An Agenda for the Study of Institutions

Author(s): Elinor Ostrom

Source: Public Choice, Vol. 48, No. 1 (1986), pp. 3-25
Published by: Springer
Stable URL: http://www.jstor.org/stable/30024572
Accessed: 12/10/2009 16:37

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=springer.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

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.

Springer is collaborating with JSTOR to digitize, preserve and extend access to Public Choice.

http://www.jstor.org
 Public Choice48: 3-25 (1986).
z 1986 Martinus
Nijhoff Printedin the Netherlands.
Dordrecht.
Publishers,

An agenda for the study of institutions*

ELINOROSTROM
513NorthPark,Bloomington,
of PoliticalScience,IndianaUniversity,
Department
IN 47405, and
Workshopin PoliticalTheory&PolicyAnalysis,IndianaUniversity,
Bloomington,
Indiana

1. The multiplemeaningsof institutions

Recently,publicchoicetheoristshaveevidencedconsiderableinterestin the
studyof institutions.WilliamRiker(1982:20) recentlyobserved,for exam-
ple, that 'we cannot studysimplytastes and values, but must studyinstitu-
tionsas well.' Littleagreementexists,however,on whattheterm'institution'
means,whetherthe studyof institutionsis an appropriateendeavor,andhow
to undertakea cumulativestudy of institutions.
Rikerdefinesinstitutionsas 'rulesaboutbehavior,especiallyaboutmak-
ing decisions'(1982:4). CharlesPlott also definesinstitutionsto mean 'the
rulesfor individualexpression,informationtransmittal,and social choice
... ' (1979: 156). Plott uses the term 'institutions' in his effort to state the fun-
damentalequation of public choice theory. Using O as an unspecified
abstractoperator,Plott's fundamentalequationis:

preferencesG institutionsO physicalpossibilities= outcomes (1)

Plott himselfpoints out, however,that the terminstitutionrefersto dif-
ferentconcepts. He ponders:

Couldit be, for example,that preferencesand opportunitiesalonedeterminethe structureof

institutions(includingthe constitution)?Thesequestionsmightbe addressedwithoutchanging
'the fundamentalequation'but beforethatcan be done, a lot of workmustbe done on deter-
miningexactlywhatgoes underthetitleof an 'institution.'Arecustomsandethicsto be regard-
ed as institutions?Whataboutorganizationssuchas coalitions?Theseare embarassingques-
tionswhichsuggestthe 'fundamentalequation'is perhapsnot as fundamentalas we wouldlike
(Plott, 1979:160;my emphasis).

* Thispaperwas deliveredas the Presidentialaddressat the PublicChoiceSocietymeetings,

Hilton Hotel, Phoenix, Arizona, March30, 1984. I appreciatethe supportof the National
ScienceFoundationin the form of GrantNumberSES 83-09829 and of WilliamErickson-
Blomquist,Roy Gardner,JudithGillespie,Gerd-MichaelHellstern,RobertaHerzberg,Larry
Kiser,VincentOstrom,RogerParks, Paul Sabatier,ReinhardSelten,KennethShepsle,and
YorkWillbernwho commentedon earlierdrafts.
4

Plott's questionsare indeedembarassing.No scientificfield can advance

far if the participantsdo not sharea commonunderstandingof key termsin
their field. In a recentvolume entitledTheEconomic Theoryof Social In-
stitutions,AndrewSchotterspecificallyviewssocialinstitutionsas standards
of behaviorratherthan the rules of the game. What Schottercalls 'social
institutions':
arenot rulesof thegamebutratherthealternativeequilibriumstandardsof behaviororconven-
tions of behaviorthatevolvefroma givengamedescribedby its rules.In otherwords,for us,
institutionsarepropertiesof theequilibriumof gamesandnot propertiesof thegame'sdescrip-
tion. Wecareaboutwhattheagentsdo withtherulesof thegame,notwhattherulesare(Schot-
ter, 1981:155).

Schotter sees his enterpriseas a positive analysis of the regularitiesin

behaviorthat will emergefrom a set of rulesand contraststhis with a nor-
mativeapproachthat attemptsto examinewhichruleslead to whichtypes
of behavioralregularities.Schotterdrawson a richintellectualtraditionthat
stressesthe evolutionof learnedstrategiesamong individualswho interact
with one another repeatedlyover a long period of time (Menger, 1963;
Hayek, 1976;1978;see also, Ullman-Margalit,1978;Taylor, 1976;Nozick,
1975). Rawls characterizesthis view of how individualscome to follow
similarstrategiesovertime as 'thesummaryviewof rules'(Rawls,1968:321).
Still another way of viewing 'institutions' is equivalent to the term
'politicalstructure.'Thisviewdiffersfromthatof Schotterin thatit doesnot
equateinstitutionswithbehavioralregularities.It differsfromthatof Riker
and Plott in that it does not focus on underlyingrules. Institutions,defined
as politicalstructure,referto attributesof the currentsystemsuch as size
(Dahl and Tufte, 1973), degreeof competition(Dye, 1966; Dawson and
Robinson;1963),extentof overlap(ACIR, 1974),and otherattributesof a
currentsystem.
The multiplicityof uses for a key termlike 'institution'signalsa problem
in the generalconceptionheldby scholarsof how preferences,rules,individ-
ual strategies,customsand norms,and the currentstructuralaspectsof on-
goingpoliticalsystemsarerelatedto one another.Overtimewe havereached
generalagreementabouthow we will use suchkeytheoreticaltermsas 'pref-
erences,''actions,' 'outcomes,''coalitions,'and 'games.'Further,we have
a generalagreementabout how these conceptsare used in our theoriesto
generatepredictedoutcomes.
The multiplereferentsfor the term 'institutions'indicatesthat multiple
conceptsneedto be separatelyidentifiedand treatedas separateterms.We
cannot communicate effectively if signs used by one scholar in a field have
different referents than the same sign used by another scholar in the same
field. As scholars, we are in our own game situation - a language generating
game. The 'solution' is the result of our choice of strategies about the use of
a set of terms to refer to the objects and relations of interest in our field.
 5

No one canlegislatea languagefor a scientificcommunity.Scholarsbegin

to use a languageconsistentlywhentermsare carefullydefinedin a manner
perceivedby other scholarsas useful in helpingto explainimportantphe-
nomena.In this presentation,I do not try to resolvethe debateover which
of the definitionsof institutionis the 'rightdefinition.'Instead,one concept
- that of rules- is used as a referentfor the term 'institution,'and defined.
I distinguishrulesfromphysicalor behaviorallawsand discussthe prescrip-
tive natureof rules. Then I show how theoristsuse rules in public choice
analysis.Twomethodologicalissuesareraised.Onerelatesto theconfigura-
tional characterof rules. A secondrelatesto the multiplelevels of analysis
neededforthe systematicstudyof rules.Inthelastsection,I proposean alter-
nativestrategythattakesinto accountthe configurationalcharacterof rules
and the need for a self-consciousstudy of multiplelevels of analysis.

2. What is meantby rules

Focusingspecificallyon the term'rule' does not immediatelyhelp us. Even

this narrowerterm is used variously.Shimanoff(1980:57) identifiedover
100 synonyms for the term 'rule' (see also Ganz, 1971). Even among
politicaleconomiststhe term is used to both refer to personalroutinesor
strategies(e.g., Heiner, 1983)as well as to a set of rulesused by morethan
one personto orderdecisionmakingin interdependentsituations.In game
theory, 'the rules of the game includenot only the move and information
structureand the physical consequencesof all decisions, but also the
preferencesystemsof all the players'(Shubik, 1982:8).
Rules, as I wish to use the term, are potentiallylinguisticentities(Ganz,
1971;V. Ostrom, 1980;Commons, 1957)that refer to prescriptionscom-
monly known and used by a set of participantsto order repetitive,in-
terdependentrelationships.Prescriptionsreferto whichactions(or statesof
the world)arerequired,prohibited,orpermitted.Rulesarethe resultof im-
plicitor explicitefforts by a set of individualsto achieveorderand predic-
tabilitywithin defined situationsby: (1) creatingpositions (e.g., member,
convener,agent, etc.); (2) statinghow participantsenteror leavepositions;
(3) statingwhichactionsparticipantsin thesepositionsarerequired,permit-
ted, or forbiddento take;and (4) statingwhichoutcomeparticipantsarere-
quired,permitted,or forbiddento affect.
Rulesarethusartifactsthataresubjectto humaninterventionandchange
(V. Ostrom, 1980)). Rules, as I wish to use the term, are distinct from
physical and behavioral laws. I use the term differently than a game theorist
who considers linguistic prescriptions as well as physical and behavioral
laws to be 'the rules of the game.' If a theorist wants only to analyze a given
game or situation, no advantage is gained by distinguishing between rules,
6

on the one hand, and physicalor behaviorallaws, on the other hand. To

changethe outcomesof a situation, however,it is essentialto distinguish
rulesfrombehavioralor physicallaws. Rulesare the meansby whichwe in-
terveneto changethe structureof incentivesin situations.It is, of course,
frequentlydifficultin practiceto changethe rulesparticipantsuse to order
their relationships.Theoretically,rulescan be changedwhile physicaland
behaviorallaws cannot. Rules are interestingvariablespreciselybecause
theyarepotentiallysubjectto change.Thatrulescan be changedby humans
is one of their key characteristics.
That rules have prescriptiveforce is anothercharacteristic.Prescriptive
force meansthat knowledgeand acceptanceof a rule leads individualsto
recognizethat, if they breakthe rule, other individualsmay hold them ac-
countable(seeHarr6,1974).Onemaybe heldaccountabledirectlyby fellow
participants,who call rule infractionto one's attention,or by specialists-
refereesor public officials - who monitor performance.The term 'rules'
should not be equatedwith formal laws. Formallaws may become rules
when participantsunderstanda law, at least tacitly, and are held accoun-
table for breakinga law. Enforcementis necessaryfor a law to become a
rule. Participantsmay design or evolve their own rules or follow rules
designedby others.
An unstatedassumptionof almost all formal models is that individuals
are, in general,rule followers.Even whentheoristslike Becker(1976)have
overtlymodeledillegalbehavior,some probabilityis presumedto exist that
illegal actions will be observed, and if observed by an enforcer, that
penaltieswill be extracted.Most publicchoiceanalysisis of the rulesin use
- or working rules as John R. Commons (1957) called them. Many in-
terestingquestionsneedexplorationconcerningthe originof rules,the rela-
tionshipof formallawsto rules,andprocessesfor changingrules.But, these
topics cannot be addressedhere.
Considerabledisputeexists over the prescriptiveforce of 'permission.'
Ganz(1971)and Shimanoff(1980)arguethat prescriptiveforceis restricted
to 'obligation'and 'prohibition'and does not include 'permission,'while
Commons (1957), von Wright (1968), V. Ostrom (1980), and Toulmin
(1974)all overtlyinclude'permission'in theirconceptionof rules. Part of
this difficultystems from efforts to predictbehaviordirectlyfrom specify-
ing rulesratherthanviewingrulesas a set of variablesdefininga structured
situation.In this rule-structuredsituation,individualsselectactionsfrom a
set of allowableactions in light of the full set of incentivesexistingin the
situation.
Instead of viewing rules as directly affecting behavior, I view rules as
directly affecting the structure of a situation in which actions are selected.
Rules rarely prescribe one and only one action or outcome. Rules specify
sets of actions or sets of outcomes in three ways:
 7

(1) A rule states that some particularactions or outcomes are forbidden.

The remainingphysicallypossible or attainableactions and outcomes
are then permitted.The rule states what is forbidden.A residualclass
of actionsor outcomesis permitted.(Most traffic laws regardingspeed
are of this type. The upperand lowerboundsof the permittedspeedare
delimitedby forbiddingtransitabove and below specific speeds.)
(2) A rule enumeratesspecificactionsor outcomesor statesthe upperand
lowerboundof permittedactionsor outcomesand forbidsthosethatare
not specificallyincluded.(Mostpublicagenciesareauthorizedto engage
in only those activitiesspecificallyenumeratedin the organicor special
legislationthat establishesthem.)
(3) A rule requiresa particularaction or outcome. (Recentefforts to con-
strainjudical discretionare rules of this type. A judge must impose a
particularsentenceif a juryconcludesthat a defendantis guiltyof a par-
ticularcrime.)

Onlythe thirdtype of rulerequiresthat an individualtake one and only one

actionratherthanchoose froma set of actions.Thethirdtypeof ruleis used
much less frequentlyto structuresituationsthan the first two.
In the everydayworld, rulesare statedin wordsand must be understood
(at leastimplicitly)for participantsto usethemin complexchainsof actions.
For analysis, however, rules can be viewed as relationsoperatingon the
structureof a situation. Rules can be formally representedas relations,
whose domainare the set of physicallypossiblevariablesand theirvalues,
and whose range are the values of the variables, in the situation under
analysis.(Seebelowfor furtherelaboration.)Viewingrulesas directlyaffec-
ting the structureof a situation,ratherthan as directlyproducingbehavior,
is a subtle but extremelyimportantdistinction.

3. How rules are used in public choice theory

Most publicchoice theorists'know' that multiplelevels of analysisare in-

volvedin understandinghow rulesaffect behavior.Butthis tacit knowledge
of the multiple levels of analysis and how they intertwineis not self-
consciouslybuilt into the way we pursueour work. Plott, for example,has
been engagedin a sophisticatedresearchprogramrelatedto the theoretical
and experimentalstudyof rules. Yet, as discussedabove, he poses the cen-
tral questionof our disciplineas a single equation, ratherthan as a set of
equations. We have not yet developeda self-consciousawarenessof the
methodologicalconsequencesof the multiplelevels of analysisneededto
study the effects of rules on behaviorand outcomes.
Most public choice theorists also 'know' that configurationsof rules,
8

ratherthan single rules, jointly affect the structureof the situations we

analyze.Again, this tacit knowledgeis not reflectedin the way we proceed.
Most of our theoreticalwork has proved theorems about the expected
resultsof the use of one rule in isolationof other rulesas if rulesoperated
separablyratherthan configurationally.
To illustrate the multiple levels of analysis and the configurational
characterof rules, I will use severalexamplesfrom publicchoiceliterature.
The first examplecombinesthe work of severalscholarswho have studied
how citizen'spreferencesfor publicgoods aretranslatedthroughtwo arenas
- an electoralarenaanda bargainingarena- into an agreementthata bureau
will producea particularquantityof goods for a particularbudget.The se-
cond exampleis from an experimentalstudy of Grether,Isaac, and Plott
(1979) of the combinationof default condition rules used in conjunction
withaggregationrules.Thethirdexampleis fromMcKelveyandOrdeshook
(1983) who conductedan experimentalstudy of the conjunctionof three
rules.

3.1 Rules as they affect outcomesin electoraland bargainingarenas

In a classicmodel of the electionarena,AnthonyDowns (1957)concludes
that electoralproceduresbasedon pluralityvote will constraina governing
partyto select (and thereforeproduce)the output-costcombinationmost
preferredby a median voter within a community.The Downsian model
predictsan optimalequilibriumin termsof allocativeefficiency. Downs's
predictionof optimalperformanceresultsfrom his analysisof the behavior
of electedofficials underthe threatof being voted out of office by a com-
petingparty. It is the presenceof a competitorreadyto snatchany advan-
tage that pushesthe governmentparty towardconstantattentionto what
citizensprefer.
When WilliamNiskanen(1971) examineshow bureaucracyaffects the
linkage between citizen preferences and government performance, he
focuses on the processof bargainingbetweenthe team of electedofficials
(called the sponsor by Niskanen)and bureauchiefs assignedthe respon-
sibility to direct agencies producing the desired goods and services.
Niskanenassumesthat a bureauchief attemptsto obtain as largea budget
as possiblein orderto securethe most privategain and to producethe most
goods and services for a community. Niskanen's elected officials, like
Downs's, know the preferencesof the citizens that elect them. So do the
bureauchiefs. However,electedofficials do not knowthe productioncosts
of the bureau.The equilibriumpredictedby Niskanenis not responsiveto
citizenpreferencessince more than optimallevels of output are produced.
The predicted result is technically efficient, but unresponsiveto the
preferencesof those served.
 9

Niskanen'smodel is based on an assumptionthat bureauchiefs could

threatenelectedofficials with no output if the officials did not agreeto the
initial demand. Romer and Rosenthal(1978) argue that a more realistic
assumptionwould be that the budgetrevertsto the status quo budget(the
one used for the previousyear) if the officials (or, the generalpublic in a
referendum)did not agree to the initial budgetaryrequest. Changingthis
assumptionin the model, Romerand Rosenthalcontinueto predictthat the
equilibriumbudget-outputcombination representsa nonoptimal, over-
supply. Their predictedoutcome is, however, less than that predictedby
Niskanen.
A dramaticchange in assumptionsis made by McGuire,Coiner, and
Spancake (1979) who introduce a second bureau to compete with the
monopolybureauchief in the bargainingarena.'Whateveroffer is madeby
one bureaucan then be challengedby the secondbureau.Overtime the of-
fers will approachthe same optimal level as predictedby Downs. If one
bureauproposestoo high a budget, the other will be motivatedto make a
counterofferof a moreoptimalbudget-outputcombination.As the number
of bureausincreasesbeyondtwo, the pressureon all bureausto offer an op-
timal budget-outputcombinationalso increases.
The above models focus primarilyon the structureof an operational
situationandonly indirectlyon the rulesyieldingthat structure.Withoutex-
plicit analysisof the rulesandotherfactorsaffectingthe structureof a situa-
tion - suchas the attributesof goods and the community- implicitassump-
tions underlyingthe overtanalysismay be the most importantassumptions
generatingpredictedresults.2In the analysis of electoraland bargaining
arenas,all theoristsusedsimilarassumptionsaboutthe natureof goods and
communitynorms. Goods are modeledas divisiblein productionand sub-
ject to a knowntechnology.In regardto norms,all presumea high level of
cutthroatcompetitionis acceptable.Theseassumptionsare not responsible
for the differencesamong predictedoutcomes.
The models have, however, different implicit or explicit assumptions
about some of the rules affecting the situation. The models developedby
Niskanen and by Romer and Rosenthal both give the bureau chief the
capacityto make a 'take it or leave it' offer. Both of these models assume
an authorityrulegivingthe bureauchief full controloverthe agenda.Both
modelsalso assumethat the aggregationrulebetweenthe bureauchief and
the sponsors is unanimity.The models differ, however, in regardto the
defaultspecifiedin the aggregationrule.Niskanenpresumedthis rulewould
allowthe budgetto revertto zero. No agreement- no funds!An aggregation
rule with such a default condition can be formally stated as:

Bt+l = ( Bb iff Bbc = B,; 0 otherwise), where (2)

Bt+l = the budget-outputcombinationfor the next period,
10

Bbc = the budget-outputproposalof the bureauchief,

Bs = the budget-outputproposalacceptedby the sponsor.
In otherwords,the aggregationruleaffectingthe structureof this situaction
requiresunanimityamongthe participantsand sets the budgetfor the next
time periodto zero if such agreementis not reached.The first part of this
rule states the outcome when there is unanimousagreement.The second
part of this rule states the outcome when there is no agreement,or the
default condition.
Romerand Rosenthalpresumedthe rulewouldbe to continuethe budget
in effect for the previousyear. No agreement- continuanceof the status
quo! Their rule can be formallystated as:

Bt+l = ( Bbc iff Bbc = Bs; Bt otherwise), (3)

where Bt is the level of the currentbudget-outputcombination.

Niskanenand McGuire,Coiner, and Spancakeagreeon unanimityand

the default conditionof the aggregationrule, but differ on the boundary
rulesallowingentryof potentialproducersinto the bargainingarena.Once
a position rule has defineda position, Si, such as a bureauchief, a formal
boundaryrule consistentwith the Niskanenmodel could be stated as:

Let Si = (1) (4)

A boundaryruleconsistentwiththe McGuire,Coiner,and Spancakemodel

would be the following:

Let Si = 11, ..., n) (5)

Assumingthat the otherrulesare similar,we can arraythe configuration

of rulesthat differ in the variousanalysesas shownin Figure1. The Down-
sianmodelis placedin theupperleft cell sincehe madea similarassumption
aboutthe defaultconditionof the aggregationruleas Romerand Rosenthal
(see Downs, 1957:69), but had to assumeimplicitlythat elected officials
controlledthe agendain their bargainingrelationshipswith bureauchiefs
(see Mackayand Weaver,1978).Consequently,the differencein the results
predictedby Downs, by Niskanen, and by Romer and Rosenthalcan be
relatedto changesin authorityrules and aggregationrules holding other
rules constant.
McGuire,Coiner,and Spancakeacceptedthe Niskanenpresumptionof
a zero reversionlevelwhilechangingthe boundaryrulesallowingproducers
to enterthe bargainingprocess.This changein boundaryrulesgeneratesa
differentsituationleadingto a predictionof relativelyoptimalperformance
 11

Figure1. Predictedequilibriumbudget/outputcombinationsunderdifferentruleconfigura-
tions

Authorityrules Boundaryrules

Aggregation Entryto bargainingprocess Allow multiplebureausto

rules restrictedto one bureau enterbargainingprocess

Openagenda Downs (1957) Parksand E. Ostrom(1981)

Reversion Equilibriumis the most
level is preferredbudget/output
statusquo combinationof the median
voter. Thus, preferences
of medianvoter dominate
decision.
Even if no direct
competitionbetweentwo
producersservingsame
jurisdiction,presenceof
comparisonagenciesin same
urbanarea will reducecosts
of monitoringand increase
pressuretowardan equilibrium
producingthe highestnet value
for the community.

Reversion No model yet developedfor No model yet developedfor

level is this combinationof rules. this combinationof rules.
zero budget

Restricted Romerand Rosenthal No model developedfor this

agendacon- (1978) combinationof rules, but given
trolledby Equilibriumis the McGuire,Coiner,and
bureauchief highestbudget/output Spancake(1979)statusquo
Reversion combinationthat provides reversionlevel can only
level is the medianvoter with at enhancetendencyof
statusquo least as muchvalue as the equilibriumto move toward
statusquo highestnet value for the
community.

Reversion Niskanen(1971; 1975) McGuire,Coiner,and

level is Equilibriumis the largest
zero budget budget/outputcombination
capableof winning
majorityapprovalin an
all-or-nothingvote.
Preferenceof medianvoter
is only a constraint.
Spancake(197)
Equilibriumtends over time
towardbudget/outputcom-
binationproducingthe highest
net value for the community.

as contrastedto Niskanen's predictionof nonoptimality.The change in

boundaryrulesopens up a new column of potentialoperationalsituations
undervaryingconditionsof authorityand aggregationrules.An effort that
Parksand Ostrom(1981)madeto examinethe effect of multipleproducers
in metropolitanareasuponthe efficiencyof publicagenciesis closelyrelated
to the rule conditionsspecifiedin the upperright-handcell. The implica-
12

tions of the situations created by the other combinations of rules

representedin the second column have not yet been explored.
In this discussionI wantedto illustratewhatI meantby a 'ruleconfigura-
tion'. Figure 1 presentsa visual displayof the configurationof rules that
areconsistentwiththe modelsof Downs, Niskanen,Romerand Rosenthal,
and McGuire,Coiner, and Spancake.The resultspredictedin a situation,
using one rule, are dependentupon the otherrulessimultaneouslyin force.
Both Niskanen,and Romerand Rosenthalassumethatonly one bureaucan
be presentin the bargaining.The boundaryruleis the same.Theirdifferent
resultsstem from the variationin the defaultconditionof the aggregation
rule. Both Niskanen and McGuire,Coiner, and Spancakeagree on the
default condition, but differ in regard to the boundaryrule. Different
resultsare predicteddependenton the configurationof rules, ratherthan
any single rule, underlyingthe operationalsituation.
Second,I wantedto illustratethe multiplelevelsof analysisinvolved.The
overtmodelspresentedby thesetheoristsareall at one level. By examining
the rulesaffectingthe structureof thesemodels, I have focusedon a second
level of analysis.

3.2 Committeedecisionsunderunanimityand varyingdefault conditions

A secondexampleof the studyof rulesby publicchoicetheoristsis a recent
set of experimentsconductedby Grether,Isaac, and Plott (1979) who ex-
aminethe effect of variousrules for assigningairportslots. Under one ex-
perimentalcondition, the Grether, Isaac, and Plott situation involves a
committeeof 9 or 14 individualsthat had to dividea discreteset of objects
('cards'or 'flags')usinga unanimityrule. Threedefaultconditionsareused
if unanimityis not reached:

(a) If the committeedefaulted,each committeememberreceivedhis/her 'initialalloca-

tion' of slots thatwasunambiguouslyspecifiedand knownbeforethe meetingbegan.
(b) If the committeedefaulted,slots wereallocatedrandomly.
(c) If the committeedefaulted,slots weretakenat randomonly fromthose withlargein-
itial allocationsand givento those with smallor no initialallocation(Grether,Isaac,
and Plott, 1979:V-2).

All threeof these rulescan be statedin a form similarto that of equations

(2) and (3) above.
While Romerand Rosenthalmake a theoreticalargumentthat the par-
ticular default condition used as part of an unanimityrule affects the
predicted outcomes, Grether, Isaac, and Plott provide evidence that default
conditions markedly affect behavior. The decisions about slot allocations
reached by committees tended to shift directly to the value specified in each
of the default conditions.
 13

Insummary,thecommitteedecisionsaresubstantiallyinfluencedif not completelydetermined

by the consequencesof default.Underthe grandfatherarrangement,'hardnosed'committee
memberswill simplydefaultratherthan take less than the defaultvalue. Socialpressuresdo
exist for those with 'large'initialendowmentsto give to those with 'small'endowments,but
evenif thereis no defaultbecauseof concessionsto social pressurethe final outcomeis not
'far' fromthe 'grandfather'alternative. On the otherhand, whenthe consequenceof default
is an equalchancelottery,the slotswillbe dividedequally,independentof the initialallocation
.... Defaultvaluesliterallydeterminethe outcomesin processessuchas these(Grether,Isaac,
and Plott, 1979:V-7).

It has frequentlybeen presumedthat aggregationrulesvariedunidimen-

sionallyacrossone continuumfrom an 'any one' rule to a unanimityrule
(BuchananandTullock, 1962).Whatshouldnow be recognizedis thatmost
prioranalysisof aggregationruleshas implicitlyor explicitlyassumedonly
one of the possible default conditionsthat work in combinationwith the
votingruleto yield incentivesin the operationalsituation.Thereis nothing
inherentlyconservativeabout a unanimityruleunlessthe defaultcondition
is the status quo.
Cumulativeknowledgefrom the analysisof these diversesituationsre-
quiresthat we understandthat Romerand Rosenthaland Grether,Isaac,
and Plott are examiningthe effect of variationsof the same rule given the
preferencesof participants.If some participantsstronglypreferother out-
comesto that statedin a defaultrule, a strongbargainercan threatenthem
withthe defaultunlessthe final outcomeis movedcloserto his own prefer-
red outcome. But when some participantspreferthe outcomestatedin the
defaultrule, they can afford to block any proposalsthat do not approach
this condition(see Wilson and Herzberg,1984).
To enhancecumulation,we need to develop formal representationsfor
rulesthemselvesas well as for the action situationson whichrulesoperate.
Mostformalanalyseslooselystatethe rulesaffectingthe structureof the ac-
tion situation: (1) in the written paragraphsleading up to the formal
representationof the situation,(2) in footnotesjustifyingwhy the presenta-
tion of the situationis modeledin a particularmanner,or (3) even worse,
leavethemunstated,as implicitassumptionsunderlyingthe formalanalysis
of the situationitself.3

3.3 PMR, germaneness,and open versusclosed informationrules

An experimentconductedby McKelveyand Ordeshookstronglydemon-
stratesthe configurationalrelationshipswhen pure majorityrule (PMR)
is combinedwith one 'germaneness'rule and two informationrules. PMR
and a loose operationalizationof a germanenessrule - a change in out-
come can be made in only one dimension on any one move - is used
throughoutthe experiment.4McKelveyand Ordeshookuse a closed or an
14

open informationrule. Undertheir 'closed'rule, membersof a five-person

committeecan speak only if recognizedby the chair, can addressonly the
chair, and can make commentssolely relatedto the particularmotion im-
mediatelybeingconsidered.Undertheir'open' rule,participantscan speak
withoutbeingrecognized,can talk to anyone,and can discussfutureas well
as presentmotions.
McKelveyand Ordeshookfind that the distributionof outcomesreached
underthe closedinformationrule,whenusedin combinationwithPMRand
their germanenessrule, to be significantlydifferentthan the distributions
of outcomesreachedunderthe open informationrule. The experimentis a
good exampleof how rules operateconfigurationally.
Rules affectingcommunicationflow and content affect the type of out-
comes that will be producedfrom PMR combinedwith a particularger-
manenessrule. McKelveyand Ordeshook,however, interprettheir own
resultsratherstrangely.Theirovert hypothesisis 'that the abilityto com-
municatefacilitatescircumventingformalproceduralrules' (p. 8). A close
examinationof theirseriesof experimentsfindsno evidenceof participants
breakingthe rules laid down by the experimenters.What they do test is
whethera rulegivingcapabilitiesor assigninglimitationson communication
patternschangesthe wayin whichPMRandtheirgermanenessruleoperate.
Theytest the configurationaloperationof rulesystems.And, they find that
the operationof one ruledependsuponthe operationof otherrulesin a rule
configuration.

4. Consequencesof the configurationalcharacterof rules on the

appropriatestrategiesof inquiry
These three examplesprovidestrong evidencefor the configurationalor,
nonseparable,attributeof rules.Thisleadsme to argueagainstan implicitly
held belief of some scholarsthat what we learnabout the operationof one
rule in 'isolation' from other ruleswill hold across all situationsin which
that rule is used. I will characterizethis view as a belief in the separable
characterof rules. I presumethat rulescombinein a configurationalor in-
teractivemanner.If rulescombineconfigurationallyratherthan separably,
this dramaticallyaffects the scientificstrategywe shouldtake in the study
of rules and their effects.
A key exampleof the problemsresultingfrom the view of the separable
characterof rulesis the way theoristshave approachedthe study of PMR
as an aggregationrule. Many scholars,who have studiedPMR, have self-
consciously formulated their models in as general a manner as possible. By
proving a theorem in a general case, it is presumed that the theorem will hold
in all specific cases that contain PMR.
The penchant for generality has been interpreted to mean a formulation
 15

devoid of the specificationof any rule, other than PMR. A set of N in-
dividualssomehowformsa committeeor legislature.Positionrulesarerare-
ly mentioned.The implicitassumptionof most of these modelsis one and
only one positionexists- that of member.No informationis presentedcon-
cerningboundaryrules.We do not knowhow the participantswereselected,
how they will be retained,whetherthey can leave, and how they are re-
placed.The participantscomparepointsin n-dimensionalspaceagainstone
point in the same space called the status quo. We have no idea how that
policy space came into being and what limits there may be on the policies
that could be adopted. (One might presumefrom the way such general
modelsareformulatedthatno constitutionalrulesprotectagainstthe taking
of propertywithout due processor prohibitinginfringementson freedom
of speech.)Authorityrulesareleft unstated.We mustguessat whatactions
individualparticipantsareauthorizedto take. Fromthe waythatthe models
are decribed,it appearsthat any participantcan make any proposalcon-
cerningmovementto any place in policy space. We do not know anything
aboutthe informationrules.Everyoneappearsto be ableto talkto everyone
and providesinformationabout theirpreferencesto everyone.PMR is the
only rule specified.
In this generalcase, in which only a single rule is formulated,theorists
typicallymake specific assumptionsabout preferenceorderings.This sug-
geststhat the conceptsof 'generality'and 'specificity'are used arbitrarily.
Specific assumptions about preference orderings are accepted as ap-
propriatein generalmodels, while efforts to increasethe specificityof the
rules in these same models are criticizedbecausethey are too specific.
The searchfor equilibriahas occurredpredominantlywithinthe context
of such 'general'models. And, in such 'general'models, equilibriaare vir-
tuallynonexistentand are fragileto slightmovementsof preferencesor the
willingness of participantsto dissemble (Riker, 1981). McKelvey and
Ordeshook(1983: 1) are willingto state that 'the principallesson of social
choice theory is that preferenceconfigurationswhich yield majorityun-
dominatedoutcomes are rare and almost always are fragile and thus are
unlikelyto be found in reality.'
If rulescombinein a configurationalmanner,however,theoremsproved
about a 'zero' institutionalarrangementwill not necessarilybe true when
other rules are fully specified. Shepsle and his colleaguesat Washington
Universityhave repeatedlyshownthat when severalotherrulesare overtly
combinedwith PMR, equilibriaoutcomes are more likely. Shepsle and
Weingast(1981)have summarizedthe effects of:

(1) Scoperulesthat operateto limitthe set of outcomesthat can be affectedat a node in

a process,e.g., amendmentcontrolrules(Shepsle,1979a;1979b),'smallchange'rules
(Tullock, 1981), rulesrequiringthe statusquo outcometo be consideredat the last
16

decision node, and rules requiringa committeeproposalto be consideredat the

penultimatedecisionnode.
(2) Authorityrulesthat operateto createand/or limitthe actionsets availableto par-
ticipantsin positions,e.g., rulesthat assigna convenerspecialpowersto orderthe
agenda(McKelvey,1979;Plott and Levine, 1978;Isaacand Plott, 1978), rulesthat
assigna full committee,such as the RulesCommitteein the House of Represen-
tatives,authorityto set the proceduresfor debateand even to excludea bill from
consideration,and rulesthat constrainthe action sets of membersin regardto
strikingpartof a motion, addinga part of a motion, and/or substitutinga part of
a motion (Fiorina,1980).

Structure-inducedequilibriaare present in many situationswhere scope

rules, that limit the outcomesthat can be reached,or authorityrules, that
constrainthe actionsof the participantsin particularpositions,arecombin-
ed withPMR. Thisleadsto an optimisticconclusionthatequilibriaaremore
likely,thanpreviouslyargued,in committeesandassembliesusingmajority
rule to aggregateindividualvotes. This substantiveoptimismis tempered
somewhatwhen one recognizesthe methodologicalconsequencesof rejec-
ting the belief that rules can be studiedas separablephenomena.
The methodologicalproblemrests in the logic of combinatorics.If we
werefortunateenoughto be studyingseparablephenomena,then we could
simplyproceedto studyindividualrulesout of contextas we havedone with
PMR. We could then proceed to study other rules, out of context, and
deriveseparableconclusionsfor eachtypeof rule. Eventually,we couldadd
our resultstogetherto build more complexmodels. This is an appropriate
scientificmethod for the study of separablephenomena.
However,if the way one rule operatesis affectedby otherrules,then we
cannotcontinueto studyeachrulein isolationfromothers.A simple,scien-
tific programis more difficultto envisiononce the configurationalnature
of rulesis accepted.A configurationalapproachaffectsthe way we do com-
parativestatics.Insteadof studyingthe effect of changeof one ruleon out-
comes, regardlessof the other rules in effect, we need to carefullystate
which other rulesare in effect which conditionthe relationshipsproduced
by a change in any particularrule. We cannot just assume that other
variablesare controlledand unchanging.We needto know the valueof the
othervariablesaffectingthe relationshipexaminedin a comparativestatics
framework.
Thus, we havemuchto do! It is morecomfortingto thinkabout proving
theoremsaboutthe effects of usingone particularruleout of contextof the
other rules simultaneouslyin effect. If, however, combinationsof rules
work differently than isolated rules, we had better recognize the type of
phenomena with which we are working and re-adjust our scientific agenda.
We do, however, need a coherent strategy for analyzing and testing the ef-
fects of combinations of rules. How can we isolate a key set of generally for-
 17

mulatedrulesthat providethe core of the rulesto be studied?How can we

build on the resultsof previousanalyticalwork in our field?

5. Multiplelevels of analysisand alternatestrategyof inquiry

I have no final answersto these questions,but I do have an initial stategy

to propose. This strategyrelatesto my earlierstresson the multiplelevels
of analysis involved in the study of rules. We have a relatively well
developedbody of theoryrelatedto the studyof situationssuchas markets,
committees,elections, and gamesin general.Thus, we alreadyknow what
variableswe must identifyto representone level of analysis.We can build
on this knowledgeas we develop the second level of analysis.

5.1 Thestructureof an action situation

The particular form of representationdiffers for neoclassical market
theory, committeestructures,and games in extensiveform. However, in
order to analyze any of these situations, an analyst specifies and relates
togetherseven variablesthat form the structureof a situation.

(1) The set of positions to be held by participants.

(2) The set of participants(includinga randomactor whererelevant)
in each position.
(3) The set of actionsthatparticipantsin positionscantake at different
nodes in a decisiontree.
(4) The set of outcomesthat participantsjointly affect throughtheir
actions.
(5) A set of functionsthat mapparticipantand randomactionsat deci-
sion nodes into intermediateor final outcomes.
(6) The amount of informationavailableat a decisionnode.
(7) The benefits and costs to be assignedto actions and outcomes.
Thesesevenvariablesplus a model of the decisionmakermust be explicitly
stated (or are implicitlyassumed)in orderto constructany formal model
of an interdependentsituation.Wecanconsiderthesesevento be a universal
set of necessaryvariablesfor the constructionof formal decision models
whereoutcomesare dependenton the acts of morethan a singleindividual.
This is a minimalset in that it is not possibleto generatea predictionabout
behaviorin an interdependentsituation without having explicitlyor im-
plicitly specified something about each of these seven variables and related
them together into a coherent structure. I call the analytical entity created
when a theorist specifies these seven variables an action situation.
The most complete and general mathematical structure for representing
18

an actionsituationis a gamein extensiveform(Selten, 1975;Shubik,1982).

The set of instructionsgivento participantsin a well constructedlaboratory
experimentis also a meansof representingan action situation.Using these
variables,the simplestpossible workingmodel of any particulartype of
situation whether a committee, a market, or a hierarchycan be con-
structed.5A change in any of these variablesproducesa different action
situationand may lead to very differentoutcomes. More complexmodels
of committees,markets,or otherinterdependentsituationsare constructed
by addingto the complexityof the variablesused to constructthe simplest
possible situations.6

5.2 An action arena:Models of the situation and the individual

In additionto the sevenuniversalvariablesof an actionsituation,an analyst
must also utilizea modelof the individual,whichspecifieshow individuals
processinformation,how they assignvaluesto actionsand outcomes,how
they selectan action, and what resourcesthey haveavailable.The model of
the individualis the animatingforce that allows the analyst to generate
predictionsabout likelyoutcomesgiventhe structureof the situation(Pop-
per, 1967). When a specificmodel of the individualis addedto the action
situation, I call the resultinganalyticalentity an 'action arena.' An action
arenathus consistsof a modelof the situationand a modelof the individual
in the situation(see E. Ostrom, 1985).
Whena theoristanalyzesan action arena,the model of the situationand
the model of the individualare assumedas givens.At this level of analysis,
the task of the analystis viewedas one of predictingthe type of behavior
and results, given this structure.Questions concerningthe presence or
absenceof retentive,attractive,and/or stableequilibriaand evaluationsof
the efficiencyand equityof these resultsare pursuedat this level. The key
questionat this level is: Giventhe analyticalstructureassumed,how does
this situationwork to produceoutcomes?

5.3 Rules as relations

Letme returnnow to the point I madeabovethatall rulescanbe represented
as relations.I can now be morespecific.Fromsets of physicallypossibleac-
tions, outcomes, decision functions, information,positions, payoffs, and
participants,rulesselectthe feasiblesets of the valuesof thesevariables.The
actionsituationis the intersectionof these feasiblesets. In regardto driving
a car for example, it is physically possible for a 13 year old to drive a car
at 120 miles per hour on a freeway. If one were to model the action situation
of a freeway in a state with well enforced traffic laws, one would posit the
position of licensed drivers traveling an average of 60 to 65 miles per hour
 19

(dependingon the enforcementpatternsof the state). The values of the

variablesin the action situationare constrainedby physicaland behavior
laws, and then, further contained by the rules in use. Most of formal
analyses,to date, are of action situations;this is the surfacestructurethat
our representationsmodel. The rules are part of the underlyingstructure
that shapesthe representationswe use.
But, how do we overtly examinethis part of the underlyingstructure?
Whatrulesshouldbe examinedwhenwe conductanalysisat a deeperlevel?
The approachI recommendis that we focus on those rulesthat can directly
affect the structureof an action situation. This strategyhelps us identify
sevenbroadtypesof rulesthat operateconfigurationallyto affect the struc-
ture of an action situation. These rules include:
(1) Position rules that specify a set of positions and how many par-
ticipantshold each position.
(2) Boundaryrules that specify how participantsare chosen to hold
these positions and how participantsleave these positions.
(3) Scope rules that specify the set of outcomesthat may be affected
andthe externalinducementsand/or costs assignedto eachof these
outcomes.
(4) Authorityrulesthat specifythe set of actionsassignedto a position
at a particularnode.
(5) Aggregationrulesthat specifythe decisionfunctionto be used at a
particularnode to mapactionsinto intermediateor finaloutcomes.
(6) Information rules that authorize channels of communication
amongparticipantsin positionsand specifythe languageand form
in which communicationwill take place.
(7) Payoff rulesprescribehow benefitsand costs are to be distributed
to participantsin positions.
Giventhe wide diversityof rules that are found in everydaylife, social
rulescouldbe classifiedin manyways. ThemethodI am recommendinghas
several advantages. First, rules are tied directly to the variables of an
analyticalentity familiar to all public choice theorists, economists, and
game theorists.From this comes a strategy,or a heuristic,for identifying
the rules affecting the structureof that situation. For each variableiden-
tified in the action situation, the theorist interestedin rules needs to ask
whatrulesproducedthe variableas specifiedin the situation.For example,
in regardto the numberof participants,the ruleanalystwouldbe led to ask:
Why are thereN participants?How did they enter?Underwhatconditions
can they leave? Are there costs, incentives, or penalties associatedwith
entering or exiting? Are some participants forced into entry because of their
residence or occupation?
In regard to the actions that can be taken, the rule analyst would ask: Why
these actions rather than others? Are all participants in positions assigned
20

the same action set? Or, is some convener,or other position, assignedan
action set containingoptions not availableto the remainingparticipants?
Are sets of actions time or path dependent?
In regardto the outcomesthatcanbe affected,the ruleanalystwouldask:
Why these outcomesratherthan others?Are the participantsall principals
who can affect any state variablethey are definedto own? Or, are the par-
ticipantsfiduciarieswho are authorizedto affect particularstate variables
within specified ranges but not beyond? Similarquestions can be asked
about each variableovertlyplaced in a model of an action situation.
Answersto these sets of questionscan thenbe formalizedas a set of rela-
tions that, combinedwith physicaland behaviorallaws, producethe par-
ticularvaluesof the variablesof the situation. I am not arguingthat there
is a uniqueset of relationsthat produceany particularmodelof a situation.
Given the pervasivenessof situations with the structureof a Prisoners'
Dilemma, one can expect that multiplesets of rules may produceaction
situationswiththe samestructure.Thisis not problematicwhenone focuses
exclusivelyon predictingbehaviorwithin the situation. It poses a serious
problemwhen the questionof how to changethat structure.To change a
situation, one must know which set of rules producethe situation.
Otherfactorsalso affect this structure.We know, for example,that rules
which generatea competitivemarketproducerelativelyoptimalequilibria
when used to allocatehomogeneous,divisiblegoods from which potential
consumerscan be excluded.The samerulesgenerateless optimalsituations
when goods are jointly consumedand it is difficult to excludeconsumers.
But the theoristinterestedin how changesin rulesaffects behaviorwithin
situationsmust hold other factors constantwhile an analysisis conducted
of changesin the rules.
Besidesprovidinga generalheuristicfor identifyingthe relevantrulesthat
affect the structureof a situation, a second advantageof this approachis
that it leads to a relativelynaturalclassificationsystem for sets of rules.
Classifyingrulesby whatthey affect enablesus to identifysets of rulesthat
all directly affect the same working part of the situation. This should
enhanceour capabilitiesfor developinga formallanguagefor representing
rules themselves.Specific rules used in everydaylife are named in a non-
theoreticalmanner- frequentlyreferringto the numberof the rulein some
writtenrule book or piece of legislation.Theoristsstudyingrules tend to
name the rulethey are examiningfor some featurerelatedto the particular
type of situationin which the rule occurs.
For systematiccumulationto occur,we needto identifywhenrules,called
by different names, are really the same rule. It is important that scholars
understand, for example, that Romer and Rosenthal and Grether, Isaac,
and Plott all examined consequences of default conditions of aggregation
rules. Proceeding to formalize the rules used by Grether, Isaac, and Plott
 21

in theirseriesof experimentswouldhelpotherscholarsidentifywhichrules,
calledby othereverydayterms,are similarto the 'grandfather'defaultcon-
dition, to the randomdefault condition, or to 'taking from the large and
giving to the small' default condition.
By payingas muchcareto the formalizationof the rulesaffectingan ac-
tion situation as we do to formalizingthe action situation itself, we will
eventuallyestablishrigoroustheoreticalpropositionsconcerningthe com-
pletenessand consistencyof rulesthemselves.From Romerand Rosenthal
and from Grether,Isaac, and Plott, we now knowthat any specificationof
a unanimityrule withoutan explicitdefault conditionis incomplete.I am
willingto speculatethat any aggregationrulewithouta defaultconditionis
incomplete.

6. Some concludingthoughts

Giventhe multiplereferentsfor the term 'institutions,'our first need is for

a consistentlanguageif publicchoicescholarsaregoingto returnto a major
studyof institutions.To begin this task, I have focused on one term- that
of rules- used by some theoristsas a referentfor the terminstitutions.My
effort is intendedto clarifywhat we mean by rules, how rules differ from
physicalor behaviorallaws, how we can classifyrulesin a theoreticallyin-
terestingmanner,and how we can begin to formalizerule configurations.
I have not answeredthe question, 'Whatare institutions?'This involvesan
argumentover whichreferentis 'the' rightor preferredreferent.Rather,I
try to clarify one referentand leave the clarificationof other referentsto
other scholars.
Secondly,I providedseveralexamplesof how publicchoiceanalystshave
studied rules. These examples illustratetwo points. First, rules operate
configurationallyratherthan separably.Second,the studyof rulesinvolves
multiplelevels of analysisratherthan a single level of analysis. The con-
figurationalcharacterof rules significantlyaffects the strategieswe use
to analyzerules. One approachhas been to posit a singlerule and examine
the type of equilibria,or absence of equilibria,likely to result from the
operationof this singlerule. Scholarshave concludedthat stableequilibria
do not existin situationsin whichindividualsuse majorityruleaggregation
procedures.This is not consistent with empiricalobservation. Further,
when scholarsintroducerulesconstrainingactions and outcomesinto ma-
jority rule models, it is then possible to predict stable equilibria. The
methodological consequence of the configurational character of rules is that
theorists need to specify a set of rules, rather than a single rule, when at-
tempting to ask what consequences are produced by changes in a particular
rule.
22

Once this conclusionis accepted, a method to identify sets of rules is

essentialif we hope to developany cumulativeknowledgeabout the effects
of rules. If more than one rule need be specified,the key questionis how
manydifferentrulesmustbe specifiedto knowthatwe haveidentifieda rule
configuration.My preliminaryansweris that weneedto identifyseventypes
of rulesthat directlyaffect the seventypes of variableswe use to construct
most of the action situationswe analyze.Whenwe analyzechangesin one
of these rules, we shouldidentifythe specificsettingof the other variables
that conditionhow the changesin the first rule affects outcomes.
The analysisof rules needs at least two levels. We can representthese
levels by reformulatingPlott's fundamentalequationinto two equations:

Structureof an Model of a = Outcomes (6)

Action Situation Decision Maker
Rules O Physical Behavioral = Structureof an (7)
Laws Laws Action Situation

Equation(6) is the one most public choice theoristsuse in their work. As

we delve somewhatdeeperinto the analysisof rules themselves,previous
workthat has focusedon actionsituationsthemselvescan be integratedinto
a broaderframework.Equation(7) involvesthe specificationof the rules,
as well as the physicaland behaviorallaws, that affect the values of the
variablesin an actionsituation.The seventhequationis the one we mustuse
when we want to analyze how rules change the structureof a situation
leading,in turn, to a changein outcomes(see V. Ostrom, 1982;1984).The
seventhequationmakesapparentthe needto studythe effectsof ruleswhere
physicaland behaviorallaws are invariant.
In light of these characteristics,much futureworkneedsto be done. We
needa formallanguagefor the representationof rulesas functionsaffecting
the variablesin an actionsituation.We also needto addressquestionscon-
cerningthe origin and change of rule configurationsin use. How do in-
dividualsevolve a particularrule configuration?What factors affect the
likelihoodof their following a set of rules?What affects the enforcement
of rules? How is the level of enforcementrelated to rule conformance?
What factors affect the reproducibilityand reliabilityof a rule system?
When is it possible to develop new rules through self-consciouschoice?
And, when are new rules bound to fail?

NOTES

1. Niskanenhad himselfsuggestedthat an importantstructuralchangethat could be made

to improvebureauperformancewas to increasethe competitionbetweenbureaus.
 23

2. See Kiserand E. Ostrom(1982)for a discussionof how rules,goods, and attributesof a

communityall contributeto the structureof a situation.
3. It is surprisinghow often one readsin a publicchoicearticlethatpriormodelshadimplicit
assumptionsthatdrovetheanalysis.A recentexampleis in MackayandWeaver(1978:143)
wherethey arguethat:
Standarddemandside models of the collectivechoice process,in which fiscal out-
comes are consideredrepresentativeof broad-basedcitizen demands, implicitly
assumenot only that a 'democratic'voting rule is employedto aggregatecitizen-
voters'demandsbut also that the agendaformationprocessis characterized by both
free accessand unrestrictedscope.
4. They do not specificallymentionthat they intendto operationalizethe conceptof ger-
maneness,butit wouldappearfromthe 'dicta'thattheythinktheyhavedoneso. However,
as Shepsle(1979a; 1979b)conceptualizedthis rule, decisionsabout one dimensionof a
policyspacewouldbe madesequentially.Oncea decisionabouta particulardimensionhad
beenreached,no furtheractionon that dimensionwouldbe possible.Allowingmembers
to zig-zagall overthe policyspace,one dimensionat a time, is hardlya reasonableopera-
tionalizationof the germanenessrule as specifiedby Shepsle.
5. Thesimplestpossiblerepresentation of a committee,for example,canbe constructedusing
the followingassumptions:
(1) One positionexists;that of member.
(2) Threeparticipantsare members.
(3) The set of outcomesthat can be affectedby the membercontainstwo elements,
one of which is designatedas the statusquo.
(4) A memberis assignedan actionset containingtwo elements:(a) vote for the states
quo and (b) vote for the alternativeoutcome.
(5) If two membersvote for the alternativeoutcome, it is obtained;otherwise,the
statusquo outcomeis obtained.
(6) Payoffsareassignedto eachparticipantdependingon individualactionsandjoint
outcomes.
(7) Completeinformationis availableabout elements(1) through(6).
For this simplestpossiblerepresentationof a committee,and usinga well-definedmodel
of the rationalactor, we know that an equilibriumoutcome exists. Unless two of the
memberspreferthe alternativeoutcometo the statusquo andboth vote, the statusquo is
the equilibriumoutcome.If two membersdo preferandvote for the alternativeoutcome,
it is the equilibriumoutcome.The predictionof outcomsis moreproblematicas soon as
a thirdoutcomeis added.Onlywhenthe valuationpatternsof participantsmeetrestricted
conditionscan an equilibriumoutcomebe predictedfor sucha simplecommitteesituation
withthreemembersandthreepotentialoutcomesusingmajorityrule(Arrow,1966;Plott,
1967).
6. A morecomplexcommitteesituationis created,for example,if a secondposition, thatof
a convener,is addedto the situation,and the action set of the convenerincludesactions
not availableto the othermembers(e.g., Isaacand Plott, 1978;Eaveyand Miller,1982).
See also Gardner(1983)for an analysisof purgesof recruitmentto committees.Gardner's
approachis very similarto the generalstrategyI am recommending.

REFERENCES

AdvisoryCommissionon Intergovernmental Relations(1974). Governmental functions and

processes:Localand areawide.VolumeIV of substateregionalismand thefederal system.
Washington,D.C.: U.S. GovernmentPrintingOffice.
Arrow,K. (1966). Social choice and individualvalues,2nd edition. New York:Wiley.
24

Becker,G.S. (1976). Theeconomicapproachto humanbehavior.Chicago:The Universityof

ChicagoPress.
Buchanan,J.M., and Tullock,G. (1962). Thecalculusof consent.Ann Arbor:Universityof
MichiganPress.
Commons,J.R. (1957).Legalfoundationsof capitalism.Madison:Universityof Wisconsin
Press.
Dahl, R.A., andTufte, E.R. (1973).Sizeanddemocracy.Stanford,Calif.:StanfordUniversi-
ty Press.
Dawson, R.E., and Robinson,J.A. (1963). Interpartycompetition,economicvariablesand
welfarepoliciesin the americanstates. Journalof Politics 25: 265-289.
Downs, A. (1957).An economictheoryof democracy.New York:Harperand Row.
Dye, T.R. (1966).Politics, economics,and the public. Chicago:RandMcNally.
Eavey,C.L., andMiller,G.J. (1982).Committeeleadershipand thechairman'spower.Paper
deliveredat the AnnualMeetingsof the AmericanPoliticalScienceAssociation,Denver,
Colo., September2-5.
Fiorina, M.P. (1980)). Legislativefacilitation of governmentgrowth: Universalismand
reciprocitypracticesin majorityrule institutions.Researchin PublicPolicy Analysisand
Management1: 197-221.
Ganz, J.S. (1971).Rules:A systematicstudy. The Hague:Mouton.
Gardner,R. (1983).Variationof theelectorate:Vetoandpurge.PublicChoice40(3):237-247.
Grether,D.M., Isaac,R.M., andPlott, C.R. (1979).Alternativemethodsof allocatingairport
slots: Performanceand evaluation.A reportpreparedfor the Civil AeronauticsBoard.
Harr6, R. (1974). Some remarkson 'rule' as a scientific concept. In T. Mischel (Ed.),
Understandingotherpersons, 143-183. Oxford, England:Basil-Blackwell.
Hayek, F.A. (1976). Themirageof socialjustice. Chicago:Universityof ChicagoPress.
Hayek,F.A. (1978).New studiesin philosophy,politics, economics,and the historyof ideas.
Chicago:Universityof ChicagoPress.
Heiner,R.A. (1983).The originof predictablebehavior.AmericanEconomicReview83(4):
560-597.
Isaac,R.M., and Plott, C.R. (1978).Comparativegamemodelsof the influenceof the closed
rulein threeperson,majorityrulecommittees:Theoryand experiment.In P.C. Ordeshook
(Ed.), Gametheoryandpoliticalscience,283-322. New York:New YorkUniversityPress.
Kiser,L., and Ostrom,E. (1982).The threeworldsof action:A meta-theoretical synthesisof
institutionalapproaches.In E. Ostrom (Ed.), Strategiesof political inquiry, 179-222.
BeverlyHills: Sage Publications.
Mackay,R.J., and Weaver,C. (1979). Monopolybureausand fiscal outcomes:Deductive
modelsandimplicationsfor reform.In G. TullockandR.E. Wagner(Eds.),Policyanalysis
and deductivereasoning,141-165. Lexington,Mass.: LexingtonBooks.
McGuire,T., Coiner,M., andSpancake,L. (1979).Budgetmaximizingagenciesandefficiency
in government.Public Choice34(3/4): 333-359.
McKelvey,R.D. (1979). Generalconditions for global intransitivities.Econometrica47:
1,085-1,111.
McKelvey,R.D., and Ordeshook,P.C. (1984).An experimentalstudyof the effects of pro-
ceduralruleson committeebehavior.Journalof Politics 46(1) 185-205.
Menger,K. (1963).Problemsin economicsand sociology. (Originallypublishedin 1883and
translatedby FrancisJ. Nock.) Urbana:Universityof IllinoisPress.
Niskanen, W.A. (1971). Bureaucracyand representativegovernment.Chicago: Aldine-
Atherton.
Nozick, R. (1975).Anarchy,state, and utopia. New York:Basic Books.
Ostrom,E. (1985).A methodof institutionalanalysis.In F.X. Kaufmann,G. Majoneand V.
Ostrom(Eds.), Guidance,control,andperformanceevaluationin thepublicsector.Berlin,
New York:de Gruyter,forthcoming.
 25

Ostrom,V. (1980).Artisanshipand artifact.PublicAdministrationReview40(4):309-317.

Ostrom,V. (1982).A forgottentradition:Theconstitutionallevelof analysis.In J.A. Gillespie
and D.A. Zinnes(Eds.), Missingelementsin politicalinquiry:Logic and levelsof analysis,
237-252. BeverlyHills: Sage Publications.
Ostrom,V. (1985).Constitutionalconsiderationswithparticularreferenceto federalsystems.
In F.X. Kaufmann,G. Majoneand V. Ostrom(Eds.), Guidance,control,andperformance
evaluationin thepublic sector. Berlin,New York:de Gruyter,forthcoming.
Parks,R.B., andOstrom,E. (1981).Complexmodelsof urbanservicesystems.In T.N. Clark
(Ed.), Urbanpolicy analysis:Directionsforfutureresearch.UrbanAffaireAnnualReviews
21: 171-199. BeverlyHills: Sage Publications.
Plott, C.R. (1967).A notion of equilibriumand its possibilityundermajorityrule.American
EconomicReview57(4):787-807.
Plott, C.R. (1979).The applicationof laboratoryexperimentalmethodsto publicchoice. In
C.S. Russell(Ed.), Collectivedecision making:Applicationsfrom public choice theory,
137-160. Baltimore,Md.: Johns HopkinsUniversityPress.
Plott, C.R., and Levine,M.E. (1978).A modelfor agendainfluenceon committeedecisions.
AmericanEconomicReview68: 146-160.
Popper,K.R. (1967). La rationalit6et le statutdu principlede rationalit6.In E. M. Classen
(Ed.), Lesfoundementsphilosophiques dessystemeseconomiques:Textesde JacquesRueff
et essais redigdsen son honneur23 acut 1966, 145-50. Paris, France:Payot.
Rawls,J. (1968).Two conceptsof rules.In N.S. Careand C. Landesman(Eds.), Readingsin
the theoryof action, 306-340. Bloomington,Ind.: IndianaUniversityPress. Originally
printedin the PhilosophicalReview4(1955).
Riker,W.H. (1982).Implicationsfromthe disequilibrium of majorityrulefor the studyof in-
stitutions.In P.C. OrdeshookandK.A. Shepsle(Eds.),Politicalequilibrium,3-24. Boston:
Kluwer-Nijhoff.Originallypublishedin the AmericanPolitical ScienceReview 74(June
1980):432-447.
Romer,T., and Rosenthal,H. (1978). Politicalresourceallocation,controlledagendas,and
the statusquo. Public Choice33(4):27-43.
Schotter,A. (1981). Theeconomictheoryof social institutions.Cambridge,England:Cam-
bridgeUniversityPress.
Selten,R. (1975).Reexamination of the perfectnessconceptfor equilibriumpointsin extensive
games. InternationalJournalof Game Theory4: 25-55.
Shepsle,K.A. (1979a).Institutionalarrangements andequilibriumin multidimensional voting
models.AmericanJournalof Political Science,23(1):27-59.
Shepsle,K.A. (1979b).Theroleof institutionalstructurein the creationof policyequilibrium.
InD.W. RaeandT.J. Eismeier(Eds.),Publicpolicyandpublicchoice,249-283. SageYear-
books in Politicsand Public Policy 6. BeverlyHills: Sage Publications.
Shepsle, K.A., and Weingast, B.R. (1981). Structure-induced equilibriumand legislative
choice. Public Choice37(3):503-520.
Shimanoff, S.B. (1980). Communicationrules: Theoryand research.BeverlyHills: Sage
Publications.
Shubik,M. (1982). Gametheoryin the social sciences:Conceptsand solutions. Cambridge,
Mass.: The MIT Press.
Taylor,M. (1976).Anarchyand cooperation.New York:Wiley.
Toulmin,S. (197).Rulesandtheirrelevancefor understanding humanbehavior.In T. Mischel
(Ed.), Understandingotherpersons 185-215. Oxford, England:Basil-Blackwell.
Tullock, G. (1981). Why so much stability?Public Choice37(2): 189-205.
Ullman-Margalit, E. (1978). Theemergenceof norms.New York:OxfordUniversityPress.
von Wright,G.H. (1968).Thelogicof practicaldiscourse.In RaymondKlibansky(Ed.), Con-
temporaryphilosophy, 141-167. Italy:La Nuava ItaliaEditrice.
Wilson, R., and Herzberg,R. (1984). Votingis only a block away:Theoryand experiments
on blockingcoalitions.Paper presentedat the Public Choice Societymeetings,Phoenix,
Ariz., March29-31.