A THEGRY 0F common-FORMAmNiiN _v.‘f‘??;If;}‘i LEGISLATURE: BARGAININGJN THE 5U. SSENATE .5 Thesis for the Degree Of Ph..Dt. . M!CHIGANSTATEUNWERSITY 1 " MARTIN DONALD LEVINE ' 1972 ’ ' IL." V v ' VD ) u LAA :‘ 1‘: ‘ o '- r r A’ ‘ .>,a.ri.‘u ' . ~. a -.. 1" f“! 'v M Zamrmmwm r "K. .12.:- i This is to certify that the thesis entitled A THEORY OF COALITION FORMATION IN IEGISLATURES: BARGAINING IN THE U. S. SENATE presented by Martin Donald Levine has been accepted towards fulfillment of the requirements for Ph.D. degreein Political Science fidififlpfi? _ Major professor Date ‘ i“ / ‘751’ 0-7639 ram; A ABSTRACT A THEORY OF COALITION FORMATION IN LEGISLATURES: BARGAINING IN THE U. S. SENATE BY Martin Donald Levine This work entails the develOpment and testing of an original theory of the legislative process. The theory of coalition formation in legislatures conceptualizes the process whereby bills are con- sidered and acted upon to be one of non-coercive bargaining. From this model we derive a series of testable hypotheses describing (a) the eXpected sizes of winning coalitions for different types of motions, (b) the voting behavior of Specific legislators on specific motions, (c) expected abstention rates among legislators under varying circumstances, and (d) the effect of agenda position on legislative outcomes. These hypotheses are then tested employing data from the U.S. Senate, 89th Congress; the findings are reported in detail; the theory is reformulated in light of the findings; and suggestions are made for further research. Because our theory describes the basis of legislative voting, we borrow from existing roll call studies. Because the model in- cludes a wide range of factors in explaining why individuals vote as they do, we rely especially heavily upon simulations and multi-variate Martin Donald Levine analyses of congressional behavior. And because our theory assumes the rational calculation of advantage on the part of each legislator in turn, it resembles other existing coalition theories. While our theory resembles many existing models, it is identical to none and is broader than most in terms of flu; sc0pe of phenomena explained. The theory of coalition formation in legislatures begins with an assumption that all representatives are rational maximizers of their respective probabilities of re-election. In deciding how to (and whether to) vote on any given motion each legislator is con- cerned only with maximizing those resources--constituent support, organizational assistance, money-~necessary for re-election. We further assume that every member of the legislature is aware of this and is also aware that a calculus of eXpected resources exists whereby each legislator may determine the amount of resources he will receive for voting "yea," voting "nay," or abstaining. For most legislators on any given roll call their resource payoffs are contingent only upon their voting a particular position. However, for any roll call there also exists at least one legislator who will receive additional resources if the motion passes or fails. These persons act as brokers, bargaining for passage/defeat of the motion through offers of resource side-payments to available members. Our model further assumes that a decision rule is operative within the legislature limiting the offers of side-payment bargains to those representa- tives who expect to receive less than a specified amount of resources from non-bargaining sources. The basis of this kind of a rule is the notion that in the long run brokers conserve resources by agreeing to Martin Donald Levine forego offers to "the other side" and thereby also avoiding having to make offers to one's own side. The effect of this decision rule is thus to limit bargaining to that sub-set of all motions on which neither broker has enough committed members in his correSponding proto-coalition to be certain of victory. Put another way, we are able to conclude from our theory that two distinct and identifiable types of roll calls exist--bargaining and non-bargaining motions-~and that on only the former will exchanges for votes occur. From this central conclusion we go on to derive predictions describing absolute and relative winning coalition sizes, abstention rates, and voting choices for different classes of legislators (typed according to the net quantity of resources which each eXpects to re- ceive from extra-legislative sources on any given motion) and different types of roll calls (typed according to whether or not bargaining is assumed to occur). The test of these hypotheses suggests that our model of the legislative process may, indeed, be accurate. On the single measure of predicting individual voting choices, of the senators for whom our model predicts such choices the predictions are correct 77.#% of the time across a set of 155 roll calls (over 31% of all re- corded votes in the 89th Senate). Limiting the sample of senatorial motions examined by issue area, procedural type and party positions taken, the predictive power of most hypotheses is improved still fur- ther, lending additional support to the conclusion that our unique theory of the legislative process is an accurate one and allowing us to more specifically identify the conditions under which our model is most applicable. Martin Donald Levine The work concludes with a discussion of the manner in which the model might be employed to consider other bodies, other points in time, and other questions of legislative choice. A THEORY OF COALITION FORMATION IN LEGISLATURES: BARGAINING IN THE U. S. SENATE BY Martin Donald Levine A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Political Science 1972 ACKNOWLEDGMENTS The author would like to acknowledge the assistance of several persons who have contributed their thoughts to this work. I first want to thank my committee, Professors David Meltz, Cleo Cherryholmes, and Paul Conn, who worked well together and well with me--a much too rare and very happy situation. Particular thanks goes to the chairman of the committee, David Meltz, who has been an intellectual mentor in the best sense of the term, contributing his ideas, accepting mine, and through it all allowing me to grow. Professors Tim Hennessey and Rufus Browning of the Department of Political Science at Michigan State University have both been in- volved in this project at different stages. I wish to express a special thanks to Lon Felker, who worked with me on a short paper that served as the genesis of the current work, and Mel Katz for his many useful comments on earlier drafts. ii TABLE OF CONTENTS LIST OF TABLES O C O O O O O C O O O O O O O O O O O O O O O 0 LIST OF CHAPTER CHAPTER CHAPTER CHAPTER CHAPTER FIGIIRES O O O O O O O O O O O O O O O O O O O O O O O I: TOWARDS A NEW THEORY OF THE LEGISLATIVE PROCESS . The Problem and Previous Formulations . . . . . . . . Coalition Theory: An Alternative Model . . . . . . . II: THE THEORY OF COALITION FORMATION IN LEGISLATURES Symbols and Definitions . . . . . . . . . . . . . . . Axioms . . . . . . . . . . . . . . . . . . . . . . . . Theorems . . . . . . . . . . . . . . . . . . . . . . Hypotheses . . . . . . . . . . . . . . . . . . . . . III: VALIDATION . . . . . . . . . . . . . . . . . . Setting: The U. S. Senate . . . . . . . . . . . . . . Setting: The 89th Congress, 1965-1966 . . . . . . . . The Roll Calls . . . . . . . . . . . . . . . . . . . . The Resource Calculus . . . . . . . . . . . . . . . . IV 0 FINDINGS O O O O O O O O O O O O O O O O O O O O V : COIJCLUS IOIJS O O O O O O O O O O O O O O O O O O 0 Revision of the Theory in Light of the Findings . . . Suggestions for Further Research . . . . . . . . . . . iii Page vii 17 35 35 37 48 58 64 65 69 72 79 109 150 150 155 iv TABLE OF CONTENTS (Cont'd) APPENDIX A: DATA SOURCE O O O O O O O O O O O O 0 APPENDIX B: MBABERS OF U. 8. SENATE, 89TH CONGRESS APPENDIX C: SAMPLE OF ROLL CALLS . . . . . . . . . BIBL IOGRAPHY O C O O O O O O O O O O O O O O O O O O Page 164 165 174 182 Table 1. 2. 3. 4. 5. 6. 7. 9. 10. 11. 12. 13. 1A. LIST OF TABLES Civil Rights Motions Included in Sample . . . . . . . . Urban Welfare Motions Included in Sample . . . . . . . . Labor Motions Included in Sample . . . . . . . . . . . . Foreign Aid Motions Included in Sample . . . . . . . . . Number of Roll Calls Sampled by Issue Area and Year . . Civil Rights Policy Preference and Party Activity by Rac¢,196l+....................... Social Welfare Policy Preference and Party Activity by Size of Residence/Income, 196A . . . . . . . . . . . . . Social Welfare Policy Preference and Party Activity by Income,1%l+oooooooooooooooooooooo Mean Difference Between Observed Winning Coalition Size and Minimum-Certainty-of-Winning Coalition Size by TypeOfRollCall................... Rate of Valence Voting by Class of Legislator and Type Of R011 call 0 O O O O O O O O O O O O O O C O O O O O O Abstention Rate by Class of Legislator and Type of R011 Call 0 O O O O O O O C O O C O O O C C O C O O O 0 Mean Difference Between Observed Winning Coalition Size and Minimum-Certaintyhof-Winning Coalition Size by Type Of R011 Call and Agenda POSition o o o o o o o o a Mean Difference Between Observed Winning Coalition Size and Minimum-Certainty—of—Winning Cealition Size by Type of Roll Call and Issue Area . . . . . . . . . . . . . . Rate of Valence Voting by Class of Legislator, Type of Roll Call and Issue Area . . . . . . . . . . . . . . Page 75 77 82 83 97 101 102 112 115 119 124 129 133 vi LIST or TABLES (Cont'd) Table 15. Abstention Rate by Class of Legislator, Type of Roll Call and Issue Area . . . . . . . . . . . . . . . . . . . 16. Mean Difference Between Observed Winning Coalition Size and Minimum—Certainty-of-Winning Coalition Size by Type of Roll Call and Information Level . . . . . . . . . . . . 17. Rate of Valence Voting by Class of Legislator, Type of Roll Call and Information Level . . . . . . . . . . . . 18. Abstention Rate by Class of Legislator, Type of Roll Call and Information Level . . . . . . . . . . . . . . . . 19. Mean Difference Between Observed Winning Coalition Size and Minimum-Certainty-of-Winning Coalition Size for Bargaining Roll Calls by Democratic Party Position . . . . . . . . Page 135 139 140 1A2 145 LIST OF FIGURES Figure Page 1. Frequency Distribution of Winning Coalition Sizes for Bargaining and Non-Bargaining Roll Calls Excluding ClOtureVOteSo00.000.000.0000.0.0000.112 2. Frequency Distributions of Winning Coalition Sizes for Bargaining and Non-Bargining Roll Calls Excluding Cloture Votes: Civil Rights and Labor Motions Only . . . . . 131 3. Frequency Distribution of Winning Coalition Sizes for Bargaining and Non-Bargaining Roll Calls Excluding Clo- ture Votes: Urban Welfare and Foreign Aid Motions Only . . . 132 4. Frequency Distributions of Winning Coalition Sizes for Bargaining Roll Calls, Excluding Cloture Votes, by DemocraticPartyPOSition...........o...... 145 vii CHAPTER I TOWARDS A NEW THEORY OF THE LEGISLATIVE PROCESS If much work has been done in the area of legislative behavior, little has been done describing the legislative rocess, and still less work has addressed itself to the development of theories of the legislative process. In this work we present and test such a theory. While our primary interest is in predicting and explaining the size of winning legislative coalitions, other derivatives of our theory describe the individual voting behavior of legislators; and the axioms from which these theorems are deduced collectively comprise a description/abstraction of the process by which collective legislative decisions are assumed to be made. This process is considered to be one of coalition formation through non-coercive bargaining. Thus, our theory borrows heavily from two minimally overlapping bodies of liter- ature: legislative behavior and the theory of games, particularly the dynamics of coalition formation. In this introductory chapter we ex- amine some exemplary studies in both areas and note the basis of our own theory in these works. The second chapter presents the theory itself, both argumentatively and as a logical system of assumptions and derivations. The third chapter includes a discussion of the set- ting in which these conclusions are tested and the manner in which necessary variables and key concepts are operationalized. Chapter IV presents the empirical findings and discusses their implications for the theory. In the final chapter we offer a revision of the theory in 2 light of the data and make suggestions for further research in the form of possible applications of the theory to alternative settings and alternative questions of legislative behavior beyond those of coalition size, voting, and abstentions dealt with here. The Problem and Previous Formulations In looking at previous theoretical formulations and empirical research in the area of legislative behavior our attention is limited. The following discussion ignores many significant legislative studies dealing with other matters (committee action, changing rules of the legislative game, etc.) and focuses exclusively on those works con- cerned with legislative roll call voting. And at that the discussion is by no means exhaustive of the field, nor is it intended to be. In examining this sub-field of legislative behavior we look at several different approaches to the problem of predicting and explain- ing individual votes and evaluate the relative strengths and weaknesses of each such approach with respect to: (a) the degree to which each offers a plausible, logically consistent explanation for legislative voting and (b) the contribution of each to the growing body of empirical findings in this area. Throughout we carefully note the insights bor~ rowed from these partial explanations and research findings in develop- ing our own theory of the legislative process. The most prevalent approach to the study of roll call voting may be loosely labeled the "cross-pressure" theory of the legislative process. This term is meant to encompass the broad range of formula- tions which predict a legislator's voting behavior on the basis of his party affiliation and his constituency characteristics. Common to most such theories is the assumption that in deciding how to vote on any 3 given motion a legislator is under possibly contradictory "pressures" from the leaders of his party on the one hand and his constituents on the other; the position the legislator eventually takes is the result of an aggregation of these pressures. As Lewis Froman, Jr. notes in characterizing recent legislative research in the federal arena: "Stu- dies of roll-call voting in Congress have tended to stress two factors: (1) the large amount of party cohesion in Congress on most issues, and (2) the importance of constituency factors in explaining deviations from party votes within parties."1 In evaluating the utility of cross-pressure theories we must recognize that formulations of this sort are quite useful in predicting roll call outcomes. A legislator's voting behavior is indeed highly correlated with his party's position and with certain characteristics of his constituency (which, in turn, are meant to reflect policy pref- erences of the persons he represents). However, if these partial theories are to provide explanations for this relationship, one must specify the manner in which these two factors--a legislator's party and characteristics of his constituency--affect his voting behavior. And it is a major criticism of these studies that they are often un- clear in specifying even the nature of the pressures which they posit to exist. At one extreme there is no attempt to explain why a congress- man should support positions favored by his party leaders and/or the personshe represents; rather these hypotheses are simply tested and we are offered such "explanations" as: Political pressure [is] a force which brings about distinctive voting behavior on the part of Congressmen, whether the force is applied through conventional types of coercion or through the appeals of loyalties and ideologies. and H It is assumed . . . that some logical relationship ordinar- ily exists between the positions of Congressmen and the dem- ographic characteristics of their districts, whether the con- nection derived from the pressures of constituents few or many, from Congressman's perceptions of their constituents' needs and demands, or from the workings of the Congressional nomination process. Furthermore, where attempts are made to specify what it is about party and constituency pressures which makes these salient in a legislator's determination of how to vote, the alternative explanations are mu- tually inconsistent. In concluding that a legislator will support his party, these theories generally make either or both of two underlying assumptions. In some cases it is imagined that the legislator uses his party for voting "cues" so as to avoid the costs he would otherwise incur by actually determining the content of a bill and evaluating its merit. The party is employed for this purpose, so the argument goes, because the representative assumes that leaders of his party share his policy preferences. John Jackson articulates this position in a study of roll call voting in the U.S. Senate. [The:]large volume of complex business, specialization, and multitudinous demands upon a senator's time shape the way he perceives the problem posed by the floor vote and the way a decision is made. The precise stimulus may be the ringing of the bells or the announcement several days in advance that a particular bill is going to be considered . . . . The routine or program evoked is a fairly simple search for in- formation about the content, implications, or even the "best" vote on the proposed amendment. These routines save the individual senator a lot of work and time, which he can then devote to his area of specialization, constituents, speeches, etc. These sources of information and cues on how to vote be- come the variables hypothesized to influence how a senator votes. It is expected that a senator's vote will correspond to the positions of his sources of information. One may speculate that, even given the "multitudinous demands" upon a legislator's time, the cost of finding out at least something about 5 a bill so small and the electoral consequences of voting "wrong" so grave that seeking cues from one's party leaders--or anyone else in the legislature for that matter--would prove a false economy. The most common alternative hypothesis is that one votes with one's party under a credible threat of some sanction. The sanction may range from social ostracism to exclusion from posts of influence within the party to withdrawal of support for re-election. Typical of the formulations which make this assumption is David Mayhew's 23331 Loyalty AmongCongressmen.5 However, Mayhew never identifies the particular sanction(s) which are employed. This failure to specify the kinds of inducements and threats legislative party leaders employ to achieve cohesion is a significant weakness of most such partial theories which hypothesize the threat of sanctions as the major basis of party pressure. For it is obvious that any assumption specifying the particular sanctions that are actually used--symbolic, material, electoral, or otherwise--will itself reflect significantly differing assumptions concerning the sorts of things for which the legislators have the greatest utility. The lack of such an assumption can only result in an incomplete theory. Turning next to explanations of relatinnships between con- stituency characteristics and roll call voting, we find considerably less disagreement as to what it is about the nature of a representa- tive's constituents which leads him to vote in certain ways on certain issues. Constituencyapressure theorists generally assume that: every legislator seeks re-election and that any representative's constituents will decide whether or not to return him to the legislature largely on the basis of the degree to which his voting record conforms to their 6 preferences. These researchers then go on to assume that particular policy preferences may be attributed to various demographically distinct groups in the p0pulation. From these assumptions it is pos- sible to derive the conclusion noted above that constituency charac- teristics will be systematically related to legislative voting pat- terns. Froman states this position in a "single complicated sentence": Since [legislative] districts . . . vary among each other on a number of characteristics (that is, are heterogeneous as a whole), and if we assume that (1) on matters of national public policy many constituents are interested in the outcomes and indicate their interest by contacting their representa- tives, and (2) congressmen wish to remain in office and have certain perceptions of their districts which lead them to act in such a fachion as to avoic damaging their chances of being re-elected, doing this by behaving in ways congruent with con- stituency pressures and their perceptions of real and potential interests within their constituencies, then we would eXpect variations in congressional behavior to be related to dif- ferences in constituencies. The "if" clauses in this statement represent the assumptions or axioms of the theory, and the then clause represents the conclu- sion derived from these assumptions. Other legislative voting research examines the relationship between a representative's region and his behavior on roll calls. The findings of such studies are fairly consistent: the most significant correlation between region and voting occurs within the vounds of each legislative party and is issue specific. The following from Daivd Truman's The Congressional Party is typical. There was among the Republicans no single and sharp line of cleavage such as that wich divided the northern from the southern wing of the House Democrats, but rather a series' of noncongruent tendencies within the party, reflecting shifting and fluid bases of agreement and, presumably, of association. There is much less agreement as to the eXplanations for such relationships. Here the independent variable--region--is not even an 7 indirect measure of a legislator's constituent interests, and it is, therefore, by no means clear that a representative "votes his region" for the same reason he "votes his district." If the following is a dissatisfying answer, it is also a typical one. It is not clear what the intervening process is that con- verts regional affiliation into roll-call vote. It is clear, however, that regional interests constitute an aduience that provides cues to congressmen at the roll—call stage.8 Of this sort of self-avowedly partial explanation one may ask: what value are these "cues" to the legislator? In following the policy preferences of his constituents a legislator seeks to maximize the possibility of his re-election, but how does voting with his region enhance his electoral chances? The link between a representative's voting in accord with his regional interests--as opposed to his con- stituent interests--and that representative's re-eletion is certainly not obvious, and rarely is the attempt made to specify what the con- ceptual link might be. As an explanation for regional voting the argument cited seems woefully inadequate. Other researchers argue that region serves the same cueing fun- tion as party, that is, as a cost-reducing mechanism allowing the legislator to avoid the expenditure of resources which would be neces- sary to evaluate a motion on its content/merits. The same criticism we noted of the party-as-cue theories may also be leveled against these region-as-cue explanations. Again, it would seem to be a false economy for a representative to accept the Opinion of a group of his colleagues on the merits of a motion over his own evaluation of its effect on his re-electoral chances. This argument is especially telling for the explanation of sectional voting noted, because here it is being assumed that the legislator will look to a group (his regional associates) which 8 has no direct influence over his future career chances (as his legisla- tive party leaders might) and on a subset of bills presumably highly relevant to him. It is especially on these motions that we would expect a representative to make his own judgment, and it is especially such an inconsequential group as his geographical neighbors that we would expect him to ignore in determining his position. What seems to be a more plausible explanation for the relation— ship found to exist between region and voting within each party is that it is not causal but spurious--that the relationship simply reflects the fact that people from both the same section of the country and the same party are likely to represent similar constituencies and thus be under nearly identical party and constituency pressures, and that it is this identity of pressures that "causes" an identity of voting behavior. Thus far among the cross-pressure theories of legislative be- havior we may observe a general underarticulation' of the reasons for the relationships found to exist betweeneilegislator's party affilia- tion, the characteristics of his constituency, and his region as independent variables and his voting as dependent variable. Where ex- planations are offered, they are mutually inconsistent to a degree which is clearly unacceptable for an integrated formal theory of the legislative process. Among the explanations for adherence to a party *The term "articulation" is used in the sense defined by Richard Rudner as follows: "To the extent that a theory has been fully articu- lated in some formulation, it will achieve an explicit deductive devel- opment and interrelatinnship of the statements it encompasses."9 Thus, a fully articulated theory is one in which the entire set of assumptions necessary to deduce the conclusions is made explicit and in which the entire set of theorems which can be deduced from the axioms is given. As noted, one of the major criticisms of existing explanations of legis- lative behavior is that they are severely under-articulated, and one of the major values assumed for the original theory presented in the follow- ing chapter is that it is much more nearly fully articulated. 9 position there is widespread disagreement, and the most satisfying of these explanations is not entirely in consonance with the generally agreed-upon explanation for voting one's constituency. The incon- sistenies become even more marked iwth the inclusion of region as a third contributing factor in a representative's voting decision. To avoid this confusion in our own theory we begin with a single assumption describing the motivation of any legislator in deter- mining whether to support or Oppose any motion. We present this as— sumption in the next chapter as the first axiom. of the theory of coalition formation in legislatures. It may be paraphrased as follows: in determining what position to take on any given motion a legislator will be concerned 22$! with maximizing his chances for re-election. This statement in consistent with the underlying assumption employed by Froman and others to explain the relationship between con— stituency characteristics and voting behavior. If we further assume that a legislator's party may offer electoral support in direct propor- tion to the extent to which the legislator votes the party line, then the axiom given above can also serve as the beginning of an explanation for the high incidence of party voting. Turning next to studies of legislative lobbying, we will see that this assumption of a representa- tive's desire to achieve re-election can also explain the generally observed relationship between interest group positions and legislative voting. Whereas cross-pressure theorists are in general accord at least as to the effects of party, constituency, and region, studies of the effect of interest groups on legislative voting evidence little agreement even as to what the effect is and still less agreement as to its causes. One * All references are to the axiom-set as presented in Chapter II. 1O reason for this disagreement over findings is the considerable variance in methodology among such studies. Few attempt roll call analyses to determine the independent effect of interest group positions on voting outcomes. Rather, most works rely primarily on surveys of lobbyists and legislators. What agreement there is among studies of this kind seems to indicate that the impact of interest group preferences in the legislator's voting calculus is marginal, if there is any impact at all. However, these formulations differ considerably in the assump- tions each offers to explain this finding. In an examination of the U.S. Senate Matthews begins with the assumption that "legislators do not like to be 'pressured'"10 and might react adversely to demands from lobbyists. From this Matthews concludes that senators perceive interest groups primarily as sources of specialized information and that their greatest impact on roll call outcomes consists of spurring those senators already committed to a position into active support of-- rather than simply voting for--specific legislation.11 In a study of the effects of lobbying in the U.S. House of Representatives, Scott and Huntlz reach a similar conclusion that interest groups are of little consequence in determining congressmen's eventual votes. Their explan- ation for this phenomenon is based on the pressure model of legislative voting discussed above and the assumption that interest groups have little with which to press: From the point of view of the interest group, the problem is that it rarely has anything to give or withhold that is of more than marginal significance to the congressman. Its bargaining position is inherently weak since the instruments available to it are not the stuff out of which 'pressure' can be fashioned.1 11 Lester Milbrath in The Washington Lobbyists11+ employs a similar kind of explanation for the apparent impotency of interest groups, though Milbrath never articulates the assumption made explicit by Scott and Hunt. Milbrath's empirical findings, however, differ from those of the other two works. Through a systematic sampling of legislators and the application of a well-formulated survey instrument-~a combination of data-gathering techniques employed in neither of the other works-- Milbrath finds that, "on broad political issues commanding considerable public attention, the major determinant [of positions taken by con- gressmen] is the desire of the public. Lobbyists can do very little to affect the outcome, [however if] the legislation is specialized and affects only a small segment of the population, lobbyists are more likely to play a larger role."15 This is clearly a more specific con- clusion than the little-orbno-effect finding of the other works. A comparison of these three Opposing theories finds no one of them to be obviously superior. In concluding that under almost no cir- cumstances will lobbyists alter the preferences of legislators because interest groups simply "preach to the converted," Matthews appears to contradict a finding common to both the other studies. And because Matthews' is by far the methodologically weakest of the three, there is reason to believe that his conclusion is the least accurate. However, his partial explanation of interest group influence is better articu- lated than that of either of the other two. The explanation offered by Scott and Hunt, on the other hand, caibe criticized for being dis- tinctly underarticulated. Their methodology is also deficient in that the survey instrument they employ appears to be biased toweard minim- izing the reported importance of interest group pressures in a congress- man's overall preference calculus.16 Of the three studies outlined 12 above, Milbrath's is the most methodologically rigorous and his find- ings, indicating the specific circumstances under which interest group pressures are salient, are probably the most accurate. However, the explanation he offers is the least articulated. In constructing our own theory we assume that Milbrath's find- ing that interest groups affect a legislator's votes only where his own constituency is unconcerned is the most accurate description of reality. Our explanation for this relationship is based on the earlier assumption of a universal desire for re-election among all legislators and on a secondary assumption, similar to the one employed by Scott and Hunt, that what lobbyists have to offer a legislator only margin- ally affects his electoral chances. That is, where constituents offer votes directly, interest groups offer primarily monetary payoffs. And money can only be indirectly transformed into votes and, at that, is important only where the candidate lacks adequate financial re- sources.* Another approach to the problem of predicting and explaining roll call outdomes conceptualizes the legislature primarily as a social grouping. These studies assume that the legislator takes his voting cues largely from other members of the group (e.g. other legislators with whom he is personally associated). What empirical evidence there I"If one assumes, as we do, that a legislator's party, consti- tuency, and interest group ties are all salient elements in his voting calculus, then one must also Specify the relative weights which he assigns to each of these three factors. We address ourselves to this question of relative weights in Chapter III when we present our own calculus of "pressures." There we will discover a variety of findings and a conceptual richness in legislative research which is not evident in this initial cursory discussion. 13 is seems to indicate that some relationship does exist between the preferences of persons with whom a representative talks and the legis- 17 lator's voting behavior. In a study cited earlier John Jackson finds that a small amount of variance among voting patterns of junior U.S. senators can be explained by the positions taken by each one's senior colleague, where both are members of the same party. Cherryhomes and Shapiro18 incorporate a similar notion of interpersonal persuasion into a computer simulation of roll call voting in the House of Representa- tives which successfully predicts a large proportion of individual votes. Inclusion of a "persuasion phase" does increase somewhat the predictive power of their model over that which is obtained by looking only at party, constituency, and regional factors alone. Though the methodology and findings of these two works differ considerably, they share a common explanation for the phenomenon they describe. Both assume that it is possible for a legislator to be "talked into" taking a particular position on a particular motion. Thus, both studies also implicitly assume that there are circumstances under which a legislator will accept an acquaintance's Opinion of the merits of a motion rather than his own evaluation of the relative re- wards to be expected in voting either position. To accept this ex— planation for the occurence of voting patterns paralleling association- al patterns would violate our primary axiom of the legislator as ration- al calculator of electoral advantage. Therefore, in our formulation we assume that a legislator may "change his mind" only when offered con- crete payoffs affecting his subsequent elecoral chances. Still other studies of roll call voting assume the psychological and attitudinal makeup of each legislator to be the prime determinant 14 and direct cause of his voting behavior. To the extent that these form- ulations further assume this makeup to be unique for each individual, such constructions seem ill-suited as the basis for a general theory of the legislative process. It should also be noted that the measurement of relevant psychological and attitudinal variables poses a problem for these studies which does not arise in the case of most of the pre- viously examined works. One means which has been used to measure the effect of attitudes on voting is to determine the extent to which a legislator's past votes can predict his future behavior independent of other factors. Cherryholmes and Shapiro19 do this indirectly by in— cluding "memory" as one of the independent variables in their simula— tion of voting outcomes, and find that the inclusion of this memory of past votes over and above the other factors incorporated in the model has little effect on the predictive power of the model in most in— stances. This finding alone serves as partial justification for not considering psychological/attitudinal factors in our abstraction of the legislative process. Furthermore, the assumption that legislators have attitudinal sets which determine their voting behavior is clearly at odds with the basic rationality assumption of our theory. For both these reasons psychological and attitudinal factors do not appear in our theory. Finally, we may deal briefly with the role theorists, those researchers who focus on a legislator's self—perception of his job and its relationship to his voting behavior. Seminal among such works is Wahlke, Eulau, Buchanan, and Ferguson's study of four state legisla— tures.2O Unfortunately, the typology they develop for classifying 15 legislators is not as conceptually precise as might be desired, and thus far no subsequent studies have been carried out to determine the exact behavioral correlates of each legislator-type. There is also some reason to believe that there is less variance in role-perception among members of Congress--where we test our theory-~than among state legislators. One attempt to improve upon this early typology is Joseph Schlesinger's Ambition in Politics.21 Schlesinger classifies all office-holders according to their political ambitions: (a) discrete (desiring only a limited number of terms in the current post before retiring politically), (b) static (desiring continual re-election to the current office), and (c) progressive (desiring eventually to secure election to a higher office). Schlesinger does discuss some of the be- havior patterns which we might expect to find associated with each of these types, but makes no attempt to empirically validate his hypothp eses.22 Our own theory assumes that all legislators have static ambi— tions or behave as if they do.* Therefore, self-perception of role appears as a constant rather than an independent variable affecting the voting behavior of representatives. Overall, our theory conceptualizes the legislative process to be one of coalition formation, and in the following section we examine coalition theory in some detail. But before moving on to that section it would be well to look carefully at one final work in legislative behavior; for this work combines many of the features Of our own theory ‘In Chapter V we discuss the manner in which the theory of coal— ition formation in legislatures might be altered to explain voting out- comes in legislative bodies with large numbers of discrete and/or pro- gressive members. 16 in a single formulation, and though the process it hypothesizes to ex- plain roll call voting differs significantly from our own, the outcomes predicted are quite similar. This study, referred to at several points 23 above, is Cherryholmes' and Shapiro's Representatives and Roll Calls. As noted, rather than an additional strictly theoretical formulation, this book is the report of a computer simulation of roll call voting in the U.S. House of Representatives, 88th Congress. However, the simulation does employ several of the factors dealt with both in the predecing theoretical/empirical studies and in our own model of the legislative process. First, in simulating the legislative process a congressmen's party, constituency characteristics, and region, among other factors, are considered in calculating the representative's "predisposition" on any vote. This predisposition is a measure of the tendency of the con- gressman to support either position on a motion, though how and why the several independent factors contribute to this tendency is left unclear. Our own formulation begins with the axiom that all legislators desire to maximize resources necessary for re-election and the secondary as- sumption that party, constituents, and interest groups can all offer these resources in some form or another. From these assumptions we are able to calculate a value (in resources) similar to Cherryholmes' and Shapiro's predisposition. The computer simulation in Representatives and Roll Calls is then programmed so that all legislators below a certain level of pre- disposition enter a "persuasion phase" in which they come into contact with other members of Congress, and through this process it is possible 17 for the original congressmen to be talked out of their initial positions. Again there is a parallel between Cherryholmes' and Shapiro's simula— tion and our theory. We assume (axiom 10) that there is some amount of resources which a legislator may expect to receive from party, con- stituents, and interest groups for taking a given position on any given motion beyond which no other legislator will attempt to bargain with him for his vote. But if a legislator may not expect to receive this amount of resources (in the pre-bargaining stage), then he mgy be sub- ject to bargains in the form of offers of side-payments. Here our model differs importantly from Cherryholmes' and Shapiro's: in their simulation those legislators who are not predetermined on a motion will always enter the persuasion phase, whereas from our theory we conclude that bargains will be Offered only when neither side on a motion has enough votes from among those legislators who are pre-determined to be certain of victory on the roll call. However, there are also signifi- cant parallels between our axiom-set and Cherryholmes' and Shapiro's program; and it is therefore encouraging (in terms of the probability of being able to empirically validate our own theoretical conclusions) to note that the predictions generated by their simulation do conform well to the actual roll call data. For, though they are by no means conceptually the same, the way in which their persuasion phase is programmed makes it very similar to our bargaining process in terms of the empirical consequences 2f each. Coalition Theory: An Alternative Model As indicated, in our own study of the legislative process we take an approach quite unlike the classic" works examined above. 18 Rather than conceiving of the legislative process as a series of indi— vidual voting decisions on the part of each legislator in turn and the roll call outcomes as the simple aggregate of all such individual choices, our theory views the process of passing and defeating motions as involving not only individual decisions of relative advantage by the legislators but also the aggregation of individual interests within the limitations that certain legislators will receive payoffs only if a given motion succeeds/fails and that the concurrence of a specified number of legislators (generally N E 1 of all members) is a necessary and sufficient condition for being certain of winning on the roll call. Our theory of the legislative process is one of coalition formation, and the model upon which our theory is based is the theory of games. In this section we first discuss the major components of game theory, noting the commonality between the axioms of our own theory of the legislative process and the assumptions of the theory of games. We then compare our formulation to a now classic theory of coalition formation which is similarly concerned with coalition size. The chapter closes with an examiniation of other applications of coalition theory to the legislative process. As a model for predicting the resolution of conflict situations game theory is highly parsimonious in terms of the number of assump- tions from which conclusions may be derived describing a broad range of 24 phenomena. It is in part this parsimony which compels us to examine carefully the few game theoretical axioms which are so valuable.25 We will see how these axioms or analogs thereof are incorporated into our own theory of legislative bargaining. The first such axiom is rationality. In their introduction to 19 to the field Luce and Raiffa note the unfortunate ambiguity of this central concept: Though it is not apparent from some writings, the term "ra- tionalfl is far from precise, and it certainly means different things in the different theories that have been developed [employing this concept]. Loosely, it seems to include any assumption one makes about the players maximizing someting, and any about complete knowledge on the part of the player in a very complex situation, where experience indicates that 36hu- man being would be far more restricted in his perceptions. As additional game theoretical formulations have been developed, ra- tionality has taken on a more limited and more nearly universally agreed-upon meaning. Disregarding the assumption of complete or per- fect information (to be dealt with below as a separate axiom of game theory), rationality has come to be defined as the choice by an in- dividual, when faced with a set of alternative actions, of that action which will result in the greatest utility payoff for him as determined by the individual independent of the act of making the choice. This is similar to Luce and Raiffa's definition of rationality: "Of two alternatives which give rise to outcomes, a player will choose the one which yields the more preferred outcome, or, more precisely, in terms of the utility function he will attempt to maximize expected utility."27 This axiom of utility maximization logically implies that the player has utility for something. And if one is to predict choice, one must specify that for which each individual has utility. That is, the utility function(s) of the relevant actor(s) involved must be given. This is a necessary step in the use of any "rational choice" model to both explain and predict outcomes of complex situations, and it neces- sarily involves a significant abstraction of reality to specify out Of the universe of possible things which any person may covet those objects, 20 both tangible and intangible, whose desired acquisition will direct the person's actions.‘ Our own theory of the legislative process assumes that each legislator is concerned with maximizing the probability of his being re-elected (axiom 1). We further assume that a legislator's desire to maximize resources necessary for re-election is longbterm (axioms l and 8). Finally, it is posited that the utility functions of all legislators for these resources are everywhere increasing (axiom 2a) and that when engaged in bargaining all legislators 'have extremely low utility for risk (axiom 13). In game theoretic terms, these axioms taken together constitute an assumption of rationality and a specifica- tion of the utility functions of the players involved. Also included in most game theoretical formulations is some assumption describing the knowledge or information which the players possess. Luce and Raiffa's generalized knowledge assumption states that, "each player is fully cognizant of the game in extensive form, i.e., he is fully aware of the rules of the game and the utility func- tions of each of the players."28 In our own theory we leave unstated the assumption that every legislator is perfectly knowledgeable as to the formal rules of the legislative game. However, we do explicitly state that each legislator is fully aware of the utility functions of his colleagues (axioms 2b and A). .All theorists using a rational-calculus approach have had to deal with the question of specifying that for which the political actor has utilityb-that which he is attempting to maximize. One of the earli- est such formal theories considers the electoral process and hypothesa izes that voters seek to maximize the similarity between their own policy preferences and the actions of the government while candidates seek to maximize votes.29 Other theories Of voting consider the ra- tional candidate to be a maximizer not of votes but of his chances of re-election.3O A recent empirical test of the voter-rationality 21 In his discussion of n-person, zero-sum games William Riker con- siders information as a potential variable in terms of its completeness and perfection. "Systematically complete information" is said to exist where every participant in an n-person game knows precisely how much the addition of each player to a coalition alters the value Of the coal- ition, i.e., how much each person is worth. Similarly, "systematically perfect information" exists where all participants know what moves have already been made, i.e., who has already joined each coalition, what 34 bargains have been struck. In our conceptualization of legislative bargaining, we assume systematically complete information without ever making the assumption explicit. This seems a reasonable assumption to make when we reCOgnize that systematically complete information in a legislature only requires that every member be aware of the equal weighting of votes on a roll call. Our theory also assumes systematic- ally perfect information on the part of the sub-set of legislators actively engaged in bargaining (axiom 12). Furthermore, we assume that the level of information is constant across roll calls.* assumptiOn of such theories extends to the components of the voter's presumed utility calculus to include such factors as candidate-person- ality characteristics, adherence to interpersonal cues from trusted as- sociates, and the coincience of the candidate's party affiliation with that of the voter.31 In a radically different setting, a theory of in- terest group behavior assumes that what eadimember of such a group is concerned witg is themaximization of some collective good at a minimum cost to him.3 A theory of bureaucratic behavior, on the other hand, has posited that man_qu bureaucrat will be interested in maximizing some combination of his career security, his chances for advancement, and the coincidence between the policy outputs Of his bureau and his own preferred position. Whatever the setting, the problem is the same, and the answer always involves a significant abstraction of reality. 1"The effect of relaxing the perfect information assumption is examined in Chapter IV where the theory as first presented and several reformulations thereof are tested. 22 The number of players involved in any game must also be speci- fied if we are to predict outcomes. The distinction generally made in the literature is among (a) one-person games against nature or decision- making situations, (b) two-person cooperative or non-OOOperative games, and (c) n-person games where some coalition of players is necessary to win and where the utility payoff associated with each outcome for each player is contingent upon his being a member of the winning coali- tion. Our abstraction of the legislative process includes elements of each sort of game. We first assume that on any given roll call most of the legislators are concerned only with the re-electoral resource pay- off associated with casting their "yea," "nay," or "abstain" and that these payoffs are not contingent upon being a member of the eventual winning coalition (axiom 1). Thus, these legislators are faced only with a complex decision-making situation. On the other hand, we as- sume that for each motion at least one legislator exists part of whose utility payoff is contingent upon winning on the roll call (axiom 7). We further assume that where more than one such actively concerned legislator or broker exists on the same side of a motion, all brokers on the same side of the motion will behave as a unit (corollary 8—9). SO, in this sense, there are never more than two brokers on any given motion. And while these brokers are bargaining agents of sorts, they are not fiduciaries for other legislators. Therefore, even on bargain- ing roll calls (a distinct sub—set of all motions) the situation re- duces to a two-person game between the brokers and a series of one- person games played by the other legislators the collective outcome of which will determine which broker wins. We may also note that the two-person game between the brokers 23 is zero-sum in that there are only two possible outcomes--passage or defeat of the motion--and the preference orderings of the two brokers across these two outcomes are precisely opposed. Finally, games are characterized as involving either a single play or iterations; that is, as being played either in the short- or long-term. Our theory of the legislative process explicitly assumes that any legislator, whether acting as broker or as individual decision- maker, seeks to maximize expected re—electOral resources over his entire term in Office (axioms 1, 3, and 8). Put a different way, the legisla- tive game is played over an uncertain number of iterations. The ex- tensive game which each legislator plays thus involves all motions for the entire period of time during which he expects to be a member of the legislature. In the preceding paragraphs we have pointed out the similarities between our theory of the legislative process and the axioms of game theory. However, there are important differences as well. For one thing, as noted, according to our conceptualization of the legislative process on any given roll call most of the members will be faced with only a simple decision-making situation rather than a competitive game. And whereas we- do assume that under specified circumstances the legis- lative process takes on the form of a complex two-person game involving coalition building, under no circumstances is the process that of an n-person zero-sum game. However, because of the similarities in axioms already noted and the similarities in conclusions which will be evident shortly, it is imperative that we carefully distinguish between our conceptualization of legislative voting and the now-classic theory of n-person zero-sum games developed by William Riker.35 While both the 24 theory of political coalitions (Riker's) and the theory of coalition formation in legislatures (our own) predict minimum winning coalitions under specified circumstances, the assumptions employed in the two models, the conditions under which minimum winning coalitions are ex- pected to occur, and the definition of a minimum winning coalition differ significantly. Whereas Riker deals with n-person, zero-sum games, our theory assumes that only for a single broker on either side of an issue will utility payoffs be contingent upon victory on the roll call. From his theory Riker derives the conclusion that: In n-person, zero-sum games, where side-payments are per- mitted, where players are rational, and where they have perfect information, only minimum winning coalitions occur. 36 The prediction is that all victories will be achieved by minimum winning coalitions. The analogous derivative from our own theory is as follows: For any motion where neither proto-coalition is large enough to be certain of victory independent of the behavior of other legislators, the eventual winning coalition will be of minimum size necessary to be certain of victory and no larger. Our theory thus posits the occurrence of limited-size winning coalitions gply on a Specific sub-set of all motions. Also, Riker's minimum win- ning coalition is by no means identical to our "coalition of minimum Size necessary to be certain of victory." In the context Of a legisla- ture, where each member is of equal weight and abstentions are allowed, a minimum winning coalition consists of just one more member than the defeated Opposition. In the same context our minimum-certainty-of- winning coalition consists of N + 1 2 legislature.* Our theory and Riker's thus predict the occurrence of of the entire membership of the ’Both these examples, of a minimum winning coalition and of a minim-certainty-of-winning coalition, apply only for those motions requiring a simple majority of members present and voting for passage 25 different sorts of limited-size coalitions, under different circum- stances, and derivable from different sets of axioms.* None of this is to imply that we see no value in Riker's theory; it is clearly among the most significant works produced in recent years. The number and variety of political phenomena to which the theory of political coalitions has been applied and for which it appears to offer a plausible explanation is large and growing. Riker himself suggests applications of his theory to such diverse situations as the "corrupt bargain" of 1825 whereby John Quincy Adams was elected President by the 38 House of Representatives, 39 to shifting electoral coalitions in American politics, to voting in the Indian Congress following independence, and to the creation and dissolution of international alliances.“ Others employ Riker's model to explain the creation of coalition governments in West Germany,l+2 France,43 Norway,44 and Japan}+5 Within the American context The Theory of Political Coalitions has been used in studies of electoral coalitions, judicial behavior on the Supreme Courtfi6 and co- 1+8 alition disintegration leading to the Civil War. The number of such studies alone is testimony to the value of or defeat. Where a different decision rule exists, such as the two- thirds majority required on certain procedural motions, each limited- size coalition obviously takes on different values. In the empirical tests of our theory both simple majority and two-thids majority roll calls in the U.S. Senate are considered. 1“Our own notion of a minimim-certainty-of—winnning coalition resembles Riker's concept of a subjectively minimum winning coalition which arises under conditions of uncertainty. The concept is presented in the size principle: "In social situations similar to n-person, zero-sum games with side-payments, participants create coalitions just as large as they believe will insure winning and no larger."3 26 Riker's work. Yet, the collective finding of these studies seems to be that Riker's theory is useful in predicting coalition behavior in some, but certainly not all, institutional contexts; and researchers are beginning to recognize the necessity of developing alternative ex- planatory models. And what is true of the work cited above is equally true of studies of legislative coalitions. Riker‘s theory obviously cannot offer a satisfactory explanation in this setting; winning legislative coalitions clearly are not all of minimum size but, in fact, range up to unanimity. Indeed, the inapplicability of Riker's model in this context first suggested to this writer the possibility of developing an alternative theory of the legislative process which could better predict and explain coalition sizes. A number of other researchers have also developed alternative theories of legislative coalitions, and before presenting our own theory in Chapter II, we will examine several of these other formulations. Wayne L. Francis was among the first to recognize the need to alter Riker's theory to make it applicable to the legislative setting.50 He sees the reason for the inapplicability of Riker's formulation as lying in the fact that the condition of perfect information is not satisfied in most legislatures. Francis contends that the imperfection of information in legislatures results in over-sized and under-sized coalitions being formed by causing defections among potential coalition members and uncertainty among coalition leaders as to the minimim size necessary to win. In the body of his work he offers a series of hypoth- eses describing the relationships between several variable factors of the legislative game and two key dependent variables: the prOpensity 27 of coalition members to defect and the pr0portion of all legislators sought by coalition leaders in bargaining. These hypotheses consider not only the perceived level of information but such additional inde— pendent and intervening variables as the size of the legislature and 51 And while the the percentage of seats held by the majority party. author claims "to interrelate these variables in a formalistic way," it is probably truer that his "analysis should be considered explora- 52 tory, and its conclusion, tentative." However, as an early attempt to adapt coalition theory to the study of American legislatures, this work is highly suggestive of at least some of the factors which must be included in any genuinely formal theory of legislative coalitions. In work done concurrently with our own, Richard Murray and Donald Lutz apply Riker's minimum—size hypothesis to a study of re- districting decisions in state legislatures.53 Intended to fill what the authors perceive as a lack of direct tests of Riker!s theory, the work begins as a straightforward empirical evaluation of the size principle. The authors limit themselves to consideration of votes on redistricting motions in American state legislatures, examining 3k8 such roll calls. The sample is intentionally limited in such a way as to most nearly approximate the zero-sum and perfect information dondi- tions of Riker's model. EVen so, Murray and Lutz find that less than one-third of the roll calls are decided by margins approximating minimal size. In light of this observation, the authors then introduce a series of controls to determine whether or not the tendency towards minimum winning coalitions is any more pronounced for some sub-sets of roll calls than for others. Considering such variables as the extent of the redistricting plan, whether it is aimed at state legislative or con- 28 gressional districts, and the size of the chamber in which the vote took place (the latter factor suggested by Francis' study), the authors find little improvement in the ability of Rikerfis theory to explain the observed outcomes. Some improvement is noted when the partisan division in the legislature is taken into account, with minimum winning coali- tions significantly more likely to occur in highly competitive bodies. Taking this cue, the authors conclude that minimum winning coalitions are less likely to be formed in legislatures characterized by "dominant faction structuring" (where some fairly permanent faction, usually the majority party, can generally control'the behavior of a majority of the members) than in decisional bodies evidencing "sub-dominant factional structuring" (where factions exist but no one of them controls a major- 1w)?!+ Our own theory of the legislative process allows one to deduce a similar conclusion, while defining dominance of factions (proto- coalitions) not only across legislatures but across motions as well according to the characteristics of each member and each motion in turn. Our theory thereby explains variation in coalition size across types of motions within the same body left largely unaccounted for by Murray and Lutz. While our two formulations differ importantly, both concern themselves with specifying the conditions under which minimem winning legislative coalitions are most likely to occur. And we cer- tainly cannot disagree with their conclusion that "more attention should be given to identifying types or classes of empirical situations where Riker's theory can be profitably applied."55 Another researcher, also working contemporaneously, who shares this conviction is David Koehler. His approach is similar to our own in that he also attempts to develop an alternative theory of legislative 29 56 bargaining to eXplain variance in the size of winning coalitions. Our abstractions of the legislative process differ, but our theor- etical conclusions are similar-~that minimum winning coalitions will be formed only for a limited sub-set of all legislative motions and that for the others we may expect to find larger-than-minimum coali- tions. Testing his propositions with data from the U.S. Congress for a 30-year period, Koehler finds them generally supported. This find- ing is eSpecially encouraging for us not only because of the similar- ity of theoretical conclusions but because of the similarity of set- ting and methodology as well.* Before presenting our own theory one final study of legisla- tive coalitions should be noted: Meltz's model of majority party bargaining.57 While Meltz does not deal explicitly with the size of the winning legislative coalitions, heAdm .m Aosmv lame Amsoev Areas sz Odom Hmampoh n.6m e.m smoaaoaaom m.mm s.m on soaeaooaao oaom Hemopmh as.mm as.em capstooson am.ms am.mm so eaoaasm oawoz uoz ommOz oamOz poz oawoz comm some .o»o> comm oaom mooauomam hamswsm henna vsoshoamsm hash Hmaopoh no sospwmom .0 some .ooom as sameness apnea sac ooaoaoeoaa season enemas Haaao m magma 98 in order for him to be coded as receiving re-electoral resources from this group for taking a pro-civil rights stand. Here we follow the practice of simulations of congressional behavior, establishing a threshold level for a particular constituency characteristic such that for any district where the threshold is met or exceeded the presumed effect of the constituency characteristic is included in the prediction of how the representative will vote on the apprOpriate roll calls. However, we deviate from the methodology of others of calculating the parameter for each district in turn and setting the threshold level for any one district at what seems to be the natural break point in the distribution.29 Instead, we calculate the parameter for the multi- state pOpulation and use that figure as our cut-off point.* For all non-southern states this statistic is 6.8% Negroes in the pOpulation in 1960. Thus, where Negroes represent 6.8% or more of a non-southern state's total pOpulation, we code blacks as a significant constituent group for the senators from that state on civil rights motions. Demo- cratic senators from such states receive two units of resources for supporting an expanded federal role in guaranteeing civil rights; Re- publican senators in similar circumstances receive one unit of pre- bargaining resources; and where Negroes represented less than 6.8% of a non-southern state's pOpulation in 1960, neither senator from the state receives any resources from constituents on civil rights motions. Corroborative evidence for the presumed relationship between . We recognize that the national population mean as the threshold of significance for a constituency characteristic is an arbitrary cut-off point. However, such a decision rule is no more arbitrary than the alternative of employing a "natural break point" where the means of arriving at the break point are not Specified. 99 the percentage of Negroes in a state's pOpulation and support by the senators of that state for civil rights legislation can be found in an analysis by Charles Andrain of roll call voting in the U.S. Senate in 1960.30 We follow a similar procedure as that outline above in desig- nating interested constituent groups for urban welfare motions. Here we assume that urban poor (Operationally defined as persons living in cities of 100,000 or more pOpulation and earning an annual family in- come of less than 33,000 in 1960) represent a concerned group favoring federal aid to cities and that high-income persons (310,000+ annual income, 1960) constitute a group Opposing such aid. The party affilia- tion of the former group is taken to Democratic, and the latter group Republican. Applying our pre-bargaining resource calculus for urban welfare roll calls, each Democratic senator is coded as receiving two units of resources from urban poor for supporting expansion of federal assistance to urban areas where the urban poor constitute a significant prOportion of the state's pOpulation; where high-income persons repre- sent a Significant prOportion of the state's pOpulation, a Democratic senator is coded as receiving one unit of resources for Opposing such legislation. Any Republican senator receives one unit of resources from the urban poor group for voting its preferred position on urban welfare roll calls and two units of pre-bargaining resources from high-income persons for supporting their position on such motions, again with the stipulation that corresponding constituent group comprises a Significant prOportion of the senator's district/state. No senator of either party is coded as receiving any pre-bargaining resources from either of these groups where the group does not constitute a significant interest in the 1OO senator's electorate. As before, the threshold of significance for each of these groups is set at the corresponding prOportion of the national pOpulation represented by each one--4.8% of a state's pOpu- lation for the urban poor and 15.1% for high-income persons. Also as before, the assumptions made here concerning the policy and party preferences of these two groups are supported by data from a survey of potential voters taken at the time of the 1964 general elec- tions. Tables 7 and 8 show the relationships between Size-of-residence and income on the one hand and preferences for federal social welfare legislation and political party affiliation on the other hand. The findings as reported clearly support our contentions.t Corroborative evidence also exists for our assertions regarding the kinds of consti- tuent characteristics which are associated with a legislator's support- ing urban welfare legislation.31 Turning next to roll calls of concern to organized labor, our concern shifts from constituent groups to private interest groups. The labor motions in our sample of roll calls were chosen precisely because they were the ones identified by the AFL3CIO as being their primary legislative objectives for the 89th Congress. The AFL-CIO is, * Weakness in the association between urban poor and Democratic party affiliation may be partially accounted for by the necessity of defining this group for the purposes of these tests as persons earning less than $3,000 annual income and living in cities of 50,000 or more pOpulation, rather than the limit of 100,000+ cities specified in the calculus. Furthermore, data restrictions forced us to use responses to the question of whether or not the federal government should be responsible for personal living standards as the indicator of policy preference on the question of urban welfare legislation. While these Operational indicators do not correSpond perfectly to the definitions we have used in our calculus, the results in Table 7 and Table 8 generally support our assumptions of policy and party preferences among these groups. 101 .hpSpm cowuooam soar .amusoo nonmemom ho>asm .m Ammsv Ammv Ameoev Asmv sz oaom Hmampoh 0.:m :.rm smowansmmm w.mm m.mn Op cospwmommo oaom amaopoh R m.mm & m.wm casewoosom R :.O: & u.mm mo pacemam nompo .ocH OOOMm zoaom nonpo .ocH 000mm 3oaom $368.8 89:80.8 osoocH Hmscc< use some .ouo> meoocH Hmdcm< cam wsw>wq mo can oesopamom mo ousm mamSflam hpamm monopfimom mo onwm usacom meannessamz cw maom Hmaopmh so soapflmom m some .osoocH a oosmmfimom mo beam as sameness speed can ooaoaoeoam season cheese; asaoom a oases 102 .hpsum sowpooam awmr .aowcoo noamomom ho>adm .m Ammov $03 8me 3:3 9: eHom Hmampoh m.em m.me escaaosaom m.mm a.es or soaeamoaao Odom amnepoh R mnwm R stem owumaoosmcm R N6: R m.mm mo snowmen Romeo .eooo.orm honpo .vooo.orm msoocH ammr .mpo> osoouH wuwbwq no use amends amassed henna assess usssem measuresamz ea oaom asaosom so mowuwmom mewmr .osoocH he mps>wpo< huamm use cocoaomoam howaom mammaoz Hmwoom w oases 103 therefore, taken to be a concerned interest group with reference to the organized labor motions; the preferred position of this group on each roll call is taken to be the one announced by the group itself. Fur- thermore, we assume that for these roll calls no concerned constituent groups exist. Finally, we assume the AFL~CIO to be affiliated predomi- nantly with the Democratic party in terms of likely candidate support. The threshold of significance of AFL-CIO presence within any state is set at the prOportion of the national work force made up of AFL-CIO members. In 1960 this figure was 21.5%.32 For foreign aid roll calls we assume that neither constituent nor interest groups are generally concerned enough with the outcomes to act as sources of pre-bargaining resources for senators. Thus, only the position (if any) taken by a legislator's party will enter into his calculations of the pr value of any foreign aid motion for him. This assertion that members of Congress are relatively free of constituent/ interest group pressures on these kinds of bills appears elsewhere in the legislative voting literature as well.33 This concludes our presentation of the pre-bargaining resource calculus first as a series of general rules and then as applied to the U.S. Senate. If the presentation has been of considerable length, it must be remembered that for this study the resource calculus is also of considerable importance. As we have noted, in testing our theory in the following chapter we are testing the validity of this calculus as well. And while questions may remain concerning some of the assump- tions underlying the calculus as operationally defined, we have attempt- ed here to make explicit and justify our indicators as fully as possible. 104 Before discussing the use of the calculus in classifying legi- slators and roll calls antecedent to testing our theory, we may summar- ize some of the unique characteristics of our measurements of legisla- tive voting pressures. First, we include organized interest groups in the calculus of extra-legislative re-electoral support. As indicated earlier, this element is often ignored in roll call analyses. Second, the amount of pre-bargaining resources which any legislator may expect to receive from a concerned constituent/interest group varies with the identity between the party affiliation of the group and that of the legislator. Third, in dealing with the overlapping six-year terms of U.S. senators we incorporate into the calculus an assumption that sup- porting the goals of any significant constituent/interest group will be more important as the congressman approaches re-election. Fourth, the amount of re-electoral resources which one's party may offer varies with the probability of the senator's being renominated.’ Finally, we employ a unique indicator of minority party position relying on the recorded votes of the party's leader and whip. Some of these char- acteristics of our calculus are peculiar to the setting being examined; others are generalizable to any legislature. In any event, it is hOped that the overall set of indicators as presented here and the theory as presented in Chapter II will allow us to more precisely predict voting and non-voting choices in the U.S. Senate than has been possible to date. We can now classify each senator on each roll call according to the pre-bargaining resource value of the motion for him. This, in turn, will allow us to determine whether any given roll call is a 105 bargaining or non-bargaining one. To do this we first calculate the amount of resources which each member of the 89th Senate can expect to receive from each relevant source for supporting its preferred position on each roll call in our sample. We then subtract the total amount of pre-bargaining resources which a senator receives for voting "nay" from the amount of resources he would receive for voting "yea" and classify each senator on each roll call as follows: Where the net resource- value of a motion for a senator is +2 or greater, the senator is deemed a member of the supporting proto-coalition on that motion; where his net resource-value is +1, he is classed as a pre-bargaining supporter; for a resource-value of zero (0), he is considered indifferent; for a net resource-value of -l, he is classed as a pre-bargaining Opposer; and all senators with resource—values of -2 or less are considered members of the Opposing proto-coalition on the roll call. All motions on which neither of the proto-coalitions consists of 51 or more mem- N + 1 2 calls, consiStent with our definition of the concept. Conversely, all bers ( of all senators) are then designated as bargaining roll motions on which one of the proto-coalitions has 51 or more members * are classified as non-bargaining roll calls. . On cloture motions (requiring a two-thirds majority for vic- tory) the minimum-dertainty-of-winning proto-coalition Sizes are 67 for the supporting side and 34 for the Opposing side. FOOTNOTES: CHAPTER III 1E. W. Kelley, "Utility Theory and Political Coalitions: Problems of Operationalization," in Sven Groennings, E. W. Kelley, and Michael Leiserson, eds., The Study of Coalition Behavior (New York: Holt, Rinehart and Winston, Inc., 1970), p. 480. 2Hayward R. Alker, Jr., "Computer Simulations, Conceptual Frameworks, and Coalition Behavior," in Groennings SE al, The Study of Coalition Behavior, p. 377. 3 Groennings 23.22, The Study of Coalition Behavior. One notable exception is: David B. Meltz, "Competition and Cohesion: A Model of Majority Party Legislative Bargaining" (un- published Ph.D. dissertation, University of Rochester, 1970). 5Joseph A. Schlesinger, Ambition and Politics (Chicago: Rand McNally & Co., 1966). 6Lewis A. Froman, Jr., "Differences Between the House and Senate," in Raymond E. Wolfinger, ed., Readinge on Congress (Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1971), P. 63. 7ibid., p. 62. The use of the closed rule is discussed with reference to revenue bills in John F. Manley, "The House Committee on Ways and Means: Conflict Management in a Congressional Committee," in Wolfinger, ed., Readings on Congress. 9For an examination of this kind of bargain see the discussion of "concerned" legislators in Meltz, "Competition and Cohesion." 10Donald R. Matthews, U.S. Senators and Their World (New York: Vintage Books, 1960), pp. 92-101. 11For roll call studies of the U.S. Senate see, among others: John E. Jackson, "Statistical Models of Senate Roll Call Voting," American Political Science Review, LXV (June, 1971); and Jeanne L. Martin, "Exchange Theory and Legislative Behavior: A Computer Simu- lation of Roll-Call Voting in the U.S. Senate" (unpublished Ph.D. dissertation, Michigan State University, 1971). 12Randall B. Ripley, Power in the Senate (New York: St. Martin's Press, 1969), p. 14. 106 -t.. 107 13Rowland Evans and Robert Novak, Lyndon B. Johnson: The Exercise of Power (Cleveland: The World Publishing Company, 1966), pp. 95-117. 1I+John G. Stewart, "Two Strategies of Leadership: Johnson and Mansfield,? in Nelson W. Polsby, ed., Congressional Behavior (New York: Random House, 1971), p. 61. 15See, among others: Jackson, "Statistical Models of Senate Roll Call Voting." 16Among these studies are: Cleo Cherryholmes and Michael Shapiro, Representatives and Roll Calls (Indianapolis: The Bobbs- Merrill Company, Inc., 1969); Lewis A. Froman, Jr., Congressmen and Their Constituencies (Chicago: Rand McNally & Co., 1963); and Martin, "Exchange Theory and Legislative Behavior." 17John Bibby and Roger Davidson, On Capitol Hill (New York: Holt, Rinehart and Winston, Inc., 1967), p. 144. 18William H. Hunt, Wilder W. Crane, and John C. Wahlke, "In- terviewing Political Elites in Cross-Cultural Comparative Research," in Samuel C. Patterson, ed., American Legislative Bheavior: A Reader (Princeton, New Jersey: D. Van Nostrand UOmpany, Inc., 1968), p. 421. 19Congressional Quarterly ,Almanac (Washington, D.C.: Con- gressional Quarterly Service, 1967), XXII, 1965, p. 6; and 1966, p. 6. 20Eugene Eidenberg and Roy D. Morey, An Act of Congress (New York: W. Norton & Company, Inc., 1969), pp. 59-69. 21CongressionalrQuarterlyrAlmanac, 1966, pp. 1302, 1347. 22See, among others: Cherryholmes and Shapiro, Representa- tives and Roll Calls; and Martin,"Exchange Theory and Legislative Behavior." 23Jackson, "Statistical Models of Roll Call Voting." See, among others, the measure used in Raymond E. Wolfinger and Joan H. Hollinger, "Safe Seats, Seniority, and Power in Congress," in Wolfinger, ed., Readings on Congress, pp. 40-41. 25See: Anthony Downs, An Economic Theory of Democracy (New York: Harper and Rown 1957) for an imaginative explanation of the appearance of these kinds of party affiliations. 26See: Donald R. Matthews and James Prothro, Negroes and the New Southern Politics (New York: Harcourt, Brace, and World, Inc., 1966). 108 27V. 0. Key, Southern Politics (New York: Alfred KnOpf, 1949), among others. 281964 Election Study, Survey Research Center (Ann, Arbor, Michigan). 29Cherryholmes and Shapiro, Representatives and Roll Calls, footnote on p. 49. 30Charles F. Andrain, "Senators' Attitudes Toward Civil Rights," in Patterson, American Legislative Behavior. 31.8ee, in this regard, an analysis of the constituency char— acteristics of the congressmen who supported the federal food stamp program in 1964: Randall B. Ripley, "Legislative Bargaining and the Food Stamp Act, 1964," in Frederic N. Cleaveland and Associates, Congress and Urban Problams (Washington: The Brookings Institution, 1 969) 9 P- 305- 32Source: U.S., Bureau of the Census, Statistical Abstract of the United States 1963, 84th edition. 33The assumption is implicit in the calculus of predisposi— tion developed in Cherryholmes and Shapiro, Representatives and Roll Calls, Chapter 7. It is also supported by the findings of James A. Robinson, Congress and Foreign PolicyeMaking (Homewood, Illinois: The Dorsey Press, 19677. CHAPTER IV FINDINGS In this chapter we test the hypotheses derived in Chapter II within the setting and employing the Operational indicators specified in Chapter III. We report the results of these tests first for the entire sample of roll calls and legislators, then for sub-samples of roll calls and legislators to determine whether or not our theory of the legislative process is a more accurate description of reality under certain more limited circumstances. Throughout we discuss the implica- tions of our findings for the theory of coalition formation in legislatures. As we did in first presenting our theoretical prOpositions, we begin here by dividing the full set of testable hypotheses into four discrete sub-sets each dealing with a different phenomenon. Within each hypothesis-set the empirical tests employed are similar; across sets the test vary considerably. The first hypothesis-set predicts ab- solute and relative sizes of winning coalitions for bargaining and non- bargaining roll calls. Here we wish to demonstrate that two distinct types of roll calls do indeed exist and that on only one type of roll call will minimum winning coalitions occur. The initial deterministic preposition is as follows: H The eventual winning coalition for all bargaining roll calls is just large enough to be certain of victory and no larger. The second proposition is in the form of a relativistic hypothesis: 109 110 H2 The mean difference between the size of the eventual winning coalition and the minimum-certainty-of-winning size is greater for non-bargaining roll calls than for bargaining roll calls. To test these hypotheses we deveIOp an index of deviation between any given winning coalition size and the associated minimum- certainty-of-winning size. This index of deviation is calculated by subtracting the minimum-certainty-of-winning coalition size from the observed winning coalition size, dividing the absolute value of that figure by the difference between the total number of possible votes and the minimum—certainty-of-winning cOalition size, and multiplying that fraction by one hundred.1 Expressed as a general formula the index of deviation for any given roll call is as follows: (observed winning coalition Size (total possible ) number of votes ) - (minimum-certainty-of-winning size) X 100 - (minimum-certainty-of-winning size) This index may take on values from zero through one hundred, inclu- sive, corresponding to winning coalitions ranging from exactly minimum- certainty-of—winning (zero) to a winnig coalition of the whole (one hundred). In our application of the measure to voting in the U.S. Senate the "total possible number of votes" is always one hundred. The "minimum-certainty-of-winning size" is 51 for most roll calls. How- ever, for procedural votes requiring a two-thirds majority of those present and voting (e.g. cloture, veto-override, etc) the minimum- certainty-of-winning size is 67 for the supporting proto-coalition and 34 Votes for the Opposing proto-coalition. We may now apply the index of deviation to our sample of roll calls from the 89th Senate in order to test the initial pair of 111 hypotheses. In so doing we calculate (a) the mean index of deviation for all 116 motions (including six cloture votes) classed as bargaining bills, and (b) the mean index of deviation for all thirty-nine roll 2 calls which our theory predicted to be of the non-bargaining type. The results of these calculations appear in Table 9 and Figure 1.* The data in the first table are generally supportive of our theory. They disconfirm the deterministic prediction of H1, as would be expected, but, at the same time, the data strongly support the rela- tivistic prediction of H While not all bargaining motions are 2. passed or defeated by an exactly minimum-certainty-of-winning coalition, there igla significant difference in the expected direction between the average deviation from minimum size for bargaining roll calls and the same statistic for non-bargaining roll calls. Indeed, the average de- viation for non-bargaining motions is nearly twice that of bargaining roll calls (as reported in Table 1). Thus, we may tentatively con- clude that our central prOposition that there are two distinct types of motions in legislative bodies is correct. Furthermore, the probability of minimum-certainty-of-winning coalitions occuring is, as we predicted, greater for one type of roll call than for the other. Figure 1, comparing the distributions of winning coalition sizes for bargaining and non-bargaining roll calls (excluding cloture votes), illustrates these phenomena dramatically. From this figure we I In this and in all subsequent tests we include recorded posi- tive and‘negative pairs as the equivilants of "yea" and "nay" votes respectively. This is based on the assumption that such pairs are known to all brokers before a roll call is taken and that they are thereby equivilant to other vote commitments in the bargaining process. Other research on congressional coalitions has similarly counted re- corded pairs as the equivilants of votes.2 112 Table 9 Mean Difference Between Observed Winning Coalition Size and Minimum-Certainty-of—Winning Coalition Size by Type of Roll Call Bargaining a Non-Bargaining Roll Calls (n) Roll Calls (N) Mean Index of 16.7 % (116) 32.5 % (39) Deviationb a. Number of roll calls. b. Figure in each cell is the mean index of deviation for all roll calls of the type specified. See text for definition of index of deviation. 40- - - % of --—-' Bargaining Roll Calls Roll Calls ----- Non-Bargaining Roll Calls 30~ 20- 1o- " «1‘ l L V I l T $40 41-47 48-54 55-61 62-68 69-75 76-82 83-89 90;: di- _ qu- —n— Size of Winning Coalition Figure 1 Frequency Distribution of Winning Coalition Sizes for Bargaining and Non-Bargaining Roll Calls Ex- cluding Cloture Votes 113 see that the modal size of winning coalitions for non-bargaining bills is considerably greater than the correSponding figure for bargaining roll calls. The modal winning coalition size for bargaining bills is within three votes of 51, as our theory suggests, while the modal win- ning coalition size for non-bargaining motions is in the 69-75 member range. Furthermore, fully 38.2% of the winning coalitions for the bar- gaining roll calls in our sample are within three votes of 51, and 73.7% of the coalitions fall within the 41-61 vote range. For non- bargaining roll calls these figures are 7.7% and 33.4% respectively. While these figures are certainly supportive of the second hypothesis, * 1. tions from minimum-certainty-of-winning coalitions among bargaining again, they fail to confirm H Possible explanations for the devia- roll calls are presented and empirically examined later in the chapter. But before looking at such explanations, we turn next to the second series of theoretical derivatives. This second hypothesis-set, H3 - H7, describes the absolute and relative rates of valence voting among different classes of legislators on bargaining and non-bargaining motions. The third and fourth prOpo- sitions state that all proto-coalition members on any roll call and all pre-bargaining supporters/Opposers on non-bargaining roll calls who vote will vote for the position for which they can expect to receive * The disconfirmation of this and the other deterministic hypo- theses is in no way unexpected. Explanations for 100% of the variance in complex social/political phenomena simply do not now exist. For the remainder of Chapter IV we shall acknowledge this fact by treating all deterministic statements as intermediate prOpositions in the derivation of the probabalistic hypotheses. Evidence relating to these determin- istic statements will be noted only in passing, and conclusions as to the validity of our theory will be based on the results of tests of the relativistic hypotheses only. 114 the greatest resource payoff from pre-bargaining sources. That is: H3 All proto-coalition members who vote vote their valences. H4 All pre-bargaining supporters and Opposers who vote on non- bargaining roll calls vote their valences. As outlined in Chapter II, these prOpositions follow from the conclu- sions that no proto—coalition member and no pre-bargaining supporter or Opposer on a non-bargaining motion will be offered any side-payments to shift his vote. A further conclusion from our theory (theorem IV) states that pre-bargaining supporters/Opposers on bargaining roll calls “El vote for the position Opposite to that for which they can expect to receive the greatest amount Of pre-bargaining resources. This theorem taken together with the pair of statements just presented allows us to derive the following testable hypotheses: H5 The prOportion of voting pre-bargaining supporters and Opposers who vote their valences is greater on non-bargaining roll calls than on bargaining roll calls. H6 Across all bargaining roll calls, the prOportion of voting protO-coalition members who vote their valences is greater than the prOportion Of voting pre-bargaining supporters and Opposers who vote their valences. H7 Across all roll calls, the prOportion of voting protO-coalition members who vote their valences is greater than the prOportion of voting pre-bargaining supporters and Opposers who vote their valences. TO test these hypotheses we calculate the prOportions Of proto- coalition members and pre-bargaining supporters/Opposers voting their preferred pre-bargaining positions on both bargaining and non-bargain- ing roll calls as percentages Of the total numbers Of each class of senator voting on each class of motion. Table 10 gives the results of these calculations for the entire sample of roll calls. As with the preceding hypothesis-set, the deterministic statements (H and H4) are 3 115 .msoam on» as myopmamwmoa Has an ammo mmpo> Hmavfi>flcsw mo Mogadz .n .ucpo> on: macaw can as who» Imamflmma mo amass: dance on» mo swapsoonmm s ms mnOflpwmom mswswmwamnnoum sophomomm pmos Macaw mom wouo> on} mucusamflmca no adonw wsflwsommmhaoo on» Has mo soaphomoam can ww HHOO some a“ Onsmflh .m lam-v Aomomv ammmmv A-amav alas mmaapo> & m.nw R m.mm &.m.ow R m.wu mOsOHs> mo mumm mflamo Haom maaco HHOm maamo HHOm maaso HHOm msfisflmmammusoz mswswwmamm mstHMMAmmlsoz wqwsflmmamm mammom90\mpmphommsm whonaoz maaaammamm1aam aoapaaaoouoaopm Hams aaom no mass was aoawamamaq co mamas an maaao> moaoam> no «ham 0- magma 116 disconfirmed while the testable probabilistic generalizations (H , H6’ and H?) are supported by the data. The Observed rates of valence voting for protO-coalition mem- bers across all roll calls and for pre-bargaining supporters/Opposers on non-bargaining roll calls are found to be 79.0% and 67.9%. While neither of these findings are equal to the 100% rate predicted, neither of the findings are unexpected, and both figures represent fairly high levels of correctly predicted voting choices. Of greater interest are the results of the tests of the fifth through seventh hypotheses. PrOposition H hypothesizes that the 5 prOportion Of pre-bargaining supporters/Opposers voting their valences will be greater on non-bargaining roll calls than on bargaining mo- tions. Our findings support this predictioi, with the respective rates of valence voting for the two groups being 67.9% and 53.9%. The sixth hypothesis, predicting a greater rate of valence voting among proto- coalition members than among pre-bargaining supporters and Opposers on bargaining roll calls, is also supported by the data (78.3%gzg. 53.9%), as is the seventh prOposition predicting a greater rate of valence voting among all protO-coalition members than among all pre-bargaining supporters/Opposers (79-0% Ego 58.9%). These results are generally supportive Of our theory. Further- more, additional manipulations Of the data presented in Table 10 may be performed to suggest that what error does exist lies not in the theory but in the indicators employed to test it. Specifically, if our theory is valid, in addition to the results described in hypotheses H5 - H7 we would also eXpect to find a greater difference between the rates Of valence voting for pre-bargaining supporters/Opposers on 117 non-bargaining roll calls and the same class of legislators on bargain- ing roll calls than between the rates of valence voting for pre- bargaining supporters/Opposers on non-bargaining motions and proto- coalition members on the same roll calls. Put another way, we would expect the voting behavior Of the pre-bargaining supporters/Opposers on non-bargaining roll calls to more closely resemble that of proto- coalition members on the same roll calls than that of other pre- bargaining supporters/Opposers on bargaining roll calls. This additional hypothesis is weakly supported by the data in Table 10 where we find the difference between the figure in column three and the corresponding figure in column to be 14.0% while the difference between the figures in columns two and four is only 32.6%. Another secondary hypothesis can be Offered to the effect that there will be a greater difference between the voting behavior Of pre-bargaining supporters/ Opposers on the two different types of roll calls than there will be between proto-coalition members on the two kinds of motions. Our data support this conclusion as well with the two differences in valence voting rates being 14.0% and 2.2% reSpectively. Still another such hypothesis which we may derive from our theory predicts a greater dif- ference between the rates of valence voting Of protO-coalition members and pre-bargaining supporters/Opposers on bargaining motions than be- tween these same twO groups of senators on non-bargaining roll calls. Again, the results reported in Table 10 conform to this expectation; we find the difference between column one and column three, 24.4%, greater than the difference between columns two and four, 12.6%1 These additional findings Offer further evidence for believing that our the- ory is an accurate abstraction of the legislative process. 118 The third set Of hypotheses deals with abstention rates. The first prOposition in this set is a deterministice restatement Of theorem VIII: H8 All indifferents on non-bargaining roll calls abstain. The ninth hypothesis describes the relative abstention rates we eXpect to Observe among the various types Of legislators on the various types Of roll calls. This generalization follows from theorems VIII and V wherein we posit that: all_indifferents on non-bargaining roll calls will abstain for strategic reasons-~reasons other than unavoidable detention; pre-bargaining supporters/Opposers and indifferents on bar- gaining roll calls may abstain for strategic reasons; and pre-bargain- ing supporters/Opposers on non-bargaining roll calls as well as proto- coalition members on any roll call will 22335 abstain as a short- or long-term resource maximizing strategy. The testable theoretical derivative is: H9 The abstention rate among indifferents on non-bargaining roll calls is greater than the abstention rate among pre-bargalning supporters/Opposers and indifferents on bargaining roll calls, which is, in turn, greater than the abstention rate among pre- bargaining supporters/Opposers on non-bargaining roll calls and protO-coalition members on all roll calls. The testing of these prOpositions is a fairly simple matter of calculating and comparing the abstention rates for each class of legislator on each type Of roll call. The results Of this analysis appear in Table 11. As is readily observable from this table, neither Of these statements is supported by the data. Indeed, the group Of senators hypothesized in H8 to always abstain evidences the lowest ab- stention rate--1.1% non-voting for indifferents on non-bargaining motions. Also, pre-bargaining supporters/Opposers and indifferents on 119 .Aopwa soapsoumnm wswpmasOHmo now ommnv mono» Hmwpscpom HO Honszz .n .msoam on» ha ammo soon o>mn canoe pmnp mouo> mo amass: annoy on» no mwwpnmoncm m as msflpo> no: vmawmm acnpwmn macaw wsfivsommmhaoc on» s« mMOpmeflmOH mo sowvaomoum on» ma HHOO some as madman .m a $3 383 Sam: :mmmv $33 33.5 :3 movmm a -.- a m.m- a o.m a -.o- a o.m a 0.: 835me mHHco Haom maamo Haom mHHOo Haom maamo Haom mHHmo Haom maamo HHOm .mnmmusoz wswsfimwhmm .mnsmlsoz wswswmmnmm .mnmmnnoz wdwstMHmm masonommwvsH mamm09m0\maoumommsm unease: waaaammaamamam aoaaaamoouoaonm Hamo HHOm MO cake was hovmamwwmq mo mmmHo hp 09mm nowusmvmn< we OHQME 120 bargaining roll calls, predicted to have the next highest abstention rate, are found to abstain at the highest rate, or 11.5%. Finally, we Observe that the group of legislators posited to vote most frequent- ly did, in fact, abstain at the rate Of only 10.1%, corresponding tO the weighted average of columns one, two, and four in Table 11. These results make it clear that at least one of the axioms from which these hypotheses are deduced must be false. Because Of the wide variation between the prediction of H8 and the corresponding Observation, our attention is drawn to theorem VIII in searching for 3 __x _ WET an eXplanation for our unexpected findings. Any such explanation will entail the designation of which axiom(s) antecedent to the invalidated theorem is (are) false. Of the sub-set Of axioms used to derive theor- em VIII there is only one which does not appear as a logical antecedent to any other conclusion. This is axiom 6 which posits a small cost in the act of voting. If we substitute for this axiom the alternative assumption that: Axiom 6'. The act of not voting involves a small net cost for any legislator on any motion.‘ we may drOp theorem VIII and hypotheses H8 and H9 and derive a single new testable hypothesis describing the relative rates Of abstention which we would now expect to find among the different groups of legisla- tors On the different class Of roll calls. Our theory so revised now predicts that the only legislators who will be willing to incur the * One may still assume that there is some cost in attending a roll call and yet conclude that the £33 cost Of abstaining is greater, where the cost Of not voting is measured in the negative effect it can have on a senator's probability of re—election by lowering his over- all attendance rate and thus projecting a "Tuesday-Thursday Club image to his constituents. 121 cost Of abstaining are those for whom an abstention serves to communi- cate a threat of non-support tO potential brokers; these legislators are the indifferents and pre-bargaining supporters/Opposers on bar- gaining bills. If these are the only persons who will abstain for strategic reasons, we would expect to find a greater overall abstention rate among them than among other legislators on types Of roll calls. This new probabilistic hypothesis may be stated as follows: H8' The overall abstention rate among pre-bargaining supporters/ Opposers and indifferents on bargaining roll calls is greater than the overall abstention rate among protO-coalition members on all roll calls, pre-bargaining supporters/Opposers on non-bargaining roll calls, and indifferents on non- bargaining roll calls. This alternative hypothesis is consistent with the data.pre- sented in Table 11 where we find an abstention rate among those sena- tors predicted to sometimes abstain for strategic reasons greater than the corresponding abstention rate among those senators predicted to ggzgg abstain for strategic reasons. The two abstention rates are 11.5% and 10.0% respectively, a 15% greater prOportion of non-voting among those legislators posited to strategically abstain on occasion. We also note that the figure in column five of Table 11 (abstention rate for indifferents on bargaining roll calls) is greater than any Of the figures in columns one, two, four, or six (abstentionrrates among those legislators posited to never strategically abstain); and the figure in column three (abstention rate among pre-bargaining supporters and Opposers of bargaining roll calls) is greater than all but one of the abstentinn rates Of those groups Of persons predicted to not ab- stain for strategic reasons. These secondary Observations would seem to lend further support to our alternative hypothesis H8'. However, 122 none of these secondary relationships are very strong, and there still remains unexplained the wide variance among groups predicted to never abstain for strategic reasons. In light Of these Observations and be- cause the initial hypotheses concerning abstention rates were found to be invalid, we will be especially careful tO note relative abstention rates as we examine sub-sets Of our sample Of roll calls later in this a chapter. The last of our hypotheses is a restatement of theorem I and follows from the assumption that all legislators are long-term resource maximizers, that during any legislator's term in Office there will be a large number of roll calls, and that each legislator's utility func- tion for the resources acquired in the bargaining process is constantly increasing. H10 The position Of a motion on the legislative agenda does not affect the processes Of bargaining and coalition formation. To test this prOposition we divide our complete sample of roll calls into four sub-sets according to the session of the 89th Congress ans the calendar-half Of each session in which each motion was voted on in the Senate. The first sub-set includes those roll calls taken from the Opening of the 89th Congress through June 30, 1965; the second sub-set Of motions consists Of those voted on during the period Of a: While the size of our current sample of roll calls does not allow it here, in further tests Of this and other theories Of non- voting it might be useful to introduce as a control variable the number Of other roll calls taken of the same day as that for which abstention is predicted. If a legislator is already on the fiber (or at least in town) for another vote the same day, it is more likely that he will al- so vote On all other motions; having incurred the cost of attending one roll call for the day, the additional cost of attending others will be minimal. Lacking the data to test it ourself, this secondary hypothe- sis remains a suggestion for further research. 123 July 1, 1965 through the close Of the first session; the third group is made up Of those roll calls recorded during the second session through June 30, 1966; and the last sub-set of motions consists of the roll calls taken during the remainder Of the 89th Senate. We chose these groupings because the creation of any additional sub—samples would seriously reduce the number and variety Of rOll calls appearing W in each one. Also, by splitting the agenda of each session at June 30/ July 1 each calendar year is equally divided and the demarcation date cOrresponds to the end Of the fiscal year and the time Of the tradi- tional Independence Day congressional recess. With our sample Of roll j calls divided in this fashion, we test hypothesis H by calculating 10 and comparing the mean indices of deviation (between Observed winning coalition size and minimum-certainty-Of-winning size) for all bargain- ing and non-bargaining motions with each sub-set. The findings appear in Table 12. These data clearly support our prediction that the process Of coalition formation is independent Of the position Of a roll call on the legislative agenda. While the strength Of the relationship varies, for each of the four sub-sets of roll calls in our sample the average percent deviation from the minimum-certainty-Of-winning coalition size is greater for non-bargaining motions than for bargaining motions. Where the relationship is weakest, for the first half Of the second session, the N's are also smallest (five bargaining roll calls and a single non-bargaining motion). But throughout, including this small * sub-set, the indices Of deviation are related as predicted. # With a larger overall sample Of motions one could also test the effect of agenda-position on valence voting and abstention rates. 124 Table 12 Mean Difference Between Observed Winning Coalition Size and Minimum-Certaintyeof-Winning Coalition Size by Type of Roll Call and Agenda Position Agenda Bargaining Non-Barg. Position Roll Calls (N) Roll Calls (N) 1965. la 1st. Half 26.3 96 (27) 38.4 % (25:) 1965. 2nd. Half 11.9 (28) 21.1 ( 9) 1966, 1st. Half 20.2 ( 5) 20.4 ( 1) 1966, 2nd. Half 14.2 (56) 24.5 ( 4) All Roll Calls 16.7 (116) 32.5 (39) a. See note Table 9. 125 Another interesting relationship is evident in Table 12 and may be noted in passing. While the mean Observed winning coalition size among non-bargaining motions is consistently greater than the same statistic for bargaining bills, the prOportion of each type of roll 3 call varies significantly and systematically with agenda-position. For our entire sample Of rOll calls the prOportion of bargaining motions is 74.8%. However, Of the sub-set of these roll calls taken from the first half of the first session of the 89th Senate only 51.9% are of the bargaining type. For the next three time periods in sequence the prOportion Of bargaining roll calls increases consistently to 75.7%, then 85.3%, and finally 93.3% for the last half Of the second session. One plausible eXplanation for this phenomenon which is not at all in- consistent with our model of the legislative process is that on certain motions the potential resource payoff to the brokers is very large, the side-payments Offered are also large, and, therefore, bargaining takes a longer time delaying votes on such motions until late in the legisla- tive session. At the same time, the less controversial non-bargaining motions may be taken up at an earlier date while the new and returning senators adjust to the reconstitution of the legislative body and the redefinition of the parameters of the legislative game. The empirical results reported thus far are supportive Of most of our theory's testable, probabilistic hypotheses; and for the one in- stance in which the Observation differs consistently from the predic- tion we are able to prOpose a tentative explanation consistent with the results. However, the deterministic prOpositions are in all instances disconfirmed (as expected), and the findings which support the relati- vistic hypotheses are not everywhere as strong as might be desired (not 126 uneXpected, but disappointing). In the remainder Of this chapter we Offer a series Of explanations for the still uneXplained Observations and, wherever possible, test these eXplanations in an attempt to increase the power of our theory. Discrepencies found between Observation and prediction may re- sult from a variety of causes. For one thing, as we have noted re- peatedly, in testing our formal theory Of the legislative process we are also indirectly testing the validity of the Operational indicatOrs employed in the analysis. It is thus possible that where the data are only partially supportive of a particular hypothesis or hypotheses, the fault may lie not in the veracity of the antecedent axioms but in some error in measurement. To test this possibility we examine separately each issue-area sub-set of our roll call sample to determine whether or not our Operationalization Of the resource calculus might be more accurate for certain classes of motions than for others. A second eXplanation for weakness in our findings, particularly where Observation has only partially supported our probabilistic hypo- theses, is that some logically antecedent axiom or axioms are accurate descriptions of reality only under certain circumstances more limited than those used to define our analysis sample. That is, our assump— tions concerning the legislative process may hold only for specific types of motions rather than across all roll calls. In the following analysis we look at the possibility that the axioms positing perfect information on the part Of legislators hold only where all senators may reasonably be eXpected to have had considerable fore-knowledge that a particular motion would be introduced and voted on. We also test whether or not the description of bargaining and coalition formation 127 presented in our theory is an accurate abstraction only where one Of the brokers can control the timing Of a roll call vote and thus control also the minimum coalition size necessary to be certain of victory. Finally, we discuss how one might indirectly test the assumption Of legislator-rationality to see whether or not only some sub-set Of all senators is behaving as resource-maximizers. TO reiterate, the following analysis attempts to improve the predictive power Of our theory by examining various sub-samples of roll calls on the assumption that: (a) our Operationalization of certain concepts is more accurate for some types of motions than for others; and/or (b) our theory, or certain parts thereof, holds only under a more limited and identifiable set of circumstances than was initially assumed. Whether or not this analysis succeeds in making our theory more predictive, it should certainly succeed in giving us a greater insight into the legislative. As already noted, the most likely source of measurement error in the preceding analysis is in the Operationalization of the resource calculus. That error in the designation of the pr calculus could pro- duce significant discrepencies in our findings may be readily seen when we recall that it is the set of values generated by the calculus which is used to classify both the senators and rOll calls in ourcsam- ple. And because the Operational indicators employed in constructing the calculus are Specific to each issue-area in our sample of motions, any measurement error involved is also likely to be issue-specific. To determine if such error has been introduced, we repeat here for each issue area separately the tests performed in validating the first three 128 hypothesis-sets.‘ The results Of these issue-specific tests appear in Tables 13 through 15 and in Figures 2 and 3. Looking first at the hypothesis describing relative expected winning coalition sizes (H2), Table 13 compares the mean indices of deviation between the minimum-certainty-Of-winning coalition size and the Observed winning coalition size for bargaining and non-bargaining roll calls by issue area. Our earlier full-sample analysis supported the prediction of hypothesis H that the mean index of deviation for 2 non-bargaining roll calls is greater than the same figure for bargain- ing motions. The issue-by-issue tests reported in Table 13 reveal: considerable variance in the extent to which the data support this hy- pothesis. Specifically, we find hypothesis H supported for three of 2 the issue areas but clearly disconfirmed in the case of urban welfare votes (where we note only two non-bargaining roll calls appear). Also, the expected relationship, the mean index Of deviation for non-bargain- ing roll calls exceeding the mean index Of deviation for bargaining motions, is only very weakly evident among foreign aid motions. At the same time the relationship is as eXpected and quite strong for the labor and civil rights bills. From these Observations we may tentae tively concludetthat our Operationalization of the pr a more accurate description of the kind and amounts Of resources which calculus Offers a senator may expect to receive from extra-legislative sources on ci- vil rights and labor motions than on urban welfare and foreign aid motions. To further test this conclusion we reconstructed the fre- quency distributions of winning coalition sizes by class of roll call s 'The test of hypothesis H1O is not repeated because the small sizes Of the sub-samples of roll calls precludes meaningful analysis. 129 Table 13 Mean Difference Between Observed Winning Coalition Size and Minimum-Certainty-Of-Winning Coalition Size by Type of Roll Call and Issue Area Issue Bargaining Non-Barg. Area Roll Calls (N) Roll Calls (N) Civil Rights 23.1 %a (16) 41.4 % (13) Urban Welfare 12.2 (34) 7.1 ( 2) Labor 11.2 (31) 36-7 ( 2) Foreign Aid 25.1 (35) 29.1 (22) All Roll Calls 16.7 (116) 32.5 (39) a. See note Table 9. 130 (as appeared in Figure 1) first for civil rights/labor motions, then for urban welfare/foreign aid motions. These new distributions are presented in Figure 2 and Figure 3, reSpectively. It is apparent from these representations that not only is the eventual winning coalition size close to the minimum-certainty-of-winning size of 51 for a great- er prOportion Of civil rights and labor motions, but also there is much less overlap between the Observed winning coalition sizes on bargaining and non-bargaining bills among these roll calls than among the urban-welfare/foreign aid votes. Again, there is evidence that in the initial Operationalization of the resource calculus more measure- ment error may have been introduced into our analysis for certain issue areas than for others. This eXplanation for our earlier findings is further tested by turning to those hypotheses which predict relative rates of valence voting. Here tOO the tests performed earlier are repeated with the results divided by issue area. Table 14 presents these data. As is the case for our entire sample Of roll calls, hypotheses H5 - H7 are all supported for each issue-area sub-set of motions. However, as we Observed above, the strength of the predicted relationships varies con- siderably across issue areas. Again we find the level Of prediction imporved for civil rights and labor motions but diminished for urban welfare and foreign aid bills. These results lend further support to our conclusion that measurement error hasbeen introduced and that the error is issue-specific. Finally, the abstention rate of each class of legislator for each class of roll call is calculated with the roll calls further di- vided by issue area. The results Of these calculations appear in 131 saso nsoaso: noses can nosmam Hasao .ncso> cannoao measaaoxm naaco aaom measasmncmusoz can wsficfimwmmm mom mcuwm sowpfiamoc wmfisswz mo mowusnfiaumfin hososthh N ohswflh soasaacoo masses: no scam Woe gyms macs made $18 51mm a3: 2.5 oaw _ _ a #1 _ Ibfl LH L _ naaco sacs weasasmanmunoz 11111 naano aacm measasmnnm \ maamo Haom M0 R ION rom {OJ 10m 132 ‘ aaso nsoaeoz cad essence can cascaos amps: "mouO> caspoao wsflvsaoxm maamo Haom wdwsflmmusmusoz was msflswmwmmm How mouwm sowuHHmOo wsflssflx mo sOfluanHapmwn hososwoam M OASMwh soaeaanoo masses: so csam w om mwumm mmnms msumm wmnmm 51mm smuwa 31 E 3 W A. 111 F a a. _ a 1L _ a no- now won .os oases Haom naano Haom weasanwnnmusoz 11111 1. so a naano Haom wsacacmncm .1111111. .om 133 .or magma Opos com .m m.sm m.mm m.ow m.ws naaco Haom Has m.-m s.-m m.ss o.mm saa smwouoh $.55 w.m: m.mm :.mw honmq r.¢m r.mm r.sm m.ms madman: amps: a as a and a was ea has Essa Hw>wo maamo HHOm maamo HHOm maamo HHOm mdamo HHOm moa< msflsflmmammlmoz wswswmmnmm msflsfimmnmmtsoz wsHGMMMAmm OsmmH mammOQQO\mamvaommsm unease: wswsfimmnmmnoam coasaanoouosona sons cannH can Hans Haom no case .noscanawca co cacao as msaeo> concans mo chem s- manna 134 Table 15. We are interested here is seeing whether or nOt the extent to which the data support hypothesis H8” varies significantly with issue area. The prediction in this revised hypothesis is that the ab- stention rate among pre-bargaining supporters/Opposers and indiffer- ents on bargaining roll calls will be greater than the rate Of non- voting among all other representatives on all other types Of motions. For the entire sample of roll calls analyzed earlier these two absten- tion rates are 11.5% and 10.0%, moderately supporting the hypothesis. In examining Table 15, we see that the strength and direction of this relationship does indeed vary with the issue area being considered. The relationship between the two abstention rates is in the expected direction and strongest for civil rights motions with the rates Of non- voting being 12.6% and 5.2%, reSpectively. However, where we found the predicted relationships in the preceding hypotheses holding more strongly among labor roll calls as well, here we Observe relative ab- stention rates on such motions which tend tO disconfirm H8'. Across all labor roll calls the senators predicted to abstain only for reasons Of unavoidable dentention evidenced a non-voting rate of 12.8%, while those predicted to abstain for strategic reasons failed to vote 10.4% of the time. On urban welfare motions the non-voting rates for the presumed high-frequency and low-frequency abstainers are 10.1% and 9.6¢; on foreign aid roll calls the figures are 13.1%1and 11.9%, re- Spectively. This means that for three of the four issue areas the ex- pected relatiOnship holds, but in only one Of these three instances does the strength of the relationship increase over what it is for the entire sample Of roll calls. And while the one sub-set Of motions for which the expected relationship is strengthened is civil rights roll 135 .rr manna macs com .m r.r N.mr 0.0 v.0r 0.0 0.rr maamo Haom HH< 1 m.mr n.0r w.:r m.0w 0.mr OH< smfloaoh . m.m- o.s- a.s m.mm s.- noses 0.0 n.0r N.w :.m m.mr 0.0 mammao3 song: a. m.- a. or? a Q: & m.m a. man was :4. nsswam Hfl>w0 mflamo Haom maamo Haom maaco Haom maamo Haom maamo Haom maamo HHOm mon< .mammnsoz massamwnmm .wummnsoz msflswmwhmm .mammusoz mswswmwnmm OsmmH mpsoaommflpsH mumm0990\mhmuhommsm muonscz msasfimmammnmpm noasaacoonoaosa woa< osmmH 0am Hamo Haom mo came .nOpmHmwmoA so nndao as comm soapsoenss mr OHQOB 136 calls, consistent with the results reported in Tables 13 and 14; the one issue area for which H8. is found not to hold is the sub-set Of motions Of concern tO organized labor, inconsistent with the improve- ment in prediction noted on these roll calls for the previous issue- specific tests. Perhaps the most consevative conclusion to be drawn from this issue-by-issue analysis is that the substantive content Of the roll calls being examined does affect the Observed strength (and, occasion- ally, the direction) Of relationships described in our theory, and that there is thus some basis for believing that issue-specific measurement error is present. For all hypotheses the Observed strength of expect- ed relationships improves when attention is limited to civil rights motions alone. Similarly, the eXpected relationships are found to be everywhere weaker for urban welfare and foreign aid roll calls. The results are mixed in the case of labor motions; however, on balance it appears that less error is introduced for these roll calls than for those dealing with urban welfare and foreign aid. In any event, it is clear from this analysis that not all of the variance between predic- tion and observation can be explained by examining issue-specific sub- samples of roll calls to test for measurement error. In what follows we consider alternative eXplanations. One such alternative explanation is that the axioms in our theory which assume perfect information do not hold across all roll calls. Obviously, the less complete and perfect the information, the less the individual and aggregate voting outcomes will resemble those predicted by our theory. The plea of imperfect and/or incomplete in- formation is Often made by researchers testing game theoretical 137 formulations where results do not conform to expectations. Indeed, William Riker, after advancing the size principle of minimum winning coalitions in n-person zero-sum games with side-payments, makes ex- plicit an "information effect hypothesis": The greater the degree of imperfection or incompleteness Of in- formation, the larger will be the coalitions that coalition- makers seek to form and the more frequently will winning coalitions actually formed be greater than minimum size. Con- versely, the nearer information approaches perfection and completeness, the smaller will be the coalitions that coalition-makers aim at and the more frequently will winning coalitions actually formed be close to minimum size.3 In the case Of our theory of the legislative process we examine the effect of information level not only on sizes of eventual winning co- alitions but also on the patterns of valence voting and abstention. _The procedural type of each roll call is used as an indirect indi- cator of the level of information on the assumption that the more fore- warning a senator has that a motion will come up for a vote, the more liekly it is that he will be able to acquire information on both the pr value of the bill for each Of his colleagues (axioms 2b and 4) and on any bargains which may already have been made (axiom 12). With this in mind we class as high-information roll calls (a) final votes on conference committee reports, (b) final votes on Senate versions of bills, (c) votes on Senate committee amendments, and (d) all clo- ture motions. All other votes including floor amendments, substitute motions, tabling motions, and motions for reconsideration are conSi- dared to be low-information roll calls. Of our full sample of 155 motions, 35 are high-information roll calls and 120 low-information roll calls. Retesting the first sets Of hypotheses for each of these two sub-samples, if our extension of the information effect hypothesis 138 is valid, we would expect the results Of our analysis to more closely conform to theoretical expectations for high-information roll calls than for low-information motions. For prOpositions H1 and H2, we find the data support the in- formation effect hypothesis. As reported in Table 16, the mean index of deviation from minimum—certainty-Of-winning coalition size is closer to zero for high-information bargaining motions than for low- information bargaining motions--15.1% 3g; 17.1%. Also, there is a greater difference between the mean indices Of deviation for bargain- ing Kg. non-bargaining roll calls among high-information motions than among low-information motions. Turning next to the hypotheses dealing with relative rates Of valence voting, the effect of information level is mixed but generally in the expected direction. Table 17 gives the rates of valence voting among the members of the 89th Senate by class Of legislator, type of roll call (bargaining or non-bargaining), and information level. We see a substantial improvement in the prOportion Of voting proto- coalition members whose votes are correctly predicted between low- and high-information roll calls. Among pre-bargaining supporters and Opposers on non-bargaining roll calls, we find a slightly higher rate Of valence voting on low-information motions as Opposed to the rate on high-information roll calls, and this latter result is counter to what the information effect hypothesis would lead us to believe. Of great- er interest are the effects of information level on the probabilistic relationships described in hypotheses H5 through H7. From Table 17 we see that all of the predicted relationships dchold for both high- and low-information roll calls, and in two of the three cases the 139 Table 16 Mean Difference Between Observed Winning Coalition Size and Minimum-Certainty-of—Winning Coalition Size by Type of Roll Call and Information Level Information Bargaining Non-Barg. Levela ' Roll Calls (N) Roll Calls (N) High b Information 15.1 % (24) 36.7 % (11) Low Information 17.1 (92) 30.8 (28) All Roll Calls 16.7 (116) 32.5 (39) a. High information roll calls include final passage votes (both Senate and conference bills), committee amendments, and clo- ture motions. Low information roll calls are votes taken on floor amendments and other procedural motions (substitute mo- tions, tabling and reconsideration motions, etc.). b. See note Table 9. g :i e g ! .‘ l: L 140 _l 1 | 11.-(1.? .or manna moon 000 .n .0? manna coo: com .m 0.00 0.mm m.0w m.m0 maamo HHOm HH< 0.wm w.¢m :.ws 0.00 sowpmshomsH 304 a eds a men a was ha 0% Saragossa swam m maamo HHOm maaso Haom maamo Haom maamo HHOm Ho>oq wowswmwammusoz mswsamwamm msflswmwammnsoz mswsfimwnmm cospmsaomsH mammom00\mh0paommsm unease: weasaemnemnono soasaaeoouonono Hosea ecasesnoasH can Hans aaom no case .noonanamoa so needs so meano> cocoan> mo comm s- cases 141 Observed relationship is stronger where the information level is greater. Only in the case Of the fifth hypothesis, where we prOpose that the rate of valence voting among pre-bargaining supporters/ Opposers will be greater on non-bargaining roll calls than on bargain- ing ones, is the relationship found to be stronger on low-information motions. And here the difference is very small indeed (68.9 - 54.8 = 14.1% for low-information roll calls as Opposed to 65.1 - 51.8 = 13.3% for high-information roll calls). Furthermore, for all those senators whose eventual votes are predicted by our theory--proto- coalition members on all roll calls and pre-bargaining supporters/ Opposers on non-bargaining roll calls--these predictions are correct 82.0% of the time on high-information motions and only 75.6%iof the time on low-information motions. This result, as well as the others cited in this paragraph, seems to support the information effect hy- pothesis as it relates to predictions of individual voting. However, in predicting abstention rates the Opposite is true. As reported in Table 18, the relationship described in hypothesis H8' holds among low-information roll calls but 222 among high-information motions. On the latter sub-set of roll calls the senators predicted to sometimes abstain for strategic reasons-~pre-bargaining supporters/ Opposers and indifferents on bargaining motions--failed to vote 7.9% Of the time, while for all other legislators on all other motions the abstention rate is 9.1%; on low-information roll calls the figures are 12.2% and 10.3%, respectively. The information effect hypothesis as Operationalized and tested here is disconfirmed. That is, there is less conformity between prediction and Observation for high-information roll calls than for low-information roll calls. 142 .ee oases coon com .c .0r canoe ouos mom .m e.- m.m- 0.0 e.oe 0.0 o.-- naado Haom Has m.r s.mr m.0 0.rr r.m 0.Nr sowpmsaomsH 30a .\e 0.0 R. 0.w & aflw & 0.0 mm m.: 0.0m m.w sown-€0.80” awam m maamo Haom maaso Haom maamo Haom maamo HHOm maamo Haom maawo Haom HO>OA .wammnsoz msfisfimmamm .mnmmumoz wowswmmamm .mnmm:soz wswsamwnmm qOfivmshOHsH mascacmmflusH mamm0000\mnouaommdm mucosa: mswaasmanmuond soasaasooioeona HQ>QH GOflpw—FHOHGH fig Hans aaom no case .noocanamoa no woman so cuss noaudcsnos we cares 143 On balance, the findings reflecting on the validity of the information effect hypothesis are ambiguous. Clearly, only a limited amount of the error in the original findings can be explained on the basis of differential applicability Of the perfect-information axioms in our theory. Other possible explanations should be considered. The next such eXplanation to be examined is that the axioms assuming all vote commitments secured by side-payments are binding (axiom 11) and all brokers have perfect information concerneing vote commitments (axiom 12) are not every where applicable. It is possible that these assumptions are somewhat inaccurate abstractions Of rality. Such an inaccuracy does not affect our predictions describing indivi- dual and aggregate voting outcomes on non-bargaining motions. However, in the case Of bargaining motions this inaccuracy, if it exists, would result in the eventual winning coalitions being larger or smaller than the minimum-certainty-Of-winning size. Furthermore, such a short— coming in our theory could eXplain the difficulties we have found in predicting abstention rates. FOr, if the fact that one side has al- ready achieved certainty-Of-winning size is not immediately known to the other broker and/or if it is possible for support acquired through bargaining to revert to the other side in the face Of alternative Offers, at least two specific results would Occur: First, some of the members of the eventual winning side might be bought away (producing a smaller than minimum-certainty-Of-winning coalition), or the broker for the eventual winning side might continue to buy members (pro- ducing a larger than minimum-certainty—of—winning coalition). Second, in this process Of "unnecessary" bargaining some of the unbought pre- bargaining supporters/Opposers and indifferents who would otherwise 144 abstain might be bought into one of the coalitions, lessening their overall non-voting rate and thus weakening the eXpected relationship between their abstention rate and the abstention rate Of other repre- sentatives. To test this possibility we divide all bargaining roll calls in our sample into two sub-sets according to whether or not the Demo- cratic party took a position on each motion. Assuming the Democratic majority party leadership--the presumed broker for at least some of the roll calls on which it takes a position--has some control over the scheduling Of votes, there will be less unnecessary bargaining on these, the Democratic-party-position bargaining motions, than on the other bargaining motions. Minimum-certainty-of—winning coalitions would occur more frequently on the Democratic-party-position bargaining roll calls because the majority party leadership in the Senate would be able to call for the vote as soon as--and if--they, as broker, achieved a coalition large enough to win. If these assumptions are valid, we would, therefore, eXpect to find less deviation from mini- mum-certainty-Of-winning size among the eventual winning coalitions on Democratic-party-position bargaining roll calls than among other bargaining roll calls. Our data support this conClusion. As indicated in Table 19, the mean index of deviation for all bargaining roll calls on which the Democratic party took a position is 12.5%, while the mean index Of deviation for all other bargaining roll calls is 21.M%. This finding is presented graphically in Figure 4. The relationship is in the expected direction and also quite strong, leading to the tentative conclusion that axioms 11 and 12 are not completely accurate and that this inaccuracy may account not only 145 Table 19 Mean Difference Between Observed Winning Coalition Size and Minimum-Certainty-Of-Winning Coalition Size for Bargaining Roll Calls by Democratic Party Position Democratic Party N2 Democratic Party Position Taken (N) Position Taken (N) Mean Index of Deviationh 12.5 % (61) 21.4 % (55) a. See note Table 9. 5o.1 % of Democratic Party Position Taken Roll Calls ------ N2 Democratic Party Position Taken 304 20- 104 I l r I I j 1 #l I 1 am 41.47 48-54 55-61 62-68 69-75 76-82 83-89 90$ Size of Winning Coalition Figure 4 Frequency Distribution of Winning Coalition Sizes for Bargaining Roll Calls, Excluding Cloture Votes, by Democratic Party Position l_.s’ 21..- . 146 for some of the error in predicting winning coalition sizes but also for some Of the error in predicting relative abstention rates among senators. While this conclusion may appear more satisfying than those drawn from the issue-Specific and information—effect analyses, it should not close our mind to other explanations for the variance Ob- served in our results. Though we do not pursue it, before concluding our discussion Of the empirical findings, we ought to note one final explanation. This is the possibility that the differentially applie cable axiom is the initial one--the assumption of legislator ration- ality. If, in fact, certain senators are "crazy" in the sense of being unwilling or unable to calculate their rational voting strate- gies, it is likely that they will be excluded from bargaining and that their eventual votes will be randomly distributed, thus distorting all of our findings. Furthermore, if this were true, we would expect the irrational senators to be the least successful in achieving re-elec- tion. One could test this secondary hypothesis by comparing the rates Of defection from protO-coalitions with re-electoral success rates, the expected result being that the legislators defecting most frequent- ly would be least likely to be re-elected having failed to vauire sufficient resources in the inter-election period. To summarize our body of empirical findings: In evaluating the validity of the hypotheses derived from our theory Of legislative coalitions, we found trends in the data to be supportive of all but one of the deterministic prOpositions, while all save one of our testable probabilistic hypotheses were supported. In this one case-- predicting relative abstention rates--we prOposed an alteration in 147 the theory resulting in an alternative derived hypothesis, this one consistent with Observation. These generally supportive findings led us to the conclusion that the theory of coalition formation in legisla- tures Offers an accurate abstraction of the legislative process in most instances and that what error does exist in our findings may have been introduced in Operationalizing the complex resource calculus antecedent to testing the formulation. Examining this possibility, it was determined that, indeed, for two of the four issue areas in- cluded in our sample Of roll calls there was consistently greater con- formity between prediction and Observation. We next attempted to improve the predictive power of our model by considering the possi- bility that the theory is an accurate abstraction Of reality for only certain identifiable sub-sets of all roll calls. Here we found that we were able to specify a priori those motions (by procedural type and majority party position) for which our theory comes closest to cor- rectly predicting individual and aggregate voting outcomes. Again, the single exception occurred in the case of predicting relative ab- stention rates among senators. This difficulty in predicting abstentions remains an empirical disappointment if not an unexpected one. Particularly encouraging in terms of the probable validity Of our theory is the fact that we found considerable and consistent evidence for the existence Of two specific sorts of roll calls on only one of which bargaining will occur and the eventual winning coalitions will approach the minimum size necessary tO be certain Of victory. Limiting our attention to those roll calls for which we hypothesized our model to be most accurate, we find that Of all the motions in our 148 sample dealing with civil rights and organized labor issues, where the information level is presumed to be high for all senators, where the Democratic (majority) party takes a position, and where the minimum- certainty-Of-winning size is 51 votes and pairs fully 67% were carried with coalitions within three votes of the eXpected size, and the other 33% of these motions were carried by winning coaltions within ten votes of 51. Furthermore, our findings in terms of correctly predicting individual votes and roll call outcomes compare favorably with other studies on congressional behavior. Across all motions in our sample, Of the senators whom we predict specific votes, 77.4% Of those paired or voting did take the eXpected position. This compares with 84% correctly predicted in a simulation Of the House of Representatives1+ and 66.2% correctly predicted in a similar simulation of voting in the U.S. Senate.5 Of all motions for which our theory predicts a winning side, e.g. all non-bargaining roll calls, these predictions are correct 97% of the time (38 Of 59 roll calls). FOOTNOTES: CHAPTER IV 1Our index Of deviation is similar to the independently arrived at "index of minimalness" developed by Richard Murray and Donald S. Lutz in "Redistricting Decisions in the American States: A Test of Riker's Minimal Winning Coalition Hypothesis," a paper read before the 1971 Annual Meeting of the American Political Science Association, Conrad Hilton Hotel, Chicago, Illinois, September 7-11, 1971. 2David H. Koehler, "Coalition Formation and the Legislative Process," a paper read at the 1971 Annual Meeting of the American Political Science Association, Conrad Hilton Hotel, Chicago, Illi- nois, September 7-11, 1971. 3William Riker, The Theogy of Political Coalitions (New Haven: Yale University Press, 1962), pp. 88489. #Cleo Cherryholmes and Michael Shapiro, Representatives and Roll Calls (Indianapolis: J-he Bobbs-Merrill Company, Inc., 1965). 5Jeanne L. Martin, "Exchange Theory and Legislative Behavior: A Computer Simulation of Roll-Call Voting in the U.S. Senate," (un- published Ph.D. dissertation, Michigan State University, 1971). 149 ‘ .. ‘35 -‘v- _ CHAPTER V CONCLUSIONS The results reported in the preceding chapter are simulta- neously heartening and discouraging. On the one hand, we find con- siderable support for our central coalition-size hypothesis; on the other hand, support for some of the hypotheses is quite weak, and in another instance the findings disconfirm one Of the original hypotheses. 0n.balance, there appears to be reason to believe that the theory of coalition formation in legislatures offers a non-Obvious and, by no means completely inaccurate abstraction of the legislative process. However, to make the derivatives Of our theory consistent with Obser- vations of voting in the U.S. Senate we have already found it necessary to alter certain parts of the formulation and to qualify the applic- ability Of others. In this concluding chapter we first note the changes made in the theory. We then discuss applications of the revised theory to other legislative settings and note further opportunities suggested by our analysis. We close with a brief evaluation of the utility Of the current work and its significance for the study of legislative be- haVioro Revision Of the Theory in Light Of the Findings Throughout the analysis in Chapter IV we note implications of our findings for the theory Of coalition formation in legislatures. 150 : rues-zit”, 151 Below we offer a reformulation of the theory, revised in light of the empirical findings. As most Of these findings are supportive Of the initial formulation, the theory given here closely resembles the one first presented in Chapter II. Where changes are made, they are duly nOtedo Symbols and Definitions: (a) (b) (c) (d) (e) (f) (g) (h) (i) R: Resources necessary to a legislator for his re-election. All R is measured by the effect it will have on a legislator's probability of being re-elected. All R must then be trans- lateable into votes, i.e. constituency support, financial con- tributions, organizational assistance, support for election to a legislative leadership position, etc. R : The amount of resources which a given legislator may egpect to receive for voting on a particular side Of a par- ticular motion independent of resources Offered in the course Of bargaining on the motion. The R value of a motion to a legislator in the pre-bargaining stage. pr: The value of pr such that no legislator expecting to receive pr or more resources for voting a particular posi— tion on a particular motion will be offered additional re- sources during bargaining in exchange for a vote commitment. Supporting protO—coalition: All those legislators with a value of pr equal to or greater than pr for supporting a given motion. Opposing proto-coalition: All those legislators with a value of pr equal to or greater than pr for Opposing a given motion. Pre-bargaining supporter: Any legislator with a value Of pr on a given motion such that O.30 swam ca.s>v HHH swam A.m>v can .vahm acacasm soumzoam msomaom noon moonwoz osoosH amps: names ocean: shonpsom AoHonqhm0 x x m 0 somxomb x x Q OhsocH m m mxmsam x x m 0 somaaom x x n Haflm m ammooasoxoam x 0 amazon 0 oxpnmm msowaom soom AOH0|ghoooz a come: a w some a snaaoaocs m a enhances x m a meacaanse: x x Q consume: a x m n A.cav mwsoa x x a A.ozv when x x x m a chooses x a x m assess muomkmnm .HOOnH AOHOév mmvflfiv’ meRmOZ “mum. M0 SOHvomHmH hvhmm honvmaaom OsOOsH some: aonmA succusom mmosommm axoz 1smam sosasemoo so open mmsoaw pompousH\psosufipmsoo posscosoo usmOflMsswam 170 x mw0r n Haom wo0r m nonsmom x m n ooOpmmm 000V 0 aomaonsoz x a somaoz m a Omesz x x x m hanks: mw0r m puss: Q moo: x m sounoz x Q coho: n smOpsoz x n mesoasoz mw0r n masosoz msomaom soom AOHDIAMv .uh .pamm hOpssOm mo sump scafimxosd wsaaaam .O .mpoumsom O>wpoommoa mo aasoamo coasomou msflsflmwamnnosm cuss house you 00 voths nos mmsomw moose mo moosoaomomm .0 .AHHH scammso oomv Ommm seasons ma anon on soapmsflaosom n m .o .000r no wm0r nonpwo as sOfipOOHOIOM comm maOpmsom nonpo HHd .n .qufiHnsmom u m .m .umaooson u a x x x a noasov asses m m A.n.zv endow x x n smoothnes x x m A.Hoav msmaaaaz x x m 0 A.b.zv memaHHfiB x x x x a nasasas x x mm0r m nozoa x m mm0r m ssosnsne x x m a ompmsame x a m a cosmeaaam msomaom soom A0H019hww has so coauflmom m.pseowmoum on» wsauhommsm Mom A.mqv msoq was vaowmmsm: mucussom cu canoes Op seasons ow mooasomoa mo was: HmsOflufippm Ono .w samsmzoz monsoon mo snow soawmxoss msaaaflh .m 6.380 .m waesooaa on 882 APPENDIX C SAMPLE 0F ROLL CALLS Issue Agenda Congress'l Proce- Demo- Repub- Constituent/ Area Position Quarterly dural cratic lican Interest Number Typea Sigiyio garty. group. nb 051 ionb 081 ionb,c Civil 1965, 49 FA 0 + + Rights 1st half 51 FA - 0 - 52 FA - 0 + 53 FA - O - 54 FA 0 - - 56 FA 0 O - 57 FA - o .. 59 FA - O - 60 FA 0 0 + 61 FA 0 O + 62 FA 0 0 + 64 FA 0 - - 65 FA - _ - 66 FA - - - 67 C10 + + + 68 FA - - - 174 175 Issue Agenda Congress'l Proce- Demo- Repub- Constituent/ Area Position Quarterly dural cratic lican Interest Number Type Party Party Group Position Position Position Civil 1965, 69 FA - - _ Rights 1st half 70 FA - - - 71 FA - - - A i 72 FA - - - 73 FA - - - 74 FA - - - 75 FA - - - 76 FA 0 - - 77 SUb + + + 78 PS + + 4. 2nd 178 PC + + + half Civil 1966, 185 Clo + O + Rights 2nd half 186 C10 + - + Urban 1965, 158 Sub O O + Wel— 2nd fare half 159 FA - + - 160 FA - + - 161 FA 0 + - 162 PS + 0 + 183 FA - + .. 184 Sub 0 + - 185 Tab 0 - + 186 Rec 0 + - 176 Issue Agenda Congress'l Proce- Demo- Repub- Constituent/ Area Position Quarterly dural cratic lican Interest Number Type Party Party Group Position Position Position Urban 1 965, 1 87 Sub O + - Wel- 2nd fare half 188 FA 0 O - 1 89 FA - + - 190 FA 0 0 - 191 FA 0 O - 192 FA 0 O _ 193 FA 0 O - 194 FA 0 + - 195 FA 0 + - 1 96 FA - O - 197 FA + + + 198 FA 0 + - 1 99 PS + O .,. 241 PC + 0 + 253 FA - + - 254 FA - + - Urban 1 966 9 33 FA + - + Wel- 1st fare half 34 CA - + - 2nd 142 FA - O - half 161 FA - O .. 163 PS + O + 201 FA 0 0 - 204 Sub + 0 - 177 Issue Agenda Congress'l Proce- Demo- Repub- Constituent/ Area Position Quarterly dural cratic lican Interest Position Type Party Party Group Position Position Position Urban 1966, 205 Tab 0 0 + Wel- 2nd fare half 206 FA 0 0 - 207 PS + + + 233 PC + 0 + Labor 1 965, 30 FA - + .. 1st . half 31 PC + + + 2nd 248 Clo + 0 + half Labor 1966, 4 FA 0 0 - 1st half 5 FC + + + 8 Clo + O + 9 Clo + 0 + 2nd 122 CA + - + half 123 CA + - + 124 CA + - + 125 FA + - + 127 Tab - + - 128 CA + - + 129 CA + - + 130 CA + - + 131 Sub - + - 132 FA 0 O + 133 FA + o + 178 Issue Agenda Congress'l Proce- Demo- Repub- Constituent/ Area Position Quarterly dural cratic lican Interest Number Type Party Party GrOup Position Position Position Labor 1966, 134 FA + O + 2nd half 135 PS + 0 + 165 FA - o - 166 Sub o o + 167 FA 0 O - 168 FA 0 + + 169 FA 0 O - 170 FA 0 o - 171 FA 0 o - 172 FA 0 o _ 173 FA - O - 174 FA 0 o - 175 FA 0 o - 176 PS + o + 184 PC + O + For- 1965, 86 FA - o O eign 1st Aid half 87 FA + o 0 88 FA 0 + o 89 FA 0 - 0 90 FA 0 0 o 91 FA + - 0 92 FA 0 0 0 179 Issue Agenda Congress'l Proce- Demo- Repub- Constituent/ Area Position Quarterly dural cratic lican Interest Number Type Party Party Group Position Position Position For- 1965, 93 FA - - 0 eign 1st Aid half 94 FA 0 + O 95 FA 0 O O 96 FA 0 - O 97 FA — — 0 98 FA - - 0 99 FA - - 0 100 FA 0 + 0 101 FA 0 + 0 102 FA - - O 103 FA - - 0 104 FA - — 0 105 FA - - 0 106 FA - - o 107 FA - - 0 108 FA - - O 109 PS + + 0 2nd 201 PC + + 0 half 233 FA - + 0 234 FA - 0 0 235 FA - - 0 236 FA - - o 237 FA - - 0 180 Issue Agenda Congress'l Proce- Demo- Repub— Constituent/ Area Position Quarterly dural cratic lican Interest Number Type Party Party Group Position Position Position For- 1965, 258 FA - o O eign 2nd Aid half 239 FA - - 0 240 PS + + 0 245 PC + + 0 For- 1966, 96 Sub + 0 O eign 2nd Aid half 97 Gs + 0 0 98 FA - + 0 99 FA - O 0 100 FA 0 - O 101 Sub O - 0 102 FA - 0 0 103 FA 0 + o 104 FA + 0 0 105 FA + - O 106 FA 0 - O 107 FA 0 - O 108 Tab 0 O 0 109 Tab 0 0 O 110 FA 0 - 0 111 FA 0 O O 112 PS + + 0 113 FA - - O 114 FA - - o 181 Issue Agenda Congress'l Proce- Demo- Repub- Constituent/ Area Position Quarterly dural cratic lican Interest Number Type Party Party Group Position Position Position For- 1966, 115 PS + 0 O eign 2nd Aid half 183 PC + O O 208 FA - 0 O 209 PC 0 + 0 a. High Information Roll Calls: Final passage, Senate bill Final passage, conference bill Committee amendment PS PC CA Clo = Cloture motion Low Information Roll Calls: FA = Floor amendment Sub Tab ReC + 0 Substitute motion Tabling motion Reconsideration motion Supporting position taken Opposing position taken NO position taken c. Constituent/Interest Group positions given are for: Negro constituents on Civil Rights motions (Southern white constituent position is Opposite of the one given) Organized labor on Labor motions Urban poor constituent positions on Urban Welfare motions (High—income persons' position is opposite Of the one given) BIBLIOGRAPHY BIBLIOGRAPHY Alker, Haywood R. Jr. "Computer Simulations, Conceptual Frameworks, and Coalition Behavior." In Groennings, SE al, The Study of Coalition Behavior. Anderson, Lee F. "Variability in the Unidimensionality of Legislative Voting," Journal Of Politics, XXVI (1964). Anderson, L. F., Watts, M. W., and Wilcox, A. R. Legislative Roll- Call Analysis. Evanston: Northwestern University Press, 1966. Andrain, Charles F. "Senators' Attitudes Toward Civil Rights.@ In Patterson, American Legislative Behavior. Barber, James D. "Leadership Strategies for Legislative Party CO- hesion," Journal of Politics,_XXVIII (May, 1966) Bauer, R. A., Pool, 1., and Dexter, L. A. American Business and Public Policy. New York: Atherton Press, 1964. Bibby, John, and Davidson, Roger. On Capitol Hill. New York: Holt, Rinehart, and Winston, Inc., 1967. Belknap, G. M. "A Method for Analyzing Legislative Behavior," Midwest Journal of Political Science, II (November, 1958). Campbell, A., Converse, P. E., Miller, W. E., and Stokes, D. E. The American Voter. New York: John Wiley & Sons, 1960. Cherryholmes, Cleo H., and Shapiro, Michael J. Representatives and Roll Calls. Indianapolis: The Bobbs-Merrill Company, Inc., 1969. Clapp, Charles L. The Congressman: His Work as He Sees It. Garden City, New York: Doubleday and Company, 1963. Clausen, A. R. "The Measurement Of Legislative Group Behavior," Midwest Journal Of Political Science, II (1967). Cleaveland, Frederic N. and Associates. Congress and Urban Problems. Washington: The Brookings InstitutiOn, 1969. Coleman, James S. "The Marginal Utility of a Vote Commitment," Public Choice, V (Fall, 1968). 182 183 Coleman, James S. "Foundations of a Theory of Collective Decisions,” American Journal of Sociology, LXXI (May, 1966). Converse, P. E. "Information Flow and the Stability Of Partisan Atti- tudes," Public Opinion Quarterly, XXIV (1962). Crane, Wilder Jr. "A Caveat on Roll-Call Studies of Party Voting," Midwest Journal of Political Science, IV (August, 1960). Crane, Wilder Jr. "DO Representatives Represent?," Journal of Poli- tics, XXII (1960). Davis, Otto A., and Hinich, Melvin J. "A Mathematical Model of Policy Formation in a Democratic Society." In J. L. Bernd, ed., Mathematical Applications in Political Science II. Dallas: Arnold Foundation, SMU Press, 1966: Davis, Otto A., Hinich, Melvin J., and Ordeshook, Peter C.f "An Ex- pository DevelOpment of a Mathematical Model of the Electoral Process," The American PoliticaIEScience Review, LXIV, No 2 (June, 1970). Downs, Anthony. An Economic Theory of Democragy. New York: Harper and Row, 1957. Downs, Anthony. Inside Bureaucraey. Boston: Little, Brown, & Company, 1967. Dye, Thomas R. "A Comparison of Constituency Influences in the Upper and Lower Chambers of a State Legislature," Western Political Quarterly, XIV (1968). Evans, Rowland, and Novak, Robert. Lyndon B. Johnson: The Exercise of Power. Cleveland: The World Publishing Company, 1966. Francis, Wayne L. "Coalitions in American State Legislatures: A Propositional Analysis," In Groennings, 22.229 The Study of Coalition Behavior. Francis, Wayne L. "Influences and Interaction in a Legislative Body," American Political Science Review, LVI (1962). Froman, Lewis A. Jr. angressmen and Their Constituencies. Chicago: Rand McNally & Co., 1963. Froman, Lewis A. Jr. "Differences Between the House and Senate." In Wolfinger, ed., Readings on Congress. Froman, Lewis A. Jr. "The Importance Of Individuality in Voting in Congress," Journal Of Politics, XXV (1963). Froman, Lewis A. Jr. "Inter-party Constituency Differences and Con- ifisfimonal” Voting Behavior," American Political Science Review, 184 Froman, Lewis A. Jr., and Ripley, Randall B. "Conditions for Party Leadership: The Case of the House Democrats," American Pol- itical Science Review, LXIX (1965). Grassmuck, G. Sectional Biases in Congress on Foreign Policy. Bal- timore: Johns Hopkins Press, 1951. Greenstein, F. I., and Jackson, E. F. "A Second Look at the Validity of Roll-Call Analysis," Midwest Journal of Political Science, VII (1963). Groennings, Sven. "Patterns, Strategies, and Payoffs in Norwegian Coalition Formation." In Groennings, SE El, The Study_of Coalition Behavior. Groennings, Sven, Kelley, E. W., and Leiserson, Michael, eds. The Study of Coalition Behavior. New York: Holt, Rinehart, and Winston, Inc., 1979. Grumm, John G. "A Factor Analysis Of Legislative Behavior," Midwest Journal Of Political Science, VII (1963). Grumm, John G. "The Systematic Analysis of Blocs in the Study of Legislative Behavior," Western Political Quarterly, XVIII (1965). '7 Hain, Paul. "The Legislative System." Unpublished Ph. D. disserta- tion, Michigan State University, forthcoming 1971. Havens, M. C.. "MetrOpolitan Areas and Congress: Foreign Policy and National Security," Journal Of Politics, XXVI (1964). Hunt, William H., Crane, Wilder W., and Wahlke, John C. "Interview- ing Political Elites in Cross-Cultural Comparative Research." In Patterson, American Legislative Behavior. Huntington, S. P. "A Revised Theory of American Party Politics," American Political Science Review, XLIV (1960). Jackson, John E. "Statistical Models of Senate Roll Call Voting," American Political Science Review, LXV, No. 2 (June, 1971). Jewell, M. E. "Evaluating the Decline of Southern Internationalism Through Senatorial Roll-Call Votes," Journal Of Politics, XXI (1959). Jones, Charles A. "Inter-party Competition for CongresSional Seats," Western Political Quarterly, XVII (September, 1964). Jones, Charles A. "The Minority Party and Policy-Making in the House of Representatives," American Political Science Review, LXII (June, 1%8)o 185 Kelley, E. W. "Utility Theory and Political Coalitions: Problems of Operationalization." In Groennings, 22.229 The Studygof Coalition Behavior. Kesselman, M. "Presidential Leadership in Congress on Foreign Policy," Midwest Journal of Political Science, V (1961). Key, V. 0. Southern Politics. New York: Alfred A. Knopf, 1949. Koehler, David H. "Coalition Formation and the Legislative Process." Paper read at the 1971 Annual Meeting of the American Politi- cal Science Association, Conrad Hilton Hoterl, Chicago, Il- linois, September 7-11, 1971. Leiserson, Michael. "Coalition Government in Japan." In Groennings, ‘23 El! The Study of Coalition Behavior. Luce, R. Duncan, and Raiffa, Howard. Games and Decisions: Intro- duction and Critical Survey. New York: John Wiley and Sons, T957. MacRae, Duncan Jr. "Roll-Call Votes and Leadership," Public Opinion Quarterly, xx (1956). Manley, John F. "The House Committee on Ways and Means: Conflict Management in a Congressional Committee." In Wolfinger, ed., Readings on Congress. Martin, Jeanne L. "Exchange Theory and Legislative Behavior: A Com- puter Simulation of Roll-Call Voting in the U.S. Senate." Unpublished Ph.D. dissertation, Michigan State University, 1971. Matthews, Donald R. U.S. Senators and Their World. New York: Vint- age Press, 1960. Mayhew, David R. Party Loyalty Among Congressmen. Cambridge: Harv- ard University Press, 1966. Meller, Norman. "Legislative Behavior Research," Western Political "Quarterly, XVIII (1965). Meltz, David B. "Competition and Cohesion: A Model of Majority Party Legislative Bargaining." Unpublished Ph.D. disserta- tion, University of Rochester, 1970. Meltz, David B. "Legislative Party Cohesion: A Model ofthe Bargain- ing Process in State Legislatures," Journal of Politics (forthcoming November, 1972). Merkl, Peter H. "Coalition Politics in West Germany." In Groennings, g£_§l, The Study Of Coalition Behavior. 186 Milbrath, Lester W. _The Washington Lobbyists. Chicago: Rand McNally & Company, 1960. Miller, Warren E., and Stokes, Donald E. "Constituency Influence in Congress," American Political Science Review, LVII (1963). Murray, Richard W. and Lutz, Donald S. "Redistricting Decisions in the American States: A Test of Riker's Minimal Winning CO- alition Hypothesis." Paper read before the 1971 Annual Meet— ing of the American Political Science Association, Conrad Hilton Hotel, Chicago, Illinois, September 7-11, 1971. Olson, Mancur Jr. The Logic Of Collective Action. New York: Schock- en Books, 1968. Patterson, Samuel, ed. American Legislative Behavior: A Reader. Princeton, New Jersey: D. Van Nostrand Company, Inc., 1968. Polsby, Nelson. Congress and the Presidency. Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1964. Polsby, Nelson, ed. Congressional Behavior. New York: Random House , 1 971 o Pomper, Gerald M. "Conflict and Coalitions at the Constitutional Conventions." In Groennings, 23 al, The Study Of Coalition Behavior. Rieselbach, L. N. "The Demography Of the Congressional Vote on For- eign Aid 1939—1958," American Political Science Review, LVIII (1964). Riker, William H. The Theory Of Political Coalitions. New Haven: Yale University Press, 1962. Riker, William H., and Ordeshook, Peter. "A Theory of the Calculus Of Voting," American Political Science Review, LXII (March, 1968). Ripley, Randall B. "Legislative Bargaining and the Food Stamp Act, 1964." In Cleaveland and Associates, Congress and Urban Problems. Ripley, Randall B. Power in the Senate. New York: St.Martin's Press, 1969. Robinson, James A. Congress and Foreign Policy-Making. Homewood, Illinois: The Dorsey Press, 1967. Rohde, David. "Strategy and Ideology: The Assignment of Majority Opinions in the United States Supreme Court." Unpublished Ph.D. dissenation, University Of Rochester, 1971. c ,. v-L 187 Rosenthal, Howard. "Size Of Coalition and Electoral Outcomes in the Fourth French Republic." In Groennings, 23.231 The Study of Coalition Behavior. Rudner, Richard S. Philosophy_of Social Science. Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1966; Schelling, Thomas C. The Strategy Of Conflict. London: Oxford University Press, 1960. Schlesinger, Joseph. Ambition and Politics. Chicago: Rand McNally 8! CO. , 1%50 Scott, Andrew M., and Hunt, Margaret A. Congress and Lobbies. Chapel Hill: University of North Carolina Press, 1965. Shapiro, Michael A. "Rational Political Man: A Synthesis Of Economic and Social Psychological Perspectives," American Political Science Review, LXIII, NO. 4 (December, 1969). Shubik, Martin, ed. Game Theory and Related Approaches to Social Behavior. New York:, John Wiley & Sons, Inc., 1963. Sorauf, Frank J. Pargngepresentation: Legislative Politics in Pennsylvania. New York: Atherton Press, 1963. Stewart, John G. "Two Strategies of Leadership: Johnson and Mans- Field." In Polsby, ed., Congressional Behavior. Truman, David B. The Congressional Party. New York: John Wiley & Sons, Inc., 1959. Turner, Julius. Party and Constituency: Pressures on Congress. Bal- timore: Johns HOpkins Press, 1951. Turner, Julius. "Primary Elections as the Alternative to Party Com- petition in Safe Seats," Journal of Politics, xv (1953). Wahlke, John C., Eulau, Heinz, Buchanan, William, and Ferguson, LeRoy C. The Legislative System. New York: John Wiley & Sons, Inc., 1962. Wolfinger, Raymond E., ed. Readings on Congress. Englewood Cliffs, New Jersey: Prentice—Hall, Inc., 1971. Wolfinger, Raymond E., and Hollinger, Joan H. "Safe Seats, Seniority, and Power in Congress." In Wolfinger, ed., Readings on Con- greSS. Yarwood, Dean. "A Failure in Coalition Maintenance: The Defection Of the South Prior to the Civil War." In Groennings,‘e£‘gl, The Study of Coalition Behavior. 188 Zeigler, Harmon. Interest Groups in American Society. Englewood Cliffs, New Jersey: Prentice-Hall, Inc., 1966. R MMMMMM i l 1111111121911111111111111111 1