You are on page 1of 34

Institutional Arrangements and Equilibrium in Multidimensional Voting Models

Author(s): Kenneth A. Shepsle


Source: American Journal of Political Science, Vol. 23, No. 1 (Feb., 1979), pp. 27-59
Published by: Midwest Political Science Association
Stable URL: http://www.jstor.org/stable/2110770
Accessed: 05-06-2015 13:42 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

Midwest Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to American Journal
of Political Science.

http://www.jstor.org

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
Institutional
Arrangements
and Equilibrium
In Multidimensional
VotingModels*

KennethA. Shepsle,WashingtonUniversity

Nearlythirtyyearsof researchon social choice has produceda large body of


theoretical results.The underlying
structureof the modelsthathave generatedthese
resultsis highlyatomisticand institutionally
sparse.Whileattention has been devoted
to the mechanismsby whichindividualrevealedpreferencesare aggregatedinto a
social choice, rarelyare otheraspectsof institutionalarrangements treatedendoge-
nously.In thispaperinstitutionalproperties
are givenmoreprominence. In particular,
I focuson threeaspectsof organization:(1) a division-of-labor arrangement called
a committeesystem;(2) a specialization-of-labor systemcalled a jurisdictional ar-
rangement; and (3) a monitoring mechanismby whicha parentorganizationcon-
strainsthe autonomyof its subunitscalled an amendmentcontrolrule. The concep-
tual languagehas a legislative
flavorbut,in fact,the conceptsare broadlyapplicable
to diverseorganizational forms.The principalthrustof thispaper is a demonstration
of the waysinstitutional arrangementsmay conspirewiththe preferences of individ-
uals to producestructure-inducedequilibrium.
Theoriesof social choice are concernedwiththeoperatingcharacteris-
ticsand equilibriumpropertiesof collectivedecision-making arrangements.
At a very general level, as exemplifiedin the famous work of Arrow
(1963), we have failed to understandhow collectivedecision-making ar-
rangementsoperate because they lack equilibriumproperties.Indeed, a
fundamentallesson of his inquiryis that an institutionalarrangement
lacking some (perhaps distasteful)constitutionalrestrictionsor failing
that,a basic value consensus,is inherentlyinexplicablein its operation.
In light of Arrow's PossibilityTheorem, one major directionof

*The authoris Professorof PoliticalScience and ResearchAssociate of the


Centerforthe Studyof AmericanBusiness,Washington Earlierversions
University.
of this paper were presentedat the annual meetingsof the AmericanEconomic
Association,New York, December,1977 and the Public Choice Society,New Or-
leans,March, 1978. The authoris gratefulto the CarthageFoundationfor sponsor-
JohnAldrichof MichiganStateUniversity,
ing thisresearchand to Professors John
Ferejohnand MorrisFiorinaof CaliforniaInstituteof Technology,Howard Rosen-
thalof Carnegie-Mellon and RobertParks,Trout Rader,Barbara Salert,
University,
and BarryWeingastof WashingtonUniversityfor useful conversationsand com-
ments.

AmericanJournalof Political Science, Vol. 23, No. 1, February 1979


X) 1979 by the Universityof Texas Press 0026-3397/79/010027-33$02.55

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
28 KennethA. Shepsle

scholarship has soughtto isolatethe characteristics of thatvalue con-


sensus.Represented mostelegantly by the theoremof Sen (1966) and
broadlysummarized in Sen (1970), Pattanaik(1971), and Fishburn
(1973), thisliterature focuseson theproperties of a collectionof indi-
vidualpreference orderingsthatinduceequilibrium statesin an aggregation
mechanism. It turnsout,however, thatin at leastone interesting choice
environment-that in whichthe alternatives are a compactsubsetof a
multidimensional Euclideanspace-the restrictions on preferences are
extraordinarily severe (Kramer,1973). Related researchon majority
ruleand otherbinarydecisionrulesin multidimensional alternativespaces
has produceda similarconclusion(Plott,1967; Sloss, 1973; McKelvey
and Wendell,1976; Slutsky,1977b; Matthews,1977). In all of these
works,in orderforequilibrium to obtain,preferences mustexhibitan
extremely precisesymmetry whichis unlikelyto emergein naturalde-
cision-making settings and, even if it should,is extremely sensitiveto
slightperturbations(McKelvey,1977; Cohen,1977). It nowappearsthat
furthercharacterizationsofthevalueconsensusrequired forchoiceequili-
briumwill,in essence,consistof "mopping up" exercises.Equilibrium of
a social choiceprocessdependson the configuration of preferences but
can onlyrarelybe guaranteed bythatconfiguration. We mustsearchelse-
whereto understand howcollective choicesare arrivedat.
In recognitionof thistheoreticalcul de sac, somerecentefforts have
been made to model arrangements
institutional more directly.As Zeck-
hauserand Weinstein (1974, p. 664) note,
appearanceof [disequilibrium
the relativelyinfrequent in the formof] cyclical
majoritiesin functioninglegislaturescannot be explainedby some geometric
property of individuals' preferences . . . . [Miatters such as procedural rules
providekey insightsinto real-worldsocial choice
constraints
and institutional
phenomena.
Two suchefforts, thoseof Slutsky(1975, 1977a) and Denzauand Parks
(1973), attempt to providethestructurefora politicaleconomyin which
tax pricesand levelsforpublicgoods,and after-tax income,prices,and
forprivategoodsare all in equilibrium
quantities (in thesensethatpublic
sectorlevelsare majoritywinners, thepublicbudgetis balanced,private
goods marketsare cleared, maximized,
profits and no citizenbankrupted).
To circumvent the surrounding
difficulties public sectorvotingequilibrium,
theyresortto restrictionson the domain of public sectorchoices.Space
does notpermitme to review theirwork more fully here,so letme only
note thattheirrespective generalequilibrium orientationstendto sub-
mergematters ofirstitutional in
structure thepublicsector, mattersI wish

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 29

to make moreprominent in thispaper.Therefore, in whatfollows,I


focusexclusivelyon theinternalstructure decision-
of a genericcollective
makinginstitution. looks verymuchlike a legisla-
Whilethisinstitution
ture,I shallarguebelowthatit sharesmanyinteresting in
characteristics
commonwithotherkindsof decision-making arrangements. The basic
structure-in
pointof thispaperis thatinstitutional theformof rulesof
andamendment
jurisdiction control-hasan importantindependent impact
on theexistenceofequilibrium withthedistribution
and,together ofprefer-
ences,co-determines of the equilibrium
the characteristics state(s) of
collectivechoice processes.

1. Motivation, Examples
Notation,
composedof n
institution
We considera genericdecision-making
members,N = {1, 2, .. . , n} and a binarychoice procedureC(x, y)
thatdetermines The space of alter-
choicesbetweenpairsof alternatives.
convexsubsetRn ofm-dimensional
nativesis a compact, Euclideanspace.
Each ieNhas a complete, binary
transitive, relation,
preference ?i, defined
on all x, y e Rmn,and represented by an ordinalutilityfunctionui:
Rm -) R whichis maximizedon Rmat xi (= , . . .,
Winners
A pointx e Rmis said to be a globalbinarywinnerif and onlyif
x e n C (x, y). Moregenerally, winner
x is said to be an A- restricted
y 6 Rm
ifand onlyifx e n C (x, y), forsomeA C Rm.Clearlyglobalwinners
yeA
winners
areA- restricted underthemostdemanding domainrequirement-
A = Rn. Shortly we givea substantive interpretationto A, relatingit ex-
to institutional
plicitly arrangements. First,however, we definetheinsti-
choicefunction,
tution's C (x,y).
Throughout thispaperwefocuson decisivemajority though
coalitions,
notnecessarilyof the puremajority ruleform. With some changes it may
be possibleto generalizeresultsto other of
collections decisive coalitions.
For BCN let IBI represent thenumber of i e B and let Ix>ylI={ilx>y}l.
Thefollowing A- restrictedwinners aredefined.'

are adaptedfromMcKelveyand Wendell(1976).


IThese definitions

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
30 KennethA. Shepsle

DEFINITION:

StrongMajorityCondorcel: E, -xEA IVyEA,y+x,


(Weak) (E2) nx>y>

(2)
StrongPluralityCondorcet: E83 xEAIVyEA,y=/x, Ix>y
(Weak) (E4) > lY >xI}
(2)
StrongMajorityCore E5 {xeA VyEA, ly >x| < }
(Weak) (E6) (?)
It is clear that a strong winner of each type is also a weak winner of that
that type E1- E2, E3 - E4, Ef -* E6. Moreover, E1 E} > Eo >

(E2 -* E4 - E6). For A = Rm, McKelvey and Wendell (1976, Theorem

1.1) demonstrate that whenever I is strictly convex-x # y and y t x


imply [t x + (1-t)y] >j x for any t E (0, 1)-all strong (weak) winner
types are equivalent:

Es E= E3 E5
Ew E2 E4 E6

More generally, for A-restricted winners we have


convexand A a convex
THEOREM 1.1: It >i is strictly
set,thenE1 = E3 =E5 (E2 E4 -E).
PROOF: We already have E1 -- E3 -+ E, from the definitions, so demon-
strating that E5 -) E1 establishes the result. Suppose the contrary-x e E,
but x 4 E1. x f E1 implies there exists a y eA with [x > yI < B. Thus,

n
Iy ? xi 2 Since A is convex, pick a y' 6 A with y' = tx + (1-t)y,
2
t e (0,1). From the strict convexity of >*, y ?j x implies y' >i x. Thus

Iy' > xi >


n
and x *
E5, a contradiction. The proof for weak winners

is developed in a similar fashion. Q.E.D.


Notice that the equivalence of the strong (weak) winners need not obtain
if A is not convex, for then it would only be by accident that a point y'
e A with the appropriate properties exists.2
2Notice also for choice processesin which simplemajoritycoalitionsare not
decisivethatstrong(weak) winnersare notequivalent.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 31

Restrictions
Jurisdictional
The focusof thispaper (and the maindeparture fromtraditional
socialchoicetheory)is on theconsequences ofconstraints on contests be-
tweencompeting proposals.Restrictions on contests derivefromtherules
by whichthe decision-making agendais constructed and are oftenmost
conspiciousin the structural arrangements into whichdecision-making
groupsconstitute themselves. Particularlynoteworthy in thisrespectare
themechanisms ofdecentralization thatare employed to expeditecomplex
decisionmaking.Examplesabound:a committee systemin a legislature;
a collection ofschools,colleges,and departments in a university; a system
of divisions in a firm;an arrangement of bureausin an agency;the"sep-
arationof powers"withina nationalgovernment; a federalorganization
forlayersofgovernment; andso on.
Whatdistinguishes thesemechanisms of decentralization is thatthey
are division-of-labor instruments.The different committees of a legislature
or departments of a universityhavedifferent (thoughnotnecessarily dis-
joint) domainsof responsibility or jurisdictions.Of course,the idea of
jurisdiction is quiteindependent of division-of-labor structural arrange-
ments.The formalagendaof an ordinary businessmeeting, withits sep-
aratecategories of activity-oldbusiness,newbusiness,officers' reports,
etc.,suggests a separation intojurisdictions
of activities withouta struc-
turaldivision-of-labor. Whilea matter ofnewbusinessmaynotbe brought
up (i.e. is outoforder)duringthesessionon old businessandviceversa,
theentiremembership of theorganization participates in bothdelibera-
tions.
In orderto keep theideas of jurisdiction and division-of-labor dis-
tinct,wedefine twofinitecoverings.3
DEFINITION: ofsetsC = {Cj) a
Call thefamily
committeesystemif it covers N = (1,2, . . . , n}
Example1.1 (Committee-of-the-Whole): Let thefamilyof setscon-
sistofa singlesetC = (N). Thisis knownas thecommittee-of-the-whole
socialchoicetheory,
and,in traditional is themainstructural arrangement,
usuallylabeledcommittee, etc.It is a trivialpartition
society,electorate,
ofN.

3 A finite
covering ofa setB is a finite
collection( ={#j, whereU3Pj= B. If,
in addition,
Pi n /f _= 0 forall Pi, P, e (, thenP is a partition
ofB. Technically,
U1P1 D B is all thatis requiredforP to coverB, butthistechnically is of no con-
sequenceinwhatfollows.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
32 A. Shepsie
Kenneth

Example1.2 (Legislativeparties):Let the familyof setsC =(C)


be thepartygroupings The family
of a legislature. partiesis
of legislative
normallya partition withone and
inasmuchas each legislatoraffiliates
party.
onlyonelegislative
Example1.3 (Committee systemof the U.S. House of Representa-
tives):Let thefamilyC {Cj} represent thecollectionof standingcom-
mittees Here C is a coveringbut
of theU.S. House of Representatives.
nota partition sincemosti e N serveon morethanone standing com-
mittee.
A committee system, is a division-of-
as theseexamplesillustrate,
laborarrangement members
thatdistributes decision-making
to structural
subunitsof theorganization.The notionof is
jurisdiction next.
defined
Let E = {el, . . ., em) be an orthogonalbasis for Rm,
DEFINITION:
whereei is theunitvectorfortheith dimension. Call thefamilyof
setsJ = {Jk) a jurisdictionalarrangement if it coversE.
Thuseach jurisdictionJk e J is a subspaceof Rm.Threeexamplesin R3
the rangeof alternatives.
will illustrate
Example1.4 (Global Jurisdiction):Let thefamilyof setsconsistof
R3, fallsentirely
a singlesetJ ({{e,e,,e,J).The space of alternatives,
Together
withina singlejurisdiction. withExample1.1, thisarrangement
thecontext
characterizes in theliterature.
ofsocialchoicethatis prevalent
Example 1.5 (Simple Jurisdictions):Let = {J1,J2,J3)with Jk
{ek). Hereeach jurisdiction of R3.
is simplya singledimension
Let J = {J1,J2)withJ1
Jurisdictions):
Example1.6 (Overlaping
{e1,e2} J2 ={e2,e,}. This jurisdictionalarrangementis descriptiveof
arrangements in the U.S. House and U.S. Senate,wherejurisdictional
overlapsof committees are considerable.
In the remainder of thispaper,attention is devotedto the case of
simplejurisdictionswithcomparisons made betweencommittee-of-the-
wholedecisionarrangements and arbitrary committee systems.The ex-
istenceof equilibrium will be established in each case withdifferences
betweentheequilibrianoted.
For an arbitrarycommittee system C and jurisdictional arrangement
J, thecorrespondence f : C -- I associates(sets of) with
jurisdiction(s)
theCj e C. Severalpoints are in order. First,each Cj e C is mapped into
at least one Jk J. It then
6 has jurisdictionover the dimensionsin Jk.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 33

Thus if some subsetof the i e N comprisesC3 and f takesC, to k=


(e,,e2) thenC1 has jurisdictionover the pointsin Rm spanned by e, and
e2; theyare of theform(x1,x2,x 0, ... 9 X0m) wherethex?i are compo-
nentsof somepredetermined statusquo (see below). Second,the cor-
respondence permits a C1 e C to be associatedwithseveralJk's. One of
specialinterest involving
to us is thesituation thecommittee-of-the-whole
and simplejurisdictions. ThereC1 N and J {(e,), (e2),..., (em)).
The pointsin Ci's jurisdiction are of the form(x1,x02,. . , x0m) or
(x01,x2,x,0 .... , x0m) or ... or (X01,X02,. .. ., xm,L, xm)points that
alterthestatusquo in at mostone dimension. Contrastthisarrangement
to the one consideredin traditional social choicetheoryin whichJ=
{{el,e2, . .. , em)): there,all pointsin Rmfall in Ci's jurisdiction.Third,
it is possiblefora particular es e E to be containedonlyin Jk forwhich
fl(k) is undefined.That is, some of thechoicespace maybe
dimensions
contained associatedwithno C, e C. Shortof revolution
in jurisdictions
or more peacefulconstitutional change,the statusquo level on this
dimension is immutable.

Agenda
The StatusQuo and the Organizational
The prevailing socialstateis a pointxo e Rm.The statusquo repre-
sentsthecumulation decisionsthathas brought
of historical theorganiza-
thecurrent
tionto xo; it characterizes levelon each of them dimensions
of the choice space. The organizational agenda for changesin xo is
controlledby theCj e C and is channeledby jurisdictionalarrangements.
The set of feasiblechangesin xo is theset of pointsthatalterxo in no
morethanone jurisdiction:
DEFINITION:A proposalis a changein xo restricted to a single
Thesetofproposalsis P = {xlx = x? + X Aset,1 5 Jk
jurisdiction.
iel
forsomeJk J} C Rm.
Committees may recommend changesin xo to the parentorganization
fromamongtheproposalsfallingwithin Com-
theirassignedjurisdictions.
in thisview,are instruments
mittees, thatgenerateproposedchanges.
by theirrespective
Theiractionsare constrained and their
jurisdictions
own internal rules.
decision-making
Control
Amendment
f(C) as thedimensions
We havedefined thejurisdiction
constituting

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
34 A. Shepsle
Kenneth

theset of proposalsavailableto C, when


of C, e C. Let g(Cj) represent
the statusquo is x0:
g(Ci) = (xlx xo + YqAej,ei e f(Cj))
SupposenowthatC, proposessomex e g (Cj) totheparentbody.4
DEFINITION: For anyproposalx E g (Cj), theset M(x) C Rn con-
sistsof themodificationsN maymakein x. M(x) is said to be an
amendment controlrule.
The idea hereis the following:C, proposes some x E g (C1) to the parent
body;beforex is comparedto xo it maybe modified by N; eitherx or
somex' E M(x) is thencomparedto xo.
An amendment the extentto whichan or-
controlrule represents
can monitor
ganization and changetheproposalsof itssubunits.In some
a parentorganization
instances, is littlemorethana holdingcompany
votingthe latter'srecommendations
forits subunits, "up" or "down":
DEFINITION: A committee's proposalis said to be governedby the
closed rule if M(x) = 0 forall x e g(CQ).
Example1.7 (Romerand Rosenthal):Romerand Rosenthalcon-
structa model in whicha group,called the agenda setter,proposes
an alternative-achangein thestatusquo-which is eitheracceptedor
or rejectedby thecollectivity.Theyhave in minddecisionsettings illus-
tratedby schooltax choicesin manyschooldistricts. Therethe agenda
setteris usuallythe schoolboard whichproposesa property tax rate.
By referendum thecitizensof theschooldistrictdecidewhether to adopt
the recommendation of the schoolboard; if theydo not,thenthe rate
to whichthe tax revertsis usuallyconstitutionally-specified.5
Example1.8 (Tax Legislationin the U.S. House): Thoughby no
meansa hardand fastrule,it has been the usual practicein the U.S.
proposalsoftheWaysand MeansCommittee
Houseto providelegislative
witha closed rule.No modifications,unlessacceptableto a committee
are permitted.
majority,
The closedrule is obviouslythe mostrestrictive
constraint
on the

4For themoment we deferconsideration of theinternal


procedure by which
Cj arrivesatx as itsproposal.
5 It maybe theprevailing rateor someotherpredetermined rate.In Oregon,
Romerand Rosenthal report,if a schooltax rateproposalfails,thenit reverts
to
somelow tax ratethatgenerates onlythe levelof revenuemandated by earlier
statute.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 35

discretion
of theparentorganization
to modifycommittee
proposals.At
theotherextremethereis the open rule:
DEFINITION: A committee's
proposalis said to be governedby the
openruleifM(x) = Rmforall x e g (Cj).
Example1.9 (U.S. SenateRiders):The rulesof procedurein the
U.S. Senatepermitthemodification of proposalsthrough theadditionof
riders.As a consequence, a majorpieceof legislation,
e.g.,theextension
of theVotingRightsAct,maybe taggedon as a riderto a fairlyinnoc-
uousbill,e.g.,a Districtof ColumbiaPublicLibraryauthorization.
Betweenthe open and the closed rule lie alternative amendment
controlruleswhichmaybe partially orderedby set inclusion.One that
deservesspecialdistinctionis thatof germaneness:
DEFINITION: A committee'sproposalis said to be governedby a
jurisdictional
germanenessruleifM(x) = {x'lx'i = xi' ifet4f(Cj) }.
The jurisdictional
germanenessrule permitsamendments of x e g(Cj)
onlyalongthosedimensions thatfallin thejurisdiction
of thecommittee
proposingx. A slightly
different
ruleof relevancy is thefollowing.
DEFINITION: A committee's proposalis said to be governedby a
proposalgermaneness ruleifM(x) = {x'lxi' = x?0ifxi x?}.
If jurisdictionsare simple(example1.5), thenproposalgermaneness and
jurisdictional germaneness are equal; otherwise proposalgermaneness is
a. propersubsetof jurisdictional germaneness. For example,if f(CJ)=
{ei, e2} and Cj proposesx = (X,x2,X03,9. . ., x0.)-i.e., onlya change
in thefirstof Cj's twodimensions - thenjurisdictional germanenessen-
tailsM(x) = {x'/x' = x? + X1e,+ X2e2},whereasproposal germaneness
requiresM(x) = {x'/x' = xo + Xle1}.In effect, germaneness rulesare
openrulesrestricted eitherto thedimensions or to the
of Cj'sjurisdiction
dimensions on whichCj's proposalaltersx?.

andPreference-Induced
2. Structure-Induced Equilibrium
somesalientaspectsof
We have,to thispoint,soughtto characterize
namelystructural
organization, and jurisdictionaldecentralization.
To
summarizebriefly, choicesituationis describedby
an institutional
(1) a statusquo, x? e Rm;
(2) a coveringC = {C} on theset of institutional
actors;

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
36 Kenneth
A. Shepsle

(3) a coveringJ= (k) on the choicespace RI, describing the


constraints
jurisdictional on challengesto x?;
(4) a correspondence, f, mappingelementsof C to (subsetsof)
elementsof J; and
(5) an amendment controlrule,M(x), definedforx e g (C), de-
againstwhichN may compareany
scribingthe alternatives
proposalmadeby a Cj e C.
In thissectiona modelof organizationpossessing is
thesecharacteristics
offeredand its states
equilibrium defined.
DEFINITION: A proposalx E Rn is called a forx? if
replacement
(i) x e g (C3)forsomeCj e C
(ii) x = Cj (x,x?)
(iii) x e C (x,x')forall x' e M(x),and
(iv) x= C (x,x) (2.1)
In orderto replacexo, a proposalmust,first, be feasiblein thesenseof
(i). It is in thissensethatjurisdictionalarrangements channel(constrain)
change.LettingCJ(-,-) be a binarychoicefunction forCj, (ii) requires
thatproposalx e g (C) be preferred to x? by a decisivecoalitionin Ci.
Thus, not onlydo jurisdictional arrangements channelproposals;they
empowervetogroupsas well. (iii) requiresthatx surivivewhencom-
paredto all admissible alterationsx' e M(x); thatis,x mustbe a maximal
element of (x) UM (x). Finally,by (iv), x mustbe preferred to xo by a
decisivesubsetofN.6
Thissuggests thefollowing equilibrium notions:
DEFINITION: The statusquo x? is vulnerable ifthereexistsa replace-
mentforit or if thereexistsan x e g (C3) n ci (x,xO) and an
x' e M (x) withx' e C (x',x) n c(x,xo).
Thatis, xo is vulnerable if thereis a pointthatdominates it in thesense
of (i)-(iv) aboveor ifthereis a proposalof someCj an amendedver-
sionofwhichdominates it.
DEFINITION: inducedequilibrium
The statusquo x? is a structure if
andonlyifitis invulnerable.
the assumptionof sincerebehavioris maintained.When called to
6 Throughout,
make a choice,each i e N consults>i and chooses in a mannerconsistent withit.
By immediateimplication,if x? is not a maximalelementof g(Cj) forsomeCj, and
if the maximalset of g(C1) is nonempty, thena proposalby Cj is forthcoming.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
ANDEQUILIBRIUM
ARRANGEMENTS
INSTITUTIONAL 37

By a somewhat circuitousroute,itmaybe shownthata structure-induced


equilibriumis a particular winnerdefined
formof A-restricted in section
to
instructive
1. It is not particularly dwell on thesedetailsexceptto
note the contrastbetweena structure-induced and a global
equilibrium
binary winner:

equilibrium
DEFINITION:The statusquo x? is a preference-induced
ifand onlyifit is a globalbinarywinner-x?E fl C (x,x?)
x e Rm
thenit is a
equilibrium
THEOREM2.1: If x? is a preference-induced
equilibrium,
structure-induced butnottheconverse.
Proof:xo a preference-induced -+ xo e n
equilibrium C(x,xo)-
xeRm
x? e C(x,x?) for any x e C, (x,x?) and x? e C(x',x?) for any x' e
M(x) forany x e C1(x,x0) -+ xo invulnerable-+ x? a structure-induced
equilibrium.The conversefailsin generalsincetherecan be a structure-
inducedequilibrium x? and anotherpointx forwhichx = C(x,x?) but
x? e g (Cj) forno C,. Q.E.D.

EXAMPLE2.1: In R3,supposeN = (1, 2, 3) withutility maxima,re-


at (1, 0, 0), (0, 1, 0), (0, 0, 1) andspherical
spectively, loci.
indifference
The Paretooptimalsurfaceis the triangleconnecting thesepoints(see
Figure1). It is easyto showthat,forC( , *) thesimplemajority rule,
no preference-induced equilibriumexists.Consider now x= (0, 0, 0),
a jurisdictional
covering J= {{el), {e2}, {e3)}, thecommittee-of-the-whole
-C = {N}, and a germanenessrule to amendments.The point xo is a
structure-inducedequilibriumsince all proposals (and amendments)are
of the formxo + Xieiand, for any such proposal with Xi =- 0, at least
two individualspreferx?. That is, xo is preferredby a majorityto all
there
proposalsof the form(x, 0, 0), (0, x, 0) or (0, 0, x). Interestingly
existpointson theshadedtriangle unanimouslypreferred to x?, butjuris-
dictionalconstraints and amendment rulesprohibittheirconsideration.
If, insteadof (0, 0, 1), 3's ideal pointwere(t, 1-t,0), t e [0, 1], then
thatpoint,a convex combinationof the ideal points of 1 and 2, would
equilibriumand, by Theorem2.1, a structure-in-
be a preference-induced
ducedequilibriumas well.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
38 Kenneth
A. Shepsie

FIGURE 1

e3

/(0,1,0()0,1
1?Q? ) (IX el

/2

e2

andGermaneness:
3. SimpleJurisdictions forEquilibrium
Conditions
schemeJ = { {ej}, {ej}, . * ,
The mainfocusis on a jurisdictional
and the germaneness
jurisdictions
{em} } of single-dimensional rule de-
finedin section1. The particularcommitteesystem remainsimplicit for
now.The following and lemmaare
definitions usefulforthe main results
ofthissection.
DEFINITION: For statusquo x? and jurisdiction ej, let theinduced
idealpointin thejh directionfori e N be x*i = (X*i, . . . * ,

where x,,*"= Xk0, k p j, andui (x*i) max[ui(xl?, . X. ,


X?m)]. Moregenerally, idealpointonan arbitrary
x*' is theinduced
on X at x - x*.
setX ifus(x) is maximized
DEFINITION: A utilityfunction quasi-concaveif
ui (x) is strictly
and onlyif,forx = y, ui (x) 2 us (y) -e ui (x') > ui (y) for
anyx' =Xx+ (1-X)y,X A (O, 1).

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL
ARRANGEMENTS
ANDEQUILIBRIUM 39

DEFINITION:Let X = {xlx = Xy + (1-X)z, y, z e R1m, X e [0,1]}


be the line connectingarbitrarypointsy and z. A preferencerepre-
sentationon X is said to be single-peakedifand onlyif,forall x e X,
x + x*', ut [a x + (1-a) x*i > ui [B x + (1-f3) x*'] whenever
0 < a < IB< 1 and x*i is the induced ideal point on X.
LEMMA3.1: If ui: Rm-e R is strictlyquasi-concaveand continuous
for i E N, then the preferencerepresentation
for i on any line X is
single-peaked.

Proof (see Figure 2): For any line, its intersectionwith Rm (a compact,
convex set) is itselfcompact and convex. Since ui is continuousit has a
maximumon thisset (see Nikaido, 1972, Theorem 1.1). The uniqueness
of the maximumfollowsfromthe strictquasi-concavityof ui. Hence an
induced ideal point,x*i, existsand is unique. Let x" = /x + (1-f/)x*i
for arbitraryx e X. By construction,ui (x*i) > ui (x") and, by strict
quasi-concavity,ut (x') > ut (x") forx' -Xx* + (1-X)x", X e (0,1).
From the definition of x", it followsthatx' - x*i + (1-X) [fx + (1-f3)
x*'] - (1-X) /3x+ [A + (1-X) (1/l3)]x*i - a x + (1-a)x*i where
a = (1-X) fl < fl.From the above definition, us is single-peaked. Q.E.D.

THEOREM3.1: LetXi* = {Xj*l, XJ*2, ... , X)*n} be theset


of jthcomponentsfromtheinducedideal pointsof thei e N
in thedirectionej fromstatusquo x?. For one-dimensional
a germanenessruleforamendments,
jurisdictions,
and anycommittee system,x? is a structure-induced
equilibriumif,forall j,
x= medianX*. (3.1)

Proof: From Lemma 3.1, the ut are single-peakedon any line of the form
X = {x/x = x? + kei}. From Black's well-knowntheoremon single-
peaked preferencerepresentations, if xjo = median X,* then xo defeats
all points in X. If condition(3.1) holds for all j, then,given the juris-
dictional constraintand the germanenessrule for amendments,xo is
invulnerable.From the earlierdefinition, it is a structure-induced
equilib-
rium. Q.E.D.

for xo to
The theoremestablishesthat condition (3.1) is sufficient

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
40 Kenneth
A. Shepsle

FIGURE 2
y
x

X*i

x'=ax+ (1-a)x*i

Ox" 3x +(I -)3) x*


x

be an equilibriumwhenthejurisdictions
are theindividual basis vectors
and a germaneness rulegovernsamendments. It is not a necessarycon-
dition,as thefollowing
corollary
suggests
and Theorem3.2 proves.
COROLLARY 3.1:(3.1) is necessaryconditionforx? to be a struc-
ture-induced
equilibrium underthecommittee-of-the-whole arrange-
ment.
The proofis omitted.The corollarysuggeststhatnot onlymusta pre-
to x? existforthe latterto be vulnerable;
ferredalternative it mustbe
proposedby a committee withappropriate or "reached,"via
jurisdiction
theamendment process,froman appropriatecommittee proposal.Under
the committee-of-the-wholeprocedure,x? is a structure-inducedequili-
briumonlyif (3.1) holds,sincex?; #&medianXj* impliesthereexistsa
y = x? + X ej withy = medianXi* and anymember ofat leasta simple
majorityof N eligibleto proposeit-i.e. the "committee" (N, itself)
will approvey = (xl?, . . . ,y., . . ., X.') which,in turn,is passed over
x' by theparentorganization
(again N).

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 41

The fullimplicationsof the veto-power of committees,and conse-


quentlytheircapacityto createequilibria,are providedin Theorem3.2.
Let
SI = {x x = x? + Xej,X #0O,x = C, (x,x?))
be thesetofmodifications of x? in itsjurisdiction
preferredto thestatus
quo by Cf.Fromthestrictquasi-concavity of theus,i e C,, and Lemma
3.1, it is straightforward
to showthatSI is an openinterval(possiblyde-
generate)forC, ) thesimplemajority ruleforcommittee j. Similar-
ly,let
Tj {xx = x? + Xel,X # O,x = C (x,x?)}

be thesetofmodifications
in x? alongespreferred
byN. Finally,define
M(x) U {x} ifxES,
Rj (x)-
0 ifSj =0
Nowwemaystate
THEOREM 3.2: For the jurisdictional
arrangement consistingof the
basis vectorsof Rm and a germaneness rule for amendments, x?
is a structure-induced
equilibriumifand onlyif,foreveryj and every
X ESj,

Rj (x) n T=o0.
PROOF: (1) Necessity. Supposex' E R, (x) n T,. SinceRj (x) is
theunionof twosets (it cannotbe emptyunderthesupposition)
thereare twocasesto consider:(i) x' = x and (ii) x' E M (x).
We establishnecessity for(i); it also holdsfor(ii) but the
proofis omitted here.Thus,ifx' x, thenthereexistsan
xefsl n Tj.xE Sj -x Eg (Cj),x- C, (x,x0).xET e-x = C (x,x?).
Therefore, accordingto (2.1), eitherx is a replacementforx? or there
is an x' e M(x) withx' e C (x', x) n C (x', x?), i.e.,x' defeatsbothx
and x?. In eithercase,x? is vulnerable. It is, therefore,
nota structure-
inducedequilibrium and necessity is established.
(2) Sufficiency. Supposex? is not a structure-induced equilibrium.
Thenx? is vulnerable. Theneitherthereexistsa C, and an x E g (C,)
satisfying (2.1)-that is, a replacement forx?--or thereexistsa
C, and an x e g (C1) withan "amendedversion"x' e M (x) suchthat
x c(x, x') n C(x', x?). In eithercase, thereis a C, and an x e S1

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
42 KennethA. Shepsle

forwhichRj (x) n T, 7 0. Sufficiencyis established. Q.E.D.


It shouldbe notedthatTheorem3.1 is, in fact,a directconsequence
of Theorem3.2, so thatthelattermaybe regarded of
as a generalization
the formerto any C - (Cj). In particular,if x0j = medianXi* forall j,
thenTj = 0 and thecondition of Theorem3.2 is satisfied,independent
of thecommittee or of therecommendations
structure of anyCj E C. In
Theorem3.1 identifies
effect, x? as an equilibrium whenit is a "partial
median"in the senseof Hoyerand Mayer(1974) .7 In theirTheorem
theyprovethata partialmedian,
2, however, intheabsenceofjurisdictional
is an equilibrium
on comparisons,
restrictions onlyifitis a totalmedian-
the medianin everydirection, not just in the directions of the basis
vectors.And,in orderfora totalmedianto exist,a strongsymmetry of
preferences-aconditionformulated elegantlybySlutsky(1977b)-is re-
on comparisons
restrictions
quired.Jurisdictional obviatethisnecessity-
exhibitsymmetry
ifpreferences within (and theyalwayswill
jurisdictions
withundimensional thenx? is an equilibrium.
jurisdictions), This is pre-
ciselywhatTheorem3.1 tellsus.
Further ofTheorem3.2 areinstructive:
consequences
COROLLARY 3.2: xo is a structure-inducedequilibrium
wheneverSji 0 forall j. In thiscase the Cj E C
are vetogroups.
COROLLARY 3.3: If M(x) entails only germanenessand
Sj =A 0, thenxo is a structure-inducedequilibrium
ifand onlyifxo = medianXj*.
Corollary3.2 indicatestherealimportof structural
arrangements. If,for
no X $& 0, thereis an x - xo + Xej which a majorityof C, prefersto
x?-that is, if xoj: median{xj*i)}then xo is invulnerable in the jth
i E Cj
importof C, "keepingthegatesclosed"in its
The normative
direction.
depends,in somesense,uponthe
agendarolefortheparentorganization
deviationbetweenmedianXj* and median{xj*i). Example3.1 suggests
i E Cj
biasesin thecommittee
thatifthereare distinct process,this
assignment
canbe considerable.
deviation

7Hoyer and Mayer definea partialmedianin termsof any basis for Rm. A
pointis a partialmedianif it is the medianalong each of them linesdefinedby the
basis vectors.The one-dimensional arrangements
jurisdictional are such a basis. Also
see Davis,Hinichand Ordeshook(1970).

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 43

FIGURE 3

. . ~~~~~~~~~~~el
Example3.1: In Figure3 therearesixvoterswithcircularindifference
contoursand idealpointsas indicated.
The pointx* is a total(and hence
a partial)median-foranyvectororiginating at x*, at mostthreevoters
prefera pointalongthatdirection overx*; x* is a weakequilibrium in
thesenseof thedefinitionin section1 and Theorem1.1. Supposenow
that thereis a committeesystemC - (C1, C2), with C1 - {1, 2, 3),
C2 - (4, 5, 6) and f(Cl) - el, f(C2) e2. The pointx? - (
- x25)
is a structure-induced
equilibrium (so is x*), sincebothcommittees will
"keepthegatesclosed"-S = S2 0. Also notethatxo does notlie in
theconvexhullof thesix ideal points-theParetoset.
In the same fashionthatTheorem3.2 generalizesTheorem3.1,
Corollary3.2 standsas a generalization of Corollary3.1. In particular,
theonlywayforSi = 0 to holdunderthecommittee-of-the-whole com-
mittee systemis when(3.1 ) is satisfied.
FromTheorem3.2,itis apparent thata structure-induced equilibrium
maysatisfy Rj (x) 0 for some j and Tj = 0 forotherj. Thatis,xj0 =
medianX,* forsomej andXko = median{Xk*i) forall remaining k. If each
i E Ck

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
44 A. Shepsie
Kenneth

C1e C hasa uniquely


defined andN hasa unique
medianinthejth direction,
medianX* foreach j, thenthereare as manyas 2mstructure-induced
equilibria(underthegermaneness
ruleand simplejurisdictions).
Example3.2: In Figure4, C1 = {1, 2, 3) has jurisdiction el and C2
-{4,5,6,7} has jurisdiction e2,and each i e N has circularindifference
contours and idealpointas illustrated.
PointC is thepointat whichTj =
0 for all j; pointB has theproperty Si = 0 forall i; pointsA and D
that
mix these properties-A has So T = 0 and D has S -.=
- 0.
Notethattwoof thefourequilibriaare notParetooptimaland thatnone
offourare,inanysense,"centrally" located.
Finally,if eithercommittee or parentorganization medianin any
jurisdiction is nonunique,therewillbe a densesetof equilibriaassociated
withthatjurisdiction; insteadof pointA,B,C,D,as in Figure4, therewill
be regions.
Example3.3: In Figure5, N {1,. . ., 8),C1 { 1 ,...,4}, C2
{5, ... , 8), each i e N has circularindifference contourswithidealpoint
as indicated, and thejurisdictionsof C, and C.,are e, ande.,,respectively.

FIGURE 4

e,2 Rx77

-"65 A \B

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 45

FIGURE 5

X
xz7
i
~~6

3?

el
The regions A,B,C,Dare thestructure-induced equilibriaforthisproblem
(theirdescriptions parallelthosein Example 3.2). In contrastto the
previousexample,"almostall" of the structure-induced equilibriaare
Paretooptimal.Indeed,a substantial proportion of the Paretosurface
consists
ofequilibria.
Corollary3.3 strengthens Theorem3.1 andgeneralizes Corollary 3.1.
It strengthens
Theorem3.1 in thesensethat,undera germaneness rulefor
amendments, unlessa Cj E C is at a maximalelement from
alongej distinct
medianX1* (in whichcase Sj = 0), thencondition(3.1) is necessary
as wellas sufficient.
It generalizes Corollary3.1 by providing thecondi-
tion,viz.Sj # 0, thatrenders(3.1) necessary forequilibrium forarbi-
trarycommittee systems(not just the committee-of-the wholearrange-
ment).
Corollary3.3, however,does strainthe assumption of nonsophisti-
catedbehaviorby committees. In particular,
it is entirely
imaginable that
a highlyunrepresentativecommittee's efforts
to movethestatusquo along
thelinex? + Xe1to thepointx' = x? + X'ejwillbackfire in thatits
proposalwillbe amendedunderthegermaneness ruleto x" = x? + X"el

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
46 Kenneth
A. Shepsle

and passedby N, wherex' > x? > x" (> is >j forall i e C,). There
Cj Cj Cj
are,then,occasionallystrongdisincentives
to proceedsincerelyin reveal-
ingthat,in fact,Sj #e0. Moreprecisely, whenever xo lies betweenthe
maximalelements8 of C, and N along xo + Xej, sincerebehaviorby
Cj is penalized.However,onlyunderratherstrongassumptions aboutin-
formation (knowing,forinstance,the distribution
of preferencesof the
parentbodyon the issueat hand) do theincentives fornonsincere be-
haviorbecomeunequivocal.Whilemuchis made in the Congressional
literature
ofthewily,sophisticated,
committee chairman,antennacarefully
attunedto the"will"of theHouse,e.g. WilburMillsof theHouse Ways
and MeansCommittee (see Manley,1969), therepresentativeness
of this
caricatureis insomedoubt.

4. SimpleJurisdictions
andGermaneness: ofEquilibrium
Existence
The results
oftheprevious sectionprovidesomeofthecharacteristics
ofstructure-induced equilibriawhentheyexist.In thissectionwe establish
theirexistence.Unlikepreference-induced equilibrium,
theirexistence does
notdependupona knife-edge assumption aboutthedistribution of pref-
erences.
In orderto illustratethe existenceresultit is usefulto distinguish
betweenxji and xj*i. The former is thejth component of i's ideal point;
it is independent of xo, the statusquo, and alwaysexists.The latteris
thesolution tothefollowing maximization problem:
Findx*i suchthatui (x*i) = Max [ui(xO1, x02, . . ., x0j-1, Xj, X0j+i,

* * *Xxom) (4.1)
That is, xj*i is the jth component of theinducedideal pointin the jth
jurisdictionwhenthelevelsforall otherdimensions are maintained as in
the statusquo. For manyof theexamplesof the last section,in which
individuals wereassumedto havecircular indifference
contours, xj*i=x.
This equality,called separability,holds undermoregeneralconditions,
and impliesthatpreference as betweentwopointsthatdiffer in at most
one jurisdiction is independent of the levelsof variablesin otherjuris-
diction.For the case of generaljurisdictions, separabilityis definedas
follows:
8 Along any line in Rm, the majorityrule choice functionis transitive,
owing
to the single-peakednessof the >,; since Rm is compact,the choice functionhas
a maximalelement.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 47

Let an arbitrary
DEFINITION: consistof (e1,
jurisdiction , es}.
definethe following
Furthermore, points:
X= (XI, . . . Xi, Z;+19 . . . * Zm)
y (Y *... yj, Z;+i, ... * Zm)
xI= (X, . . . , Xi, Z'+9 . .. ,
Znm)
Y= (yi, . yj' Zj+1, . . . Z' m).

A preference
representation if and
ui is separableby jurisdiction
only if us (x) 2 ui (y) -* us (x') 2 ui (y').

This conceptwill be utilizedin the corollaryto the existencetheorem


below.
The issueof existenceof a structure-induced reducesto
equilibrium
the followingquestion:Does a point exist thatis simultaneously in-
Theorem4.1 answersthisquestionin the
in all jurisdictions?
vulnerable
affirmative.
THEOREM of each i e N are
4.1 (Existence):If the preferences
representablebya strictly continuous
quasi-concave, utility
function,
if thebasis vectorsof Rmconstitutecommittee and if
jurisdictions,
a germaneness rulegovernstheamendment process,thenstructure-
inducedequilibriaexist.
We provethistheoremshortly. In orderto obtainsomeintuition about
thetheorem corollary:
and its proof,we firstestablishthefollowing
COROLLARY 4.1: If thepreferences of each i e N are representable
by a strictly
quasi-concave, continuous utility
functionseparableby
if the basis vectorsof Rmconstitute
jurisdiction, committeejuris-
and if a germaneness
dictions, rulegovernstheamendment process,
thena vectorof mediansa = (,1X Jh2 . . * tm), where kj =
median{(, xj2, . . . X,"}, existsand is a structure-induced
, equili-
brium.
Proof:The medianvector, existssince,byconstruction, V existsforeach
i e N (a strictlyquasi-concave, continuous function has a maximum on
a compactset) and themedianof a well-ordered setexists(though
finite
it neednotbe unique).FromTheorem3.1, p* = (,1*1 / 2*? . . . M *)
where,pj*= median{Xj*l, xj*2, . . , xj*n}, is a structure-induced
equilibri-
Y/=xj*i so that,u
um. But, by jurisdictionalseparability, ,*. Thus u
is a structure-inducedequilibrium and, as demonstrated above,,u exists.
Q.E.D.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
48 Kenneth
A. Shepsle

In effectthiscorollarystatesthatwhenpreference "types"are re-


strictedto thoserepresentable by a utility separable
function byjurisdiction
either
(essentially spherical indifferencecontoursor ellipsoids themajor
andminoraxesofwhichareparallelto thebasis vectors),then a structure-
inducedequilibrium exists.In thisinstance,jurisdictionalarrangements
notpreference createequilibrium.
distributions This standsin starkcon-
trastto thestrictrequirements on preferencesrequiredfora preference-
inducedequilibrium. The restrictionofpreference "types"in corollary 4.1
inherent in theseparability condition confusestheissuebecausethecon-
clusiongeneralizes to a muchbroaderclass of preference characteristics.
Thisis theimport ofTheorem4.1 whichwe nowprove.
The strategy for provingTheorem4.1 is firstindicated.For any
jurisdictionej themaximization problem(4.1) is solvedforeach i E N
forall combinations of levelsforotherjurisdictions. This tracesout the
Xj*isurfaceforeach i E N andej E E. The strictquasi-concavity and con-
tinuityoftheuj and thecompactness ofRmimplythat(4.1) has a unique
solution.Fromthesesurfaces, themedianin thejth jurisdiction X,M
median {Xj*l, . . ., X"*n} may be traced. The intersectionof the XjM
surfacesj = 1, m, is a structure-inducedequilibrium.
Example4.1: The procedure forthreeindividuals
is illustrated and
twodimensions in Figure6, parts(a) and (b). For variouslevelson the
e2 dimension,each i's solutionto themaximization problem(4.1), based
on hisindifferencecurvesin part(a), tracesout his most-preferred level
on the el dimension. These are drawnin part (b) as XI*', XI*2, and
X1*3.NoticethatMr. 3, whosepreferences a fixed
are separable,prefers
level on e1, no matterwhatthe e2 level. By a similarprocedure, for
variousel levels,themaximization problemyieldsa solutionin e2,tracing
outthecurvesX2*1, X2*2,X2*3.For variouslevelsof e2,medianXl*=
ieN
X1Mand,forvariouslevelsof el, medianX2*' = X2M. Theseare simply
iEN
segments oftheindividual curves.The pointx? e X1M n X2M is theunique
Fromthedefinition
pointof intersection. of theX,M curves,it is obvious
thatxo is a structure-induced
equilibrium-since it lies on bothmedian
curvesit is invulnerablein bothjurisdictions.Notice,in part (a), the
disparitybetweenx? and /,u(the vectorof mediansof theideal points).
Whenseparability does nothold (onlyMr. 3 has separablepreferences),
,u no longer has properties.
equilibrium
The proofof Theorem4.1, then,requiresthatwe establisha non-
emptyintersection fortheXjMsurfaces.In Figure6 thereis a singleton

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
a

-~~ ~~~~~~

X~~~~~~~~
00
_
?-~~~~~~~~~~
X
0 * .00

* :~:0I(%.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
50 A. Shepsle
Kenneth

thisfollowsfromthefactthat,in thatparticular
intersection; example,the
medianmappings X1M e2 -+ e, and X2M: e, -. e2 are single-valued.
However, XjMneednotbe single-valued (eventhough maxi-
theindividual
mizationsurfaces,X*i, are). XjM,thatis, is generallya correspondence,
in Figure7, a contingency
as illustrated thatmay occur,forexample,
whenthereare an even numberof voters.In thiscase the equilibria
areweak(see section1).
TheproofofTheorem bythreelemmas.
4.1 is facilitated
LEMMA and single-
4.1: Xi*i is continuous
valued foreach i E N in each es E E.
Proof:X,*i is the function whichgraphssolutionsto the maximization
problem(4.1), fori e N, forall combinations of levelsof ek (k = j).
Sinceu is strictly quasi-concave, Xi*' is single-valued.Since ui is con-
tinuous,so is thesolutionto (4.1). Q.E.D.
LEMMA 4.2: X,M is uppersemicontinuous forall j.
thenit is continuous.
If it is single-valued,

X2

Equilibria

MIGREe

xl

FIGURE 7

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 51

The proofis omitted, but severalremarksare made.Whenever, forany


combination of levelson otherdimensions, thereis an odd numberof
distinctimagesof the Xj*i, thentherewill be a uniquemedian-X,M
willbe single-valued forthatcombination of levelson otherdimensions
(See Figure6). X1Mwillsimplybe "pieces"of continuous functions(the
Xi*i's). It is clearlycontinuousalongeach piece,as well as wheretwo
pieces"join."Henceit is everywhere continuous.Even in thoseinstances
whereXAmis a correspondence, mappinga particularcombination of
levelson otherdimensions intoa rangeof pointsalong ej, thatrange
willbe closed.Hence it is uppersemicontinuous.9
LEMMA4.3: n x,m 0.
Proof:The argument behindthisproofexploitswell-known fixedpoint
theorems.In particular,
if theXjMare continuous and single-valued and
Rn is compact,thentheconditions of theBrouwerFixedPointTheorem
are met.This theorem impliestheexistence of a fixedpoint,i.e. a non-
emptyintersectionof theX,M. If, on theotherhand,theXjmare upper
semi-continuouscorrespondences,
and Rmis compact,thentheconditions
of the KakutaniFixed Point Theoremare met,implying a nonempty
intersection.10 Q.E.D.
Proofof Theorem4.1
Lemmas 4.1-4.3 guaranteethe existenceof points in n X,M.
3
Let x be one of thosepoints.Considerx' = x + Xej,,foranyj'.
x Efn X1M -e x e Xj,M. FromLemma3.1, theui are single-peaked
on x +
Xej,.Therefore,x e X,,M - x = C (x, x') forallj - x is invulnerable-* x
is a structure-induced
equilibrium. Q.E.D.
The mainpointofthisessayhas nowbeenestablished.
Jurisdictional
and rules of procedure(amendmentcontrol) create equili-
arrangements
brium.Theorem4.1 providesthe groundsforconcluding
thata social
9 Upper semi-continuityis defined in most topology and mathematical econo-
mics texts,e.g., Berge (1963, p. 109), and Nikaido (1968, p. 65). A correspon-
dence b: X -* Y (Y compact) is said to be uppersemi-continuous if (xi) is a
sequencein X withxi -* x and {yi} is a sequencein Y such thatyi e 4 (xi) and
yi-> y impliesy e '1 (x). This is equivalentto requiringthe graphof N(x) to be
closed-see Figure7.
10After constructingthis argumentI discoveredit was nearlyidenticalto a
resultprovedelegantlyby Kramer (1972), Theorem 1', thoughfor different pur-
poses.The detailsof the proofare foundthere.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
52 KennethA. Shepsle

choicearrangement of anycollectionof individual


consisting preferences.
each representableby a continuous,strictly
quasi-concave function,
utility
possessesan equilibrium stateundermajority rule.This is trueforany
committee system forwhichthejurisdictions and a
are single-dimensional
germaneness ruleforamendments is in effect.
This resultcan be extendedmodestly as follows:
COROLLARY 4.2: If Xtm is single-valuedfor all j,
thenthe XjM intersectin a unique point and that
pointis a strongequilibrium.
does not implythatit is a unique
The uniquenessof the intersection
equilibrium.Let {x}==
structure-induced X.im,withx = (x,, . . , xm)
of XjM,xj = median {xi*i}.
and, by definition
lEN
Let yj = median {xj*i} and y = (xl, . . ., x-1, yj, xj?+] . .m. xM). The
i e Cj
pointmaybe a structure-induced equilibrium.Certainly
ifpreferencesare
theny is a structure-induced
separableby jurisdiction The
equilibrium.
factwe exploitedin establishing is thatthereis alwaysa point
existence
fromwhichno majority of the parentorganizationdesireschange(in any
jurisdictionally
permissibledirection).FromTheorem3.2 and its corol-
laries,however,we knowthatequilibriamayariseforotherreasons.
A finalgeneralization
is suggestedby some earlierworkof Slutsky
(1975, 1977a). Let V - {v1, . . . , Vk) be any collectionof linearlyin-
k < m. The vj e V
vectorsin Rm.Fromlinearindependence,
dependent
arepermissible
directions
of change.
COROLLARY 4.3: If the preferencesof i E N are representableby a
continuous,strictlyquasi-concave utilityfunctionin the subspace
and
of Rm spanned by the vj e V, if the vj constitutejurisdictions,
ifa germanenessrulegovernsamendments, then
(i) structure-induced equilibriaexist,and
(ii) a pointx? is a structure-induced equilibrium
ifand onlyif,forall vj e V and all
x E Sj, Rj (x) n T= 0 (where
R1,Sj, and Tj, are definedas in Theorem3.2).

5. Discussion
sectionthereare severalloose ends to be dealt
In thisconcluding

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 53

with.First,the questionof sensitivity is raised.The model of social


choiceoffered in thispaperis moreinstitutionallydetailedthanits pre-
decessors.Indeed,thatis the centralthrustof the paper-institutional
detailsmatter.Buthowmuch?How sensitive ofequilibrium
is theexistence
and its properties to the specificsof institutionalarrangements?This
raisesa second,relatedquestion.What are the effects of institutional
"reforms"? In particular,andthisis theobverseofthepreviousquestion,
underwhat conditionsare equilibriainvariant underinstitutional"re-
forms"?Finally,the prospectsforexploiting the generalframework of
section2 to establishresults
formorecomplexor troublesome institutional
arrangements arebrieflyexplored.

Robustness
ofStructure-Induced
Equilibria
Withsingle-dimensional and a germaneness
jurisdictions rulegovern-
ingtheamendment process,anycollectionofvoterswhosepreferences are
representedbystrictly quasi-concave,continuous utility
functions,andany
committee system, yieldsstructure-induced
equilibria.
1. Existenceis insensitive to changesin the distribution of ideal
pointsor to changesin otherproperties oftheutilityfunctions:In contrast
to preference-inducedequilibrium,theexistence ofstructure-inducedequi-
librium doesnotdependon preference Changesinpreference
distributions.
distributionsor otherutilitypropertiesdo notendanger the existenceof
equilibrium,thoughtheymaychangeitslocation(see below). Even some
oftheproperties assumedabovefortheproofofexistence-continuity and
strictquasi-concavityof theui-may be relaxedto somedegreeso long
as theuppersemicontinuity ofeachX3Mis preserved.
2. The locationof structure-inducedequilibriamaydependon pref-
erencedistributions: As Theorem3.2 and its corollariessuggest,the
equilibrium properties of a statusquo, x?, dependon the relationship
of itsjurisdictional x03,to median{x*") and median{x*')}.
projections,
i eN ieCj
To theextentthatchangesin preference altereitherof these
distributions
thenthelocationof equilibrium
relationships, shifts.However,distribu-
intactwillnotaffect
tionalchangesthatleavetheserelationships theloca-
tionofequilibrium."

and thosethatfollow,dependon the otherconditions


11Theseassertions, of the
analysis"is conductedone
relevanttheoremsholding.In otherwords,this"sensitivity
conditionat a time.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
54 A. Shepsle
Kenneth

3. The existence does notdependupontheparticular


of equilibrium
system
committee C ={Cj}:
The pointx withx, = median{x1*} alwaysexistsand is a structure-
ieN
no matterwhatcommittee
inducedequilibrium, Par-
systemis in effect.
produceadditional
systems
ticularcommittee butassociatedwith
equilibria,
everycommitteesystemis theequilibrium above.
pointidentified
4. The locationof equilibrium by thecommittee
is affected system
Corollary3.1 indicatesthat (3.1) is a necessaryconditiona
in effect:
pointmustsatisfy in orderto be an equilibrium underthecommittee-of-
the-wholearrangement. It is notnecessary committee
underalternative ar-
rangements.The setofequilibria associatedwitha givencommitteesystem
from,
is different though sharescommonelements with,thesetofequilibria
associatedwithsome othercommittee system.In thissense,structural
arrangements matter.
dependon the rule
5. Both existenceand locationof equilibrium
governingtheamendment straightforward
process:Considerthefollowing
result.
ofTheorem
THEOREM 5.1: Iftheconditions 4.1
holdexceptthatM(x) = Rm-theopenrule-
andifSj #0 forsomej,thenx? is a structure-
ifandonlyifitisa preference-
inducedequilibrium
inducedequilibrium.
For a givencommittee, Cj, Sj # 0 whenever xj? :L median{xj*i}. With
i E Cj
our assumption of nonstrategicbehavior,if thisconditionholdsforany
Cj, thenit will proposean alternative to x?. WithM(x) = Rm,any
amendment to theproposal,evena nongermane one, is in order.Thus,
whenever Sj 7 0 for some Cj, committees and jurisdictions no longer
serveto structuresocialcomparisons. Social intransitivitiesare no longer
mitigatedbystructuralor jurisdictional
arrangements. Genericcycling pre-
dominates unlessthesevereconditions thatassurea preference-induced
equilibriumprevail.The onlyroleplayedby thestructural arrangement is
associatedwiththeconditions Sj = 0, viz.,ifthisholds,and onlyifthis
holdsforall Cj, thenstructure-inducedand preference-induced equilibria
arenotequivalent.
On the otherhand,for the closed rule-M(x) = 0-it may be

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 55

shownthatstructure-induced equilibriaalwaysexistand, in fact,those


producedby thegermaneness ruleare a propersubsetof thoseproduced
by theclosed rule.The closed rulegivesmuchstronger veto powerto
conumittees-a pointrecentlyexaminedby Romerand Rosenthal(forth-
coming)-and,as a consequence, producesa large(and alwaysnonempty)
equilibriumset.12
6. Bothexistenceand locationof equilibrium dependon thejuris-
dictional
arrangement: The unidimensional natureofjurisdictions
has per-
mittedus toexploitthe"goodbehavior" ofsimplemajoritychoicefunctions
whenindividual preferences
are single-peaked.
Now considerthefollowing
extreme,butotherwise straightforwardresult.
THEOREM 5.2: If theconditions
ofTheorem4.1
holdexceptthat,fora particular
Cj e C,
f(Cj) = {el, . . . , e.} (that is, g(C,) = Rm),
then,ifSj #&0 forthisCj, thenx? is
a structure-induced
equilibriumifandonlyif
itis a preference-induced
equilibrium.
Thisresultonceagainunderscores theimportant "channeling"roleplayed
by jurisdiction(and structure). If thebinarycomparisons theparentor-
ganizationis allowedto make are not sufficiently restrictedthen the
prospectsforequilibrium increasingly dependuponpreference distribution
requirements. In theextreme(Theorem5.2), theyare necessary and suf-
ficient.
One line of inquiryworthpursuing is thenatureof cyclingas a
function of jurisdictionalrestrictions;thisline has been pursued,in the
absenceof anyjurisdictional considerations, by McKelvey(1976, 1977),
Cohen(1977), andSchofield (1977).
Thisbrieftreatment ofrobustness suggests that,evenwhenstructural
and jurisdictionalmatters are givenmoreattention, equilibriumis stilla
delicateaffair.The resultsof section2, 3, and 4 needto be generalized,
a pointwe turnto shortly. In anothersense,however, thelackof robust-
nessfurther underscores themainthesisoffered here.Structuraland juris-
dictionalarrangements offer theprospect ofequilibrium; theyare alterable
propertiesof institutions(as preferences probablyare not) and thusare
potentialinstruments of equilibrium (disequilibrium) in thehandsof in-
stitutional
designers; and finally,evenif theydo not assureequilibrium,
theyundoubtedly affect(constrain)theformof disequilibrium.
12 For a comparison
of amendment
rules,see Shepsle(1978).

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
56 A. Shepsle
Kenneth

Equilibrium
Structure-Induced Reform
andInstitutional
Traditional socialchoicetheory, to theextentthatitsmainelements
are limited to individual preferences, a preference aggregation mechanism,
and decisivecoalitions, is unlikelyto have muchbearingon substantive
debatesaboutinstitutional reform. The latterdealmostly withthestructure
of choices permitted an organization (as well as its decisive coalitions)
(see Rohde and Shepsle, 1978). Hence a theory which elevatesinstitu-
tionalfeatures is to be welcomed if it can contribute to these debates an
understanding of how institutional practices work (and to whose advan-
tage). WhileI do not claimto have providedthistheory, the structure
offered in section1, or somevariation, is one promising line of attack.
For now let me focuson one issueinvolving institutional reform.
In bothhousesoftheUnitedStatesCongress, thelastfewyearshave
witnessedthe salienceof jurisdictional realignments for standingcom-
mittees. To obtainsomepurchaseon thisproblem, considerthejurisdic-
tionalarrangement J? = {Jk}. witheach Jk e J? a singlebasis vectorin
E? = {e,, . . . , em}. Associate,under the conditionsof Theorem4.1, a
equilibrium,
structure-induced x?. Question:Whathappenswheneach
by0 degrees?
vectoris rotated
jurisdiction Call theassociatedjurisdictional
arrangement,basis, and equilibrium, J@,Ee and x@. From
respectively,
Corollary4.3, xe exists.But whatis the relationshipof xe to x?? We
havethefollowingresult:
THEOREM 5.3: If theconditions of Theorem4.1 hold,
and ifx? is a preference-induced then
equilibrium,
rotations.
underjurisdictional
it is invariant
A preference-induced equilibrium (whichis a structure-induced equilib-
rium-see Theorem2.1) remainsone underrotationsof jurisdictional
vectors;it dependsin no wayon jurisdictional arrangements. In general,
however, x? #Fxe. Letting e: J -* Rm be the correspondence whichas-
sociatestheequilibrium state(s) identified in Theorem 4.1 withjurisdic-
tionalarrangement J,thefollowing questionis posed: Under what condi-
tionsdoes e possesscontinuity-like properties? That is, forexample, when
is lime(Je) = e(J0)?13 Put insubstantive terms, thissort questionseeks
of
e--o
theclass of jurisdictional
to distinguish thathave an impacton
reforms
outcomes are
fromthosethat merely cosmetic.
13 One could pose a relatedquestionforthecorrespondence e thatassociatesall
the structure-induced arrangement.
equilibriaof a particularjurisdictional

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGEMENTS AND EQUILIBRIUM 57

ConcludingObservations
The theoreticalconceptsof section1 have been offered in orderto
properties
elevateinstitutional in thedebateon socialchoice.The overly
atomisticrepresentationsof traditionalsocial choicetheoryand general
equilibriumtheory are troublesome whosee a world
to politicalscientists
ofindividualswhosechoicesare constrained by theoperating characteris-
and
ticsofpolitical economic institutions.14
CurrentlyI ampreparing papersonalternative amendment procedures
(see Shepsle,1978), complexjurisdictions, jurisdictionalchange,and
hierarchies In eachcase theideais toidentify
ofdecentralization. equilibria
iftheyexistor to tracethepathof disequilibrium iftheydo not.Under-
lyingtheseprojectsis theexpectation thatweakening rarely
an otherwise
explored15axiom of Arrow-the Social Completeness Axiom-in the
"right"way,andineffect capturing thewaysinwhichinstitutional structure,
andotheroperating
jurisdiction, channelandconstrain
characteristics social
comparisons, social choicein institutionally-richcontextscan be under-
stood.

20 December1977
submitted
Manuscript
received
Finalmanuscript 1 May1978

REFERENCES
Arrow,KennethA. 1963. Social choice and individualvalues. 2nd ed. New York:
Wiley.
Berge,Claude. 1963.Topologicalspaces.New York: Macmillan.
Cohen, Linda. 1977. The structureof maximummajorityrule cycles. Paper pre-
sentedat the Annual Meetingof the AmericanEconomic Association,New
York,December26-30, 1977.
Davis, Otto A., Melvin J. Hinich, and Peter C. Ordeshook.1970. An expository
development of a mathematicalmodel of the electoralprocess.AmericanPo-
liticalScienceReview,64 (June1970): 426-449.
Denzau, Arthurand RobertParks. 1973. Equilibriumin an economywithprivate
and spatialpoliticaldimensions.Paper presentedat the WinterMeetingof the
EconometricsSociety,1973.
Fishburn,PeterC. 1973. The theoryof social choice. Princeton:PrincetonUniver-
sityPress.
theoremswithoutthe social completenessaxiom. Eco-
. 1974. Impossibility
nometrica, 42 (July1974): 695-704.

14See mycomments in Shepsle(1977).


of whichI am awareis Fishburn(1974).
15 The one exception

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
58 Kenneth
A. Shepsle

Hoyer,R. W. and LawrenceS. Mayer. 1974. Comparingstrategies in.a spatialmodel


of electoralcompetition.AmericanJourtnal of Political Science, 18 (August
1974): 501-523.
Kramer,Gerald H. 1972. Sophisticated votingover multidimensional choice spaces.
Journalof Mathematical Sociology,2 (1972): 165-180.
. 1973. On a class equilibriumconditionsfor majorityrule. Econometrica,2
(March 1973): 285-297.
McKelvey,RichardD. 1976. Intransitivities in multidimensional votingmodelsand
some implicationsfor agenda control.Journalof Economic Theory,12 (June
1976): 472-482.
. 1977. Generalconditionsfor global intransitivities
in formalvotingmodels.
Paper presented at AnnualMeetingof the Public Choice Society,New Orleans,
March,1977.
and RichardE. Wendell.1976. Votingequilibriain multidimensional choice
spaces. Mathematicsof OperationsResearch,1 (May 1976): 144-158.
Manley,John.1969. WilburD. Mills: A studyin congressional influence.American
PoliticalScienceReview,63 (June1969): 442-464.
Matthews,Steven. 1977. The possibilityof votingequilibria.Unpublishedpaper,
CaliforniaInstituteof Technology.
Nikaido,Hukukane.1963. Convexstructures and economictheory.New York: Aca-
demicPress.
Pattanaik,PrasantaK. 1971. Votingand collectivechoice. Cambridge:Cambridge
University Press.
Plott,Charles R. 1967. The notionof equilibriumand its possibility undermajority
rule.AmericanEconomicReview,57 (September1967): 787-806.
Rohde,David W. and KennethA. Shepsle.1978. Thinkingabout legislativereform.
In Leroy Rieselbach,ed., Legislativereform.Boston: Lexington.
Romer,Thomas and Howard Rosenthal.Forthcoming. Politicalresourceallocation,
controlledagendas,and the statusquo. Public Choice, (Winter1978)
Schofield,Norman. 1977. Local acyclicityand the null dual set in dynamicgames.
Paper presentedat MathematicalSocial Science Board Conferenceon Game
Theoryand PoliticalScience,Hyannis,Mass. July,1977.
Sen, AmartyaK. 1966. A possibilitytheoremon majoritydecisions.Econometrica,
34 (April 1966): 491-499.
. 1970. Collectivechoice and social welfare.San Francisco: Holden-Day.
Shepsle,KennethA. 1977. The futureof publicchoice: A politicalscientist's views.
Banquet address,Meetingsof the Public Choice Society,New Orleans,1977.
. 1978. Institutional
structure and policychoice: Some comparativestaticsof
amendmentcontrolprocedures.Paper presentedat Conferenceon Political
Scienceand the Studyof Public Policy,HickoryCorners,Michigan,May 1978
and at the AnnualMeetingof the AmericanPoliticalScience Association,New
York,September1978.
Sloss, Judith.1973. Stable outcomesin majorityrule votinggames. Public Choice,
15 (Summer1973): 19-48.
Slutsky,StevenM. 1975. Majorityrule and the allocationof public goods. Unpub-
lishedPh.D. dissertation, Yale University.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions
INSTITUTIONAL ARRANGBMENTS AND EQUILIBRIUM 59

. 1977a. A votingmodel for the allocationof publicgoods: Existenceof an


equilibrium.Journalof Economic Theory,14 (1977): 299-325.
. 1977b. Equilibriumunder pluralityvoting.Paper presentedat Annual
Meetingsof the Public Choice Society,New Orleans,March, 1977.
Zeckhauser,Richard J. and Milton C. Weinstein.1974. The topologyof Pareto-
optimalregionswithpublic goods. Econometrica,42 (July 1974): 643-666.

This content downloaded from 142.58.129.109 on Fri, 05 Jun 2015 13:42:16 UTC
All use subject to JSTOR Terms and Conditions

You might also like