EXCHANGE-THEORY AND LEGISLATIVE BEHAVIOR: : \ 
A COMPUTER SIMULATION OF NULLaeALL » f
VOTING IN THE US. SENATE * V f

' Thesis for the Degree of Ph. D.
MICHIGAN STATE UNIVERSITY
JEANNE LOUISE MARTIN ‘
1971

This is to certify that the

thesis entitled

EXCHANGE THEORY AND LEGISLATIVE BEHAVIOR:
A COMPUTER SIMULATION OF ROLL-CALL
VOTING IN THE U. S. SENATE

presented by
Jeanne Louise Martin

has been accepted towards fulfillment
of the requirements for

Ph.D. degreein Political Science

 

Mqior '

Date April 2, 1971

0-7639

 

V 7““
LI ‘11-

- ‘
Mien; ‘11 State a

Uni» crsity I]

 

 

 

@Wé
WI
W27

 

 

 

 

 

 

ABSTRACT

EXCHANGE THEORY AND LEGISLATIVE BEHAVIOR:
A COMPUTER SIMULATION OF ROLL-CALL
VOTING IN THE U.S. SENATE

BY

Jeanne Louise Martin

In a political system, a number of decisions are
made on the allocation of values or utilities for the
society. Inasmuch as these decisions have different con-
sequences for the members of the society, individuals vary
in the intensity with which they desire particular outcomes
of political decisions. In a society where resources for
affecting political decisions, or political power, are
broadly distributed, rules provide for the aggregation of
power in order to make collective decisions.

Aggregation of power is carried out in a representa-
tives system by reallocating the political power of indi-‘
viduals to politicians who, in turn, make political decisions.
This is done through elections in which voters choose
representatives from among candidates for representative
offices. Thus, a representative political system makes

allocations of social utilities and, as a derivative of

“r v w WV —.—.—v

 

Jeanne Louise Martin

its decision rules, political power. It does this through
a set of interlocking institutions.

The political power of office-holders in political
institutiéns varies in relation to the political power
reallocated to them by election. Hence, the political
power of a representative varies with the size of his
electorate, those citizens for whom he may make collective
decisions. As electorates increase in size, fewer poli-
ticians can occupy positions of aggregated power. Taken
together, then, the offices in a representative political
system form a hierarchy or opportunity structure of poli-
tical power attendant upon holding office in the system's
institutions. Political parties are organizations of
individuals whose goal is to win political power allocated
in the representative system: in the political system we
describe, there are two political parties.

This political system models five conditions:

1. There are a number of issues to be decided in the
political system.

2. Actors vary in the intensity by which they desire
particular outcomes of different political-
decisions.

3. Resources for affecting the outcome of political
decisions are broadly distributed.

4. There are rules by which political decisions are
made; and these rules are accepted by the
participants.

5. Two political parties attempt to gain political
power.

 

 

Jeanne Louise Martin

In addition, we make two assumptions:
1. Actors are rational.

2. Actors make exchanges of resources affecting poli-
tical decisions. ‘

From this theory in conjunction with decision rules
which tell what resources may be used and how they may be
allocated, we derive hypotheses explaining behavior of
voters and candidates in campaigns and behavior of legis-
lators between elections.

Most of these hypotheses are corroborated by
empirical findings of political research. Furthermore,
these hypotheses parallel most of the hypotheses programmed
into a simulation model by Cleo Cherryholmes and Michael
Shapiro. This simulation model has been fairly successful
in predicting roll—call voting in the U.S. House of Repre-
sentatives.

In the light of the theory, criteria of explanatory
power of simulation models, and considerations of applying
a simulation model in comparative research, we revise the
simulation and test both the Cherryholmes-Shapiro model
and its revision on data describing characteristics of the
U.S. Senate in the Eighty-Eighth Congress (1963-1964).

Although predictive accuracy of neither simulation
model is as great for the Senate as the Cherryholmes-Shapiro
model was for the House of Representatives, many of the
discrepancies reflect patterns of correspondence and non-

correspondence in the model construction itself. Many of

 

Jeanne Louise Martin.

    
 
 
    
  
  
  
  

.. .L‘

rt? é discrepancies between the predictions of the simulation
9lj§nd actual behavior, then, contribute to our confidence in
3 f {the validity of the theory. Furthermore, the revision of
°,Ig.the Cherryholmes-Shapiro model, in addition to being more
A f; parsimonious, more closely related to a theory which explains
‘ it. and more easily applied in comparative research, is
higher in explanatory power than the original model.

Finally, we offer suggestions for revision of the
.model to more adequately assess the correspondence between
theory and reality, and for application of exchange theory

to phenomena other than roll—call voting in American

legislatures.

 

EXCHANGE THEORY AND LEGISLATIVE BEHAVIOR:
A COMPUTER SIMULATION OF ROLL-CALL

VOTING IN THE U.S. SENATE

By

Jeanne Louise Martin

A THESIS

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY
Department of Political Science

2; 1971

     

"Hos/(a

 

 

ACKNOWLEDGEMENTS

Many people took part in the completion of this
project, and I thank them all. In particular, my disserta-
tion committee, Joseph Schlesinger, Timothy Hennessey, and
Cleo Cherryholmes, furnished many ideas at the inception
of the research and useful criticisms for the revision of
the thesis. The chairman of my dissertation committee,
Cleo Cherryholmes, deserves special mention in this regard:
he has been a friend as well as an academic advisor, demon-
strating that academic communities fostering creativity
and human development are possible.

A valuable friend, Elizabeth Powell, assisted me
in the technical and methodological aspects of the research.
Our dialogue is a source of great satisfaction in my con-
tinuing education.

Finally, I offer my deepest thanks to my parents
who, by allowing me freedom and.support, encouraged me to

pursue my interests.

ii

 

TABLE OF CONTENTS

Page

   
  
  
    
  

.ACKNOWLEDGEMENTS . . . . . . . . . . . . . ii
LIST OF TABLES . . . . . . . . . . . . . . v
LIST or FIGURES . . . . . . . . . . . . . vii
LIST OF APPENDICES . . . . . . . . . . . . viii

Chapter

I . INTRODUCTION 0 O O O C O O O O I O I 1

Research on Legislative Behavior . . . . . l
Explanations of Roll-Call Voting . . . . . 5
An Introduction to Retroduction . . . . . 13
Simulation . . . . . . . . . . . . 21
The Setting . . . . . . . . . . . . 26

II. EXCHANGE THEORY: AN APPLICATION TO POLITICS . 27

The Political System . . . . . . . . . 27
Axioms O O O O O O O O O O O O I 2 9
The Election Game . . . . . . . . . . 34
The Legislative Game . . . . . . . . . 51
Summary . . . . . . . . . . . . . 76

FN- III. THE SIMULATION MODEL . . . . . . . . . 34
' Theory-M063 1 o o o I o n o I o o I 8 4
Made l-BehaVior n o o o o o o o o o 10 5
MOdBl'outCOIne o o o o o o o o o o o 113

V‘IV. VALIDATION AND FINDINGS . . . . . . . . 120
J

Validation of Simulations . . . . . . . 120
System Performance . . . . . . . . . 126
Individual Performance . . . . . . . . 137
Conclusion . . . . . . . . . . . . 160

iii

 

     

_ T

. Page ‘

" s. MUSIW I O O C O O O C O O C C U 161 f
l

I

awry o o o o o o o o o o o o o 161
Recommendations ..........164

  
    
 
 
   

M

‘ ~;r~- ‘

LAW

   

i V‘ .

+51 404'} ~. - .-
E . A
: frt3-b.

LIST OF TABLES

i”-_—_.—~.~—

Table Page

1. Comparison of Propositions in the Cherryholmes-
Shapiro Inventory and Deductions from Exchange
Theory 0 I I O O O I I I I I O I I 7 8

2. Motions on Issue of the Federal Role . . . . 114

3. Coding for SIMEX Simulation of Motions on the
Federal Role 0 O I I O I C I O I O I 116

4. Actual and Simulated Yea-Nay Split on Twenty-
Three Federal Role Motions: SIMEX . . . . . 127

5. Actual and Simulated Yea-Nay Split on Twenty-
Three Federal Role Motions: Cherryholmes-
Shapiro Model . . . . . . . . . . . . 128

6. Actual and Simulated Outcome on Twenty-Three
Federal Role Motions: SIMEX . . . . . . . 129

7. Actual and Simulated Outcome on Twenty-Three
Federal Role Motions: Cherryholmes-Shapiro
Model I O O O O O O C 0 I I O I O 130

8. Coalition Size. Difference Between Number
Voting with Winning Coalition and Number
Voting with Losing Coalition: SIMEX . . . . 134

9. Coalition Size. Difference Between Number
Voting with Winning Coalition and Number
I Voting with Losing Coalition: Cherryholmes-
' Shapiro Model . . . . . . . . . . . . 135

10. Proportion Correct for Individuals: SIMEX . . 138

11. Proportion Correct for Individuals:
I Cherryholmes-Shapiro Model . . . . . . . 140

i 12. Proportion Correct for Highly Interested
i Senators: SIMEX . . . . . . . . . . . 144

 

   

Page

Predictive Accuracy with Interaction and
without Interaction: SIMEX . . . . . . . 146

Predictive Accuracy with Interaction and

without Interaction: Cherryholmes-Shapiro

M0661 I I o o I O O I o I o ‘ o I C 148
Effect of Party on Predictive Accuracy: SIMEX . 150

Effect of Party on Predictive Accuracy:
Cherryholmes-Shapiro Model . . . . . . . 152

Effect of ”Memory" on Predictive Accuracy:
Cherryholmes—Shapiro Model . . . . . . . 155

Effect of Executive Position on Predictive
Accuracy: SIMEX...........157

vi

 

’ LIST OF FIGURES

Figure Page
i i 1. Validation Scheme . . . . . . . . . . 25
2. Flowchart: Party . . . . . . . . . . 87
3. Flowchart: Constituency . . . . . . . . 88
4. Flowchart: State and Region . . . .. . . . 92
} 5. Flowchart: Sponsorship and Committees . . . 93

6. Flowchart: Conclusion of the Predisposition
Phase I O I O I I O O I I I I C I 9 4

7. Flowchart: Interaction with Leadership . . . 97

8. Flowchart: Interaction with Members of
Reporting Committee . . . . . . . . . . 98

} 9. Flowchart: Interaction with Members Sharing
Committee Assignments . . . . . . . . . 98

10. Flowchart: Conclusion of Interaction Phase . . 99
i 11. Format: Attributes of Legislators . . . . . 110

12. Format: Attributes of Motions . . . . . . 110

 

vii

   

 

 

 

 

 

CHAPTER I

INTRODUCTION

Research on Legislative Behavior

 

Some decisions on public policy are made in legis-
latures. This decision-making may be studied by focussing
on various aspects of the legislative process, such as
committee decision—making, leadership influence, communica-
tions and roll-call voting. Each focus may be interesting
and may contribute to our understanding of the legislative
process; however, roll—call votes express both the legis—
lator's final decision on the policy in question and the
legislature's collective decision. If one is interested
in the outcome of decisions made in the legislature, then,
the roll-call vote is one dependent variable which research
on legislative decision—making should explain.

Not only is the roll-call fully justifiable as a
phenomenon to be explained, but there is a wealth of fairly
reliable data at hand to use in research. Furthermore,
students of legislative behavior have developed elaborate

techniques for manipulating this data.1 One can identify

 

1An examination of many methods and their properties

may be found in Lee Anderson, Meredith Watts, and Alan
Wllcox, Eegislative Roll-Call Agelysis (Evanston: North-
western University, I966); and Duncan MacRae, Jr., Issues

and Parties in Legislative Votin : Methods of StatistIcal
EPGIXSIs (New York: Harper and Row, 19755.

1

 

 

 

2
agreement and disagreement of legislators by generating
blocs, measuring cohesion and conflict, and scaling issue
postures.2 One can gauge continuity and discontinuity of

public issues by generating factors, dimensions, and scales.3

 

2Classic methods of handling roll-calls originated
with Stuart Rice, "The Behavior of Legislative Groups," Poli-
tical Science Quarterly, 40 (1925), pp. 60-72; Rice, "The
Identification of Blocs in Small Political Bodies," American
Political Science Review, XXI (September, 1927), pp. 619-627;
Herman C. Beyle, Identification and Analysis of Attribute-
Cluster-Blocs (ChIEEgo: University of Chicago, 1931). For
a comparison of measures of cohesion, see Aage R. Clausen,
"The Measurement of Legislative Group Behavior," Midwest
Journal of Political Science, XI (May, 1967), pp. 212—224.

3Scale analysis is appropriate for differentiating
among both individuals and legislative issues: G. M. Belknap,
"A Method for Analyzing Legislative Behavior," Midwest Journal
of Political Science, 2 (1958), pp. 377-402. Scaling may Be
used in conjunction with cluster-bloc analysis: Duncan
MacRae, Jr., "A Method for Identifying Issues and Factions
from Legislative Votes," emerican Political Science Review,
LIX (December, 1965), pp. 909—926. Studies using scales
include: Duncan MacRae, Jr., Dimensions of Con ressional
Voting, Publications in Society and SOCiaI Institutions, No.
3 (Berkeley: University of California, 1958), pp. 203-390;
Charles Andrain, "Senators' Attitudes Toward Civil Rights,"
Western Political Quarterly, 17 (1964), pp. 488— 503, Paul
Dempsey, "Liberalism-Conservatism and Party Loyalty in the
U. S. Senate," Journal of Social Psychology, 56 (1962), pp.
159-170; Charles H. Gray, "A Scale Analysis of the Voting
Records of Senators Kennedy, Johnson and Goldwater, 1957-
1960," American Political Science Review, LIX (September,
1965), pp. 615- 621; Factor analyses include Stephen Cimbala,
"Foreign Policy as an Issue Area: A Roll- Call Analysis,"
American Political Science Review, LXIII (March, 1969), pp.
II§-156; John G. Grumm,’“A Factor Analysis of Legislative
Behavior," Midwest Journel of Political Science, VII (November,
1963), pp. 336-356; Gerald Marwéll,’“Party, Region and the
Dimensions of Conflict in the House of Representatives, 1949—
1954," American Political Science Review, LXI (June, 1967),
pp. 380-399. DimenSIOnal analysis is utilized in a compara-
tive manner by Aage Clausen and Richard Cheney, "A Compara-
tive Analysis of Senate-House Voting on Economic and Welfare
Policy 1953-1964," American Political Science Review, LXIV
(March, 1970), pp. I‘fi-isz. For exploring continuity over
time see, Aage R. Clausen, "Measurement Identity inc the
Longitudinal Analysis of Legislative Voting," Amer
Political Science Review, LXI (December, 1967), pp. CIO 020- 1035.

 

 

 

 

A

.. ~—‘ -w... A-

 

 

Seldom, however, has the roll-call vote been used
as a dependent variable.4 Patterns of cohesion have been
described by the common characteristics of cohesive indi-
viduals,5 and roll-calls have been identified by their
underlying commonality of content and by the characteristics
of legislators who support and oppose them.6 As a result,
we have many hypotheses about the determinants of roll-call
votes and about differences across issues, but rarely have
these hypotheses been tested on new data.

Furthermore, each roll-call presents the legislator
with a number of conditions which Vary from motion to
motion. These differences have often been obscured by
lumping together large numbers of roll-calls. Roll-call
votes have been aggregated into sessions, parties and issue

areas, and disaggregated into blocs and policy dimensions.

 

4Notable exceptions to this generalization are the
more formal treatments, mathematical, statistical, and simula-
tion models of roll call voting which have been tested against
data; Cleo Cherryholmes and Michael Shapiro, Re resentatives
and Roll Calls (New York: Bobbs-Merrill, 1969); DonaId
Matthews and James Stimson, "Decision-Making by U.S. Repre-
sentatives: A Preliminary Model," 1968 (Mimeo.); David B.
Meltz, "Competition and Cohesion: A Model of Majority Party
Legislative Bargaining" (unpublished Ph.D. dissertation,
University of Rochester, 1970); Warren E. Miller and Donald
E. Stokes, "Constituency Influence in Congress," American
Political Science Review, LVII (1963), pp. 45-56.

5David Truman, The Congressional Party (New York:
John Wiley, 1959).

6Andrain, "Senators' Attitudes Towards Civil Rights;'
Cimbala, ”Foreign Policy as an Issue Area;" Clausen and
Cheney, "A Comparative Analysis of Senate-House Voting;"
Grumm, ”A Factor Analysis of Legislative Behavior;“ Marwell,
”Party, Region, and the Dimensions of Conflict in the House
of Representatives." ‘

 

kw"

4

In the glut of hypotheses about correlates of individual
and group behavior, such as party, constituency and region,
we find few insights into how or why the individual legis-
lator reconciles these influences. Moreover, observers
infer interpersonal influence and group processes from
patterns of cohesion and conflict,7 and legislatures have
been described as social systems.8 Few hypotheses, however,
attempt to explain how we may link interaction to roll-call
votes independently of, or in conjunction with other influ-
ences on legislative behavior.

Unless we account for both the individual's decision-
making process and the effect of group decision-making, we
cannot explain roll-call votes.

We propose to approach explanation of roll-call votes
as decisions made by individuals in the context of an ongoing
collective decision—making process. We will develop a theory
and a method appropriate for validating it which are useful

in comparative research.

 

7Truman, The Congressional Party.

8Raymond A. Bauer, Ithiel deSola Pool, and Lewis A.
Dexter, American Business and Public Polic (New York:
Atherton, 1964); Wayne L. FranCis, "Influence and Inter-
action in a State Legislative Body," American Political
Science Review, LVI (December, 1962): Pp. 953-960; Samuel
C. Patterson, "Patterns of Interpersonal Relations in a
State Legislative Group: The Wisconsin Assembly," Public

92inion Quarterly, XXIII (Spring, 1959), pp. 101-10 .

 

'0 7‘

In the next section we shall examine two attempts
to explain individual roll-call votes for their adequacy
and for the insights they bring to the nature of scientific

explanation.

Explanations of Roll-Call Voting

There are two types of explanation: the pattern
type or factor theory fits what is to be explained into a
set of known relationships. That is, a series of factors‘
or influences all partially explaining one behavioral ‘
phenomenon are postulated. This set of propositions
embodies a pattern of relationships which converge on
the observation to be explained- The deductive type or
hierarchical theory shows that the observation can be
deduced from more general considerations. That is, from
a set of assumptions, or lawlike generalizations, we derive
hypotheses which predict the observed outcome.9

In general, the deductive model of explanation is
to be preferred over the pattern model. First, it is gener-
alizable to any system for which its assumptions are
plausible; frequently the pattern model is bound to a
specific situation by it component propositions. Further-

more, the deductive explanation circumscribes the realm of

 

9Abraham Kaplan, The Conduct of In ui (San
Francisco: Chandler, 196 , pp. 327-3 6; Quentin Gibson,

The Lo ic of Social En ui (London: Routledge and
Kegan Paul, 1960).

 

 

admissible evidence; the pattern model may admit any grounds
for explanation. Finally, if predictions are logical conse-
quences of a set of assumptions, errors in prediction must
stem from those assumptions; the inability of a factor
theory to explain may stem from a mistaken conponent
hypothesis or from a blank or hole in the pattern.

We have examples of both types of explanation for
legislative voting. One, developed and tested by Donald
Matthews and James Stimson, purports to be a deductive
explanation; the other, a.model developed and tested by
Cleo Cherryholmes and Michael Shapiro, is a pattern explana-
tion. Both have been tested by computer simulation on data

from the United States House of Representatives.10

The Matthews-Stimson Model

Matthews and Stimson assume that representatives
are rational or goal-directed in their actions. They
maintain that rational behavior is "general (like cases
are handled alike), regular (reasonably stable over time),
and consistent (decisions on different issues are logically

"11

compatible). They also assume that the legislator is

faced by a multiplicity of complex issues, the impact of

 

10Matthews and Stimson, "Decision-Making by U.S.
Representatives;" Cherryholmes and Shapiro, Representatives
and Roll-Calls.

11Matthews and Stimson, "Decision-Making by U.S.
Representatives," p. 4.

which (relative to his goals) he is uncertain. Matthews
and Stimson argue:

. . . when a member is confronted with the necessity

of casting a roll-call vote on a complex issue about

which he knows very little, he searches for cues pro-'

vided by trusted colleagues who——because of their

formal position in the legislature or policy special—

ization—-have more information than he does and with

whom he would probably agree if he had the time and

information to make an independent decision.12

To operationalize their theory, Matthews and Stimson
designate nine possible cue-givers: the state party delega-
tion, party leadership, the President, the chairman of the
committee reporting the bill, the ranking minority member
of the committee reporting the bill, the "conservative
coalition" (two-thirds of the Southern Democrats voting
the same way), the leadership of the Democratic Study Group,
the party majority, and the House majority (two-thirds of
the House voting the same way).13
From roll-calls in the first half of a congressional

session, they calculate for each member and each cue-giver
the ratio of the number oftimesthe member voted with the
cue-giver to the number of times the one was available and
the member was voting.14 Since a ratio of one-half indicates

chance agreement, deviations from chance are interpreted

as propensities to vote with or against the cue-giver. The

 

121bid., p. 11.

13Ibid., pp. 14-16.

F 14Ibid., p. 16.

 

 

(cue-givers for each legislator are arranged in the order of
the deviation from chance and the propensity to agree is
adjusted in direct relationship to its deviation from .5
to new values between +.5 and -.5.

On each roll-call simulated in the model, the
t observed outcome is analyzed to determine which cues are
available and whether the cuesgiver voted yea or nay.
Then, for each legislator, the model multiplies the legis-
lator's propensity to agree with his highest available
' cue-giver by the direction of the cue (+1 for yea, —l for
) nay). If the product is +.5 or -.5, the legislator votes
yea or nay, respectively. If the product is not +.5 or
-.5, the process is repeated for the legislator's second
highest available cue-giver. The criterion for voting is

whether the sum of the first and second products is as

 

I great as +.5 or —.5. If it is not, the third highest
available cue-giver is processed and algebraic sign of

’ the sum of the three products determines the legislator's

vote. \

In spite of the restrictions of the assumptions
and the simplified,.somewhat arbitrary, selection of cue-
givers, Matthews and Stimson report a high mean level of
predictive accuracy (85.7% to 89.3%).15

The Matthews and Stimson model does not, however,

. Provide us with an explanation of the decision-making

 

 

lSIbid., p. 25.

 

 

 

process because it uses roll—call votes exclusively as
independent variables, including the roll-call vote to
be explained! The simulation operates within a virtually
tautological system of relationships. While it generates
outcomes which substantially match observed behavior, it
does so by gathering information from that very data and
not from independent observations.
Joseph Hanna distinguishes between the descriptive

and explanatory power of models:

In brief, predictive or explanatory power is identi-

fied with information transmitted by empirical

factors which can be determined prior to the data,

while descriptive power is identified with total

transmitted information, including information trans-

mitted by the data.16
We sometimes take information.which enables us to predict
behavior from observation of that behavior itself, either
through parameter estimation or by direct use of measures
of that behavior as independent variables...Whenever we do
this, our ability to predict stems from the descriptive
power or our model. On the other hand, information which
enables us to predict behavior determined prior to measure-
ment of the dependent variable, such as the selection of

variables and the direction of influences, is explanatory

power.

 

16Joseph F. Hanna, "Explanation, Prediction,
Description, and Information Theory,"'Synthese, 20
(October, 1969), pp. 308—309.

 

 

10

The Matthews-Stimson model takes all its information
except the designation of possible cue-givers from the data
to be explained. Its ability to predict, then, stems almost

entirely from its descriptive power.

ThegCherryholmes-Shapiropygdgl

The Cherryholmes-Shapiro simulation is based on a
pattern model of explanation. Their model results from an
extensive review of empirical research on legislative
behavior and attitude consistency and change. This model
relates a number of explanatory variables to legislative
behavior, particularly roll—call voting. It also describes
the norms and communication patterns of the legislature as
a social system.

Sixty-eight propositions are generated. Many of
the propositions are explicitly programmed into the computer
model as relationships between the attributes of representa-
tives and the attributes of the roll-call to be explained,
exclusive of the outcome of the vote.

There are two phases in the Cherryholmes-Shapiro
model. First, all representatives are processed serially
through the predisposition phase in which the effects of
party sponsorship, regional, state, and constituency
interests, and public position (individual sponsorship and
agreement with committee majority and minority reports)
are computed for each congressmen. The first phase is

deterministic, that is, with a given set of conditions,

 

 

11

only one outcome is possible. If these factors generate a
strong predisposition for the member of the legislature,
his vote is determined at this point. '

If his predisposition is weak, the legislator enters
the stochastic communication phase of the model where, with
varying probabilities, he may encounter every other member's
predispositions and the.position of the President. The
probabilities are computed by adding probabilities associated
with the other actor's characteristics relative to the
characteristics of the legislator whose vote is being pre—
dicted. The computer generates a random number between
zero and one: if it is greater than the computed proba—
bility of interaction, they do not interact; if it is less
than the probability of interaction, they do interact. The
larger the probability, then, the greater the likelihood a
random number will be less than the Specified probability
of interaction. The outcome of this phase, then, depends
partially on chance.

The congressman's final disposition to vote is one
half the sum of his own predisposition and the average of
all the predispositions of those he encountered in the
communications phase. His vote is the sign of the resulting
disposition, yea if positive, nay if negative.

This model predicts at a rate nearly equal to the
Matthews-Stimson model (84%). Moreover, it is superior as

an explanatory instrument to the Matthews-Stimson model:

 

 

12

most important, it does not use votes on the same motion
to predict a roll—call. As we have noted, the explanatory
power of a model is that which stems from its theoretical
base, the information it draws from the environment prior
to observation of the behavior to be predicted. The
Cherryholmes-Shapiro model, then, while not deductive, is
higher in explanatory power than the Matthews-Stimson
model. The variables and direction of influence are
determined largely by propositions included in their
inventory.

Furthermore, the Cherryholmes-Shapiro model fits
more comfortably with notions of the influences of party,
constituency, and communication within the legislature
which have structured other analyses of legislative behavior.

There are several problems associated with the
Cherryholmes—Shapiro model. One is that the probabilities
added in the communication phase may not be independent:
neither autocorrelation nor interaction of effects is
explored. Secondly, the model uses past voting on similar
legislation, the "memory score," as an input of the model;
Cherryholmes and Shapiro interpret this as personal ideology
and ideological consistency relative to the President's
program. Although the "memory score" does not contribute
much to the prediction level of the model, it would be
preferable to exclude from the model direct measures of

roll-call voting. Finally, although the Cherryholmes-Shapiro

 

 

13

model is a reasonably good pattern explanation, we would

prefer to have a deductive model.

An Introduction to Rgtroductiqn

Abraham Kaplan suggests that the difference between
the two types of explanation is primarily one of maturity:
"The pattern model may more easily fit explanations in
early stages of inquiry, and the deductive model explana-
tions in later stages."17 In other words, sometimes a
pattern model of explanation can be mapped onto a deductive
model. Progress in scientific explanation is gained by
arrival at a hierarchical theory, one which is articulated
as a deductive model of explanation.

. . . we can subsume pattern explanations under the
deductive model. Fitting something into a pattern has
explanatory force insofar as thereby we are enabled

to show how what is being explained can be deduced
from more general considerations.

Furthermore, Norwood Hanson maintains that this
progression from data to theory has a logic of its own,
retroduction. He contrasts this form of inference with
induction and deduction (hypothetico-deductive, or H-D

accounts):

The form of the inference is this:

1. Some surprising phenomenon P is observed.

2. P would be explicable as a matter of course
if H were true.

3.. Hence there is reason to think that H is true.

 

17Kaplan,'The Conduct of Inquiry, p. 332.

1'81‘b'id., p. 338.

 

 

14

H cannot be retroductively inferred until its
content is present in 2. Inductive accounts expect
H to emerge from repetitions of P. H-D accounts
make P emerge from some unaccounted-for creation of
H as a 'higher—level hypothesis.'19

We do not, then, arrive at hierarchical theories
from a priori knowledge of general laws but from what seems
plausible in the light of our knowledge about what is to
be explained.

The Cherryholmes—Shapiro model suggests that a theory
of legislative behavior should explain two things: the
generation of predispositions and.the patterns of inter-
action in a legislature. 70ther students of legislative

behavior have also distinguished between the "outside model"

and the "inside model" of explanation,20 the "concerned"

21 22

and the'flefectors," the committed and the uncommitted.
Political research in the recent past has also pursued
the notion that bargaining is central to politics, especially

legislative behavior. Gordon Black says,

 

19Norwood Hanson, Patterns of Discovery (Cambridge:
University Press, 1958), p. 86.

20Heinz Eulau and Katherine Hinckley, "Legislative
Institutions and Processes," in James Robinson, ed.,
Political Science Annual, 1966 (Indianapolis: Bobbs-
Merri , 66 .

21Meltz, “Competition and Cohesion."

, 22Donald Matthews, U.S. Senators and Their World
(New York: Random House, 1960).

 

15

Explicit bargaining pervades the entirety of the
political process, from the development of electoral
alliances to the formation of legislative coalitions,
and at the center of these processes stands the
politician.23

Riker notes that rules of the political "game" have a

crucial effect as they structure the bargaining process:
This fact should inspire political scientists to the
continuation of the most important of their traditional
functions, which is the explication of the effect of
particular rules in occasioning deviation from abstractly
expected strategies and solutions.

There is a theory which accounts for the generation
of both the "inside" and the "outside" influences on legis-
lative behavior as well as bargaining and the rules of the
game. Furthermore, this theory is in the growing tradition

of the application of economic analogues to politics.

Economic Analogies
Analogues based in economic theory have been applied

to interest groups,25 political parties and elections,26

 

23Gordon Black, "A Theory of Professionalization in
Politics," American Political Science Review, LXIV (September,
1970), p. 865.

24William Riker, "Bargaining in a Three-Person
Game," American Political Science Review, LXI (September,
1967), p. 656.

25Mancur Olson, Tee Logic of Colleptive Action (New
York: Schocken Books, 1968); Robert H. Salisbury,“ An
Exchange Theory of Interest Groups,"‘Midwest Journal of
Political Science, XIII (February, 19 9 , pp. -3 .

26Anthony Downs, An Economic Theo of Democra (New
York: Harper and Row, 1957); Duncan Black, The Theogy of
Committees and Elections (Cambridge: UniverSity Press,
9 .

 

16

7

constitution—making,2 and the making of public policy.28

Sociologists have also drawn parallels between social

interaction and economic transactions.29
James S. Coleman has been the chief proponent of

exchange theory in sociology: he has explored the model

he suggests through formal analysis and experimental games.3o

Coleman's experiments are particularly suggestive of a use

for exchange theory to explain legislative behavior, since

he simulates a legislature as an example of collective

social decisions.

The Legislative Game

In Coleman's legislative game, the objective of the
players is to be reelected in designated districts by as
large a majority as possible. Districts are represented
by a number of precincts which are identified by their
preferences on legislation. The player is reelected if

the legislation preferred by his constituents is passed.

 

27James M. Buchanan and Gordon Tullock, The Calculus
of Consent (Ann Arbor: University of Michigan, 1962 .

28R. L. Curry and L. L. Wade, A Theo of Political
Exchange (Englewood Cliffs, N.J.: Prentice-Hall, 1968).

2gGeorge C. Homans, "Social Behavior as Exchange,"
American Journal of Sociology, 63 (1959), pp. 597-606.

- 30James S. Coleman, "Collective Decisions,"

'Sociological Inguiry, XXXIII (Winter, 1963), pp. 166-181;

Foundations for a Theory of Collective Decisions,"

'American Journal of Sociology, LXXI (May, 1966). Pp. 615—

; Games as Vehicles for Social Theory," American
Behavioral Scientist, XII (July-August, 1969), PP. 2-6;
"The PossiBiIity of a Social Welfare Function," American
Economic Review, 56 (December, 1966), pp. 1105—1I22.

 

 

 

17

Each player has one vote on each measure which is presented.
Not all issues, however, have consequences for the legis-
lator's chances of being reelected. Coleman assumes that
each player is rational, that is, "wholly concerned with
pursuit of his own interest."31
Coleman predicts,
Faced with a situation of lack of power over actions
which interest him, together with a surplus of power
over actions which interest him little or not at 2
all, the rational man will make an exchange of power.
This prediction is born out in the playing of the legis-
lative game. Coleman finds,
Typically, in the playing of this game, the principal
behavior of the players is exchange of votes: each
player giving up votes on issues which are of little
interest to him for votes on those which are of much
interest. Other types of exchange exist also, involving
the subsidiary types of powers held by each of the
actors over the collective decisions: the order in
which issues come to the floor, and the vote on tabling
an issue. 3
Other observations on the playing of the game
indicate that this process is similar to legislative
behavior. For example "pork barrel" legislation which
affects only one constituency elicits more vote trading
than other types because the constituency is intensely
interested in the outcome while opposing interests are

widespread.-

 

31Coleman, "Collective Decisions," p. 168.

321bid., p. 172.

33Ibid., p. 173.

 

 

18

Furthermore, norms developed over time, particularly
the norm of reciprocity. In early sessions of the game
defection from bargains occurred frequently. As the sessions
went on, however, defectors responded to sanctions and
demonstrated themselves trustworthy.

Certain actors gained reputations of being unreliable,
with the consequence that others made agreements with
them less frequently. As the sessions continued, 34
however, all actors became quite trustworthy . . .

Coleman also reports variations in the game's
structure of decision rules. Of particular interest rela-
tive to the correspondence between the game and American
legislatures is the addition of committees.to the decision-
making process.

. . . this enriched greatly the amount and kinds of
the resources of the collectivity members . . .

Much of the bargaining, negotiation, and exchange
was now directed to obtaining a positive action in
the committee . . . ; the smaller size of the com-
mittees . . . made the committee action much more
dependent upon specific individuals and thus concen-

trated the control of particular actions much more
in the hands of a few people.

Exchange Theory

We will use Coleman's work as the basis.of a theory
of legislative decision—making. .The game structure models

several general conditions or givens of this theory. These

 

are:

34Ibid., p. 176.

35Coleman, "Games as Vehicles for Social Theory,“
9. 6.

 

19

1. There are a number of issues to be decided
collectively.

2. Actors vary in the intensity by which they
desire particular outcomes of different col—
lective decisions.

3. Resources for affecting the outcome of col-
lective decisions are widely distributed.

4. There are rules by which collective decisions
are made; and these rules are accepted by the
participants.

Coleman assumes one axiom. underlies the behavior

of actors in this situation:

1. The actors are rational; that is, they pursue
their own interests as defined by the game.

Coleman's experiments confirm his prediction which becomes
our second axiom explaining the dynamics of collective
decision-making:

2. Actors engage in exchange of resources.
Specifically, Coleman finds the rational actor increases
"control over a decision for which control makes a great
deal of difference in his expected utility, . . . in return
giving up to another person control over a decision for
which control makes little difference in his expected

"36

utility. (Assumptions 1 and 2)

 

36Coleman, "The Possibility of a Social Welfare
Function," p. 1113.

uh vvu—w—

 

20
Application of Exchange Theory to
Legis ative BehaVior

As Coleman has suggested, we can apply exchange
theory to collective decision-making in elective legis—
latures. These are situations in which a number of col-
lective decisions are made by specific decision rules and
resources are broadly distributed. Moreover, legislatures
and elections which select legislative actors are inter-
locking institutions, themselves part of the decision rules
for making certain collective decisions for the society.

We will apply exchange theory to the decisions made in
elective legislatures by considering the implications
elections have for the legislator's utility relative to
outcomes of collective decisions.

In order to claim that this theory explains roll-
call voting, we will do three things. First, we will specify
the theory, i.e. define the assumptions and conditions in
such a way that derivation of hypotheses is possible.
Secondly, we must operationalize a model of derived
hypotheses which can.generate predictions. Finally, we
must test the model against empirical data.

We, like others who have attempted explanations
of roll—call votes as individual.and.group decisions, will
utilize computer simulation to test this theory. This is,
in part, a practical matter: we can in this way more easily

compare the adequacy of our explanation with others' along

-'3

 

n “

21

various criteria. The choice of method, moreover, is

appropriate for several reasons.

Simulation

A simulation is an operating model of the process
to be investigated which generates consequences or outcomes.
When we construct a computer simulation, we make statements
about relationships among variables in computer language.
Since a computer uses a set of logical rules to interpret
the set of statements, or program, the program must be
internally consistent to generate outcomes. The relation-
ships stated in the program thus embody a logically con-
sistent theory. The output of a computer simulation is
the set of dependent variables we wish to explain; the
input, the values associated with variables which the
program manipulates to generate consequences, are our
independent variables.

Simulations are controlled experiments. In this
respect, they are superior to most other types of data
manipulation. The experimenter, the simulation designer,
is assured that his predictions result from his independent
variables and the process by which the simulation manipulates
them. Furthermore, he can change aspects of the simulated

environment to compare the results of different model

‘instanCes, the same process operating with different

 

.initial conditions.

 

22

Simulations, as Raymond Boudon explains, are no
different from other formal models in their content, but
the medium allows us to observe the consequences of the
model at a variety of points.

Among a system of verbal propositions and a simulation
model the relationships are of the same type. In any
case, we manipulate these propositions in such a
manner that it is possible to derive from them new
consequences. One can obtain a simulation model from
a verbal theory by a sequence of specifications,
simplifications, and, eventually, reductions. The
difference is that the consequences of a simulation
model are obtained, not by deduction, but by the
observation of an artificial system which corresponds
to a physical realization of the model.37

Simulation is appropriate when we are studying a
system and the behavior of parts of the system, such as
collective and individual behavior, simultaneously. Through
simulation we can investigate the behavior of collectivities
under varying conditions. Furthermore, we can study the
behavior of members of the collectivity, each of whom may
exhibit a unique but explicable behavior under varying
conditions. In such cases, Coleman says, our aim is

. . . to program into the computer certain theoretical
processes, and then to see what kind of behavior system
they generate. The aim is to put together certain
processes at the individual and interpersonal level--

and then to see what consequences they have at the
level of the larger system.3

 

37Raymond Boudon, "Reflexions sur la logique des
modeles simules," Archives Euro eenes de Sociolo ie, VI,
I (1965), p. 4. (Author‘s transIation)

38James S. Coleman, "Analysis of Social Structures

and Simulation of Social Processes with Electronic Computers,"

Educational and Psychological Measurement, 21 (1961), p. 216.

A

V‘w"

 

23

Just as there are two different types of explanations,
we may construct simulations from two different types of
theories. We may test a hierarchical theory by a simulation
as a substitute for a mathematical model. However, Boudon
observes,

. . . one also encounters examples where the simulator
abandons parsimony of hypotheses characteristic of
mathematical models so as to produce an artificial
system which incorporates all known propositions about
the real system which he wishes to study.39
The Cherryholmes-Shapiro model is of this type. As we move
from a pattern model of explanation to a deductive model,
we shall introduce changes in the simulation to make it
reflect more accurately the predictions of the theory it
is intended to represent. In.making.these-changes, we may
sacrifice some predictive accuracy since the theory will
circumscribe our range of admissible evidence. However,
we hope to gain the advantages of a deductive explanation,
that is parsimony, generalizability, and the ability to
isolate sources of errors.

We validate simulations and, derivatively, the
theories they represent by comparing the output of the
simulation with empirical observations. The simulation
model, however, is not identical to the verbal theory it

represents. Nico Fridja admonishes,

 

39Boudon, "Reflexions sur la logique des modeles
simules,“ p. 4.

 

 

 

 

1.7“. -

24

According to some formulations in the literature, a

program is a theory. This seems an incorrect way of

putting things. Rather, a program represents a

theory. It does this with the help of a number of

mechanisms which are irrelevant to the theor or

which the theory might explicitly disclaim.4

Abraham Kaplan, too, warns us of the "danger in

the use of models . . . map reading: the failure to
realize that the model is a particular mode of representa-
tion, so that not all its features correspond to some

characteristic of its subject matter."41

Theories may
have leei e; ignorance, aspects of the referent process
which are neglected from necessity or convenience in
formulating the model.42
For example, frequently stochastic processes will
be included in a simulation program, as Cherryholmes and
Shapiro have done, to circumvent a locus of ignorance.
we should investigate the consequences of this and other
approximations for the performance of models and their
implications in terms of theory.
Perhaps more important from the standpoint of

validating the theory, however, are aspects of the theory

which are not included in the model. With the addition

 

4oNico H. Fridja, “Problems of Computer Simulation,"
Behavioral Science, 12, l (1967), p. 60.

41Kaplan, The Conduct of Inguiry, p. 284.

42Joseph Hanna, "Information-Theoretic Techniques
for Evaluating Simulation Models," Michigan State University,
Department of Philosophy, n.d. (Ditto.).

 

 

25

of the simulation model, our problems of validating the
theory relative to the behavior we wish to explain are

complicated.

THEORY (—~—) MODEL

BEHAVIOR F————> OUTPUT
Figure l.—-Va1idation Scheme.

Since we validate the theory by the relationship
of the output of the model to the behavior, the relationship
of the theory to the model is crucial. If the model is not
an exact reproduction of the theory, we can expect that the
output will be distorted relative to observed behavior.

We can predict the patterns of correspondence and
non—correspondence in the output relative to the observa-
tions from our knowledge of the aspects of the theory which
are absent or roughly approximated in the model. If these
predictions are born out, this renders our model, hence
our theory, more valid.43

The difficulty of evaluating the output of a
simulation model is compounded by the inapplicability of

statistics in most cases. Fridja suggests, "The construction

 

43Paul Smoker, "Social Research for Social Antici-
pation," American Behavioral Scientist, XII (July-August,
1969), pp. 7-I3.

26

of alternative models seems the only control, and a necessary
one at that."44
In the validation of the simulation presented in
this project we shall pursue both routes: we shall predict
patterns of correspondence and compare output and behavior
along these dimensions. We will also run alternative
models: in our case we cannot only run several instances
of one model, testing subhypotheses, but we can also compare
the output of the Cherryholmes-Shapiro formulation with

SIMEX, the revised model we shall develop in the light of

our application of exchange theory.

The Setting

Cherryholmes and Shapiro tested their model on
foreign affairs and federal role issues voted on in the
U.S. House of Representatives in the Eighty-Eighth Congress
(1963-64). We shall employ data describing the attributes
and behavior of U.S. Senators as they voted on twenty-three
roll-calls on motions having to do with the issue of the
nature and extent of the federal role in domestic affairs
in the Eighty—Eighth Congress. In this way we will hold
constant situational factors such as the level of government,
the President, and the issues brought before the legislature,
while adding perspective on the differences between the two

chambers.

 

44Fridja, "Problems of Computer Simulation."

CHAPTER II

EXCHANGE THEORY: AN APPLICATION TO POLITICS

The Political System

 

Let us define a situation which meets the structural
conditions of exchange theory. In a political system a
number of decisions are made on the allocation of values
or utilities for the society. Inasmuch as these decisions
have different consequences for the members of the society,
individuals vary in the intensity with which they desire
particular outcomes of political decisions. In a society
where resources for affecting political decisions, or politi-
cal power, are broadly distributed, rules provide for the
aggregation of power in order to make collective decisions.1

In a representative system, aggregation of power is
carried out by reallocating the political power of individuals
to politicians who, in turn, make political decisions. This
is done through elections in which voters choose representa-
tives from among candidates for representative offices.

Thus, a representative political system makes

allocations of social utilities and, as a derivative of

 

1For a theoretical exposition of rational strategy
of the citizen in formulating decision rules, or constitu-
tions, see Buchanan and Tullock, The Calculus of Consent.

 

27

28

its decision rules, political power. It does this through
a set of interlocking institutions. Occasionally, we shall
refer to these institutions as games, since they have
rules, participants, and outcomes, rewards for winning the
game and penalties for losing it.

The political power of office-holders in these
institutions varies in relation to the political power
reallocated to them by election. Hence, the political
power of representative varies with the size of his elec-
torate, those citizens for whom he may make collective
decisions. When a representative shares control over
decisions with other representatives, a situation we find
in legislative chambers, the power inherent in his office
is proportional to the average size of the electorates
represented.2 As electorates increase in size, fewer
politicians can occupy positions of aggregated power.
Taken together, then, the offices in a representative
political system form a hierarchy or opportunity structure
of political power attendant upon holding office in the

system's institutions.3

 

‘r

2Political power at one level of aggregated power may
vary with other attributes of office, such as the scope of
decisions made (the number and impact of utilities allocated),
their finality (whether they can be revoked by other decisions),
and the length of tenure of the office (how long aggregated
power may be held by the politician between elections).

3For further elaboration and research on the existence
of an opportunity structure in the United States, see Joseph A.
Schlesinger, Ambition and Politics (Chicago: Rand McNally,
1966).

 

29

Political parties are organizations of individuals
whose goal is to win political power allocated in the
representative system. Let us further specify that in

the political system to which we will apply exchange

theory:
Two political parties attempt to gain political
power.
Axioms
Rationality

 

To avoid a tautological definition of rationality
which would declare all action rational on the part of the
actors, we define rationality in terms of the situation or
game in which collective decisions are made. We assume
that: (1) Actors intend to win whatever game or games
within the collective decision-making process they are
playing, i.e. they are "intendedly rational." (2) Actors
will utilize their resources in a manner that they believe
will enable them to win the game, i.e. they are "strategically

rational."4

 

4This breakdown of rationality was suggested by
Paul H. Conn, David B. Meltz, and Charles Press, "Rationality,
Effectiveness, and Political Calculations," Michigan State
University, August, 1970 (Mimeographed). Strategic ration-
ality is utilized in Arthur S. Goldberg, "Social Determinism
and Rationality as Bases of Party Identification, American
Political Science Review, LXIII (March, 1969), pp. 5-25.

30

Since the political system allocates both social
utility and political power there are two kinds of rational
intentions. The citizen's rational goal is to gain the
highest benefit from collective decisions on the allocation
of social utilities. The rational goal of the politician
is to gain political power. The politician, then, is
motivated by political ambition.

This does not mean that the politician does not
have personal preferences about the outcome of collective
decisions, but that these utilities are lower in priority
relative to his political ambition.5

Because of its dependence upon strategic choices,
the rationality assumption renders theories of which it
is a part difficult to utilize in predictions. Rationality
does not necessarily imply effectiveness. Goldberg says,

. . . being rational in a decision situation consists
in examining the alternatives with which one is con-

fronted, estimating and evaluating the likely conse-

quences of each, and selecting that alternative which
yields the most attractive set of expectations.

However, there is room for error in these calcula-

tions. Individuals may, therefore, vary in their
effective rationality.6

 

 

5The thesis of Schlesinger, in Ambition and Politics,
is that ambition is central to the explanation of the behavior
of politicians. For research using ambition as a central
concept, see Gordon S. Black, "A Theory of Professionaliza-
tion in Politics," American Political Science Review, LXIV
(September, 1970), pp. 865-878; Michael L. Mezey, "Ambition
Theory and the Office of Congressmen," geurnal of Politiee,
32 (August, 1970), pp. 563-579; John W. SouleTvflFuture Poli-
tical Ambitions and the Behavior of Incumbent State Legis-
lators," Midwest Journal of Political Science, XIII (August,

1969), pp. 439-2540

6Goldberg, "Social Determinism and Rationality,"
p. 5 (emphasis added).

31

As Quentin Gibson points out, rational (effective)
action is composed of both a subjective element, the
actor's belief in his efficacy, and an objective element,
the foundation of this conviction.7

Gibson maintains the inclusion of the rationality
assumption precludes derivation of a closed system of laws;
rationality requires correction and modification of predic-
tions at each step of the deductive system in the light of
the influences we know to work counter to effectiveness.

On the other hand, this quality of rationality invites the
use of logic and evidential argument by the theorist. The
theorist's ability to argue the efficacy of an action

toward specified ends is one of the most attractive aspects

of the rationality assumption.8

Exchange

Although political exchanges invite an analogy to
the economic marketplace, this resemblance should not be
taken literally. In several crucial ways, they are dis-
similar. In the political marketplace, the exchange process
takes place over time. This introduces a contingency
problem. Each actor operates under some uncertainty that

his bargains will be fulfilled. Thus, exchanges promising

 

v

7Gibson, The Logic of Social Enquiry, p. 156.
8

 

Ibid., pp. 164-166.

32

certain gains may take precedence over more profitable, but

 

risky, alternatives.

Furthermore, rules sometimes anchor decisions in
time so that the most rational bargains may not be com-
pleted. The realization of bargains structured by time
may require that one party or the other consent to an
exchange that falls short of the return he might expect:
under unlimited bargaining. 'Z, 9““j““%¢°70 én764$fl

Social marketplaces also deviate from the economic
marketplace in the flexibility of transactions. In the
economic marketplace, presumably the seller of a good does
not care what the buyer of that good does with the exchanged
value. In political exchanges, frequently the seller of
control over a decision continues to have a stake in the
use of the political resources. Consequently, preferences
among alternative low-priority decisions influence the.
choice of partners to an exchange.9

Furthermore, the exchange currency of political
transactions is frequently not divisible}0 Curry and
Wade make a persuasive argument supporting the thesis that

11

all policy is divisible. Policy may be among the more

 

9For further discussion of this point, see
Coleman, "The Possibility of a Social Welfare Function."

10For further discussion of this and other limita-
tions of social exchange, see Black, The Theogy of
Committees and Elections.

11Curry and Wade, A Theory of Political Exchange:
Economic Reasoningin Political Analysis.

33

divisible of political resources, but this generalization
has its limitations. One of the most common resources for
political actors, the vote, is not divisible among options.
The rational actor will exchange resources, including
control over decisions slightly affecting his expected
utility, for control over decisions which greatly affect
his expected utility. When there is.a choice of alternative
possible exchanges, he will make the one which brings him
the greatest expected utility and.which.costs him the least
in terms of resources. His costs include the opportunity.
foregone to use his resources in other ways. That is, the
rational actor maximizes profits, or benefits-minus-costs,

including Opportunity costs.

 

Furthermore, since all actors attempt to maximize
gain in transactions, parties to an exchange may not
immediately agree on the terms of the bargain. The
resources of time and investment of effort expended in
negotiating decisions are bargaining costs; these expendi-
tures will also be minimized by the rational actor.

Since knowledge of the outcome of different options
conditions the ability of the actor to choose the correct

strategy, he must take into account information costs, the

 

resources expended in gathering information on alternative
courses of action. Information-gathering may be inhibited
in two ways. One encumbrance to information-gathering is

the availability of resources to invest. These include

34

money, energy, and time. The second inhibition is the
diminishing marginal return from further investment in
information. Each actor deems a finite amount of informa-
tion sufficient to make a strategically rational choice.
An actor will engage in a limited, non-random search
procedure designed to give him sufficient information at
the least cost. His search procedure will be based on

the likelihood that, given the structure of the situation,
the information gained will narrow his choices to the

Optimal strategy and rule out non-optimal strategies.

The Election Game

In this representative political system, some
collective decisions are made in the legislature. Partici-
pants in the legislature, legislators, are selected from
among candidates for political office by voters, who
constitute the electorate of his district. Legislative
districts are geographically-defined subdivisions, or
enclaves, of the total electorate. The decisions of the
legislature must be approved by the executive, an elected
representatives whose constituency is the entire electorate

for whom collective decisions are made.12

 

12Behavior of representatives of enclaved districts
and their relationship with representatives of the larger
electorate is a topic discussed in Schlesinger, Ambition
and Politics.

 

35

Electoral Decision Rules

1.

Elections for each office are held at set intervals
Of time, limiting the length Of time the winners
Of elections can hold power. .

Each adult in the constituency has one vote for
each office, a vote he casts in secret.

Each district is single-member, that is only one
candidate can be elected. .

All candidates are nominated by political parties.
Political parties Operate by decision rules which
include that nomination for election in legislative
districts are independent Of one another and Of

the nomination Of the executive.

The candidate receiving a majority Of the votes
case is elected tO Office.

Candidates may spend money only for staff and
communications media. (They cannot buy votes.)

The executive is elected in the same election as
the legislator but the voter may make his choices
separately on the ballot.13

Electoral Exchanges

 

The citizen cannot win the electoral game per ee.

He can, however, utilize his vote in a way that will give

him the highest reward among possible outcomes Of the

larger collective decision-making game.

14 The politician

seeks political power through representative Office; his

 

13Election rules 4 and 7 are among the three condi-

tions Schlesinger names as requisites Of his "ambition
theory," Ibid., p. 126.

14Peter Blau makes the point that some exchanges

are carried out via chains Of intermediate transactions
in Exchange and Power in Social Life (New York: John

 

36

primary resources are his actions in that Office. The
citizen has resources, votes, which the politician needs
in order to be elected. Furthermore, the candidate is in
competition for these votes with another candidate who
also desires political power. Since the immediate outcome
Of elections is more valued by the candidates than the
voters, the candidates bear the.bargaining costs Of the
campaign.

The candidate.and the voter exchange resources,
legislative activity for votes. This exchange is not
explicit and binding, but implicit and contingent. The
temporal structure Of the political system imposes uncer-
tainty upon both participants in the exchange.

The ongoing political system we describe is
temporally divisible into election intervals and legis-
lative intervals. This has implications for the exchanges.
made in each interval and the information available to
the actors in the political system. Since the electiOn
decides who will make collective decisions in the succeeding
time period, the voter exchanges votes for future action.
However, as we shall see, the voter's best information for
making his choice consists largely Of knowledge Of poli-
ticians' past behavior. While past behavior.does not bind
a politician to like future behavior, the voter can make a
new decision in the next election on the basis Of action

in the interval between this and the next election.

37

Candidate uncertainty in campaigns stems from the
necessity to make implicit bargains through public media
and other efforts of his campaign organization. Voting
is secret and discrete from the campaign, not a direct
exchange, for example, the collection of proxies by the
candidate or his organization. The candidate campaigns
for power to be held in the future. Usually his best
resources, however, are his past actions--his own if he
is an incumbent or those Of members Of his party if he is
non-incumbent.

After election, the politician is not bound by
campaign promises, neither explicit pledges Of action,
nor implicit ones in his heralding Of achievements Of
himself or his party in the past. In the interval between
elections the focus Of exchanges he makes is the next time
period, the next election. The politician may use informa-
tion Of past behavior tO gauge the effectiveness Of alterna-
tive strategies. Nevertheless, activity between elections
is an extension Of campaign strategy for the next election

not the last election.

The Voter

 

The citizen has only three alternatives: to vote
for one or the other Of the two candidates or tO abstain.
TO cast any vote requires some expenditure Of resources.
If he abstains, however, the citizen foregoes the Opportunity

tO use his limited political power tO affect collective

38

decisions. Hence, the citizen who votes must see at least
marginal differences in the probable outcome Of collective
decisions if one candidate is elected rather than the
other.
V-l: The less difference in outcomes between his
alternatives perceived by the citizen, the
less likely he will vote.
Since the perception Of the choice depends upon information
possessed by voters, a corollary to the preceding hypothesis
is

V-l.l: The less information possessed by the citizen,
the less likely he will vote.

In non-competitive constituencies, i.e. where the candidate
Of same party wins nearly every election, the citizen may
not perceive that he has an.effective choice, that his vote
makes any difference in the outcome. Consequently,

V-l.2: The less competitive his district, the less
likely a citizen will vote.

Corroborative evidence.--Campbell, et al., find that

intensity Of partisan preference is positively related to

voting turnout among the respondents in their sample.15

They also discover a direct relationship between concern

16

over election outcome and voting turnout. Furthermore,

turnout increases monotonically with political involvement,

 

15Angus Campbell, et al., The American Voter,
abridged edition (New York: John Wiley, 1964): p. 53.

16

 

Ibid., p. 57.

39

a measure combining interest in campaigns, concern about
election outcome, political efficacy, and sense Of citizen
duty.17
Converse shows that information levels are related
tO non-voting; out Of 201 non-voters in the Survey Research
Center's sample, 72% possessed a medium level Of informa-

18

tion, 11% a high level Of information. Froman shows that

non-competitive congressional districts tend to exhibit

lower voting turnout than more competitive ones.19

The citizen can best utilize his vote by voting for
the candidate who he believes will use aggregated power to
make collective decisions he prefers.

Every voter may have different preference orderings
Of collective decisions. Consequently, neither candidate
is likely to promise policies which match every preference
of every voter. The vote, moreover, is not divisible. In
order to maximize his utilities, the voter must choose
between the candidates on the basis Of his highest prefer-
ences among policies Offered by the candidates.

In voting for one candidate, the citizen foregoes

the Opportunity to vote for the other. Since the candidate

 

17Ibid., p. 62.

18Philip E. Converse, "Information Flow and the
Stability Of Partisan Attitudes," in Elections and the
’POlitical Order, by Angus Campbell, et al. (New York:
363p Wiley, I966), p. 139. “"‘

19Lewis A. Froman, Jr., Congressmen and Their
Constituencies (Chicago: 'Rand McNally, 1963), p. 30.

 

40

makes exchanges for future benefits, the citizen must assess
the likelihood that each candidate will in fact pursue his
interests with the political power allocated tO him by the
election.

The most readily available information on the
politician's behavior is the candidates past behavior. If
the candidate is an incumbent in political Office, the voter
can assess his performance directly. If he has a record
Of supporting the voter's interests,.the candidates's
credibility in current exchanges is enhanced.

On the other hand, if the candidate is not an
incumbent, the voter must rely on related past experiences.
In the past, partisan representatives in power enacted
collective decisions affecting the voter. The voter can
assess the likelihood that the current candidates will
enact his preferences by examining the behavior of poli-
ticians Of the two parties in Office. If a party's candi-
dates have pursued his interests in collective decisions in
the past, then the candidate's value and credibility are
increased.' Consequently, the voter decides not only on
the basis Of candidates' campaign promises, but perhaps
primarily, on the basis Of his experience with the candidate
and members Of his party in Office as well.

V-2: For each Office, the voter chooses tO vote for
the candidate who

(a) promises collective decisions which bring
him the highest anticipated utility,

41

(b) is most credible because Of his past behavior
or because Of the policies enacted by the
party which nominated him or his past be-
havior.

Corroborative evidence.--Campbell, et al., demon-

 

strate that voters hold evaluations Of parties that are
somewhat independent Of their evaluation Of candidates.20
Key finds that although four-fifths to seven-eighths Of
the voters consistently vote for the same party, they are
rational in this behavior; that is they do so in conformity
with their agreement with the (presidential) party's poli-
cies. Key also shows that voters switched their support Of
party's nominees from election to election in accordance
with their approval or disapproval Of public policies enacted
by politicians Of the two parties.21.
Although the level Of ideological conceptualization
among the sample analyzed in The American Voter is not high,
the bulk (47%) Of the voters had some idea Of group benefits
associated with political parties. Fifteen and one-half per
cent Of the voters had a more coherent ideological concep-

22

tion Of party policies. Furthermore, Campbell, et al.,

hypothesize on the basis Of the survey responses:

 

20Campbell, et al., The American Voter, pp. 21-25.

 

21V. 0. Key, The Responsible Electorate (Cambridge:
Belknap, 1966). .

22

 

Campbell, et al., The American Voter, p. 144.

42

Other things equal, then:

(1) Persons for whom a value is more important will
be more likely tO express intense Opinions than
will persons for whom the same value is less
important.

(2) Persons who perceive issues to be more relevant
for their values will be more likely to express
intense Opinions.23

Finally, voters react tO past events. Key asserts,
The patterns Of flow Of the major streams Of shifting
voters graphically reflect the electorate in its great,
perhaps principal, role as an appraiser Of past events,
past performance, and past actions . . . VOters . . .
are not likely tO be attracted in great numbers by
promises Of the novel or unknown.2
Over time, if candidates Of one party consistently
use their Office to effect the voter's preferences in col-
lective decisions, the voter may reduce his information
costs in campaigns by voting with little information beyond
the party affiliation Of the candidates. The proportion Of
such partisan voters in a district constitutes the "normal
party vote" for each party in that district.25

V-3: Some voters consistently vote for the candidates
for Office Of the same party.

Corroborative evidence.--The data presented in The

 

American VOter reveal that, over time, a fairly consistent

proportion Of the electorate identify themselves as partisans

 

23Ibid., p. 104.

24Key, The Responsible Electorate, p. 61.

 

25Elaboration Of the "normal vote" can be found in
Elections and the Political Order, by Campbell, et al., pp.
5'39.

43

I I O I 2
Of various intenSities. 6

Furthermore, the more strongly
a respondent identifies himself with a party, the more
likely it was that he reports voting "always or mostly for

the same party."27

As we have noted, however, Key shows
that the ”standpatters," those who do not switch parties,
generally remain loyal in conformity with their approval
Of the policies enacted under their party's administra-

tions.28

The Candidate

 

As we have noted, the candidate's electoral resources
are his actions in Office. In order to be a candidate, the
politician exchanges a pledge to support his party's quest
for political power in return for nomination for Office.
While this promise applies to the competition for political
power in the legislature and the support Of the party's
quest for power in the larger electorate, the requirements
Of party loyalty cannot supercede the immediate Objective
Of winning the election.

C-l: The candidate promises party loyalty in order
to gain nomination.

C-l.l: Party loyalty is limited by the strategy the
candidate must pursue to be elected.

 

26Campbell, et al., The American Voter, p. 69.

 

27Ibid., p. 69.

28Key, The Responsible Electorate.

 

44

Corroborative evidence.--Kingdon's interviews with
candidates show that some politicians attribute more value
tO party nomination than others. Furthermore, winners who
encountered little Opposition attributed their victories
to party considerations far more than politicians who were
elected with stiff competition. The politicians, then, who
had fewer strategic decisions to make in order tO gain
election had more respect for the role of the political

party in their victories.29

The executive candidate runs under.the same party
designation or label as the legislative candidates.
Furthermore, the candidates for these Offices make policy
promises in return for votes on the same set Of collective
decisions. If the candidate judges that the policy promises
made by the executive are inimical to his chances Of
winning the election, he will dissociate his exchanges
from those Of the executive candidate; if the policy
promises Of the executive are congruent with his own
electoral exchanges, he will attempt to associate himself
with those exchanges.

C-l.2: The candidate will dissociate himself from
other party nominees if the other nominees
are making exchanges with the electorate that

conflict with the exchanges he is making,
that inhibit his chances Of being elected.

 

29John Kingdon, "Politicians' Beliefs About Voters,"
American Political Science Review, LXI (March, 1967), pp.

137-145.

45

Corroborative evidence.--The findings of Robert A.

 

Schoenberger support his hypothesis, "Those legislative
candidates most closely identified with their party's
Presidential candidate will achieve (or suffer) the latter's
electoral fate in a direct relationship to that identifica-

"30 Schoenberger found that Republican congressional

tion.
candidates who dissociated themselves from Goldwater's
policy stance were more successful at resisting the Demo-

cratic landslide in 1964 than those who did not.

Once nominated, the candidate exchanges promises
on collective decisions in return for votes. These promises
may be explicit, or implicit in the politician's record.
The politician knows that individuals' preference orderings
vary with respect to collective decisions. Consequently,
he knows that he is unlikely to satisfy a winning propor-
tion Of voters on every policy. Among a finite number Of
collective decisions, however, some groups Of individuals
will value the same decisions highly. Hence, these groups
will react similarly tO election campaigns. Thus, in order
tO reduce his bargaining costs,

C-2: The politician appeals to groups rather than
' to individuals for their votes.

The groups with whom a candidate makes exchanges are his

electoral coalition.

 

30Robert A. Schoenberger, "Campaign Strategy and
Party Loyalty," American Political Science Review, LXIII
(June, 1969): p0 516.

46

Corroborative evidence.--Blau asserts, "Political
parties compete for social segments rather than individual

«31

votes. Buchanan and Tullock also maintain that in large

groups where interaction and mutual adjustment dO not take
place directly and explicitly, vote trading is implicit
rather than explicit.
The political 'entrepreneurs' who Offer candidates or
programs to the voters make up a complex mixture Of
policies designed to attract support. In so doing,
they keep firmly in mind the fact that the single
voter may be so interested in the outcome Of a par-
ticular issue that he will vote for the one party
that supports this issue, although he may be Opposed
to the party stand on all other issues.3
Kingdon provides us more direct evidence: although only
fifty-six per cent Of the candidates interviewed admitted
making "group appeals," well over seventy-five per cent Of
those candidates designated groups which comprised their

electoral coalition.33

The fewer promises the candidate must make, the
easier it is to communicate them to constituents, and the
less the probability that they will be inconsistent within
voters' preference orderings. Hence,

C-3: The politician will make as few promises on

collective decisions as he can in order to be
elected.

 

31Blau, Exchange and Power in Social Life, p. 248.

32
134-135.

33John Kingdon, Candidates for Office, Beliefs and
Strategies (New York: Random House, 1968).

Buchanan and Tullock, Calculus Of Consent, pp.

47

Corroborative evidence.--This hypothesis is related

 

to the size principle, the prediction that, within n-person,
zero-sum games in which actors are rational and have perfect
information, minimum winning coalitions will be formed.34
Its veracity in an electoral situation is demonstrated by

Rosenthal's research on French electoral coalitions.35

Campaigns are costly in the money, time, and energy
required to gain nomination and communicate policy promises
to voters. In order to maximize his opportunity tO win the
election, the candidate will collect information that will
enable him to choose a winning strategy in terms Of campaign
behavior and to make his promises of.future behavior credible.
The least costly information.for the candidate is
the past behavior Of his constituency and his opponent's
current behavior. If the candidate is an incumbent in
political Office, he has made exchanges with constituents
in the past. In the present election, he can reduce bar-
gaining costs by remaking exchanges with the same constituency
groups, using his past behavior as evidence Of his value and

credibility. If he has not acted in his constituents'

 

34William H. Riker, The Theogy Of Political Coali-
tions (New Haven: Yale, 1962).

35Howard Rosenthal, "Voting and Coalition Models
in Election Simulations," in Simulation in the Study Of
Politics, ed. by William Coplin (Chicago: Markham, 1969),
pp. 237-285.

 

 

48

interests, the rational strategy for his opponent is to
exploit the incumbent's record in Office in order tO
discredit him.

C-4: The incumbent running for reelection uses
his record in Office as a campaign resource.

C-4.l: The challenger to an incumbent attempts tO
discredit an incumbent by exposing his voting
record to constituents if that record demon-
strates that the incumbent did not act in the
interests Of his constituents.

Corroborative evidence.--The advantages Of incumbency
are demonstrated by the high rate of reelection Of incumbents
who run for consecutive terms. Kenneth Prewitt, studying
city councilmen, found that over a ten-year period, eighty

36

per cent Of the incumbents who sought reelection won. The

success Of congressional incumbents is even more dramatic,
90% Of the incumbent representatives who ran between 1924

37 Barbara Hinckley Observes that

and 1956 won reelection.
in Senate elections, " . . . one finds a statistically
significant relationship between incumbency and strong
versus moderate deviations from the base party vote, with
this relationship holding for both presidential-year and

Off-year elections."38

 

36Kenneth Prewitt, "Political Ambitions, Volunteerism,
and Electoral Accountability," American Political Science
Review, LXIV (March, 1970), pp. 5-17.

37David A. Leuthold, Electiopeering in a Democraey
(New York: John Wiley, 1968), p. 127 (cited in Prewitt,
Ibid., p. 9).

38Barbara Hinckley, "Incumbency and the Presidential
Vote in Senate Elections: Defining Parameters in Subpresiden-
tial Voting," American Political Science Review, LXIV
(September, 1970), P. 840. ’

49

The consequence Of a record in Office displeasing
to constituents is illustrated by an example described by
Miller and Stokes. In 1958, Brooks Hays, who had gained
a reputation as a civil rights moderate was defeated by a
write-in candidate, Dale Alford, in Arkansas' Fifth District.
Furthermore, interviews Of constituents show that they were

aware of this record, and that is why he was defeated.39

If the candidate is not an incumbent, he must make
new exchanges. Candidates Of his party have made exchanges
with constituency groups in the past. The candidate will
attempt tO remake those exchanges, appealing tO the behavior
Of party affiliates in Office as evidence Of his credibility.
If the candidate's party has received large majorities in
past elections, the current candidate will simply renew the
exchanges made by those candidates in the past.

C-S: Candidates of the majority party in non-competitive
constituencies will make exchanges with little
policy content beyond that necessary tO gain the

normal party vote.

Corroborative evidence.--Kingdon's study shows that

 

candidates who won by wide margins named party the most

important consideration in their victory in contrast to

. . 4
marginal Winners. 0

 

39
Congress."
40

Miller and Stokes, "Constituency Influence in

Kingdon, "Politicians' Beliefs About Voters."

50

The less successful party candidates in the past
have been, the more the current candidate attempts tO make
exchanges in addition tO past exchanges to assure election.
If candidate Of his party have been in the minority, the
candidate attempts to make exchanges with voters who have
previously supported the majority party by Offering a policy
product similar to the majority party's candidate. The
candidate Of a minority party may also attempt tO make
radically different exchanges with the electorate, that
is introduce new preferences into the voters' priorities.
This is risky, however, because it may alienate the support
otherwise easily obtainable for the candidate. He may also
have difficulty making such promises credible. Thus,

C—6: In a two party system, the minority party
candidate's policy promises converge on the
majority party candidate's promises.

Corroborative evidence.--While Downs maintains that
parties will converge ideologically in.a two-party system,41
Schlesinger modifies this proposition:

The condition Of dominance, then, within the two-

party system introduces different kinds Of tensions

for the two parties. The tension tO converge exists
primarily for the minority party. The majority party

seeks instead to differentiate itself from its
competitor.42

 

41Anthony. Downs, An Economic Theory_of Democracy,

pp. 117-132.

42Schlesinger,‘Ambition and Politics, p. 125.

51

In a competitive district, where parties are evenly
balanced, candidate uncertainty is elevated. Each candidate
will try to raise the value Of voting for himself relative
tO that Of his rival by promising more in terms Of policy
to voters. Therefore:

C-7: In competitive constituencies, the campaigns Of
both candidates are more issue-oriented than in

non-competitive constituencies.

Corroborative evidence.--Kingdon finds that candidates

 

who won in competitive constituencies are more likely than
those in non-competitive constituencies tO name issues and
candidate characteristics rather than party as the most
important factor in their victories.43
As a consequence Of the electoral process, the
politician makes two kinds Of exchanges, one with the party,
to support its quest for political power, and one with his

constituents, tO support their interests in the allocation

Of social resources.

The Legislative Game

 

The legislature is a body Of individual politicians,
legislators, who meet face-to-face in a legislative chamber
or "house." Legislative decisions are made by voting on
motions in the legislative chamber, or "on the flOOr."

Motions are made for passage Of bills, legislation or

 

43Kingdon, "Politicians' Beliefs About Voters."

52

collective decisions. Other motions affect the outcome Of

bills:

motions for specific amendments, for recommitting

the bill to committee for further consideration, for

"tabling" a motion--de1aying a decision for a definite

or indefinite interval Of time. Committees are ongoing

organizations composed Of a small number Of legislators who

are allowed tO revise the content Of bills and report a

recommendation Of passage or defeat Of bills to the entire

legislature.

Legislative Decision Rules

 

1.
2.

All legislators are elected tO their Office.

Any member Of the legislature may introduce a
bill, an amendment tO a bill, or a procedural
motion affecting the outcome of a bill.

Each legislator has one vote on each motion brought
tO the floor Of the legislative chamber.

The motion wins if a given proportion, usually a
majority Of legislators, votes 'yea.‘ It loses
if that proportion votes 'nay.‘

The executive must agree tO a bill's passage before
it becomes law.

The legislature is organized into the majority
party and the minority party.

a. Party leaders are selected by majority vote by
party members and serve at the will Of the
party members, i.e. they can be replaced at
any time.

b. Party leaders make assignments Of rank and
file party members tO committees.

All bills originated in the legislative house are
sent to committees for consideration.

53

a. Committees have their own rules for making
decisions which may include assigning legis-
lation tO subcommittees Of the committee.

b. Committee chairmen allocate resources within
the committee, such as the time available for
hearings, assignment tO subcommittees, and
staff for gathering information on the impact
Of collective decisions.

c. Committee chairmen are members Of the majority
party.

d. Majority party members hold the same proportion

Of seats on committees as they hold in the
legislative chamber as a whole.

Legislative Exchanges

 

Legislators are politicians, whose rational inten-
tion is the acquisition Of political power. As we have
noted, the legislative Office is only one in the political
Opportunity structure. Since the legislator is interested
in gaining political power, he will attempt to use his
resources to gain political power in larger constituencies.
Opportunities become attenuated at higher levels in the
structure Of political Offices; the legislator, then,
cannot be certain about his chances to gain higher Office.
Because the legislator is uncertain about advancement tO
positions Of greater power in the political system, he will
attempt tO maintain his present position by using his re-
sources tO assure reelection. Uncertainty does not preclude
a legislator's pursuing both reelection and higher Office.

Since the legislator has many opportunities to affect

54

collective decisions, he may use his resources tO different
ends at different points in time.

The most abundant resource the legislator has is
his vote on collective decisions. He will use this resource
as a currency with which to make exchanges with extra-
1egislative constituencies by seeking enactment Of legis-
lation in their interest and by blocking legislation
inimical to their interest. The legislator will also
exchange his votes and other resources with other members
Of the legislature for control over decisions which affect
his power within the legislature and his Opportunities for

extra-legislative exchanges.

Constituency

 

In order to be elected, the politician must make
exchanges with the electorate in his district. Whether he
seeks election to a higher Office or reelection to his
present Office, his record in Office is a valuable resource.
For reelection purposes, a legislator's future constituency
is virtually identical tO the constituency which elected
him to his present Office.

L-l: The legislator supports legislation which is
in the interest Of his constituents, and
Opposes legislation which is inimical to
the interests Of his constituents.
Some Of the interests Of the legislator's constituents may

be relatively easy to infer, such as economic advantage.

Thus,

55

L-l.l: The legislator supports legislation which is
congruent with the economic interests Of his
constituents and Opposes legislation which
is inimical to their economic interests.

Some collective decisions may affect the political power Of
his constituents. Hence,

L-l.2: The legislator supports legislation which
enhances the political power Of his constituents
and Opposes legislation which curtails the
political power Of his constituents.

Furthermore, it follows from the campaign strategy Of
candidates in competitive districts that the legislator
who is uncertain about his chances for reelection attempts
to raise the value Of voting for himself relative tO future
Opponents by vigorous pursuit Of the interests Of the
voters.

L-l.3: If the legislator was narrowly elected, his
support Of constituency interests is stronger

than if he was elected by a wide margin.

Corroborative evidence.--Rustow says, "The desire

 

for economic improvement has been an important motive in
the rise Of political movements. But the quest for poli-
tical power and social status has been an even more potent
force."44
Several studies show that, while the Democrats and

the Republicans are generally supported by different types

Of constituencies, legislators who represent districts

 

44DankwartmRustow, The Politics Of Compromise (New
York: Greenwood, 1969), PP. 229-230.

 

56

atypical Of party strength are likely tO deviate from their
party toward the party typical for their districts.45
Miller and Stokes find that agreement with constitu-
ency Opinion on roll-calls is issue-specific. Cnudde and
McCrone, upon reanalysis Of party Of the Miller and Stokes
data, attribute agreement with constituency Opinion tO the
legislators' voting with their perception Of constituency
Opinion rather than his sharing the attitudes Of his
constituents.46
Mayhew found that "interested" congressmen Of both
parties voted with their constituents' interest on farm,
city, labor, and "Western" issues.47
Evidence on the hypothesis relating competitiveness
Of district tO roll-call voting is mixed. MacRae asserts

48

that this is the case. Froman speculates that this may

 

45Lewis A. Froman, Congressmen_end Their Constitu-
encies, pp. 94-95; Duncan MacRae, Jr., "The Relation Between
Roll-Call Votes and Constituencies in the Massachusetts House
Of Representatives," American Political Science Review,
LXVI (1952), pp. 1046-1055; Frank J. Sorauf, Party and
Representation: Legislative Politics in Pennsylvania (New
York: Atherton, 1963), p. 52.

46Miller and Stokes, "Constituency Influence in
Congress;" C. F. Cnudde and D. J. McCrone, "The Linkage
Between Constituency Attitudes and Congressional Voting
Behavior: Causal Model," American Political Science
Review, 60 (1966), pp. 66-73.

47David Mayhew, Party Loyalty Among Congressmen
(Cambridge: Harvard University, 1966).

48MacRae, "The Relation Between Roll-Call Votes and
Constituencies in the Massachusetts House Of Representatives.‘

57

not be true at the national level since competitive districts
are more likely to be heterogeneous as well. Heterogeneity
would make it less likely that the representative could
accurately assess his constituents' interests on issues

arising in the legislature.49

Some constituency interests may be relatively unknown
to the legislator. The legislator cannot predict precisely
which groups will support him in future elections since
electoral exchange do not result from face-to-face bargain-
ing. The legislator does know, however, how his constituents
respond to other campaigns.

The executive, like the legislator, must act in
Office in such a way that he can use his record to make
electoral exchanges. Thus, the executive will support bills
which promote his ability to make exchanges with future
constituents and oppose bills which will inhibit his
Opportunities tO win the votes Of a majority in the total
electorate. The legislator, then, can infer enclaved
constituency desires in part from its past response to
the executive's campaign. We anticipate,

L-l.4: If the executive was successful in his dis-
trict, the legislator supports legislation
supported by the executive; if the executive

fared poorly in his district, the legislator
Opposes legislation supported by the executive.

 

49

Froman} Congressmen and Their Constituencies,
p. 117. ’

 

58

Corroborative evidence.--Evidence on hypothesis

 

L-l.4 is mixed. Sorauf finds that competiveness for

governor is related to deviation from party majorities.50
LeBlanc also finds that among state senators, the party
vote for governor or president is a relatively good pre-

dictor Of roll-call Voting.51

Among Southern Democrats,
Flinn and WOlman find that the vote for Kennedy in 1960
(positively) and the vote for the States' Rights Party in
1948 (negatively) correlate with Kennedy Support and Party
Unity measures in the First Session of the Eighty-Eighth
Congress. Furthermore, these variables are more highly
related tO the dependent variables than demographic
variables (per cent urban, per cent Black, per cent work
force in mining and manufacturing, and for rural districts,

2 Waldman investigates this rela-

per cent farm tenancy).5
tionship within parties in the Eighty-Ninth (for Democrats
and the Eighty-Seventh (for Republicans) Congresses.

Among Democrats he finds that the size Of the presidential

vote and the relationship Of the congressman's vote relative

tO the President's (whether he led or lagged) is related to

 

50Sorauf, Party and Representation, p. 141.

 

SlHugh LeBlanc, "Voting in State Senates: Party
and Constituency Influences," Midwest Journal Of Political
Science, XIII (February, 1969), pp. 33-57.

52Thomas A. Flinn and Harold L. WOlman, "Constituency
and Roll-Call.Voting:7 the Case Of Southern Democratic
Congressmen," Midwest Journal Of Political Science, X (May,
1966): Pp. 192-199.

59

liberalism Of voting on selected roll-calls. Those who
lagged behind the President's vote were higher in liberalism;
within groups discriminated by this criteria, those in whose
districts Kennedy received large majorities were more
liberal. This relationship is much weaker and inconsistent

53 Other studies show, however, that the

for Republicans.
size Of the executive vote in a district is related tO roll-
call behavior before the election, casting doubt on a

causal chain from the executive's electoral margin to roll-

call votes.54

Sponsorship

Whatever future Office the legislator desires, he
has the Opportunity to introduce legislation designed to
make credible exchanges with future constituents. Conse-
quently, we can infer high interest in the success Of
legislation from the sponsorship Of motions.

L-2: Legislators support motions they Sponsor with
their votes on those decisions.

Party

In order tO be nominated for Office, the politician

must support his party in its drive for political power.

 

53Loren K. Waldman, "Liberalism of Congressmen and
the Presidential Vote in Their Districts," Midwest Journal
Of Political Science, XI (February, 1967), pp. 73-86.

54Sarah McCally, "The Governor and His Legislative
Party," American Political Science Review, LX (December, 1966),
pp. 923-942; Marvin Weinbaum and Dennis R. Judd, "In Search
Of a Mandated Congress," Midwest Journal Of Political Science,
XIV (May, 1970), pp. 276-302.

60

Furthermore, since a majority of seats on committees and
the chairmanships are allocated to majority party members,
it is the advantage Of the members Of each party to have
as large a proportion Of the seats in the legislature as
possible. Hence, party members will use their resources
to support legislation proposed by members Of their party.
Further, a legislator will undermine the opportunities Of
the other party by opposing legislation proposed by members
of the rival party. We anticipate, then,
L-3: Party members support members Of their party
on motions present to the legislature and
Oppose motions presented tO the legislature
by members Of the opposite party.
Since the executive is also a partisan politician, we add
the following hypothesis:

L-3.l: The executive's position contributes tO the
definition Of motion as a party sponsored
motion.

As a consequence Of support elicited by party exchange, in

the aggregate we expect

L-3.2: Party affiliation is highly related to legis-
lative voting.

Since the impact Of constituency exchanges may conflict
with party loyalty, we hypothesize,
L-3.3: Party loyalty varies across different issues.
L-3.4: Party loyalty tends to be reinforced when it
' is congruent with constituency influence and

tends tO be weakenedwhen it conflicts with
constituency factors.

61

Corroborative evidence.--Gerald Marwell finds party

 

influence highly evident in the factor structure Of congres-
sional voting. Marwell excludes roll-calls with a .9 (or
higher) correlation with party affiliation and unanimous

and nearly unanimous roll-calls which indicate high bi-
partisan support. Nevertheless, he finds, " . . . party

still defines the first factor . . . ."55

Leadership

Although power is equally distributed in the legis-
lature in the sense that every legislator can introduce
motions and has one vote on each motion, the form Of a bill
can be greatly influenced in committees. A committee seat
Offers an added Opportunity to the legislator to affect the
outcome Of decisions. The legislator, then, prefers tO sit
on committees which consider legislation affecting the
utility Of his future constituents. The party leaders
also hold disproportionate power by virtue Of their role
in committee assignments.

The members Of a party elect the party leaders by
voting for their choice among contenders.for that position.
The resources Of aspirants for party leadership posts are

their power over committee assignments and their votes on

 

55Gerald Marwell, "Party, Region and the Dimensions
Of Conflict in the House Of Representatives, 1949-1954,"
American Political Science Review, LXI (June, 1967): p. 381.

 

62

collective decisions. The candidate for party leader can
exchange promises tO support party member's ambitions in
return for his position Of added power. In Office, the

party leader makes committee assignments desired by the
individual party members for their pursuit Of extra-legislative
exchanges. The party leader's resources in this respect are
limited by the number Of committee seats available for
distribution. Furthermore, committee subject matter may

vary in the interest it-hOldS for.extra-legislative constitu-
encies, especially for larger constituencies than the ones
from which legislators were elected. Furthermore, some
committees may hold subsidiary power over subject matter

Of other committees. Given the variations in attractiveness
Of committees, the leader satisfies every ambition he can.
However, in order to preserve his position Of power, the
leader will support the ambitions Of individual members Of
his party by supporting the legislation they propose. In
order to distinguish himself from the rest Of the party
members,

L-4: The party leaders exhibit greater party loyalty
than the rank-and-file party members.

Corroborative evidence.--MacRae finds that party

 

leaders are more highly loyal tO party than the rank-and-

56

file. Ripley says,

 

56Duncan MacRae, Jr., "Roll-Call Votes and Leader-
ship," Public Opinion Quarterly, 20 (1956), pp. 543-558.

 

63

Party leaders make few requests Of the members; the
bargaining position Of the individual senators is
good, even when weighed against that Of the leaders.
Senators can use the leaders to help them attain
their own individual ends.57

Committees

 

Committees have their own rules and decision-making
processes. As a result Of their bargain with the party
leadership, most members have high interest in the outcome
Of decisions made by the committees Of which they are
members. The type Of bargaining that goes on in committees
is influenced by the subject matter which is considered.

(1) If interests Of committee members conflict, decisions
will be difficult to reach, and unanimous decisions will

be virtually impossible to achieve. (2) If the interests
Of committee members are diverse, but not conflicting,

then it is likely that all members can have their interests
satisfied in the committee's recommendation. (3) If the
subject matter is continuously divisible, e.g..money matters
such as appropriations or taxes, then committee bargaining
is likely to occur which is similar to that Of economic
transactions. The resulting recommendation is likely tO

be a compremise figure, a mutually satisfactory decision

for all members Of the committee.

 

57Randall B. Ripley, Power in the Senate (New York:
St. Martins, 1969), p. 228.

64

As a consequence Of his high interest in the
substantive matters considered by the committees of which
he is a member,

L-S: The legislator has an interest in the bills
considered by his committee.

L-5.l: If a bill revised in committee promotes his
ambitions, the legislator supports it by
voting for a favorable recommendation and
for its passage in the legislature.

L-5.2: If the revised bill inhibits his ambitions, the
legislator opposes a favorable recommendation,
and votes against its passage in the legis-
lature.

Corroborative evidence.--GOOdwin shows that all the
committees which are nationally-oriented in their substantive
concerns are more prestigious than those which are oriented

58

toward narrower clienteles. Masters asserts that electoral

considerations rank high in importance among the criteria
utilized in committee assignments.59

Ralph K. Huitt concludes that committee members are
advocates Of policy positions rather than.impartial judges:
committee members take sides in debate and their perception
Of facts is affected by their predilections toward the

subject matter.6.0

 

58George Goodwin, The Little Legielatures (Amherst:
University Of Massachusetts, 1970), p. 115.

59Nicholas A. Masters, "House Committee Assignments,"
'American Political Science Review, LV (June, 1961): PP.
345-358.

 

 

6oRalph K. Huitt, "The Congressional Committee: A
Case Study,"‘Ame;ican Political Science'Review, XLVIII (June,
1954), PP. 340-365.

65

Rustow differentiates among coalition strategies
according as they are employed for reaching decisions on
different kinds Of issues. He names these strategies
exclusive, inclusive, and split-the-difference. Exclusive
strategies are followed in cases where a variety Of inter-
ests cannot be accomodated because all participants are
highly interested in the outcome, and they disagree. The
resultant coalitions reflect the gross strength Of the
conflicting groups. Inclusive strategies are those per-
mitted by a variety Of interests, all Of which can be
accommodated without conflict. A split-the-difference
strategy is possible where the issue is monetary and a
compromise position can be reached through bargaining.61

Case studies indicate that these strategies which
parallel the three predictions about intra-committee
bargaining from exchange theory accurately describe the
behavior in committees Of Congress.

1. The House Labor and Education Committee is
characterized by an exclusive coalition pattern. The
subject matter it handles is redistributive, social class-
oriented, and approximates a zero-sum game. The cleavages
in this committee stem from disputes over the kind and
quality Of government action rather than the extent Of

particular programs. Bargains are difficult tO make since

participants are unwilling to make exchanges. Hence,

 

61Rustow, The Politics Of Compromise.

66

conflict within this committee is great, reports contested,
and cohesion on roll-call voting poor.62
2. The House Agriculture Committee follows the
inclusive strategy. Product differentiation among constitu-
encies of committee members constitutes a variety Of
compatible interests. Virtually all members can be satis-
fied in their desires for legislation.63
3. Richard FennO suggests that
The subject matter . . . keeps [Appropriations]
Committee members relatively free agents, which
promotes intra-committee maneuvering and, hence,
conflict avoidance. Members do not commit them-
selves tO their constituents in terms Of precise money
amounts, and no dollar sum is sacred--it can always
be adjusted without conceding that a principle has
been breached.64
The mutual adjustment process-allows widespread satisfaction
and high cohesion among committee members both in their
reports and in their behavior on the floor Of the legis-
lature.
Hypotheses L-l through L-5 predict the generation

Of interests in a motion presented tO the legislature.

 

62Richard Fenno, Jr., "The House of Representatives
and Federal Aid to Education," in New Perspectives on the
House Of Represeptatives, Robert Peabody and Nelson POIsEy,
eds. (Chicago: Rand McNally, 1963), pp. 195-236.

63Charles 0. Jones, "The Agriculture Committee and
the Problem Of Representation," in New Perspectives on the
House Of Representatives, Robert Peabody and Nelson POlsby,
eds. (Chicago: Rand McNally, 1963), pp. 109-128.

64Richard Fenno, Jr., "The ApprOpriations Committee
as a Political System," in New Perspectives on the House Of
Representatives, Robert Peabody and Nelson POlsby, eds.
TGnicago: Rand McNally, 1963): p. 84.

 

 

 

 

 

67

Since some Of these interests conflict with one another,
it is likely that the legislator will pursue his interests
serially over the decisions made in the legislature rather
than simultaneously. The full impact Of his interests are
likely to be exhibited on passage Of bills, since these
have the greatest interest for extra-legislative constitu-

encies.

Legislative Bargaining

 

Over the gamut Of collective decisions made in the
legislature, some legislators have little interest in some
and great interest in others. On any particular decision,
the legislators who have great interest in the outcome
attempt tO trade resources affecting decisions which mean
little to them for votes on the current decision. Likewise,
the legislators who have a surplus Of power over a decision
which is Of little interest to them attempt to trade their
vote for other resources which affect decisions which are
Of greater value to them. At a stage immediately prior to
the roll-call, the legislators can be diVided into buyers
and sellers Of resources. The buyers are those.who, because
the collective decision is Of value to their ambition, have
high interest in the outcome. The sellers are those for
whom the vote has little value except as a resource to
trade for greater control over other decisions. A decision
may have little value for a legislator either because it

does not affect his ambitions or because it may both promote

68

and inhibit his ambition depending upon the situation in
which he will make exchanges in the future.

L-6: The value Of a collective decision tO a
legislator is proportional to the value it
has to him in making exchanges for political
power.

L-6.l: Those who have great value for the outcome
Of a collective decision exchange control over
decisions the outcome Of which is less valuable
in return for support Of their pOsition on the
collective decision which is Of greater inter-
est.

L-6.2: Those who value collective decisions little
exchange support on that decision in return
for greater control over decisions they value
highly.

Corroborative evidence.--Matthews characterizes

activities exhibiting the norm Of reciprocity in the

65

Senate a "game." He also says,

Every senator, at one time or another, is in a posi-
tion tO help out a colleague. The folkways Of the
Senate hold that a senator should provide this
assistance and that he be repaid in kind. The most
important aspect Of this pattern Of reciprocity is,
no doubt, the trading Of votes.55

Although later in his work, Matthews describes the inter-
action in the Senate as advice or cue giving, reciprocity
remains the dominant thrust Of his account.67

POlsby also Observes that consequences Of fragmented

power--mutual interdependence Of legislators.

 

65Donald Matthews, U.S. Senators and Their World
(New York: Random House, 1960), p. 101.

661bid., p. 99.

67Ibid., p. 252.

69

. . . virtually nO senator in the Eighty-Eighth Con-
gress save a sprinkling Of freshmen, was without some
institutional base which guaranteed him a dispropor-
tionate say, either in some substantive area Of public
policy or in the behind-the-scenes management Of Senate
business . . . . The need for cooperative effort, and
uncertainty about the precise composition Of any par-
ticular winning coalition, makes senatorial bargaining
necessary . . .

Ripley, likewise, stresses the prevalence of bargaining in
the Senate, the trading Of votes as well as other resources
affecting the outcome Of legislation.

The general stress on mutual aid that characterizes
the Senate, particularly in committee, results in

a great deal Of bargaining and trading Of credits.
Trades Of one specific item for another are not
Often made. But if a senator helps another senator,
he anticipates that when he needs help at some future
time he will be able to Obtain it from the person he
is presently helping.69

Coalitions

 

The sellers can only sell votes on the current
collective decision, but the resources for which they can
exchange votes may be votes or other resources held by
legislators, such as action in committee. The buyers Of
votes will not, Of course, expend more resources than
necessary to assure the outcome they desire. That is

L-7: Proponents Of both passage and failure Of a

motion attempt to gain a minimum winning
coalition.

 

68Nelson W. POlsby, Congress and the Presiden
(Englewood Cliffs, N.J.: Prentice Hall, 1964), p. 89.

69Ripley, Power in the Senate, p. 175.

70

Corroborative eVidence.--In the legislature, as
in elections, we have a situation in which Riker's size
principle is Operable in the formation Of coalitions. The
size principle in legislative situations is demonstrated
by Axelrod's work on coalition formation in the Italian

Chamber Of Deputies.70

Patterns Of Interaction

 

Since the trading Of resources takes place under
the pressure Of time, information on the alternative
possibilities for exchange is very important for both
buyer and seller. The legislator seeks information that
will enable him to minimize bargaining and Opportunity
costs. His information must include who is willing to
make exchanges, i.e. who is interested in the outcome of
a decision. Furthermore, the legislator wants tO reach
bargains with a minimum Of cost in terms.of the time it
will take to negotiate bargains. Finally, he will attempt
to make bargains with individuals to whom he must relinquish
the least in terms Of Opportunities foregone to use his
resources in ways which would be Of greater benefit to him.

The party leader, because Of his position as

supporter Of party members' goals, has information about

 

70Robert Axeerd, "Derivation Of a Coalition Theory
Based on Conflict Of Interest with an Application tO Italy"
(paper prepared for delivery at the Sixty-Fifth Annual
Meeting Of the American Political Science Association,
1969).

71

exchanges which can be made. Consequently, some exchanges
will be made through the party leader as intermediary.

His willingness to carry out this function may be tempered
by his own extra-legislative political interests in the
bill. We predict,

L-8: Some legislators exchange votes for resources
affecting political decisions through the party
leader.

Corroborative evidence.--Polsby comments on Lyndon
Johnson's performance as majority party leader in the U.S.
Senate:

Johnson was in a position to know more about the rela-

tive intensities Of senators' positions on a variety

Of issues, and in this way could create coalitions Of

senators who would never have thought to get together

on their own, but who, under Johnson's guidance,

could be brought together to help one another on

projects important to them. In return they would

give Johnson support on items that for them.mattered

less.71

Secondly, the members Of committees which have
reported the bill have an interest in it; hence, they will
trade resources for votes in support Of their positions on
the bill. We anticipate, then,

L-9: Trading Of legislative resources takes place
place between those uninterested in the outcome
Of bill and the members Of the committee(s)
reporting the bill.

Cerroborative evidence.--Francis finds that influence

concerning substantive matters in a state legislature is

 

71POlsby, Congress and the Presidengy, p. 45.

72

attributed to members Of committees dealing with those
subjects.72 Ripley, tOO, finds that among U.S. senators,
". . . most automatically vote the position taken by
their fellow party members on the standing committee
handling the bill."73
Members Of the legislative committees have experience
bargaining with one another. Because they have had more
interaction directed towards reaching collective decisions
which promote their political ambitions, members Of the
same committees are likely to be familiar with the prefer-
ence Of other members Of their committees. We predict,
L-lO: Trading Of legislative resources takes place
between legislators uninterested in the out-
come Of the motion and interested legislators
who share their committee assignments.
Corroborative evidence.--Ripley points out the
wealth Of possibilities committee sharing provides for
legislators: '
Because senators sit on several committees the
chances and necessity for negotiation are increased.
The senators' remembering that they will have to
deal with other specific senators on a number Of issues
facilitates accommodation.74
Thibaut and Kelley also discuss the likelihood that indi-

viduals who are spatially close make exchanges.7S

 

72Wayne L. Francis, "Influence and Interaction in
a State Legislative Body."

73Ripley, Power in the Senate, p. 122.
74Ibid., p. 118.

75John Thibaut and Harold Kelley, The Social Psychology
Of Groups (New York: John Wiley, 1959), p. 39.

 

73

The participants in an exchange can minimize
Opportunity costs by making exchanges with legislators
who require the least in terms Of Opportunities foregone
to use resources to affect decisions which are Of greater
value to them. The buyer Of a vote does not wish to
promise action on a collective decision which is likely
tO be inimical tO his political Opportunities. If he has
any preference about the outcome, the seller Of a vote
hopes to exchange his support with members Of the coalition
who are voting in accordance with his small interest.
Among individuals in groups within which we expect trans-
actions tO take place, we eXpect trades will take place
among individuals who are similar to one another along some
dimension relating to the focus of their political ambi-
tions. One dimension along which similar individuals
would not conflict is demographic composition of the
constituency which elected them.. We chose this instead
Of party because in campaign strategies, constituency
considerations come before party.

L-ll: Trading Of legislative resources takes place
between legislators uninterested in the out-
come and legislators similar tO them in com-
pfieition of the constituency which elected

Corroborative evidence---In a system where electoral

success is more rewarding to political ambition than party

74

loyalty, Sorauf finds, "Where the influence Of party and

constituency conflict, constituency tends to win out.”6

As a consequence of this bargaining hypothesis,
party groups simulated in our model should be cohesive
insofar as constituent districts are similar. This
variable characteristic Of simulated party voting cor-
responds tO Observations Of comparative studies Of state
legislatures in the United States. Comparing eight
legislatures, Jewell finds,

One factor leading tO greater party voting in states
with a higher urban concentration is that in most

of these states party strength follows more consist-
ently an urban-rural division. In the larger urban
states the parties represent more clearly defined
groups Of interests, with the Democrats coming mainly
from the largest cities and metropolitan areas, and
the Republicans more representative Of the smaller
cities and rural areas. 7

LeBlanc finds a similar pattern in a comparison Of
twenty-six state senates. He concludes,

. . . socio-economic interests Of constituencies are
more likely perceived as differentiated by senators
from the more industrialized states who are inclined
tO vote their constituencies. This Often results in
party votes . . . . Less homogeneity and more
ambiguity characterized the relation Of constituency
factors and legislative voting in the less partisan
states.78

 

7680rauf, Party and Representation, p. 145.

77Malcolm Jewell, "Party Voting in American State
Legislatures," American Political Science Review, XLIX
(September, 19537, p. 786.

78LeBlanc, "Voting in State Senates: Party and
Constituency Influences," pp. 51-53.

75

Axelrod points out that Riker's theory Of political
coalitions does not predict that minimum winning coalitions
will be formed which are connected, i.e. members adjacent
in an ordinal policy space. Connectedness is related to
rationality in the Axelrod links it tO ease in bargaining.
The reason is that negotiations for coalitions that
have low conflict Of interest will simply be easier
to conclude successfully, and hence these coalitions
can be expected to be more likely to form--even if
the political leaders are not able to identify them
beforehand. Likewise a coalition with low conflict
Of interest can be eXpected tO last longer once formed
than an average coalition, just because disputes within
such a coalition will be easier to resolve. 9

The Vote Decision

The legislator, the vote seller, surveys the informa-
tion that is likely to narrow his choices tO those which
will cost him the least and benefit him the most. The
legislator must take into consideration the relative value
Of voting each way in terms Of how much his vote is worth
in exchange with interest legislators. That is, he will
maximize his gain from the trade. He will also minimize
his Opportunity costs, the Opportunity foregone to vote in
accordance with his own interests on the bill, if any.

L-12: The legislator exchanges his vote with an
interested legislator who Offers him the

most in terms Of resources in exchange for
his vote.

 

79Axelrod, "Derivation Of a Coalition Theory Based
on Conflict Of Interest with an Application to Italy," p. 9.

76

L-12.l: The legislator takes into account his own
interest, if any in the outcome when he makes
a decision to exchange his vote for resources
affecting the outcome Of other decisions.

Summapy

Table 1 allows comparison Of the propositions
generated by Cherryholmes and Shapiro and the hypotheses
related to legislative behavior deduced from exchange
theory in a two party system. Clearly exchange theory
explains much Of the pattern exhibited in the Cherryholmes-
Shapiro model. That is, we can subsume their pattern
explanation under exchange theory.

The chief differences between the two sets Of
hypotheses are not apparent from the table because they
concern hypotheses in the Cherryholmes-Shapiro inventory
not comparable to predictions Of exchange theory.

1. Cherryholmes and Shapiro include propositions
relating ideology and ideological consistency tO roll-call
voting:

The political predispositions or ideologies Of
congressmen are related to roll-call voting even
when the effect Of party, region, and.constituency
are controlled.

Legislators tend tO be very consistent in their
roll-call votes from one Congressional session to
the next.80

2. The Cherryholmes-Shapiro model does not include

the influence of.the.executive's position in the manner

 

8oCherryholmes and Shapiro, Representatives and
Roll-Calls, p. 35.

77

hypothesized in hypothesis 1.4. It does include proposi-

tions relating the presidential position to voting behavior

through communication patterns.81
3. Cherryholmes and Shapiro hypothesize high inter-

82 On the

action rates among legislators Of the same party.
basis Of exchange theory we hypothesize vote trading among
party members will be high if they represent similar
constituencies.

4. Cherryholmes and Shapiro list a series Of
propositions that generate conformity Of position among
leaders and the organization leaders use to disseminate
information. The Cherryholmes-Shapiro model funnels leader-
ship communication tO party members predominantly through
the regional whips in the House Of Representatives.83
Although the party leaders in a legislative chamber the
size Of the U.S. House Of Representatives cannot communicate
directly with the rank-and-file on every roll-call, the
regional whips are only intermediaries, emissaries Of party
leaders. For greater generality, especially since most
legislatures in the United States are not even half the
size Of the U.S. House Of Representatives, we ignore such
institutionalized communication networks as whip organiza-

tions and hypothesize direct dissemination Of information

through the elected party leaders.

 

81 82

Ibid., pp. 69-710 Ibid., po 77.

83ibid., pp. 74-75.

78

.Aom .ov mnoeeemom Heoemoaoooe
oEoHexo onee 0e mo>eeeenomonoon
How honoonoe one omeoeooo mooo enn
mnHeO> mesem oe ooeeaou on oe Boom
eon mooo mo>eeeenomoemom mo omnom
one ne mmono>eeeeomnoo menemluoenH

.Aom .ov mnoeeemoo HeoemOHoooe
oEoeexo onexee none meoeeamemoa
meanenne one moeo> menem HoSom

oe ooeeHoH on oe monoe moeneeamemoa
oeeem ne mmono>eeeeoofioo meeeoleoenH

.Aem .ev Hoe>nemn Hemoueeou oe canteen
one meoeeemeo o>eeeamemoa mo moeeme
leoeoeueno OenmeemOEoo one Heoeeeaom

.iem .ee

omeem HHeOIHHOH one ee weeemoa weeem
mmoH oonooem Oe onoe nemnoeem heeem
MO Heoemwee meoenemeo o>eeeamwmoq

e.Aom .mv

eoweemeo one ooemeno wheemee no

we oononamne emoxeo3 one one eoeeemeo
one neneHB meeemwe nO one oononemne
emoeeoem mo meoee one ”cemeooom onmwe
on 0e monoe Hoe>enon HHeOIHHOH one
wononeeemnoo noozeon menmnOeeeHoH one

.nemeen ooea e
an ooeooao me3 on me nene Homnouem we
memoeoene wononeeemnoo mo euoomnm men

.ooeooHo >H3ouuen me3 Hoeeamemoa one eH

.menoneeemnoo man no Hozom

enoeeeeon one neemeuso noes; done
seamemoa momommo one menoneeemnoo men
mo Ho3om Heoeeeaoo one moonenno noen3
nOHeeHmHmoH meeooonm Hoeeememoa one

.memosoene Oeaonooo neone oe
HeOHEene we noenz noeeeamemoa monommo
one menoneeemnoo men mo memouoene
Oenonooo one nee3 enonsmnoo me noenz
nOeeeHmemoH meeomonm Hoeeamemoa one

.menoneeemnoo men
mo memoeoene one oe HeOeEene we noen3
nOeeeHmemoH momoooo one menoneeemnoo
men mo emoeoenw one ne we noen3
noeeeamemoa menoomnm Hoeeamemoa,ona

.m.H

.~.H

 

ouemenm one monaonweuono

muoone omnenoxm

 

.auoone omnenoxm none mnoeeonoon
one >eoeno>nH ouemenmnmoEHoanuono one ne mnoeeemomosm mo nomeuemnooul.a memes

79

.Amm .mv monmmw
enosommeo mmoeoe weaewoa mo moonmoo
enoeoemeo moneEEOO noeeeeeeeee meuem

.Amm .mv oneeo> HHeOIHHOH Oe
emenemn seamen we aoeeneeeeen spend

flmoe .o .mmnmm .mon oeoooe Eenmoeog

Oeemenm one mofiaonmenonu

.monmme
enoeoweeo mmonoe moeHe> meHemOH xenon

.mneeo> o>eeeamemoa Oe
penned“ seamen we aoeeneeeeen spend

.nOeeOE ooeomnommlmeeem
e we noeeofi e no noeeenemoo one Oe
moenneeenoo nOeeemom m.o>eenooxo one

.osneeamemoa one oe enomoum
mone mnOeeOS no wenem oeemoooo one no
mHonEoE omoomo one oeneeememoa one oe

enomoem hone mnOeeofi no heeem neone
mo meonfioa eeoomnm mHonEoE meeem

.mnOemeooo
omone no moeo> Heone nefl3 HOmnOQm
hone mnoeeofi esommnm mHOeeHmemoA

.o>ee
Inooxo one we ooeeomonm nOHeeHmwmoH
momommo Hoeeamemoa one .eoeeemeo men

ne mauoom oosem o>eenooxo one we xo>ee

Inooxo one >n ooeuommnm noeeeamemoa
meuommnm Hoeeamemoa one .eoeeemeo

men ne Hnmmmooonm me3 o>eenooxo one NH

mnoona omnenoxm

.m.m

.N.m

.m

 

 

.oonneenooll.a manna

80

.Ame .mv moeo> aHeOIHHOH

no mnonEoE neone mnofie noemonoo
noen oonooem eene mono ooe>osm
eon 0o mooeeeenoo o>eeeamemoa Han

.Amw .mv meson o>eeeamemoa one no

mHonEoE mnoae mneeo> nee3 ooeeHoH

IHOO eaomeo>ne on Oe monoe meonEoE
ooeeeEEOO mnone noemonoo mneeo>

.Amv .mv omeem

HHeOIHHOH one ee oo3OHHOe on 0e
onoe eene meonEoE Oe mono mneeo>
o>em mooeeeEEOO ooeeemoene HHo3

.Amm .ee meme

IEoE oeee one nnee nene woOHoooH
one nOeeeNenemHO weeem one oe

HeNOH oHOE on oe onoe meooeoa meeem

.Amm .mv moonoeone enoeaem an
oomommo non3 oomeoeooo me one monone
:Hemnoo one noemoe we nonm .meoeoem
enoeaem an ooenommnm we meeem non3
oomeoeone on Oe monoe heeeeoa eeuem

.oHneeHmemoH one ne omemmem wee emneeme

moeo> one .nOeeeonoEEoooH oHneHO>em
e monommo noeeememoa one .mnoee

.1enen nee neeneeae Heep pene>du one on

.oeneeememoa
one ne omewmem wee How one noeeeonon
Isooon oHneeO>em e How oneeo> en ee

mesomonm Hoeeamemoe one .mnoeeenne men
moeosoum.ooeeennoo ne oome>oe Haen e we

.ooeeeE
:EOO men an oonooemnoo maaen one
ne emoeoene nmen men Hoeeamemoa one

.oHHM
lonelnnee one nene meaemoa heHeQ
noeeoum eenenxo muooeoa xenon one

.memoeoene mononeeemnoo

neeB meoeamnoo ee non3 oonoxeor

on Oe monoe one memouoene mono
Ineeemnoo neez enonemnoo me ee non3
ooOHOeneoH on oe monoe weHemOH xenon

.~.m

 

oeeoenm one monaoneeuonu

esoone omnenoxm

 

 

.oonneenOUII.H mnmne

81

.Aow .mv mnonenmom ooneeeee ononexo
mmoe nena meenoe>none eene oononaenn
oe oennemoomnm mmoa one mnoeenmom
ooneeeee oEonexo onoe nen3 maenon>nonn

.Amm .mv oononaenw mo meomnee on oe
onoe wnonmeooo mneeo> onomemom on3
omone one mnoononaene on oe haonea
onoe one mnoemnooo unneo> waneo oxen
on3 moononoeono omnoenn nens mnomnom

.Aem .el

eonaenoo monnmmonm enonaem non3 nene
nonaneo oer on oe onoe mnonmnooo
.enomno>noo one mnnxenlnoemeooo
Heoeeeaom oe enonHem monnmmonm nonz

.Amm .ov mnoneemom oeenooon oaon
on3 mHenon>nonn one nene moononaenn
InoennOO oe oaneemoomnm mmoH one
mnoeeemom oEonexo oeon onz mnOmnom

.iee .al .mnnee

nOneemoomnoonm one onenno nene

one nenz noeeeenonenoo men onenno
mmoao>oo on eene noeeemommeoonm
one e0 noeeonnm e we HHHn no>em

e no nonmeoHHOO men no>o oononamne
Hewenoeom m.o>neeenomonmon noem:

.>Hnmnn onae> hone mnonmeooo no>O
Honenoo noeeonm now nnneon nn nOnm
Inooo eene no enoomnw omnenoxo oHeenH
nOemeooo o>eeooHHOO e onae> onB omone

.emonoene noeeonm no we nonnB
nonmnooo o>neooHHOO one no none
Inmom neone e0 enommnm now nnneon nn
oenenee> mmoH we noenz e0 onooeno
one mnOnmnooo no>O Honenoo omnenoxo
nonmnooo o>eeooHHOO e no onooeno

one now onHe> eeonm o>en on3 omone

.no3om Heoneneom now momnenoxo

mnenefi ne Ben Oe men ee onHe>

one Oe HenOnenomonm we noeeamnmoa e
0e nOnmeooo o>eeooHHOO e e0 onHe> one

.m

 

onemenm one monaonennono

enoone omnenoxm

 

 

.UOaneGOUII.H mamfle

82

.loe .ae naoeenoeaaesoo no memen menaoee

Inomonomno e o>nooon mnonnon onennen
nmen one nonneeno ooeenEEOO one

.Aom .ov momeono on noen3 nn
mnoeeeoennenoo one onOE one .nooeoH

emeomnm one no game one enema: wee

.Ame .av manneonemne can mannen we
Haea en eene Om noemMm one noemnnnnoe
oe mnooeoe neone eoomxo wnOeeHmemon

.Ame .de
nOneeEnoene mo noneennnommeo one
we eonaenoo o>HOmon one monmme mnooe

oe mnooeoH neone eoomxo mnoeeamnmon

.Amm .ov mnoeoenoene nmnn one
mnoononeene nmnn noozeon onnmnoeeeaon
o>nenmom .nmen e on oe monoe onone

.Aom .mv on> men onennoo

:nOO oononemnn nonennm oe oenneooo
none on oe eaoxeann we nOneooneo
oEem one nn on> men mneononeene
mononeeemnoo one heneo me none monnm
Imonm onOE no O3e nen3 nOeeHmemoH n

.nan one maeenoamn Anemmeenesoo one «0
mnonnon one one Haen one no oEooenO one

ne ooemonoennnn omone nooseon ocean
monee moonnOmon o>eeeHmnmoH no oneoene

.mnooeoH wenem one nononne mnonm
leooo Heoeeeeom oneeoomme moonnOmon
now moeo> omnenoxo mnoeeamemoe oEOm

.nOnenHeOO mnnnne3 Ennenefi e
nnem oe eoEoeee noeeon e no onnaeem
one ooemmeo neon mo menonomonm

 

onemenm one monaonennono

enoone omnenoxm

 

 

.emscneaooun.n memes

83

.Ammma aflaennozlmnnom

 

unnoe 3ozv mHHeOIHHOm one.mo>eeeenowonmom .Oneoenm Hoenonz one monaonennono
OoHo nn mnOeenmomonm omone mo noneeOOH one oononomon mnonnnn omeme

 

.Amm .mv nonmeoeaoo men one
enooemonm one neez mnoeeeonnnnnoo.mnn

.mnoemeooo

noneo mo oEooenO one oneeooeee
moonnOmon now oeo> men omnenoxo Oe
noemeooo e moxen on non3 oEooeno
one ne .ene we .emonoene n30 men

 

 

 

ne menonenoo on eene onO ooeno>e ennoooe Oene moxee nOeeHmnmoH one .H.NH
one one noeeemommeoonm Hennmnno
men noozeon me3umeen nOeenmom e .oeo> own now omnenoxo ne moonnomon
oe Henn one one3oe nOeenmommnoonm mo mfinoe ne emon one Enn mnommo
Hennmeno men momneno Haen e no on3 noeeamemoa ooemonoene ne nee3
moeeoennEEOO on3 o>eeeenomonmon n on> men momnenoxo nOeeHmemoH one .NH
.Eone ooeooeo nonna
mononeeemnoo one no noneemoonoo
nw Eone Oe neeenem mnOeeHmeooe
.Ame .mv mo>eeeenomonmon one oEooenO one ne ooemonoenenn
onone noneeonnnEEOO mo oonnom mnOeeHmnmoH noozeon ooeHm moxee
e we mononeeemnoo mo meeneeefiem moonnOmon o>neeemnmoa mo mnnoene .HH
.menofinmemme ooeenEEOO neone
onenm on3 mnoeeameooe ooemonoenn one
.Ame .mv menonnmemme nceeon one e0 oEooenO one ne ooemo
ooeeeEEOO neone onenm on3 nonmeoaaoo Inoennnn onone nooseon ocean monee
nee3 oeeoennEEOO mo>eeeenomonmom moonnOmon o>eeeameooe e0 annoene .OH
Onnmenm one monaonhnnono enoone omnenoxm
.oonneenooll.a mnmne

CHAPTER III

THE SIMULATION MODEL

In this chapter we move from our theory to the
simulation model. As we mentioned at the conclusion Of
the first chapter, the validation Of a theory by the use
Of’a simulation model engages a complex relationship among
behavior, theory, model, and model output or predictions.
Our main considerations in this chapter are the links
between our theory and the model.

In this project, we simulate behavior in the U.S.
Senate. For purposes Of operationalizing the model, we
will explore the links between the model and behavior.
These include selection and measurement Of variables and

parameter estimation.

Theory-Model

 

Our model, SIMEX, is a revision Of the Cherryholmes-
Shapiro model. The revisions are baSed on the deductions
from exchange theory in a two-party system, criteria Of
explanatory power, and pragmatic considerations Of the use
Of the simulation as an instrument in comparative research.
SIMEX models the decision-making calculus Of the legislator

as he decides tO vote for or against motions affecting

84

85

collective decisions on the basis Of exchanges he makes
within the legislature and with extra-legislative elector-
ates.

In the predisposition phase, the legislator assesses
information about the bill and its consequences for his
political ambition. In the interaction phase, the legis-
lator who remains uncommitted after the predisposition
phase surveys his trading Options according to hypothetical
bargaining patterns and decides which coalition he will
join. These bargaining patterns take into account the
strategies Of both buyer and seller insofar as these
concern bargaining and Opportunity costs. That is, the
vote-sellers' Options are constrained by the preferences
Of vote-buyers.

Exchange theory is based on the notion that the
legislator is motivated by a desire tO be approved by
extra-legislative constituencies. Communication to the
voter Of the outcome of all motions, however, is not
uniform. In general, passage votes receive more media
coverage than other motions. Since the final outcomes
are, then, likely to be more salient for constituents
than amendments and procedural motions, we expect the
simulation based on exchange theorvaill predict passage
motions more accurately than amendments and procedural

motions.

86

Model hypothesis 1. Predictions Of the simulation
will be more accurate with respect to final votes
on legislation than on amendments or procedural
motions.

Simulations can be graphically represented by flow
charts which diagram the progression Of changes that take
place in the Operating model. We will demonstrate the
operationalization Of hypotheses in the model with flow

charts in the following section.1

Predisposition Phase

 

Party exchange are represented in the flow chart
in Figure 2: hypothesis L-3 for all legislators and L-4
for party leaders.

Figure 3 represents hypothesis L-l and its subsidiary
hypotheses, the generation Of legislators' interests in
bills according to the characteristics Of their constitu-
encies.

For the purpose Of explaining roll-call voting in
the Congress Of the United States, we shall extend our
definition Of constituency tO regional and state interests.

Occasionally, programs benefit particular states

more than others. It is not, then, surprising that Truman

 

1The flow charts are not necessarily in the same
order corresponding statements appear in the program;
however, every step is illustrated in a figure appearing
in this chapter. Hypotheses and corresponding record
numbers with a listing Of the program appear in Appendix A.

87

 

 

Enter Bill Enter legislator
Pred. = 0a

 

 

 

 

 

 

 

 

 

Is bill supported
by either party

 

 

no

 

 

 

yes

 

 

Is bill supported
by legislator's
party?

 

 

 

yes no

.1 .1\

Is legislator Is legislator i
a party leader? a party leader?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

yes no
nO yes

+1 -1

 

 

 

 

 

 

1

Figure 2.--Flowchart: Party.

aSigned integers change the value Of the predisposi-
tion.

88

 

 

Is constituency interest
for the bill?

 

 

./

I

 

Isvconstituency

 

the bill?

interest against‘

 

 

1

yes

-1

 

 

Is constituency
competitive?

 

 

 

 

no yes
l

/_-'-l

 

 

yes

+1

 

 

 

Is constituency
competitive?

 

 

 

I [

yes no

 

 

 

 

 

 

 

I

Is executive for
the bill?

 

 

 

 

no

1

 

 

Was executive's
plurality in
legislator's

district in the

election high?

 

 

 

no yes

in

 

 

med?

no \y-
l
e 45

 

 

 

 

7

yes

I

 

 

Was executive's
plurality in
legislator's

district in the

last election
high?

 

 

 

med?

 

 

 

 

yes

B.
n n

 

 

Figure 3.--Flowchart:

l

Constituency.

89

finds that in the House Of Representatives state delegations
increased in cohesion as party cohesion declined on roll-
call votes.2
From hypothesis L-l, we would argue that the racial
composition and economic status Of Southern districts should
elicit a sympathetic response to social and economic problems
Of blacks from Southern representatives. However, the
greatest Opposition tO the economic, social, and political
advancement Of blacks springs from the "black belts,"
areas in the South where forty per cent and more Of the
population is black.
Southern sectionalism, Key explains, lies deeply
rooted in the history, the economy, the social structure
and the political culture Of the South.
The South's heritage from crises Of the past, its
problem Of adjustment Of racial relations on a scale
unparalleled in any western nation, its poverty
associated with an agrarian economy which in places
is almost feudal in character, the long habituation
Of many Of its people tO non-participation in political
life--all these and other social characteristics both
influence the nature Of the South's political system
and place upon it an enormous burden.3
The South, moreover, has two related attributes

which make behavior Of its representatives somewhat unpre-

dictable. As a consequence Of the Reconstruction, the

 

2Truman, The Congressional Party. Cherryholmes
and Shapiro also hypothesize the effect Of state delega-
tions and state interests, Representatives and Roll-Calls,
p. 39.

» 3V. 0. Key, Southern Politics (New York: Knopf,
1949), p. 4.

90

white South votes overwhelmingly Democratic even though

it has a predominantly rural population. Consequently,
Southern Democratic districts, while not competitive, are
atypical in some respects of Democratic districts nationwide.
Furthermore, lack Of competition is accompanied by fac-
tionalism within the Democratic party.4 Consequently,
Southern representatives are relatively independent Of

their legislative party and Of many Of the groups residing
in their constituencies.

The West has also been characterized by its own
sectional interests, although they are not as problematic
as those Of the South. The West's problems have been
primarily those Of the frontier, Of economic and pOpulation
growth; and these have been congruent with the growth Of
federal power and services. Mayhew finds that western
issues, mostly problems related to resource development
such as water reclamation, were supported by Westerners
and also by the Democratic party and its elected leader-
ship.5 Furthermore, regional cooperation in the Senate
is fostered by an organization which monitors communication
about legislation Of interest tO the West, the Conference

Of Western Democratic Senators.6

 

4Ibid., especially Chapter 14.

5Mayhew, PartnyoyaltyiAmong Congressmen.

6Neal Maxwell, Regionalism in the United States
Senate: the West (Salt Lake City: Utah University, 1961).

 

91

Figure 4 diagrams the extension Of constituency
interests to the state and regional level for the purposes
Of explaining behavior in the Senate and the House Of
Representatives.

The consequences Of the legislator's choice Of
specialization and sponsorship, hypotheses L-2 and L-5
are represented in Figure 5.

Figure 6 diagrams the separation Of legislators
into protocoalitions supporting passage and defeat Of the
motion and those less interested in the outcome Of the
collective decision.

Discussion.--The predisposition phase Of SIMEX is

 

identical tO that Of the Cherryholmes-Shapiro model with
the exception Of the relationship of legislators' positions
tO the executive program, hypothesis L-l.4 diagrammed in
Figure 3. Our model does not include a prediction Of the
effects Of personal ideology including in the Cherryholmes-
ShapirO model. Cherryholmes and Shapiro include the
following propositions in their inventory:
The political predispositions or ideologies of congress-
men are related to roll-call voting even when the
effects Of party, region, and constituency are con-

trolled.

Legislators tend to be very consistent in their roll-
call votes from one Congressional session to the next.

 

7Cherryholmes and Shapiro, Representatives and
Roll-Calls, p. 35.

92

i

 

 

Is the legislator's region
for the measure?

 

 

 

 

 

 

 

 

 

 

 

 

 

 

no yes
Against? +1
yes no
-1
Is the legislator's state for

 

the measure?

 

 

Figure

 

nO yes

H

.L

 

 

 

4.--Flowchart: State and Region.

93

1.

Is legislator
a sponsor?

 

 

 

 

 

 

yes nO

+2

 

 

 

Is legislator on a committee
reporting the bill?

1

yes no

 

 

 

 

 

Minority reporter?

 

 

 

 

yes no

51 +1

 

 

 

 

 

Figure 5.--Flowchart: Sponsorship and Committees.

94

.nOneeHnsnm one nn on> noneon one no ooeo> Meaeneoe on3 onone manoe

.omenm noneemommeoonm mo nonmnaonooll.m onnmnm

e

omnenoxo o>eeeamnooa \\
lenenn now omenm

 

 

noeeoenoene noenm%:aez= oonnnnoeoo oeo> \ /e..eoe= oonnnnoeoo oeo>\\

 

 

 

on no» mo»

 

 

 

mm: Oe Henoo no nene

 

mmoa nOneemommnoonm mH

 

m~+ oe Henoo no

0G III Ewe—u. Hwflmwhm COHflHmOQmHUOHQ mH

 

 

 

 

 

 

 

 

//noneemommnoonm oeennoaeok\\

 

95

These propositions are operationalized by relating the
"memory score," the proportion Of occasions on which the
legislator voted in favor Of an expanded federal role in
the past session tO a propensity to vote for or against
the simulated motion which is described as for or against
expansion Of the federal role.
. . . roll calls from the previous session Of Congress--
the Eighty-Seventh--were used tO represent personal
ideology or memory . . . . A representative's ideo-
logical predisposition was calculated in the model on
the basis Of the number Of times he voted in support
Of foreign affairs or for an increase in the federal

role as a proportion Of all such bills on which his
vote was recorded.8

We have noted that politicians may have personal
;preferences about the outcomes Of collective decisions
.apart from their use Of legislative action as a political
:resource. We assume, however, because Of the game played
by legislators, that these preferences are lower in
priority relative to their desires for political power.

While we concede the likelihood that the legis-
lator's personal ideology may be marginally reflected in
voting behavior, its measurement for the Cherryholmes-
Shapiro model is questionable. The measure Of "memory"
or ideology is highly related tO the output, i.e. votes
in a particular issue area. Thus, its interpretation is

Prohlematic; it may simply reflect the combined effect Of

 

 

8Ibid., p. 52.

96

all the variables which influence the vote, including
ideology in an unknown proportion.

According tO Cherryholmes and Shapiro, the "memory
score" reflects ideological consistency over time. Consist-
ency Of action, however, is not implied by exchange theory
except insofar as past behavior has been rewarded and insofar

as future ambitions are similar tO past ambitions.

Interaction Phase

 

When a legislator interacts with another legislator
in the interaction phase Of the simulation, he in fact
takes note Of the second legislator's predisposition, defined
as the value that passage or failure Of the motion has for
the second legislator (hypothesis L-6).

Figure 7 represents hypothesis L-8, the intermediary
role played by party leaders in the exchange process.

Figure 8 illustrates the operationalizatiOn Of the
conjunction Of hypotheses L-9 and L-ll, the tendency Of
the legislator to interact with members Of the committee
reporting the bill who represent similar constituencies.

Figure 9 depicts the legislator's tendency to
interact with legislators who share committee assignments
and represent similar constituents, the conjunction Of
hypotheses L-10 and L-ll.

Figure 10 illustrates the operationalization Of
hypotheses 12 and 12.1, the legislator's assessment Of the
value Of his vote in exchange and the value of the motion

for his own political interests.

97

J.

Each undecided legislator L______

considers another legislator
until house membership is E
exhausted.

 

 

 

 

 

 

Is second legislator Of same party?

 

 

 

 

no yes

 

 

Is he a party leader?

 

 

 

no yes

/ Note predisposition k—

 

 

 

 

 

 

Figure 7.--Flowchart: Interaction with Leadership.

98

J.

 

 

Is he a member of a committee
reporting the bill

 

 

 

 

Does he represent a
similar constituency?

 

 

 

 

 

no yes

 

 

 

/Note predispositionV

 

 

 

 

V

Figure 8.--Flowchart: Interaction with Members Of
Reporting Committee.

LL

 

 

Is he a member on one Of the
same committees as the first
legislator?

 

 

 

 

no yes

Does he represent a
similar constituency?

[ |

no yes

/[NOte predisposition \r——

Figure 9.--Flowchart: Interaction with Members
Sharing Committee Assignments.

 

 

 

 

 

 

 

 

{I

99

.noneeannnm one nn on> noneos one no oon> eaaeneoe on3 onone Nance

.omenm noeeoenoenH no noemnnonoo "enenosonmnn.oe onnmnm

/ on> 02 \ Memen on> \ /eeo>. oeo> \

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

on mom mom
o>neemon , on o>neemom
eenmon we eenmon mH
.omenm

noneoenoene one ne ooeon men
on onoeeemommeoonm mo omeno>e
one one omenm emnee one no ono
one ee noeenmommnoonm mo Enm
"noeeemommeo Henne oeeHnOHeU

 

100

Discussion.--There is a limited relationship between

 

the interaction phase Of SIMEX and the communication phase

Of the Cherryholmes-Shapiro model. Every hypothesis incor-
porated in SIMEX is in the Cherryholmes-Shapiro model.
However, the interaction phase Of SIMEX is far simpler than
the communication phase in the Cherryholmes-Shapiro model.

The communications phase Of the Cherryholmes-Shapiro model

is very complex and highly structured by party considerations.
In the U.S. House Of Representatives, it was apparent that
the model overdetermined party influence except for Democratic
administration bills.9 Many Of the influences Operating in
the communications phase may overlap one another and Often
they repeat influences in the predisposition phase: party,
region, constituency, state delegation, presidential posi-
tion. Interpreting these as communications at some stage

Of the decision-making process is probably correct, although
interpreting them as influences on voting independent from
those in the predisposition phase and independent Of each
other may be in error.

The interaction phase Of SIMEX is deterministic,
while the interaction phase Of the Cherryholmes-Shapiro
model is stochastic. Because Of the complexity Of the
communications phase Of the Cherryholmes-Shapiro model,
parameter estimation, the designation Of probabilities

Of interaction, is difficult. Cherryholmes and Shapiro

 

91bidol PP- 109-110.

101

based their probabilities on an apportionment Of the average
number Of communications expected for non-leaders, leaders,
and the President.10

The success Of the communications phase in predicting
roll-call votes in the House Of Representatives lends validi-
ty to these parameters. However, when the model is applied
to another legislature, the question Of Optimum parameter
values is reOpened. Readjustment for the Senate is relatively
simple: the values for intraparty and interparty communica-
tion are altered to allow the same average number Of contacts
as in the House Of Representatives; leadership communica-
tions are changed in consideration Of the absence Of regional
whips. These adjustments, however, are the minimum for
comparability with performance on House Of Representatives
data. We may still question their interpretation and
whether they are sufficiently precise.

Joseph Hanna suggests that stochastic models have
advantages over deterministic models. Of these, the
recognition that simulation models contain lgei_g£
ignorance--either neglected causal factors or errors in
measurement--is among the most important.

. . . the random elements occasioned by both loci
Of ignorance--neglected factors and measurement
errors--are explicitly included in the [stochastic]
model. In general, the less precise one's measure-

ments, and the more significant the neglected
causal factors, the greater variance the resulting

 

loIbid., pp. 69, 74.

102

probability distribution will have. In the limit, if
nothing is known about the causal factors influencing
behavior, the model should attribute equal likelihood
to all possible responses.

A stochastic model can generate a probability
distribution Of outcomes rather than a single determined
prediction. In a model which has two possible outcomes,
for example, a stochastic model can generate a probability
for one outcome, pl; the probability Of the other, p2,
equals (l-pl). Alternatively, we may repeatedly run a
stochastic simulation which generates a simple prediction
and calculate the proportion Of times each outcome occurs
to Obtain the probabilities Of each outcome. With this
information, we cannot only reject a model outperformed
by a random prediction (one which generates a probability
Of .5 for each outcome), but we can also make better
comparisons among model instances, versions Of the same
process with different initial conditions. We would, then,
prefer the model or model instance which generates the
highest probability for the Observed outcome.

Unfortunately, the Cherryholmes-Shapiro model does
not generate probabilities Of the predicted outcome. Even
though it is stochastic and the outcome may vary across

runs Of the same model instance, it generates only a simple

prediction. Furthermore, the second method Of Obtaining

 

11Hanna, "Information-Theoretic Techniques for
Evaluating Simulation Models."

103

probabilities involves a prohibitive number Of replications
Of the model.

Although the SIMEX model is deterministic, we can
regard the likelihood that an individual will vote in the
way predicted as an increasing function Of intensity Of
preference. Those legislators with high predispositions
should, then, include prOportionately fewer errors in
prediction than those with low predispositions. The
simulation model predicts the votes Of those with high
predisposition values at the end of the predisposition
phase. Hence, we anticipate,

Model hypothesis 2. Those legislators whose votes are
determined at the end Of the predisposition phase
will be predicted more accurately than those
predicted at the end Of the interaction phase.

The referent process for exchange theory is bar-
gaining and exchange. The interaction phase Of SIMEX models
the search for information on alternative exchange possibi-
lities by those legislators who are undecided at the end
Of the predispoSition phase. Their information consists
Of the predispositions Of those with whom they interact,
defined as the value Of a vote for their positions. The
resources exchanged for votes, sidepayments, are not
present in the model; rather, they are a hypothetical

construct assumed tO be proportional to the value Of the

104

motion generated in the predisposition phase for each
legislator.12

In the final step, the values noted by the un-
decided legislator are averaged. This approximates the
value Offered by one member Of the protocoalition Offering
him the most in terms Of resources. Averaging makes it
unlikely that a number of interactions with legislators
having low predispositions will change the legislator's
vote from the direction Of his predisposition at the end
of the first phase Of the model.

Although the interaction patterns are based on
the strategies Of both buyers and sellers, only the sellers
Of votes seek information in the model. The simulation
models nO information feedback which would enable members
Of protocoalitions tO gauge the size Of their coalition
and tO limit sidepayments tO those necessary to win with
the least expenditure Of resources (hypothesis L-7).

With complete absence Of information feedback on

the size Of coalitions, we expect SIMEX will generate larger

than Observed coalitions. This neglected aspect of the

 

12The value Of a motion for a legislator, a re-
definition Of a calculated numerical value, is an "inter-
vening variable." The resources, not present in the model,
but essential tO its interpretation according tO the theory,
are a "hypothetical construct." For a discussion Of this
distinction, see Kenneth MacCorquOdale and Paul E. Meehl,
"On a Distinction Between Hypothetical Constructs and
Intervening Variables," Psychological Review, 55 (March,
1948)] PP- 95-107.

105

theory, or locus Of ignorance, is present in both SIMEX
and the Cherryholmes-Shapiro model. Since the stochastic
Cherryholmes-Shapiro model produces conformity among
groups within certain probabilistic limits, however, it
builds in a brake on the overestimation Of winning coali-
tions. Hence, we anticipate,

Model hypothesis 3. SIMEX will generate larger than
Observed winning coalitions.

Model hypothesis 3.1. The Cherryholmes-Shapiro model

will generate smaller winning coalitions than
SIMEX.

Model-Behavior

 

Senate Constituencies
In legislative chambers Of small membership repre-
senting a large total electorate, districts Of legislators
are likely to be heterogeneous with respect to interests.
Consequently, the tension to effect the wishes Of any
particular group Of voters is lessened. It is likely,
then, that constituency interests as defined by demographic
characteristics will be less evident in the voting behavior
Of smaller legislative chambers than in the behavior Of
larger ones. We predict,
Model hypothesis 4. The constituency hypothesis will
have less positive effect on predictive accuracy
in the Senate than in the House Of Representatives.
Corroborative evidence.--Clausen and Cheney discover

that U.S. senators respond more to party relative tO

constituency, but they are also more independent Of party

106

constraints on voting compared to U.S. representatives.
They suggest that the greater heterogeneity Of Senate
constituencies and the ability Of senators to build a
reputation independent of their party through the media
damp the influence Of both party and constituency.13
The senator, according tO Matthews, is more free
tO define his constituency than representatives from
smaller, more homogeneous constituencies.
. . . a constituency as large as a state can be
represented in many different ways. Up to some point
the senators are free to choose which groups and
interests within their formal constituency they will
champion. The senators' party, in two-party states,
largely commits them to one approach in satisfying
their constituents, yet this is not the case for the
men from one-party or modified one-party areas . . .
Senators from socially and economically heterogeneous
states also enjoy greater maneuverability along with
greater risks. 4
Senate Interactions
The hypothesized constraints on bargaining rest
on the assumption Of limited availability Of information
on possible exchanges. The smaller the legislature,
however, the more likely such shortcuts will be necessary.

In smaller legislative chamber, then, we can expect more

idiosyncratic patterns Of vote trading.

 

l3Clausen and Cheney, "A Comparative Analysis Of
Senate-House Voting on Economic and Welfare Policy 1953-1964."

14
237-238.

Matthews, U.S. Senators and Their WOrld, pp.

107

Model hypothesis 5. The Senate will exhibit less
predictable patterns Of interaction than the
House Of Representatives.
Corroborative evidence.--Truman corroborates this
hypothesis by comparisons Of bloc patterns in the U.S.
Senate and House Of Representatives. U.S. representatives,
in contrast tO senators, appear dependent on internal struc-
tures: their patterns Of bloc voting are more regular than
those Of senators. Truman asserts that because Of its larger
size, the communication system is more standardized in the
House than in the Senate.15
Clausen and Cheney contrast the Senate and the House
Of Representatives in the dimensional structure Of their
voting: the Senate exhibits greater complexity than the

House.16

Variables

 

Each legislator in SIMEX is represented by eighteen
variables: identification number, party, state, region,

nine constituency characteristics, leadership position,

17

and up to four committee assignments. All variables

 

15Truman, The Congressional Party, pp. 194-195.

 

16Clausen and Cheney, "A Comparative Analysis Of
Senate-House Voting on Economic and Welfare Policy 1953-1964."

17In order tO operationalize the more complex com-
munications phase Of the Cherryholmes-Shapiro model, more
attributes are required: regional and assistant whips (if
available), region as defined by the whip organization,
chairman and the two ranking members Of each party for each

108

except constituency characteristics are measured at the
nominal level and can be identified by the assignment Of
a numerical value tO each classification. (Appendix B
lists the codes assigned tO each value Of each variable.)
Demographic constituency variables, per cent urban,
rural non-farm, rural farm, non-white, owner-occupied
dwellings, white collar employment, and median income,18
are collapsed into ordinal scales according tO Observed
clusters in their distribution across districts.19 In
the absence of parameter estimation routines, higher level
measurement for these variables is not useful.
Political constituency variables were collapsed
into ordinal scales according tO breaking points which
seem reasonable in the light Of the information they were
intended to convey to the legislator, electoral competition

and the executive's plurality in his district.20

 

committee, and seniority groups. Cherryholmes and Shapiro
also utilize the "memory score" as an attribute Of the
legislator utilized in the predisposition phase.

18Cherryholmes and Shapiro did not include median
income, but they dO include per cent foreign stock.

19Source: U.S. Bureau Of the Census, U.S. Census
Of POpulation: 1960, Vol. 1: Characteristics Of Popula-
tion (Washington, D.C.: U.S. Government Printing Office,
1964).

20Source: Governmental Affairs Institute, America
Votes: 1962, VOl. 5, ed. by Richard M. Scammon (Pittsburgh:
University Of Pittsburgh, 1964).

109

Bills simulated in SIMEX are identified by party
sponsorship, committee reporting the bill, and individual
sponsorship. Minority reporters were coded on the bill
to operationalize hypothesis L-S. Descriptions of bills
in the Congressional Quarterly Almanac are used to designate
for each motion the constituency groups and states which
would benefit, and regional interests. The executive's
position, obtained from the C.Q. Almanac, is coded for or

21 When the executive is against a

against the motion.
motion sponsored by a member of his own party, party Spon-
sorship is coded "without party" to operationalize hypothesis
L-3.l.

The actual votes on roll-calls are arranged across
columns of cards following bill attribute cards. The actual
result is used only for assessing accuracy of the predic-
tions of the model.

Figure 11 and Figure 12 illustrate the format on

attribute cards for legislators and bills, respectively.

 

. 21The Cherryholmes-Shapiro model also uses informa-
tion on the president's party, his regional and state
base. Although not defined as such, the federal role
orientation of the bill is utilized as the president's
position on the motion. That is, on the roll-calls simu-
lated in the Eighty-Eighth Congress, Kennedy and Johnson
always favored expansion of the federal role, and opposed
its diminution. The C.Q. Almanac confirms that this
congruence is in most cases justified for the Eighty-
Eighth Congress.

110

 

 

mucoacmfimmm
mouuwafioo

coflpwmom mflnmumomma

moflumfluouomumno
wocmsuflumcoo
coammu

mumum
manna

r// coaumofluancmofl

Figure ll.--Format

 

Attributes of Legislators.

 

 

mumuuommn
wuflnocflfi

mcfluuommu
mmoupflfieoo

pmpmmnoacw
mmflocmsuflumcoo

woummuoucw mcoflmou

Umummumucfl mmumum

maanOmcomm muumm

Card 1

 

 

 

 

cofluwmom
m.ucooflmmnm

mHOmcomm

Card 2

 

 

Attributes of Motions.

Figure 12.--Format

lll

Parameters

 

We have adopted the parameters used by Cherryholmes
and Shapiro in the predisposition phase of SIMEX. These
are reasonable in the light of the value which separates
the highly interested from the less interested at the end
of the first phase of the model, a predisposition (positive
or negative) that equals or exceeds 2.

In his simulated calculus, party alters the legis-
lator's predisposition by a value of l for rank-and-file,

2 for leaders. Constituency interests change the predis-
position by l, by 2 if the constituency is competitive.

State and regional interests alter the value of the pre-
disposition by 1. Sponsors move to a strong prOpensity

to agree; the majority of reporting committee(s) to moderate
agreement, the minority of reporting committee(s) to moderate
disagreement.

The executive's plurality in the legislator's
district generates a tendency for the legislator to agree
if it is moderate or high, to disagree with the executive
if it is low. Our assignment of parameters gives the
executive's position an advantage in the outcome of the
bill. This aspect of the parameter values is somewhat
questionable. However, we anticipate that because of the
advantage the president has in media coverage to popularize
his program, the legislator will tend to favor the presi-

dent's position.

112

In view of the findings of Jewell and LeBlanc (see
page 74, above), we chose urbanness as the relevant constitu-
ency dimension in the interaction phase. The simulation
model, then, should generate cohesion within legislative
parties which represent mostly urban or mostly rural con-
stituencies. Since Senate constituencies are heterogeneous,
any one constituency variable may not adequately delineate
bargaining patterns. Because of its smallness, a correlate
of heterogeneity, we have already predicted idiosyncratic
bargaining patterns in the Senate (see p. 106, above).

Unquestionably, more precise parameters are desirable.
However, since the theory and the model are still in the

developmental stages, these approximations are appropriate.

Motions on the Federal Role

 

We chose twenty-three roll-calls from among votes
on motions affecting the eighteen bills identified by the
Congressional Quarterly Service as those which represented
action on the nature and extent of the federal role in

22 .
Nine passage votes, one vote on a

domestic government.
conference bill, one vote on authorization of appropriations,
and one vote on extension of a program (library services)

to urban areas are considered final votes. Amendments and

 

22Congressional Quarterly Service, Congressional
Quarterly Almanac, Vol. 19: 88th Congress, lst Session;
Vol. 20: 88th Congress, 2nd Session (Washington, D.C.:
Congressional Quarterly, Inc., 1963, 1964).

113

procedural motions included eight amendment votes, one vote
on a motion to consider a bill, one vote on cloture, and
one vote on a recommittal motion.

We purposely chose roll-calls which give a sampling
of types of legislative action and variation in sponsor-
ship. We also chose motions on which conflict was evident,
on which at least fifteen per cent of the participants on

23 We chose those with

the roll-call were in Opposition.
the greatest number of participants where possible. Table
2 lists the motions in our sample and the observed outcome.

Table 3 lists codes used in describing the motions.

Model-Outcome

 

In order to test our hypotheses, the model must
produce outputs by which we can assess the validity of our
theory and our predictions of patterns of correspondence
between model and theory. Basically, the simulation generates
individual predictions. We compare predictions with actual
behavior to compute individual accuracy. We can also
manipulate predictions of individual votes to assess
system-level performance--passage or defeat of a motion,
the generation of protocoalitions, and size of winning

coalitions.

 

23This criterion was not met on two roll-calls
because of the lack of roll—calls on these issues which
did meet the standard.

114

Haflm sums

 

.mmumm mmmmmmm coflumuumflceswm mmfimmm Asmmfiemm ma
HHflm sums
mmuom ngmaucmsm couumuumflcflacm mmammm lemmaemv ea
Haem sums
omuma seasonmsm :oflpmunmflcuswa mmammm Aemmavmv ma
smuvm moamummcoo aha mmmHHoo meflwmm Ammaavfifim NH
hams :oflmcmpxm mmofl>umm sumunflq mmmmm Ammmavmom HH
mauom mummmmm was mmmHHoo msflmmm Ammmaemma OH
chHumwnmoummm Honucoo
Haumm mwfluoensm coupsaaom “mums meow Ammmavwmfl m
mmnae anammma maflanmna umzoaamz HmmHm Amwmavmmfl m
mmuou
vases mmmwmmm moa>umm Hmnoflnmz Hmmam Amomaemoa a
mmuoo
mauve Hmpflesoomu mofl>umm Hmcoflumz Hmmam AmmmHVNOH m
smuom mmmmmmm namesoamsm annoy Hm Ammmaemm m
sends “casccmsm ncmssoHasm gases Hm Ammmfivam a
avnmm mommmmm cowumunommcmne mmmz mm Ammmavam m
nvnvv “masocmfim coflumpuommcmua mmmz mm Amwmavma m
nmlav unmaocofim cOHpmuHommcmua mmmz mm Ammaavma H
mEoouso cowuod .oz Hamo Hones:
om>ummno mo mama mfioz IHHOM.O.U cowpoz

 

.oHom Hammock any mo mommH so maoquZIl.N manda

115

 

 

maums mSMmmma manomfimama Avoaflemmsmm “emmavmam mm
lemmas

mmuoo mmmmmmm mumOAems mmmaamm lemmaemmm mm

hmnms mmmmmmm mpemhm HH>Ho «masmm lemmavoma Hm

saumm usafimcmfim museum Ha>flo mmasmm lemmalmaa om

amuse ucmeeamsm mpnmflm HH>Ho mmaamm Asmaavms ad

mmnas muspoHo museum Hfl>flo mmasmm lemmavmp ma

meuam ucmeecosm mngmflm HH>Ho amasmm lemmaemm pH

.Hmmuflmcoo

pause on coupes mpsmflm HH>Ho mmaamm lemmavmm we
ofioopoo cowuofl .oz HHMU Hones:
cm>ummno no mass msmz :HHom.o.o combos

 

 

.omssflucovll.m mnmda

.116

now mEOUCH .me

 

vh.am.ma.ha 30H .Eumw Honsu uwcwmmm Hmmam
How .ma.oa .m mh.vm ma noes .cmnns nmfln nusom owumuoosmo m
mm.mm
.nm.mm.vh
.mm.mm.wa How
.m~.ma.mfl meoocfl .eoe umcflmmm Hmmam
HOM .oa.m.m.~ vm.ow ma 30H .cmnuo cmflc cpsom oaumuoosmo n
uwcflmmm
umcflmmm mm mEooca .pwE Hmmam
30H .cmnus nmfls cmoflanommm w
vm.mm.>m
.Nm.vn.nm “0m Hoaaoo ouflc3
.mv.a¢.mm 30H .mEooca .pwE
.mm.mm.am 30H .muflnzlcoc umcflmmm am
new .mH.mH.N mm.vm.om.ma MH gmflc .cmons swan nusom UHuMHOOEmQ m
umcflmmm HoaaoolouHLB
30H .meooCfl .omE
30H .muflx3lco: _ Hm
umcflmmm NH nae; .cmnu: nmfln cmoflansmmm q
mm.mm.am -
.mm.mm.am umgummm
.mm.nm.mm om.mm.mn AEMMmucoc paw
. .mm.na.ma .mw.vm.~m.me gummy Hanan can: mm
Hoe .ma.mH.H .om.nm.ma.m m.m .uOM can“: emu: oeumHUOEmo m
HOm AEHMmucoc paw
EHMMV HMMSH 30H: mm
umcflmmm we .umCHmmm canes now: cmoflansmmm m
MOM Afinmwucoc pom
shame Hanan noes mm
umcwmmm hm .umcflmmm cmnus now; H
coauamom nmcfluMOQGm Hones:
m.ucmpflmmum QmHOmcomm >uwuocflz mmwpuHEEoo mocmsuflumcoo coflmmm wuumm cofluoz

 

.oHom Hmuoomm on» so mcofluoz mo coflumaseflm xmsz HON mcwoooll.m mqmma

117

 

umCHmmm How mmanmm

umcflmmm mm muflcsucoc some gusom ma

“Ow umcwmmm mmHnmm

new Hm.mm muflzauco: emu: nusom cmmfluumaflm ma

umCHmmm How mmahmm

nmcflmmm so muaez-coc emu: bosom cmoflansamm AH

mow umcflmmm mmanmm

new Hm mufln3|coc nae: cusom UHuMHoOEwQ ma
ecu

mEoocH .UmE 30H HOw mmammm

MOM mm.mm.am N .Eumw Hmnsu gown cuoom UHuMHOOEoQ ma
umcflmmm

oEoocH .poE 30H umcammm mmammm

umcflmmm mm .EHMm amusu 50H: nusom va
umcfimmm oEoocH

.UmE 30H .um:flmwm um2flmmm mmammm

umcflmmm so shah Hanan can; nusom cmoflansamm - ma

MOM mvammm

new vo.ma ma umHHoo muses some oflpmuoOEmo NH

mommm

How mm ma MOM songs amen oeumuooEmQ Ha

HOw mvaomm

“oh ma.vm ma umHHoo wages can: onumMUOEmo oH
MOM Afiumwlcoc

paw EHMwV HMHDH mvmm

new mm.m mm ma swan .cmnus some caumHUOEmo m

coflufimom omcwuuomwm nomad:

m.ucwoammum anOmcomm muwuocflz mmmuuHEEou xocmsuflumcoo coammm mumum muumm cofluoz

 

 

.owscHMCOUII.m mqmde

.118

.muoumcwm mCflpcommwuuoo How m xflpcmmmfl o» momma

Q

.mmwuquEoo weepcommouuoo new m xflocmmmfi on ummmmm

 

 

mm

.mm.vn.mn .ccoe

.me.mm.nm ..m>.3..m>
.wm.mm.mm MOM ..mm.0Hco mmhmm
HOW .H~.o~.mH em.mm.ma ma nu20m ..o.z..o2 :mmfluummflm mm
MOM mmmamm
How m mEoocH .me mapofle oaumuooEmQ mm
. ms.on.vo sou umcflmmm mmHemm
“Om .Nw.hv.ma.m ma.m mufln3|coc nae: spoom OHpMHOOEmQ Hm
umcflmmm How mmanmm
umcflmmm Ha opflnzucoc nus; cusom om
coeuflmom QmCHuHomom Hones:
m.ucwoflmmum QmHOmcomm hufluocflz mmmuuflfifiou mucosuflumcoo coamom mumum hunch coauoz

 

.omscflucoouu.m mqmde

119

By manipulating the coding of bills, we can test
subhypotheses of the model. For example, we can perform
a test of the sensitivity of the model to the effect of
party by coding all bills without party. We will perform
sensitivity tests on the major hypotheses of the theory--
those concerning party, constituency, region, "memory"
(for the Cherryholmes-Shapiro model), and executive position
(for SIMEX). We can test the effect of interaction by
predicting the votes of all legislators (not only those
with high predispositions) at the end of the predisposition
phase. We will also examine output in view of our model

hypotheses predicting patterns of correspondence.

CHAPTER IV

VALIDATION AND FINDINGS

Validation of Simulations

 

Abraham Kaplan suggests three norms of validation
for theories--correspondence, coherence, and pragmatism.
In Chapter II we showed how exchange theory coheres with
past empirical research and theoretical formulations by
citing corroborative evidence of the theory's predictions
from the literature on political behavior. In Chapter III,
we demonstrated how the theory can be of practical use in
empirical research. Kaplan points out, however, that
correspondence of predictions and behavior is basic to
the validation of theories.1

In discussing validation of simulations, Charles
Hermann says, "The process of determining how well one
system replicates properties of some other system is

2

called validation." The methodology of evaluating simula-

tions, however, has evolved somewhat beyond the notion of

 

1Abraham Kaplan, The Conduct of Inquiry (San
Francisco: Chandler, 1961), pp. 312-313.

2Charles Hermann, "Validation Problems in Games
and Simulations," Behavioral Science, 12 (1967), p. 216.

 

 

120

121

_strict replication and prediction to one in which some
distortions in model output relative to behavior are not
only permissible, but may contribute to our confidence in
the worth of the simulation model. Of primary importance
in considering validation criteria is a recognition that
our standards are inseparable from the purposes of the
simulation. Abelson points out, in "social simulation

. . . one is not interested in literal prediction, but

rather in validation of the model for theoretical purposes."3

Internal Validity

 

Hermann suggests that inter-run reliability is a
measure of internal validity for stochastic models.4 This
is a measure of internal validity in that it allows us to
make inferences from the results of a simulation. If a
stochastic model varies greatly in its predictions across
runs, it contains little information about the process
simulated. To paraphrase Joseph Hanna, whom we quoted
earlier, if the model provides us no information, it will
predict that all outcomes are equally likely; this is
equivalent to a random prediction (see above, p. 102).

In other words, if a model contains no information, the

 

3Robert Abelson, "Simulation of Social Behavior,"
The Handbook of Social Psychology, ed. by Gardner Lindzey
and Elliot Aronson, 2nd ed., II (Reading, Mass.: Addison-
Wesley, 1968), p. 339.

 

4Hermann, "Validation Problems in Games and Simula-
tions."

122

probability distribution of outcomes it generates is hori-
zontal, or flat.

Cherryholmes and Shapiro used this standard of
internal validity for their model. On data from the House
of Representatives, the model varied little in its pre-
dictive accuracy across runs of the fully-coded model.

The same model, however, exhibits more inter-run variance
on Senate data. Apparently, since it is a smaller legis-
lative chamber, it is less likely that errors in prediction
of individual votes will be evenly distributed across

runs. Although the average of predictive accuracy on all
the motions simulated with Senate data did not vary across
runs, accuracy of individual motions did.5 Consequently,
when we compare smaller subsets of the motions across

runs, it is difficult to make inferences about the perfor-
mance of the model.

We do not encounter this difficulty with SIMEX.
Hermann's test of internal validity is not applicable for
a deterministic model. Since deterministic models generate
only one outcome, the distribution of possible outcomes is
single-peaked. It is impossible, then, for a deterministic

model which generates outcomes to be internally invalid.

 

5On three runs of the fully coded Cherryholmes-
Shapiro model, the average prediction level was 68.7 per
cent, 69.0 per cent, and 69.1 per cent. Variation for
single bills was frequently higher. For example, motion
1 was predicted with 65.3 per cent, 56.1 per cent, and
62.2 per cent correct on the three runs.

123

External Validity

 

While external validity of a theory by definition
is a question of matching prediction and behavior, Paul
Smoker suggests that we refine our criteria for external
validity of models by recognizing the ways in which the
model may correspond well or poorly to behavior depending
upon the behavioral consequences on which we focus our
attention. Smoker labels this accounting recognition of
"patterns of correspondence."

Criteria of validity might well shift from demanding
correspondence between simulation and reality to
defining patterns of correspondence and noncor-
respondence in terms of the model construction
process itself.
Patterns of correspondence and non-correspondence imply
that there are loci of both information and ignorance in
the model, even though loci of ignorance may not be incor-
porated in stochastic processes.

First of all, we expect the theory itself, hence
the model, to be more accurate in predicting certain types
of behavior. Since the outcome of amendments and proce-
dural motions is less visible to extra-legislative constitu-
encies than passage of legislation, we expect our model to

predict votes on these motions less accurately than on

final votes on legislation.

 

6Paul Smoker, "Social Research for Social Anticipa-
tion," American Behavioral Scientist, XII (July-August,
1969), p. 8.

124

Model hypothesis 1. Predictions of the simulation
will be more accurate with respect to final votes
on legislation that on amendments and procedural
motions (p. 86).

Secondly, although SIMEX generates only one predic-
tion, we do not regard all the predicted outcomes equally
likely. That is, the lower the interest in the bill, the
less likely we will be correct in our prediction of the
individual's vote. Accordingly, we have predicted that
the model will be more accurate for those legislators for
whom the model generates high predispositions than for
those with low predispositions.

Model hypothesis 2. Those legislators whose votes
are determined at the end of the predisposition
phase will be predicted more accurately than
those predicted at the end of the interaction
phase (p. 103).

Furthermore, by explicit recognition of the rela-
tionship between theory and model, we can anticipate
inaccuracies built into the model. Specifically, in the
interest of theory relevant validation, if our inaccuracies
result largely from discrepancies in the model relative to
the theory, then we may claim correspondence of theory and
behavior, even when the model does not predict behavior
in all cases. Smoker maintains,

Validity and correspondence are no longer conceptually

linked in a simple additive way such that the greater

the number of correspondences the greater the validity.

Nor does high correspondence necessarily imply high

validity. Validity is the degree to which the predicted
and actual correspondence patterns coincide.

 

7Ibid., p. 8.

125

Because the model does not simulate all aspects of the
intra-legislative bargaining process, we anticipate that
winning coalitions will be generated which are larger
than observed.

Model hypothesis 3. SIMEX will generate larger than
observed winning coalitions (p. 105).

Finally, since the model is related through the
theory to behavior, we can extend this criterion to aspects
of the model which may vary in their effect when the model
is used in different settings. We expect some hypotheses
included in the model to vary in validity when the simula-
tion is used to predict behavior in legislative chambers
of different sizes.

Model hypothesis 4. The constituency hypothesis will
have less positive effect on predictive accuracy

in the Senate than in the House of Representatives

(p. 105).

Model hypothesis 5. The Senate will exhibit less
predictable patterns of interaction than the House

of Representatives (p. 107).

These anticipated non-correspondences, of course, are
verifiable only in comparative research.

We shall compare the Cherryholmes-Shapiro model
output to observed behavior overall and with respect to
predicted patterns of correspondence and non-correspondence.

We shall also compare the performance of the two models to

assess the value of the two models relative to one another.

126

System Performance

 

We will examine two aspects of system-level per-
formance. First, we shall assess the accuracy of the out-
come of the motion, the predicted collective decision of
the legislature. Secondly, we shall examine the structure
of the behavior system generated by the simulation models,

the generation of interests, interactions, and coalitions.

Collective Decisions

 

Exchange theory is a theory of collective as well
as individual decision-making. We should, then, be able
to predict the outcome of collective decisions with our
simulation model.

Tables 4 and 5 show the actual and simulated yea-
nay splits for each model with variations in the coding of
the bills. Tables 6 and 7 show the result in terms of
passage and defeat of actual and simulated motions. SIMEX
with full coding predicts the outcomes of 21 of 23 motions
in our sample correctly (91.3 per cent). Table 7 shows .
that the Cherryholmes-Shapiro model predicts the outcomes‘
of 20 of 23 motions correctly with full coding (86.7 per
cent).

Relative to the fully coded model instance, the
without-constituency SIMEX model predicts collective
decision outcomes more accurately, 22 of 23 (95.6 per cent)
compared to 21 of 23 (91.3 per cent). We anticipated this

with model hypothesis 4: because constituent districts

127

TABLE 4.--Actual and Simulated Yea-Nay Split on Twenty-Three
Federal Role Motions: SIMEX.

 

No. Actual Fully Without Without region
coded constituency or constituency

 

1. S6 41-57 57-41 38-60 38-60
2. 86 44-47 34—57 21-69 21-69
3. sea 52-41 62-31 75-18 75-18
4. 51 41-47 8-78 19-69 18-69
5. 31a 50-34 79- 5 65-19 70-14
6. 51321 44—48 l6-76 21—71 21-71
7. 31321a 47-44 73—18 68-22 73—18
3. 31331a 41-26 63- 3 52-14 55-11
9. S6493 69-11 78- 1 65-14 65-14
10. HR6143a 60-19 63-15 61-17 61-17
11. $2265a 89- 7 80-15 78-17 73-17
12. HR61433 54-27 67-13 65-15 65-15
13. HR6196 12-80 8-78 19-67 20-65
14. HR6196 30-63 7-86 28-64 35-58
15. HR6196a 53-35 82- 4 73—13 70-16
16. HR7152 67—17 71-13 63-19 68-16
17. HR7152 51-43 17-72 27-63 19-70
18. HR7152 71-29 99- o 99- o 99— o
19. HR7152 40-59 27-71 44-52 38-60
20. HR7152 25-74 27-71 43-53 38-60
21. HR7152a 73-27 80-19 74-25 78-21
22. HR11865 60-28 78- 8 68-18 68-18
23. $27823 43-13 57- o 57- o 57- o

 

aFinal vote on legislation.

TABLE 5.--Actual and Simulated Yea-Nay Split on Twenty-Three
Cherryholmes-Shapiro Model.

Federal Role Motions:

128

 

 

No. Actual Fully Without Without region
coded constituency or constituency
l. 86 41-57 53-44 47-51 47-51
2. 56 44-47 31-60 32-57 32-57
3. 56a 52-41 64-28 62-31 62-31
4. 81 41-47 13-75 28-59 28-59
5. 51a 50—34 69-15 50-33 57-26
6. $1321 44-48 17-74 33-59 34-57
7. 81321a 47-44 66-25 51-40 60-31
8. $1831 41-26 61- 5 38-27 47-19
9. S649 69-11 55-23 58-21 55-24
10. HR6143a 60-19 57-20 53-25 54-24
11. $2265a 89- 7 75-20 65-30 68-25
12. HR6143a 54-27 63-17 58-22 59-20
13. HR6196 12-80 17-68 27-68 31-56
14. HR6196 30-63 8-84 22-71 37-54
15. HR6196a 53-35 65-21 68-18 58-28
16. HR7152 67-17 64-20 54-29 58-25
17. HR7152 51-48 22-68 33-56 27-61
18. HR7152 71-29 99- 0 94- 5 97- 2
l9. HR7152 40-59' 25-73 42-54 41-55
20. HR7152 25-74 74-24 40-58 36-62
21. HR7152a 73-27 96- 3 67-30 67-31
22. HR11865a 60-28 57-26 61-25 61-24
23. 82782a 43-13 57- 0 57- 0 57- 0

 

aFinal vote on legislation.

129

TABLE 6.--Actual and Simulated Outcome on Twenty-Three

 

 

 

 

Federal Role Motions: SIMEX.
No. Actual Fully Without Without region
coded constituency or constituency
l. 86 fail pass fail fail
2. S6 fail fail fail fail
3. S6a pass pass pass pass
4. $1 fail fail fail fail
5. 813 pass pass pass pass
6. 81321 fail fail fail fail
7. 81321a pass pass pass pass
8. $1831a pass pass pass pass
9. S649a pass pass pass pass
10. HR61433 pass pass pass pass
11. $22653 pass pass pass pass
12. HR6143a pass pass pass pass
13. HR6196 fail fail fail fail
l4. HR6196 fail fail fail fail
15. HR6196a pass pass pass pass
16. HR7152 pass pass pass pass
17. HR7152 pass fail fail fail
18. HR7152 pass pass pass pass
19. HR7152 fail fail fail fail
20. HR7152 fail fail fail fail
21. HR7152a pass pass pass pass
22. HR11865a pass pass pass pass
23. 82782a pass pass pass pass
Per cent
correct 91.3% 95.6% 95.6%
aFinal vote on legislation.

130

TABLE 7.--Actua1 and Simulated Outcome on Twenty-Three Federal
Role Motions: Cherryholmes-Shapiro Model.

 

 

 

No. Actual Fully Without Without region
coded constituency or constituency
1. S6 fail pass fail fail
2. S6 fail fail fail fail
3. 86a pass pass pass pass
4. S1 fail fail fail fail
5. 81a pass pass pass pass
6. $1321 fail fail fail fail
7. S1321a pass pass pass pass
8. $1831a pass pass pass pass
9. S649a pass pass pass pass
10. HR6143a pass pass pass pass
11. 82265a pass pass pass pass
12. HR6143a pass pass pass pass
13. HR6196 fail fail fail fail
l4. HR6196 fail fail fail fail
15. HR6196a pass pass pass pass
16. HR7152 pass pass pass pass
17. HR7152 pass fail fail fail
18. HR7152 pass pass pass pass
19. HR7152 fail fail fail fail
20. HR7152 fail pass fail fail
21. HR7152a pass pass pass pass
22. HRll865a pass pass pass pass
23. 52782a pass pass pass pass
Per cent
correct 86.9% 95.6% 95.6%

 

aFinal vote on legislation.

131

are heterogeneous, the Senate responds little to constitu-
ency pressures operationalized by demographic group con-
centration. Without constituency interests coded, the
Cherryholmes-Shapiro model also predicts 22 of 23 motions
correctly (95.6 per cent). The without region or constitu-
ency model, which we shall call the basic model, also
predicts 22 of 23 (95.6 per cent) outcomes correctly for
both SIMEX and the Cherryholmes-Shapiro model, neither
validating nor invalidating the hypothesis of regional
interests at this point of correspondence.

All errors in prediction of passage and defeat
occur on amendments, corroborating our prediction that
the models will be more accurate with respect to passage
votes than amendments and procedural motions (model

hypothesis 1).

Structure: Protocoalitions

In the referent process of our legislative decision-
making model, protocoalitions form, consisting of those
highly interested in the outcome of a roll-call at a point
in time preceding the vote. In our simulation, these
protocoalitions consist of those whose votes are deter-
mined at the end of the predisposition phase.

Both SIMEX and the Cherryholmes-Shapiro model
generate protocoalitions of "determined" senators which

are fewer in number than the final number voting. Fully

132

coded, SIMEX averages 55 determined, ranging from 31 to 74
at the end of the predisposition phase. The Cherryholmes-
Shapiro model determines an average of 52, ranging from 34
to 82 at the end of the first phase. Other model instances
for both simulations generate slightly smaller protocoali-
tions: SIMEX 46 and 49 on without-constituency and basic
models, respectively: the Cherryholmes-Shapiro model 49

and 45 on without-constituency and basic models, respec-

tively.

Structure: Interactions

 

Cherryholmes and Shapiro estimated that representa-
tives would talk to twenty-five to thirty of their col-

8 The Senate is a smaller

leagues before each roll-call.
chamber than the House of Representatives and more roll-
calls are taken. These conditions restrict the oppor-
' tunity for inter-personal communication among senators.
A reasonable approximation of the number of senators who
interact, then, would be somewhat less than the estimate
for representatives.

On the average, SIMEX generates seventeen inter-
actions between an undecided senator and other senators.

The Cherryholmes-Shapiro model generates an average of

fourteen interactions between undecided senators and

 

8Cherryholmes and Shapiro, Representatives and
Roll-calls, p. 75.

133

other legislative actors. Both models, then, predict a
reasonable number of interactions for a legislative body

the size of the Senate.

Structure: Coalition Size

Tables 8 and 9 show the actual and predicted dif-
ference in size of coalitions supporting passage and defeat
of each motion for SIMEX and the Cherryholmes-Shapiro
model. Minimum coalitions, of course, would average a
difference of one member. Because of uncertainty, Riker
concludes that coalitions will be greater than minimum.9
There are reasons other than payoffs for joining coalitions:
in the absence of profitable sidepayments, any difference
in the values of passage and defeat would lead a legislator
to join a coalition.

The average of actual differences between winning
and losing coalitions in our sample of roll-calls is 29.
We have predicted that SIMEX will generate larger than
observed coalitions (model hypothesis 3). On the fully
coded model, the average difference is 61, more than twice
the actual average difference. The without-constituency

SIMEX model generates winning coalitions closer to observed

coalitions in average size, differing by 47. The basic

 

9Riker, The Theory of Political Coalitions.

134

TABLE 8.--Coalition Size. Difference Between Number Voting
with Winning Coalition and Number VOting with
Losing Coalition: SIMEX.

 

No. Actual Fully Without Without region
coded constituency or constituency

 

1. S6 16 16 22 22
2. S6' 3 23 48 48
3. S6a 11 31 S7 57
4. 81 6 70 51 51
5. 81a 16 74 46 56
6. S1321 4 60 50 50
7. 81321a 3 55 46 55
8. S1831a 15 60 38 44
9. S649a 58 77 51 41
10. HR6143a 41 48 44 44
11. 82265a 82 65 61 61
12. HR6143a 27 54 50 50
13. HR6196 68 70 48 45
14. HR6196 33 79 36 23
15. HR6196a 18 78 60 - 54
16. HR7152 50 58 44 52
17. HR7152 3 55 36 51
18. HR7152 42 99 99 99
19. HR7152 19 44 8 22
20. HR7152 49 44 10 22
21. HR7152a 46 61 49 57
22. HR11865a 32 70 50 50
23. S2782a 3O 57 57 57

 

aFinal vote on legislation.

135

TABLE 9.--Coalition Size. Difference Between Number Voting
with Winning Coalition and Number Voting with
Losing Coalition: Cherryholmes-Shapiro Model.

 

No. Actual Fully Without Without region
coded constituency or constituency

 

1. S6 16 9 4 4
2. 56 3 29 25 25
3. 56a 11 36 31 31
4. s1 6 62* 31 31
5. 81a 16 54 17 31
6. S1321 4 57 26 23
7. 51321a 3 41 11 29
8. 81831a 15 56 11 28
9. 8649a 58 32 37 31

10. HR6143a 41 37 28 3o

11. 82265a 82 55 35 43

12. HR61433 27 46 36 39

13. HR6196 68 51 35 25

14. HR6196 33 76 49 17

15. HR6196a 18 44 50 3o

16. HR7152 so 44 25 33

17. HR7152 3 46 23 34

18. HR7152 42 99 89 95

19. HR7152 19 48 12 14

20. HR7152 49 50 18 26

21. HR7152a 46 93 37 36

22. 8811865a 32 31 36 37

23. 527823 30 57 57 57

 

aFinal vote on legislation.

136

SIMEX model generates slightly larger winning coalitions,
differing from losing coalitions by 48 members.10
We also anticipated that the Cherryholmes-Shapiro
model, because its interactions phase is stochastic, will
generate smaller coalitions than SIMEX (model hypothesis
3.1). Table 9 shows that the fully coded Cherryholmes-
Shapiro model predicts coalitions which differ in average
size by 52. While this is larger than observed differences,
it is smaller than SIMEX predictions. Like SIMEX, the
Cherryholmes-Shapiro model generates coalitions closer in
average size to observed coalitions in the without-
constituency model; the simulated average difference is
32 compared to the actual average difference of 29. The
basic model instance, like SIMEX, on the average generates
slightly larger coalitions than the without-constituency
model. Overdetermination of the winning coalition in the
Cherryholmes-Shapiro model is less consistent than in
SIMEX. While on the average this model generates smaller
winning coalitions, it would be difficult to alter the
interaction phase to conform more closely to exchange
theory, since the model as it is now formulated under-

estimates winning coalitions more frequently than SIMEX.

 

loOnly accurately predicted outcomes are considered
in the comparison of coalition size.

137

Individual Performance

 

Individual Predictions

The prediction of individual votes is a much more
rigorous test of the validity of the model than the pre-
diction of collective outcomes. The fully coded SIMEX
model predicts an average of 66.2 per cent of participants'
votes on roll-calls in our sample correctly (Table 10).
Subdividing the motions, we find that mean predictive
accuracy is 68.8 per cent for passage of legislation,
64.0 per cent for amendments and procedural motions. This
corroborates model hypothesis 1, the anticipation that the
simulation model will predict passage roll-calls more
accurately than roll-calls on other types of motions.

Comparing the results of twenty-two motions coded
with constituency interests, we find that prediction
improves in the without-constituency model from 66.2 per
cent to 67.4 per cent. This pattern accords with model
hypothesis 4, which predicts that the coding of constituency
interest will have less positive effect on model performance
with Senate data than with House of Representatives data.

Although the average difference is small, differ-
ences are not uniformly distributed across motions coded
with constituency interests. The prediction of votes on
two bills which are solely urban in nature, S6, Mass
Transportation, and $2265, Library Services, were aided

by the coding of constituency. Water Pollution Control,

138

TABLE lO.--Proportion Correct for Individuals: SIMEX.

 

 

 

 

No. Fully Without Without region
coded constituency or constituency
1. $6 .571 .684
2. S6 .582 .689
3. 86a .720 .688
4. 81 .640 .655
5. 51a .631 .702 .667
6. 81321 .630 .685
7. 81321a .648 .700 .670
8. $1831a .591 .636 .652
9. S649a .873 .861
10. HR6143a .667 .692
11. 52265a .832 .811
12. HR6143a .600 .625
13. HR6196 .826 .791 .800
14. HR6196 .688 .620 .667
15. HR6196a .616 .651 .663
16. HR7152 .643 .598 .607
17. HR7152 .584 .622 .584
18. HR7152 .717 .717 .717
19. HR7152 .551 .573 .520
20. HR7152 .541 .490 .449
21. HR7152a .606 .626 .586
22. HR11865a .698 .698
23. $2782a .772 .772
Average
correct .662 .674 .643
aFinal vote on legislation.

139

a Democratic bill, received high bipartisan support: coding
of broad constituency support enhanced predictive accuracy
of the vote on authorization of appropriations. Amendments
to the Farm Bill, HR6196, are also aided by constituency
coding. Since this bill had been carefully balanced to
include several interests, rejection of amendments which
would have changed its provisions was important. Passage
of the Farm Bill, however, was very nearly a party vote.

Several bills--Youth Employment, Manpower Training,
and the National Service Corps--were predicted better
without constituency coded. These bills were intended
to help disadvantaged, unemployed and delinquent individu-
als. It may be that the constituency groups benefitted
by these bills are poorly mobilized, have little effect
on the outcome of elections, and, therefore, elicit little
response from their representatives.

Fully coded, the Cherryholmes-Shapiro model predicts
an average of 69.0 per cent correct on the individual
level (Table 11). Passage votes are predicted with greater
accuracy than amendments and procedural motions, 71.2 per
cent compared with 66.2 per cent. This corroborates model
hypothesis 1.

Although coding of constituency interests aids
predictive accuracy of the Cherryholmes-Shapiro model on
Senate passage votes slightly, the improvement falls short

of that obtained with this model on data from the House of

140

TABLE ll.--Proportion Correct for Individuals: Cherryholmes-
Shapiro Model.

 

 

 

 

No. Fully Without Without region
cOded constituency or constituency
1. $6 .653 .735
2. $6 .769 .670
3. 86a .753 .720
4. 81 .636 .670
5. 51a .702 .690 .667
6. $1321 .707 .707
7. $1321a .747 .780 .725
8. Sl83la .652 .727 .591
9. S649a .759 .797
10. HR6143a .654 .615
ll. $2265a ~300 .695
12. HR6143a .650 .563
13. HR6196 .802 .742 .667,
14. HR6196 .667 .667 .570
15. HR6196a .698 .686 .756
16. HR7152 .583 .571 .524
17. HR7152 .589 .644 .556
18. HR7152 .717 .747 .697
19. HR7152 .714 .786 .673
20. HR7152 .449 .735 .694
21. HR7152a .707 .606 .515
22. HR11865a .686 .733
23. $2782a .772 .772
Average
correct .690 .696 .646
aFinal vote on legislation.

141

Representatives. On eleven passage votes coded with
constituency interest, predictive accuracy declines without
constituency coding to 69.2 per cent from 71.0 per cent
with constituency coding. This decrease is smaller than
the decrease of 7 per cent, from 85 per cent to 78 per

11 On all

cent, reported by Cherryholmes and Shapiro.
twenty-two motions coded with constituency, performance

of the Cherryholmes-Shapiro model improved without constitu-
ency to 69.6 per cent from 68.6 per cent.

On the individual level the hypothesis of regional
interests has marginal validity for SIMEX. Comparing
thirteen motions with regional interest with the basic
SIMEX model we find that predictive accuracy declines from
65.3 per cent to 64.3 per cent. This pattern of regional
effect varies across motions. In particular, Southern
coding does not aid predictive ability on any votes con-
cerning the Administration Farm Bill. Southerners generally
did oppose amendments and support passage: however, constitu-
ency and party interests coincide for Southerners on these
motions. Apparently, the addition of region masks varia-
tion which is already explained by constituency and party.

Regional coding for the Cherryholmes-Shapiro model

improves performance over the basic model. Thirteen

motions coded with region are predicted with an accuracy

 

11Cherryholmes and Shapiro,'Representatives and

142

of 70.4 per cent; without region, predictive accuracy on
these motions declines to 64.6 per cent.

SIMEX and the Cherryholmes-Shapiro model are
similar in their performance. In every case, the Cherryholmes-
Shapiro model predicts individual performance slightly better
than SIMEX. Since the Cherryholmes-Shapiro model does not
use median income as a constituency characteristic, we
compared the difference between with- and without- constituency
model instances for both models to find out if this variable
accounts for the difference in constituency effect between
the two models. On nine motions coded with median income
for SIMEX but not for the Cherryholmes-Shapiro model, we
find that without-constituency improves performance 1.9 per
cent for SIMEX, and 1.1 per cent for the Cherryholmes-
Shapiro model. While this difference is slight and cannot
account for the difference between the two models, the
value of median income as an indicator of constituency
interest should remain suspect before further testing.
Since Southern states are characterized as low on median
income, this variable may not predict well on the national
level, yet still be a salient characteristic for use in

prediction of roll-calls in subnational legislatures.

Prediction of Protocoalitions
Model hypothesis 2 states that legislators highly

predisposed to vote a given way on a roll-call will be

143

predicted more accurately than those who enter the inter-
action phase. Table 12 shows the assessment of SIMEX
accuracy for highly interested senators alone. The fully
coded model predicts an average of 70.9 per cent of these
participants' votes correctly; among the total number
predicted, only 66.2 per cent were correct. On the twenty-
two roll-calls coded with constituency interest, predic-
tive accuracy for highly predisposed participants rises
from 70.5 per cent to 71.8 per cent correct when constitu-
ency coding is eliminated. This corroborates model
hypothesis 4. Furthermore, removal of regional coding
decreases predictive accuracy for highly interested legis-
lators. On thirteen motions coded with regional interest,
SIMEX predicts only 63.6 per cent correctly when region is
eliminated, whereas it predicts 73.6 per cent correctly
with region coded. Only the basic SIMEX model, without
region or constituency coding, predicts better for the
total number voting than for the highly predisposed.

Although output from the Cherryholmes-Shapiro model
does not allow us to test this model hypothesis, the pre-
disposition phases of the two models are nearly identical.
The Cherryholmes-Shapiro model is, in general, more accurate
on the individual level than SIMEX; it is likely, then,
that highly predisposed legislators are predicted with

even higher accuracy in the Cherryholmes-Shapiro model.

144

TABLE 12.--Proportion Correct for Highly Interested

 

 

 

Senators: SIMEX.
No. Fully Without Without region
coded constituency or constituency

1. S6 .672 .571

2. $6 .667 .694

3. S6a .778 .781

4. 51 .685 -.690

5. 81a .672 .804 .724

6. 81321 .706 .746

7. S1321a .776 .849 .767

8. $1831a .609 .750 .644

9. S649a .946 .868»

10. HR6143a .650 .643

11. $2265a .797 .770

12. HR6143a .607 .588

13. HR6196 .891 .786 .780
14. HR6196 .824 .917 .833
15. HR6196a .741 .737 .741
16. HR7152 .565 .532 .491
17. HR7152 .576 .667 .583
18. HR7152 .683 .783 .667
19. HR7152 .613 .727 .333
20. HR7152 .727 .636 .357
21. HR7152a .581 .561 .500
22. 8811865a .736 .714

23. 52782a .814 .850
Average

correct .709 .718 .636

 

aFinal

vote on legislation.

145

Interaction

 

We have predicted that in smaller legislative
chambers, patterns of interaction will be more idiosyncratic
since there is a greater likelihood that legislators will
know more about one another's interests (model hypothesis
5). Furthermore, familiarity reduces bargaining costs
between all members of the smaller legislative house. we
can eliminate the interaction phase by predicting the votes
of all legislators at the end of the predisposition phase:
in this case, the total number voting in the simulation
usually does not reach the number actually voting.

On the fully coded SIMEX model, bypassing the inter-
action phase actually increases predictive accuracy of
those voting from 66.2 per cent to 67.3 per cent. Examina-
tion of alternative formulations of SIMEX shows that inter-
action does aid prediction for the without-constituency and
basic model instances (Table 13). Since these versions of
SIMEX generate smaller protocoalitions, leaving more votes
to be determined after interaction, we may claim some
validity for the hypothesized interaction patterns.
Predictive accuracy increases from 65.9 per cent to 67.8
per cent for the without-constituency model when inter-
actions are permitted, from 65.8 per cent to 67.1 per cent
for the basic model. The slightly greater difference
between predictions with and without interaction with

region coded probably reflects the greater accuracy of

146

TABLE l3.--Proportion Correct with Interaction and Without

 

 

 

 

 

Interaction: SIMEX.
Fully Without Without region
No. coded constituency or constituency
Interaction
with without ‘ with Without with without
1. S6 .571 .640 .684 .670
2. S6 .582 .612 .689 .662
3. 56a .710 .726 .688 .728
4. s1 .640 .620 .655 .667
5. 51a .631 .685 .702 .681 .667 .697
6. 51321 .630 .680 .685 .706
7. S1321a .648 .714 .700 .733 .670 .746
8. $1831 .591 .672 .636 .604 .652 .633
9. S649 .873 .878 .861 .841
10. HR6143a .667 .652 .692 .625
11. 82265a .832 .782 .811 .712
12. HR6143a .600 .619 .625 .576
13. HR6196 .826 .812 .791 .776 .800 .758
14. HR6196 .688 .663 .620 .644 .667 .663
15. HR6196a .616 .700 .651 .710 .663 .708
16. HR7152 .643 .592 .598 .484 .607 .492
17. HR7152 .584 .597 .622 .571 .584 .563
18. HR7152 .717 .681 .717 .642 .717 .661
19. HR7152 .551 .571 .573 .590 .520 .531
20. HR7152 .541 .489 .490 .482 .449 .454
21. HR7152a .606 .583 .626 .542 .586 .530
22. HR11865a .698 .696 .698 .712
23. 52782a .772 .804 .772 .804
Average
correct .662 .673 .678 .659 .671 .658
'aFinal vote on legislation.

147

this model in predicting the votes of thehighly interested
senators.

Table 14 presents the results of with- and without-
interaction versions of the Cherryholmes-Shapiro model.
The fully coded model with interaction predicts 69.0 per
cent correctly, whereas when it bypasses the interaction
phase it predicts only 65.1 per cent correctly. Approxi-
mately the same difference obtains for the without-
constituency and basic model instances. Even though
the interaction phase does exhibit some positive effect
on predictive accuracy, the improvement falls short of
that observed in eXperiments with data from the House of
Representatives. Cherryholmes and Shapiro found that
communications improved predictive accuracy from 78 per
cent to 84 per cent, a difference of 6 per cent compared

to only 3.9 per cent for the same model on Senate data.12

Party
The effect of political party affiliation on legis-

lative voting has repeatedly been a focus for students of
legislative behavior. There is no doubt that party exerts
pressure on its members to conform. However, the proper

simulation of this influence is of interest to us because

it may help us answer the question of how and under what

 

121516., p. 117.

148

TABLE l4.--Proportion Correct with Interaction and Without

 

 

Interaction: Cherryholmes-Shapiro Model.
Fully Without Without region
No. .coded_ constituency or constituency
Interaction

with without. with. without with Without

 

 

 

1. S6 .653 .541 .735 .592
2; S6 .769 .659 .670 .670
3. 56a .753 .699 .720 .710
4. 51 .636 .648 .670 .705
5. 51a .702 .726 .690 .738 .667 .702
6. $1321 .707 .728 .707 .707
7. 51321a .747 .769 .780 .769 .725 .736
8. 51831a .652 .697 .727 .742 .591 .682
9. 8649a .750 .671 .797 .671
10. HR6143a .654 .654 .615 .615
11. 52265a .800 .642 .695 .579
12. HR6143a .650 .575 .563 .537
13. HR6196 .802 .709 .742 .652 .667 .533
14. HR6196 .667 .602 .677 .516 .570 .495
15. HR6196a .698 .616 .686 .651 .756 .605
16. HR7152 .583 .560 .571 .619 .524 .488
17. HR7152 .589 .600 .644 .678 .556 .578
18. HR7152 .717 .677 .747 .657 .697 .626
19. HR7152 .714 .694 .786 .684 .673 .663
20. HR7152 .449 .500 .735. .633 .694 .622
21. HR7152a .707 .606 .606 .657 .515 .556
22. HR11865a .686 .674 .733 .674
23. 82782a .772 .737 .772 .737
Average
correct .690 .651 .699 .661 .666 .630
aFinal vote on legislation.

149

conditions party exerts its influence. Party alone predicts
63.5 per cent of the votes of senators on motions in our
sample. This is lower than the "just party" prediction in

13

the House of Representatives, 78 per cent. Clausen and

Cheney also observed that on all roll-calls, senators

14 Table

respond less to party than do representatives.
15 shows SIMEX results with party (fully coded), without
party, and the party-only prediction. Comparing party-only
and with party-model instances, we find that party-only
predicts an average of 63.5 per cent correct whereas the
fully-coded model instance predicts 67.8 per cent correct.15
Subdividing the sample, we find that the difference in the
party-only prediction and the fully coded model prediction
is greater for passage votes than for amendments and
procedural motions. Cherryholmes and Shapiro found that
recommittal motions exhibited relatively more inter-party
conflict than passage votes in the House of Representa-

16

tives. Insofar as recommittal motions are one type of

bill included in our selection of amendments and procedural

 

l3Ibid., pp. 106-110.

14Clausen and Cheney, "A Comparative Analysis of

Senate-House Voting on Economic and Welfare Policy 1953-1964.

15Only motions coded with party are used in the
comparison of alternative party formulations.

16Cherryholmes and Shapiro, Representatives and
Roll-Calls, p. 110.

150

TABLE 15.--Effect of Party on Predictive Accuracy: SIMEX.

 

 

 

No. Fully coded Fully coded Party only
(With party) (without party)
1. $6 .571
2. S6 .582 .516 .648
3. 86a .720 .667 .645
4. 51 .640 .517 .670
5. 51a .631 .590 .631
6. S1321 .630 .598 .674
7. 81321a .648 .626 .670
8. 51831a .591 .636 .552
9. 8649a .873 .861 .738
10. HR6143a .667 .654 .575
11. 82265a .832 .726 .667
12. HR6143a .600 .612 .506
13. HR6196 .826 .837 .674
14. HR6196 .688
15. HR6196a .616 .612 .727
16. HR7152 .643 .590 .500
17. HR7152 .584 .544 .551
18. HR7152 .717 .546 .707
19. HR7152 .551
20. HR7152 .541
21. HR7152a .606 .626 .484
22. HRll8GSa .698 .694 .667
23. 52782a .772 .737 .772
Average
correct .678 .632 .625

 

aFinal vote on legislation.

151

motions, we find that their result obtains in the Senate
as well. As a corollary to model hypothesis 1, then, we
might add that party contributes proportionately more to
procedural motions and amendments than other variables.

When we compare the fully coded SIMEX model with
the without-party model, this conclusion is reinforced.
The without-party SIMEX model predicts 64.2 per cent of
the votes correctly, whereas the fully coded model pre-
dicts 67.8 per cent of the votes correctly. Most of this
difference, however, occurs on results for amendments and
procedural motions. Prediction level for passage votes
declines 1.7 per cent, from 68.8 per cent to 67.1 per
cent; for amendments and procedural motions, however,
prediction level declines 6.7 per cent, from 66.0 per
cent to 59.3 per cent.

We have established that party contributes to the
predictability of roll-call votes, especially procedural
motions and amendments. A comparison of the Cherryholmes-
Shapiro model with party and the party-only prediction
affirms this (Table 16). The party only prediction is
63.5 per cent whereas the with-party prediction of the
model is 70.4 per cent. More of this difference is attri-
butable to passage votes which decline from 71.5 per cent
correct in the fully coded Cherryholmes-Shapiro model to

63.7 per cent correct in the party-only prediction.

152

TABLE l6.--Effect of Party on Predictive Accuracy:
Cherryholmes-Shapiro Model.

 

 

 

No. Fully coded Fully coded Party only
(with party) (without party)
1. S6 .779
2. S6 .769 .712 .648
3. 86a .753 .872 .645
4. $1 .636 .677 .670
5. 81a .702 .640 .631
6. $1321 .707 .774 .674
7. S1321a .747 .678 .670
8. 51831a .652 .707 .552
9. S649a .759 .737 .738
10. 886143a .654 .721 .575
11. 82265a .800 .842 .677
12. 886143a .650 .673 .506
13. HR6196 .802 .736 .674
14. HR6196 .720
15. 886196a .698 .580 .727
16. HR7152 .583 .655 .500
17. HR7152 .589 .674 .551
18. 887152 .717 .714 .707
19. 887152 .682
20. 887152 .595
21. 887152a .707 .775 .484
22. 8811865a .686 .848 .667
23. 82782a .772 .744 .772
Average
correct .704 .719 .635

 

aFinal vote on legislation.

153

Amendments decline from 68.6 per cent correct in the fully
coded model to 63.2 per cent in the party-only prediction.
When we compare the fully coded prediction of the
Cherryholmes-Shapiro model with the without-party predic-
tion, however, we find a surprising difference. Without
party the Cherryholmes-Shapiro model predicts at a higher
level than it does with party. Prediction level rises
from 70.4 per cent to 72.4 per cent when party is elimi-
nated. Furthermore, this difference is equally apportioned
between passage votes and votes on other types of motions.
This suggests that the Cherryholmes-Shapiro model is over-
determined with respect to party, as Cherryholmes and

17 This becomes most evident in the

Shapiro suggested.
Senate, in which members appear to be less bound by party
considerations. Not only is party over-determined, but
some variable in the Cherryholmes-Shapiro model contributes
more to predictive accuracy in the absence of party influ-
ence than in its presence. The interaction phase of the
Cherryholmes-Shapiro model is highly structured by party
likeness. Furthermore, the memory score prediction matches
73 per cent of party only predictions on individuals. The
memory score, moreover, is a likely candidate for an
influence superior to party, since it is directly related

to the measure of model output, votes cast on federal role

measures in the past session.

 

17Ibid., pp. 109, 116.

154

Effect of "Memory"

 

Table 17 shows that "memory" alone predicts 69.5
per cent of the votes in our sample. Overall, this is
higher than the prediction of the full model. Furthermore,
when we compare the without-memory model to the fully
coded model, we find that predictive accuracy falls off
to 64.5 per cent without "memory" from 69.0 per cent with
"memory." Compared to results of the House of Representa-
tives, "memory" contributes more to predictive accuracy of
the Cherryholmes-Shapiro simulation run on Senate data.
Cherryholmes and Shapiro found that without memory, predic-
tive accuracy fell only 2 per cent short of predictive
accuracy with memory. In the Senate, the difference is
3.5 per cent. Since the fully coded model on Senate data
predicts less well, generally, memory contributes even
more to variation explained than the difference in per-
centage values indicates.

Examining subsets of the sample, we find that
"memory" alone predicts amendments and procedural motions
better than passage motions, 70.5 per cent compared to
68.6 per cent correct. Furthermore, without memory,
predictive accuracy is reduced less for passage motions
than for amendments and procedural motions.

These findings are not unanticipated. we have
found that senators are more independent of constituency

and party than representatives. Furthermore, their patterns

155

TABLE l7.--Effect of Party on Predictive Accuracy:
Cherryholmes-Shapiro Model.

 

 

 

No. Fully coded Fully coded "Memory"
(with "memory") (without "memory") only
1. S6 .653 .520 ..733
2. S6 .769 .571 .753
3. 56a .753 .656 .732
4. 81 .636 .602 .785
5. 81a .702 .679 .853
6. $1321 .707 .679 .802
7. S1321a .747 .626 .775
8. S1831 .652 .561~ .783
9. $649 .759 .759 .643
10. 886143a .654 .654 .681
11. 82265a .800 .832 .553
12. 886143a .650 .587 .592
13. HR6196 .802 .767 .548
14. 886196 .667 .677 .542
15. 886196a .698 .698 .558
16. HR7152 .583 .619 .733
17. HR7152 .589 .578 .756
18. HR7152 .717 .717 .701
19. HR7152 .714 .602 .744
20. HR7152 .449 .337 .655
21. 887152a .707 .737 .690
22. 8811865a .686 .651 .722
23. 52782a .772 .772 .654
Average
correct .690 .645 .695

 

aFinal vote on legislation.

156

of interaction are more idiosyncratic than the correspond-
ing processes in the House of Representatives. As we
suspected, it is likely that the memory score reflects a
number of influences that are not replicated by other
processes in the Cherryholmes-Shapiro model. This summary
of influences may include ideological preferences; however,
since amendments and procedural motions are relatively more
influenced by "memory," it seems unlikely that "memory"
exclusively measures ideology. It is likely that Senate
voting is more idiosyncratic than voting in the House of
Representatives for several reasons. Hence, memory is
essential to the Senate performance of the Cherryholmes-
Shapiro model because it inculcates variables and effects
of patterns of interaction not included in the model.
Without memory, the fully coded Cherryholmes-
Shapiro model predicts individual votes less accurately
than the fully coded SIMEX model. Since SIMEX includes
the effect of the president's position in the predisposi-
tion phase which is not present in the Cherryholmes-Shapiro
model, we shall examine the consequences of this addition

for predictive accuracy of SIMEX.

Effect of President's Position
The results presented in Table 18 show that the
executive's position does not contribute to predictive

accuracy of the SIMEX model over all motions. Predictive

157

TABLE 18.--Effect of Executive Position on Predictive

 

 

 

Accuracy: SIMEX.
No. Fully coded Fully coded
(with exec. pos.) (without exec. pos.)
1. $6 .571 .531
2. S6 .582 .622
3. 86a .720 .667
4. S1 .640 .636
5. 81a .631 .595
6. $1321 .630 .652
7. 51321a .648 .670
8. S1831a .591 .606
9. 5649a .873 .861
10. 886143a .667 .654
11. 52265a .832 .830
12. 886143a .600 .612
13. 886196 .826 .779
14. HR6196 .688 .685
15. 886196a .616 .651
16. HR7152 .643 .639
17. HR7152 .584 .578
18. HR7152 .717 .717
19. HR7152 .551 .592
20. HR7152 .541 .745
21. 887152a .606 .667
22. 8811865a .698 .640
23. 82782a .772 .772
Average
correct .662 .670

 

aFinal vote on legislation.

158

accuracy rises from 66.2 per cent to 67.0 per cent when
the effect of the president's position is eliminated.
When we examine subsets of the motions, however, we find
that this variable contains negative information only for
amendments and procedural motions. The results of the
fully coded model are 63.4 per cent correct on amendments
and procedural motions, but 65.2 per cent correct when
the executive's position is eliminated. For passage
votes, the level of prediction declined only marginally
when presidential position is eliminated from the fully
coded SIMEX model, from 68.8 per cent to 68.5 per cent.

Theoretically, this variable is utilized as an
item of constituency information, operationalized by the
president's plurality in the senator's state in conjunc-
tion with the president's position. Its marginal positive
effect on passage motions but not on amendments and pro-
cedural motions may reflect the greater consideration given
to future exchanges with constituents by legislators on
those legislative outcomes most likely to be salient to
future voters.

This variable aided predictive accuracy on twelve
of the twenty-three bills. It did not help for passage
votes on the National Service Corps, Manpower Training,
College Aid, the Administration Farm Bill, or the Civil
Rights Bill. It is likely that the president does not

equally stress passage of all legislation he favors. We

159

might, then, attempt to find a better measure of the
president's position than the one provided for us by the
Congressional Quarterly Service.

Our decision rules state that the legislator and
the executive run for office in the same elections; however,
six year terms prohibit a senator's running in a presiden-
tial election year twice in succession. Thus, the electoral
cycle of the senator relative to that of the president may
attenuate the relationship between the senator's voting
record and the president's plurality in his state (hypothe-
sis L-l.4). Furthermore, the U.S. senator enjoys a position
of great importance and'prestige in the American political
system. Consequently, news media publicize his activities
more than any other American legislator. Because of this
opportunity to popularize issues, a senator may preceive
more electoral independence than other legislators. Barbara
Hinckley shows that the incumbent senator does experience
independence of other electoral forces in his reelection
campaigns.18

Although the executive's position as we have engaged

it is not satisfactory, we should test it on other legisla-

tive bodies before discarding the hypothesis altogether.

 

18Barbara Hinckley, "Incumbency and the Presidential
Vote in Senate Elections," American Political Science Review,
LXIV (September, 1970): pp. 836-842. Clausen and Cheney also
hypothesize that senators are relatively more free of party
than representatives because of their opportunity to build a
reputation independent of their parties. Clausen and Cheney,
"A Comparative Analysis of Senate-House Voting on Economic
and Welfare Policy 1953-1964."

160

Conclusion

 

Both the Cherryholmes-Shapiro model and the revised
simulation model, SIMEX, predict the outcomes of collective
decisions quite well. However, their performance on Senate
data reveals that they do not predict individual behavior
as well as simulations tested on data from the House of
Representatives. Nevertheless, since patterns of corre-
spondence and non-correspondence in predictions of the
models relative to behavior conform to our prognosis of
these distortions, we may conclude that exchange theory
provides a better explanation of legislative behavior than
the models provide predictions.

In comparing predictions of the Cherryholmes-
Shapiro model with SIMEX, we find great similarities.
Without memory, the fully coded Cherryholmes-Shapiro model
fails to predict as well as the fully coded SIMEX model,
64.5 per cent for the former compared with 66.2 per cent
for the latter. Since the memory score predicts well, but
provides little information we can determine apart from
results of roll-calls themselves, we must conclude that
SIMEX has greater explanatory power than the Cherryholmes-

Shapiro model.

CHAPTER V

CONCLUSION

Summary

In this project we moved in two directions from the
point Cherryholmes and Shapiro concluded their work on the
U.S. House of Representatives. First, we expanded vertically,
subsuming their model and the research findings upon which
it is based under a hierarchical theory, one postulated as
a deductive explanatory model. Based on this theory, the
simulation designed by Cherryholmes and Shapiro was modified
to accord more closely with this theory. Secondly, the
simulation experiments were conducted on another legislative
chamber, the U.S. Senate. Furthermore, in extending the
analysis of legislative behavior in these two directions,
we acknowledged both the limitations of the theory and the
sources of variations in the simulation's predictive
accuracy in various settings.

In this concluding section, the contributions of
this exercise are summarized and directions in which
research might continue are suggested. Much of the dis-
crepancy between predictions of the simulation and observed
behavior is anticipated. The reasons for these results lie

in the nature of a theory based on individual rationality

161

162

and in the distortions built into the simulation model.

we must, then, operate between the theory and the simula-
tion model to enable us to claim better correspondence; if
theoretically-based adjustments in the simulation model do
not enable us to predict with greater accuracy, we should
hold in abeyance our confidence in the theory and search
for one which more closely accords with observed behavior.

Kaplan, as we have noted, says that correSpondence
is basic to the validation of theories, but we must also
consider the norms of coherence and pragmatism in evaluat-
ing theories. In Chapter II, we demonstrated the coherence
between predictions of the theory and relationships observed
in a number of research settings. we also suggested at
several points the aspects of exchange theory that cor-
respond to ambition theory and theories of political
coalitions. The making of sidepayments, central to coali-
tion theories, is integral to the assumption that individu-
als make exchanges of resources as part of their
strategies. Schlesinger's central axiom, that ambition
is the primary motivation for politicians, is explicit in
our definition of rationality.

Although the rationality assumption and the com-
plexity of the situation as specified by the decision rules
inject an element of uncertainty into the precision of our
predictions, we demonstrated that exchange theory can be

defined in such a way that testable hypotheses can be

163

derived from it. In Chapter II, we derived such hypotheses
and in Chapter III we illustrated one method for testing
these hypotheses, a simulation model. This demonstrates
the pragmatic validity of the theory, our ability to use
it in empirical research.

The model we designed based on exchange theory,
SIMEX, has several advantages over the simulation model
used by Cherryholmes and Shapiro. SIMEX requires fewer
input variables than the Cherryholmes-Shapiro model.
Internal processing in the interaction phase of SIMEX is
simpler than that in the "communications" or interaction
phase of the Cherryholmes-Shapiro model; more specifically,
SIMEX requires estimation of fewer parameters than the
Cherryholmes-Shapiro model. Furthermore, the SIMEX model
uses no direct measures of the outcome, roll-call votes,
to predict behavior; the Cherryholmes-Shapiro model includes
"memory," a direct input of the individual's past behavior.
Since "memory" is an issue-specific measure, we might add
that operationalization of SIMEX is not dependent on the
isolation of broad issue dimensions. Because it is easier
to use for research, SIMEX scores more highly on the
criterion of pragmatic validity than the Cherryholmes-
Shapiro model.

We have noted that explanatory power is that
associated with information determined prior to the measure-
ment of the dependent variables, while descriptive power

is that associated with information drawn from the data to

164

be explained. "Memory" is a measure of the dependent
variable: while the Cherryholmes-Shapiro model with
"memory" predicts votes better than the fully-coded SIMEX
model, the "memory score" contributes to descriptive, but
not explanatory power of the model. In fact, ”memory"
alone predicts individual behavior better than any instance
of either simulation model. This finding clearly demon-
strates the descriptive nature of this variable. Without
"memory,” predictive accuracy of the Cherryholmes-Shapiro
model lags behind that of SIMEX; hence, we conclude that
SIMEX is higher in explanatory power than the Cherryholmes-
Shapiro model.

SIMEX predicts collective outcomes very well, and
performs less well for individual votes. we find, however,
that distortions in predicted outcome relative to actual
behavior result from known distortions in the construction
of the model. Many of our inaccuracies in prediction,
then, contribute to our confidence in the explanation

provided by the theory.

Recommendations

 

Strictly speaking, the model is not a sufficient
test of the theory. We suggest two major types of modifica-
tion: (1) Since our primary concern is theory-building,
we should first work between the model and the theory to
render comparison of simulation output and behavior a more

adequate test of correspondence, (2) The operation of the

165

model depends upon our choice and measurement of variables
and the estimation of parameters. We should, then, attempt
to test the value of variables which are justifiable

theoretically and to estimate parameters with more accuracy.

Interaction Phase

 

The most blatant discrepancy in the SIMEX model
relative to exchange theory lies in the interaction phase:
the rough approximation of bargaining and exchange generates
larger than observed coalitions. Concomittantly, the
interaction phase results in less accurateprediction of
votes of individuals who are processed through it than
the prediction of votes at the end of the predisposition
phase. The subsidiary prediction that an undecided legis-
lator will interact with seventeen other senators may be
somewhat high. Fewer interactions would still be reasonable.
and might heighten the accuracy of this phase of the simula-
tion. There are several ways in which the interaction phase
of the model might be modified to fit the theory while
maintaining the basic patterns we have hypothesized.

1. We could alter the interaction phase to allow
legislators to encounter only those legislators who are
highly interested in the outcome of the collective action.
The simulation model could store this information at the
end of the predisposition phase as a screening device for
the interaction phase. This alteration is unlikely to

affect the outcome greatly, since averaging predispositions

166

encountered prevents a series of marginally interested
legislators from influencing the undecided legislator's
vote decision.

2. A more profitable approach might be to dis-
tinguish between those undecided because of cross-pressures
and those indifferent to the outcome in the model's pre-
diction. This accounting could also be stored as an
additional variable generated by the simulation. Since
the indifferent operate under little risk, buying their
votes would be easier for the interested protocoalition
than bargaining with a legislator who is more uncertain
about the consequences of his action. After bargaining
with this group of legislators is completed, the model could
check if either coalition is winning, has more than fifty
per cent of the total number of possible votes. If one
coalition is winning, then the model could predict the rest
of the votes on the basis of the algebric sign of whatever
small interest the rest of the legislators (the cross-
pressured) have. If neither coalition is winning, the
model could continue the interaction phase for the rest
of the undecided legislators.

3. A third way of limiting coalition size is to
revert to a stochastic process in the anticipation that
fewer interactions would reduce the consensus-formation
aspect of the deterministic process. One way of estimating

probabilities of interaction is to adjust them inversely to

167

the size of protocoalitions determined in the first phase
of the model. In this manner, the fewer the number deter-
mined in favor of a position in the first phase, or the
greater the need for additions to their coalitions, the
greater the probability of interaction between legislators
for the purposes of exchanging resources.

4. Finally, we could adjust the "offer," or the
value of an additional vote for a coalition inversely in
relation to the size of the coalition already committed
to a particular outcome. In this alternative formulation,
the larger the protocoalition, the less likely that a
legislator's vote will be "purchased" by a member of that

coalition.

PrediSposition Phase

The success of the interaction phase, no matter how
much it conforms to the theory and the theory to reality,
depends upon the precision with which we estimate parameters
in the first phase of the model. Without taking information
directly from the data, we can refine the estimation of
some parameters.

First of all, we reason in Chapter II that the
legislator probably pursues various goals serially over
the multiplicity of motions decided in the collective
context. In the simulation outcomes we find that party

considerations are more salient on the less visible

168

amendments and procedural motions. Constituency considera-
tions appear more salient when the legislator votes on
passage of legislation. We might gain more accuracy by
reducing the value of the party parameter on passage
motions and raising it on amendments and procedural motions.
Correspondingly, we might raise the value of the constitu-
ency parameters on passage motions and reduce them on
amendments and procedural motions.

Further precision might be introduced by judicious
choice of constituency variables. The legislator acts on
his prospects of making exchange with future constituents.
Past governmental action, particularly financial aid, is
the current status quo for an electorate. Legislators
who plan to seek votes in a district upon which governmental
programs have already made an impact might, then, be more
likely to vote for increases or expansion of programs than
those legislators whose future constituents are less well-
acquainted with such action. There is some evidence to
this effect: Ripley finds that of the congressmen who
were elected from districts in which the food stamp pro-
gram had been, thirty-one of thirty-three supported it in
1964 (two did not vote).1

Ultimately, we must invest in parameter estimation

procedures such as regression or factor analysis to more

 

1Randall B. Ripley, "Legislative Bargaining and the
Food Stamp Act, 1964," in Congress and Urban Problems,
by Frederic Cleaveland, et a1. (Washington, D.C.: Brookings,
1969), pp. 279-310.

169

accurately assign values to parameters in the predisposi-
tion phase of the model. Parameter estimation takes
information from the data to enhance predictive accuracy

of model output. At this point additional predictive power
is descriptive and not explanatory. Nevertheless, parameter
estimation provides us much more meaningful information
than a variable such as the "memory score" because it
allows us to infer relative importance of independent
variables. Only after parameter estimation and theoretical
adjustments of the model can we begin to make educated
guesses about what proportion of the remaining variance
must be explained by unique and idiosyncratic variables

such as personal ideology.

ngparative Use of the Model

We have arrived at a point similar to the conclusion
of Cherryholmes and Shapiro's research on the House of
Representatives. Further comparative research is necessary.
Testing the simulation on roll-call votes spanning a
number of congressional sessions should prove theoretically
interesting and informative. The postwar period provides
us with variations in presidential party, majority party
in Congress, and combinations of presidential party and
congressional majority. If these variations affect the
performance of the model, it would inform us of ways in

which we might further refine or elaborate our theory.

170

Since the theory of exchange in a two-party system
is a comparative theory and data is readily available on
state legislatures, broad testing of the model should be

relatively easy.

Beyond Simulation

In our deductions from exchange theory, we have
focused on the behavior of legislators as they decide to
allocate their votes on each motion. Simulation is appro-
priate for carrying out such research on behavior of
individuals in an organization: depending on his constitu-
ency, party and position in the legislature, the deter-
minants of each of the legislator's votes may vary; the
organization itself exhibits varying characteristics under
changing conditions.

The application of exchange theory is not, however,
confined to its use in simulation of roll-call votes in
American legislatures. Two related areas of research to
which we might apply it are legislative behavior other
than roll-call voting and legislatures and collective
decision-making in non-American settings. We might focus
on selected aspects of legislative behavior other than the
individual legislator's roll-call vote. We have suggested
in passing that hypotheses are derivable from exchange
theory explaining behavior of committees and party leaders.

Preliminary evidence of the resources and their distribution

171

in these areas could be sought in recent work on party

leadership change and legislative committees.2
Secondly, by altering the conditions, particularly

electoral decision rules and the number of political

parties, we can apply exchange theory to behavior in

non-American elective legislatures. Cabinet stability

and instability may reflect convergent and divergent

tensions generated by episodic elections and the desire

to participate in a coalition government.

 

2Robert L. Peabody, "Party Leadership Change in the
House of Representatives," American Political Science Review,
LXI (September, 1967), pp. 675-693; andTSenate Leadership
Change: 1953-1970" (paper prepared for delivery at the 66th
Annual Meeting of the American Political Science Association,
1970). Randall B. Ripley, MajorityParty_Le§dership in
Congress (Boston: Little Brown, 1969); and Party Leaders
In the House of Representatives (Washington, D.C.: Brookings,
I967).

 

BIBLIOGRAPHY

172

BIBLIOGRAPHY

Abelson, Robert P. "Simulation of Social Behavior." The
Handbook of Social Psychology. Edited by Gardner
Lindzey and Elliot Aronson. '2nd ed. Vol. II.
Reading, Mass.: Addison-Wesley, 1968.

Anderson, Lee F.; Watts, Meredith W., Jr.; and Wilcox,
Allen R. Legislative Roll-Call Analysis. Evanston:
Northwestern University, 1966.

Andrain, Charles F. "Senators' Attitudes Toward Civil
Rights." Western Political Quarterly, 17 (1964).

Axelrod, Robert. "Derivation of a Coalition Theory Based
on Conflict of Interest with an Application to
Italy." Paper prepared for delivery at the 65th
Annual Meeting of the American Political Science
Association, 1969. (Mimeographed.)

Bauer, Raymond A.; Pool, Ithiel deSola; and Dexter, Lewis A.
American Business and Public Policy. New York:
Atherton, 1964.

Belknap, George M. "A Method for Analyzing Legislative
Behavior." Midwest Journal of Political Science,
2 (1958).

 

Beyle, Herman C. Identification and Analysis of Attribute-
Cluster-Blocs. Chicago: University of Chicago,
1931.

 

Black, Duncan. The Theory of Committeesiand Elections.
Cambridge: University Press, 1958.

Black, Gordon S. "A Theory of Professionalization in
Politics." American Political Science Review,
LXIV (September, 1970).

Blau, Peter. Exchang§_and Power in Social Life. New York:
John Wiley, 1964.

 

Boudon, Raymond. "Reflexions sur la logique des modeles
simules." European Journal of Sociology, VI, 1
(1965).

173

174
Browning, Rufus. "Computer Programs as Theories of Political
Processes." “Journal of Politics, 24 (1962).

Buchanan, James M., and Tullock, Gordon. The Calculus of
“Consent. Ann Arbor: University ofMichigan,1§62.

 

Campbell, Angus; Converse, Philip; Miller, Warren; and
Stokes, Donald. ‘The American Voter. Abridged
edition. New York: John Wiley, 1964.

 

Cherryholmes, Cleo, and Shapiro, Michael. Representatives
and Roll-Calls. New York: Bobbs-Merrill, 1969.

 

Cimbala, Stephen. "Foreign Policy as an Issue Area: a
Roll-Call Analysis." American Political Science
Review, LXIII (March, 1969).

Clausen, Aage R. "Measurement Identity in the Longitudinal
Analysis of Legislative Voting." American Political
Science Review, LXI (December, 1967).

Clausen, Aage R. "The Measurement of Legislative Group
Behavior." Midwest Journal of Political Science,
XI (May, 1967).

Clausen, Aage R., and Cheney, Richard B. “A Comparative
Analysis of Senate-House Voting on Economic and
Welfare Policy 1953-1964." American Political
Science Review, LXIV (March, 1970).

Cleaveland, Frederic, et a1. 'Congress and Urban Problems.
Washington, D.C.: BroSkings, 1969.

Cnudde, C. F., McCrone, D. J. "The Linkage Between Constitu—
ency Attitudes and Congressional Voting Behavior:
A Causal Model." American Political Science Review,
60 (March, 1966).

 

Coleman, James S. "Analysis of Social Structures and
Simulation of Social Process with Electronic
Computers." Educational and Psychological Measure-
ment, 21 (1961).

Coleman, James S. "Collective Decisions." Sociological
Review, XXXIII (Winter, 1963).

Coleman, James S. "Foundations for a Theory of Collective
Decisions." American Journal of Sociology, LXXI
(May, 1966).

Coleman, James S. "Games as Vehicles for Social Theory."
American Behavior Scientist, XII (July-August, 1969).

 

175

Coleman, James S. ”The Possibility of a Social welfare
Function." ‘American Economic Review, 56 (December,
1966).

 

Conn, Paul H.; Meltz, David B.; and Press, Charles.
"Rationality, Effectiveness, and Political Calcula-
tions." Michigan State University, August, 1970.
(Mimeographed.)

Converse, Philip E. "Information Flow and the Stability
of Partisan Attitudes." Elections and the Political
Order. Edited by Angus Campbell, Philip Converse,
Warren Miller, and Donald Stokes. New York: John
Wiley, 1966. ,

Cummings, Milton C. Congressmen and the Electorate.
Washington, D.C.: Brookings, 1966.

Curry, R. L., and Wade, L. L. A Theoryyof Political
' Exchangy: Economic Reasoning_in Political Analysis.
Englewood Cliffs, N. J.: Prentice-Hall, 1968.

Cyert, Richard M., and March, James G. A Behavioral Theory
of the Firm. Englewood Cliffs, N. J.: Prentice-
Hall, 1963.

Dempsey, Paul. "Liberalism-Conservatism and Party Loyalty
in the U.S. Senate.” Journal of Social Psychology,
56 (April, 1962).

Downs, Anthony. An Economic Theory of Democracy. New York:
Harper and—Row, 1957.

Dye, Thomas. "A Comparison of Constituency Influences in
the Upper and Lower Chambers of a State Legislature."
western Political Quarterly, XIV (June, 1961).

Eulau, Heinz, and Hinckley, Katherine. "Legislative
Institutions and Processes.” Political Science
Annual, 1966. Edited by James A. Robinson.
Indianapolis: Bobbs-Merrill, 1966.

 

Fenno, Richard. "Congressional Committees: A Comparative
View." Paper prepared for delivery at the 66th
Annual Meeting of the American Political Science
Association, 1970. (Mimeographed.)

Fenno, Richard F., Jr. The Power of the Purse: Appropria-
tions Politics in Congress. Boston: Little, 1966.

176

Flinn, Thomas A. "Party Responsibility in the States:
Some Causal Factors." American Political Science
Review, LVIII (March 1964).

Flinn, Thomas A., and Wolman, Harold L. "Constituency
and Roll-Call Voting: the Case of Southern
Democratic Congressmen." Midwest Journal of
Political Science, 10 (May, 1966).

 

 

Francis, Wayne L. "Influence and Interaction in a State
Legislative Body." ‘American Political Science
Review, LVI (December, 1962).

Fridja, Nico H. "Problems of Computer Simulation."
Behavioral Science, 12 (1967).

Froman, Lewis A., Jr. Congressmen and Their Constituencies.
Chicago: Rand McNally, 1963.

Froman, Lewis A., Jr. "The Importance of Individuality in
Voting in Congress." Journal of Politics, II
(May, 1963).

Froman, Lewis A., Jr. "Inter-Party Constituency Differences
and Congressional Voting Behavior." American
Political Science Review. (March, 1963).

Froman, Lewis A. "Organization Theory and Explanation of
Important Characteristics of Congress." American
Political Science Review, LXII (June, 1968).

Froman, Lewis A., Jr., and Ripley, Randall B. "Conditions
for Party Leadership: the Case of the House
Democrats." American Political Science Review,
LIX (March, 1965).

Gibson, Quentin. The Logic of Social Enquiry. London:
Routledge and Kegan Paul, 1960.

Goldberg, Arthur S. "Social Determinism and Rationality
as Bases of Party Identification." American
Political Science Review, LXIII (March, 1969).

Goodwin, George. The Little Legislatures. Amherst:
University of Massachusetts, 1970.

Gray, Charles H. "A Scale Analysis of the Voting Records
of Senators Kennedy, Johnson and Goldwater, 1957-
1960." American Political Science Review, LIX
(September, 1965).

177

Grumm, John G. "A Factor Analysis of Legislative Behavior."
Midwest Journal cf Political Science, VII (November,
I963).

Hanna, Joseph F. "Explanation, Prediction, Description,
and Information Theory." Synthese, 20 (October,
1969).

Hanna, Joseph F. "Information-Theoretic Techniques for
Evaluating Simulation Models." East Lansing,
Michigan: Michigan State University, Department
of Philosophy, n.d.

Hanson, Norwood R. Pattern§_of Discovery. Cambridge:
University Press, 1958.

Hermann, Charles F. "Validation Problems in Games and
Simulations with Special Reference to Models of
International Politics." Behavioral Science, 12
(1967).

Hinckley, Barbara. "Incombency and the Presidential Vote
in Senate Elections: Defining Parameters in
Subpresidential Voting." ‘American Political
Science Review, LXIV (September, 1970).

Homans, George C. "Social Behavior as Exchange." American
Journal of Sociology, 63 (1959).

Homans, George C. "Bringing Men Back In." American
Sociological Review, 29 (December, 19 .

Huitt, Ralph K. "The Congressional Committee: A Case
Study." American Political Science Review, XLVIII
(June, 1954).

Huitt, Ralph K. "Democratic Party Leadership in the
Senate." American Political Science Review, LV
(June, 1961).

Huitt, Ralph K., and Peabody, Robert L. Congress: Two
Decades of Analysis. New York: Harper and Row,
1969. ‘—

Jewell, Malcolm E. "Party Voting in American State
Legislatures." American Political Science Review,
XLIX (September, 1955). ‘

Kaplan, Abraham. The Conduct of Inquiry. San Francisco:
Chandler, 1964. ‘7'

178

Kesselman, Mark. "Presidential Leadership in Congress on
Foreign Policy: A Replication of a Hypothesis."
Midwest Journal of Political Science, IX (November,
1965.

 

Key, V. 0. American State Politics: an Introduction. New
York: Knopf, 1966.

Key, V. O. The Responsible Electorate. Cambridge: Belknap,
1966.

 

Key, V. 0. Southern Politics. New York: Knopf, 1949.

 

Kingdon, John W. Candidates for Office, Beliefs and
Strategies. New York: Random House, 1968.

 

 

LeBlanc, Hugh L. "Voting in State Senates: Party and
Constituency Influences." Midwest Journal of
Political Science, XIII (February, 1969).

 

 

Lehnen, Robert G. "Behavior on the Senate Floor: An
Analysis of Debate in the U. S. Senate." Midwest
Journal of Political Science, XI (November, 1967).

 

Lowi, Theodore J. "American Business, Public Policy, Case
Studies, and Political Theory." world Politics,
XVI (July, 1964).

MacCorquodale, Kenneth, and Meehl, Paul E. "On a Distinc-
tion Between Hypothetical Constructs and Intervening
Variables." Psychological Review, 55 (March, 1948).

MacRae, Duncan, Jr. "A Method for Identifying Issues and
Factions from Legislative votes." American Political
Science Review, LIX (December, 1965).

MacRae, Duncan, Jr. Dimensions of Congressional VOting.
Berkeley: University of California, Publications
in Society and Social Institutions, Vol. 3, 1968

MacRae, Duncan, Jr. Issues and Parties in Legislative
Voting: Methods of Statistical—Analysis. New
York: Harper and Row, I97 .

 

MacRae, Duncan, Jr. "The Relation Between Roll-Call votes
and Constituencies in the Massachusetts House of
Representatives." Amerigainolitical Science
Review, XLVI (December, 1952).

MacRae, Duncan, Jr. "Roll-Call Votes and Leadership."
Public Opinion Quarterly, 20 (1956).

179

MacRae, Duncan, Jr., and Schwartz, Susan Barker. "Identify-
ing Congressional Issues by Multidimensional
Models." 'Midwest Journal of Political Science,

XII (May, I968).

Martin, Roscoe.” The Cities and the Federal System. New
York: Atherton, 1965.

Marwell, Gerald. "Party, Region and the Dimensions of
Conflict in the House of Representatives, 1949-
1954." American Political Science Review, LXI
(June, 1967).

Masters, Nicholas A. "House Committee Assignments."
American Political Science Review, LV (June, 1961).

Matthews, Donald R. U.S. Senators and Their werld. New
York: Random House, 1960.

Matthews, Donald, and Stimson, James. "Decision-Making by
U.S. Representatives: A Preliminary Model."
University of South Carolina, 1968. (Mimeographed.)

Mayhew, David R. Party Loyalty Among Congressmen. Cambridge:
Harvard University, 1966.

Maxwell, Neal A. Regionalism in the United States Senate:
the West. University of Utah, InstiEute 0
Government Research Monograph No. 5. Salt Lake
City: University of Utah, 1961.

McCally, Sarah. "The Governor and his Legislative Party."
American Political Science Review, LX (December,
1966).

Meller, Norman. "Legislative Behavior Research." Western
Political Quarterly, 13 (1960).

 

Meller, Norman. "Legislative Behavior Research Revisited:
a Review of Five Years' Publications." Western
Political Quarterly, 18 (1965).

 

Meltz, David B. "Competition and Cohesion: a Model of
Majority Party Legislative Bargaining.” Unpublished
Ph.D. dissertation, University of Rochester, 1970.

Mezey, Michael L. "Ambition Theory and the Office of
Congressmen." Journal of Politics, 32 (August,
1970).

Miller, Warren E., and Stokes, Donald E. ”Constituency
Influence in Congress." American Political Science
Review, 57 (1963).

 

180

Munger, Frank J., and Fenno, Richard F., Jr. National
‘gglitics and Federal'AidytO'Education. Syracuse:
Syracuse University, 1962.
Olson, Mancur, Jr. The Logic of Collective Action: Public

' Goods and the Theory of Groups. New York: Schocken
Books, 1968.

 

Patterson, Samuel C. "Comparative Legislative Behavior:
a Review Essay." 'Midwest Journal of Political
Science, XII (November, 1968).

 

Patterson, Samuel C. "Patterns of Interpersonal Relations
in a State Legislative Group: the Wisconsin
Assembly." Public Opinionguarterly, XXIII (Spring,
1959).

 

Peabody, Robert L. "Party Leadership Change in the House
of Representatives." American Political Science
Review, LXI (September, 19673.

Peabody, Robert L. "Senate Leadership Change: 1953-
1970." Paper prepared for delivery at the 66th
Annual Meeting of the American Political Science
Association, 1970.

Peabody, Robert L., and Polsby, Nelson W., eds. New
Prespectives on the House of Representatives.
Chicago: Rand McNally, 1963.

 

Polsby, Nelson W. Congress and the Presidency. Englewood
Cliffs, N. J.: Prentice Hall, 1964.

 

Polsby, Nelson W. "The Institutionalization of the U.S.
House of Representatives." American Political
Science Review, LXII (March,'i968).

 

 

Polsby, Nelson W.; Gallaher, Mirian; and Rundquist, Barry
Spencer. "The Growth of the Seniority System in
the U.S. House of Representatives." American
Political Science Review, LXIII (September, 1969).

 

 

Prewitt, Kenneth. "Political Ambitions, Volunteerism, and
Electoral Accountability.” American Political
Science Review, LXIV (March, 1970).

Rice, Stuart. "The Behavior of Legislative Groups."
Political Scienceguarterly, 40 (1925).

181

Rice, Stuart. "The Identification of Blocs in Small
Political Bodies." 'American Political Science
Review, 21 (1927).

Riker, William H. "Bargaining in a Three-Person Game."
‘American Political Science Review, LXI (September,

1967).

Riker, William H. The Theory of Political Coalitions.
New Haven: Yale University Press, 1962.

 

 

Riker, William H., and Ordeshook, Peter C. "A Theory of
the Calculus of Voting." American Political Science
Review, LXII (March, 1968).

 

Ripley, Randall B. Majority Party Leadership in Congress.
Boston: Little Brown, 1969.

 

Ripley, Randall B. Party Leaders in the House oprepresenta-
tives. Washifigton D.C.: Brookings, 1967.

 

Ripley, Randall B. "Power in the Post-World War II Senate."
Journal of Politics, 31 (May, 1969).

 

Ripley, Randall B. Power in the Senate. New York: St.
Martins, 1969.

Rosenthal, Howard. "Voting and Coalition Models in
Election Simulations." Simulation in the Study
of Politics. Edited by William D. Coplin. Chicago:
Markham, 1969.

 

Rustow, Dankwart A. The Politics of Compromise. New York:
Greenwood, 1969.

Salisbury, Robert H. "An Exchange Theory of Interest
Groups." Midwest Journal of Political Science,
XIII (February, 1969).

Schlesinger, Joseph A. Ambition and Politics. Chicago:
Rand McNally, 1966.

Schoenberger, Robert A. "Campaign Strategy and Party
Loyalty: the Electoral Relevance of Candidate
Decision-Making in the 1964 Congressional Elec-
tions." American Political Science Review, LXIII
(June, 1969).

Shapiro, Michael. "The House and the Federal Role: a
Computer Simulation of Roll-Call Voting." American
Political Science Review, LXII (June, 1968).

182

 

Smoker, Paul. "Social Research for Social Anticipation."
American Behavioral Scientist, XII (July-August,
1969).

Sorauf, Frank J. Party and Representation: Legislative
POlitics in Pennsylvania. New York: Atherton,
I963.

 

Soule, John W. "Future Political Ambitions and the Behavior
of Incumbent State Legislators." Midwest Journal
of Political Science, XIII (August, 1969).

 

Thibaut, John W., and Kelley, Harold H. The Social
Psychology of Groups. New York: John Wiley,
1359. i-

 

Truman, David B. The Congressional Party. New York:
John Wiley, 1959.

Turner, Julius. Party and Constituency. Baltimore: Johns

Wahlke, John C.; Eulau, Heinz; Buchanan, William; and
Ferguson, LeRoy C. The Legislative System. New
York: John Wiley, 1962.

Waldman, Loren K. "Liberalism of Congressmen and the
Presidential Vote in Their Districts." Midwest
Journal of Political Science, XI (February, 1967).

Weinbaum, Marvin G., and Judd, Dennis R. "In Search of a
Mandated Congress.“ Midwest Journal of Political
Science, XIV (May, 1970).

White, William S. Citadel: The Story of the United
States Senate. New York: Harper, 1956.

 

APPENDICES

183

APPENDIX A

COMPUTER PROGRAM

184

COMPUTER PROGRAM

Following a summary list of propositions appearing

in Chapter II predicting the behavior Of legislators, we

present a list of the program statements exactly as they

appear in the operating simulation program. Beside each

hypotheses we give the location Of corresponding statements

in the program.

1.

1.1.

1.2.

1.3.

1.4.

3.1.

The legislator supports legislation which is 87-
in the interest of his constituents, and he 121
opposes legislation which is inimical to the

interests of his constituents.

The legislator supports legislation which is coding
congruent with the economic interests of his guide
constituents and opposes legislation which is

inimical to their economic interests.

The legislator supports legislation which coding
enhances the political power of his constitu- guide
ents and opposes legislation which curtails

the political power of his constituents.

If the legislator was narrowly elected, his 113-
support of constituency interests is stronger 120
than if he was elected by a wide margin.

If the executive was successful in his district, 71-86
the legislator supports legislation supported by

the executive; if the executive fared poorly in

his district, the legislator opposes legislation
supported by the executive.

Legislators support motions they Sponsor with 55-58
their votes on those decisions.

Party members support members of their party coding ‘
on motions they present to the legislature guide
and oppose members of the Opposite party on 38-54
motions they present to the legislature.

The executive's position contributes to the coding
definition of a motion as a party-sponsored guide
motion. '

185

186
Party affiliation is highly related to
legislative voting.
Party loyalty varies across different issues.
Party loyalty tends to be reinforced when it

is congruent with constituency influence and
tends to be weakened when it conflicts with

-constituency factors.

Party leaders exhibit greater party loyalty
than the rank-and-file.

The legislator has high interest in the bills
considered by his committee.

If a bill revised in committee promotes his
ambitions, the legislator supports it by
voting for a favorable recommendation and
for its passage in the legislature.

If the revised bill inhibits his ambitions,
the legislator opposes a favorable recommenda-
tion, and votes against its passage in the
legislature.

The value of a collective decision to a
legislator is proportional to the value it
has to him in making exchanges for political
power.

Those who have great value for the outcome of
a collective decision will exchange control
over decisions the outcome Of which is less
valuable in return for support of their posi-
tion on the collective decision which is of
greater interest.

Those who value a collective decision little
exchange support on that decision in return
for greater control over decisions they
value highly.

Proponents of both passage and failure of a
motion attempt to gain a minimum winning
coalition.

38-54

38-121
38-121

41-43
52-54

60-70

62-64

67-69

122-219

121-174

187

Bargaining

8.

10.

11.

12.

12.1.

Some legislators will exchange votes for
resources affecting political decisions
through the party leader.

Vote trading takes place between those
uninterested in the bill and the members of
the committee reporting the bill.

VOte trading takes place between legislators
uninterested in the bill and interested
legislators who share their committee assign-
ments. '

Trading of legislative resources takes place
between legislators uninterested in the out-
come and legislators similar to them in
composition of the constituency which elected
them.

The legislator exchanges his vote with an
interested legislator who Offers him the
most in terms of resources in exchange for
his vote.

The legislator takes into account his own
interest, if any, in the outcome, when he
makes a decision to exchange his vote for
resources affecting the outcome of other
decisions.

179-184

185-200

202-216

195-200
212-217

221-222

221-222

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

«no .on.~v£3mcmm
y-ene - :- u4--.1-1.-:1-z-:-:.:.---: .eu111ccewa - -----:-
ox: m.au1 e no
-- III!!! 0N0 ill-(flip I-l---:.-:-:..-1.1- - -.- -. mazwhzouh , - -1
BNO .ouhnvzam»<»m
omoii. vii: |£¥§:i¥!-i!si-i! .an~V_ch<»m. .-i;
mNO om~4u— 0 OD
-Illill- (I ll vmc ll!--l5ll:li£?ln-- {I it-- %-: mazuhzoo N - -
. NNO .o»11.o.z;mzou
l-tlf:.-lt:-!x- é-mNo. .|?, Ir!!! {ll-21:: i It- 1 - - -1 1. 7;:- .onA—efivkm300
HNo 4.4": N as
ituust -.oNe - . (11--11:1-c1- - --:- I .:- - c.4u1 N on -
oNa mazfiezou a
:::st--:ati:;;i.2;!ulteuz;tll-i;|:i;rmao (iii--r-i- 1-x- -; - -;- - - ..ouAHszmced -
nae .eu111emcac
.s:--.:azr:p 1::- :e.cHs --z-ar::--.I.it: I: - - - 1-1. - - - - ~.4u .fl no --
mac -:. . - - oneoazzwm -
- ow z|:::-:.-I.-!-- eae.-:I-1::;-.x- 1; e :ezza -- 4-- .cuwec>zam -- -
:+Ii-.nw.-.:a::.li.i::s!ttss.:-l- -9 nae-:I:-- ue-I:- - - - -- z .ouerOw - -

.ouememo
.euemczam
.enecoz

. .ouozzcm
...ouezmm>
.onmm>rnm
mandzou mm34<> mN1J<1»_7H
omeo>1..moco.domeo.aeeemnm
x..1n.«u1..1.zouoc.1o.eulq

op<ymo..a<dn .mon.ooca<mm och

mJgum z“ mama u 5.1--.

193N.Nu1.1c.Nuo.xo.1.xzam1c..v.auo..n.1.xou11.11.a<wn1q
»<»m_.1~.x<d1.._ca.. .NoN.occaamm
mmceacmlcwn do mmesmlxeea 21 game

u

 

 

-lzsei.:; a: a: 21:- -- oao
- -1 - ::I.=l: coo -- --- - -
coo
- z:;:-t.»uls».;-!-2 |r- - a-..nQo.- .- ., -
z<moomd z. a
r-.stzii!z ;$ln-!s:;1;--t :--.--coc:;-. p14dm.mopzaz.ieoa.fin1.111
.dxad1.1ma.au1.._.zcdm-c.1na.a.1.111zoo
--- :. -.34-1--;¥- .- ..1czcun...m.au_..1cummw.1.a.au1.fi_
--1:-;:.:s- -: -1: :..-=:_:- moo: . . - -
..a.Nue.ie.Hezoonc.11.08e1.~..
- -‘:|.z::-.laa. -;nl.s.¢ea -4:.. --;Il-:--.- ,-- : -
- .3-1Iinlzl:n:!-- -:IIEMco:2-
.;::. Nee .eeacomzm.No.mczamzcua.wcrn . .
|€¥!041!|to{ilillils:l!-1-!!erwUmaomvhxpdpm.amvhzamz..vaaqumwmvhwmw
I. Ii‘-.(fltl.(-(l(-l!((lul(! (I mnHvDWFO>~ s AnHvZCUYQ A ODva<Q~ snmflv , L.
Jilin autumn.“ zoom

I
«dd:;:;:r-co o.oodczooa..o

.pndeao- cavowmu.

a0 ma

 

<e.esezlneo waaeweo

exez name..;.

ao>zxsm.zoc<z:w anoma7~
dsmcaem.lchsmaaa.1a.e1ean.;:.;-y----.5- - -..
M81q084.181ewae.laceaemsN
5 1c oouvx2<m_..co

dwhdhm~.500HvQ— 7OHvO<wJ~H ::(-5.::s(:

(III-Ill, In." .
.l-I...-,-1-;.:-. ,2- -11: 13.150.1312 . 122121
1; (((11! v x .
, - .-;a:;/:;:/ .; -79 z m 1..
l(-I.I J l‘l'll’l'lt 1' 11,.074 I! I1...” I:
Ii: III . l!il!!ti .I:!!/ noomm I11;
"I l! Ill-D! (VI... a
ti’!( {AI , r] "I'l’l’l ".'l"l"l I
)l‘l’l’.) lf’ /I“ I/ I ’l I,
(I( {I I 1”} I I".
I 1!; Ir/ I'll/'1‘ o l '1: (I’ll: {’5'
I, I- ' l/nlll/v‘ (: ((117 .(- I I! I.
(.1 III" 1’} (ii/(7| . i. It I I}...
I1:(((( )1! all II Ill'lt 1?. }/
Ilv III A, I (Cu/IX I; II“: I 6.} (..l I .4] vi!
I! (ll/[II ”/ I. l.- l' i /l '1' I. I'll.
(ll: Ill [/14 / (III: /.
((/ll..’ / ll 1’ 5,! lo. ‘ l
/ i ll" ( l I l J
I r I)( I ('/-/.:I.I z/{w(! .
(If ’1 rl
I I (ll. .9 I
I, /‘7 I

.-1u

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

. .03.. gel 351 Z on
mac -.--s xsnx- . .,- : Lezlezoo ma ----1
see -1. 1]: ani- . :a.1meo>1cemca .me5511ewaa ea-
ses a : -z. 8a ce ac . -:i..-;_ . :-
.:- Nee . am o» co ..eczoce.cm.11.mec>_czco_1 c1
ass a a- a ma ea .--|.-l.- .s,-., .-
- ,111ese- .--.§- .--.1 4 8-1 ma ea
. . mcwecocua >e1eczls cz< Homamez1 wmeelxzcc we<szonao -u.- . -. -
ame.:||x- .Illll:-:..ia|i;é;. mszleycu ma
cae z.a-;.N..meo>11cw¢861mce>11emea Na .----;- -.
ems ea ce ec
cme . Na 3 8 11_.zodmw.cw.1m8>:31 c1 :1:- -l1 --
mne ma a-1 ma on an
emmawcz1 macmzoam deaoscnao Lu -.1 -1 :
..lanr ems .a..ueo>1ccwcc-1meo>1deuce sea
was .---:-. 4“ ea ac -.- - ,1- ..
Nae cow ce cc .e.eo.1meo>11e<1511 cl
mm fine ...N..meo>1.cmmdu .meo>11emea ea .u. ..l.--
11 one 1H 0» ac
N.e .uyalizl:.a.1mec>1.cmmaulmee>1cowma aea.7----:.-.--
I moo .3 or 00
N49 - I-coa.s~;ce 1e 8 ..weo>lcc<wdlc cl
: ewe .a.lmec>1camec-awec>_1cwma o
mes made 1a” ca 8.51 oh so a. x]: ........... l -
3.9 a." o» no
nee --;.N..meo>11emmau1mec>_.emmd sea :1. I;- .
Nse ..y 48 as ea
Nee no” 0» as .e. eo.lweo>1.cams11 11 1:11- -i:- ---z
ave .N...weo>11cmea- lmec>11e1ea e
ass a 01 ac, -:::-1.I- :- .--
1: new - a o» co leaaa.cm.1wec>_cx<ilc c1
>ee<d IP13 mczaxcxw ceacsoneu ----- u \z-.-:-;1.
“M“ ::.:.,-s -.--,-, . “auiwco>11omaa
.:y1IIIIaI1yllplllrylullulllnlunnlu.I -;-:- DON H.whc>1 Ne on c. - --;x - - - -
- mes -;:|::|«ulnillln|: (mmc+mDmMcmgam+-wau«w.w>1eaomx nummma u

 

 

...... m,ch-oer1v.ec.
e114.
(:1;:1::::a::::aar:
cam-mm-cc.-m551m .weoz

(Ill/(f. y- . - wazuhzou v
M.,/4(1)) (.- -1-
// (117-X
i/Vll/l (ill/1

com OH ou.mra.z-:Iri-:t

o o» no

mo. H. H4. m<lfiv .-
o-z .we< Lani» “w

 

 

 

 

 

 

 

 

 

 

 

 

 

. o.1u1 mm on
-1 .MMW 1 1.11.!.:.. 11 . --;-...1 -. 3u41m1 -- --.--
mkmmmw1z1 m1zw7111mzou w1<JDUJqu o
--1 .001, 1 11111111..11--1 “9:11zou nn.. - -
103° .1.1m1o>1.awmau1wpa>11ow«a an
11. mac -1 1111.111 . an o» ow - 1-
- noo111111111111311111:. 11-1smn o» oo 11111<1mu.om.1m»o>111<1m11 11
. 1 can .-1... - . 1.1n1 Mn oc
mkmmxm1z1 mm1<1m w1¢JDUJ1u u
- 111 1 1-11 11 11111111111111}: 111111..mao 111.-:1.-...11..1..11111.. i111l111- -- . . , 3322.200 an .
vac .1;1m1o>11ommau1m1o>1vaumm on

 

moo . 1m 01 no,
.mao. . . . . ..-. _. on o» oo 1oa.om.1mpo>11omm1u111owxw. 11 om
«mo . . : 1» C1 oo
.1.1mho>1.9mmau1w1o>1vomxa mm
Ono ., . ....1 ....- .: . ... ..;- . am C» on
was cm 0» cu 111vommn.cw.1m»o>11uwxa1 11
111 nu ---::1mo.1111111-1-1.11111111121111-1 .::; . m11u1 1» on - ..1-
a, ~mmmm1z1 J«zo1omm m1<J:UJ<u u
. . - ..019 1. ., .. . . -.: ; .. .... . ”:211zoo 1N . .
moo .1.1m1°>1.amxau1w1o>1.owxa om
1-1111111.:111 1-1111111111191111evno 111111111111-111111111111111 -1111.1:- :1. 1m 01 oo
nae . .1.1m1o>11amzau1m13>1vawma mm
.ma: 1- .. .... ,. ... .1. . :1 z,.. -.: . moms 1oN.m~1 o» oo em ..1.-.
15° 1N 01 on
.cno ... . . .1 .. .; .:., -.1.1w1o>1.awmmu1m1o>11ammm nm 1
01¢ 1m 01 oo
Q19. ; - . . . 1- ; .... . .1.1m»o>1.owmau1mho>11amma mm
“no mean 1n~.-1 o» oo 1m

 

 

 

 

. ‘1‘. It- li'll1 . .Illlllo'll' I‘ll-1‘1!!! I'll-IIIIIIII ll ‘1'"! III.|'I||,III Ill-11:10. «1 II. I 1|: IE

01¢ 1m 01 co .,.

muo . .- :. ... ... .. ,. .;1~.1w»o>11owxaa1mha>1vomua ow . ..

.11, . : . - 1m 01 oo . -.. 1 -
1 -n1o . .~.1m1o>11ow¢au1m13>11owma o1
:111111111111111- - ~1a-11-1-1111111111 211111z1--- 1 ,1,- --. moan 11~.o11 01 no m1 -;,1-.

1 1 . 11¢ 111, -111; 1o.mpo>1vzoo1 1q~11~1m11 e» co

,ono. . zo111wog1mpzwa1mwma 21-1mmmmkz1-w1<43044u;.11.u11.1111r1.1.-
11n11111111111111111 ..-:-1. . . m3z1hzoo “a
-.111 11111111-mwm-- 111111111111111511f111 w 1mho>11nmmmu~mpo>11omm¢.oH :1:19,.:

.--1- 11 111111111111111111 a“ d».do.Ndwqxoum.om.~m»o>%wnww.mw1z.1-11

 

 

 

 

 

..‘il'l / , (1"

If #13 {1 1:01.» |
j I‘- Iffltl 0 .
. Ill.
11(1)! II. {I I'llll/
I- 1"‘ll! {ll/l 1| I I,
It}. lv|l I {I I’ll: 1'1/ I 'Ill'l'
Ill/II; ’{ '1 ‘1 If .
j ’1'! 1 O. (I!)
1 .{I-I ‘
I; ’0 j’l 1 II 1 "ll [1"
1 [1‘11 I / ''/l .1 I1 I. I, I
[1" "ll'llll I II! t I 1| I’D» ‘1’.
.1114! I [’1’ I111] . 0" 1 :11.
.l 1))! /; . 1l/ I III-ll I. I!
f. a ‘ Ill: All 41 ’1
II '/J |’/, I] II I! I I I
'[l' I .111! '/ 1 JA/ 1" I/ '
If) 1'1” [I II '
If, /1’ ’1 (All I I {1 O
// l. /.I ll. ;
I!) O} 11 .I ll I
I’ll I1 I / .' I l 11! I III’
1 I? [01! [1].! . 1.
11/ II I j! I!) ’lit’!’
(1.! III: .1 Ir: I.’
F/ I 11 I I ‘1‘}
f I, (I 1 1". I
[I I1” I
/!- l
I
I 1.11
111! 1 Jr
1!,
1 1 l ’I.
I I!)
/ '1’

 

 

w

_

.1
. 1C1 CC C1 CC 11C1C>11CC1C>1. CC.C1C>C1 C1

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

1 Ca" .11C1C113C- C1C>CCC ..:1111.11 1 . 1
CC1 .1o1CCzCC11CCCCC
_ Cm1 -1 1 ...-. -.1.mm1zCCuCC1ICC. 11-111 .
. 1C1 1»C1C>C me
_ CC1 . CC C1 CC . 11-- . -
1 1. CC1 av C1 CC 1. m. CC. 1C1C>11meav 11
h . «m1 .1o1ox1z3mowmu 1Cz1::wcma ....... .1
.. nm1 111 .1.1Cx1xam1<1m-1vazam1<1w
_ mm“ ...1.1<xvzamm<mu1<:1:sma«m .. 1
1 .1n1 . mCz112Cu we
C Cmn . .wc1~.axvxamZCUu11.szzzmzco : 1111111 11-
o~1 1:11111 .11;,11.C1o>112CC1uCz
1 C~1 . C.1u1 CC CC 1111111111 ..
C 1~u .1,;11-111 . 1m1C>11CuC11Cz
C~1 . ...1-1.1 1C1C>111<1C11mx .111. 1--
mm1 1 111.111 1w1o>11a«q1u<z Cc
C m“ ¢C1 .112111- 1-. a C1 C1 CC .11 111-. 1
11111 CC1 CC C11 CC1MW1.CCHMC1C>1~CM1C>1 CC.1.CC.1C1C>11CC1C>11 C1
-1 CC1 1n C1C>1 11 CC .11111.1 1 ..
11, mCC<> >m<zzam m1aazoo u
. 111 1111111.. CCz112CC NC .11111: 11111- .
1 _ Cxfi .ao1w1C>11¢mzan 1m_C>11CwCa 1v
1 . . C11 CC C1 CC 111..111-1. . .-
C11 11 C1 CC 1C.Cm.1C m1C>112CC11 11
111.1.1m1C>11Cmmau1m1o>11CwCC Ce 111.1111 .:
. CHM we 01 CC
. C11 1-.1 1m1C>11CmCCu1m1C>1CCCCC on 1 1. :. .1. 1 .
v11 111111.-:1 . NV C1 CC
C11 CC1C1 CC.1C.Cm.1C.m1C>112CC11 11 1111..11..11.1 :
~11 .111C1CC11CCCC1 1C1C>11CCCC CC
111 CC. NC CC 144C11 11 .-11 1.- .
C11 . . mCz112CC 1C
oC1 114<m1.41m1 on 11.1 . 1.-- ..1
C01 11 C1 CC
5 c n {‘1'} 1 ,. ..1.
1.1111111111111111111111111 0C1 . on CC CC 11117CCC. CC. p.11. m1C>112CC11 11.mn.1:111.11 1..:-1
111 mC11111111111111111111111111J1- 111 1
111111n11111111111111111111111111wC1 . 1-111.111.1--1..-. 1» C1 CC
an1-aoq<m1u4<m~ vn 111111 111.1

1. . vn.o11ou
. 111 CC C1
. 11 11111111111111111111111111111111111111111111.I1111111111111111111111111memwm1CC.11.m1C>112CC11.mw 111111111

 

 

I IllIIIhI: II III vI'ic.Il III“ I 'I'IITIIII1'IIIIII II
III! .I 1.il IIIII illl'0l1l'll.lI.Illillf-|1'i III I‘IIII ‘ II

                                 

 

 

.IIIII III I

C11

n1xmmmC<m1 >1m<a :11: 2C11<C1zazzou

..... 11 -C11 1. 11111111-111111; . CCH Hux1111 CC CC
111 1: 111-11.. C11z1sz
1.1111. .011. -I:1- Cnxou<sz
C11 . .-111111111.:111é...-:-.1:V .--..C»CCCsz
..111111111111 1 11 .11- .111111111111.111::-:-1:11;:- 1 11meooma 2C11<C1zszCC 21Cmm
. v11 ... 11 01 CC
:111-;:11111111 1111111:11-C11 11111111111u1o.>11Cm1C>1. m1C>C 1C1C>11Cwmn.1m1o>11C1 1CCN. C1111CC
N1" . m3z11zou

. . ... .1 ..- 1. 1.1111. C211mzouu 11. Cz11zzou
o1“ . 11 w10>11zou1n C2
C11 .1.- . . . 1 .1-:.- ,.1...-.-- o.1n1 CC CC
CC1 . : .C+1C1CCu1x1mC
1C1- :... . . .1..1 : 1. 1. .-. .111szu1CCz
0C1 1.: .111C111CCCCC1C211CCCC
.1111 .11-m01131lt11 11.11.-1!-1:11111 : .1 -;-.11.1m211m1C1mu1mz11a141m
% 1.3 . ... .111111111111211115152
111111-; 111111-1.CC1.11111-11111111111111.11- : - . -.1 1 . .111msCmu1szm
~01 . Cm C1 oc
11:1..-. . ...11 . - .-, . ...., 1n C1 CC 11m1C>11CU1C>1. om. w1C>C1 C1
C01 .11w1C>meuw1C>sz
.1111111111111-11.. 11--CC1-. . .- - -1; -.. . . - . .-s 1 .:- - ..«11CCCCCu1CCsz
CC1 . .11szCCuszCm
1C1 1.. 1-1:-. _. 1 1.1-- - . . , ~um1C>C
CC C1 CC
CC1 CC C1 CC 1.11.m1.1C1C>11C1111 11

 

 

 

 

 

 

 

 

 

 

 

1|. IIIIII .IIII..II1IIII ..IIII..II:II19 -. .... I

IIIIIIllull I": -' 1I|II I- «1‘. 1 I 11I. II

1 111 111 111111:111.nm11.mz1 111-Imm1o>11Cm1C>1. m1C>C11m1C>11Cme 1m1C>11C1 1nC~ «C1m11m:
Nmfi mDZ~hzou
111111 1 1:1 1m1111111111112111:11:11.111;.11- . 11 .H.11. C111mzCCu11. C111CCCC
Cm“ , 1..1 111m1o>11zCC1u Cz

CC1 . o .111 1c CC

.113 , . .1.1C11C.1111CC
or“ 1111 111111111111 1111111111111. .111CCC1.1CCC1

 

 

 

 

 

 

111- 1 cc“ -11 .1111 1111111111111111-::111 .Hu1mz11m141m 1
. 11 . 1111111111 -1 1 mzv1m
www:1111111111..1111x111111u11211nmuw

‘I‘I’III
’JIII'I

1 [XX/111+:- 111 11-11- -...1I1111111.11-».1§CC.1§CC

 

mn1or1oo1hpnrunwx441-11m.11. C2<. 1x1«111m<¢1.om. 1w1o>111<l11 11-

.IIIIIJI. 11111111111;- ov 0» cal: 115-.

mm
mm

am

on
0v

cw

1111111111111111m11 . .1+1C211¢Cmmu1Cz11wam1111111:.. ..-

'.0

CC

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

i-i. ZN on o» ow
oflw --a.-e-:. --a.zou<zamuxaoazam - -s - -
.- mflm -:::ii:|;:::-ni-- -.|- .«.zou::muzouzam
vaw - s----.14<»~.omma.uz~xaanz~zam mo. - -
n3 i - on: 3 co
«.3 mo 3 oo 2.134:«zoniamfiafiko::83. .: do .- - -- -
- - «an i i :3i oo 0» oc
. afiw - -1-.-..i-.-d§:oo<zamuzou<znm - i
i- i- iii oow-sli-i-i-I-.-.-.,:-i..3-:-..51:-- - - i J.:ouznmuxournm
0cm . - “x4<»~.owmaokz_zamuLz~zam no
saw ¢o o» on
com :--s._».o 3-813.5..533333 .2 No- i
- .mon-igiae.;;-I---i.-;_- se:-s... - -:- - . co“ 0» ow
«cwwo o» ou.fixux4<»~.zoo~.om.n_.w~o>~.xou~.oz<.a.»o.._.m»o>~vzoo~v a" -
mom. i «33. no." on
New iiiiiii-i. - s-ri-::-.-za.s -.: c.an. ooa 09.? .:- :- -;
.s-z -, wwwph~zzco ukd>_ o
.5 How - -i: -::-ii-:-i--- . u:z_»zou «o ; -:----
- mw i.eou oo o» ow
. co” - sz-- -- -:-- -.. “ozou<z3muzau<zam ...--:
- on" . .d.zou::muzouzam
sofl :azs:s!:is:i-.iis-sis -aiizis- 5:44huvamma.uz~zamu‘z~z:m co -- - - -
i .oom iili.ii. - - . . «o o» oo
ma“ on Oh-do:-«.x4<gw.zou~.ow..n.m»o>_vzoo~u.u-om -1- z -
.- uivofi no o.— 00
no“ as:ii :-i- -1-- «+xou<x:muzau<zam --;: .:
«ow .«.xouzamuzouzam
Ha“ i- .- i -. 33¢:.nwmafizuznmumizam am --i- -
-- paw -s am 0» no
on” ; ---on-o~-ou .o.»o..14‘-.o<mg_. u” an- : -i---
i s .00.” i-i---3 -i--- 40 o... oo
bx“ um ow cu ..~.zou1.ow..x.xg<»..zoo_. h_ - .-; - :
- .2...“ ii... :i iii-E.-. v33. 3 on
ma" ».Hu~ do on om -
iii- _- i «iii-iii!!! iii-Em 03:2: mECéS-E; 22222228 0
-- i12aiaa- iii-ii:a- .inmmlius fl cu : oo o» no i;.. :;i:- z
- :- mm .:i-:;i----u;:iz; ----- - i -. +2 «z u.
iii,-ifi-ii-si--iiiiiiwmmii:-ii-i-:iii .1 .Yzouzmmummwnmm , -.
iiiii-ii - i if; $.n-.w¢E.,-L.~z3mu5:3 mm
iiiii. Iii!!! II. -
liliiilllilliiiiliiiiilii iiiiiiiiii onopoug_-?i -
-ii iii .I/ii iiilviil o
iii. iiiiiiiiiirri. .iitiiu:- -i;---i
riitiiiiiziiiiliiiiiiMHHHWiii/iii:iii- .iiii-i-iiii-zii: 1:1:i- ---------
iii-iiii-ii/ii/i Iii / Iii/II! . il-Il/ i4 v.- ai .
{XII i: /ifi/ iii: {ill/l
i/Xii- ii - -/i-/- - if .......
///II III: ’- Il/ uu/i-/ . -
Il/s/ii/o/i: iii/if Juli-il- J..-

 

I | Iui ill... I‘i'; ‘0': «I I ii ll Ir :1! ii 'ill 0 i b I ii in I. i I til: i
c
-lll|'-.llln- "l'ili I-II.. Iliunl 1i ..i ..III.. I . . II III! 1.. Pl! . .0 viii: 9 .t l|'-n¢-

 

 

 

i

 

. lulu! Itiill ii! '1 "iinl. It‘ll!

 

 

 

on“

- :a.iuii-aI-i:: -- mnw

vnm

1-1:;i.- -- lizili:-i -- - nww
mnw

Iii-Ii :--ni-::nfliiiiei- anm
onw

:LIIII:-s- .i i.i-l!omm
cxw

aiiaxti--i-iii-!+iiili-:i:-- iii-swm
cww

--ii..---ii::- :-i:ii:.:- -us| mmw
vww

-:--ilii111£:i .iiznmw

 

 

 

nh-o» oo:.fimho>_.am»o>~.cw.mho>w. u“

 

 

 

mazwh?ou mm
:ii:;i-:-ii:i;-,:- -41H+A~Jozvznmzoou.~.oz.z:mzou-
.fi-m»o>_.zau_uoz
o.au_ NNH on
a::;-i§,12!-is-.a+6uxvzamowxu.uzvz:mcma.
.a..mxvznm»<»munmyvzamhqpm
-iii-s-:ii-.T:a!-- -.- -|1 - .«+A<zvz3mm<1u.qzvz:wa«a
.«.wpo>zamnmpc>:am
:-i-- e--i- -- : - .a.; -- - . .Hoozromnczxam
mumpo>1 wk

. us 6» cu.
sou«:3m.a
:pmJDmmm.Amho>~.owpo>_.mho>a..m»o>_vommm.hm»o>~Va" .¢o~.4o.wh_m3 an
m:z_hzou oo

 

 

 

 

a

iii-Iii, vi . .---.. ..w+:.ozvhmzoun:.axvhzzou -
A_.who>~vzoq~uoz
-1-:-i---- in----:--;:ii: .-s.---i - - o.fiu_ on on
.H.»mmm>uhmmm>
-.iii-- -iz-iz i - -zli. - - .«+.uz.»momau.ozv»mcma
.a.amzvhxh<kmu.mz.kahqpm
xi.i--sii-I:-;::;ie-:i::-Ii-a -ii- .1-i- .«*.«z.»ma«au.<z.»mm«a
. .H.»my3mu»cram co
.iiiaii-i;:ii-ii:i:aIs-;:su-i- --w ;-:s- i -. - a“ CF ow - -
no 0» oo ..mpo>~vom»o>~.ow.wyo>nv a“
z -- - ..i!. :---:-s- - maz.»7ou hem
.a....oz.xnmzouu.~.az.z:myou
t-iliiriii-;:;ii-:-:-:- ii:sl: -;.:: .1 . .~.m»o>_vzau_uoz
o.an~ New on
ie,-raili-i iii-iii iiiiii-i. Iii-i - - .«oaozvzsmommufiuz.z:muwa
.a+.mxvznm»<»muamz.zam»<pm
ssiii-z-2-x:uiiuziis-i-s-i :; .i;s - -- .fi..<xvz:mm«au.«zv23mm<a -. i.
: .a.who>zamuwpo>zjm
.a.mm>r3mumm»x3m - - -
is i. --::;---s- --uo.o~.-- m. Huwho>a so
i i. -‘--i--.-i---.iv i .éuxnwwzzarfio: swamfiwwmw
.ixsii:::n:iiii-ii---:-:a-- x-i-ii. p~moam~owma 4‘z~u w»<qsog«o ..:.u.
.Iiiiiii-riii:iiiiiiiiiiiiiitii. .,--.s-;::;; wwn~pzoo co
iiiiiiii:iii:ii Iii:ii:!s!ii:i!.:-rrii . thou coas;- -
{LII i1! aililll/i I’liill-i i 1 «III iii/Ill
-iiiiiit- : iiiiiris- i-i:-
iiiiiliiiiiii iiiii-rs ,- :21; -i:
fiii/ I/II/ ll-l/p/i Iii/4. i/l’
lii l i I iii/ii/-/ {-lliliiiali l.) I I :
{iii/Ilia i I ili/ iii Iv Iiii/i- r-i -
- I!- - - - I I i tI-i}- /o/i.i

 

 

 

 

 

 

 

 

 

 

 

 

 

 

iii. mow i- --i- maz_pyou cow
com -a,..-;:.;--;- , -cafi CF 00 .... - - --
--: now .2!:- . p.41m.mmm::z .mam‘fio. w»_«3
Now Fmaomm.»mom.oza.mm»a .Hfi~.Ho. mh_m: .-,::
do“ --.- ; -:, -s.m»c>zam\haz:muhalcaa
co“ pmozamxhapwauhmaa -.,. - . -;
- i.oow -:::::ia---i;. :- -.i , oyzam\paozuC7a
. mom .- - mm>zamxkamw»umw>a -----s
-n-ai. i:::-;s- “mm:iiiuirii.sia:z+o>zx3mwwho>x3m.pmoz:m.ozzam .mm>:2m .ofl~.eo. wh_«3

 

omw .waz:m.»m»mo.»aoz.»mmm> “oo~.ao.w»~¢z
-i-momi- -::-s- -.-.--:- uazxkzou fie
en“ . -..:---zoum.1.u .oom.aovwh_wz-
.nmw - -, . . ,s , .~.w.z:mzoux.~.1.szcuuzooa
New . --;-. v.fiuw do on
«cw . . ... . ,3 -. - o.au~ am on
own 1 . - .-- u:z_»zou ow _
-; zi,-onw- --e , uwm¢._ .ho~.4o.mh_mz -
mum . ...-..-:a .fivznmomm\.~.»aumxuomma . -.- s
: . - -. - s - m.wu_ cm on
on“ -.-- .. - mszprou on ; ;
s i-i-;.mnw-ezi-iiis--:-: -2-i.i.-i:-s-.,-,.._;--i-- m<au.~ .mo~.flov mkfimz
«um “nvzamm<m\.~v»am<1um<ma
»NWIs ......... iii-:-.--s:--iis- ;:::. a---. iii,--- . ~.en~ mm on
th a . mnz~hzou-n
as“, - .-Mw»o>~.ow»o>wnmho>nw.Mho>~.omma..m»o>~yo” Ano~.ao.m»_a3
cum ao»o>zx3mupo>zzwm
.ai .ils-iiiillii;iiiis- cow-riiiz-lull-an-a-ssii;-1 ---ii.Is-. i.e-: --- oumpo>n on
on“ _ .-.- - - mu 0% no - - -.- - - -
is in now-Iii . --.: - -. .-z.- -: 20042:».H
pm49mx¢..n»o>~.owpc>~.wpc>a.~upo>~vowma‘.m»o>_.n_ .¢o~.ao.m»~m3.m~.
i-e.-iii iii.--i---i - cow-i ii i --lli Ii... .-i. iii- - - - mazwpzou tn
mow s.a ...«..~.oz.»xzouu.~.az.»xzoo
-;- -usii-:i:i::i-:-:-vowiiizziiials-ialiii:iiiwaial- ...mpo>~.zou~uoz
Ili-...II i -ll - .-i-----i- i liii - : O‘Huh v“ on -.. .-.. . iii - - «-
wow - .Hopxozubanz
«ow .a--|-:-s;:i:ii-.a.~uz.»mwmmunoz.»zuwm---:in.::i---.i-..--
cow- . -inlz!||II;I.-- .H+.mzv»mp<»mu~mzvpahqhm
omw.iiii!:ilii i iiins.- ii; --i:-.«..¢zv»mm<mufi<z.»mm<a----;- :- .-. --
.ii-::Iiifli-i:: omw - i i. .a.»mzomu»mxam nu
5mm . :I-i::-.-:m-op ou-.;:s. -: .2

 

 

 

 

L95

 

 

1|. li-‘i-Oi- AI

 

 

 

 

 

 

 

 

 

 

 

 

 

 

ilIl-lllll,l ll .Iu..-|IIII

 

 

 

 

 

 

 

. It... l.||'li." I‘ll"

 

aw:-
mom

-
now

now

01!- iii

.n.ou.

omho>~

 

'nlll'

row.«~.

:N».mu.
.-un-mu.;.km oz moxm xv«.n.mu. #1 mm» mommxnfl.xoqvh<zmou Hflm

«on-:i-n-.:s-anmh. p10"
.n.mu.

awym>wsuxafi.au.,.mko>ms row».ou. awma xo.n~. ouzn.xoav~<zaou ¢om

. czw.-
_ - . . .oa<.m<._<zmok mam
4(PO» coma xoa.n.mk. hm kmm com; zmfifl

.nw. mpo>z xm.o.nu. w.c>z:mxcfl.xmd

-Jo.numi-pwazamzo.xm.o.»uns-ozzamxu.x¢.o.nu. mm>zamxo.xc«.p<zqou cam

.o.»u. pxo_mr:mlofi.xn.o.»ua

.xw.»m-Fmaxo.xm‘o.nu.xm.hm oz:m.x¢.o.nu.x~.»m mw>xo.xC4._<xaou can
pxonm zonpzoaomm za~.«~. 44> xm.«~. zouzv.xoav+<zcou com

z zo~kmoaoa¢. xdm.aa. -zo~uwmzo.xD4vb<zaou hm“

~1o_a zouhmoaomm IHN.H_. >pm<azm.xoavp<zmou mam

.»~.x~.n.mu. Fm¢qowm¢ Imfi.fifl
.HH

-.xd.«wwmwdquwm.wnupxw_m.Nm.afiwnwmaxnunfi.-n.flfio.-m.~_h.awvhqzaou Nam

!.' 'I‘II'I I'liil'ii'il II'I Ill i .Ili' I1-| i! 0-

mpo>w xo.».¢u.i,awma

..rm.n~. ouzn.x°avkqsaou now-:---
.oa<.m<.xov.«~oa .

s~w~..~.nflfi~.fl.o.a..mm.fl~.n~.»<zaou «om

Iii-i " a!

 

 

 

 

 

 

ig'. It'll!
'lll‘c‘l‘lll
I {.II 'II-llll' n:
{Allin-In-“
'I 'I..- l. .
..- .all' IV
i 1 I... I - 1

.- i- achmi-iii? - - -

APPENDIX B

CODEBOOK

197

Coding Individuals

 

 

Col. Variable Value
1-3 ideatifi- l.Dodd 31.Hickenlooper
33;;22 2.Ribicoff 32.Mi11er
3.Muskie 33.Carlson
4.Smith 34.Pearson
5.Kennedy 35.Humphrey
6.8altonstall 36.McCarthy
7.McIntyre 37.Long(Mo.)
8.Cotton 38.8ymington
9.Pastore 39.Chrtis
10.Pell 40.Hruska
11.Aiken 4l.Burdick
12.Prouty 42.Young(N.D.)
13.Boggs 43.McGovern
14.Williams(Del.) 44.Mundt
15.Willians(N.J.) 45.Byrd(Va.)
16.Case 46.Robertson
17.Javits 47.Hi11
18.Keating 48.Sparkman
19.Clark 49.Fulbright
20.Scott SO.McClellan
21.Douglas Sl.Holland
22.Dirkson 52.Smathers
23.Bayh 53.Russell
24.Hartke 54.Talmadge
25.Hart 55.Ellender
26.McNamara 56.Long(La.)
27.Lausche 57.East1and
28.Young(0hio) 58.Stennis
29.Nelson 59.Ervin
30.Proxmire 60.Jordan(N.C.)

198

199

 

 

Col. Variable Value
61.Johnston 81.Mansfie1d
62.Thurmond 82.Metcalf
63.Yarborough 83.Bib1e
64.Tower 84.Cannon
65.Cooper 85.Anderson
66.Morton 86.Mechem
67.Brewster 87.Moss
68.Beall 88.Bennett
69.Edmondson 89.McGee
70.Monroney 90.Simpson
71.Gore 91.Engle
72.Kefauver 92.Kuchel
73.Byrd(W.Va.) 93.Morse
74.Randolph 94.Neuberger
75.Hayden 95.Jackson
76.Goldwater 96.Magnuson
77.Allott 97.Bartlett
78.Dominick 98.Gruening
79.Church 99.1nouye
80.Jordan(Idaho) 100.Fong

200

 

 

Col. Variable .Value--
4 Party 1=Democrat
2=Repub1ican
5-6 State 1-50, in alphabetical order
7 Region l=East (Connecticut, Delaware, Maine,
Massachusetts, New Hampshire, New
Jersey, New York, Pennsylvania,
Rhode Island, Vermont)
2=South(Alabama, Arkansas, Florida,
Georgia, Kentucky, Louisiana,
Maryland, Mississippi, North
Carolina, South Carolina, Oklahoma,
Tennessee, Texas, Virginia, west
Virginia)
3=Midwest(Illinois, Indiana, Iowa, Kansas,
Michigan, Minnesota, Missouri,
Nebraska, North Dakota, Wisconsin,
South Dakota, Ohio)
4=Mountain(Arizona, Colorado, Idaho,
Montana, Nevada, New Mexico, Utah,
Wyoming)
S=Pacific(California, Oregon, Washington,
Hawaii, Alaska)
8 Urban 1=high (70-100%)
2=medium (51-69%)
3=low (0-50%)
9 Rural 1=high (26-100%)
non-farm 2=low (0-25%)
10 Rural 1=high (7-100%)
farm 2=low (0-6%)
11 Non- 1=high (5-100%)
white 2=low (0-6%)
12 Owner- 1=high (71-100%)
occupied 2=medium (63-70%)

dwellings

3=low (O-63%)

201

 

 

_ Col. Variable value
13 White- 1=high (43-100%)
collar 2=medium (39-42%)
occupied 3=low (O-38%)
14 Median 1=high ($6000.+)
income 2=medium ($5000.-$5999.)
3=low (0-$4999.)
15 Competi- 1=high plurality (61-100%)
tion 2=medium plurality (55-60%)
3=low plurality (0-54%)
16 Presi- 1=high (SS-100%)
dent's 2=medium (46-54%)
plurality 3=low (O-4S%)
17 Leadership l=speaker or president pro tem
position 2=majority leader
3=minority leader
4=majority whip
5=minority whip
18-25 Committee l=Aeronautica1 and Space Sciences
assignment 2=Agriculture and Forestry
(four two- 3=Appropriations
column 4=Armed Services
fields) 5=Banking and Currency

6=Commerce

7=District of Columbia
8=Finance

9=Foreign Relations
10=Government Operations
ll=Interior and Insular Affairs
12=Judiciary

l3=Labor and Public Welfare
14=Post Office and Civil Service
15=Public Works

16=Rules and Administration

 

202

Coding Motions

 

 

position

card cOl. Variable Value
,no. no.

1 1 Party 1=Democratic
2=Republican
3=Bipartisan
4=no party

1 2-25 State (seven state code=for
two column empty=no effect
fields)

1 16-25 Region (five regional code=for
two-column fields regional code+10=against
corresponding empty=no effect
to regional
codes 1-5)

1 26-34 Constituency code=for
(nine one code+5=against
column fields) empty=no effect

1 35-40 Committee committee code=reporting
(three two bill
column fields)

1 41-79 Minority identification number=
fiifiiitlfg corresponding legislator
nine three a minority reporter
column fields)

2 1-45 Sponsors identification number=

corresponding legislator
a sponsor of motion
2 50 President's 1=for

2=against