A Critique of the Ruling Elite Model

Author(s): Robert A. Dahl

Source: The American Political Science Review, Vol. 52, No. 2 (Jun., 1958), pp. 463-469
Published by: American Political Science Association
Stable URL: http://www.jstor.org/stable/1952327 .
Accessed: 20/11/2014 08:07

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.

American Political Science Association is collaborating with JSTOR to digitize, preserve and extend access to
The American Political Science Review.

http://www.jstor.org

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions
 A CRITIQUE OF THE RULING ELITE MODEL
ROBERT A. DAHL
Yale University

A greatmanypeople seem to believethat "they" runthings:the old families,

the bankers,the City Hall, machine,or the party boss behind the scene. This
kind of view evidentlyhas a powerfuland many-sidedappeal. It is simple,
compelling,dramatic, "realistic." It gives one standingas an inside-dopester.
For individualswith a strongstrainof frustratedidealism,it has just the right
touch of hard-boiled cynicism.Finally, the hypothesishas one very great
advantage over many alternativeexplanations:It can be cast in a formthat
makes it virtuallyimpossibleto disprove.
Considerthe last point fora moment.There is a type of quasi-metaphysical
theorymade up of what might be called an infiniteregressof explanations.
The rulingelite model can be interpretedin this way. If the overtleaders of a
communitydo not appear to constitutea rulingelite, then the theorycan be
saved by arguingthat behind the overtleaders thereis a set of covertleaders
who do. If subsequentevidence shows that this covertgroup does not make a
rulingelite, then the theory can be saved by arguing that behind the first
covertgroupthereis another,and so on.
Now whatever else it may be, a theorythat cannot even in principle be
controvertedby empiricalevidence is not a scientifictheory.The least that
we can demand of any rulingelite theory that purportsto be more than a
metaphysicalor polemicaldoctrineis, first,that the burdenof proofbe on the
proponentsof the theoryand not on its critics; and, second, that there be
clear criteriaaccordingto whichthe theorycould be disproved.
With these points in mind, I shall proceed in two stages. First, I shall try
to clarifythe meaningofthe concept"rulingelite" by describinga verysimple
formof what I conceive to be a rulingelite system.Second, I shall indicate
what would be requiredin principleas a simple but satisfactorytest of any
hypothesisassertingthat a particularpolitical systemis, in fact,a rulingelite
system.Finally, I shall deal with some objections.

I. A SIMPLE RULING ELITE SYSTEM

If a rulingelite hypothesissays anything,surelyit assertsthat withinsome

specificpolitical system there exists a group of people who to some degree
exercisepower or influenceover otheractors in the system.I shall make the
followingassumptionsabout power:'
1. In orderto compare the relativeinfluenceof two actors (these may be
individuals,groups,classes, parties,or what not), it is necessaryto state the
scope of the responsesupon which the actors have an effect.The statement,
I See RobertA. Dahl, "The ConceptofPower,"BehavioralScience,Vol. 2 (July1957),
pp. 201-215.

463

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions
464 THE AMERICAN POLITICAL SCIENCE REVIEW

"A has morepowerthanB," is so ambiguousas to vergeon themeaningless,since

it does not specifythe scope.
2. One cannot comparethe relativeinfluenceof two actors who always per-
formidenticalactionswithrespectto thegroupinfluenced.What thismeans as a
practicalmatteris that ordinarilyone can test fordifferences in influenceonly
wherethereare cases of differences in initial preferences.At one extreme,the
difference may mean that one group prefersalternativeA and anothergroup
prefersB, A and B beingmutuallyexclusive.At the otherextreme,it may mean
that one group prefersalternativeA to otheralternatives,and anothergroup
is indifferent. If a political systemdisplayed completeconsensusat all times,
we should findit impossibleto constructa satisfactorydirecttest of the hypo-
thesis that it was a rulingelite system,although indirectand ratherunsatis-
factorytests mightbe devised.
Consequently,to knowwhetheror not we have a rulingelite,we must have
a political systemin which there is a difference in preferences,fromtime to
time, among the individual human beings in the system.Suppose, now, that
among theseindividualsthereis a set whosepreferences regularlyprevailin all
cases of disagreement,or at least in all cases of disagreementover key political
issues (a term I propose to leave undefinedhere). Let me call such a set of
individualsa "controllinggroup."In a full-fledged democracyoperatingstrictly
accordingto majorityrule,the majoritywould constitutea controllinggroup,
even though the individual membersof the majoritymightchange fromone
issue to the next. But since our model is to representa rulingelite system,we
requirethat the set be less thana majorityin size.
However, in any representativesystemwith singlemembervotingdistricts
where more than two candidates receive votes, a candidate could win with
less than a majorityof votes; and it is possible,therefore,to imagine a truly
sovereignlegislatureelected under the strictest"democratic" rules that was
nonethelessgovernedby a legislativemajorityrepresenting the firstpreferences
ofa minorityofvoters.Yet I do not thinkwe wouldwant to call such a political
systema rulingelite system.Because of this kind of difficulty, I propose that
we excludefromour definitionof a rulingelite any controllinggroup that is a
productof rulesthat are actuallyfollowed(that is, "real" rules) underwhicha
majorityof individualscould dominateif theytook certainactionspermissible
underthe "real" rules.In short,to constitutea rulingelite a controllinggroup
must not be a pure artifactof democratic rules.
A rulingelite,then,is a controllinggroupless than a majorityin size that is
not a pure artifactof democraticrules. It is a minorityof individualswhose
preferencesregularlyprevail in cases of differencesin preferenceon key
political issues. If we are to avoid an infiniteregressof explanations,the
compositionof the rulingelite must be moreor less definitelyspecified.
II. SOME BAD TESTS

The hypothesiswe are d.1ing-with wquld run aloig these 1ines; "Such and
such a politicalsystem(the U. S., the U.S.S.R., New Haven, or the like) is a

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions
 CRITIQUE OF THE RULING ELITE MODEL 465

rulingelite systemin which the rulingelite has the followingmembership."

Membershipwould then be specifiedby name, position,socio-economicclass,
socio-economicroles, or what not.
Let me now turn to the problem of testinga hypothesisof this sort, and
begin by indicatinga few tests that are sometimesmistakenlytaken as ade-
quate.
The firstimpropertest confusesa rulingelite with a group that has a high
potentialforcontrol.Let me explain. Suppose a set of individualsin a political
systemhas the followingproperty:thereis a veryhighprobabilitythat if they
agree on a key political alternative,and if they all act in some specifiedway,
thenthat alternativewill be chosen.We may say of such a groupthat it has a
highpotentialforcontrol.In a largeand complexsocietylike ours,theremay be
many such groups. For example, the bureaucratictriumvirateof Professor
Mills would appear to have a high potential forcontrol.2In the City of New
Haven, withwhichI have some acquaintance, I do not doubt that the leading
businessfigurestogetherwith the leaders of both political parties have a high
potentialforcontrol.But a potential forcontrolis not, except in a peculiarly
Hobbesian world,equivalent to actual control.If the militaryleaders of this
countryand theirsubordinatesagreed that it was desirable,they could most
assuredlyestablish a militarydictatorshipof the most overt sort; nor would
theyneed the aid ofleaders of businesscorporationsor the executivebranchof
our government.But they have not set up such a dictatorship.For what is
lackingare the premisesI mentionedearlier,namelyagreementon a key politi-
cal alternativeand some set of specificimplementingactions. That is to say,
a group may have a high potential for controland a low potentialfor unity.
of a groupis a functionofits potentialforcon-
The actual politicaleffectiveness
trol and its potentialforunity. Thus a group with a relativelylow potential
forcontrolbut a highpotentialforunitymay be morepoliticallyeffectivethan
a groupwith a highpotentialforcontrolbut a low potentialforunity.
The second impropertest confusesa rulingelite with a group of individuals
who have more influencethan any othersin the system.I take it forgranted
that in everyhuman organizationsome individualshave more influenceover
key decisionsthan do others.Political equality may well be among the most
Utopian of all human goals. But it is fallaciousto assume that the absence of
politicalequality provesthe existenceof a rulingelite.
The thirdimpropertest,whichis closelyrelated to the precedingone, is to
generalizefroma single scope of influence.Neither logically nor empirically
does it followthat a groupwith a high degreeof influenceover one scope will
necessarilyhave a high degreeof influenceover anotherscope withinthe same
system.This is a matterto be determinedempirically.Any investigationthat
does not take into account the possibilitythat different elite groupshave dif-
ferentscopes is suspect. By means of sloppyquestionsone could easilyseem to
discoverthat thereexists a unifiedrulingelite in New Haven; forthereis no

2 C. Wright Mills,, The Power Elite (New York, 1956), pa-sim.

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions
466 THE AMERICAN POLITICAL SCIENCE REVIEW

doubt that small groupsof people make many key decisions.It appears to be
the case, however,that the small group that runsurban redevelopmentis not
the same as the small group that runs public education, and neitheris quite
the same as the two small groupsthat run the two parties.Moreoverthe small
groupthat runsurban redevelopmentwith a highdegreeof unitywould almost
certainlydisintegrateif its activitieswere extendedto eithereducation or the
two political parties.
III. A PROPOSED TEST

If testslike these are not valid, what can we properlyrequire?

Let us take the simplestpossible situation. Assume that there have been
some number-I will not say how many-of cases where therehas been dis-
agreementwithinthe political systemon key political choices.Assume further
that the hypotheticalrulingeliteprefersone alternativeand otheractorsin the
systempreferotheralternatives.Then unlessit is truethat in all or verynearly
all ofthesecases the alternativepreferred by the rulingeliteis actuallyadopted,
the hypothesis(that the systemis dominatedby the specifiedrulingelite) is
clearlyfalse.
I do not want to pretendeitherthat the researchnecessaryto such a test is
at all easy to carryout or that communitylifelends itselfconvenientlyto strict
interpretationaccordingto the requirementsof the test. But I do not see how
anyonecan supposethathe has establishedthedominanceof a specificgroupin a
community or a nationwithoutbasinghis analysis on thecarefulexaminationofa
seriesof concretedecisions.And these decisionsmust eitherconstitutethe uni-
verse or a fairsample fromthe universeof key politicaldecisionstaken in the
political system.
Now it is a remarkableand indeed astoundingfact that neitherProfessor
Mills nor ProfessorHunter has seriouslyattempted to examine an array of
specificcases to testhis major hypothesis.3 Yet I suppose thesetwo worksmore
than any othersin the social sciencesof the last fewyears have soughtto in-
terpretcomplexpoliticalsystemsessentiallyas instancesof a rulingelite.
To sum up: The hypothesisof the existenceof a rulingelite can be strictly
tested only if:
1. The hypotheticalrulingelite is a well-definedgroup.
2. There is a fair sample of cases involvingkey political decisionsin which
the preferencesof the hypotheticalrulingelite run counterto those of
any otherlikelygroupthat mightbe suggested.
3. In such cases, the preferencesof the elite regularlyprevail.

IV. DIFFICULTIES AND OBJECTIONS

Several objectionsmightbe raised against the test I propose.

First, one mightargue that the test is tooweak.The argumentwould run as
follows:If a rulingelite doesn'texistin a community,then the test is satisfac-

3 Mills, PowerStructure(Chapel Hill, 1953).

op. cit.; Floyd Hunter, Community

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions
 CRITIQUE OF THE RULING ELITE MODEL 467

tory; that is, if every hypotheticalrulingelite is compared with alternative

controlgroups,and in fact no rulingelite exists,thenthe test will indeed show
that thereis no minoritywhose preferencesregularlyprevail on key political
alternatives.But-it mightbe said-suppose a rulingelite does exist. The test
will not necessarilydemonstrateits existence,since we may not have selected
the rightgroup as our hypotheticalrulingelite. Now this objection is valid;
but it suggeststhe point I made at the outset about the possibilityof an in-
finiteregressof explanations.Unless we use the test on everypossiblecombina-
tion of individualsin the community,we cannot be certainthat thereis not
some combinationthat constitutesa rulingelite. But since there is no more
a priorireason to assume that a rulingelite does existthan to assume that one
does not exist,the burdenofproofdoes not restupon the criticofthehypothesis,
but upon its proponent.And a proponentmust specifywhat group he has in
mind as his rulingelite. Once the groupis specified,then the test I have sug-
gested is, at least in principle,valid.
Second,one could object that the testis toostrong.For supposethat themem-
bers of the "ruled" groupare indifferent as to the outcome of various political
alternatives.Surely (one could argue) if thereis anothergroup that regularly
gets its way in the face of this indifference,it is in fact the rulinggroupin the
society.Now my reasons forwishingto discriminatethis case fromthe other
involvemorethan a mere question of the proprietyof using the term "ruling
elite," which is only a term of convenience.There is, I think,a difference of
some theoreticalsignificancebetween a systemin which a small group dorni-
nates over anotherthat is opposed to it, and one in which a group dominates
overan indifferent mass. In the secondcase, the alternativesat stake can hardly
be regardedas "key political issues" if we assume the point of view of the in-
different mass; whereasin the firstcase it is reasonableto say that the alterna-
tives involvea key politicalissue fromthe standpointof both groups.Earlier I
refrainedfromdefiningthe concept "key politicalissues." If we were to do so
at this point, it would seem reasonable to require as a necessary although
possibly not a sufficientconditionthat the issue should involve actual dis-
agreementin preferencesamong two or more groups. In short,the case of
"indifference vs. preference"would be ruled out.
However, I do not mean to dispose of the problem simply by definition.
The point is to make sure that the two systemsare distinguished.The test
forthe second,weakersystemofelite rulewould thenbe merelya modification
of the test proposed forthe firstand more stringentcase. It would again re-
quire an examinationof a series of cases showinguniformlythat when "the
word" was authoritativelypassed down fromthe designatedelite,the hitherto
indifferent majorityfell into ready compliancewith an alternativethat had
nothingelse to recommendit intrinsically.
Third, one mightargue that the test will not discriminatebetween a true
rulingelite and a rulingelite togetherwith its satellites.This objection is in
one sense trueand in one sense false. It is true that on a seriesof key political
questions,an apparentlyunifiedgroupmightprevail who would, accordingto

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions
468 THE AMERICAN POLITICAL SCIENCE REVIEW

our test, therebyconstitutea rulingelite. Yet an inner core might actually

make the decisionsforthe whole group.
However,one oftwo possibilitiesmust be true. Eitherthe innercore and the
frontmen always agree at all timesin the decisionprocess,or theydo not. But
if they always agree, then it followsfromone of our two assumptionsabout
influencethat the distinctionbetween an "inner core" and "frontmen" has
no operational meaning; that is, there is no conceivable way to distinguish
betweenthem. And if they do not always agree, then the test simplyrequires
a comparisonat those points in time when they disagree.Here again, the ad-
vantages of concretecases are palpable, forthese enable one to discoverwho
initiatesor vetoes and who merelycomplies.
Fourth, it mightbe said that the test is eithertoo demandingor else it is
too arbitrary.If it requiresthat the hypotheticalelite prevails in everysingle
case, then it demands too much. But if it does not requirethis much, then at
what point can a rulingelite be said to exist?When it prevailsin 7 cases out of
10? 8 out of 10? 9 out of 10? Or what? There are two answersto this objection.
On the one hand, it would be quite reasonableto argue,I think,that since we
are consideringonly key political choices and not trivialdecisions,if the elite
does not prevail in everycase in which it disagrees with a contrarygroup,
it cannot properlybe called a rulingelite. But since I have not supplied an
independentdefinitionof the term "key political choices," I must admit that
this answeris not whollysatisfactory.On the otherhand, I would be inclined
to suggestthat in this instanceas in many otherswe oughtnot to assume that
political realitywill be as discreteand discontinuousas the conceptswe find
convenientto employ. We can say that a systemapproximatesa true ruling
elite system,to a greateror lesser degree,withoutinsistingthat it exemplify
the extremeand limitingcase.
Fifth,it mightbe objected that the test I have proposedwould not workin
the most obvious of all cases of rulingelites, namely in the totalitariandic-
tatorships.For the controlof the elite over the expressionof opinion is so
great that overtlythereis no disagreement;hence no cases on whichto base a
judgmentarise. This objectionis a fairone. But we are not concernedherewith
totalitariansystems.We are concernedwith the application of the techniques
of moderninvestigationto Americancommunities,where,except in very rare
cases, terroris not so pervasivethat the investigatoris barredfromdiscovering
the preferencesof citizens.Even in Little Rock, forexample, newspapermen
seemed to have had littledifficulty in findingdiverseopinions;and a northern
politicalscientistofmy acquaintance has managed to completea large number
of productive interviewswith White and Negro Southernerson the touchy
subject of integration.
Finally one could argue that even in a societylike ours a rulingelite might
be so influentialover ideas, attitudes,and opinionsthat a kind of false con-
sensuswill exist-not the phonyconsensusofa terroristic totalitariandictator-
ship but the manipulated and superficiallyself-imposedadherence to the
normsand goals of the elite by broad sectionsof a community.A good deal of

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions
 CRITIQUE OF THE RULING ELITE MODEL 469

ProfessorMills' argumentcan be interpretedin this way, although it is not

clear to me whetherthis is what he means to resthis case on.
Even morethan the othersthisobjectionpointsto the need to be circumspect
in interpreting the evidence. Yet here,too, it seems to me that the hypothesis
cannot be satisfactorilyconfirmedwithoutsomethingequivalent to the test I
have proposed. For once again either the consensus is perpetual and un-
breakable, in which case there is no conceivable way of determiningwho is
rulerand who is ruled. Or it is not. But ifit is not, thenthereis some point in
the processof formingopinionsat whichthe one groupwill be seen to initiate
and veto, whilethe restmerelyrespond.And we can onlydiscoverthese points
by an examinationof a series of concretecases wherekey decisionsare made:
decisions on taxation and expenditures,subsidies,welfareprograms,military
policy, and so on.
It would be interestingto know, for example, whetherthe initiationand
veto of alternativeshaving to do with our missile programwould confirm
ProfessorMills' hypothesis,or indeed any reasonable hypothesisabout the
existenceof a rulingelite. To the superficialobserverit would scarcelyappear
that the militaryitselfis a homogeneousgroup,to say nothingof their sup-
posed coalitionwith corporateand political executives. If the militaryalone
or the coalitiontogetheris a rulingelite, it is eitherincrediblyincompetentin
administeringits own fundamentalaffairsor else it is unconcernedwith the
success of its policies to a degreethat I findastounding.
However I do not mean to examine the evidence here. For the whole point
of this paper is that the evidencefora rulingelite,eitherin the United States
or in any specificcommunity,has not yet been properlyexaminedso far as I
know.And the evidencehas not been properlyexamined,I have triedto argue,
because the examinationhas not employed satisfactorycriteriato determine
what constitutesa fairtest of the basic hypothesis.

This content downloaded from 157.182.150.22 on Thu, 20 Nov 2014 08:07:08 AM

All use subject to JSTOR Terms and Conditions