Interest Groups and the Size of Government
Author(s): Dennis C. Mueller and Peter Murrell
Reviewed work(s):
Source: Public Choice, Vol. 48, No. 2 (1986), pp. 125-145
Published by: Springer
Stable URL: http://www.jstor.org/stable/30024587 .
Accessed: 28/10/2011 18:46
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 [email protected].
Springer is collaborating with JSTOR to digitize, preserve and extend access to Public Choice.
http://www.jstor.org
Public Choice48: 125-145(1986).
c 1986MartinusNijhoff Publishers,Dordrecht.Printedin the Netherlands.
Interestgroups and the size of government*
DENNISC. MUELLER
PETERMURRELL
Departmentof Economics,Universityof Maryland,CollegePark, MD 20742
1. Introduction
The size of government, both absolutely and as a percentage of gross national product, has in the last decade reached unprecedently high levels in
all Western countries (Nutter, 1978). Although in most cases growth in
government began long before World War II, it is only in recent years that
the level of government activity has reached such proportions as to cause
widespread concern and discussion in the political arena and in academia.
In the economics literature, this concern has led to an increasing interest in
positive analysis of the size of government (Borcherding, 1977; Brunner,
1978; Frey, 1982; Meltzer and Richard, 1978, 1981; Peltzman, 1980; Fratianni and Spinelli, 1982). The present paper is a contribution to that
analysis.
Among the many factors explaining the size of government, mention is
often made of the potential role of interest groups. Yet, surprisingly little
has been done to develop and test hypotheses concerning the impact of interest groups on government size (but see McCormick and Tollison, 1981).
This paper begins to remedy this deficiency. In Section 2, we discuss the impact of interest groups on government size. The hypotheses to be tested are
formulated in Section 3. Section 4 presents single equation estimates aimed
at testing these hypotheses. In Section 5, we embody interest group activity
and voting behavior in a rudimentary simultaneous equations model determining the size of government and we estimate that model. Conclusions are
drawn in Section 6.
* Financial support for the work in this paper was provided by the Sloan Foundation grant
to the University of Maryland to support a workshop in Public and Urban Economics, and
the International Institute of Management/Industrial Policy in Berlin. Extremely helpful comments on an earlier draft were obtained from Mark Pauly, Joe Oppenheimer, and Robert
Tollison.
126
2. Interestgroupsin the politicalprocess
We envisagea politicalprocessin whichthe government,the executivethat
is, is formedby the winningmajoritycoalitionin parliament.Whena single
partywins a majorityof the seatsin parliamentin an election,it forms the
governmentitself. Whenno partysecuresa majority,a coalitionof parties
with a majorityof seats must come togetherto form a government.This
type of systemcharacterizesmost democraciesin the worldtoday, the most
importantexceptionbeing the U.S.A.
Interestgroupsattemptto win favors for their membershipby offering
to supplya partywithvotes. An interestgroupmayendorsea party,supply
campaignvolunteers,or contributefunds to the party'scampaign.Eachof
these translatesinto votes whichthe interestgroupattemptsto 'trade'with
a given partyin exchangefor a promisedfavor shouldthe partysucceedin
becomingthe government,or a part thereof.
Someof the favorsinterestgroupsseek, suchas a quotato protecta given
industry,do not havelarge,directimpactson governmentsize. Others,like
a depletiontax allowancefor a particularindustry,may actuallyreducetax
revenues.However,many programslike urbanmass transitsubsidies,job
retrainingand the constructionof damsand otherpublicworksinvolveexpanded governmentactivity which directly benefits given economic or
geographic interest groups. Such activities will be introduced into the
governmentbudget when the benefits to interestgroups can be targeted
moreefficientlythroughtheseprogramsthanby meansof 'costless'regulations or tax subsidies.'We hypothesizethat on averagethe favors sought
by interestgroups from governmentrequirean expansionof tax revenues
and expenditures.
The supplyof legislationto specificinterestgroupscomesabout as parties
attemptto maximizetheirexpectedvotes and win elections(Downs, 1957).
A competition for interest group support among parties is assumed in
which, at least in the early phases, each-partyis inducedto increasethe
numberof interestgroups supportingit in responseto an increasein the
numberof interestgroupssupportingits opponents.Whetherthis competition leadsto the absorptionof all interestgroupsinto the list of supporters
of one or the otherpartiescannotbe deducedwithouta moreformalmodeling of politicalcompetition.Whatseemsquiteintuitive,however,is thatthe
numberof interestgroupsabsorbedinto the politicalprocessis an increasing functionof the numberof interestgroupsexistingin the polity. Thus,
the effective demandfor governmentprogramsfavoringinterestgroupsis
greater, the greater the number of interest groups in society.
Competition for interest groups takes place prior to an election. After the
election one or more of the parties controlling a majority of the seats of
parliament form a government. This party or coalition of parties governs
127
untilthe next election.Duringthis periodthe party(ies)in the majoritycontrol both the executiveand the parliament,and effectuate the bargains
struckwith its (their)interestgroup supportersduringthe election.
The nextelectionbringsa newcompetitionfor interestgroups,most likely some reshufflingof interestgroupsupportamongthe parties,perhapsa
new government.Almost certainlythe set of interestgroupsrepresentedin
the newgovernmentwill not be identicalto those in the previousone. Some
of these may have felt the previouslevel of governmentexpendituresexcessive. Thus, one cannot predict that this new set of interest groups
representedin parliament favors an increase in expendituresover the
previouslevel. Onecan predicta levelof governmentoutlaysthat is greater,
the greaterthe numberof interestgroupsin the society. Our theoryof interestgroup politics predictsexcessivelevels of governmentexpenditures,
not necessarilyexcessivegrowth in the size of government.
In the foregoingdiscussion,the number,size and othercharacteristicsof
interestgroupsare treatedas exogenous.The hypothesisproposedherecan
be used to predictgrowinggovernmentsize in an era in whichthe number
of organizedinterestgroups grows, since new interestgroups bring with
themdemandsfor publiclyfundedgoods. But a full, dynamicmodelingof
the growthof interestgroupsand governmentis beyondthe scope of this
paper. We do allow for the possibleendogeneityof interestgroupsin our
empiricalwork, however(see Section 5).
3. The size of governmentequation
The basic hypothesisemergingfrom the precedingdiscussionis that the
relativesize of governmentis positivelyrelatedto the numberof organized
interestgroups.In testingthis hypothesis,we shallincludein our equations
severaladditionalvariablesgleanedfromthe publicchoiceliterature.Wedo
so not in the pretensethat we aretestingthesealternativemodelsof government againstour own, but underthe assumptionthat the impactof these
otherfactorsis additive.We thus makethe strongpredictionthat the effect
of interestgroupson sizeof governmentremainsas predictedin the presence
of additionalinstitutionalcomplexity,and beginto test this assumptionby
addingseveralof the variableswhichotherstudieshavepositedto be determinantsof the size of government.
3.1 The basic equation
We seek to explain the relative size of government across countries. No one
measure may fully capture the concept of governmental size. We shall,
therefore, employ several alternative dependent variables that measure
128
5
by
a
in
definition
Variable
level
10 population
and
divided
x
income
parliament
the country
the
of
of
each
political
election
in income
government
GDP GDP
for
GDP of
the
of
capita
members members general
central
listed
a
per group
began
household
theread
households
in
GDP.
at to 0.
percentage
of
U.S.selectedgroupsselected
of a
=
9
percentage
of
a
as
voting country
able
average
a
as
of
quintile
collected
interest party
ethnolinguistic
randomly randomly
processes
Variable
which
countries
of
percentage
by in
percentage
a
twosame
twosame
middle
population
population
a
work
revenues
as consumption
other
government as thatthe
thatthe theproportion
years
number
1;
a
adult
taxadult
millions
of
of
final
the of of =
the
in GDP from
from as of multiplied
revenue
5 of modernization
be of reference
be
income
outlays
tax
probability
probability
not
not
capita
count
Population
Percentage
Variable
Average
Percentage
Percentage
Switzerland
Total
Government
PerThewillA standard
ThewillTotal
distribution
economic
Total
in
sign explaining
government
of
coefficient
Predicted
size
of equation
- + -
+
+
+/-
+ +
groups
centralization
income
level
interest
enfranchisement
of
fractionalization
of
variable
modernization
of income
government
government
government
fractionalization
income
of
of of of
Concept
measured
Dummy
Political
Degree
Median
Skewness
Start Governmental
SizeSizeSizePopulation
Mean
EthnicNumber
distribution
Educational
variables
The
1.
Table Variable
name
LITSWI
INC
MINC
DATECENT
TAX
CONS
POP
EFRCNIG PFRCSKEWVOTE
EXP
1. 2. 3. 4. 5. 6.
7.
8.
9.
10. 11. 12. 13. 14. 15.
129
trade
and
counmost
DATEearlier
because
an
place. for
For
industry
are
groups
The
data.
1973.
1973.
Fund
Fund
the
in
problem
took data
no
is
giving
countries.
others
source
a and
between transformation'
deal
great
a
social
and
than
variable
rather
this
vary
1973.
Monetary
Monetary
Nations,
included etc.,
were
work
this
International
International
United
in
5 5 5
chosen,
'economicbetween
was
and
source
mismatch
households,
This
leadership'
temporal
1970s.
income,
1960s.
1968.
The
andandand
of
listed
late
to
the
4 4 4
in modernizing
the
of 2,000.
in others.
groups1963
Tables
Tables
Tables
in
point
definitions
All
in in in
than
point
(5).
minus
that some
point
some
to
(1973).
stability.
'consolidation
periods
column
some
estimates
estimates
estimates
for countries
warned
income.
to
refer
the twoare
for for for
be
commerce.
more
59;43;58;
data
of refer
temporal
data
for
which
232-237,
these
p. p. p.
shouldThe
of The
data
pp.
data
post-transfer
during
exhibits
1982a
1980a
1982a
The
reader
chambers
years had
1978 Wirtschaftsverbande
271-274.
54-56. (1982). 229-235.
derand48. Thepretax
which
number
OECD
OECD
OECD
pp.
p.
pp.periods sourcepp.
140-144.
on
Fund
3, 3, 3,
Heston
pp. 1972
this 1972
1972
1972 beginning
unions,23-25.
and
variable
variable
based
the
Table
Table
Table
14,
1972
identified a
Verzeichnis
by in in in
of
is
labor pp.are has Monetary
because
1,
Hudson Hudson
Hudson
Hudson
data
Summers,
1976
Nations
and
and
and
and
average
Table
the (1966)
period,
estimates
estimates
estimates
to
the
ForForForUnited
Kravis,
Sawyer
Black
Taylor
associations,
tries
is International
Internationales
Taylor
Taylor
time
centralization
Taylor
1. 2. 3. 4. 5. 6. 7.
8. 9.
10. 12. 13.
14.
Notes
130
governmentsize as a percentageof total economicactivity.2Thesevariables
arelistedanddefinedin Table 1 togetherwithall othervariablesusedin this
study. The data are for 1970, unless otherwisenoted.
The traditionaldiscussionof the role of governmentviewsit as a provider
of publicgoods. By definitionpublicgoods have significantscaleeconomy
attributes.The pricepercapitaof an army,of a judicialsystem,or of a central governmentshouldfall as the populationof a countryincreases.Thus,
as populationincreasesthe relativecost of publicgoods shoulddecline.As
the demandfor publicgoods is likelyto be priceinelastic,probablyinfinitely so for goods such as legislativeactivity, the fraction of total income
devoted to governmentshould declineas populationincreases.For some
goods, however,demandmayshift outwardas populationincreases:for example, the threatof aggression(demandfor defense), the level of crime,
etc., may increasewith countrysize. The populationvariablethus captures
the net effect of shiftingpublicgood demandand fallingpublicgood price.
McCormickand Tollison (1981)assumeall governmentactivityconsists
of wealthtransfers.Theyhypothesizethat interestgroupshavemoresuccess
using governmentto make these transfers,the less diligentare citizens in
policinggovernment,i.e., the more citizen free-ridingthereis. Since freeridingincreaseswith populationthey predicta positivecorrelationbetween
populationand governmentsize. The coefficienton populationcan be used
to test whethertotal governmentactivityappearsmoreas a publicgood or
a wealthtransfer.
In additionto relativecost (as proxiedby population),theoryleads us to
expect a positive relationshipbetween income and public good demand.
Since we seek to explain the relativesize of nationalgovernmentexpenditures,a positiverelationshipbetweenincomeandgovernmentimpliesthat
the incomeelasticityof a nation'sdemandfor publicgoods exceedsthe income elasticityof its demandfor privategoods, which is the assumption
usuallyreferredto as Wagner'sLaw (see Pryor, 1968:50). Both meanand
medianincome have been used in previousstudiesand each is tried in the
presentwork.
We employeda seconddemographicvariablebesidespopulationto capturethe degreeof ethnicfractionalizationin a country.Ethnicfractionalization might be viewedas a form of tastes variable,but we view it as more
relatedto the transactioncosts of reachingcollectivedecisions.The greater
the ethnicfractionalization,the greaterthe difficultyof reachingcollective
decisions, and the smalleris the expectedsize of government.
We turn now to those variablesthat emergeexplicitlyfrom the public
choice literature.
131
3.2 Public choice variables
The discussionin Section2 arguesthat the governmentsuppliesservicesto
specialinterestsin exchangefor politicalsupport.The greaterthe number
of interestgroupsin a country,the greaterwill be the numberof programs
arisingas a resultof bargainsbetweengovernmentand specialinterests.We
test for the influenceof interestgroupson governmentsize by includinga
count of the numberof interestgroupsin each countrylisted in a standard
referencework (see notes to Table 1).
Intuition suggests some sort of weightingof interest groups by their
potentialinfluence.Butthe most obviouschoicesof weightsareunsatisfactory. For example,interestgroupinfluenceis not a functionof the number
of membersin any simple way. An industrytrade associationmay have
relativelyfew members,but exerta largeimpactthroughsubstantialfinancialcontributionsto a party.Citizensgroupsof similarsizemayvarygreatly
in their impactsdependingon the intensityof the membersconcerns.For
this reason,and due to the non-availabilityof suitablealternativedata, we
have chosen to use the absolute number of interest groups formally
operatingin a countryas the measureof interestgroup strength.In doing
so, we makethe implicitassumptionthat the expectedimpactof a singleinterestgroup is the same across countries.
While we emphasize the importanceof organized interest groups in
addingadditionalexpenditureitemsto the publicbudget,one mightargue
that politicalpartiesare also a meansby whichdifferentvoter interestsare
weightedin the political process, and by analogy that governmentsize is
greaterthe greaterthe numberof politicalparties.This analogyseemsmost
plausiblewhenpartiesarecloselyassociatedwithgiveninterests(a farmparty, a labor party),but may hold more generally.We test for this potential
roleof numberof partiesby includinga measureof partyfractionalization,
i.e., the probabilitythat two randomlyselected membersof parliament
belong to differentparties.
It is often argued,however,that multipartysystemsare less stable, and
thuslesseffectivethantwo partysystems.To the extentthatthis is truethere
may be an offsettingeffect of the numberof politicalpartieson the size of
government.Stateswith multipartyparliamentsmay make more promises
to moreinterestgroups,but be less effectiveat deliveringon theirpromises.
We hypothesize,nevertheless,that governmentsize is greater,the greater
the degreeof party fractionalization.
The most frequentlyused publicchoice model in studiesof local government expenditures is the median voter model.3 The median voter theorem
predicts the outcome from a simple majority rule vote over a set of single
dimensional issues when voters have single-peaked preferences (Mueller,
1979: 40-42). As its assumptions are stretched considerably even when the
132
medianvoter model is usedto predictexpendituresby city governments,it
is not likelyto be applicableto a cross-sectionstudyof nationalgovernment
size.4Thus, we do not explicitlyattemptto use the medianvoter theorem
in our empiricalwork. We do test to see whethermedianincomeperforms
betterthanmeanincomeas a measureof averageincome,andwe also tested
an admittedlycrudeproxyfor the tax priceof the medianvoter. The latter
had no impacton any of the measuresof governmentsizeand no resultsfor
this variableare reportedhere.5
Two studieshaverecentlyappearedthat relyon the medianvotertheorem
and explicitlyseek to explainthe size, or growth in size, of government.
Both assumethat all governmentactivityinvolves only redistributionand
that the amountof redistributionis relatedto the skewnessof the distribution of income. Meltzerand Richard(1981)use the medianvoter theorem
to arguethat more redistributiontakes place(and thus moregovernmental
activity)the lower the income of the medianvoter relativeto averageincome. Peltzmanpredictsthe reversesign and claimsempiricalsupportfor
his hypothesis.Thus, we includea measureof the relativeincomeof the median voter. Our intention, however,is not to conducta test of these rival
theoriesbut ratherto ensurethat our resultson interestgroupinfluenceare
not biased by omittingrelevantvariables.
An importantelementof Meltzerand Richard'saccountof the growthin
governmentis the extensionof the voting franchiseto increasingnumbers
of voters,whoseincomefalls below the mean. We test for this enfranchisement effect directly by includingas a separateexplanatoryvariablethe
percentageof the adultpopulationwhichvotes. Oursuppositionhere, based on considerableempiricalsupport,6is that lowerincomegroupstend to
be disproportionatelyexcludedfrom voting de facto if not de jure. Thus,
higherpercentagesof votersin a populationmeanhigherpercentagesof low
income voters relativeto high income voters, and should lead to greater
redistributionand governmentsize.
Following Niskanen (1971) many economists have argued that the
strengthof the bureaucracyis importantin determiningthe size of the
government. We were able to construct a rather crude measure of
bureaucraticstrengthand test for its influence on a small subsampleof
countries.Giventhe small size of the samplefor this test, and the lack of
significanceof the bureaucracyvariable,the resultsfor this variableare not
reportedhere.
133
4. Ordinaryleast squaresresults
4.1 OECD countries
Thehypothesesput forwardpertainto developedcountriesin whichinterest
groupshave the potentialfor influencinggovernmentdecisions.A natural
choice of samplemeetingthis criterionis the OECD countries.
Whiledata for OECDcountriesare moreplentifulthan for others,even
for these, observationson all variablesare not available.We thus confront
a trade-offbetweennumberof observationsand numberof variablesin any
equation.Ratherthan arbitrarilyselecta givensubsetof variablesand subsampleof countries,we havechosento presenta spectrumof resultsrunning
from maximumnumberof observationsand fewest explanatoryvariables
to fewest observationsand maximumnumberof variables.The readeris
thus free to make his own trade-off. Table 2 lists the 24 OECD countries
and indicateswhich were deleted from the varioussubsamples.
The first3 equationsin Table3 providethe benchmarkfor measuringthe
influenceof interestgroups and the other public choice variableson the
relativesize of government.Populationhas a negativecoefficient in each
equationconsistentwiththe hypothesisthattotal governmentoutputhas on
averagegood characteristics.7Both income and ethnic fractionalization
Table 2. Countries in Sample
OECD countries
Australia 1, 2
Austria1,2, 3,4, 5, 6, 7
Belgium1,2, 3, 4, 5, 6, 7
Canada1,2, 3, 4, 5, 6, 7
Denmark1,2, 3, 4, 5,6, 7
Finland1,2, 3, 4, 5,6, 7
France1,2, 3,4, 5, 6, 7
F.R. Germany1,2, 3, 4, 5, 6, 7
Non-OECDcountries
Chile4, 5, 6, 7
Costa Rica4, 5, 6, 7
Israel4, 5, 7
Venezuela4, 5, 6
Key
I 2 3 4 -
Luxembourg1
Netherlands1,2, 3,4, 5,6, 7
New Zealand1,2, 3, 4, 5, 6, 7
Norway 1, 2, 3, 4, 5, 6, 7
Portugal1
Spain 1,2
Sweden1,2, 3,4, 5, 6, 7
Switzerland1,2, 3, 4, 5, 6, 7
Turkey1,2, 3,4, 5, 6, 7
UnitedKingdom1,2, 3,4, 5, 6, 7
UnitedStates 1,2, 3, 4, 5, 6, 7
Jamaica4, 5, 6
Mexico4, 5, 6
Panama4,5,6,7
Singapore4, 5, 6
Trinidad4, 5, 6
Uruguay5,6
Greece 1, 2
Iceland1,4, 5, 6, 7
Ireland1,2, 3,4, 5,6, 7
Italy 1,2, 3, 4, 5, 6, 7
Japan 1,2, 3, 4, 5,6, 7
included in equations 1-6 Table 3.
included in equations 7-9 Table 3.
included in equations 10-12 Table 3.
included in equations 1-2 Table 4.
5 - included in equation 3 Table 4.
6 - included in equation 4 Table 4.
7 - included in equations in Table 5.
134
0.23(4.36)
0.09(1.71)
0.28(3.48)
VOTE
SKEW
0.12(0.66)
-0.47 -.14
(-0.36)(-1.73)
PFRC
0.136
0.058
0.01(0.13)
0.002 0.129
0.109
0.008 0.079
0.065
(1.10)
(1.81)
(1.15)
(1.70)
(1.40)
(1.09)
(-0.05)
(-0.25)
NIG
0.005
0.004
0.002
0.005
0.002
0.003
0.006
0.006
0.005
(1.63)
(3.48)
(4.17)
(3.17)
(3.56)
(2.44)
(4.03)
(2.76)
(2.54)
0.02(1.83)
0.01(1.37)
0.02(3.36)
0.01(2.03)
0.06(0.41)
0.02(2.55)
MINC
INC
parentheses)
in
EFRC
(t-statistics
sample
POP
0.07(1.18)
0.06(0.59)
0.14(1.83)
0.29(3.04)
0.11(2.35)
0.30(4.22)
.07 .014
.170 .062
.050
.019 .161
-.100
-.10 - 1.79)-.149
(2.84)
-.117 -.106 (3.21)
(-0.35)
(-1.55)(-1.85)(-0.52)(-2.50)
(-1.47)
(-3.38)(-1.98)(-1.56)(-
.11
.003
.045
3.29)
2.17)-.14 -.05 -.12 -.137
-.136 -.051 -.143.24)
-.055
-0.04
((-2.30)
(-3.60)
(-5.03)(-1.66)
(-0.21)
(-4.55)
((-4.38)
(-2.76)
(-1.35)(-
OECD
squares,
1.16(0.15)
8.03(2.96)
19.3(1.62)
18.5(3.43)
10.8(4.19)
18.4(3.30)
15.5(3.66)
25.0(2.87)
24.2(5.03)
20.6(5.92)
-0.80 -3.27
(-0.07)(-0.43)
least INTERCEPT
ordinary
.24 .42 .11 .51 .67 .32 .52 .67 .50 .70 .84 .41
R2 S.E.E.
7.02 5.33 3.41 5.61 4.04 2.99 5.67 4.16 2.57 4.19 2.70 2.71
using
of
results
Number
24
24
observations
24
24
24
24
21
21
21
18
18
18
Regression
3.
EXP TAX CONSEXP TAX CONSEXP TAX CONSEXP TAX CONS
1.
3.
8.
5.
7.
2.
6.
9.
4.
10. 11. 12.
Table Dependent
variable
135
have the predictedsigns in all three equations, although only income is
statisticallysignificantin each.
The fit is improvedconsiderablyby the inclusionof the two publicchoice
variables,numberof interestgroups and political fractionalization.The
formeris significantat the 99 percentlevel in all 3 equations,politicalfractionalizationis significantat the 5 percentlevel (one tail test) in one equation. The performanceof both populationand ethnic fractionalizationis
noticeablyimprovedby the additionof the numberof interestgroupsand
politicalfractionalization.The performanceof mean income is worsened.
For 21 countrieswe wereable to measuremedianincome and skewness
of the income distribution.When medianincome is introducedits coefficientis positivein all 3 equationsand significantin 2. In those 3 equations,
and the following 5, we tried mean income and median income as alternatives and median income performedbetter, in terms of t-values, all 8
times. We reportthe resultsfor only medianincomethroughoutthe restof
the table.
and Peltzmantheoriesposit a relationshipbeBoth the Meltzer-Richard
tweenthe pre-transferskewnessof the incomedistributionand government
size. Ourskewnessmeasureis post-transfers,andthus our resultsarebiased
theorypredicts
awayfromthe negativecoefficientthat the Meltzer-Richard
and towardthe positivecoefficientPeltzmanexpects.The negativecoefficienton this variablein two equationsin spiteof this biasmightbe regarded
as weaksupportfor the Meltzer-Richard
hypothesis.The statisticalperformanceof this variableremainedweakin the remainingequationsalso, and
it is omittedto save a degreeof freedom.
The performanceof the other variablesin eqs. 7-9 is similarto that in
eqs. 4-6, but with lower t-valuesas can be expectedgiven the additionof
a variablewhichperformspoorly. The numberof interestgroupsremains
significantin all 3 equations,however.
To add VOTE, the percentageof the populationwhichvotes, another3
countriesmustbe droppedfromthe sample.Eqs. 10-12 presentthe results
withthe VOTEvariableincluded.Its coefficientis positiveas predicted,and
significantin all 3 equations. Its inclusion increasesall R2's visibly, and
generallyincreasesthe t-statisticson the othervariables.For the first time,
the interceptis not significantlydifferentfrom zero, as one's intuitionsuggests should be the case. Eqs. 10-12 are clearlythe best specificationsof
the governmentsize equationin the table.
Sincethe scale of the interestgroup variablewill not be knownto most
readers,it will be usefulto give moreinformationto facilitateinterpretation
of coefficient estimates. This information is best conveyed through elasticity estimates. Thus, in equation 10, for example, the elasticity at the sample
mean of the interest group variable is 0.18. (The equivalent statistic for
equation 11 is 0.15 and for equation 12, 0.13.) This indicates that, ceteris
136
paribus, in a country which is at the sample mean (government expenditure
35% of GDP), a 10% increase in numbers of interest groups will lead to an
extra 0.7%0of GDP flowing through the government sector.8
4.2 An expanded democratic country sample
There are two fairly obvious criticisms of the results presented so far, (1)
the sample is small and (2) some of the right-hand side variables in the
estimated equations may be related to the dependent variable in other relationships and therefore will not be exogenous. We take up the first objection
here, the second in the following section.
In extending the sample we faced the problem that most non-OECD
countries have significantly lower levels of economic development than the
OECD countries and only a small fraction have political institutions sufficiently democratic that one might reasonably expect to find the kind of
trading of interest group support for government programs the theory
predicts. We used Bollen's (1980) democracy index to decide which countries had sufficiently democratic institutions,9 and we omitted any country
with a per capita income less than Turkey's, which has the lowest figure for
any OECD country. Using these criteria, we were able to find 10 countries,
in addition to the OECD ones, for which at least one of the three dependent
variables and the most important independent variables could be obtained
(see Table 2).
Table 4 presents the ordinary least squares regression results for the expanded sample. The equations include the variables shown by the OECD
sample to be important: population, number of interest groups, voter participation, and income. Mean income replaces median income and political
fractionalization has been dropped due to the lack of data. These results can
be compared with those for equations 10-12 in Table 3. The results for the
government final consumption variable (equation 3) are disappointing. Inspection of the data revealed the newly added observation on Israel to be
an outlier. The size of that country's defense expenditures causes Israel's
government final consumption observation to be 75% higher than the next
largest observation. When Israel is excluded (equation 4), the results for this
dependent variable fall in line with those for the other two and resemble
those of Table 3.
The coefficient for the number of interest groups is positive and significant in all 3 equations; that of population is negative and significant in all
3. Once again, the model explains government consumption less well than
total expenditure or tax revenue. Both income and voter participation are
highly significant in the expenditure and revenue equations and of the right
sign in the consumption equation. Ethnic fractionalization is of the right
sign in all three but is not significant. It performed even more weakly in
137
VOTE
0.27(2.76)
0.21(3.52)
0.09(1.28)
0.05(1.22)
NIG
0.005
0.004
0.002
0.002
(2.25)
(2.98)
(2.71)
(1.06)
INC
0.02(2.66)
0.02(4.66)
0.004
0.003
(0.72)
(0.91)
EFRC
0.002
-0.11 -0.06 -0.02
(-1.49)(-1.38)(-0.03)
(-0.66)
parentheses)
in
POP
0.04 0.052.44)
-0.12 -0.11 (-2.67)(-3.98)(-1.22)
(-
(t-statistics
sample
1.92(0.26)
5.69(1.00)
8.93(2.86)
2.14(0.47)
INTERCEPT
expanded
squares,
least
R2 S.E.E.
.64
.80 .08 .38
0.43 3.94 4.93 2.69
ordinary
using of
results Number
observations
28
28
29
28
Regression
4.
EXP TAX CONSCONS
1.
2.
3.
4.
Table Dependent
variable
138
initialestimatesof the simultaneousequationsmodelof the nextsectionand
is (therefore)droppedfrom those results.
5. Towarda simultaneousequationsmodel of governmentsize
While feedbacks from the level and composition of governmentexpendituresto the level of nationalincomeand populationsize can be envisioned, such feedbacks are likely to unfold sufficiently slowly that we are
justified in ignoringthem. More serious, perhaps,are possible feedbacks
from governmentsize to the two political variableswith the greatestexplanatory power, interestgroups and voter participation.We allow for
these feedbacksin this section by estimatinga threeequationmodel, with
governmentsize, numberof interestgroupsand voter participationas the
dependentvariables.
The first equationof our model is, of course, the one developedin previous sections.The secondequationfocuseson the determinantsof interest
group formation. Peter Murrell(1984) has examinedeleven hypotheses
regardingthe formationof interestgroups. Of the many variablesused to
test these theoriesonly three, population,decentralizationof government,
and lengthof time of modernization(see Table 1), had significantexplanatory power. These three variablesare used to explain variationsin the
numberof interestgroups across countrieshere.
Countrieswith largerpopulationscan be expectedto have more heterogeneouspopulations.Thus, largercountriesrequiremoreinterestgroupsto
representthe diverseinterestsof the polity. If we assumethereare some fixed costs, or scale economies to interest group formation, then holding
heterogeneityconstant,the largerthe populationthe more interestgroups
of optimalsize a societycan accommodate(Pauly, 1967).Thus, population
size should have a positiveimpacton the numberof interestgroups.
MancurOlson (1982)hypothesizesthat the formationof interestgroups
is fostered by periods of democratic stability. We incorporate this
hypothesis by including the date when modern political and economic
developmentin a countrybegan. Its sign should be negative.
Salisbury(1975:200) arguesthat the more decentralizedpoliticalpower
is, the more potentialfor interestgroup influencethere is and the greater
the number of interest groups there will be. A negative coefficient is
predictedfor politicalcentralization.
Finally,we test whetherlargegovernmentsize spursthe formationof interest groups by including a measure of government size in the equation explaining the number of such groups.'0
To find the determinants of voter participation to be included in our third
equation, we turn first to cross-section studies of voter behavior in the
139
UnitedStates.Thesestudiesfind educationand incometo be importantexplanatoryvariables(see Note 6). We tried both of these variablesin our
equation, proxyingeducationallevel by the literacyrate.
To the extentthat greatervoterparticipationis a resultof largerturnouts
by the enfranchisedpoor, ratherthandifferencesin enfranchisementacross
countries,one might expectthe poor to participatein greaternumbersthe
greaterthe competitionfor their supportamong existingparties. In turn,
more party competitioncan be expectedthe greaterthe numberof viable
politicalparties.Thepoliticalfractionalizationvariablemeasuresthe degree
of multipartycompetitionin a country, and we includeit to capturethis
possibleeffect.
In 1970,womendid not havethe rightto vote in Switzerland.One would
expectthat, if this fact werenot takeninto account,therewouldbe an overpredictionof Switzerland'svoterparticipationgivenits literacyand income
levels. Indeed, we have already observed Switzerland'soutlier status in
some equations(see Note 8). An interceptdummy for Switzerlandis included in the VOTEequation.
Bothof the otherdependentvariablesareincludedin the VOTEequation.
As hypothesizedfor NIG, largergovernmentsize might inducegreaterinterestin politicsand greaterpoliticalparticipation.The numberof interest
groupsis expectedto havea negativeimpacton voterparticipation,if it has
any impactat all. Interestgroupsare in parta substitutefor directpolitical
participation.The strongerthe interestgroup structurein a country, the
smaller the direct participationof voters in the political process. The
negativesignon this variablemay also reflectsome voterdisillusionmentin
polities whereinterestgroups are very strong.
Thesehypotheseslead to a fully identified,threeequationsystem.Table
5 presentstwo-stageleast squaresresultsfor this systemusingthe 23 countries for whichdata were availablefor all variables.The first 3 equations
specifya linearrelationshipamongall of the variables.We reportonly the
resultsfor the tax revenuemeasureof the sizeof government,sincethe other
measuresof governmentsize give similarresults,once Israelis excludedfor
the reasondiscussedabove.
In equation 1 of Table 5 we see that the numberof interestgroupscontinuesto havea strongpositiveimpacton governmentsize whenallowance
is made for the endogeneityof this variable.The other variablesperform
as before. In equation2, we see that the POP, DATE, and CENTvariables
performas predictedfrom Murrell's(1984)study, althoughthe t-valuefor
CENTis lowerthanMurrellfound for the OECDcountriesalone.1"Of particular interest is the low t value on tax revenue in this equation. Government size does not appear to affect the formation of interest groups in this
linear formation of the model. Thus, the estimates presented in Tables 3 and
4 are not contaminated by simultaneous equations' bias, at least with
respect to the NIG variable.
140
Both income and political fractionalizationperformedweakly in the
VOTEequationandareomittedfromthe reportedresults.Thus,on a crossnationalbasis,higherincomesdo not leadto greatervoterparticipation,nor
does an increasein the numberof politicalpartiesinducegreaterparticipation as a result say of greater competition for votes. Switzerlandhas
significantlylower voter turnoutas a percentageof the populationas expected. Literacyis positivelyrelatedto voter participation.The numberof
interestgroupsshows a strongnegativerelationshipto voter participation
indicatingthatindirectparticipationthroughinterestgroupsdoes substitute
for directparticipationto someextent.Thereis someindicationthatgreater
governmentsize elicitsgreatervoter participation.The hypothesesare not
formulatedin such a way as to determinewhich is the most appropriate
functionalform. Thus, all equationspresentedin linear form throughout
the paperwerealso estimatedin log-linearform and resultsweregenerally
similarfor both specifications.With respectto the equationpredictingthe
numberof interestgroups,however,an importantdifferenceemerged.The
estimatesof the log-linearversionof the model are thus given in Table 5.
The tax revenuevariableobtaineda t-valuegreaterthan2 whenall variables
were included in log form (see eq. 5). Thus, whetherone can treat the
numberof interestgroupsas exogenousor not whenexplaininggovernment
size dependsupon whethera linearor logarithmicspecificationis imposed.
Whileeq. 5 suggestsa betterfit underthe logarithmicspecification,equations 4 and 6 indicateweakerfits. Nevertheless,essentiallythe samepattern
of resultsemergeswhen all variablesare measuredin logs.
6. Conclusions
Theresultsof the previoussection,estimatesof a threeequationmodelfrom
23 observations,must obviously be regardedas tentative. The consistent
positiverelationshipbetweennumberof interestgroupsand size of governmentobservedwithchangingsetsof includedindependentvariables,changing samplesof nations, and treatingthe numberof interestgroupsas either
exogenousor codetermined,does imply ratherunequivocallythat interest
groupsare able to influencepublicpoliciesin such a manneras to lead to
increasedgovernmentsize. Beyondhelpingto reinforcethis conclusion,the
results of the previous section should be regardedas first steps in the
developmentof a model of the polity that can explainparticipationin the
political process by interest groups and citizens as well as the size of
government.
The two most important variables explaining government size other than
the number of interest groups proved to be population and the percentage
of the population voting. The consistently negative relationship between
141
log
1.54
(-7.28)
50.15.65)
(-
SWI
natural
0.32(1.39)
0.37(1.79)
LIT
the
is
form
DATE
9.90
(-2.82)
-5.44
(-2.61)
log-linear
parentheses)
in
CENT
-2.24
(-3.72)
-11.9
(-1.29)
the
in
which
INC
(t-statistics
0.02(2.48)
0.20(1.11)
forms
-.13 17.5(5.42)
(-3.12)
-0.069 0.42(4.72)
(-1.59)
POP
log-linear
TAX
and
3.73(0.13)
0.47(1.25)
SWI
except
1,
Table
1.55(2.24)
0.29(1.49) in
listed
linear VOTE
0.25(3.44)
0.40(3.06)
0.005
(2.34)
2.45)
-0.004 0.13(1.75)
(-
both
in
variables
NIG
the
-0.04
(-1.66) of
model
logarithms
10679
1.06(0.19)
2.10(2.72)
0.28(0.48)
83.3(2.98)
INTERCEPT
(2.47)
29.2(2.17)
three-equation
natural
of
R2 S.E.E. 0.724.200.78617.0.737.910.760.140.890.550.810.12 the
areform.
of
estimates
linear
23
observations
23
Number
squares
23
23
23
23
equations
the
in
least
log-linear
variable
Log-linearthe
Log-linear
Log-linear
form LinearLinearLineara
a
a
Functional
in the
Two-stage
plus
5.
one
TAX NIG VOTETAX NIGI VOTE
Variables
1.
Table Dependent
of
5.
a
6.
3.
2.
variable
4.
142
relativegovernmentsize and populationis noteworthysince severalrecent
papershaveassumedthatthe only governmentoutputis redistribution.The
negativerelationship,implyingthatan increasein populationleadsto a less
than proportionateincreasein the size of government,shows that government expenditureexhibitsa most basic publicgood characteristic.
The percentageof the population voting, which probably is closely
relatedto the proportionof voters with incomes below the median, consistentlyhas a positiveandsignificantimpacton the sizeof government.The
Meltzer-Richard
hypothesisthatgreaterparticipationby low incomevoters
leads to more redistributionand greater government size is strongly
supported.
The inclusionof both the interestgroupand voterparticipationvariables
in the governmentsize equationrelies on theoriesrelatedto redistributive
activities.The voterparticipationvariableposits a directresponsivenessof
governmentoutcomesto voterpreferencesthroughthe operationof the median voter theorem, and implies rich-to-poorredistribution.The interest
grouptheoryposits increasinggovernmentsize throughthe additionto the
publicwealof expenditureson goods withdisproportionatebenefitsfor certain interest groups. Such expenditureshave distributionalimplications
since in the absence of governmentprovision the interestgroups would
eithergo withoutthe goods or have to providethem themselves.Whilethe
theory makes no explicitpredictionabout the directionof this redistributional flow, sincethe largestsinglecategoryof interestgroupsin most countries by far is industrytradeassociations,12one might expect poor-to-rich
redistributionas the most likely consequenceof interestgroup influence.
Thus, the possibilityexists that the influenceof the two variableson the
distributionof incomemightbe largelyoffsetting, whiletheirinfluenceon
the sizeof governmentis cumulative.Disaggregatingthe effectsof theseand
other public choice variablesis a promisingavenue for future research.
NOTES
1. This propositionis rigorouslyderivedalong with othersconcerninginterestgroupsand
governmentsize in Muellerand Murrell(1983). Becker(1983)stressesthe point that the
most efficientmeansfor supplyingbenefitsto interestgroupswill be employed.
2. Thesevariablesall measureaggregategovernmentsize.A directconsequenceof ourtheory,
however,is thatinterestgroupswill havemoreeffecton somecomponentsof government
thanon others.Unfortunately,we werenot ableto carryout ourtestson disaggregatedata
becauseof the unavailabilityof sufficientnumbersof observationson componentsof
governmentspending.
3. See Bergstromand Goodman(1973), Borcherdingand Deacon (1972), Deacon (1978),
Pommerehne(1978).
4. For a lengthycritiqueof the medianvoterliteratureas appliedto explaininglevelsof local
governmentexpendituressee Romerand Rosenthal(1979),and Mueller(1979:106-111).
143
5. For the resultsusingthis tax-pricevariableand for the results,discussedat the end of this
section, on the bureaucraticstrengthhypothesis,see Muellerand Murrell(1983).
6. See Frey (1971), Tollison and Willett(1973), and Verbaand Nie (1972), and references
therein.
7. McCormickand Tollison(1981: Ch. 3) find that populationsize is positivelyrelatedto
regulationactivitiesat the statelevelconsistentwiththeirhypothesisthatlargepopulation
sizeleadsto less vigilantcitizenpolicingof governmentandthusmorewealthtransferactivitiesby government.The two resultsneed not be contradictory.Some governmentactivities of a particularlyredistributivenaturemay grow largeras populationincreases,
becomerelativelysmaller.Ourresultsindicate
whilethosewithpublicgood characteristics
that the lattertendencydominatesfor total governmentoutputat the nationallevel.
8. Wetestedfor thepresenceof heteroscedasticity
usinga modifiedversionof the Glejserprocedurein whichthe log of the squarederrorsfromeqs. 10-12 was regressedon the log of
each independentvariable(see Pindyckand Rubinfeld,1981:123-126). The hypothesis
that the residualsweredrawnfrom a homogeneousdistributioncould not be rejectedfor
withthe residualsinverselycoreqs. 10and 11, but the test did indicateheteroscedasticity,
relatedto the numberof interestgroups,for eq. 12. Multiplicationof all of the variables
Giventhe samplesize, however,
in (12) by NIG successfullyremovedheteroscedasticity.
is not too surprising.Wealso examinedthe
the failureto findsignificantheteroscedasticity
residualsfor outliers.No residualwasgreaterthantwicethe S.E.E. in absolutevalue, for
equations10-12, 1-3, and5. Inthe remaining5 equations,Switzerlandappearedas a lone
outlier.Switzerland'soutlierstatusis discussedand allowedfor below whenwe estimate
the simultaneousequationsmodel.
9. Wechosea scoreof 70%on Bollen'sindexfor 1965as a cutoff. Thisseemedto be a natural
cut-offpoint.Whilemanycountriesfallin the70s, few appearin the 50sand60son Bollens
index.Venezuela(73)andMexico(74)arethelowestscoringcountriesincludedwhileZambia (67) and Brazil(61) are the highest-scoringcountriesexcluded.
10. We triedthe variableVOTEin the equationexplainingNIG and it had no effect andthus
these resultsare not reported.
11. The readeris reminded,however,that the t-statisticsfor 2SLSestimatesfrom samplesof
this size are only suggestiveof whetherthe coefficientis significantor not (see Maddala,
1977: 231-233, 237-242).
12. Interestingin this connectionis Fratianniand Spinelli's(1982)observationthat special
governmentprogramscateringto businessinterestgroupsare becomingincreasinglyimportantin Italy.
REFERENCES
Becker,G.S. (1983).A theoryof competitionamongpressuregroupsfor politicalinfluence.
QuarterlyJournalof Economics98: 371-400.
Bergstrom,T.C., and Goodman,R.P. (1973). Privatedemandsfor publicgoods. American
Economic Review 63: 280-296.
Black, C.C. (1966). The dynamicsof modernization:A study in comparativehistory.New
York:Harperand Row.
of politicaldemocracy.American
Bollen,K.A. (1980).Issuesin the comparativemeasurement
SociologicalReview45(3):370-390.
Borcherding, T.E. (Ed.) (1977). Budgets and bureaucrats. Durham: Duke University Press.
Borcherding, T.E., and Deacon, R.T. (1972). The demand for the services of non-federal
governments. American Economic Review 62: 891-901.
144
Breton, A. (1974). The economic theory of representative government. Chicago: Aldine.
Brunner, K. (1978). Reflections on the political economy of government: The persistent growth
of government. Schweizerische Zeitschrift Volkswirtschaft und Statistik 114: 649-680.
Buchanan, J.M. (1967). Public finance in democratic process. Chapel Hill: University of
North Carolina Press.
Deacon, R.T. (1978). A demand model for the local public sector. Review of Economics and
Statistics 60: 184-192.
Downs, A. (1957). An economic theory of democracy. New York: Harper and Row.
Eckstein, H. (1963). The determination of pressure group politics. In H. Eckstein and D.E.
Apter (Eds.), Comparative Politics: A reader. New York: The Free Press.
Fratianni, M., and Spinelli, F. (1982). The growth of government in Italy: Evidence from 1861
to 1979. Public Choice 39: 221-243.
Frey, B.S. (1971). Why do high income people participate more in politics? Public Choice,
101-105.
Frey, B.S. (1982). The public choice approach to the explanation of collective consumption.
In R.C.O. Matthews and G.B. Stafford (Eds.), The Grants Economy and Collective Consumption, 43-65. London: Macmillan.
Frey, B.S., and Schneider, F. (1975). On the modeling of politico-economic interdependence.
European Journal Political Research 3: 339-360.
International Monetary Fund (1982) Government finance statistics yearbook. IMF;
Washington, D.C.
International Monetary Fund (1973). International financial statistics. IMF; Washington,
D.C.
Internationales Verzeichnis der Wirtschaftsverbiinde. Miinchen K.G. Saur Verlag.
Kravis, I.B., Heston, A.W., and Summers, R. (1978). Real GDP per capita for more than one
hundred countries. Economic Journal 88: 215-242.
Maddala, G.S. (1977). Econometrics. New York: McGraw-Hill.
McCormick, R.E., and Tollison, R.D. (1981). Politicians, legislation, and the economy. The
Hague: Martinus Nijhoff.
Meltzer, A.H., and Richard, S.F. (1978). Why government grows (and grows) in a democracy.
Public Interest 52: 111-118.
Meltzer, A.H., and Richard, S.F. (1981). A rational theory of the size of government. Journal
of Political Economy 89 (5): 914-927.
Mueller, D.C. (1979). Public choice. Cambridge: Cambridge University Press.
Mueller, D.C., and Murrell, P. (1983). Interest groups and the size of government. IIM Working Paper 1983-1, International Institute of Management, Berlin.
Murrell, P. (1984). An examination of the factors affecting the formation of interest groups
in OECD countries. Public Choice 43 (2): 151-171.
Niskanen, W.A., Jr. (1971). Bureaucracy and representative government. Chicago: AldineAtherton.
North, D.C., and Wallis, J.J. (1982). Longer-run determinants of the economic role of the
U.S. Government. American Economic Review 72 (May): 336-340.
Nutter, G.W. (1978). Growth of government in the West. Washington: American Enterprise
Inst.
Olson, M. (1965). Logic of collective action. Cambridge: Harvard University Press.
Olson, M. (1982). The rise and decline of nations: Economic growth, stagflation, and social
rigidities. New Haven: New Haven.
Organization for Economic Cooperation and Development (1980a). Revenue statistics of
OECD member countries. Paris: OECD.
Organization for Economic Cooperation and Development (1982b). Employment in the public
sector. Paris: OECD.
Organization for Economic Cooperation and Development (1980b). The tax benefit position
145
of selected income groups in OECD member countries: 1974-8. Paris: OECD.
Organization for Economic Cooperation and Development (1982a). Historical statistics:
1960-80. Paris: OECD.
Pauly, M.V. (1967). Clubs, commonality, and the core: An integration of game theory and the
theory of public goods. Economica 35: 53-65.
Peltzman, S. (1980). The growth of government. Journal of Law and Economics 23 (2):
209-288.
Pindyck, R.S., and Rubinfeld, D.L. (1976). Econometric models and economic forecasts. New
York: McGraw-Hill.
Pommerehne, W.W. (1978). Institutional approaches to public expenditures: Empirical
evidence from Swiss municipalities. Journal of Public Economics 9: 163-201.
Pryor, F.L. (1968). Public expenditures in communist and capitalist nations. London: George
Allen and Unvin.
Romer, Th., and Rosenthal, H. (1979). The elusive median voter. Journal of Public Economics
12: 143-170.
Sawyer, M. (1976). Income distribution in OECD countries. OECD Occasional Studies. Paris.
Taylor, Ch.L., and Hudson, M. (1972). World handbook of political and social indicators.
New Haven: Yale University Press.
Tollison, R.D., and Willett, T.D. (1973). Some simple economics of voting. Public Choice 16:
59-71.
Verba, S., and Nie, N.H. (1972). Participation in America - Political democracy and social
equality. New York: Harper and Row.
United Nations (1972). Demographic yearbook.
United Nations (1973). Yearbook of national accounts statistics. New York: United Nations.