely signed loadings by chamber. In the post Revolution untransformed sample (Table 4 2 ) this fin ding holds. However, in the pre Revolution untransformed sample the third dimension exhibits instability. Notice that in Table 4 1 the component loadings for the Senate are both negative and modestly sized while the loading for the House majority is also negative (but close to zero). For unambiguous evidence of bicameral conflict the latter loading should be positive and modestly sized along with its House minority counterpart (as it is in the post Revolution sample). Though this feature may be a manife station of the smaller sample size, it suggests that in the first half of the postreform period bicameral conflict was not as salient as during the second half of the postreform period. Though far from definitive, this finding supports conclus ion that conflict between the House and Senate has increased in salience over the postreform period Taken as a whole, the findings from this section suggest that the historical patterns in conference outcomes exhibit stability across the two periods. Tho ugh roll call voting in Congress is more reliably partisan in the latter half of the postreform period, resolving differences remains first and foremost a process of compromise and concession. Nonetheless, there is some limited evidence that partisan conf lict and bicameral conflict
174 have increased in salience over the postreform period. A closer inspection in the next sect ion will reveal what exactly, these developments tell us. The Winners and Losers: Mapping Conference Outcomes T o explore conference out comes at the bill each successful conference committee in the dataset These scores, discussed briefly at the close of the Chapter 3 identify a point in the three dimensional resolving differences coordinate sys tem corresponding to the dimension or dimensions that explain the bill level conference outcome ( Rabe Hesketh and Everitt, 2007) This is in contrast to the earlier multivariate results which looked solely at the aggregate patterns of resolving difference s. This methodology allows us to create a visual map of the shift parameters a policy effects of a conference committee. Recall from Chapter 3 that in the first dimension negative values indicate that our four actors had a decrease in their policy utility after a conference committee while positive values indicate that our four had an increase in their policy utility In the second while positive values indicate the Fi gures 4 1 and 4 2 map all conference outcomes in three dimensions for the pre Revolution and post Revolution periods (respectively). 6 At the end of Chapter 3 these two periods were combined into a single Figure to observe dimensionality in resolving diffe rences. I scaled the values in each Figure so that the origin (the point at X=0, Y=0, 6 As with earlier, both figures reported here were created using MATLAB.
175 Z=0) has a real interpretation it is the point in the space where all shift parameters equal zero 7 Thus, the origin indicates that the changes made during conference n egotiations had absolutely no affect on the initial House and Senate roll call pat terns for our four actors (the House and Senate majority and minority) This typically occurs when the c hanges made in conference were very minor relative in scope to the fu ll bill. As we saw in Chapter 3 in Figure s 4 1 and 4 2 most of observations are located around the origin of the three dimensional coordinate system. This indicates that major changes to bills reconciled by conference committees are less frequent than minor changes. And as with earlier we can see visual evidence of multidimensionality in the spatial scores. However, even a cursory examination reveals an obvious difference in conference outcomes across these two periods In the post Revolution sample (Figure 4 2) there is greater variability in the spatial coordinates compared to the pre Revolution sample (Figure 4 1). That is, where conference outcomes are located around the origin (0,0,0) in the pre Revolution Figure in the post Revolution Figure t here ar e a greater number of points in three dimensions toward the edges of the coordinate system. Substantively, this increased variability suggests that the process of resolving differences (writ large) is affecting greater modifications to the initial House and Senate proposals in the contemporary Congresses as compared to thirty years ago Where minor changes to bills passed by both houses typically satisfied competing chambers and parties before the Repub lican revolution, larger modifications are nec essary to satisfy our various actors in the contemporary period. This is evidence that the process 7 Fortunately an observation in the sample has four usable roll call votes and identical pre and post observations.
176 of resolving differences is under greater strain in the modern Congresses, supporting the notion that competing parties and chambers have come into greater conflict over this period. To assess whether this increased variability is statistically meaningful from the origin using the formula: Higher values indicate greate r distance from a point in the policy space that indicates no change in the passage stage roll call vote. According to the data reported in Figures 4 1 and 4 2, the average Euclidian distance from the origin in the pre Revolution sample is 1.02 whereas in the post Revolution sample the Euclidian distance from the origin is 1.54. 8 With standard errors of .12 and .13 (respectively) the difference between the two periods is statistically significant at the .01 level. Thus, we can safely conclude that con ference committees are affecting greater changes to pre conference House and Senate passed legislation in the post Revolution period We can obtain a more refined look at patterns in the po licy space for both periods by parsing the three dimensional Figu re into a series of three two dimensional plots. Figure 4 3 presents these plots In the sub sections (representing the pre and post Revolution periods) the top Figure plots reconciliation (X axis) against partisan conflict (Y axis), the middle Figure p lots reconciliation (X axis) against dimension bicameral conflict (Y axis) and the bottom Figure plots partisan conflict (X axis) against bicameral conflict (Y axis). Notice that in these figures not all reconciliation outcomes are located exclusively on one dimension or the other; some bills are located in the corners of the 8 These values have no substantive interpret ation.
177 coordinate system. This highlights the usefulness of multidimensional spatial modeling in this application al confli two or more dimensions. For example, bills located in the upper left corner of the first plot in Figure 4 3 are those that experienced reconciliation in the negative direction and partisan confl ict in the positive direction. Substantively, this indicates that all actors sacrificed provisions in conference (reconciliation), with the minority party sacrificing the greatest share of those provisions ( a pro majority outcome). Conversely, bills loca ted in the lower right corner of the first plot in Figure 4 3 experienced reconciliation in the positive direction and partisan conflict in the negative direction. Substantively, this indicates that all actors gained provisions in conference (reconciliati on) while the minority party gained the greatest number of provisions relative to the majority (pro minority outcome). Again we can see greater variability in the conference outcomes in the post Revolution mappings. The additional bit of information revea led by the two dimensional maps is that we can see that the increased variability over time appears to occur in all three dimensions simultaneously. Though these two dimensional mappings provide a more refined look at each dimension relative to the three dimensional mapping, additional trends are challenging to spot. The only other trend apparent in these figures is a shift over time in the partisan dimension (dimension 2) in the negative direction Though there is increased variability in the second dim ension over time (including larger values in the positive direction), the observations in the negative direction appear to be both greater in proportion and greater in magnitude. This suggests that in the contemporary Congresses there are more frequent an d more consequential pro
178 minority policy modifications made during conference negotiations. This is an intriguing finding given the increasing strength of parties over the post reform period. In other words, it is counter to what the conventional wisdom suggests. Finally, t o look in even greater detail at patterns of resolving differences over the postreform period, I estimated the median coordinate within each dimension for every Congress in the sample (95 th to 110 th ). Because of the unique nature of reconciliation I also estimate the median absolute coordinate for this dimension. The median coordinate reveals in which direction resolving differences has shifted within each dimension over time (if any) while the median absolute coordinate in the firs t dimension reveals trends in reconciliation writ large ( whether it be greater compromises or greater concessions ). The temporal patterns are presented in F igures 4 4, 4 5 4 6 and 4 7 Also included is a linear trend. T wo trends are apparent in th ese fi gures First, and as noted earlier, there appears to be a shift over time in the partisan dimension in the negative direction. That is, the negative values in the latter half of the time series i n Figure 4 6 indicate that resolving differences is increas ingly generating pro minority policy outcomes in the contemporary period. The second trend is an increase in absolute value of reconciliation (Figure 4 5) This suggests that resolving differences is affe cting greater compromises and concession s in the c on temporary Congresses Note that there does not appear to be any systematic change in the directio n of reconciliation (compromise or concession ) as evidenced by Figure 4 4 There also do not appear to be any patterns in bicameral conflict over the postr eform period as evidenced by Figure 4 7 Though earlier results suggest that bicameral conflict has increased in salience over time, these
179 results suggest that this trend has not occurred in one direction (pro House or pro Senate) exclusively. Overall, these two findings (along with the increased variability noted earlier) fit into a developing narrative that over the postreform period the so This fits the overal l narrative presented in Chapte r 2 But i n contrast to Chapter 2 which found a similar growth in the bicameral hurdle with respect to bicameral gridlock and bicameral disagreement, here th e effects of this hurdle are growing challenges resolving differences (including greater reconcil iation and pro minority conference outcomes) Taken as a whole there is a latent symmetry to the historical patterns in bicameral disagreement and reconciliation. Determinants of Conference Outcomes Finally, we can use the principal component scores prese nted in the previous figures to model the determinants of conference committee outcomes. This approach follows the conventional lens though which prior work has assessed conference committees and bicameral bargaining. Recall from the review of the litera ture that the earliest conference committee researchers (e.g. Steiner 1951; Fenno 1966; Manley multivariate spatial models ident ified bicameral conflict as the third dimens ion (in order from most to least important) the results confirm that conflict between the House and Senate is indeed a salient characteristic of resolving differences. More recent work has addressed a different question: Which party rence? The conventional wisdom is that the majority party possesses a kind of
180 attributed to the fact that the Speaker and Majority Leader wield the power to name conference managers (Lazarus and Monroe 2007 Vander Wielen and Smith n.d.). Thus, the majority has the ultimate power to amend or kill legislation in conference. And finally, this dissertation argues that the process of reconciliation marked by consensu s and compromise rather than conflict plays a critical role as well. In fact, unlike prior work on conference committee outcomes, I have shown that reconciliation is the leading dimension or pattern in resolving differences. But because the previous resu lts showed that the process of resolving differences is multidimensional, it does not make sense to Rather conference, we will be looking the c onditions under which one actor wins at the expense of the other. This allows us to assess conference c ommittee outcomes with respect to conventional theoretical debates about legislative organization. Modeling the First Dimension: Reconciliation The multivariate spatial model reported in Chapter 3 revealed a first dimension marked by reconciliation where positive coordinates indicate a conference committee outcome where all four actors increased their vote share ( sometimes called a logroll) while negative values indicate all four actors d ecreased their vote share (a moderated bill). These coordinates serve as the dependent variable in this section. However, because we are interested in greater or lesser reconciliation, irrespective of the particular compromise reached, we want to model t he absolute value of this dimension. 9 9 The raw coordinates were analyzed using the covariates listed in this section. Such an approach models whether House and Senate negotiators produced a compromise where all sides lost or gained vote share (i.e. why conferees engage in log rolling versus bill moderation). The effect of Budget approached statistical significance and was positive, indicating that during times of budgetary surpluses all four sides gain favorable provisions in conference negotiations while during times of budgetary def icits all four sides lose favorable provisions in conference
181 T he response is therefore coded such that higher values indicate greater changes were made to the original pre conference bills (greater compromises or greater concessions) while lower values indicate fewer conference changes. Because of this transformation the response is not normally distributed. Thus, t o properly model the data I estimated a series of generalized linear models (GLMs). After multiple specifications it was determined that the best fit was provided by a gamma family and log link function. Because the residuals are heteroscedastic, robust standard errors were used. 10 Two bill conference proposal. Midpoint Divergence captures ideolog ical differences in the (1998) common space scores. Recall that this factor was the de pendent variable bicameral disagreement in Chapter 2 The expected effect is negative, indicating that when the House and Senate pass ideologically different bills conferees are forced t o make greater compromises and concessions in order to create a work able solution (i.e. greater overall reconciliation). The second bill level covariate, Partisan is an index of added for both chambers. 11 Substantively, this facto conference formal and informal powers over House and Senate outcomes. A value of negotiations. Though this effect makes theoretical sense, it was insignificant. Ultimately, greater work is needed exploring the determinants of compromise versus concession. 10 This was determi ned by a Breusch Pagan/Cook Weisberg test ( hottest in STATA 10 ) 11 http://www.voteview.com/
182 2.0 indicates that both chambers passed perfectly pro majority partisan bills while a value of 0.0 indicates that both chambers passed per fectly bipartisan bills. 12 Negative values indicate both chambers passed pro minority pre conference bills. From the perspective, if conference committees are easily manipulated by the majority (as the conventional wisdom holds) we would expect a neg ative effect, indicating that for pro majority bills conferees enact few changes thereby preserving the initial House and Senate passed bills In other words, if the pre conference bills passed by both the House and Senate are party median (e .g. Aldrich and Rohde 1997 ; Rohde 1991; Cox and McCubbins 1993, 2005) little to no change in conference reflects a pro majority outcome. However, if conference committees produce majoritarian or centrist outcomes the expected effect is positive, indicatin g that partisan bills experience greater reconciliation in conference. The remaining variables tap the effect of institutional and contextual factors on patterns of reconciliation in conference. The variables Divided Government Budget Party Mandate an d Mood Lag are the same as those used in Chapter 2 (Binder 1999, 2003). 13 The effect of divided party control is expected to be positive. Such an effect 12 he majority party voted 55 0 and 255 0 in 0 and 180 0 against majority voted 28 27 a nd 128 voted 23 22 and 90 Figure is an 435 members of the House vote. 13 Divided Government Budget measures the size of the federal governme outlays. The variable Party Mandate records the number of prior Congresses a new majority was in the minority averaged for both chambers. And Mood Lag public mood policies.
183 committee is typical ly the last stage in the policy process where changes can be made), the hurdle created by legislative executive checks and balances has a greater effect on conference negotiations. Thus, I expect that divided government is associated with greater reconcil iation. The remaining three effects are expected to yield negative effects. When the budgetary situation is favorable, the majority has a popular mandate and the public mood favors activist government, we would expect conferees to affect fewer changes to the original House and Senate bills. Finally, I include the effects our three sources of bicameral co nflict identified in Chapter 2: Bicameral Distance Intra party Bicameral Distance and Filibuster House Distance Recall that because the distance betwe en the parties across chambers and the filibuster pivot distance are so highly correlated we must estimate these effects in separate models (Models 1 and 2, respectively) Consistent wi th the findings from Chapter 2 the expectation is that these factors representing compositional differences between the House and Senate a re associated with greater reconciliation as they create bicameral hurdles which conferees must overcome. However, it is very likely that these effects are attenuated by the bicameral d isagreement measure ( Midpoint Distance ) Indeed, Chapter 2 showed us that these three factors cause greater bicameral disagreement at the passage stage Thus, it is very likely that these factors affect reconciliation by creating greater bicameral disagr eements which, in turn, affects the need for confer ees to propose compromises and concessions. Including these factors will determine if compositional asymmetries have independent effect s on
184 resol ving differences or whether their effects are simply antece dent (funneled through bicameral disagreement) The GLM estimates of reconciliation are contained in Table 4 9. Overall the models perform adequately as they explain between 12% and 11% of the variation in reconciliation outcomes In both models the effe ct of divided government is statistically significant and positive, as expected. This indicates that in conference negotiations the president plays an important role in fostering compromises and concessions. We know from historical accounts as well as pr ior work on that sometimes the president often plays a direct role in negotiating with House and Senate conferees as well as House and Senate leaders (Longley and Oleszek 1989). influence can be indirect as legislatio n in for his signature or veto) Thus, legislative executive constraints play a role in how conferees resolve disagreements (whether direct or indirect) The results in both models also show that the ideological distance between the pre conference House and Senate bills increases the magnitud e of reconciliation outcomes. This is as expected. Substantively, when the House and Senate pass ideologically different bills before a conference is con vened, greater compromises and concessions are needed to resolve disagreements. This is consequential because it provides a clear answer to the increase in reconciliation outcomes over the postreform period. This increase can be attributed, at least in part, the growing pre conference House and Senate c onflict (reported in Chapter 2 ). However, t he effect of our sources of bicameral conflict Bicameral Distance Intra party Bicameral Distance and
185 Filibuster House Distance are all positive (as expected) but insignificant. As was speculated at the outset, these factors affect the nature of the initial pre conference disagreement and this factor affects the nature of resolving differences. Thus, the significant positive effect of bi cameral disagreement captures the effect of increasing compositional asymmetries on reconciliation outcomes. Compositional asymmetries, in other words, have an antecedent effect in this domain. The remaining factors are all insignificant. This includes t he effect of Partisan which was considered to be a n important factor prior to the estimation U ltimately there is no evidence to support either the partisan or majoritarian perspectives when it comes to determining reconciliation outcomes. Of course, th is effect is most likely to manifest in the partisan dimension. Modeling the Second Dimension: Partisan Conflict The multivariate spatial model reported in Chapter 3 revealed a partisan dimension where positive coordinates indicate a pro majority conferen ce outcome while negative coordinates indicate a pro minority conference outcome. These coordinates serve as the dependent variable in this section. Because the response is approximately normally distributed, the estimates were derived via OLS. And b eca use the residuals are heteroscedastic, robust standard errors were used. 14 Both bill level factors reported previously are included in the forthcoming models as there exist strong theoretical expectations that they have important effects on partisan outcom es. Midpoint Divergence is expected to yield a negative effect, indicating that when the House and Senate pass ideologically different bills the majority 14 This was determined by a Breusch Pagan/Cook Weisberg test (hottest in STATA 10)
186 party sacrifices favored provisions in conference (resulting in a pro minority outcome) On the effe ct of Partisan if conference committees produce majoritarian outcomes the expected effect is negative, indicating that partisan bills experience greater pro minority conference outcomes. The partisan perspective, on the other hand, would predict either a positive effect or a null effect. 15 Thus, the majoritarian perspective has a much larger hurdle to overcome. In two auxiliary models I disaggregated Partisan into each House Partisan and Senate Partisan ). I also includ e an interaction between these two covariates ( X Partisan ). This interaction modeling strategy allows us to explore differences across the two chambers as well as model the effect of partisan divergence other chamber passes a bipartisan bill) Indeed, we know that the House passes partisan legislation to greater extent than the Senate (where bills routinely pass with supermajorities). ural power over legislation at the passage stage (either in committee or on the floor), two additional committee outcomes. For example, Lazarus and Monroe (2007) maintain that when the Speaker has reason to believe a conference delegation composed solely of jurisdictional committee members will produce a party damaging conference report he or she uses his or her appointment power to select a preferable conference delegation. Co nsistent with modern partisan theories of lawmaking, this strategic maneuver, 15 The expectation of a null effect follows the logic that if a conference commi ttee resolves differences by making no change to the initial House and Senate bills the aggregate outcome is pro majority (as indicated by the pre conference roll call vote).
187 maintains the norm of naming conferees that are familiar with the bill while simultaneously allowing the Speaker to put the party stamp on legisla tion. Recent work by Vander Wielen and Smith (n.d.) finds evidence that Senate conference committees are similarly biased. Thus, following these studies the variables House Outsiders and Senate Outsiders record the percentage of non jurisdictional commi ttee conferees serving in conference for the House and Senate. This was achieved by, first, identifying the standing committee(s) that received referral of the bill and, second, coding whether each named conferee was a member of this committee(s). 16 Highe and, presumably, a pro majority conference outcome It is important to point out that this prior work (Lazarus and Monroe 2007; Vander Wielen and Smith n.d.) examines the conference delegations as the unit of analys is rather than the actual conference outcome Thus, the link between packing the conference and a pro majority outcome has been inferred rather than empirically verified. This study is the first to take this critical next step. The remaining variables ta p the effect of institutional and contextual factors on patterns of partisan conflict in conference. As with earlier, I include Divided Government Budget Party Mandate and Mood Lag The effect of divided party control is expected to be negative. Such an effect indicates that when the president is of the rival party to the majority the hurdle created by legislative executive checks and balances yields pro minority conference outcomes. The remaining three effects are expected to yield positive effects. When the budgetary situation is favorable, the 16 th to 102 nd Congresse s and Stewart rd to 110 th Congresses. Both datasets are publically available http://web.mit.edu/17.251/www/data_page .html
188 majority has a popular mandate and the public mood favors activist government, we would expect the majority to wield greater power over conference outcomes. Finally, I also include our measures of House and Senate compositional differences : Bicameral Distance Intra party Bicameral Distance and Filibuster House Distance Recall that because the distance between the parties across chambers and the filibuster pivot distance are so highly correlated we must es timate these effects in separate models (Models 3 and 4, respectively) And a s was discussed earlier, there is reason to believe that the bicameral disagreement measure will attenuate any effects of these variables as Chapter 2 showed that the latter caus es the former Nonetheless, the expected effects are all negative indicating greater pro minority conference outcomes as the Table 4 10 contains the main estimates while Table 4 11 contains the estimates disag gregating Partisan into its constituent parts and interacting them All four models perform very well as they explain 20% to 25% of the variation in pro majority conference outcomes. In all four models the effect of divided government is statistically s ignificant and negative, as expected. This effect indicates that in conference negotiations the president plays an important role in terms of structuring partisan outcomes. In some cases the majority shepherds a bil l through the House and Senate knowing that the bill(s) cannot survive a presidential veto with the intention of moderating the proposal during the post passage stage (Longley and Oleszek 1989) Other times the president plays a more direct role, negotiating with conferees over the compromise policies (Longley and Oleszek 1989). Whether the influence is direct or indirect, when the
189 president is of the rival party conference outcomes shif t in a pro minority direction as we would expect I also find that the effect of party mandates is positive and significant, also as expected. This effect shows that House and Senate majorities who were in the minority for an extended period of time wield greater power over conference outcomes (due to a unified agenda and popular mandate) Such was the case w ith Republicans in 104 th Congress. 17 Though frequently vetoed by Clinton, Republicans were able to draft a number of partisan conference outcomes (as evidence d by the o riginal shift parameters). The Work Opportunity Act of 1995 (HR 4) is one example. 18 One of more consequential results in Tables 4 10 and 4 11 are evident with conventional wisdom I find no evidence in any of the four models that the procedure majority outcome. The percentage of House outsiders has a negative effect on pro majority outcomes in all four models, contrary to popular accounts. And though the percentage of Senate outsiders has a positive effect i n all four models the coefficients are small relative to their associated standard errors. This is in direct contrast to the conventional wisdom which maintains that the power afforded the Speaker and Senate Majority Leader over the composition of confer ence committees affords the majority party greater influence over bicameral 17 I have repeatedly argued that the Republican Contract With America was an example of significant bicameral conflict and, ultimately, had mixed successes. Thus, it is important to remind the reader that this effect captures all party mandates, not just the 104 th Congress. 18 The Work Opportunity Act of 1995 loads strongly onto the second dimension because Senate Democrats defected en masse from the conference report on HR 4. Where eleven Democrats voted against HR 4 at the passage stage, 45 voted a gainst the conference report.
190 bargaining. Recall that these prior studies tested delegation under the assumption that such strategic behaviors affect outcomes in the partisan direction. This study is the firs t to directly test this presumed link, finding no evidence to support it. conference agenda control has a negative effect on conference outcomes. The results unequivocally support a majoritarian perspective. In all four models the covariates indicating how partisan the House and Senate bills were at final passage are all neg ative and statistically significant. This indicates that partisan legislation is moderated in conference, producing a pro minority outcome. In models three and four in Table 4 10 the effect of Partisan is negative and significant, while in models 5 and 6 in Table 4 11 each of the individual terms ( House Partisan and Senate Partisan ) are negative and significant as well. Thus, partisan legislation is biased in a centrist direction in conference. Of course, a counterclaim to the previous discussion is that the opposite direction of the prior effects is evidence for a pro majority conference outcome: that bipartisan legislation emerges from conference shifted in a non centrist or partisan direction. L ogically, this argument undermines itself. Indeed, such an argument forces us to assume that the majority has little power over chamber outcomes (i.e. at the passage stage) in order to derive the logic of a partisan conference outcome. In other words, the necessary argument would be that parties are strong in conference but not in standing committees or on the House and Senate floor s Nonetheless, this interpretation is rejected by the data as well. First we need to describe the variable Partisan and
191 calculate its predicted effects on partisan outcomes. Figu re 4 8 presents a histogram of the partisan passage data. The mean value is 0.52 with a standard deviation of 0.54. 19 aisle (as we would expect) Moreover, 83% of the obse rvations are above 0 (the location o f a bipartisan outcome). This means that, given the negative coefficients on the partisan passage variables, 83% of the observations are estimated to yield a pro minority outcome. But even among the 17% of observat ions where the effect of Partisan contributes to a pro majority outcome the estimated pro majority shifts are very small A s we can see from Figure 4 8 that the negative values on Partisan are small in absolute value. For example, of the pro majority pa rtisan bills ( Partisan > 0) the average estimated effect in Model 3 is a pro minority conference outcome of 0.31 in magnitude (a 31% change in the roll call) F or the pro minority partisan bills (Partisan < 0) the average estimated effect in Model 3 is a pro majority conference outcome of only .05 in magnitude (a 5% change in the roll call). Thus, though the negative effect on Partisan can be interpreted (mathematically ) to mean that moderate legislation is skewed in a pro majority direction in conference these outcomes are not only rare (17%) but very small in magnitude. The two models reported in Table 4 11 disaggregating the effect of partisan into separate covariates for the respective chambers lends additional support to the prior discussion. Howeve r, these models add some nuance to the overall effect of partisan pre conference legislation on conference outcomes Th e positive and significant 19 coalition. This would indica te a party unity vote if 75% of the majority voted against 50% of the minority.
192 interaction term in both models indicates that when one chamber passes a partisan bill and the other chamber passes a bipartisan bill the outcome is increasingly pro minority. This is further evidence in favor of a majoritarian perspective of conference outcomes because, in reality, the Senate frequently passes bills in a bipartisan direction while the House fre quently passes bills in a much more partisan direction. But the converse is indeed true in this case. When the leadership in both chambers coordinate their legislative proposals and pass simultaneous partisan bills, there is an attenuating effect on pro minority outcome s carefully because the aggregate outcome once we calculate the overall effect remains in the direction of a pro minority shift. Figure 4 9 plo ts estimated outcome of Model 5 fo r bills passing each chamber with identical partisan roll call votes. For this Figure I calculated the estimated effect for identical House and Senate partisan bills from .20 (an extreme pro minority bill in the sample) to 1.0 (an extreme pro majority b ill in the sample). This variation i s captured on the X axis. The Y axis is the estimated effect on partisan conference outcomes according to Model 5. We can see that around 0.6 on the X axis the estimated effect switches directions becoming positive at values larger than 0.6 (where a positive value indicates a pro majority outcome) But even when the House and Senate pass identical and perfectly pro majority partisan bills (X= 1 .0) the aggregate outcome is still a pro minority conference outcome. As w as discussed earlier, the Figure rev eals that pro minority House and Senate pre conference bills do experience pro majority shift s in conference (the upper left quadrant of Figure 4 9 ). However, according to the data, in only nine out of 149 cases (6%) di d the House and Senate pass bills that were both on the minority side. For the same
193 reasons discussed earlier, the frequency and magnitude of pro majority conference outcomes are minor given the distribution of the primary independent variable. Modeling t he Third Dimension: Bicameral Conflict The multivariate spatial model reported in Chapter 3 revealed a third dimension marked by bicameral conflict where positive coordinates in this dimension indicate a pro House conference outcome while negative coordina tes indicate a pro Senate conference outcome. These coordinates serve as the dependent variable in this section. Because the response is approximately normally distributed, the estimates were derived via OLS. The only prior variable included in the esti mates of bicameral conflict is Budget The expectation is that this effect will be positive in the present estimates, indicating that when the budgetary situation is favorable the House gets more of what it wants at the expense of the Senate. Indeed, Fenn o (196 6) showed us that the Senate typically appropriates larger sums of money compared to the House and that this feature appropriations amounts is politically easier than cuttin g appropriations amounts ). Based on this logic, it follows that when the budgetary situation is strong the Senate loses some of its typical leverage (thus a pro House conference outcome). The remaining variables used in the previous two sections are not included as there are no theoretical reasons to include them in a model of bicameral conflict. Included in the estimates are a series of additional bill level factors. The first is a dummy variable indicating whether the bill was initiated by the House. The effect of House Initiate of the Senate on bills it passes first (and vice versa). This expectation stems from prior
194 research which finds that the chamber which initiates legi mover Rogers 1998 2005). Also included are bill level variables derived from the pre conference roll call record In the previous models the pre conference roll call served as an important factor as it relates to the poli There we were interested in partisan divisions. Here we are interested in how widely a bill passes each chamber, irrespective of party. In a strategic sense, when one chamber passes a bill by a wide margin the second c hamber has a strategic advantage the risk of killing the entire bill. Stated in the opposite direction, when a bill passes by a narrow margin in one chamber details of final bill. Thus, House Passage and Senate Passage are coded as the raw percentage of each chamber voting to pass the bill. The expected effect is negative with respect to a pro chamber outcome (ne gative for the House and positive or the Senate). The effect of the filibuster pivot on House Senate negotiations is subsumed in this measure. Indeed, we would expect the effect of the House passage variable to be greater in magnitude than the Senate pas sage variable, indicating that a narrowly passed Senate bill will experience less moderation than a narrowly passed House bill. An auxiliary model was estimated where divergence in the passage rate was modeled. This additional factor combined the two pas sage variables by using the differences between House Passage and Senate Passage Thus, for Passage Divergence higher values indicate the House passed its bill by a larger margin than the Senate. Finally there are two contextual, Congress level factors i ncluded in the estimates of bic ameral conflict. Both factors House Majority and Senate Majority tap the size
195 of the majority party in both chambers. These variables are simply the percentage of majority party lawmakers to total members. Larger majoritie s are expected to enhance The findings are presented in Table 4 12. Overall the models perform extremely well as they explain between 46% and 47% of the variation in bicameral conflict. In both models the substantive r esults are identical. The coefficient on Budget is statistically significant and positive, as expected. This indicates that during periods of budgetary surpluses the House tends to secure favorable conference outcomes while during periods of poor budgeta ry surpluses the Senate tends to secure favorable conference leverage. Surprisingly, the results do not support the claim that the chamber which initiates legislation secures fa vorable outcomes in conference negotiations. Of course, the prior work in this area suggests that the chamber that acts first secures favorable outcomes in the aggregate Here we are looking exclusively conference outcomes. Ultimately, this null result does not invalidate the previous findings though it is contrary to expectations In both models the results show that when one chamber passes a bill with a large favorable provisions at the conference stage. The negative coefficient on House Passage in M odel 7 indicates that as the passage rate increases the conference outcome shits in a pro Senate direction while the positive coefficient on Senate Passage indicates that as the passage rate increases the conference outcome shifts in a pro House direction. The same conclusions hold for M odel 8 where divergence in the
196 passage rate for both chambers was modeled. The negative coefficient on Passage Divergence indicates that as the House passes a bill by wider margins than the Senate the outcome shift in a pro Senate direction (and vice versa for when the Senate passes a bill by wider margins than the House). These effects are attributable to the strategic advantage of passing a bill by a narrow margin. Additional tests reveal that the effect of House passage is greater in magnitude than Senate passage (f=3.11, p<.05). Thus, the House is predicted to lose a greater share of provisions when it passes a bill by a 60% margin comp ared to an identical bill passing the Senate with a 60% margin. This Rule 22 ). Overall, these effects are majoritarian in nature, as conference outcomes benefit the chambe r with the fewest pre conference votes. Finally, the results show that the larger the majority in each chamber the more likely that chamber is to prevail at the conference stage. The positive coefficient on House Majority indicates that as the majority p conference outcomes shift in a pro House direction while the negative coefficient on Senate Majority conference outcomes shift in a pro Senate direction. Thus, th e strength of parties over conferenc e outcomes in this dimension is not due to some procedural advantage (i.e. naming partisan to serve in conference) but its numerical size. Discussion In Chapter 3 we exam ined the macro level patterns for how the House an d Senate res olve policy disagreements in conference The present chapter took a second look at this process, examining conference outcomes at the bill level
197 In the first part I reexamined the multivariate spati al model presented in Chapter 3 estimati ng separate spaces for the pre Republican Revolution and p ost Revolution periods. I found that the first and second dimensions (reconciliation and partisan conflict respec tively) exhibit remarkable stability over the postreform period Thus, even in the contemporary Congresses where partisan roll call patterns are more pronounced, resolving differences remains a multidimensional process governed by consensus and compromise first and foremost. I also reported some modest evidence that partisan conflict an d bicameral conflict have increased in salience in the post Revolution period, consistent with findings. The second section explored each dimension at the bill level using spatial mappings of conference outcomes. The spatial scores used in th ese mappings identify a point in the three dimensional policy space corresponding to the dimension or dimensions that explain the individual conference outcome. The first finding in this section was that most conference outcomes are located around the ori gin of the three dimensional coordinate system. Second, the average Euclidian distance of each observation from the origin has increased significantly between the pre Revolution and post Revolution periods. Thus, there has been a significant increase in the variability of conference outcomes over time, indicating greater changes to the pre conference House and Senate passed bills. Consistent with this is evidence that bicameral bargaining is under greater strain in the modern Congre sses, supporting the notion that competing parties and chambers have come into greater conflict over time Third, both two dimensional scatter plots of conference outcomes as well as the line plots of the median spatial location of conference outcomes in each dimension
198 indicate that over the postreform period : (1) conference outcomes have increasingly shifted in a pro minority direction over the postreform period ; and (2) there has been an increase in reconciliation suggesting that resolving differences i s aff ecting greater compromise and concession in the contemporary Congresses. Overall, the findings from the second section suggest that over the postreform period the so called in the realm of resolving differences (in addition to in the realm of bicameral conflict) and that the effects of this hurdle have created the greatest challenges for the majority party in both chambers. Finally, the third section of the analysis used the multivariate spatial coor dinates to examine the factors structuring conference outcomes. The first finding is that the greater the extent of disagreement betw een the House and Senate at the passage stage the greater the manifestation of reconciliation compromise agreements or con cessions. Second, there are unambiguous majoritarian trends in partisan conference committee outcomes. The main finding is that when partisan House and Senate passed legislation goes to conference it typically emerges resolved in a pro minority direction A t the same time, there is no evidence in any of the four models that majority outcome. Though published work has examined the composition of conference committees, finding pro majori ty biases in the H ouse and Senate, no work has directly tested this important claim. And third, pro House and pro Senate outcomes are determined largely by how widely the pre conference bills pass each chamber. The more widely a bill passes a chambe r rel ative to the other chamber, the more the conference outcome shifts in the direction of the rival chamber. This is further evidence of majoritarian
199 conference outcomes. However, I do find that chambers with larger majority party cohorts gain favorable con cessions in conference, suggesting that strong parties play a role in the bicameral dimension. findings as evidence of strategic, pro majority behaviors. In the simplest of term s, opponent s of the prior two chapters might argue that the majority party introduces and passes legislation in each chamber beyond their median ideal point so that, in negotiations with the other chamber, they are able to concede a few relatively minor pr ferred policy. Chapter 5 explores this issue in greater detail.
200 Figure 4 1. 3 D spatial map of resolving differences in the pre Revolution era The dots repres ent how conferees modified the original House and Senate passed bills according to the principal components analysis in Table 4 1 The point at patterns.
201 Fi gure 4 2. 3 D spatial map of resolving differences in the post Revolution era The dots represent how conferees modified the original House and Senate passed bills according to the principal components analysis in Table 4 2 The point at 0,0,0 has been c patterns.
202 Figure 4 3. 2 D spatial map of resolving differences by era The dots represent how conferees modified the original House and Senate passed bills according to the principal components analysis reported in Tables 4 1 and 4 2 The point patterns.
203 Figure 4 4 Reconciliation median coordinate The data were calculated as the me dian spatial coordinate in the first dimension. Higher values indicate greater compromise and lower values greater concessions. Figure 4 5 Reconciliation median absolute coordinate The data were calculated as the median absolute spatial coordinate in the first dimension. Higher values indicate greater reconciliation (compromise or concession) and lower values less reconciliation.
204 Figure 4 6. Partisan conflict median coordinate The data were calculated as the median spatial coordinate in th e second dimension. Higher values indicate pro majority conference outcomes and lower values indicate pro minority conference outcomes. Figure 4 7. Bicameral conflict median coordinate The data were calculated as the median spatial coordinate in th e third dimension. Higher values indicate pro House conference outcomes and lower values indicate pro Senate conference outcomes.
205 Figure 4 8 Distribution of partisan pre conference bills Observations to the right indicate pro majority House and Senate passed bills while figures at 0 indicate perfectly bipartisan House and Senate passed bills.
206 Figure 4 9 Estimated effect of identical House and Senate partisan bills on pro majority conference outcomes
207 Table 4 1. Component loadings (u ntransformed Data, 95 th to 96 th and 100 th to 103 rd ) Shift parameter Dimension 1 Dimension 2 Dimension 3 House majority .5832 .4553 .0047 House minority .4564 .2749 .8213 Senate majority .6051 .1693 .4391 Senate minority .2923 .8297 .3642 Interpretation : Reco nciliation Partisan conflict --Notes: N=78 The loadings were estimated via principal components analysis. The component loadings represent the correlation between the shift parameter and the dimension. The magnitude and dir ection of the loadings help us interpret the dimension. Table 4 2. Component loadings (untransformed d ata, 104 th to 106 th and 108 th to 110 th ) Shift parameter Dimension 1 Dimension 2 Dimension 3 House majority .4078 .5873 .6697 House minority .4684 .5 704 .3816 Senate majority .5124 .4568 .5820 Senate minority .5931 .3480 .2591 Interpretation Reconciliation Partisan conflict Bicameral conflict Notes: N=93 The loadings were estimated via principal components analysis. The component l oadings represent the correlation between the shift parameter and the dimension. The magnitude and direction of the loadings help us interpret the dimension. Table 4 3. Component loadings (transformed d ata, 95 th to 96 th and 100 th to 103 rd ) Shift parame ter Dimension 1 Dimension 2 Dimension 3 House majority .4786 .5911 .5978 House minority .5614 .2861 .4707 Senate majority .5456 .3129 .4079 Senate minority .3976 .6862 .5047 Interpretation : Reconciliation Bicameral conflict -Notes: N= 78 The loadings were estimated via principal components analysis. The component loadings represent the correlation between the shift parameter and the dimension. The magnitude and direction of the loadings help us interpret the dimension.
208 Table 4 4 Component loadings (transformed d ata, 104 th to 106 th and 108 th to 110 th ) Shift parameter Dimension 1 Dimension 2 Dimension 3 House majority .4707 .4264 .6780 House minority .4927 .6093 .3290 Senate majority .4843 .5807 .4343 Senate minority .5488 .3213 .4935 Interpretation : Reconciliation Partisan conflict Bicameral conflict Notes: N=93 The loadings were estimated via principal components analysis. The component loadings represent the correlation between the shift parameter and the dimension. The magnitude and direction of the loadings help us interpret the dimension. Table 4 5 Explained variance by dimension (u ntransformed Data, 95 th to 96 th and 100 th to 103 rd ) Dimension Eigenvalue Explained variance SE 95% confidence interval Dimension 1 R econciliation 1.63 .41 .05 .32 .50 Dimension 2 Partisan conflict 1.02 .25 .04 .18 .33 Dimension 3 --0 .83 .21 .03 .15 .27 Notes: N=78. The explained variances were estimated via principal components analysis. T test Dim1 to Dim2 = 2.6253 (p<.01); T test Dim2 to Dim3 = .9809 (p=.16); T test Dim1 to Dim3 = 3.6490 (p<.001) All significance tests are two tailed. Table 4 6. Explained variance by dimension (untransformed d ata, 104 th to 106 th and 108 th to 110 th ) Dimension Eigenvalue Explained variance SE 95% confidence interval Dimension 1 R econciliation 1.40 .35 .04 .27 .43 Dimension 2 P artisan conflict 1.12 .28 .03 .21 .35 Dimension 3 B icameral conflict .79 .20 .03 .14 .25 Notes: N=93. The explained variances were estim ated via principal components analysis. T test Dim1 to Dim2 = 1.3106 (p<.10); T test Dim2 to Dim3 = 1.7517 (p<.05); T test Dim1 to Dim3 = 3.0433 (p<.001) All significance tests are two tailed.
209 Table 4 7 Explained variance by dimension (t ransformed D ata, 95 th to 96 th and 100 th to 103 rd ) Dimension Eigenvalue Explained variance SE 95% confidence interval Dimension 1 R econciliation 1.53 .38 .04 .29 .47 Dimension 2 B icameral conflict 1.09 .27 .04 .20 .35 Dimension 3 --0 .77 .19 .03 .13 .25 No tes: N=78. The explained variances were estimated via principal components analysis. T test Dim1 to Dim2 = 1.9153 (p<.05); T test Dim2 to Dim3 = 1.6925 (p<.05); T test Dim1 to Dim3 = 3.6170 (p<.001) All significance tests are two tailed. Table 4 8. Ex plained variance by dimension (transformed d ata, 104 th to 106 th and 108 th to 110 th ) Dimension Eigenvalue Explained variance SE 95% confidence interval Dimension 1 R econciliation 1.69 .42 .04 .34 .50 Dimension 2 P artisan conflict 0 .95 .24 .03 .17 .30 Di mension 3 B icameral conflict 0 .83 .21 .03 .15 .26 Notes: N=93. The explained variances were estimated via principal components analysis. T test Dim1 to Dim2 = 3.5649 (p<.001); T test Dim2 to Dim3 = .6957 (p=.24); T test Dim1 to Dim3 = 4.2696 (p<.0 01)
210 Table 4 9. Determinants of reconciliation outcomes Model 1 (GLM ) Model 2 (GLM ) Divided 0.82*** 0.20 0.82*** 0.20 Mood lag 0.01 0.06 0.04 0.04 Party mandate 0.02 0.03 > 0.01 0.03 Budget 1.03 2.55 1.96* 1.16 Midpoint distance 0.54** 0.24 0.5 5** 0.24 Partisan 0.17 0.16 0.16 0.16 Bicameral distance 2.15 2.19 2.34 1.60 Intra party bicameral distance 1.55 5.85 Filibuster House distance 0.69 1.01 Alpha 1.77 2.73 3.02 1.95 R squared .12 .11 N 149 149 Notes: *** P<.01, ** P<.0 5, P<.10. All significance tests are two tailed. The response is coded such higher values indicate greater reconciliation (either greater compromise or greater concession) while lower values indicate less reconciliation Standard errors are robust.
211 Table 4 10. Det erminants of partisan outcomes Model 3 (OLS ) Model 4 (OLS ) Divided 0.66*** 0.17 0.70*** 0.16 Mood lag 0.09 0.07 0.04 0.05 Party mandate 0.13*** 0.04 0.10*** 0.04 Budget 0.66 3.07 2.23* 1.23 Midpoint distance > 0.01 0.21 0.02 0 .20 Partisan 0.43** 0.19 0.47** 0.19 House outsiders 0.44 0.39 0.48 0.40 Senate outsiders 0.14 0.30 0.13 0.30 Bicameral distance 5.95** 2.52 6.39*** 1.88 Intra party bicameral distance 2.31 7.80 Filibuster House distance 1.44 1.22 Alpha 2.31 3.37 2.02 2.9 R squared .20 .21 N 149 149 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The response is coded such that higher values indicate a pro majority outcome and lower values indicate a pro minority ou tcome Standard errors are robust.
212 Table 4 11. Determinants of partisan conference outcomes Model 5 (OLS ) Model 6 (OLS ) Divided 0.64*** 0.17 0.69*** 0.16 Mood lag 0.07 0.07 0.05 0.05 Party mandate 0.12*** 0.04 0.11*** 0.04 Budget 1.20 2.86 1.93 1.20 Midpoint distance 0.04 0.21 0.05 0.20 House Partisan 0.72** 0.35 0.73** 0.36 Senate Partisan 1.90*** 0.47 1.83*** 0.53 X Partisan 2.14*** 0.70 2.03** 0.81 House outsiders 0.39 0.40 0.42 0.41 Senate outsiders 0.06 0.30 0.06 0.30 B icameral distance 6.16*** 2.28 6.33*** 1.79 Intra party bicameral distance 0.93 7.25 Filibuster House distance 0.77 1.33 Alpha 3.05 3.24 2.13 2.35 R squared .24 .25 N 149 149 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The response is coded such that higher values indicate a pro majority outcome and lower values indicate a pro minority outcome Standard errors are robust.
213 Table 4 12. Determinants of bicameral conflict conference outcomes Model 7 (OLS ) Model 8 (OLS ) Budget 3.12*** 1.01 3.05*** 1.02 House Initiate 0.05 0.13 0.07 0.13 House Passage 4.40*** 0.45 Senate Passage 3.93*** 0.48 Passage Divergence 4.22*** 0.44 House Majority 7.94** 3.33 7.89** 3.32 Senate Majority 15.08 *** 5.33 14.89*** 5.36 Alpha 4.19*** 1.35 3.73 1.32 R squared .47 .46 N 206 206 Notes: *** P<.01, ** P<.05, P<.10. All significance tests are two tailed. The response is coded such that higher values indicate a pro House outcome and negative values indicate a pro Senate outcome Standard errors are robust.
214 CHAPTER 5 RESOLVING DIFFERENCE S: FACILITATOR OF, OR IMPEDIMENT TO MAJORITY PARTY AGENDA SETTING ? In Chapter 2 I reported that bills passed by the House and Senate have experienced gre ater bicameral gridlock over the postreform period. On its face, this governance. A c ounterargument might be that this trend is agenda con trol to block bipartisan legislation (allowing pro majority legislation to pass across the two chambers unimpeded). I proceeded to show that this intuition does not hold and, in fact, the opposite is true: partisan House passed legislation has become incr easingly vulnerable to bicameral gridlock over the postreform period. Chapter 4 reported a number of findings regarding how the House and Senate resolve disagreements when they arise. Key among those findings were the conclusions that over the postreform period there has been an increase in pro minority conference majority outcome and strong evidence that partisan legislation is most likel y to emerge from conference shifted in a pro minority direction. A general counterargument to these findings is that they actually reflect strategic, pro majority agenda control. In simple terms, opponents of Chapter 4 might argue that the majority party introduces and passes legislation in each chamber beyond their median ideal point so that, in negotiations with the other chamber, they are able to concede a few relatively minor provisions and ultimately emerge from conference with thei preferred policy. This chapter explores this issue in greater detail. The central question of this chapter is thus: How does a bicameral sequence formal forms of resolving differences affect the capacity of parties to control the
215 legislative agenda? Virtually al l existing research, though only implicitly, treats this form of institutional variation as inconsequential. The importance of this question has been chamber models of legislative organization a re under specified, and empirical evidence on key features of chamber I address the question of agenda control using a modified version of the popular Lawrence, Maltzman and Smith 2006). The primary observation of interest is the rare, but nonetheless revealing, occurrence of agenda setters agents who propose policy (Cox and McCubbins 2002) voting agai nst their own proposals at final passage. The measure is designed to capture bill level variation in agenda control, allowing us to straightforwardly compare bills that do not traverse a bicameral sequence where unicameral specifications are most appropri ate 1 with those that do traverse a bicameral sequence where intercameral bargaining and reconci liation almost certainly matter Competing formal models of agenda control are then compared in terms of their explanatory power. The goal is not to challenge the conventional wisdom; in fact, the present chapter lends support to a number of the conventional claims (Lawrence, Maltzman and Smith 2006) Rather, this chapter offers an important qualification to our understanding, highlighting how parties manage, s uccessfully or unsuccessfully, varying institutional hurdles Moreover, the findings reveal that the effects demo nstrated in 1 research shows that, even without direct inter chamber interaction, the presence of a second chamber creates strategic hurdles each chamber must consider when proposing legislation (Taylor 2008).
216 Chapter 4 stem from processes of bill moderation and bicameral reconciliation rather than strategic majority party behaviors. Theo retical Foundations for Agenda Control The capacity of parties to manipulate legislative rules, procedures and individual lawmakers in pursuit of non majoritarian (pro majority) outcomes is aptly documented in the contemporary congressional literature. Th is ability is derived in a myriad of ways: naming members to positions within the leadership, approving committee assignments and chairmanships, scheduling legislation, adopting rules of procedure, managing debate and wielding the power to defeat or amend any motion or bill on the chamber power stems from their ability to manipulate the agenda the collection of policy proposals considered by the chamber. In response t o Keith Krehbiel (1993), Gary Cox As elaborated by Gary Cox and Matthew McCubbins (1993, 2005, 2007), institutional authority by party controls the consideration of policy known as negative agenda control through powers such as scheduling and amending. In this way t he cartel model posits that the majority advances legislation to final passage when the proposal is preferred by its members to the status quo. This allows lawmakers within the majority to simultaneously vote their sincere preferences while fostering an e lectorally beneficial party record. In addition to their formal work, Cox and McCubbins marshal empirical evidence supporting the cartel model. For example, Cox and McCubbins (2002) demonstrate that that is, when a majority of the party votes on the
217 losing side of a formal vote that ultimately passes occurs on less than .07% (or 4 out votes where the leadership of both parties voted against each other to show that there has been no decline in the level of party voting throughout the post New Deal era. Independent tests have yielded compleme ntary results ( Carson, Monroe and Robinson 2009) often coming at the expense of a lternativ e theoretical perspectives ( Lawrence, Maltzman and Smith 2006). Though typically constrained to the House, recent work finds empirical support for the cartel model in the Senate (Campbell, Cox, and McCubbins 2002; Gailmard and Jenkins 2007). In fact, one study found statistically indistinguishable levels of party power in both chambers (Gailmard and Jenkins 2007). Though sharing a number of empirical predictions, the theory of conditional party government (Aldrich 1995; Rohde 1991) emphasizes a f undamentally different causal mechanism underlying party agenda control: intra party preference homogeneity and inter party preference divergence. 2 rank and file delegates greater authority to the leadersh ip which, in turn, exploits institutional rules and powers to ensure members of the party act in a way consistent CPG what is referred to as positive agenda control the abil ity to manipulate outcomes, usually on the cha mber floor ( Finocchario and Rohde 2008; Sinclair 1995). Though the effect of conditional party government is more pronounced in the House, powers wielded by the Senate majority party such as the right of first recognition and the ability 2 Sinclair (1995) argues that the majority leadership frequently employs both strategies.
218 to bypass committees suggest that both chambers are affected by polarized preference arrangements. These points are reviewed in a forthcoming section. The Bicameral Sequence Congressional scholars have long noted that legisla tive sequence matters. In fact, sequence is fundamental to mathematica l models of agenda setting ( Romer and theoretic form: (1) a member introduces a bill; (2) the bill is reported to the f loor by a committee and; (3) the floor votes on the bill. As is often the case, especially on non controversial approving it verbatim, without change. 3 In these cases a uni cameral model is adequate to understand legislative politics (though see Taylor 2008 and footnote 1 in this proposal, as is often the case, a bicameral sequence is initiated. Su ch sequences are characterized by one of two formal forms of bicameral reconciliation 4 ; where the third option is simply the preservation of the exogenous status quo. As reviewed in Chapter 1 the two ways in which the House and Senate formally resolve p olicy disagreements are the shuttling of legislation (also referred to as Shuttling is characterized by the two chambers agreeing to identical versions of legislation afte r messaging the original bill back and forth with amendments. After the 3 For example, of all public laws enacted during th e 110 th Congress originally introduced in the House, 69% were enacted without Senate disagreement. This statistic was compiled by the author using Thomas.com 4 Though both can occur on a single bill, this rarely occurs.
219 initial amendment, the modified proposal returns to the initiating chamber where the bill furthe reconciliation process of shuttling is analogous to what game theorists call sequential bargaining. If at any time during the legislative process one chamber insists on its pro form an ad hoc joint conference committee (for an excellent review see Longley and O leszek 1989). Conferees from both chambers act in what is similar to a unicameral system with majority rule. That is, once conferees have completed negotiations, a report a do cument detailing the proposed compromise. Both chambers then vote on (Oleszek 2007). Despite the centrality of sequence for research on agenda setting, hardly any wor k considers a bicameral sequence. One exception is a recent study by Sean Gailmard and Jeffery A. Jenkins (2007). Gailmard and Jenkins explore agenda setting in the context of Senate bills, confirmation reports and conference reports. Conceptually, vari ation in the rules governing these three legislative vehicles provides a natural experiment to test how constitutional features affect agenda control. 5 Using congressional level roll rates as the dependent variable and analogous datasets for the House and Senate, Gailmard and Jenkins (2007) find significant variation in agenda 5 For example, Article I Section 7 requires conference reports to have the approval of both chambers while confirmations, because of Article II Section 2, require only Senate approval.
220 control across legislative vehicles. Of note for the present chapter is their finding that confere nce reports and the president on confirmations. However, on Senate bills, the Senate majority party is unconstrained by either of these two institutional features. In testing this latter finding against an analogous model for House originated bills, 6 Gai in terms of keeping unwanted measures from receiving floor consideration, the Senate majority party is no less successful than the House majority party when it faces disagreement fr article, and likely due to space limitations, they afford little discussion to the fact that control over the agenda. regarding conference committees. But rather than simply reestimating their model, the present paper adopts a more wide ranging test. This occurs in two ways. First, Gailmard and Jenkins test the effect of conference committees as a dummy variable (sometimes referred to as Binder 1999 2003 )). This confines the effect to only a few Congresses and obfuscates a more theoretically control even when the same party controls both chambers. That is, the effect of a c onference committee on the policy process transcends partisan arrangements. 6 They were unable to model presidential confirmations since this is not a prerogative of the House.
221 Second, I test the effect of shuttling proposals between the chambers, a feature unexplored by Gailmard and Jenkins (2007). At the macro level, the results speak to larger questi setting power and intercameral bargaining. As the subsequent two sections will show, there is disagreement about these dynamics in the congress ional literature. In addition the results address whether the findings in the previous chapters concerning the importance of reconciliation and limits of partisanship in conference are due to strategic pro majority control over the congressional agenda. Bicameral Sequence and Agenda Control This s ection reviews research which, directly or indirectly, advocates that the majority is able to maintain or extend agenda control when legislation traverses a bicameral policy sequence. The first part evaluates this matter regarding conference committees, w here the literature speaks to the issue directly, while the second part discusses the shuttling of legislation, where the literature is more general in this regard. Conference Committees : committee agenda control. Because standing committee members are almost always named as managers in conference, and because conference agreements are considered u nder what amounts to a closed rule in the House, the standing committees wield the ability to modify any policy altered on the floor back to its original committee approved position (or simply defeat it). Nagler (1989), using a revised version of Shepsle model, linked the selection of conferees by the Speaker and the outcome of conference committees. The ultimate power conferred by the ex post veto, and by extension the
222 sts with the majority, according to Nagler (1989). Lazarus and Monroe (2007) advanced this line of work further. They demonstrate that under certain conditions the Speaker will name partisan loyalists as conferees in addition to committee members. This strategic party obtains its preferred outcome in conference negotiations. More recent work finds that this selection power, though less prominent, exists in the naming of S enate conferees as well (Vander Wielen and Smith n.d.). Amendment Trading : Agenda Control in the House and Senate Though published research on conference committees is limited, even less research explores the process exchanging amendments between the chamb ers. discussion about the power of the majority in the House and Senate more generally. Because of space constraints, I adopt a cursory review of this power in the Hou se, since the literature has been well rehearsed elsewhere, in order to provide more discussion of the Senate, where expectations are less well established. In the House, the majority party selects a Speaker who has the power to schedule legislation, rec ognize members on the floor and suggest members to serve on the Rules Committee as well as standing committee members and their chairpersons (Oleszek 1989). Conceptually, then, party leaders and committees should be responsive to the demands of their indi vidual members. But perhaps the most significant power exercised by the majority party in the House stems from the Rules Committee. Whether acting independently (Dion and Huber 1996) or as an arm of the leadership (Cox and McCubbins 2005), the Rules Comm ittee can limit the terms of
223 debate, a significant departure from the Senate, and limit amending activity on the floor (Romer and Rosenthal 1978). For example, special rules have been shown to confer significant procedural control over the agenda ( Finocch ario and Rohde 2008; Romer and Rosenthal 1978). At the same time, increased party homogeneity, characteristic of the modern House, allows the majority to coordinate policymaking through shared preferences and the delegation of greater power to the leaders hip ( Aldrich 1995; Rohde 1991 ). In the contemporary era, this means the majority party has greater positive agenda power to go along with their historically st rong negative agenda power ( Cox and McCubbins 2002). In the Senate, recent research has posited a significant degree of agenda control through procedures such as filling the amendment tree, tabling motions, policy coordination, the use of reconciliation and through non referral. The first two right of first recognition while chamber, the filibuster pivot and standing committees) that might compete with the party over the agenda. The procedure known offers an amendment to a bill and subsequently offers amendments to those amendments (known as second degree amendments). Because of first recognition, filling the amendment tree allows the majority to restrict the agenda in a manner amendments ( Sinclair 2000 a ). Filling the amendment tree is especially important for the
224 shuttling process since both processes op erate according to the rules governing amending activity A second power of the majority in the Senate over the agenda is through the motion to Table A motion to Table which, unlike a roll call to defeat an unfavorable amendment is not subject to unli mited debate, can be used to quickly dispose of any pending question a simple majority of senators want to avoid taking up. Thus, it represents a relatively costless way to constrain the agenda (Den Hartog and Monroe 2008). Crespin and Monroe (2005) repo rt that Table motions occur regularly one third of all roll call votes during the 101 st 104 th were Table motions and that the majority party is rolled only about 10% of the time. As they maintain, the Senate floor is not an for as is commonly believed (16). Third, the majority may mitigate the attenuating effects of bicameral sequence because of greater policy coordination and shared preferences. When preferences within the majority are homogenous and divergent from the minori ty (Aldrich 1995; Rohde 1991), when the chambers allocate power arrangements in accordance with each other (Sin and Lupia 2004), or when party leaders strategically coordinate the introduction of proposals (Taylor 2008) we may expect a narrow, non majorita rian range of policy proposals as well as greater use of restrictive rules and procedures to protect Indeed, others have pointed out that it is common for drafts of legislation to be circulated for comment and strategies for passage coordinated in both chambers (Oleszek 2004, 257; Taylor 2008).
225 Of course, the majority can use reconciliation to circu mvent the constraints Republican tax cuts during the Bush era as well as health care reform in 2010. The majority party can also circumvent committees by placing a House passed bill directly on the calendar without committe e review (Evans and Oleszek 2000 ). Thus, two institutional actors often thought to compete with the majority party in the Senate the filibuster pivot and committees can be wholly circumvented. How Bicameral Se quence Aids the Majority Party In summary, scholars have put forward a number of good reasons to expect that the majority party is able to exercise significant control over the policy agenda even when the policy sequence involves resolving differences. Re garding conference committees, the existing literature suggests that the strategic selection of conferees allows the majority to skew policy outcomes toward its preferred location and/or defeat damaging proposals. In this way, the majority has an effectiv e veto over legislation at the conference change and can manipulate the range of policies considered as solutions to inter chamber disagreement. Regarding the shuttling of legislation, rules and procedures in each chamber favorable to the majority (specif ically those that govern ing floor activity) afford them significant control over policy outcomes. That is, once a bill is passed by the initiating chamber, the majority party in the second chamber ents or rival bills, or amend it in a manner favorable to the majority in both chambers. In sum, in the context of Chapter 3 and Chapter 4 an argument can be made that the majority uses reconciliation to its advantage and that the votes lost by the major ity after inter chamber bargaining are the result of losing relatively minor provisions strategically included in legislation to facilitate
226 bicameral reconciliation. As is clear at this point in the manuscript, the theoretical argumentation and prior find ings predict the opposite. Formal Models of Agenda Control In this chapter, I test four models of agenda control following Lawrence, Maltzman and Smith (2006). The first two models are w ithout an effective party role (or preference based ) These models a ssume that simple majorities dictate outcomes. In distance from some pivotal lawmaker(s) determines their likelihood of securing favorable outcomes. The Median Voter m odel assumes that when the chamber median prefers some policy to the status quo, that policy is introduced and summarily adopted. Because a simple majority is responsible for consequential agenda setting outcomes such as adopting rules, electing the Speak er or discharging a bill from committee the floor median wields significant power over outcomes. The second model makes the same simplifying assumptions but incorporates the constitutional requirements of a presidential signature. That is, according to th e Pivotal Voter model, successful proposals must be preferred by the president or the lawmakers needed to circumvent the president. Members located between these two pivotal actors are expected to uniformly support successful proposals relative to the sta tus quo; the expected utility for all other members declines linearly relative to their distance from the nearest pivotal actor. The third and fourth models of legislative behavior posit parties are the main factor affecting the behavior of lawmakers. The Majority Party model holds that introduced and enacted policies are favored by members of the majority. Parties use such as campaign resources, favorable committee assignments and
227 favorable policies as well as an electorally beneficial party label to persuade members to vote with their cohort. This model predicts uniform party effects. The second party Agenda Control model, when t he majority party favors a proposal to the status quo the policy advances to the floor for a vote. Minority party members closest to the chamber median have a greater likelihood of joining the majority and securing some preferred outcome. The predictions are thus non uniform. Figure 5 1 presents the theoretical implications of these four theories. The Y Axis represents the likelihood a lawmaker will vote to approve some proposal at final passage. Each model assumes the policy space is unidimensional. As Lawrence, Maltzman and Smith (2006) note, these four models make unique predictions about the nature of agenda control. The goal is to identify the theory that best represents actual outcomes Data and Methods Agenda setting studies typically analyze tr the percentage of bills a majority of the majority party opposed but nonetheless passed. Aggregated at the congress level (though see Carson, Monroe and Robinson 2009), a low roll rate is evidence that the majority party infrequent year period. In their study, Lawrence, Maltzman and Smith (2006) developed an analogous measure constructed at th e individual in rates measure record s the percentage of times in a given t wo year Congress a member voted with the winning side on all final passage roll call votes. Unfortunately, these existing measures are not designed to address the questions raised by the present chapter. Instead, the central research questions require
228 a measure sensitive to bill l evel institutional variation, such as intercameral policy sequences, as well as sensitive to subtle, but nonetheless consequential, changes in the range of policies being considered by each chamber. Indeed, the conventional meas ures are non dynamic, final stage operationalizations. Constructed via backward induction, these measures as sume the final roll call vote reveals the optimal outcome preferred by the majority party relative to the status quo. Fortunately there is an int uitive way to develop a bill level measure of agenda effects analogous to the conventional measures. The comparability of these two proposes some policy consideration (Cox and McCubbins 2002). Methodologically, we need to identify two things: (1) members who have made agenda setting proposals and (2) the agenda eference for the final product at the end of the legislative sequence. For the latter, I follow convention and use final passage roll call votes. If the proposal is not modified by the second chamber this is simply a final passage vote. For legislation shuttling between chambers this is th e final passage vote immediately preceding If the proposal goes to conference, I use the vote approving the conference report. If the proposal was ing, I use the last final passage vote. To identify agenda setting proposals, I use the slate of bill sponsors and cosponsors. These individuals are, quite literally, agents who have proposed some policy for consideration by the chamber ( before action by committees or floor amendments )
229 reasons. First, research shows that party leaders base their agenda setting decisions on the signals provided by bi ll sponsors and cosponso rs ( Kessler and Krehbiel 1996; Koger 2003; Wawro 2000). Of course, sponsoring and cosponsoring is an institutionally costless endeavor; there is no agenda control over who sponsors or cosponsors legislation. However, if we restrict the data collection to bills with final passage votes, the various institutional gatekeepers (committees, party leadership, floor, etc.) favored the proposal to the status quo based on the simple fact the bill was enacted into law. oposal was favored by gatekeepers at various stages of the policy process or the proposal was successfully modified to satisfy those gatekeepers. Thus, in these cases, the majority party has exercised significant agenda control. In addition to satisfy ing the needs of the present study, using sponsorship and cosponsorship as the revealed preferences of agents who proposed some policy has a number of virtuous qualities. First, because no formal institutional constraints exist, the act of sponsoring or c osponsoring legislation more accurately captures the revealed preferences of lawmakers compared to roll call votes. Second, and most important, because cosponsorship occurs very early in the legislative process 7 it allows the researcher to model the entir e legislative sequence, rather than just the final moment ( Woon 2008, 207). 7 Ho committee authorized to consider and report the bill or resolution reports it to the House or is
230 Of course, this coding scheme assumes that the act of sponsorship and cosponsorship reveals information about the propos Woon 2008). This position is supp orted by a number of existing studies which argue that cosponsorship reflects either convergence between the proposal and the ideology of the member or a member signaling the ideological content of the bill (Kessler and Krehbiel 1996; Krehbiel 1995; Wawro 2000; Woon 2008). 8 Using this coding scheme, the primary observation is whether a member voted for (coded 1) or against (coded 0) their initial proposal at the final stage in the sequence Conceptually, this variation distinguishes favorable from unfavor able agenda effects. This methodology mirrors a coding scheme recently employed by Harward and Moffett (2010) used to assess the costs associated with cosponsoring legislation. Rather than 8 Of course, some studies demonstrate that members may also engage in position taking when cosponsoring legislation (Koger 2003; Woon 2008). Since the present paper cannot fully address this question due to space constraints, a more pertinent question is: How will position taking, whether ra re or frequent, affect the empirical results? I believe there are two options: (1) position taking is constant from sponsorship to the time of final passage and thus white noise or (2) based on the results of existing research, position taking will be bia sed against the thrust of the present paper. Whichever is true (or, if both are true), position taking will represent an additional hurdle the primary hypotheses must overcome and not type one error. On the fist point, Assume for a moment that sponsorship is, to some degree, a form of position taking. Following this argument to its logical conclusion, once a member has staked his or her electorally beneficial position, it is unlikely that they would reverse that position on the final vote (where their dec ision is even more visible and consequential). Thus, position taking is not bias because it is constant from t 1 to t 2 and appears in the statistical results as white noise. On the second point, The answer to the question concerning how position taking will affect the empirical results depends on which party, if either, is most likely to engage in position taking when sponsoring legislation. I believe the intuitive answer is that members of minority, because of the difficulty they have effecting bill co nsideration on the floor (Cox and McCubbins 2005), are more likely to sponsor symbolically when the proposed policy is favored by their median Using data on means that any bias in the dependent variable (that is, if we accept that sponsor ship represents and thus work against the proposed hypothesis.
231 aggregate win rates (Lawrence, Maltzman and Smith 2006), t he meas ure indicates individual level agenda setting wins. The raw data for every House and Senate sponsor and cosponsor for the 94 th to 110 th 9 For the House ta to match each sponsor and cosponsor to their final passage roll call vote. This was done by first matching each bill Votes database, 10 importing the sequential roll call vote number and ess by congress roll call matrices. 11 Three additional pieces of information were recorded: (1) if the bill became law (2) if the final passage vote was on a conference report and (3) if the bill was shuttled between the c hambers. The public law data are website. 12 The remaining pieces of information (i.e. whether either form of bicameral reconciliation was used) were compiled by the author. Unfortunately, collecting parallel data for every sponsored or cosponsored Senate bills is not possible; existing datasets containing all required pieces of information do not currently exist. Manually collecting this data for every bill with a final 9 http://jhfowler.ucsd.edu/ 10 Rohde, David W. Roll Call Voting Data for the United States House of Representatives, 1953 2004. Compiled by the Political Institutions and Public Choice Program, Michigan State University, East Lansing, MI, 2004. 11 These data http://voteview.com/ 12 The data are available at: http://www.policyagendas.org/ The data were originally collected by Frank R. Baumga rtner and Bryan D. Jones, with the support of National Science Foundation grant number SBR 9320922, and were distributed through the Department of Government at the University of Texas at Austin and/or the Department of Political Science at Penn State Univ ersity. Neither NSF nor the original collectors of the data bear any responsibility for the analysis reported here.
232 passage vote is beyond the scope of this chapter. Fortunately, the primary f ocus on bills with a bicameral policy sequence limits the data collection to a much more manageable scope. Thus, for bills that had a conference committee or amendment exc hange the corresponding data were manually collected using Thomas.com and the Databa se of Historical Congressional Statistics 13 Using the same process as described for the House, each sponsor and cosponsor was matched with their final passage vote by congress roll call matrices. 14 Since this measure is an an comparison, I replicate Lawrence, Maltzman and the one hand, replication is a key component of scientific research and the study by Lawrence, Maltzman and Smith is an important one. Thus, uncovering the same results will meaningfully add to the literature in its own right. At the same time, their analysis directly tested various formal models of agenda setting. Thus, replication represents a straightforward test of the previou s theoretical discussion. Consistent with their approach, I construct independent variables discriminating the predictions of the agenda models described previously. The Median Voter model is operationalized as the absolute distance of each member from t he House median. The model predicts that agenda setters closest to the chamber median are most likely to win at final passage. The Pivotal Voter model is operationalized using two variables. The first variable is the distance from the House 13 The Database of Historical Congressional Statistics Compiled by Elaine K. Swift, Robert G. Brookshire, David T. Canon, Evelyn C. Fink, Jo hn R. Hibbing, Brian D. Humes, Michael J. Malbin, and Kenneth C. Murtis. 14 http://voteview.com/
233 median for m embers on the opposite side of the ideological spectrum from the president. This coefficient is predicted to be negative. The second variable is the distance from the veto override pivot for members on the same side of the ideological spectrum as the pre sident. Members to left (right) of the veto pivot during liberal (conservative) majorities are expected to have a decreased win rate while members located between the pivot and the chamber median should have an increased win rate. The Majority Party mode party. The coefficient is predicted to be positive. Finally, the Agenda Control model is operationalized using two variables. The first variable is the distance from the House median for mem bers of the minority while the second variable is a dummy variable for negative coefficient on the first term and a positive coefficient on the second term. All operat ionalizations use the exact same construction as Lawrence, Maltzman and Smith (2006). The raw preference data consists of first dimension common DW NOMINATE scores (Poole 1998). Since the dependent variable is dichotomous, all models were estimated via l ogit. 15 Standard errors were clustered by Congress to account for temporal heterogeneity. 16 In the House, 1864 sponsors and cosponsors voted against their initial proposals at final passage. Though this reflects only a small fraction of the population 15 Given the low variation in the dependent variable I also estimated each model using rare events logit (King and Zeng 2001). None of the substantive results changed. Since the rare events specification is unable to estimate information criterion (AIC and BIC), a decision was made to report the results of a standard logit mod el with cluster robust standard errors. 16 For example, one source of heterogeneity is a rule prior to the 96 th Congress that limited cosponsorship to 25 members.
234 of interest (about 2%), there are two important points to keep in mind. First, this paper does not dispute the fact that the majority party has strong powers over the agenda. This low variation, much like the very low variation in the roll rate, reflects th is fact. But more importantly, instances in which sponsors and cosponsors vote against their own proposals are substantively important and provide a clean look at agenda effects. The 9/11 Commission Bills: An Example As an example of what the dependent va riable captures recall the legislative history of the act (HR10 and S2845) in the 108 th Congress. The House bill was introduced by Speaker Hastert with 26 Republican cosponsors including Roy B lunt (MO), Tom Delay (TX), John Boehner (OH), James Sensenbrenner (WI) and Duncan Hunter (CA). These members favored the House bill in general, but also a number of specific provisions they helped shepherd through committee. There were no Democr atic cosp onsors. The pro majority location of the House bill contrasted sharply with the bipartisan location of the parallel Senate bill, which had a mix of Republican and Democratic cosponsors including John McCain and Hillary Clinton. The salient policy disagre emen ts between bill centered on the authority afforded the proposed national intelligence director with the Senate favoring a more authoritative director and sweeping immigration changes contained within the House bill. Formal and informal bicameral negotiations proved difficult and broke down at least twice (Kady 2004) Unable to reach a consensus, the bill went to a conference committee The final conference agreement effectively split the difference on these two key issues, limiting the power of the new intelligence director while striking some of the more controversial immigration provisions. These changes were unfavorable to many conservative
235 lawmakers, who happened to be in the majority. Only after intervention by the Bush administr ation did Speaker Hastert bring the conference report to the House floor for a vote (Kady 2004) in the House (67 of which were Republicans) were five of the initial cosponsors, including James Sensenbrenne r ( R WI) Phil Gingrey ( R GA), Mark Green ( R WI), Darrell Issa ( R CA) and Scott McInnis ( R CO). As the prior example shows, despite the low varia tion in the dependent variable, the present measure cleanly captures variation in agenda control from introduct ion to final passage (including after a conference or amendment exchange). In this regard, passage; thus, the conventional measures would cite the 9/11 Commission law as an example of majority party agenda control, though intuition tells us that the majority party sacrificed some of its preferred proposals namely immigration provisions. Findings Th e findings are contained in eight tables. Tables 5 1 through 5 6 present the resu lts for the House while T ables 5 7 and 5 8 present the r esults for the Senate. Tables 5 1, 5 2 and 5 3 contain pooled estimates for all Congresses in the dataset (94 th 110 th ) while T ables 5 4, 5 6 and 5 7 Re th 110 th ). For the House data, three specifications are tested: (1) all final passage votes (intended as both a reference group and for repor t final passage votes; (3) all amendment exchange final passage votes. Recall that the data for all Senate final passage votes is not available. Also, it w as discovered after the data were collected that the occurrence of Senate sponsors and cosponsors
236 v oting against their proposals at fina l passage is significantly rare compared to in the House. 17 exchange. Thus, estimation of the Senate models was only possible for conference report votes Two Senate t ables are reported, all conference report votes for the pooled sample (5 7) and all conference report votes for t he post revolution sample (5 8). The first set of results in T ables 5 1 and 5 2 for the House confirm almost exactly the findings reported by L awrence, Maltzman and Smith (2006). When we explore all final passage roll call votes three of the four models have significant and correctly signed coefficients. Only the non Lawrence, Maltzman and Smith. The only difference between the present results and LMS is that the present analysis uncovers a (correct) negative coefficient on both pivotal voter terms. All other results are exactly statistically and substantively those reported by Lawrenc e, Maltzman and Smith. In both the pooled model (5 1) and the post Revolution model (5 2) I find that both party based models outperform the preference based models according to the information criterion. Of these two, the majority party agenda control m odel outperforms the uniform majority party model. These results are important for two reasons. First, it demonstrates the robustness important study for our understanding of the U.S. Congress. Inde ed, despite the contribution of this project and its frequent majoritarian (non partisan) conclusions, I can confirm that in the aggregate that the majority wields significant agenda control. This is consistent with my 17 disputes (Fenno 1966; Manley 1970; Strom and Rundquist 1977; Vogler 1970) and because much of the Senate is governed by unanimous consent agreement (thus there are not usable final passage votes).
237 qualification that the results of th is dissertation should not be interpreted to mean that Second, it demonstrates the validity of the dependent variable used in this study. Though the constructions of my dependent variable diffe rs from convention out of necessity, these findings suggest that my response and the response used by Lawrence, Maltzman and Smith tap the same construct. The larger contribution of the present chapter is the results reported for the conference report and amendment exchange models ( T ables 5 3 through 5 6) These results speak to the debate concerning the effect of resolving differences on the ability of legislative actors to control the congressional agenda. The results clearly support the view that resol ving differences the use of conference committees and the shuttling of legislation between chambers is majoritarian in nature. Whether we examine the entire post reform era (94 th 110 th ) or simply the post Revolution era (104 th 110 th ), the median voter model outperforms all rivals on these bills. 18 Substantively, the main finding is that when the House passes legislation that is sub sequently amended by the Senate and the two chambers engage in bicameral negotiations, the agenda shifts in a direction tow ard the House median (rather than away from the House median) This supports the intuition that bicameral bargaining is dominated by the space connecting the medians of the House and Senate rather than the ideal points of the majority party. Fo r example, in the pooled sample for conference report votes ( Table 5 3) I find that House lawmakers two standard deviations removed from the chamber median are three times more likely to vote against their proposal at final passage compared to the median 18 However, none of the models for the 104 th 110 th subset achieve statistical significance in the amendment exchange analysis.
23 8 member when the bill goes to a conference committee. For bills exchanged between the chambers as amendments ( Table 5 4) lawmakers two standard deviations removed from the chamber median are about two times more likely to vote against their proposal at final passage The uniform majority party model performs second best across all four specifications. The model has statistically significant and correctly signed coefficients in both conference committee models. In the pooled analysis, the majority party model outperforms the pivotal voter model, which is also statistically significant and correctly signed, as indicated by the information criterion. This is the only model in which both pivotal voter variables obtain statistical significance. In none of the spe cifications does the majority party agenda control model reach statistical significance. The results in T ables 5 7 and 5 8 for the Senate lend some support to the majoritarian findings reported regarding conference committees. In the full analysis for the 94 th to the 110 th Congresses (Table 5 7 ) only the median voter model is statistically significant and correctly signed. I find that Senate lawmakers two standard deviations removed from the chamber median are two times more likely to vote against their p roposal at final passage compared to the median member when the bill goes to a conference committee. However, these results are not statistically significant when we constrain the analyses to the post Republican Revolution era (Table 5 8) In this analys is, only eight Senators voted against their proposal so the insignificant results are probably the result of inadequate statistical power Comparing Conferencing and Amendment Trading One final issue that we can address with the data is whether agenda sett ing differences exist between conferencing and amendme nt trading. As we have seen, for
239 both forms of r esolving differences the median voter model outperforms all rivals. This indicates that resolving differences is a majoritarian process first and foremo st. But is one form more majoritarian than the other? There are strong theoretical reasons to suspect the answer is yes With an amendment exchange, though the challenges of resol ving differences remain such as the difficulty satisfying pivotal actors, multidimensionality in the policy space, the uncertainty and the risk involved etc. there is one critical advantage enjoyed by majority party that does not exist with respect to conferencing Specifically, an amendment exchange is governed by the same ru les in each chamber that confer organizational and agenda setting advantages to the party leadership (namely control over amendments). For example, the majority in the House can con trol amendments to a Senate bill with a special rule from the Rules Commit tee or by simply defeating the amendment on the floor. In the Senate, the majority can control amendments to the House bill by filling the amendment tree or by using tabling motions. In a conference, by contrast, the majority delegates their authority to a group of actors who, in return, make an unamendable proposal. Thus, we would expect the median voter model to perform better during a conference committee as compared to shuttling To test the previous hypothesis I ran a n additional logit model with al l observations for shuttling and conferencing combined In addition to the median voter variable, I included a dummy variable for whether the bill went to a conference (labled Conference ) and an interaction between the median voter variable and the confer ence dummy variable ( Chamber median distance Conference ). The prior hypothesis predicts that the interaction effect will be statistically significant and negative. Such an effect would
240 indicate that lawmakers furthest from the chamber median are signif icantly less likely to vote for legislation they cosponsor after a conference committee compared to an amendment exchange. The results are presented in Table 5 9 The results presented in Table 5 9 confirm the previous hypothesis. In both the full sample (94 th to 110 th ) and the post Revolution sample (104 th to 110 th ) the interaction term is negative and statistically significant. This indicates that representatives closest to (furthest from) the chamber median are more (less) likely to vote for legislati on they cosponsor after a conference committee as compared to an amendment exchange. In simple terms, though both forms of postpassage bargaining are majoritarian in nature, conference committees have more pronounced majoritarian effects on policy outcome s To see the differences in these effects I calculated the predicted probability of a cosponsor signing their legislation on the final roll call vote using representatives ranging from 0 .0 in the first dimension ( those the chamber median) to 0 .5 (a fair ly conservative or liberal lawmaker) 19 Conference committees are represented by squares while amendment exchanges are represented by diamonds. The solid lines denote the full sample while the dotted lines represent the post Revolution sample. The estimat ed effects tell two interesting stories. First, we can see that at about 0. 4 in the full sample and 0 .55 in the post Revolution sample the estimated probability of a cosponsor voting for their legislation is the same for a conference committee and an amen dment exchange. In these two samples lawmakers to the left of these points (closer to the chamber median) are more likely to vote for their legislation after a 19 between bet ween 0.545 and 0.587 in the first dimension throughout Congressional history.
241 conference committee compared to a shuttle while lawmakers to the right of these points (closer are less likely to vote for their legislation after a conference committee. This confirms the main hypothesis. The other interesting effect is the slightly positive slope on the estimated effect for shuttling in the post Revolution sample. Compared to the entire sample, where the shuttling estimate has a slight negative slope, this indicates that amendment exchanges have become slightly non centrist in recent Congresses. Thus, as the parties have become stronger ov er the postreform period amendment trading has become an increasingly pro majority procedure. This adds further support to the main hypothesis tested in this section. However, it is important to qualify these statements with the fact that T able 5 6 shows that this effect is statistically insignificant. Discussion This chapter has sought to advance our understanding of agenda control by considering the topic from a bicameral perspective. Indeed, the contemporary literature focuses almost exclusively on un icameral, House centric models. The central question was: How does a bicameral sequence specifically, formal forms of bicameral reconciliation affect the capacity of parties to control the legislative agenda? Using a ted at the bill level, I found that conference committees and the shuttling of legislation between chambers produces suboptimal when considering the topic of agenda contr ol as well as highlights how parties manage, successfully or unsuccessfully, varying institutional hurdles. These findings are especially important given the findings reported in Chapter 3 and Chapter 4. Chapter 3 reported a number of findings concerni ng how the House and
242 Senate resolve disagreements when they arise. Key among those findings were the conclusions that over the postreform period there has been an increase in pro minority conference outcomes, greater strain in the reconciliation process, a lack of evidence majority outcome and strong evidence that partisan legislation is most likely to emerge from conference shifted in a pro minority direction. A general counterargument to these findings is that the evidence actually reflect strategic pro majority agenda control. In the simple terms, opponents of might argue that the majority party introduces and passes legislation in each chamber beyond their median ideal point so that, in negotiations with the other chamber, they are able to concede a few relatively minor provisions and ultimately emerge from conference with majority party typically concedes some of its preferre d policies when legislation goes to conference or is resolved via amendment trading. Though the main contribution of this study concerns the limits of majority party agenda control, this should not be regarded as a dismissal of the strength of the majority in the contemporary era. In fact the initial portion of the study confirmed that on all final passage votes the majority exhibits significant agenda control (Lawrence, Maltzman and Smith 2006). At the same time, however, the results demonstrate that we cannot generalize across bills. That is, lumping together bills that were engaged in a bicameral sequence with those that did not experience a bicameral sequence ignores important institutional variation in this regard.
243 Table 5 1. All House final pass age votes, 94th 110th Congresses Median v oter Pivotal v oter Majority p arty Agenda c ontrol Chamber median distance 2.02*** (.26) Chamber median distance (non pres. side ) .32 (.50) Distance from v eto Pivot ( pres. side ) 4.51*** (.84) Majority party membership (1/0) 1.20*** (.14) Majority party side of median (1/0) .79*** (.29) Minority distance from chamber median .72* (.43) 4.90*** (.19) 4.38*** (.16) 3.46*** (.13) 3.87*** (.19) N 99481 99481 99481 9 9481 AIC/BIC 15525 / 15544 15521 / 15550 15357 / 15377 15339 / 15368 Notes: *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a co sponsor voting for their bill on the final roll call vote and 0=a cosponsor voting against their bill on the f inal roll call vote. C luster robust standard errors in parentheses. Table 5 2 All House final passage votes 104th 110th Congresses Median voter Pivotal voter Majority party Agenda control Chamber median distance 2.48*** (.34) Chamber medi an distance (non pres. side) .70 (.58) Distance from veto pivot (pres. side) 7.52*** (1.35) Majority party membership (1/0) 1.36*** (.24) Majority party side of median (1/0) .58* (.34) Minority distance from chamber median 1.21*** (.37) 5.39*** (.36) 4.68*** (.25) 3.61*** (.17) 4.40*** (.20) N 61825 61825 61825 61825 AIC/BIC 8334 / 8352 8332 / 8359 8312 / 8330 8290 / 8317 Notes: *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a cosponsor voting for their bill on the final roll call vote and 0=a cosponsor voting against their bill on the final roll call vote. C luster robust standard errors in parentheses.
244 Table 5 3 All House conference report votes, 94th 110th Congresses Median voter Pivotal voter Major ity party Agenda control Chamber median distance 3.19*** (.62) Chamber median distance (non pres. side ) 1.78** (.74) Distance from v eto p ivot ( pres. side ) 4.46*** (1.26) Majority party membership (1/0) 1.30*** (.33) Majority party side of median (1/0) .04 (.41) Minority distance from chamber median 2.41*** (.73) 4.85*** (.31) 4.24*** (.24) 2.96*** (.24) 4.24*** (.38) N 6600 6600 6600 6600 AIC/BIC 1390 / 1403 1432 / 1452 1404 / 1418 1387 / 1407 Notes : *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a cosponsor voting for their bill on the final roll call vote and 0=a cosponsor voting against their bill on the final roll call vote. C luster robust standard errors in parentheses. Table 5 4 All House shuttle final passage votes, 94th 110th Congresses Median voter Pivotal voter Majority party Agenda control Chamber median distance 1.81** (.88) Chamber median distance (non pres. side ) 3.42 (2.92) Distance from veto p iv ot ( pres. side ) .97 (3.97) Majority party membership (1/0) .26 (1.00) Majority party side of median (1/0) .28 (1.30) Minority distance from chamber median .90 (1.18) 5.00*** (.77) 3.96*** (.76) 4.13*** (.49) 4.67*** (1.09) N 3982 3982 `3982 3982 AIC / BIC 557 / 570 552 / 571 566 / 579 568 / 587 Notes: *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a cosponsor voting for their bill on the final roll call vote and 0=a cosponsor voting agains t their bill on the final roll call vote. C luster robust standard errors in parentheses.
245 Table 5 5 All House conference report final passage votes, 104th 110th Congresses Median v oter Pivotal v oter Majority p arty Agenda c ontrol Chamber median d i stance 4.03*** (1.03) Chamber median distance (non pres. s ide) 1.20 (.94) Distance from v eto pivot (p res. s ide) 8.78*** (1.86) Majority p arty m embership (1/0) 1.68** (.69) Majority p arty s ide of m edian (1/0) .66 (.79) Minority distance from c hamber m edian 3.62*** (.69) 5.69*** (.62) 4.51*** (.46) 3.06*** (.45) 5.40*** (.52) N 2563 2563 2563 2563 AIC/BIC 461 / 472 473 / 491 479 / 491 469 / 487 Notes: *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a cosponsor voting for their bill on the f inal roll call vote and 0=a cosponsor voting against their bill on the final roll call vote. C luster robust standard errors in parentheses. Table 5 6 All House shuttle final passage votes, 104th 110th Congresses Median v oter Pivotal v oter Majority p arty Agenda c ontrol Chambe r m edian d istance 1.12 (1.22) Chamber median d istance (non p res. s ide) 4.64 (3.64) Distance from v eto pivot (p res. s ide) 4.76** (2.40) Majority party m embership (1/0) .67 (1.28) Majority p arty side of m edian (1/0) 3.00 (1.93) Minority distance from c hamber m edian 3.48** (1.58) 4.70*** (1.11) 3.68*** (.86) 4.71*** (.62) 7.04*** (1.55) N 2754 2754 2754 2754 AIC / BIC 411 / 422 396 / 414 410 / 422 409 / 427 No tes: *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a cosponsor voting for their bill on the final roll call vote and 0=a cosponsor voting against their bill on the final roll call vote. C luster robust standard errors in parentheses.
246 Ta ble 5 7 All Senate conference report votes, 94th 110th Congresses Median v oter Pivotal v oter Majority party Agenda c ontrol Chamber median d istance 2.13** (.94) Distance from filibuster p ivot .36 (2.00) Distance from veto p ivot .14 (1 .59) Majority party m embership (1/0) .55 (.80) Majority party side of m edian (1/0) 1.70 (1.80) Minority distance from chamber m edian 5.13** (2.53) 5.51*** (.52) 4.77*** (.69) 4.45*** (.69) 6.70*** (1.63) N 1630 1630 163 0 1630 Notes: *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a cosponsor voting for their bill on the final roll call vote and 0=a cosponsor voting against their bill on the final roll call vote. C luster robust standard errors in parentheses. The AIC and BIC are not reported since only one model performs according to expectations. Table 5 8 All Senate conference report votes 104th 110th Congresses Median v oter Pivotal v oter Majority party Agenda c ontrol Chamber median d istance .99 (1.00) Distance from f ilib uster p ivot .89 (2.13) Distance from veto p ivot 1.74** (.81) Majority party m embership (1/0) 1.06 (.97) Majority p arty side of m edian (1/0) .27 (1.31) Minority distance from chamb er m edian .94 (.60) 4.83*** (.36) 4.20*** (.89) 3.80*** (.68) 3.92*** (.30) N 705 705 705 705 *** p<.01, ** p<.05, *p<.10; cluster robust standard errors in parentheses. The response is coded such that 1=a cosponsor voting for their bill on the final roll call vote and 0=a cosponsor voting against their bill on the final roll call vote. The AIC and BIC are not reported since none of the models perform according to expectations.
247 Table 5 9 Conference and shuttle interaction effects Median v oter 94 th 110 th Median voter 104 th 110 th Chamber median distance .39 .56 .44*** .16 Conference 1.21 .73 2.56*** .81 Chamber median distance Conference 2.80** .83 4.47*** 1.08 3.64 .60 3.13*** .49 N 10582 5317 Notes: *** p<.01, ** p<.05, *p<.10. The response is coded such that 1=a cosponsor voting for their bill on the final roll call vote and 0=a cosponsor voting against their bill on the final roll call vote. C lu ster robust standard errors in parentheses.
248 Figure 5 1. Formal models of agenda control
249 Figure 5 2 Conference and chamber median interaction effects. This Figure was created based on the estimated reported in Table 5 9.
250 CH APTER 6 CONCLUSIONS The Marriage of the House and Senate in Historical Perspective This project has been the first to investigate two decisive stages in the legislative process bicameral disagreement and reconciliation jointly. One of the central premises animating the research was that the formal interaction of the House and Senate is a critical, but often neglected, feature of organizational structure. As such, t he scope of this project differs from our tendency as congressional researchers t o examine a single chamber in isolation. Though such studies a re valuable because they closely examine key features of legislative politics, they often overlook what I have argued is the central feature of congressional organization: legislative checks an d balances. And if there were one all encompassing conclusion from this dissertation, it is that inter chamber cooperation and conflict are variable evolving in consequential fashion throughout history rather than constant. Indeed, the marriage of the Ho use and Senate has changed considerably since 1787. The Honeymoon Period From 1789 ( the firs t session of the first Congress) until around the 1830s, the marriage of the House and a Senate was characterized by a legislatively dominant House or Representativ es and a less active and less prestigious Senate early career is a fitting illustrat union Clay, one of the most prominent figures in congressi onal history, began his federal legislative caree r in th e Senat e before running for electi on to the House in 1810 Though such Senate to House career paths are unthinkable today, it was quite common in the early Congresses; senators were just as likely to leave for the House as were representatives
251 to leave for the S enate (Baker 2008, 16). Because of its connection to the public through direct elections, stronger internal organization and authority over revenue bills, the House played a larger ro le in debating and crafting national policy (Carmines and Dodd 1985; Bak er 2008). Indeed, in the first few decades of the Republic, national politics was focused on domestic usually agrarian issues (Carmines and Dodd 1985). According to one estimate, legislative activity in the early House outpaced that of the Senate by thre e to one (Binder 1995). Taken as a whole, the House and initial relationsh ip was commensurate in many respects wit h Constitutional design. In fact many of the Framer s believed that the Senate should (and would) be politically quiescent and funct ion primarily as a legislative of the House. 1 Indeed, Binder (2003, 16) points out that there was no discussion of the opposite relationship during the Founding era that the House might be designed to simply ons of the Senate. The Emergence of an Activist Senate Of course, the House and Senate began to evolve vis vis each other soon after the first session of the first Congress Emblematic of this transition, Henry Clay returned to the Senate in 1831, a m the Missouri Compromise in 1820 was a critical juncture in this evolution (Baker 2008; 1 In Federalist no. 62 The necessity of a senate is not less indicated by the propensity of all single and numerous assemblies to yield to the impulse of sudden and violent passions, and to be seduced by factious leaders into intemperate and pernicious be remarked is, that a body which is to correct this infirmity ought itself to be free from it, and consequently ought to be less numerous. It ought, moreover, to possess great firmness, and consequently ought to hold its authority by a tenure of considera ble
252 but see al so Wirls 2007 ). Not only d proposal, but, after a conference committee met, it was the Se nate that drafted and passed the amendment establishing the southern border of Missou ri as the dividing line between free and slave states ( Woodburn 1893, 258 265). Thus, passage of the significant role in debating the Missouri Compromise, position in national politics. While some have characterized this period as the so cal led Senate debate (e.g. de Tocqueville ), more recent work characterizes the House and Senate of this period as coequal bodies in terms o f debating national political issues ( Wirls 2007 ). While the Senate developed an active role in national politics the rapidly expanding size of t he House growing from 59 members in 1789 to 186 members by 1820 made restrictions on debate inevitable 2 On the one hand, t his physical development eroded (Baker 2008) though, on the o ther hand, the development of the previous question rule which limited debate in the House and ability to act decisively Following the Civil War, rapid economic growth made Congress both the House and S enate, a more powerful political institution. This expansion of authority, coupled with the rising salience of foreign affairs issues further extended the vis the House. As Carmines and Dodd (1985 ) explain, Congress was a more dominant national institution and bicameralism 285 ). 2 The previous que stion rule was adopted in 1811.
253 In sum, the marriage of the House and Senate underw ent significant change from 1789 to the dawn of the 20 th Century: from a politically subordinate Senate that tended to review the policies passed by the more prestigious House to what may be described as an equal partnership. A short, probably made interchamber bargain ing over the terms of policy change a more central and frequency component of the legislative process (18, emphasis in original). Indeed, w between the two chambers (quoted in Baker 2 008, 20). The End of Bicameralism as the Founders Conceived It The relatively equal partnership of the House and Senate in the early 20 th century was dr amatically altered in the 1910s. There were two major developments. The first major development wa s, of course, the passage of the 17 th Amendment. The 17 th Amendment to the Constitution establishing the direct election of senators made the Senate a more democratic and politically responsive body. Bernhard and Sala (2006) show that after the 17 th Amen dment passed senators were much more likely to moderate their policy positions in response to looming elections while Crook and Hibbing (1997) find that the direct election of senators led to a less aristocratic upper chamber 3 and made the Senate more resp onsive to the public mood. Cook and Hibbing (1997, 852 853) aptly summarize the effects of the 17 th Amendment on the this way For better or for worse, direct election rendered the Senate less sedate and more closely tied to the people, synchronizing it with the House and the presidency; in fact, it often looks much like a smaller version of 3 In particular, they find that s enators had fewer characteristics resembling the state lawmakers who previously appointed them.
254 six of the eleven changes in party control were the use initiated the transition, compared to only one of nine changes after 1914 ( Alford and Hibbing 2002 ). The second major development in the early 20 th century was legislative reform s Prior to 1910, and despite the emergence of a politically ac tive Senat e, the House retained some of its leverage over the Senate because of its more robust rules and procedures. This structure was strengthened in both 1889 and 1903 during the influential speakerships of Thomas Reed and Joseph Cannon. In 1889, House reforms undermined the ability of the minority to block legislation (Rohde 1991) Among the changes implemented in the 51 st of and gave the Speaker the power to refer legisla tion directly to committee ( Forgette 1997; Rohde 1991). Reed further consolidated his power by appointing loyal Republica ns as chairmen of influential committees such as Ways and Means and Appropriati ons (Rohde 1991). Then in 1903, with the election of Joseph Cannon as Speaker of the House, organizational power was further centralized in the hands of the majority As Speak influential Rules Committee. As Rohde (1991, 4) explains Committee permitted him to determine which bills go to the floor, and his powers as presiding officer enabled hi m generally to dictate their fate once there Thus, f rom 1889 to 1910 the House retained some of its leverage in the legislative process because of the power enjoyed by the Speaker and
255 action Though the Senate experien ced an increase in majority party rule during this period, its organizational structure was not nearly as robust (Smith and Gamm 2009). However, in 1910, the organizational power of the majority in the House was undermined Joe Cannon. This watershed moment in congressional history transformed the relationship between the House and Senate in relatively short order. The coup against Cannon was driven by within party ideological divisions; namely, progressive Republicans join ed Democrats in enacting a series of reforms which removed the Speaker from the influential Rules Committee and deprived the Speaker of the authority to name committee chairmen and committee members (Baker 1973; Rohde 1991). What emerged after the reforms was a system of of committee government were less hie rarchical leadership, the diffusion of power across the House particularly among the committee chairmen and growing regional divisions within the Dem ocratic party (Cooper and Brady 1981; Dodd and Oppenheime r 1977; Rohde 1991). As a result, the legislative process was marked by greater bargaining on the House and Senate floor and more frequent cross party coalition formation at final passage (Smith and Gamm 2009). Though the House and Senate experienced similar transformations during this period, the Sena te reforms were much less dramatic and consequential (Carmines and Dodd 1985). Thus, a major element keeping the House and Senate in a coequal relati onship at the turn of the century structure was undone in the 1910 s Carmines and Dodd (1985 286 ) aptly summarize the alteration of the bicameral partnership in the following way: This weakening of the power of con gressional leaders marked the end of
256 central leadership and the subsequent rise of committee government, fragmentation and immobilism resulted. Rather than a restraint on an aggressive Congress, bicameralism now became an inducement toward weakness, a constitutional provision that helped increase the leverage of the executive and judiciar y over the legislative branch. The Textbook and Reform Eras Compar ed to the developments in each of the earlier periods, from 1789 to 1913, 2 1968) there were no major changes in the relationship of the House and Senate. Both houses remained nearly coequal partners, with the Senate still regarded as it was in the early 20 th century as the more prestigious chamber (because of its longer terms, role in foreign affairs, smaller size and greater visibility). Moreover, the Senate remained a more nationally focused body while the House maintained its paro terms, role in foreign affairs and greater prestige). This is not to say that consequential institutional developments never occurred during the textbook era: quite the contrary. Rather, the major development s of this period occurred in similar fashion within each body. For example, both chambers experienced a period of committee dominance, where senior and disproportionately south ern committee chairmen exerted unequal influence over policy outcomes (Rohde 19 91; Sinclair 2005). Of course, the House was marked by committee government to a greater extent than the Senate, which maintain ed an individualistic character ( Sinclair 2007). Another major development was the general decline in governing capa city in relation to the president: a change experienced jointly by the House and Senate ( Carmines and Dodd 1985; Cooper 2005; Dodd 1981; Sinclair 2007). Finally, both chambers experienced a growth in incumbency during textbook period (Ansolabehere and Sny der 2002; Mayhew 1974 ) In sum, the major institutional
257 developments of textbook period occurred along similar trajectories in each body Compared to the earlier periods, the marriage of the House and Senate w as relatively stable. Calls for reforming the textbook era Congresses began as early as the 1950s (Dodd an d Oppenheimer 1997; Zelizer 2006 ). The impetus for these efforts was a outhern Democrats to block civil rights legislation, and a decline in institutional legitimacy in relation to the Supreme Court and president (Cooper 2005; Carmines and Dodd 1985; Dodd 1981; Rohde 1991; Sinclair 2007). A number of political sci entists joined the calls for reform, advocating institutional changes designed to foster ideological coherence within the two parties and create the conditions for greater effici ency in the policy process ( Schattschneider 194 2; Key Jr. 1942). With an infl ux of new, and more liberal, lawmakers in both the House and Senate in 1958, formal efforts to consolidate organizational power with in the parties and undermine the committee system begun. These efforts culminated in the reforms of the 1970s (documented m ore fully in earlier sections of this dissertation) T he reforms implemented in the early to mid 1970s undermined the seniority system and weakened the power of committee leaders (which had empowered southern Democrats in their efforts to stymie civil ri ghts structures (particularly the role of the Speaker) and made party leaders more responsiv e to the rank and file (Dodd 1986; Rohde 1991; Zelizer 2006 ; Dodd 1979; Cooper 2005; Deering and Smith 1997; Stewart 2001). Thus, following the reform
258 period (1968 1974), congressional researchers identify a redistribution of organizational power away fro m committees and their chairmen and an increase in the organizational capacity of the parties It is within the context of these reforms and the succeeding era in Congressional history, that the present research is positioned The Postreform Period In ea rlier chapters ( Chapter 1 and Chapter 2 in particular) I highlighted two consequential developments during the reform (1968 1974) and postreform (1974 present) periods that affected the formal interaction the so of the House and Senate. First, House reforms in the mid 1970s and in 1995 enhanced the organizational power of the majority to a greater extent than the parallel reforms which occurred in Senate. For example, though the Senate made major changes to its committee system in the 19 70s in particular, Senate reformers cut the number of committees and sub committees and reorganized committee jurisdictional borders no (Davidson 1981; Deering and Smith 1997). Th is is in contrast to the House reforms of the 1970s which revived the Steering and Policy Committee (including the Speaker as its chair), expanded the whip system and empowered of the Speaker to name members to the influential Rules Committee ( Davidson 198 1 ; Deering and Smith 1997; Dodd 1979; Dodd and Schott 1979; Rohde 1991; Sinclair 1988, 1989; Shepsle 1989; Zelizer 2006). A similar narrative applies to the reforms implemented after the 1994 congressional mid term. Though reforms implemented in 1995 beq ueathed the Senate an organizational structure resembling the House (particularly with the adoption of a secret ballot to name committee chairs), the implemented reforms were again much less consequential i n terms of strengthening the parties com pared to t he House reforms
259 ( Davidson 1981; Deering and Smith 1997; Zelizer 2006). 4 As Smith and Gamm (2009, a centralized and efficient House where the majority can easily exploit rules and procedures to pass legislation while the Senate process more unpredictable because the chamber is marked by great individualism as well as partisanship (Sinclair 2005, 2007). T he second major development of the postreform period in the relationship between the House and Senate has been the growth of polarization within Congress. Though primarily an external development, electoral geographic polarization has affected the marriage of the House and Senate due to constitutional variation in the selection and apportionment of the two bodies. As I argued in Chapter 2 due to basic features of constitutional design we would expect the House and Senate to polarize very differently: with the House polarizing at a faster rate than the Senate. Indeed, the Framer s established a system of legislative checks and balances w here the House and Senate were designed to respon d to exogenous factors very differently. And though the 17 th Senate (Alford and Hibbing 2002 ), the fact that only 1/3 rd of the Senate is replaced each cycle maintains polarization asy mmetries The empirics presented in Chapter 2 confirmed this effect a factor I labeled polarization asymmetry. Though no work that I 4 Term limits, designed to enhance the responsiveness of commi ttees and leaders to the rank and file, were less stringent in the Senate compared to the House. While the House cut the number of committee staff by one third, the Senate made no corresponding cuts. And the House leadership was given enhanced power over the selection of committee chairs while no such powers were afford the Senate leadership.
260 am aware has labeled this phenomenon, p olarization asy mmetry is consequential because, as I have shown, it has caused gre ater tension between pivotal members in each chamber (between the chamber medians, party medians, and the filibuster pivot). The effect s of these compositional tensions has been greater bicameral conflict and growing challenges resolving differences Mor eover, I have shown at repeated points in this dissertation that these challenges have been experienced disproportionately by the majority party. How the House and Senate Coevolve Viewed through the lens of over two hundred years of institutional develop ment, the marriage of the House and Senate has been characterized by remarkable stability and gradual, but consequential, change. The former point stands in contrast to E dmund Randolph b could never long co exist ). This stability is due, in part, to the fact that the Constitution prevents the two chamber s from radically altering th e terms of their original union, as the Article 1 explicitly codifies key f actors of the bicameral partnership (with the 17 th Amendment). Nonetheless, throughout Congressional histo ry, two areas of change have consistently affected the relationship of the two houses. The first area of bicameral change stems from the fact that the Constitution is silent on the rules governing each chamber. Article 1 Section 5 of the Constitution stat es in unequivocal language Each House may determine the Rules of its Proceedings This has given the House and Senate great leeway over the terms of their union. Internal reforms most notably in 1910 and in the mid 1970s have led to
261 important devel opments in the terms of the bicameral partnership by altering the balance between centralization and decentralization Specifically, these reform periods altered the bicameral relationship by weakening and strengthening (respectively) the organ izational c apacity of the majority in the House: one of sources o f power vis vis the Senate. The second area of bicameral changes stem from exogenous developments which have affected the relationship of the House and Senate. One feature o f this effect has been socio structural. Changes in the physical size of the House throughout the 1800s undermined the deliberative and democratic foundation of the lower chamber while the rise of major national issues (the Missouri Compromise, the Civil War, WWI, etc.) spurned to action the more nationally focused Senate. The other feature of exogenous development has been the fact that exogenous changes have manifested very differently in each chamber These differences in the effects of external devel opments are due to fundamental institutional differences between the House and Senate Indeed, unlike some bicameral legislatures, 5 the U.S. Congress is not ma rked by The growth of polar ization, in particular, has manifested ver y differently in each chamber resulting in growing tensions between the House and Senate. dominance in the early 1800s, which created a coequal relationship between the cha mbers in the mid to late 1800s and gave rise to a dominant Senate in the 1910s, the major historical changes I note in this dissertation stem from the confluence of 5 (Tsebelis and Money 1997).
262 exogenous developments (polarization) and internal reforms. The key effect of this conflu ence has been more frequent and more severe bicameral conflict. Overarching Conclusions Summarized in a paragraph, the findings and implications of this dissertation are as follows. While it is true that parties in Congress have enjoyed enhanced organizat ional capacities over the postreform period, leading to a surge in pro majority policy outcomes, the majority has simultaneously faced growing bicameral constrain ts because of an increase in House and Senate compositional differences. As the empirical cha pters have shown, these compositional differences have caused a growth in both the frequency and severity of bicameral disagreement which have, in turn, created challenges when the House and Senate attempt to resolve differences. These developments are p aradoxical because they suggest that ideological polarization simultaneously empowers the majority for the reasons outlined by the theory of Conditional Party Government as well as creates a larger bicameral hurdle that the majority must contend with when the two chambers attempt to not just pass but enact legislation. All told, polarization has had what may be on policy activity and outcomes Moreover, the empirics have repeatedly shown that the process of resolving differen ces is majoritarian in nature marked by consensus and compromise first and foremost rather than partisan. As with bicameral conflict at the passage stage, the findings reveal that the majority party has become increasingly constrained at the postpassage s tage. The main implications, discussed more fully in a moment, are as follows. First, this paradoxical relationship offers a parsimonious solution to a clear contradiction in the con gressional literature. F or over a hundred to
263 active government ( Wilson 1885, Schattschneider 1942; Key Jr. 1942) while more recent work has cited polarization as the cause of growing stalemate ( Binder 1999, 2003; Dodd and Schraufnagel 2009 ). The present work reconciles these two views Thus, as a practical matter, the present work helps us understand a fairly intuitive observation: while the majority is strong within the House and Senate, occasionally the preferred policy is m oderated or fails outright because of disagreements across the chambers ( even during unified government ) Because of these findings and implications, I believe it is apt to characterize the modern legislative process as one of great risk Policy outcomes and policy production have become bipolar, marked by growing pro majority outcomes as well as more frequent failures and pro minority outcomes. Specific Conclusions Chapter 2 is motivated by the overarching question: How have growing ideological polarizat ion and the strengthening of parties in Congress affected the operation of inter chamber bargaining ? Relying on an original dataset that records both the frequency and severity of policy disagreements between the Hous e and Senate, Chapter 2 reveals that o ver the postreform period the two chambers have come into greater conflict when significantly over the last thirty years. Furthermore, the results show that pro majo rity or I find evidence that this effect is due to increasing distance s between the medians of each chamber as well as growing intra party disagreements across the Hous e and Senate The congressional literature has been largely silent on these developments. Noting that polarization has occurred at a faster rate in the House than in the Senate, I
264 attribute the increase in bicameral conflict to growing compositional asym metries between the two chambers. These compositional asymmetries include: (1) greater intra party bicameral differences, and (3) a growing gap between the House median and Senate filibuster pi vot. Chapte r 3 explores (at the macro level) how the House and Senate resolve developing a unified typology of how the House and Senate might resolve differences; on e that contributes to our broader understanding of this process. Using roll call data from all conference committees convened from the 95 th to the 110 th Congresses, I operationalize the typology using multidimensional spatial modeling. There are two prim ary findings in this section. First, I uncover evidence of multidimensionality in the process of resolving differences. In particular, I find that three qualitatively distinct dimensions reconciliation, partisan conflict and bicameral conflict explain ov er 80% of the variation in the conference committee roll call patterns. Second, I find that the first dimension of resolving differences (i.e. the one that explains the greatest amount of cess of resolving of the literature over the past few decades, I do not find that partisanship is the leading determinant of post passage bargaining. This finding is par ticularly consequential when juxtaposed with findings about the growing bicameral hurdle facing the majority party.
265 Where Chapter 3 examines the aggregate patterns in post passage bargaining, Chapter 4 takes a look at the micro level, examinin g individual conference committee outcomes. Following Chapter 3 the first section estimates separate spatial models for the pre Republican Revolution and post Republic Revolution periods. These separate models show that the first and second dimensions r econciliation and partisan conflict, respectively exhibit stability over time, indicating that even in the contemporary Congresses (where partisan roll call patterns are more pronounced) the process of resolving differences remains a multidimensional proce ss governed by consensus and compromise first and foremost. The second section explores the individual dimensions at the bill level using spatial mappings of conference outcomes. I find that over the postreform period there has been: (1) an increase in t he variability of conference outcomes, (2) an increase in pro minority conference outcomes, and (3) an increase in the extent of compromise and concession in conference. Overall, these findings suggest that over the postreform period the so increasing salient when competing parties and chambers attempt to resolve differences and that this hurdle has increasingly constrained the majority party. Finally, the third section of Chapter 4 uses the multivariate spatial coor dinates to examine the factors structuring conference outcomes. The main findings are as follows. First, greater pre conference disagreement necessitates greater reconciliation. Second, partisan legislation is moderated in a pro minority direction in co nference. And at the same time, yields a pro majority outcome. Third, the more widely a bill passes one chamber relative to the other chamber the more the conference outcom e shifts in the direction of
266 the rival chamber. Taken as a whole, the findings from this section suggest that conference committees operate in a majoritarian rather than distributive fashion. Chapter 5 concludes the empirical chapters by examining th e process of resolving differences through formal theories of agenda setting. Opponents of the prior chapters may argue that the majority introduces and passes legislation in each chamber beyond their median ideal point so that, in negotiations with the o ther chamber, they are able to concede a few minor provisions and ultimately emerge from conference with central question of Chapter 5 is thus: How does resolving differences affect the capacity of parties to control the legislative agenda? I capture bill level variation in agenda control, allowing us to straightforwardly compare bills that do not traverse a bicameral sequence with those that do traverse a bicameral sequence. Armed with this data, competing formal models of agenda control are compared. The results show that partisan models of agenda control perform best when we examine all final passage votes. However, when we restrict our analysis to bills that went to a conference committee or were shuttled between the chambers in disagreement, a non partisan, majoritarian model outperforms all rivals. This highlights how parties manage, successfully or unsuccessfully, varying instit utional hurdles and shows that the majority party typically concedes some of its preferred policies when legislation goes to conference or is resolved via amendment trading. Additional Empirics T wo additio nal sources of information speak to the postreform developments in how the House and Senate resolve differences. These two are: (1) the rate with which legislation in disagreement is referred to a conference committee and (2) the rate with
267 which conference committees fail to reach a compromise. What foll ows is a limited discussion of these topics; both could be the subject of an entire article or chapter. Despite this limited discussion, the descriptive statistics I present dovetail quite nicely with the findings and implications of this dissertation. Th e first additional source of information is the rate with which legislation passed by the House and Senate in disagreement is referred to a conference committee. Recall that the convening of a conference committee is a voluntary decision, one made almost exclusively by the majority party (specifically the party leaders and the jurisdictional committee chairmen). And because there are other avenues to reso lving differenc es including amendment trading or passing a new bill it is logical to regard the decisi on to appoint conferees as a strategic one. It is no doubt curious, then, that the data collected for this study shows a clear decline in the use of formal post passage bargaining strategies both conference committees a nd amendment trading. Ryan (2010) n otes this trend as well. For example, my data reveal that in the 95 th Congress, out of the 634 bills enacted into law, 227 (36%) were enacted after amendment trading while 144 (23%) were enacted after conferencing. In the 110 th Congress, by comparison, o ut of the 460 bills enacted into law, 79 (17%) were enacted after amendment trading while only 9 (2%) were enacted after conferencing. 6 Thus, not only has the use of conference committees decl ined over the postreform period but that decline has been fairl y steep. The ques tion becomes: What has caused this decline? Ryan (2010) correctly, in my opinion, attributes the decline to growing ideological polarization in Congress. He 6 The data for all laws enacted come from the Re sume of Congressional Activity.
268 notes that with greater polarization there have been fewer bipartisan votes and smaller winning coalitions on bills typically referred to a conference committee (making bicameral negotiations more challenging and non conference options more att ractive). In addition, he argues that the Senate minority has increasingly used procedural motions to block or delay the move to conference (such as filibust ering the naming of conferees ). Ultimately, Ryan suggests that the majority is intentionally avoiding conference committees and amendment trading. This sentiment was echoed by Steven Smith in his recent testimony before the U.S. Senate. Smith (2010, 167) testified that conference mechanisms to avo The declining use of conference committees has important implications for this dissertation First, my central claims in Chapters 3, 4 and 5 are : (1) that resolving differences is a process of compromise and co ncession first and foremost; (2) that resolving differences is becoming increasingly strained and ; (3) that this strain has resulted in greater pro minority conference outcomes. These claims stand in direct contrast to the conventional wisdom which holds that the conference managers yields pro majority outcomes. If the pro majority outcomes were the dominant pattern in conference we would see more bills going to a conference, not fewer Indeed, backward induction tells us that the strategic decision to avoid a conference committee stems from the suboptimal outcome s that result Thus, the decline in conferencing is further evidence in favor of th arguments Second, in Chapter 2 I presented eviden ce that within party bicameral
269 differences have become worse over the p ostreform period creating a growth in bicameral hurdles. Such a development is problematic for the majority. According to foster an electorally free rider problems. Managing the party label is the prim ary collective action problem conference reconciliation options reconciles these two matters. T hat is, non conference reconciliation options which are informal and thus hidden from public view serves t promoting a beneficial party record while satisfying the need to merge two bills in disagreement In colloquial terms, non conference reconciliation options hide the The second additional source of informa tion is the rate with which conferees appointed to resolve House and Senate disagreements fail to reach a compromise. Though such instances are rare, they are dramatic cases of policy deadlock. On the the notion that there are significant incentives to compromise at t he post passage stage. A s I ar gued in Chapter 1 once legislation reaches the conference stage there are enormous resources time spent drafting legislation, committee hearings, debate, vo te trading, etc. invested in pending legislation. Simply put, the costs associated with failure are significant. All in all, the rarity of conference failure dovetails with my general c haracterization of conferencing: that it is a process of compromise and concession rather than distributive politics first and foremost. But on the other hand, we can
270 examine the conference failure rate as an additional piece of evidence to help us characterize postreform developments in resolving differenc es. The partisan theories of lawmaking particularly conditional party government would predict that the failure rate has declined over time as the parties have become stronger (in particular, as the parties have become more internally homogenous). Howeve thesis that bicameral conflict has grown over the postreform period and that this conflict has been especially problematic for the majority would suggest an increase in the failure rate. Of course, even though bicameral conflict has grown over th e postreform period, its effect on the conference failure rate is muted. 7 Nonetheless, we should be able to glean some results. named conferees but a confer ence report was never filed. Next, using the Library of Congress Thomas .com website, I coded if a similar bill existed. 8 For these observations the initial bill and the similar bill I identified and read CQ Weekly articles detailing the legislative histo ry of each bill. Based on those contemporaneous accounts, 9 7 Indeed, as chapter four has shown, we have seen greater pro minority shifts in conference over this period, suggesting that there has been an increase in policy moderation during the conference stage. Such moderation will not affect the failure rate even though growing moderation is clear evidence of weakened majority party influence at the conference stage. Thus, an examination of the failure rate is biased against the central hypotheses of this project. 8 http://www.thomas.gov/ 9 A major failure is one where conferees were unable to draft a workable compromise and no similar bill was enacted into law. Thus, the conference failure killed the entire proposed policy. A moderate failur e, by comparison, is one where conferees were unable to draft a workable compromise but a similar bill was subsequently enacted into law. This code was typically applied in two scenarios: when a moderated version of the original bill was enacted at a late r date or major elements of the original bill were enacted as part of another bill. And finally, the
271 Because CQ Weekly Online was used as the primary historical source, the data can only be extended back to the 98 th Congress. The raw data for conference failures are reported in Table 6 1. What we can see immediately is that conference committees rarely fail to produce a compromise. I find that 6% of conference bills end in moderate failure while o nly 2 % end in major failure. Also, and most importantly, there are no visible historical trends in the data. For example, before the 104 th Congress the average moderate failure rate was 7% while after the Republican Revolution the average moderate failure rat e was 6%. The average major failure rate was 2% in the pre Revolution period and 3% in the post Revolution period. These differences are not statistically significant according to a difference of means test. Thus, it would appear that the failure rate o ffers no way to discriminate between our two perspectives. However, we need to consider one addition al piece of information: the fact that major legislation has become larger in magnitude over the postreform period. This pattern includes conference legisl ation. Figure 6 1 presents the average public law length (measured in pages) for all laws referred to a conference committee. 10 Unfortunately, these data are only available beginning with the 101 st Congress. Nonetheless, the trend is clear even in this l imited historical time span. In the 101 st Congress the average public law referred to a conference committee was 86.6 pages in length. In the 110 th Congress, the average public law referred to a conference identical or nearly identical bill was enacted into law (for e xample, if the original bill was wrapped into an omnibus bill). 10 To obtain this data, I downloaded the search results in the Lexis Congressional Public Law query. Contained within the results is the length of each law. The actual public law length (in pages) was matched to each conference bill.
272 committee was 196.3 pages in length. Thus, publ ic laws referred to a conference committee have more than doubled in size from 1989 to 2008. T he overall conclusion is this: Even though conference committees have failed at roughly the same rate over the postreform period, there is good reason to suspec t that those laws which do fail are much larger on average in the latter half of this period The those drafted and passed by both chambers are failing to make their way into law. This is consequent ial because many representatives, senators, staff, lobbyists, interest groups, etc. put effort into passing these failed policies (which were very close to enactment). The overall conclusion is that the postpassage stage has become incre asingly costly ove r the postreform period for all political actors, but particularly the majority As a simple exercise, consider the political, economic and social implications if the following (recent) bills that went to conference had failed at the post passage stage th HR 1), Medicare Prescription Drug, Improvement, and Modernization Act of 2003 th Energy Independence and Security Act of 2007 th HR 6). Implications Strong Parties or Weak Parties? Over the pas t two decades the most debated issue in the congressional literature has been the effect(s) of political parties, if any, on legislative organization. During this period, Keith Krehbiel (1991, 1993, 1996, 1998) has been the most influential critic of the so called overarching conclusion is that parties have no effects on congressional behavior independent individual ndividual legislators vote with fellow party members in spite of their disagreement about the policy
273 in question, or do they vote with fellow party members because of their agreement us researchers existing work currently sides with those who find important partisan effects in both the House and Senate (Aldrich 1995; Aldrich and Rohde 1997 ; Bind er, Lawrence and Maltzman 1999; Carson, Monroe and Robinson 2009; Cox and McCubbins 1993, 2005, 2007; Finocchario and Rohde 2008; Kiewiet and McCubbins 1991; Lawrence, Maltzman and Smith 2006; Rohde 1991). While the original aim of this project was to ass ess the formal interaction of the House and Senate in the policy process, inevitably the findings speak to the parties in Congress debate. In this first section I review the implications of this partisan theories of lawmaking. findings are especially relevant for the theory of conditional party government (Aldrich 1995; Aldrich and Rohde 1997 ; Rohde 1991). Conditional party government the dominant explanation of historical variation in the strength of parties posi ts that the distribution of preferences within the two chambers is consequential for constructs: intra party preference homogeneity and inter party preference heterogeneity. T rank and file will delegate greater authority to the their leaders who, in turn, exploit institutio nal rules and powers to ensure all members behave in a way consistent with ideologically distinct and the majority party is homogenous there are serious
274 consequences for a l egislative victory by the minority. In this context there is an incentive for the rank and file to cede some of their authority to the leadership a solution to the proverbial collective action problem. The opposite pattern, l ess delegation to party leade rs, holds for periods where the majority is heterogeneous and has overlapping preferences with the minority. Thus, according to CPG, there is a positive relationship between ideological polarization within Congress and the strength of the two parties. W hat, then, are the implications of the present research for our understanding of conditional party government and the parties in Congress literature? First, as a technical matter, nothing in the dissertation is a direct challenge to conditional party gove rnment. Indeed, the theoret ical tenets of CPG are predicated on a single chamber c onceptualization of congressional organization; nothing is the preceding chapters challeng es the utility of CPG in the House or Senate exclusively In fact, CPG informs muc h of this dissertation. For exampl e, recall that within party bicameral distance has increased over the postreform period and that this development has contributed to growing bicameral disagreement and gridlock. According to CPG, as t he governing parties in each chamber enhance their formal powers, they also become more intractable in their policy positions because of greater homogeneity. Thus, because of greater within CPG will only exacerbate these very bicameral tensions. This is especially true if one chamber the House is highly polarized and has centralized party structures while the other chamber the Senate is less polarized and more individualistic. And for the e xact same reasons, we can look to CPG for an explanation of the growing gap
275 has become increasingly centralized in the House over the postreform period, bicameral negotiat aisle) have become increasingly contentious. The opposite is also true; if the Senate passes legislation that satisfies the filibuster pivot, that proposal is less likely to pass the House when the majority is homogenous and polarized. The point, in both cases, is that we can look to CPG and its theoretical tenets to understand growing bicameral disagreement over the postreform period. The extent to which conditional party g overnment informs the present work is unsurprising; both are based on the logic of spatial theory and posit, as a core premise, that the distribution of preferences within Congress matters greatly for legislative organization and policy outcomes. However, there remain crucial differences. Most obvious is the fact that where CPG looks at variation in the distribution of preferences within each chamber, the present work explores variation in the distribution of preferences across the chambers. Because of these different empirical vantage points, the findings are not simply different but countervailing This is the main implication of the present work for the partisan theories of lawmaking. Specifically, the findings show that when we examine the distribu tion of preference s acros s, rather than within t he House and Senate, what we see are patterns of growing bicameral conflict and challenges to resolving differences over the postreform period. As the empirics have shown, though increased bicameral conflic t has created greater hurdles and challenges for governing capacity writ large, this burden has fallen disproportionately
276 on the majority party. Furthermore, I have shown that one of the key determinants of this effect is the growth in within p arty bicameral distance. But while this dissertation and CPG work in tandem in a number of respects, there is at least one important challenge or admonition to CPG contained within this work. cartel model (Cox and McCubbins 1993, 2005), CPG hinges on the concept of positive agenda control the ability of lawmakers to manipulate policy outcomes, usually on the chamber floor, in a proactive manner. Crafted with the reform s of the 1970s in mind, CPG is modeled after the efforts of liberal Democrats to wrestle power away from their southern brethren who stymied civil rights reform in the 1960s. Thus, institutional reforms were implemented to weake n the southern committee ch airme n, diffuse power throughout the House and Senate, and empower the party le adership in the House. These reforms embody po sitive agenda control because they were enacted with t he goal of helping Democrats pass new legislation, rather than simply block policies they oppose. In the context of bicameralism and legislative checks and balances we would expect cohesive parties to act unilaterally and coordinate legislation with their cohort in the other chamber (with whom they presumably have similar policy views). But the present work has shown, fairly unequivocally, that there has been an increase in both the frequency and severity of bicameral conflict over the postreform period. Even e in bicameral same party controls both chambers of Congress, we cannot simply assume that the two ough these findings
277 and implications do not negate conditional party government it does require at a minimum that we acknowledge the existence of a secondary process that has come to rival or limit the organizational str ength of the majority. Certainly o ne could argue that laws not bills (though see Cox and McCubbins 1993, 2005). Laws, after all, alter the status quo: bills do not. Of course, the prior discussion would be moot in a num ber of respects if this project had examined bicameral disagreement in isolation. Indeed, a central premise of this research is that to understand the operation of the House and Senate, we need to explore bicameral disagreement and reconciliation. The lo gic for this is that, if there has been an increase in bicameral confl ict over the postreform period but those disagreements are resolved in a pro majority fashion at the post passage stage, the implications of Chapter 2 for our understanding of parties in Congress would be limited (though, of course, bicameral stalemate remains a consequential issue in such a scenario). And while the literature on resolving differences is underdeveloped (Hines and Civettini 2004, 1; Krehbiel 1991, 194; Longley and Oleszek 1989, 2), some recent work has advocated in favor of the pro majority theory. While link between a biased conference delegation and partisan conference outcomes is indeed logical, no study has empirically verified it. 11 Chapter 4 did exactly this, findin g no evidence for this link. In many ways, the opposite seems to be true. On the one hand, this evidence negates a potential coun findings that bicameral disagreements at the passage stage are simply resolved in a pro majority fas hion as the post passage stage. But on the other hand, these findings 11 Nagler (1989) offers some descriptive evidence.
278 have implications for the cartel theory (Cox and McCubbins 1993, 2005). As with the prior discussion of conditional party government, these implications are not direct challenges in pa rt because cartel theory is based on a single chamber model but important qualifications or admonitions. In point of fact a significant portion Chapter 5 confirmed that on all final passage votes the majority exhibits significant control o ver the policy a genda ( Lawrence, Maltzman and Smith 2006), partly confirming the cartel theory. However, the results also showed that when we restrict our analysis to bills that went to a conference committee or were shuttled between the chambers in disagreement, a non p artisan, majoritarian model outperforms all rivals. Thus, in the over the agenda. At the risk of defeat or delay, and because resolving differences is complex, the majority cedes some of its negative agenda control at the postpassage stage and allows unfavorable modifications to the original chamber passed bill The not absolute and that t here is consequential variation in agenda control at the committee level (in this case, conference committees). Let me repeat an earlier statement: I do not regard parties as organizationally weak or ineffective. Nonetheless, the implication is, yet agai n, that while parties exert significant influence in many areas of the policy process, especially on the chamber floor at the passage stage, at the postpassage sta ge partisan theories of lawmaking have considerably less explanatory power So what is the o verall narrative for contemp orary legislative politics ? Does this dissertation advocate based on the present research, is that the
279 accordance wit h the tenets of conditional party government. However, the overall narrative I advocate in this dissertation is that while the organizational capacity of the majority in the House and Senate has grown over the postreform period, this does not mean that le gislating is somehow easy or perfunctory: quite the contrary. Rather, I believe that conditional party government and the present work jointly make sense of what we know intuitively that while the majority is strong within the House and, to a lesser exten moderated or fails outright bec ause of bicameral disagreements (even during unified government) What we need to acknowledge, in my opinion, is that th or weak parties glosses over Instead, I believe a proper characterization of modern lawmaking is that the legislative process is marked by significant legislative risk Risk, in economics, refers to the amount of chance or volatility in the distribution of economic outputs. Economic exchange, under conditions of high risk, only occurs when the average or expected return is higher than the costs incurred. With respect to legislative politics, I believe that in the modern Congress parties do indeed receive higher payoffs on most bills consistent with conditional party government and the partisan theories of lawmaking In the cont ext of this work these high reward outcomes occur on those bills that pass the strength and ideological homogeneity are paramount. In the introduction, I noted that of the non commemorative public laws enacted from the 95 th to the 110 th Congress, 34%
280 experienced reconciliation (either a conference committee or amendment trading). 12 Thus, 65% of bills are unaltered by the second chamber. At the same time, Sinclair (2000b) s hows us that the likelihood of a bill becoming law increases dramatically when restrictive rules are used in both the House and Senate. But a failure rate of 34% is a sizable Figure Moreover, the number of bills that enter into a bicameral sequence onl y increases when we consider landmark legislation (Smith 1989; Longley and Oleszek 1989). In these cases, growing polarization and has contributed to inc reasingly costly policymaking; f or these bills we have seen either greater bicameral stalemate or grow ing pro minority outcomes. It is timely, with the previous discussion in mind, to address two important issues one an additional implication and one a limitation of the present research. Both hinge on how we conceptualize negative agenda control. Recal l that negative agenda control, the primary con struct animating the cartel theory (Cox and McCubbins 1993, 2005, 2007), posits that the leadership of the majority controls the consideration of policy and only advances legislation to final passage when the proposal is preferred by a sufficient number of its members relative to the status quo. Likewise, and most importantly, the leadership blocks proposals that are hostile to the interests of the majority. The primary the percentage of times a majority of the majority party votes on the losing side of a formal vote that ultimately passes. For example, Cox and McCubbins (2002) demonstrate that instances when the majority is (or 4 out of 5628) of committee reports. The two 12 database. The data on all conference committees and amendment tradin g in this period was compiled by the author.
281 issues that I would like to highlight with respect to negative agenda control are best explained by way of example. Recall from Chapter 5 discussion of the first 9/11 Commission Recommendations Act in th e 108 th Congress (HR 10 and S 2845 ). At the passage stage, the House bill and Senate bill contained major differences. What emerged from conference was a moderated bill were the majority particularly in the House lost a handful of preferred provisions. In particular, the conference report lacked a number of Republican backed provisions imposing more stringent immigration policies. Voting on the conference (compared to only eight at final passage) Now j uxtapose this case with a policy that occupies the exact same issue space: the second 9/11 Commission Recommendations Act in the 110 th Congress (HR 1 and S 4 ). The second time around, now with Democratic majorities in both chambers, the House and Senate once again passed different versions of legislation. Both chambers passed their respective bills with a majority of Democrats in voting favor of the proposal and a majority of Republicans in opposition. 13 Conference negotiators had to work out compromises where, in some cases, the House, Senate and administration each had different positions ( Yoest 2007) One of the major compromises struck in conference was approval of a provision, f avored by Republicans, protecting individuals who report suspicious activity to law enforcement from criminal and civil lawsuits (Yoest 2007) Voting on the conference report, only one Democrat switched his original yes vote to a no vote. However, 150 Re publicans in the House and 37 in the Senate now 13 The vote on the House bill (HR 1) had 231 Democrats in favor and 128 Republicans against while the vote on the Senate bill (S 4) had 49 Democrats in favor and 38 Republicans opposed.
282 supported the legislation where, prior to conference negotiations these members were opposed to the legislation an increase of 45% and 61% in each chamber, respectively. Based on the prior discussion the two implications I want to highlight are these. First, the Recommendations Act in the 108 th Congress is a clear example where the majority ceded or lost some degree negative agenda control. The changes made in conference s initial proposal s tripped the immigrat i on provisions that were clearly critical for many Republicans defection on the conference report). However, and I believe this is an important point, if we were to examine on ly the final vote on the conference report we would conclude th setting studies). Indeed, despite the clear moderation of the bill, Republicans voted 152 in favor of the conference repo rt an d 62 against. The implication is that the roll rate is in certain ways too blunt a measure to capture the more subtle (but nonetheless consequential) features of agenda control. An examination of policy modifications made in conference, and the meas ure s I have used in this d issertation, show a greater degree of variation in agenda control than t he conventional wisdom reveals. More research is needed examining agenda control using measures other than the roll rate. The second implication, unlike the first, is an importa nt weakness of the present work. In the second Recommendations Act ( during the 110 th Congress ) while it is clear that the minority gained a favorable conference outcome, it is not clear that the majority s of the bill (though of course they made some sacrifices). Indeed, the second Recommendations Act implemented most of the remaining 9/11 Commission recommendations a major Democratic campaign pr omise in 2006. In the
283 end, Democrats delivered on many of their stated goals including distributing more funding to state and local governments with a greater risk of terrorist attack and expanding the screening of international cargo. Thus, a limitation of the measures used in this study is that, in some cases we see postpassage bargaining where the aggregate outcome s may be ambiguous. That is, I can show critical changes as legislation traverses the bicameral sequence and discuss patterns in those changes, but those changes sometimes tell us little about how policy was. The Constitution, Political Parties and B icameralism In addition to the previous implications for the parties in Congress debate, the findings offer a parsimonious solution to an inconsistency in the con gressional literature. On the one hand, political parties have been cited for over a century as a solution to the policymaking inefficiencies built into the Constitution. From Woodrow Wilson (1885), E.E. Schattschneider (1942), V.O. Key Jr. (1942), to mo re recent political scientists like Cutler (1988), Kernell (1991) and Sundquist (1988), the conventional view is that strong parties bridge the gaps created by the Constitution and facilitate active government. But while researchers have found support for this claim with respect legislative executive policy productivity and change (Binder 1999, 2003; Conley 2002; Edwards and Barrett 2000; Cameron 2000; Alt and Lowry 1994; Cutler 1988; McCubbins 1991; Ragusa 2010; Kelly 1993 ; Edwards Barrett and Peak 1997) no work has explored this process vis vis House Senate policy productivity. However, a clear paradox exists, revealing the limits of our current understanding. More recently, a handful of researchers have cited too much polarization a key determinant of strong parties as one cause of growing stalema te and gridlock ( Binder 2003; Dodd and Schraufnagel
284 predicted the exact opposite. Two researchers have attempted to reconcile this lacuna. Dodd and Schraufnagel (2009) propose that the relationship between polarization and policy activity is curvilinear. At the extremes high and low polarization governing institutions are increasingly stalemated because of a de cline in delibera tion ( Dahl 1967). Binder (2003), citing work by Fiorina (2001), maintains that the negative effect of polarization on policy activity is due to a decline in the number of moderates. With the disappearance of moderates, the population of congressmen willi ng to compromise with political rivals has disappeared as wel l ( Binder 2003, 80). The findings reported here offer a slightly different perspective on the determinants of the paradoxical relationship between polarization, strong parties and stalemate. T he explanations offered by Dodd and Schraufnagel (2009) and Binder (2003) posit that the link between polarization and gridlock exists at the passage stage. That is, stalemate occurs caused when lawmakers are unable to pass legislation out of the House or Senate (either because of too few moderates or too little deliberation). The present research, by comparison, portends the existence a causal mechanism at the post passage stage. This postpassage perspective reconciles the core premise of the earlier wor k that polarization has contributed to gridlock in recent decade s with the fact that legislating has not declined over the postreform period in either chamber. Earlier I showed that, in the Senate, roughly the same amount of introduced legislation is pass ed in the most recent congresses as during the initial post reform congresses. And in the House, I showed that legislation is actually passing at a greater rate. This
285 makes good sense as in the House, with greater polarization and greater intra party ho mogeneity, we would expect less stalemate over time, not more (consistent with the tents of conditional party government). What reconciles this matter is the finding that polarization has occurred at different rates in each chamber over time and the evide nce that this factor has caused an increase in the frequency and severity of bicameral conflict. Thus, though legislative executive conflict certainly plays a role in growing stalemate ( Conley 2002), I find clear evidence that postpassage conflict between the House and Senate plays a central role as well. In summary, much like the implications with respect to the parties in Congress literature, the findings offer a solution to some contradictory historical patterns. But unlike the academic strengt h of p arties debate, the relationship between polarization and bicameral conflict has very real implications. Namely, this issue gets at the heart of how our macro level institutional arrangements function The normative im plications can be viewed as either g ood bad depending on how one views stalemate. Either the constraining effect of bicameral conflict is a positive, because the limits on single party control enhance representative lawmaking and foster stable public policies, or a negative, because t hampers its democratic spirit. In the following pages I present both views and leave it up to the reader to form their own opinion. On the positive side, one might conclude that our sy stem of checks and balances at least with respect to legislative checks and balances is functioning in a manner consistent with what the Framers envisioned. Indeed, many delegates to the Constitutional C a label used to de scribe like minded
286 political coalitions, including political parties. For example, in his farewell address to the the spirit of party are sufficient to make it the interest and duty of a wise people to discourage and And Gordon Wood summarizes the mindset of Constitutional delegates, particularly those who favored scrapping the Articles of Confederation, this way measure of a free government has become its ability t the Framers, factions writ large including organized parties, wer e seen as problematic. In fact John Paul Stevens cited Federalist 10 in California Democratic Party v. Jones 14 And a bicameral legislature composed of dissimilar men would help accomplish the ends of mixed government and ens sheltered from the vices of majority rule ( Wood 1998, 503). if their constitutional decisions were based on instrumental rationality alone Nonetheless, it is worthwhile to juxtapose how the Framers envisioned the operation of our bicameral legislature with the trends and effects this dissertation has uncovered. In particular, it is an interesting parallel that empowerment of the parties o ver the postreform period due in large part to growing polarization that fundamental differences in the distribution of preferences across the House and S enate have increased in salience as a check on the organizational power of parties. Indeed, differen 14 California Democratic Pa rty v. Jones 530 U.S. 567, 592 (2000).
287 selection, apportionment and the length of terms) were thought to be a crucial check on the vices of faction (e.g. parties). Thus, to the extent that one favors limited government and legislative constraints on single party rule, the conclusions from this dissertation might be regarded as welcome news that our system of government at least in this area is functioning One is reminded of Wi This conclusion may be welcome news for many citizens, in fact, as public opinion polls show that Americans prefer checks on government power ( Lewis Beck and Nadeau 2004 or Sigelman, Wahlbeck, and Buell, 1997, 880). Therefore even if parties capture both houses of Congress and the presidency, are ideologically motivated and empowered by the rules of the House and Senate, our Madisonian system presents additional hurdles. In contrast, the conclusions of this dissertation are normatively troublesome to the extent that one favors active government over limited government. Indeed, while the Framers implemented obvious constraints on simple majority rule, they also envisioned an effective government. A s Charles O. Jones (1995, 2) that the Founders were seeking to de The Constitutional Convention is inherent proof of this. Indeed, the reason for convening the Convention in the first place was to modify or undo the Articles of Confederation and establish a more centralized and active form of government. I n Federalist no. 40, James Madison, a Let [opponents of the Constitution] declare, whether it was of most importance to the happiness of the people
288 of America, that the Articles of Co nfederation should be disregarded, and an adequate Woodrow Wilson was perhaps the first political scientist to articulate the limitations of constitutional checks on active government. Wilson, of course, lamented our system of limited government, favoring a constitution with an explicit party role. In one passage, Wilson (18855) government is a living, organic thing, and must . work out a close synthesis of active parts which can exist only when leadership is lodged in some one man or group of men. You cannot compound a successful government out of antagonisms Ranney 1951 35, emphasis added). Subsequent political scientists echoed these calls, advocating reforms to promote more active government (Burns 1963; Schattschneider 1942; Ke y Jr. 1942; Cutler 1988; Kernell 1991; Sundquist 1988 ). Thus, the negative implication of the link between polarization, strong parties and bicameral constraints is that the resulting growth in House Senate conflict is problematic because it fosters legislative inaction. Such inaction is bad for two reasons. On the one hand, legislative inaction can lead to an abdication of governing authority. Needless to say, d iminishing congressional a uthority can be problematic for representative democracy. Such an argument has been applied most famously to the textbook era Congresses (1912 1968). For example, after the reforms which undermined the organizational capacity of Congress in 1910, Carmine s and Dodd (1985, 280 bicameralism now became an inducement toward weakness, a constitutional provision that helped increase the leverage of the executive and has more immediate and tangible effects when government is unable to respond to a social or
289 economic crisis. response to the 2007 2008 housing and financial crisis is a case in point. By January of 2008, the U.S. housing market had been in a 2 year decline due in large part to the collapse of the subprime mortgage market. In May of 2007, the year Democrats regained control of the House and Senate for the first time since 1995, the House acted in response to the crisi s with the passage of HR 1427, a bill to regulate mortgage firms Fannie Mae and Freddie Mac and create a housing trust fund to bolster the housing market The House bill died in the Senate, the second effort in as many years. Indeed, despite unified Repu blican control in the previous Congress similar legislation failed because of disagreements between the House and Senate proposals ( In November of 2007, the House again passed legislation in response to the ho using crisis. This second bill, HR 3915, was designed to curb predatory lending practices and regulate mortgage brokers through uniform standards. Though Chris Dodd Chairman of the Senate Banking, Housing, and Urban Affairs committee introduced a similar bill in the Senate, there were proposal came close to addressing the crisis a third time when bot h houses passed separate bills expanding the power of the Federal Housing Administration to guarantee mortgage loans (HR 1852 and S 2338). Despite the obvious crisis, the bills were never reconciled in conference or via amendment trading in large part bec ause of major disagreements over an affordable housing fund included in the House bill This fund was also passed in a separate House bill (HR 2895)
290 though it died in the Senate. In sum by the end of 2007, in ad dition to the bills stuck at various stages within each chamber, the House had cleared four separate bills responding to the housing crisis whil e the Senate passed one. None made it to the postpassage stage in large part because of significant bicameral d ifferences By February of 2008, the House and Senate began drafting a larger reform package combining elements of the bills drafted and/or passed in 2007. The Senate acted first, clearing its package (HR 3221) on April 10 along bipartisan lines (84 12). 15 The central element of the Senate package was a series of reforms to the FHA mortgage program. These provisions were similar to previous efforts that failed in House and Senate negotiations (HR 1852, S 2338) (Ives 2008) The House acted second, subst ndments, clearing its package on May 8 th As with the earlier efforts, there were key differences between 16 In particular, the House plan inc luded a new federal regulator for Fa nnie and Freddie as well as provisions containing the original housing trust fund. The Senate acted third, offering a (second degree) amendment to the House proposal in June of 2008. The two chambers were now engaged in formal reconciliation via am endm ent trading. The substitute proposal passed by the Senate was a compromise worked out by Dodd and Richard Shelby (the Ranking Member of Senate Banking, Housing, and Urban Affairs Committee ) Though the House and Senate proposals were more consonant than in February, critical differences remained 15 HR 3221 began as an energy bill. The Senate deleted the original text of HR 3221 and inserted the reform package. 16 The key differences centered on the fact that the House bill increased the limit the FHA could in sure loans to over $700,000 while the Senate bill would increase the limit up to $417,000 (Ives 2007).
291 exchange was disrupted on July 13 th when Henry Paulson the U S Treasury Secretary asked Congress for sweeping authority to invest capit al into, and potentially take over, Fannie and Freddie. Though his request represented a significant abdication of Congressional authority, further Congressional inaction was potentially cataclysmic. The seriousness of the housing and financial crisis sp urned congressiona l action and eping authority was added to HR 3221 In lar ge part because of the escalation of the housing and financial crisis compromise s on the remaining disagreements were quickl y worked out between key House and Senate lawmakers. The House inserted the compromise plan into HR 3221 and passed it 272 152. The Senate followed suit, passing the same bill on July 26 th by a vote of 72 13. Of final version passed by both chambers Ba is inconceivable to me that anybody would like everything in this bill it is a product of a very significant set of compromises It is true that in the 110 th Congress partisan division s between Democrats and Republicans House all played a role in sluggish response to the housing crisis. However, even when there was wide bipartisan agreement on the core pro visions of a Congressional response disagreements between legislation passed by House and Senate remained. Thus, more than a year after the first bill had passed the first chamber, the House and Senate finally reached an acceptable compromise. That comp romise was reached only after competing proposals were merged into one larger package and a lengthy process of amendment trading took place. Surprisingly, t his
292 package did little more than combine the earlier House and Senate proposals died However, in one of the biggest crises of the past fifty years, the House and Senate were marked by signif icant disagreement resulting in delayed action. Moreover, Congress delegated significant power to the executive branch (the Treasury) : unwelcome news for Democrat s, Republicans, representatives and senators alike The Nature of Resolving Differences The findings of this dissertation also have implications of for how we characterize conference committees and postpassage bargaining Indeed, it is my hope that the mu ltivariate results presented in Chapter 3 and 4 provide a descriptive richness to the complex patterns by which competing chambers and parties merge bills passed by the House and Senate in disagreement. But w hile these implications exist with respect to o ur basic understanding how conference committees function, the implications are also applicable to our larger theories of congressional organization. As mentioned throughout this manuscript, we lack a strong understanding of th ese matter s The oldest pe rspective on conference committees and resolving differences is the bicameral conflict paradigm. During the 1960s and 1970s, the pervasive question posed in research on conference committees was: Which chamber dominates conference outcomes? As reviewed a t multiple points in this dissertation, the consensus suggests that the Senate wields greater leverage in the conference process (Fenno 1966; Manley 1970; Vogler 1970; Strom and RUndquist 1977). A second perspective falls under a larger class of theories known as distributive politics ( Weingast 1979; Tullock 1967; Shepsle and Weingast 1987a 1987b) and foremost.
293 In order to best represent their district or state and distribute legislative goods to their constituents, lawmakers self select onto beneficial committees. As a result, committees are unrepresentative of their chamber. Behavior among mem bers according to distributive politics is typically cooperative in the production of public policy, particularly behavior such as logrolling (Tullock 1967). Regarding conference committees, Shepsle falls under this general class of theory. Shepsle and Weingast identified the bicameral sequence as the foundation for committee agenda control. Because standing committee members are almost always named as managers in conference, and because conference agreements are considered under what amounts to a closed rule in the House, committees wield the ability to modify any policy altered on the floor back to its original committee approved position (or simply def eat it). Thus, conference outcomes are determ ined by the committee (rather than the majority or chamber). posits that legislative institutions are organized in order to collect, analyze and transmit policy information back to the chamber. Based on principal agent theory where the principal is the chamber median and the agents are the committees the basic logic is that information is costly to obtain and evaluate (a problem committees solve). Policy outcomes, accordin g to information theory, will be chamber median. Krehbiel has analyzed information theory with respect to conference committees. Krehbiel (1991,
294 committee. But, if we apply an informational lens and consider the role of expertise in conference committee 216). Finally, partisan theories of organization propose that legislative institutions are organized to benefit the majority party. These theories emphasize the fact that the majority party controls consequential organizational decisions in the House and Senate such as the naming of committee and subcommittee chairs, the appointment of committee members, co ntrol over the legislative agenda, and (in the House) the power to elect the Speaker who, in turn, names members to the powerful Rules Committee. Like the informational model, the partisan perspective sees committees as agents in the principal agent rela tionship. However, unlike the information theory, the principal is the chamber. In this way, like distributive theory, committees are composed of preference outliers and p olicy outcomes are non majoritarian. With respect to conference to name managers yields the naming of a biased conference and thus a pro majority outcome (Lazarus and Monr oe 2007; Vander Wielen and Smith n.d.). The view of conference committees their primary characteristics and patterns explicated in this dissertation differs in many ways from the prior perspectives. It is important to stress that the theoretical tenets gu iding the present work do not represent a full elaboration of legislative organization. Rather, the work is specific to conference committees and postpassage bargaining. Though I make a v ariety of theoretical
295 arguments the view I present is perhaps best labeled a What I mean by this label is that the primary role of a conf erence committee is one of successfully and efficiently merging bicameral proposals passed in disagreement. Thus, my view emphasizes compromise and majoritarianism a s the overarching characteristics of conference outcomes. First, I believe the methods of resolving House and Senate disagreements are typically complex in terms of the political divisions among pivotal actors and multidimensional in the issue space. The findings in Chapter 3 using multivariate spatial modeling to uncover the latent dimensions of resolving differences support these two tenets finding that, indeed, conference committee roll call patterns exhibit evidence of high dimensionality. The rather simple poi nt is that straightforward, uni dimensional outcomes are less frequent in conference than difficult policy changes and broad compromises. In addition to these inherent challenges, I argued that resolving differences is a process marked by uncert ainty, significant transaction costs and, untimely, risk averse legislative decision making. That is, there is great uncertainty about what piv otal actors in each chamber and party will accept as legitimate compromises (hence, why bicameral disagreement e xists in the first place). And because the legislative transaction costs have already been incurred when a bill reaches the postpassage stage (such as time spent studying, drafting, marking up and passing legislation) there is great risk associated with c onferencing. Simply put, if a conference bill dies, or the agreement causes significant delays in adoption of a conference report, the cost to lawmakers who voted to pass the bill or invested their limited resources in the policy outweigh the costs of pro sing a simple compromise.
296 Because of these issues, I have argued there is an incentive ceteris paribus to adopt an efficient solution one that quickly satisfies competing actors. Compromise, in other words, is an efficient, utility maximizing strategy e ven for the majority Though there are certainly cases in my dataset of pro majority or pro chamber outcomes, the empirical evidence across the chapters supports the notion that majoritarian outcomes are the rule rather than the exception in conference. The implications for how we characterize postpasasge bargaining are clear. Beyond the descriptive account of conference committee patterns, the main implication is that we have tended to overlook what is the central feature of conference committee politi cs: the role of compromise. This is despite the claims of some of our leading scholars: and take, compromise, horse wonder that each side claims victory, because most everyone does win e possible and the essence of the legislative process is compromise then the conference committee is the ep (Jewell and Patterson 1986, 169) 1989, 2)
297 onstrate that recognition of the cooperative behavior between the House and Senate conferees can lead to a much different interpretation of data on conference decision But less clear are the implications for how we think abo ut key features of legislative organization. One implication is that, contrary to our tendencies are researchers, not all aspects of legislatures are organized or operate an identical fashion at the micro level In fact, key features of legislative organ ization need not operate in complimentary fashion but can, instead, rival one another. Indeed, and as I have stated, my personal view is that the respective chambers are organized in a partisan manner. Parties control too many consequential organizationa l decisions to have limited effect s at the passage stage. However, I have argued in this dissertation that the same perspective does not explain the operation of conference committees (both their general patterns of operation and historical developments). Thus, an implication for future research is that we need not adopt universalistic perspectives but should, instead, consider the possibility that institutional variation at lower levels of aggregation create kaleidoscopic patterns ( Schickler 2001) as no t all bills are subject to the same institutional arrangements or the same organizational factors (Maltzman 1998) In one area of the policy process one perspective may dominate (such as the partisan perspective at the passage stage) while in other areas of the policy process another perspective may dominate (such as a non partisan perspective at the postpassage stage).
298 Table 6 1. Conference committee failure rate Congress Moderate failure Major failure Total conferences Moderate failure rate Major fai lure rate 98 2 4 90 0.02 0.04 99 7 0 74 0.09 0 100 7 2 84 0.08 0.02 101 8 2 96 0.08 0.02 102 3 1 87 0.03 0.01 103 2 4 71 0.03 0.06 104 2 1 65 0.03 0.02 105 2 1 50 0.04 0.02 106 1 3 60 0.02 0.05 107 3 0 38 0.08 0 108 6 2 47 0.13 0.04 109 2 1 31 0.07 0.03 110 3 0 22 0.14 0 Notes: Compiled by the author using CQ Weekly summaries.
299 Figure 6 1. Average public law length for bills referred to conference This data was compiled by the author from Lexis Nexis Congressional.
300 LIST OF REFEREN CES "2000 Legislative Summary: Managed Care." CQ Weekly Online (December 16, 2000): 2907. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/weeklyreport106 000 000187589 ; Accessed 10/13/2010. CQ Weekly Online (December 16, 2000): 2891. http://library.cqpress.com.lp.hscl.ufl.ed u/cqweekly/weeklyreport106 000000187642 ; Accessed 11/15/2010. 2004 Legislative Summary: Corporate Tax Overhaul." CQ Weekly Online (December 4, 2004): 2868. htt p://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/weeklyreport108 000001444264 ; Accessed: 12/2/2010. "2004 Key Votes: A Mix of Hits and Misses For Republican Leadership." CQ Weekly Online (December 11, 2004): 2912 27. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/weeklyreport108 000001453470 ; Accessed 11/15/2010. Redistricting for U ncompetitive Elections." PS: Political Science and Politics 39 (1): 87 90. American Journal of Political Science 41 (3):718 73. Aldrich, John H. 1995. Why Parties?: The Origin and Transformat ion of Political Parties in America Chicago, IL: University of Chicago Press. Political Science Quarterly 11 2 (4): 541 567. The Journal of Politics 62 (1): 1 33. Aldrich, John H. and David W. Rohde. 2001. "The Logic of Conditional Party Government: Revi siting the Electoral Connection." In Congress Reconsidered Seventh Edition, eds. L. C. Dodd and B. I. Oppenheimer. Washington, D.C.: CQ Press. ionalism, ed. B. I. Oppenheimer. Columbus, OH: The Ohio State University Press. American Political Science Review 88 (4): 8 11 828.
301 in U.S. Elections: An Analysis of State and Federal Offices, 1942 Election Law Journal 1 (3): 315 38. "Attempts to Rein In Mortgage Giants Stall." CQ Almanac 200 5 61st ed., 3 11 3 12. Washington, D.C.: Congressional Quarterly. http://library.cqpress.com.lp.hscl.ufl.edu/cqalmanac/cqal05 766 20105 1042414 ; Accessed 2/10/2011 Aufderheide, Patricia. 1999. Communications Policy and the Public Interest: The Telecommunications Act of 1996 New York, NY: Guilford Press. The Journal of American History 60 (3 ): 679 691. Baker, Ross. 2009. House and Senate, Fourth Edition. New York, NY: W.W. Norton and Company. Journal of Politics 68 (2): 345 57. Early Congress, 1789 1118. American Political Scie nce Review 93 (3): 519 33. Binder, Sarah A. 2003. Stalemate: Causes and Consequences of Legislative Gridlock Washington, D.C. : Brookings Institution Press. Extensions, A Journal of the Carl Albert Congressional Research and Studies Center Fall 2005. Political Analysis 16 (2): 213 225. Bishop, Bill, and Robert Cushing. 2008. The Big Sort: How the Clust ering of Like Minded Americans is Tearing us Apart New York, NY: Houghton Mifflin Company Bond, Jon R., and Richard Fleisher. 1990. The President in the Legislative Arena Chicago, IL: University of Chicago Press. Bond, Jon R., and Richard Fleisher. 2000. Polarized Politics: Congress and the President in a Partisan Era Washington, D.C. : Congressional Quarterly Press. Borg, Ingwer and Patrick Groenen. 2005. Modern Multidimensional Scaling: Theory and Applications Second Edition New York, NY: Springer Ver lag.
302 Brady, David W., and Craig Volden. 1998 Revolving Grid lock: Politics and Policy from Carter to Clinton Boulder, CO: Westview Press. Brady, David W., and Craig Volden. 2006. Revolving Grid lock: Politics and Policy from Jimmy Carter to George W. Bush, Second Edition Boulder, CO: Westview Press. Burns, James MacGregor. 1963. The Deadlock of Democracy: Four Party Politics in America. Englewood Cliffs, N.J.: Prentice Hall. Cameron, Charles. 2000. Veto Bargaining: Presidents and the Politics of Negative Power New York, NY: Cambridge University Press. Party, Process, and Political Change in Congress: New Perspectives on the History of Congress ed D. W. Brady and M. D. McCubbins. Stanford, CA: Stanford University Press. Carney, Dan. 1995. "Talks Begin To Show Progress, But Agreement Looks Elusive." CQ Weekly Online (November 11, 1995): 3451. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/WR409366 .; Accessed 11/12/2010. Carrol, Royce, Jeffrey B. Lewis, James Lo, Keith T. Poole, and Howard Rosenthal. NOMINATE Ideal Point Estimate s via the Parametric Bootstrap." Political Analysis 17(3): 261 275. Congress Reconsidered Third Edition, eds. L. C. Dodd and B. I. Oppenheimer. Washi ngton, D.C.: CQ Press. Carmines, Edward G., and James Stimson. 1989. Issue Evolution: Race and the Transformation of American Politics Princeton, NJ: Princeton University Press. Carson, Jamie L. Michael Crespin, Charels Finocchario and David Rohde. 200 7. American Politics Research 35 (6): 878 904. Unpacking Agenda Control in Congress: Individual Roll Rates and the Political Research Quarterly 64 (1): 17 30. 2002 meeting of the Northeastern Political Science Association. Chiou, Fang Yi, and Lawrence S. Rothenberg. 2003. When Pivotal Politics meets Partisan Politics. American Journal of Political Science 47 (3): 503 22.
303 Chiou, Fang Legislators and Political Analysis 16 (2): 197 212. Chiou, Fang Political Analysis 16 (2):226 233. CQ Weekly online (October 13, 1984): 2623 2623. http://librar y.cqpress.com.lp.hscl.ufl.edu/cqweekly/WR098403753 Accessed: 10/11/2010. More Responsible Two American Political Science Review 44 (3): Supplement. Con ley, Richard S. 2002. The Presidency, Congress and Divided Government: A Postwar Assessment College Station, TX: Texas A&M University Press. Connelly, William F., and John J. Pitney. 1997. "The House GOP's Civil War: A Political Science Perspective." P S: Political Science and Politics 30 (4): 699 702. Politics in Late Nineteenth Congress Reconsidered Eighth Edition, edited by L. C. Dodd and B. I. Oppenheimer. Washington, D.C. : CQ Press. Cox, Gary W., and Mathew D. McCubbins. 1993. Legislative Leviathan: Party Government in the House Berkeley, CA: University of California Press. he U.S. House of Representatives, 1877 Party, Process, and Political Change in Congress: New Perspectives on the History of Congress ed. D. W. Brady and M. D. McCubbins. Stanford, CA: Stanford University Press. Cox, Gary W., and Mathew D. McCubb ins. 2005. Setting the Agenda: Responsible Party Government in the U.S. House of Representatives Cambridge, MA: Cambridge University Press. Cox, Gary W., and Mathew D. McCubbins. 2007. Legislative Leviathan: Party Government in the House, Second Edition New York, NY: Cambridge University Press. and Bills: Appropriations Journal of Politics 72 (4): 976 989.
304 Critchlow, Donald T. 2004. The American Congress: The Building of Democracy New York, NY: Houghton Mifflin Company. Crook, Sara Brandes, and John R. Hibbing. 1997. so distant Mirror: The 17th American Political Science Review 91 (4): 845 54. Presidential Studies Quarterly. 18 (3): 485 492. Dahl, Robert A. 1 967. Pluralist Democracy in the United States: Conflict and Consent Chicago, IL: Rand McNally. Davidson, Roger H. 1981. "Two Avenues of Change: House and Senate Committee Reorganization." In Congress Reconsidered, Second Edition ed. L.C Dodd and B. Oppen heimer. Washington, D.C. : Congressional Quarterly Press. Deering, Christopher J., and Steven S. Smith. 1997. Committees in Congress, Third Edition Washington, D.C. : Congressional Quarterly Press. f Majority Status: The Legislative Studies Quarterly 33 (1): 63 84. The Journal of Politics 58 (1): 25 53. System: The Ninety Third and Ninety Congressional Studies 7(1): 27 56. Dodd. 1981. Congress, the Constitution, and the Crisis of Legitimat ion. In Congress Reconsidered Second Edition. ed. L. C. Dodd and B. I. Oppenheimer. Washington, D.C.: CQ Press. Dodd, Lawrence C. 1986. A Theory of Congressional Cycles: Solving the Puzzle of Change. In Congress and Policy Change edited by G. C. Wright J r., L. Rieselbach, and L. C. Dodd. New York, NY: Agathon Press. Divided Democracy: Cooperation and Conflict Between the President and Congress ed. J. Thurber. Washington, D.C. : Congressional Quarterly Press. Congress Reconsidered Second Edition ed. L.C. Dodd and B. Oppenheimer. New York, NY: Praeger.
305 Dodd, Lawren Congress Reconsidered Sixth Edition, edited by L. C. Dodd and B. I. Oppenheimer. Washington, D.C. : CQ Press. Dodd, Lawrence C., and Richard L. Sch ott. 1979. Congress and the Administrative State New York, NY: Wiley. Congress Reconsidered, Ninth Edition ed. L.C. Dodd and B. Oppenheimer. Washington, D.C. : Congressional Quarterly Press. Polarized Politics: Congress and the President in a Partisan Era ed. J.R. Bond and R. Fleisher. Washington, D.C. : Congressional Quarterly Press. American Journal of Political Science 41 (2): 545 563. Evans, C. Lawrence, and W Esteemed Colleagues: Civility and Deliberation in the Senate ed. B. Loomis. Washington, D.C. : Brookings Institution Press. Fenno, Richard F., Jr 1966. The Power of the Purse. New York, NY: Little, Brown. Ferejohn, John. 1975. Who Wins in Conference Committee? The Journal of Politics 37 (4): 1033 1046. CQ Weekly Online (July 24, 2004): 18 15 17. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/weeklyreport108 000001270290 ; Accessed 6/14/2011. for the Floor: Partisan Theory and Legislative Studies Quarterly 33 (1): 35 61. Fiorina, Morris. 1996. Divided Government. Boston, MA: Allyn Bacon. Congre ss Reconsidered, Seventh Edition (ed.) L.C. Dodd and B.I. Oppenheimer. Washington, D.C. : Congressional Quarterly Press. Polity 29 (3): 375 396.
306 Gailmard, Sean, Intra cameral Journal of Politics and House: Fingerprints of Majority Party P ower." Journal of Politics 69 (3): 689 700. Gimpel, James. G., and Jason E. Schuknecht. 2003. Patchwork Nation: Sectionalism and Political Change in American Politics Ann Arbor, MI: The University of Michigan Press. Green Donald, Bradley Palmquist, and E ric Schickler. 2002. Partisan Hearts and Minds New Haven, CT: Yale University Press. Majority Rule." CQ Weekly Online (November 12, 1994): 3226 27. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/WR103406359 ; Accessed 10/13/2010. H Legislative Studies Quarterly 35 (1): 177 143. Demand for Attributes with An Empirical Application to Estimating the Preferences of RAND Journal of Economics 28 (special issue): 142 189. Legislat ive Studies Quarterly 22 (4): 485 516. The New York Times. Committee Appointments in the U.S. House, 104 the 2004 Annual meeting of the Midwest Political Science Ass ociation. Ives, Benton. "Mortgage Relief Moves Forward." CQ Weekly Online (April 7, 2008): 880 84. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/weeklyrepor t110 000002697936 ; Accessed 2/10/2011. Jackson, J. Edward. 2003. New York, NY: John Wiley & Sons. Polarized Politics: C ongress and the President in a Partisan Era ed. J.R. Bond R. Fleisher. Washington, D.C. : Congressional Quarterly Press. Jewell, Malcolm E., and Samuel C. Patterson. 1986. The Legislative Process in the United States Fourth Edition. New York, NY: Random H ouse.
307 (3): 435 458. ress, American The American Political Science Review 89 (1): 1 9. Political Research Quarterly 54 (1): 125 141. Kady II, Martin. "Intelligence Overhaul's F ate Still Mired in Turf Disputes." CQ Weekly Online (November 20, 2004): 2746 2748. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/weeklyreport108 0000014305 85 ; Accessed 3/5/2010. Polity 25 (3): 475 495. The Politics of Divided Government, ed. G. Cox and S. Kernell. Boulder, CO: Westview Press. American Political Science Review 90 (3): 555 566. Key, V. O., Jr. 1942 Politics, Parties, and Pressure Groups New York, NY: Thomas Y. Crowe ll. Kiewiet, D. Roderick, and Mathew D. McCubbins. 1991. The Logic of Delegation: Congressional Parties and the Appropriations Process Chicago, IL: University of Chicago Press. Political Analysis 9 (2): 137 163. Legislative Studies Quarterly 28 (2): 225 246. Koopman, Douglas L. 1996. Hostile Takeover: The House Republican Party, 1980 1995 Lanham, MD: Rowman & Littlefield. Krehbiel, Keith. 1991. Information and Legislative Organization Ann Arbor, MI: University of Michigan Press. 235 266. American Journal of Political Science 39 (4): 906 23.
308 Krehbiel, Keith. 1996. "Institutional and Partisan Sources of Gridlock: A Theory of Divided and Unified Government." Journal of Theoretical Politics 8 (1): 7 40. Kr ehbiel, Keith. 1998. Pivotal Politics: A Theory of U.S. Lawmaking Chicago, IL: University of Chicago Press. Legislative Studies Quarterly 21 ( 1): 33 69. Committee Appointments in the 97 th through 106 th Political Research Quarterly 60 (4): 593 606. ate Apportionment: 24. Lee, Frances E., and Bruce I. Oppenheimer. 1999. Sizing Up the Senate: The Unequal Consequences of Equal Representation Chicago, IL: University of Chic ago Press. Leege David C., Kenneth D. Wald, Brian S. Krueger, and Paul D. Mueller. 2002. Politics of Cultural Differences: Social Change and Voter Mobilization Strategies in the Post New Deal Period. Princeton, NJ: Princeton University Press. Lewis Beck, M Ticket Voting: The Effects of 112. Lijphart, Arendt. 1984. Democracies: Patterns of Majoritarian and Consensus Government in Twenty One Democracies Ne w Haven, CT: Yale University Press. Longley, Lawrence R., and Walter J. Oleszek. 1989. Bicameral Politics: Conference Committees in Congress New Haven, CT: Yale University Press. Maltzman, Forrest. 1998. Competing Principals: Committees, Parties, and the Organization of Congress Ann Arbor: University of Michigan Press. McCown, Ada. 1927. The Congressional Conference Committee. New York, NY: Columbia University Press. Madison, James. 1987. Notes of Debates in the Federal Convention of 1787 New York, NY: Norton Press. American Journal of Political Science 52 (2): 252 267. Manley, John. 1970. The Politics of Finance Boston, MA: Little, Brown. Mayhew, David R 1974. Congress: The Electoral Connection New Haven, CT: Yale University Press.
309 Mayhew, David R. 1991. Divided We Govern: Party Control, Lawmaking, and Investigations, 1946 1990 New Haven, CT: Yale University Press. Mayhew, David R. 2005. Divided We Go vern Second Edition. New Haven, CT: Yale University Press. McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2006. Polarized America: The Dance of Ideology and Unequal Riches. Cambridge, MA: MIT Press. .S. Budget Deficits: Divided Government and Fiscal Stalemate." In Politics and Economics in the Eighties ed. A. Alesina and G. Carliner. Chicago, IL: University of Chicago Press. Roll Call (October 4, 2004): 340. http://www.rollcall.com/issues/50_41/ 7084 1.html ; Accessed: 12/2/2010. Amer ican Politics Quarterly 17 (1): 54 79. Oleszek, Walter J. 2004. Congressional Procedures and the Policy Process, Sixth Edition. Washington, D.C.: Congressional Quarterly Press. Oleszek, Walter J. 2007. Congressional Procedures and the Policy Process, Seven th Edition. Washington, D.C.: Congressional Quarterly Press. Congress Reconsidered Eighth Edition, ed. L.C. Dodd a nd B.I. Oppenheimer. Washington, D.C. : Congressional Quarterly Press. Ornstein, Norman J., and Amy L. Sche nkenberg. 1995 Poli tical Science Quarterly 110 (2): 183 206. Ota, Alan K. "Technology: With Clinton Agreeing to Caps on Damages, Y2K Liability Legislation Clears Senate." CQ Weekly Online (July 3, 1999): 1616 17. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/WR19990703 27Y2K001 Accessed 3/5/2010. CRS Report for Congress Order Code 98 853 GOV. Parks, Daniel J. 1999. "Banking and Finance: Vote Margin on Financial Services Rewrite Gi ves House Leverage on Privacy Issues." CQ Weekly Online (July 3, 1999): 1618 19. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/WR19990703 27BANKING001 ; Accessed 10/13/2010. Polsby, Nelson W. 2004. How Congress Evolves New York, NY: Oxford University Press.
310 American Journal of Political Science 42 (3): 954 993. Poole, Keith T., and Howar d Rosenthal. 1997. Congress: A Political Economic History of Roll Call Voting Oxford University Press. Poole, Keith T., and Howard Rosenthal. 2007. Ideology and Congress New Brunswick, NJ: Transaction Publishers. Rabe Hesketh, Sophia, and Brian Everitt. 2007. A Handbook of Statistical Analyses Using Stata, Fourth Edition Boca Raton, FL: Chapman & Hall/CRC. Rabe Rabe Hesketh, Psychometrika 69 (2): 167 190. Rabe Likelihood Estimation of Limited an d Discrete Dependent Variables M odels with Journal of Econometrics 128 (2): 301 323. Repeals to Landmark Legislative Enactments, 1951 American Politic s Research 38 (6): 1015 1051. Party System: A The American Political Science Review 45 (3): 488 499. International Political Scie nce Review 13 (1): 101 16. PS: Political Science and Politics 28 (4): 703 707. Dimensionality of Congressio delivered at the 2008 Annual Meeting of the American Political Science Association. American Journal of Political Science 42 (4):1025 1060. Journal of Theoretical Politics 13 (2): 123 51. es Quarterly 30 (1): 29 42.
311 Rohde, David. 1991. Parties and Leaders in the Postreform House Chicago, IL: University of Chicago Press. In U.S. Senat e Exceptionalism, ed. B. I. Oppenheimer. Columbus, OH: Ohio State University Press. Public Choice 33 (1): 27 43. Rossiter, Clinton (ed.). 1961. The Federalist Papers New York, NY: New American Library. Schattschneider, E. E. 1942. Party Government: American Government in Action New York, NY: Rinehart and Co. Schickler, Eric. 2001. Disjointed Pluralism: Institutional Innovation and the Deve lopment of the U.S. Congress Princeton, NJ: Princeton University Press. Shepsle, Can the Government Govern? ed. J. Chubb and P. Peterson. Washington, D.C. : Brookings Institution Press. Shepsle, Kennet The American Political Science Review 81 (1): 85 104. Preference for Divided Government: L American Journal of Political Science 41(3): 879 894. Sin, Gisela, and Arthur Lupia. 2004. How the President and Senate Affect the Balance of Power in the House: A Constitutional Theory of Leadership Bargaining Unpublished Manuscript deli vered at the 2003 Annual Meeting of the Midwest Political Science Association. American Journal of Political Science 32 (2): 276 301. Si nclair, Barbara. 1989. The Transformation of the U.S. Senate Baltimore, MD: Johns Hopkins University Press. Sinclair, Barbara. 1995. Legislators, Leaders, and Lawmaking: The U.S. House of Representatives in the Postreform Era Baltimore, MD: Johns Hopkins University Press.
312 Sinclair, Barbara. 2000 a In Esteemed Colleagues: Civility and Deliberation in the Senate ed. B. Loomis. Washington, D.C. : Brookings Institution Press. Sinclair, Barbara. 200 0b. Unorthodox Lawmaking, Second Edition. Washington, D.C.: CQ Press. Congress Reconsidered Eighth Edition, ed. L. C. Dodd and B. I. Oppenheimer. Washington, D.C. : Congressional Quarterly Press Sinclair, Barbara. 2007. Unorthodox Lawmaking: New Legislative Processes in the U.S. Congress Washington, D.C.: Congressional Quarterly Press. Smith, Stephen. 1989. Call to Order: Floor Politics in the House and Senate Washington, D.C.: Brookings Instit ution Press. Hearings before the Committee on Rules and Administration, United States Senate, 111 th Congress, Second Session May 19, 2010. Washington, D.C. : Government Printing Office. Smith, Steven S., and Congress Reconsidered, Ninth Edition ed. L.C. Dodd and B. Oppenheimer. Washington, D.C. : Congressional Quarterly Press. StataCorp. 2009. STATA Multivariate Statistics Reference Ma nual Release 11 College Station, TX: Stata Press. Steiner, Gilbert Y. 1951. The Congressional Conference Committee: Seventieth to Eightieth Congresses. Urbana, IL: University of Illinois Press. Stewart III, Charles. 2001. Analyzing Congress. New York, NY : W.W. Norton & Co Stimson, James A. 1999. Public Opinion in America: Moods, Cycles, and Swings, Second Edition. Boulder, CO: WestviewPress Stonecash, Jeffrey M.,Mark D. Brewer, and Mack D.Mariani. 2003. Diverging Parties: Social Change, Realignment, and P arty Polarization Boulder, CO: Westview Press. American Political Science Review 71 (2): 448 53. Sundquist, James L. 1983. Dynamics of the Party System: A lignment and Realignment of Political Parties in the United States Washington, D.C. : Brookings Institution Press. Sundquist, James L. 1988. "Needed: A Political Theory for the New Era of Coalition Government in the United States." Political Science Quarte rly 103 (4): 613 35.
313 Dimensional Journal of Politics 64 (3): 864 891. and the Sequence of American Politics Research 36 (3): 451 474. "Tightening Reins on Mortgage Lenders." CQ Almanac 2007 63rd ed., 7 5 7 6. Washington, D.C.: Congressional Quarterly. http://library.cqpress.com.lp.hscl.ufl.edu/cqalmanac/cqal07 1006 44916 2048045 ; Accessed 2/10/2011 "Treasury Gets Keys to Fannie, Freddie." CQ Almanac 2008 64th ed., 7 9 7 13. Washington, D.C.: Congressional Quarterly. http://library.cqpress.com.lp.hscl.ufl.edu/cqalmanac/cqal08 1090 52026 2174917 ; Accessed 2/10/2011. Tsebelis, George. 2002. Veto Players: How Institutions Work Princeton, NJ: Princeton University Press. Tsebelis, George, and Jeanette Money. 1997. Bicamera lism New York, NY: Cambridge University Press. Tullock, Gordon. 1967. The General Irrelevance of the General Impossibility Theorem. Quarterly Journal o f Economics 81 (2): 256 70. Esteemed Colleagues: Civility and Deliberation in the Senate ed. B. Loomis. Washington, D.C. : Brookings Institution Press. Van Beek, Stephen D. 1995. Post Pas sage Politics: Bicameral Resolution in Congress Pittsburgh, PA: University of Pittsburgh Press. The Influence of Conference Committees on Policy Outcomes." Legislative Studies Quarterly 35(4): 487 518. Vander Wielen, Ryan J., and House Dominance in Congressional Conference Midwest Journal of Political Science 14 (2): 303 2 0. Wawro, Gregory. 2000. Legislative Entrepreneurship in the U.S. House of Representatives. Ann Arbor, MI: University of Michigan Press. Journal of Public Policy 6 (4): 371 98.
314 Weingast, Barry R. 197 9. A Rational Choice Perspective on Congressional Norms. American Journal of Political Science 24 (2): 245 263. Woodburn, James A. 1893. "The Historical Significance of the Missouri Compromise," American Historical Association Annual Report 251 297. Woon, Journal of Politics 70 (2): 201 216. Cambridge, MA: Belknap Harvar d University Press. Wilson, Woodrow. 1885. Congressional Government: A Study in American Politics Cambridge, MA: Harvard University Press. Wirls, Daniel. 1998. "The Consequences of Equal Representation: The Bicameral Politics of NAFTA in the 103rd Congr ess." Congress and the Presidency 25 (2): 129 145. Wirls, Daniel. Studies in American Political Development 13 (1): 1 30. Wirls, Dani el. 2007. "The 'Golden Age' Senate and Floor Debate in the Antebellum Congress." Legislative Studies Quarterly 32 (2): 193 222. Wood, Gordon. 1998. The Creation of the American Republic 1776 1787 New York, NY: W.W. Norton and Co. Yoest, Patrick, and Tim Starks. "Lawmakers Clear Sept. 11 Compromise." CQ Weekly Online (July 30, 2007): 2284 85. http://library.cqpress.com.lp.hscl.ufl.edu/cqweekly/weeklyreport110 0000 02561346 ; Accessed 3/5/2010. Zelizer, Julian. 2006. On Capitol Hill: The Struggle to Reform Congress and its Consequences, 1948 2000 Cambridge, MA: Cambridge University Press.
315 BIOGRAPHICAL SKETCH Jordan Ragusa received his Ph.D. from the Univ ersity of Florida in the summer of 2011 His research and teaching interests are in the fields of American politics and institutions, political methodology, empirical theory, and political behavior with specific substantive interests in bicameralism, the p olicy process, legislative procedure, and political parties. In the fall of 2011 he joined the Department of Political Science at the College of Charleston at the rank of assistant professor