Upload
jessica-voigt
View
221
Download
0
Embed Size (px)
Citation preview
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 1/21
Ph.D. Program in Political Science of the City University of New York
http://www.jstor.org/stable/4150172 .
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 formsof scholarship. For more information about JSTOR, please contact [email protected].
.
Ph.D. Program in Political Science of the City University of New York is collaborating with JSTOR to digitize,
preserve and extend access to Comparative Politics.
http://www.jstor.org
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AM
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 2/21
ComparativeFederalismandDecentralization
OnMeaningndMeasurement
JonathanRodden
The basic structureof governance is being transformed in countries around the
world as authorityand resourcesmigratefrom central to subnationalgovernments.
Politicalscientistsandeconomistshavedevelopeda wealthof theoriesto explainthecauses andconsequencesof these shifts, but systematicempirical testing has laggedbehind. Researchers increasingly supplement case studies with analyses of largecross-nationaldata sets that exploit both diachronic and cross-national variation.
While the constraintsof datacollection necessitate simple propositionsat relatively
high levels of abstraction,it is appealingto test hypotheses about federalism and
decentralizationwith data drawnfrom aroundthe world over several decades. Yet,while breadthand generality are the greatest assets of cross-country regressions,
they can also createpitfalls that threaten o obscurerather hanclarifythe facts.
Reliablecross-nationaldataon decentralizationand federalismarescarce,andthe
concepts are often assumed to be complementary or even interchangeable. The
emergingview of decentralizationshows an organic, intertwinedtransferof politi-
cal, fiscal, and policy autonomy. Some cross-national studies seek to explain
endogenousfiscal decentralization.'Others treatdecentralizationand federalismas
exogenous and attemptto measuretheir effects on economic growth or proxies for
accountability,corruption,or the qualityof governance.2Still othershave examined
the implications for deficits, inflation, and macroeconomic stability, while older
studies examine the size and growthof government.3Each of these studies uses a
simple measureof fiscal decentralization,a binary distinctionbetween federal and
unitarypolitical systems, or both.The most clearly discernibleLeitmotivin these studies is a growing disappoint-
ment with decentralizationand federalism,especially among developing countries.
Optimistictheories,startingwith Montesquieuand continuing throughmodernwel-
fare economics, stress advantagesof informationrevelation and accountability in
more decentralizedgovernancestructures.4Public choice theories explorethe possi-
bility thatmobility in a decentralized,multijurisdictioncontext can facilitate better
matching of citizen preferences and governmentpolicy through"sorting"and can
lead to smaller,more efficient, less corrupt governmentand undersome conditions
481
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 3/21
ComparativePolitics July2004
more secure markets and faster growth.5However, recent empirical studies take
issue with these theories;they find thatdecentralizationand federalism areassociat-
ed with higher levels of perceived corruption,largergovernment,macroeconomic
instability,and under some conditionslowergrowth.6They often concludeby castingdoubtuponthe benefits of decentralizationand federalism.
However,distinctions between various shades of decentralizationand federalism
have not been taken seriously. Questions about the design, content, and form of
decentralizationareglossed over not because the theoriesandhypothesesof interest
areundifferentiated, ut becausemore refined dataaredifficult to collect. The blunt-
ness of these measuresis often acknowledgedbut defendedas the cost of achievinga large enough sample to make reliable inferences. But just how high are these
costs? Do the favored indicatorsof decentralizationactually measure the conceptsaddressed n the relevanttheories?Some basic questionsaboutdefinitions andmea-
surementneed to be asked. The links between theory and empirical analysis havebeen quitetenuousin the first generationof empiricalstudies.
New data can provide a fuller conceptual and descriptive account of forms of
decentralization and federalism and thus several important clues as to why the
results of previous empirical studies are so dissonantwith normativetheories and
often with one another. Above all, ratherthan enhancingthe independentauthorityof state andmunicipalgovernments,decentralizationoften creates a more complex,intertwinedform of governancethat bears little resemblanceto the forms of decen-
tralizationenvisioned in textbookson fiscal federalism or in public choice theories.
Inlight
of thesefacts,
it isnecessary
to reassesses what has been learned from the
first generationof cross-nationalstudies and to emphasize ways of improvingdata
collection,theory,andthe linksbetween the two.
Decentralization
Decentralization s often viewed as a shift of authoritytowards local governmentsand away from central governments,with total government authorityover societyand economy imagined as fixed. Attempts to define and measure decentralization
have focusedprimarilyon fiscal andto a lesser extentpolicy andpolitical authority.
Fiscal Decentralization Mostempiricalstudiesof decentralizationocus exclusivelyon the balanceof expendituresandrevenuesbetween governments.They rely on the
IMF'sGovernment inanceStatisticsYearbooko calculatethe combinedregionaland
local shareof totalgovernment pending.The first column of Table1 presentsexpendi-turedecentralization ata for all countriesfor which the IMF provides good coverageover the 1990s.Well over halfof public expenditures remade attheregionalandlocal
levels in decentralized ederationsike CanadaandSwitzerland,while the figure is less
482
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 4/21
Jonathan Rodden
Table 1 Fiscal DecentralizationVariables Averagesover 1990s)
1 2 3 4 5 6 7 8
Sm Gn-Own-ource Gr?+ rev.
Own-•r Stoocal tax SMtbou xnooral Gu~rarwaw - a.rvot, rev mavtereve Boowrlg
expotaexp bcall en•u i reoM• rv.
ra(otarvo•n (rnt andbase ionomyrmv/toal. bo m. tenloloctonnomy) lonoAmy)
Source GFS" GFS OGFS" Rodden002 Rodden2002 OECD* OECD* Rodden002
Argentina 0.44 0.56 0.18 4.0Australia 0.50 0.40 0.32 0.37 0.33 2.5Austria 0.34 0.27 0.27 0.58 0.16 0.008 0.004 1.6
Belgium 0.12 0.56 0.06 0.048 0.004Boivia 0.21 0.09 0.18 0.43 0.11 1.5Botswana 0.03 0.83 0.01 0.84 0.01 1.0Brazil 0.41 0.34 0.28 0.36 0.27 4.5
Bulgaria 0.19 0.35 0.15 0.92 0.02 1.0Canada 0.65 0.26 0.51 0.32 0.47 0.321 0.299 2.7Colombia 0.38 3.0Denmark 0.54 0.43 0.31 0.43 0.32 0,174 0 1.5Finland 0.41 0.34 0.33 0.36 0.31 0.196 0 3.0France 0.19 0.35 0.13 0.39 0.12 3.0
Germany 0.45 0.25 0.34 0.70 0.13 0.0090 0.0002 2.5Guatamala 0.10 0.65 0.04 0.867 0.03 2.0
Hungary0.10 0.018 0
Iceland 0.23 0.09 0.23 0.184 0India 0.49 0.43 0.33 0.42 0.34 2.5Indonesia 0.13 0.73 0.03Ireland 0.29 0.72 0.09 0.74 0.08 1.8Israel 0.14 0.40 0.08 0.38 0.09 2.4
Italy 0.23 0.66 0.09 0.80 0.05 2.5
Malaysia 0.14 0.19 0.16Merdco 0.26 0.11 0.20 0.59 0.09 0.018 0.018 2.6Netherlands 0.30 0.70 0.09 0.70 0.09 0.010 0 2.3
Nigeda 0.48 0.86 0.09 0.86 0.09 1.0
Norway 0.35 0.40 0.22 0.39 0.22 0.004 0 1.6Peru 0.23 0.73 0.07 0.05 0.23 2.5
Philippines 0.08 0.46 0.05 0.41 0.06 1.0Poland 0.17 0.31 0.13 0.54 0.09 0.032 0 2.0
Portugal 0.09 0.48 0.05 0.40 0.06 0.011 0.009 2.5
Paraguay 0.02 0.01 0.23 0.01 2.0Romania 0.11 0.44 0.07 0.44 0.07 1.0
Spain 0.36 0.60 0.16 0.56 0.17 0.053 0.022 2.5Sweden 0.37 0.18 0.33 0.20 0.32 0.183 0.006 3.0Switzerland 0.55 0.25 0.41 0.19 0.45 0.179 0.113 3.0Thailand 0.06 0.28 0.05
UK 0.29 0.71 0.09 0.74 0.08 0.040 0 1.5USA 0.53 0.30 0.42 0.34 0.39 0.177 0.177 3.0*and authorscalculations
thanfourpercent n some highlycentralizedAfrican countries.More important,hesedatamake tpossibleto tracedevelopmentsn fiscal (de)centralizationver time.Figure
1a displaysaverages or a groupof twenty-nine ountries or whichtime series dataareavailableback to 1978. It demonstrates strikingupwardtrend. In 1978 on averagearound20 percentof expenditureswere made at subnationalevels,andby 1995 thefig-ure hadjumpedto over32 percent.By no means is this trenduniversal,however. Insomecountries, hebalanceof spendingauthority hiftedbackslightlytowards he cen-
tralgovernmentn the 1990s,while fiscal decentralization as been mostpronouncednSpainandmuchof LatinAmerica.
However,hese datado not inspiremuchconfidencein theirusefulnessas a compos-ite measureof decentralized uthority.Forinstance,Denmark s the third-mostdecen-tralizedcountry n the worldaccording o Table1-even more decentralized hantheUnitedStates-though the centralgovernmentightlyregulatesvirtuallyevery aspectoflocalgovernmentinance.Nigeria appearsas number even,eventhoughthe statesdur-
ing thisperiodof militaryrule were little morethan administrativeutpostsof the cen-tralgovernment.
483
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 5/21
ComparativePolitics July2004
In short,it is difficult to know whatto make of expendituredecentralizationdata
withoutadditionaldataon theregulatoryrameworkor subnationalinance.Mosttheo-
retical argumentsstressing efficiency gains from decentralization along with more
recentarguments ointingoutits dangers) mplicitlyassumethatimprovementsdeteri-orations)in responsivenessor stronger ncentivesfor effort (self-seeking) stem from
increasedautonomyfrom centralcontrol.As with industrialorganization heory,the
essenceof decentralizationn moststudies s thatit grants ocal governmentswiderdis-
cretionwhile limitingthe center'saccess to informationandcurbing ts abilityto over-
turn local decisions ex post.7 In comparing irms, if the regionaldivisions in firm A
spendmorethanregionaldivisionsin firm B, it may not reflectthatA is more decen-
tralized n any real sense if A'sregionalmanagersare tightlycontrolledby the central
office andtheir decisionsaresubject o frequent xpost reversalswhile B is essentiallya holding company. n the sameway,expendituredecentralizationwithingovernments
maycommunicate erylittle about helocus of authority.An importantconsideration s whetherexpendituredecentralizations fundedby
intergovernmentalrants,revenue hat s sharedwith the centeraccording o a fixed for-
mula,or the mobilizationof own-sourcerevenuethrough ndependentaxes,user fees,
andborrowing.Untilrecently,virtuallyall cross-country tudieshaveignoredthesedis-
tinctions.The Government inance StatisticsYearbook oes includea line in its subna-
Figure 1 SelectedTime Series Indicatorsof Decentralization
la: AverageExpenditureDecentralization, 9 1d: Percent of countries with decentralized
countries primary ducation policy (43 countries)
0.5 0.6
0.4
0.20.2
0.4
0 0
1965 1975 1985 1995 1965 1975 1985 1995
Ib: Percent of countries where center annot 1: Percent of countries withdecentralized
easily override SNG 43 countries) Infastruture policy43 countries)
0.8 0.6
0.40.4
0.20.2
0
1965 1975 1985 1995 1965 1975 1985 1995
ic:ercentfountriesithlectedubnationalf. ercentfountriesithecentralizedolicegovernments43ountries) policy43ountres)
064->06-0.6-
1965 1975 1985 1995 1965 1975 1985 1995
Sources: - - - Sharedbetweencenterand subnational ovts.la: GFSlb: Henderson 2 - Sharedbetween 2 or more ubnationaliers
Ic: Henderson nd WorldBank,WorldDevelopmentReport 000
ld-lf: HendersonOne subnationalieralone
484
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 6/21
JonathanRodden
tional revenueaccountscalled"grants,"utformanycountries heydo not includecon-
stitutional evenue-sharingrograms.This line canbe usedto calculateoverall ransfer-
dependence;averagesare presentedin the second column of Table 1. Moreover,an
alternativemeasureof fiscal decentralizationanbe obtainedby calculatingown-sourcerevenueas a shareof totalgovernment evenue(the third columnin Table1). While a
useful sourceof variationover time, one shouldbe carefulaboutdrawinginferences
basedon cross-sectionvariation ince the codingof grantsand subnational wn-source
revenuesseemsnotto be consistentacross countries.
One wayto improveuponthis measureis to drawon countrysourcesto develop a
measureof subnationalrevenueautonomythatdoes not code automaticdistributions
fromrevenue-sharing chemes as own-sourcerevenue.The fourth column of Table 1
presentsa measureof grants plus shared revenuesas a shareof subnationalrevenue,and the next column includes a measure of own-source revenue as a share of the
total.8The latteris an alternative o the simple expendituredecentralizationvariable.
It attemptsto measurethe shareof total governmentrevenuethat is actuallyraised
throughrevenueeffortsof subnationalgovernments.However, ven this variable everelyoverestimateshe extent of subnational evenue
autonomy.While subnationalgovernmentsmay collect the revenues labeled as own-
source,the centralgovernmentmayneverthelessmaintain hepowerto set the rateand
the base, leaving the subnationalgovernmentsas mere collectors of centrallydeter-
minedtaxes.A recent OECD studytacklesthis complex issue, but unfortunately nlyfor a smallnumberof countries.9From this study, t is possibleto calculatetwo addi-
tional variables: he shareof total tax revenuesover which subnationalgovernmentspossess full autonomy o set theirown tax ratesand their own ratesand base. These
variables,also presented n Table1, painta very differentpictureof subnational iscal
autonomy.Severalcountries n which subnational overnmentsaccountforlargeshares
of totalspending column1) andtaxation column 5) haveverylittleautonomyovertax
rates and base (column 7). In fact, the study makes it clear thatthe United States,
Canada,and Switzerlandare in a class by themselveswhen it comes to autonomous
subnationalrevenue authority.An importantgoal for furtherdata collection is the
improvementnd extensionof theOECDstudy,especiallyto developingcountries.
Table2presents
a matrixof pair-wisecorrelation oefficientsfor all of the variables
discussedin this article.The numbersof cases uponwhich the correlationsare based
arepresented n italics. The own-sourcerevenuedecentralization ariablescalculated
from the Government inanceStatisticsYearbookndcountrysourcesarehighlycorre-
lated at .9. In the first column there are reasonably high correlations(.87 and .84)betweenexpendituredecentralizationndthe own-sourcerevenuedecentralizationari-ablescalculated rom the Government inance StatisticsYearbook ndcountrysources.
However, or the smallerOECDsamplethe correlationbetweenexpendituredecentral-
izationandeitherof the autonomousocaltaxcollectionvariables allsto around 64.
485
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 7/21
ComparativePolitics July2004
Additionally, entralgovernmentsmight attempt o restrictthe fiscal autonomyof
subnational overnmentsnot onlythroughconditionalgrantsandregulationsgoverninglocaltaxation,butalsothrough ormal imitationson subnational orrowing.Theability
to access creditmarketsor other sourcesof deficit finance independentlys an impor-tantcomponentof subnationaliscal autonomy.An indexcreatedby the Inter-American
DevelopmentBank considersdebt authorization equirements,numericallimits, and
restrictions n the use of debtimposedby the centralgovernment, longwiththe abilityto borrow hroughbanksandpublicenterprisesownedby subnational overnments in
1995).Thisvariable,whichrangesfrom 1 to 5, is presented n the lastcolumnin Table
1. Table2 shows thatborrowingautonomyis positively correlatedwith measuresof
expenditure nd revenuedecentralizationndnegativelycorrelatedwith transfer-depen-dence.10
Policy Decentralization The decentralization of policy autonomy is rarelyaddressed by empirical scholars because it is difficult to measure. Fortunately,Vernon Henderson has recently taken up the task.11 First of all, Henderson asks
whetherthe centralgovernmenthas the legal rightto overridethe decisions andpoli-cies of lower levels of government"with an ease that calls that very authorityinto
question."'12 lthough the coding involves a fair amount of discretion, Figure lb
shows that in 1975 21 percent of the central governments in Henderson's samplelacked overrideauthorityand thatby 1995 the figure soaredto 60 percent. Second,Hendersonasks which level of governmentis responsible for decision making in
each of threepolicy areas:primaryeducation(controlof curriculumand hiring/fir-ing of teachers),infrastructurelocal highwayconstruction),and local policing. The
plots in Figures 1dthroughIf show an unmistakable rendtowardsincreasinginflu-
ence for local andregionalgovernments n eachpolicy area.
Perhapshemost strikingaspectof thesecharts s theprevalenceof sharedauthority.Theportionof the samplein whichthe centralgovernmentandone or more local gov-ernmentshavejointpolicy authoritys plottedwitha dashed ine.Veryrarelydo central
governmentsully cede autonomy o subnationalgovernments. n the vast majorityof
cases,decentralizationntailsa move fromcompletecentraldominance ojoint involve-
mentof the centerand one or more subnationalier.Evenin the cases wherethe central
governments not involved,authoritys often sharedbetweentwo or more subnationaltiers (plottedin normalfont). Finally,situations n which a single subnational ier is
involved npolicymaking plotted n boldfont)areextremelyrare.Studiesof fiscal fed-
eralismstartwith"thepresumptionhattheprovisionof publicservices shouldbe locat-
ed atthelowestlevel of government ncompassing,n a spatialsense,the relevantbene-
fits and costs."'3The normativeprescriptionand ultimatelythe positive theories that
flow from it envisiona neatcompartmentalizationf authorityaccording o somethinglikethe subsidiarity rinciple,withoccasionalsharedauthority or tasks wherethe rele-
486
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 8/21
JonathanRodden
Table 2 CorrelationCoefficientsfor DecentralizationVariables,1990s
Owm-sourceraft+rv.
Ownoouro Steo-p(aotax Stx1l-4)&t&X
Stbt4ocal Gantte- sta$-bcclsharing/s(te- state-ocal rvfttotl rev rv/tots rev Borrowing Poicy Subnotionl
btal exp ocal revenue rv/otal rev.local rev. revnvtaolrev (rate (rafetandbas utonmy aulonomy eleions
(GFS) (R2002) auono my) autonomy)
-0.1958
GranssWtatelocarevewue 35
ownsoucstate
oclDrev/otal
0.87296 -0.5908*rev. GFS) 36 35
Grants+
rev.
sh
adta,-ocl
-0.1668 0.4600' -0.3724'rev. 32 30 31
Owsourcstateoca revwotn 0.8419' -0.4182* 0.8995* -0.5979'rev Rodden2002) 32 30 31 32
stoteocal tax evnot evrate 0.6407' -0.3817 0.7257' -0.6557' 0.8698*
autonomy) 19 18 18 16 16
swtateocaaxrvot.
vrate 0.6399' -0.2278 0.6806* -0.434 0.7027* 0.6797*andbse autonomy) 19 18 18 16 16 19
0.4126' -0.3545 0.4953* -0.4510* 0.5126* 0.22460.2741
32 30 31 34 32 16 16
0.5130* 0.1499 0.3402 0.0833 0.1067 0.3417 0.3117 0.1038Poy23 19 19 19 17 10 10 19
0.5475* -0.1299 0.4313 -0.2409 0.3461 0.3285 0.335 0.2533 0.7775*
23 19 19 19 17 10 10 19 47
-0.6128* -0.0716 -0.5331 0.3941 -0.4829 -0.2381 -0.2208 -0.5574 -0.6011' -0.5663'
12 11 11 12 11 7 7 12 13 13
0.6220* -0.2561 0.6039* -0.0325 0.5047' 0.0679 0.5102* 0.4204* 0.3611' 0.3994'
39 34 35 34 32 18 18 34 47 47
SirnW parWse oorotatoncoefient
Number f ovwpelan cas In tat
Isigni an t .06
vantbenefitsand costs fall between evels.However, he data n Figure1 showthatcen-
tral andregionalgovernmentsareoften involvedeven in the determination f the cur-
riculumandthe hiringandfiringof elementary chool teachersandin decisionsabout
local road construction. t is possibleto turn the policy datain Figure1 into a simplemeasureof decentralizedpolicy autonomy.A reasonablestrategy s to give countries
two points if they lack centraloverrides,two points for each policy area controlled
exclusively by regionalor local governments,and one point for policy areasin which
they arejointly activewith the center.For 1995 the index rangesfrom zero to seven.
Table2 showsthatthisvariable s correlated t .51 withexpendituredecentralizationut
is notsignificantlycorrelatedwithanyotherdecentralizationmeasures.
Political Decentralization It is possible to get some insight into political decen-
tralizationby trackingregionaland local electionsover time. Figure 1c plots the por-tion of the sample in which regionaland local governmentswere popularlyelected,
again showing a dramatictrend towarddecentralization.While only 30 percent of
local governments n the samplewere elected in 1970, by 1999 the figure had grownto 86 percent.Formanyof the countries he shift to subnationalelections was partof
a broader hift fromauthoritarianismo democracy n the late 1980sandearly 1990s.
A simple index of politicaldecentralizationangingfrom zero to two assigns one
487
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 9/21
ComparativePolitics July2004
pointfor each subnationalevel atwhich executiveswerepopularlyelectedin 1995.Not
surprisingly, able2 shows thatpopularlyelected local officials have higherlevels of
policyautonomyhanappointed fficialsandpresideoverlargersharesof publicexpen-
ditures.Whilethebinarydistinctionbetweenelected andappointedofficials is useful, sub-
tler variation n political decentralization mong systems featuringpopularlyelected
subnationalofficials is moredifficult to quantifybutjust as important.Above all, it is
importanto assess the relationshipbetweenthe centraland subnationalelectoralare-
nas.Forinstance,the slate of candidatescompetingin local electionsmightbe chosen
by centralgovernmentpartyofficials. At the other end of a spectrum,state or local
officials might play a key role in selecting candidatesfor centralgovernmentelec-
tions. Forexample,federalpartylists are drawnup by statepartyleadersin Australia
andGermany,andthe statesplaya key role in selectingpresidentialcandidates n the
U.S. Although such issues are examined in a few case studies, a worthy goal forfutureresearchwouldbe to compilea cross-nationaldataset.14
Moreabstractly,uchvariablesareinterestingn assessingthe relative ndependence
(or interdependence) f the centraland subnational lectoral arenas.Forinstance,state
gubernatoriallectionsintheUnitedStatesare known to be influencedby voters' evalu-
ationsof the presidentandhis copartisans.Unfortunategovernorsmight sufferfroma
midtermpunishment ffect,while luckygovernorsmightget a free rideon the coattails
of a popularpresidentduring imesof economicprosperity.n a more extremeexampleof externalitiesbetween centraland subnationalelections, voters view German state
elections as federal midterm contests becausethey directly
determinethe partisan
make-upof the powerfulupper egislativechamber.On the otherhand,David Samuels
suggeststhatcoattails n Brazil,where the states arethe federalelectoraldistricts,oper-ate in the oppositedirection: ederalelectionsareheavilyinfluencedby state-levelpoli-tics in whichgovernorsplaya crucialrole.15
Thesuccessof a subnational fficialhasa component hat s basedon evaluationsof
local performance nda component hatis based on evaluationsof the nationalparty.In spiteof coattails,several studies demonstrate learlinks between measuresof state-
level fiscal and macroeconomicperformanceand gubernatorial lection results in the
United States.Thus,the reelectionprospectsof U.S. governorsclearlyhaveboth nation-
al andstatecomponents.An estimationof the size of these components or subnationalofficials in differentcountrieswould tell a greatdeal abouttheirincentives,for exam-
ple, forintergovernmentalooperation,ocaleffort,andfiscal discipline.A less precise but more manageableassessmentof the relationshipbetween cen-
tral and subnational electoral arenas was first suggested by William Riker and
RonaldShapsand has been implementedfor a sample of fourteenfederationssince
the 1970s:the numberof stateor provincialchief executives sharingthe partyaffili-
ationof the federalchief executive.16This figure conveys informationaboutfluctua-
tions in political (de)centralizationover time and facilitates useful cross-national
488
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 10/21
JonathanRodden
comparisons.For instance, it tracks the dramaticdecline of Congress party domi-
nance in India,the gradualerosionof PRI dominancein Mexico, the fragmentationof the Brazilianfederalsystem, andNigeria's spurtsof authoritarianism nd democ-
ratic federalism.Althoughthe numberof overlappingcases is low, Table2 suggeststhaton averagein the 1990s this index of "partisanharmony" s negativelycorrelat-
ed with measuresof fiscal andpolicy decentralization.
Federalism
Federalismis not a particulardistributionof authoritybetween governments, but
rathera process-structured by a set of institutions-through which authority s dis-
tributedand redistributed.Federalismcan be traced back to the Latinfoedus, or
covenant.The word eventuallywas used to describecooperative,contractualagree-ments between states, usually for the purpose of defense. Covenantsand contracts
imply mutuality-to serve any purpose,both partiesmust fulfill some obligationto
one another.If the centralgovernmentcan get everythingit wants from local govern-ments by simple acts of administrative iat, it makes little sense to see the two as
engaged in a contractual,or federal,relationship.Federalismimplies that for some
subset of the centralgovernment'sdecisions or activities it is necessaryto obtainthe
consentor activecooperationof the subnationalunits.
Before filling in the details on the institutionsthatunderlie federalcontracts,it is
importanto understandhow and
whyfederal contractsare made in the first
place.Both the definition andthe operationof federalismare wrappedup in the historical
conditions that give rise to the original contract. William Riker posits that modern
federations originated as bargains aimed at achieving military defense against a
common enemy, though one might add other collective goods like free trade and a
common currency. 7Alliances and loose confederationsare often plagued by insta-
bility, free riding, and collective action problems,but if the incentives for coopera-tion are strongenough and political incentives are properly aligned,representativesof the entitiesmightnegotiatea new governancestructure eaturinga centralgovern-ment with strongerenforcementpowers and decision rules that require something
less thanunanimity.Once this contract s made, it takes on a life of its own and con-tinues even after the enemy has been defeated or a common market or currencyachieved.Alfred Stepanidentifies a second pathto federalism.18Manymultinational
statesareformednotby voluntarybargainsbutby the vagariesof conquestand colo-
nialism.A federalbargainbecomes necessaryto hold the multinationalstatetogetherandassuagefearsof interethnicexploitation.
Inbothscenarios heoriginal ederalbargain s anagreementabout the compositionandpowersof the centralgovernment ndthe rules thatwill structureuture nteractions
betweenit and the units.The aftermathof the original bargainmakes federationsdis-
489
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 11/21
ComparativePolitics July2004
tinctfromunitarysystems.The unitswill not cede authorityo the centerwithout safe-
guardsagainst utureexploitation, itherby the center tselforby otherstates.Thusfed-
eralbargainsgenerally nclude constitutionalanguageprotecting he sovereigntyand
autonomyof the units,in some cases includingclauses thatvest them with all residualpowersnot explicitlydesignated o the center. The credibilityof such languageoften
hinges on the presenceof a strong, ndependent onstitutional ourt.More important,federalbargains equiremajoritiesand oftensupermajoritiesf the territorial nits fora
wide rangeof policy changes, especiallychanges in the basic vertical distributionof
policyand fiscal authority rthe constitutiontself.
FromPhiladelphian 1776 to theNice Summitof the EuropeanUnion in 2000, it is
clear that in strikinga federalbargainthe most serious concerns aboutexploitationcome from small territoriesthat would be consistently overwhelmedif votes were
apportionedccording o population.Thus,smallstates nsiston representationchemes
based on territory,while large states arguefor population-basedrepresentation.The
compromiseusually nvolves apopulation-basedowerchamberand a highlymalappor-tionedupperchamber hatoverrepresentsmall states.19Comparative ata collectedbyDavidSamuelsand RichardSnyder howthat federationsdemonstratemuchhigher ev-
els of upperchambermalapportionmenthanunitary ystems.20Thus, federalism s a form of preferenceaggregation hat often leans on bargains
among territorialgovernmentsor their representativesas opposed to majorityrule
amongindividuals.21Butby no means is thisconcept binary.Table3 depictsa continu-
um thatreflects the role of territorial overnmentsn the centralgovernment'spolicy-
making process.At the far left, decisions are made
by majoritiesof individuals.
Territoriallyased districtsplayno role. Perhapshe bestmodemexamplesarereferen-
daor townmeetingsin ruralNew Englandand Switzerland.Toprovidea morerealistic
national-level xample, n Israel,withonly one nationwideelectoraldistrict, ower level
governmentsdo not maponto anyterritorial nitthatplaysa formalrole in the central
government's ecisionprocedure.Further longthe continuumarelegislatureswithrep-resentativeselected from territorialdistricts,the model for most modernlegislatures.Evenif these districtsdo notcorrespondo the boundariesof territorial overnments asinWestminster-styleystemswhereredistrictings frequent),moreterritorial argaining
mightbe expected han n systemswithoutdistricts.Yetfurtheralongthe continuumare
legislaturesn which theboundaries f territorial overnments orrespondwith those ofelectoraldistrictsbut seatsare allocatedby population,as in the Italianupperchamber.
The nextslot in Table3 is occupiedby theupperchambersof most modernfederations,
including heUnited States.Eachterritoryhas a similarnumberof directlyelectedrep-resentatives, egardlessof population.But the next slot on the continuum s even more
federal.In the originalU.S. Senate andthe modern GermanBundesrat, epresentativesareappointedby the constituentgovernments.Finally,atthe farrightof Table3 areleg-islaturesn whichdelegatesareappointed,mallstatesareoverrepresented,ndchangesfromthe statusquo requiresupermajoritiesr,at the extreme,unanimity.For constitu-
490
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 12/21
JonathanRodden
tional changes,the GermanBundesratfalls into this slot, as does the EU council of
ministers ormost important olicy issues.With each moveto the rightalongthis spec-trum,territorial overnmentsakegreaterprecedenceas the relevantunits in construct-
ing legislativemajorities, nd onemightsaythatrepresentations more federal.The representationof states in central government policymaking is part of the
essence of federalism.22Yeta strictdefinition would seem to exclude Canada,since
the provinces are not formally representedas veto players in the decision-making
process of the federalgovernment. Nevertheless, the federaland provincial govern-ments are clearly locked into an ongoing process of intergovernmental ontractingthat takes place primarilyoutside of centralgovernmentinstitutions.The Canadian
centralgovernment goes so far as to sign formal, contractualagreementswith the
provinces. Even though the Canadian central government need not obtain the
approvalof the provincial governmentsin orderto make policies, it often can not
implementthem withoutcajoling, striking bargainswith, andmaking side paymentsto provinces.The Russianand Spanishcentralgovernmentsengage in directbilateral
and multilateralbargainingwith regional governments,and a variety of rather for-
mal, policy-specific multilateralbargainingbodies includingthe states andthe cen-
tralgovernmenthas evolved in Germanyand Australia.
In sum, federalcontracting s largelya productof institutionalincentives arisingfrom previous bargains,but sometimes the relevant institutionsare not identified in
the constitution. Some countries,like Germany,Brazil, and the United States, pos-sess virtuallyall of the abovequalities.India,Austria,andCanada,however,aregen-
erallyconsideredfederationseven
thoughtheir
upper legislativechambers are nei-
ther strong nor highly malapportioned.India's federal credentials are sometimes
challengedbecause of the prime minister's constitutionalauthorityto dismiss state
governments,but this power has been used with diminishing frequency over time
withoutconstitutionalchange.In spiteof these grayareas,previousattemptsat cross-nationalmeasurement reat
federalismas a binaryconcept.Countlesscross-nationalstudies drawon the classifi-
cations of constitutionalscholars who identify federationsby relying more on com-
mon sense andexperiencethanrigorouscoding criteria.23 hebottom row of Table2
displays correlation coefficients for the simple federal dummy variable with the
otherdecentralizationvariables. It is not surprisingto find that federalism is posi-tively correlated with measures of expenditure, revenue, borrowing, policy, and
political decentralization.Contraryto common wisdom, federations are indistin-
guishablefrom unitary systems in levels of transfer-dependence.Furthermore, he
positive correlationcoefficient for subnational ax autonomyis drivenby three fed-
erations:the U.S., Switzerland,and Canada.The other federations for which data
were availablepossess very low levels of tax autonomy.The traditionalclassification of federationsis quite expansive, lumping together
countriesas diverseas Switzerland ndPakistan. talso masksimportant ariationover
491
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 13/21
ComparativePolitics July2004
Table3 The Role of TerritorialGovernments n Legislatures
Unitary Federal Confederal
Elected Electeddistrict ps district op Electeddistrict Distrterapa Districtrepo
Decision- thatdornot that reps that thatare thatare
making units corrtpond correspond correspond o appointedby appointedby
toterritory toterritory territtoryovts. territoryovt. tentory govt.
govt. govt.
Apportionment one person one erson mal- ma- ma
of districts n onevote one vote pportionmentapportionment
apportionment
Requirement for simple simple simple Qualified- r
policy change majority majority mjorty super-maoprity
UK talian upper UnitedStates German EU Councl ofExample townmeeting Parliament chamber Senate Bundat Minister
time,for
example, Nigeriaas it
slipsin and out of
militarydictatorshipand India as
president'sulefades into obsolescence.There s roomforimprovementn themeasure-
ment of federalism.Coefficientson federaldummyvariables n cross-countryregres-sionsshouldbe regardedwithskepticism.Althoughperhapsnotmuchbetter,a potentialavenuefor morerefinedmeasurementwouldbe the creationof a continuousmeasure
by allocatingscoresbasedon the dimensionsof federalism isted aboveandthe repre-sentationcriteriadisplayed n Table 3. Betteryet, any attemptto measurefederalism
shouldbe carefullycalibratedo the theoreticalargument f interest.
Linking Theory and Data Analysis
Several lessons from this exercise in concept clarification and measurementstand
out. While expendituredecentralizationand federalism are correlated with some
alternativemeasures,they areblunt and potentiallymisleading proxies for many of
the phenomenaaddressedin leading theories aboutthe causes and consequences of
decentralization.Even more important, he datapresentedabovepaintvery different
picturesof decentralization ndfederalismfrom those implicitlyassumedin much of
the theory.Fiscal andpolicy decentralizationdo not often entaila shift in some fixed
amount of authorityor money from the center to regional or local governments.
Rather,it usually entails the additionof new layers or additionto the resources or
responsibilities of existing lower-tier governments in the context of overlapping
spheresof authority.Political decentralizationalso addsto the complex and intertwinednatureof mul-
titieredgovernance.Whenpolicy authorityand finances areintertwined, t is not sur-
prisingthat voters'evaluationsof centralandsubnationalofficials arealso interrelat-
ed. Studies of normativefiscal federalism and American constitutionaltheories of
dual federalismhave cast a long shadow. Too often economists and political scien-
492
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 14/21
JonathanRodden
tists theorize aboutdecentralizationas if it entails a neat division of tasks, with the
center only stepping in to providenational collective goods and correct spillovers.
Prevailing notions of federalism in political science create a similar problem.
Federalismdoes not necessarilyentail greaterindependentauthorityfor subnationalgovernmentsover taxes, expenditures,or anythingelse. Federalismdoes not implythat the centerand states are sovereigns,each protectedfrom the interferenceof the
other.On the contrary, ederationshave evolvedas ongoing incompletecontractsand
by theirvery natureare underconstantrenegotiation.In most federationsthe center
often dependson the provincesto implementand enforcemany of its decisions, and
the centercan not change the statusquo in certain areaswithout the consent of the
constituentunits.
Buildingon these lessons, the remainderof this articlerevisits severalbroadtop-ics addressedin cross-countryempiricalresearch.For each area of research, it dis-
cusses limitationsof existing studies,ways to improvethe link between theory and
data,andnew directionsfortheory.
Endogenous (De)centralization The key intuitionof fiscal federalismtheory is
that the benefits of decentralizationare positively correlatedwith the geographicvariancein demandsfor publiclyprovided goods.24This line of argumentbuilds on
the logic expressedby Montesquieuand Rousseauin stressingthe benefits of decen-
tralizationwhen territories are large and populations vast. Although they do not
make explicit the political process throughwhich demands for decentralizationare
transformednto policy, they maintainthat excessivelycentralized
systemsin
large,heterogeneouscountrieswill face overwhelming pressureto decentralize, lest theyfall apart hroughsecession or civil war.AlbertoAlesina andhis collaboratorsextend
this logic and examine a basic trade-offbetween the benefits of large jurisdictionsand the costs of heterogeneity n largepopulations.25But large size comes at a cost:
the difficulty of satisfying a more diverse population. Another group of studies
emphasizesa relatedtrade-offbetween the benefits of coordinationand economies
of scale andthe benefits of settingtax ratesand determiningredistributive ransfers
locally in societies with heterogeneous income levels (and distributions) across
regions.26 n all of these models, sufficiently high levels of heterogeneitygenerate
demandsfor decentralizationor even secession.One challenge for testing these arguments s that they requiremeasurementsof
regional heterogeneity of preferences. The proxies used in empirical studies byWallace Oates and Ugo Panizza are land area and ethnolinguistic heterogeneity,which are found to be positively associated with expenditure decentralization.
Moreover,these studies find that wealth and democracy are positively associated
with expendituredecentralization.
However,as arguedabove,expendituredecentralizations likely a poor proxy for
the policy and political devolutionthat these theories address.When ethnic or lin-
493
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 15/21
ComparativePolitics July2004
guistic heterogeneitygeneratesdemandsfor decentralization,demandsare likely to
focus on the locus of decision making about educationalcurriculumor policing or
the introductionof local elections. ForBolton andRoland'sargument,which focuses
on preferences over the locus of tax authorityderiving from income, the correctindependentvariablewouldrequireregionaldataon income levels and interpersonaldistribution,andthe correctdependentvariablewould be somethinglike the OECD
tax autonomyvariablesdescribedabove. However,the move from theory to testingwill be quite difficult;the type of subnationalautonomyovertaxationenvisioned in
this theoryis extremelyrare,even amongthe most developedcountries.
In fact, the datapresentedabove suggest that theories of endogenous centraliza-
tion, if they aim to shed light on recenttrends,shouldgo beyondmedianvoter theo-
ries in which decentralizations conceptuallyequivalent o secession. Itmay be more
relevantto ask why politicians choose to make the state more intertwinedand com-
plex. For instance, what incentives lead central governmentpoliticians to give upindependentresponsibility for the provision of certain public goods and begin to
share responsibility with local governments by funding decentralized provision
throughgrants?Part of the answermight have to do with the potential for reducingcentral government budget deficits and shifting blame for poor performance.
Conversely,whatexplainsthe dominanttrendof the first half of the twentiethcentu-
ry in which independent fiscal and policy autonomy, especially over taxation,
migrated from states and provinces to federal governments or, more recently, to
intergovernmentalbodies? Since the pathbreakingwork of Fritz Scharpf and his
associates,very
little theoreticalorempirical
attentionhas been given to questions
addressing he increasingly ntertwinednatureof central,regional,and local govern-mentdecisionmaking.27
In short, beyond a handful of studies of expenditure decentralization, cross-
nationalworkon endogenoustax, policy, andpolitical decentralization s virgin ter-
ritory. The most promisingavenuefor the next generationof research s to addressa
widerrangeof decentralizationvariablesand supplementmedianvoter models with
morerealisticassumptionsaboutinstitutionsandpolitics.
Accountability, Corruption, and Good Governance If decentralizationactually
resembledthe clean transferof authorityenvisioned in fiscal federalism theory, itmightbring governmentcloser to the people andenhanceinformation,accountabili-
ty, and responsiveness to citizens. However, when decentralization amounts to
adding layersof governmentand expandingareasof sharedresponsibility,it mightfacilitateblameshiftingor creditclaiming,thusreducingaccountability.Even worse,in countries already suffering from corruption,it might lead to competitive rent-
seeking and "overgrazing"of the bribe base.28Previous studies of corruption, in
which the exogenous variableswere measures of expendituredecentralizationand
the federalismdummy,were not well suited to distinguishbetween these possibili-
494
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 16/21
Jonathan Rodden
ties.29However,more recent work by Treismanattemptsto deal with the potential
problems associated with overlappingauthority by using a variable counting the
numberof governmentaltiers and a measure of decentralized(and by assumption
overlapping) udicial and law enforcementauthority.30 erhapsfuturestudies mightimprove upon this approach by assessing the extent to which different tiers have
overlappingpolicy or, better,regulatoryauthority.Some of the most compelling arguments linking decentralizationto enhanced
accountabilityand lowercorruptionconcernmobility and intergovernmentalompe-tition that reduce the ability of governments to extract bribes. However, neither
expendituredecentralizationnor federalism is a reasonableproxy for intergovern-mentalcapitalmobility and competition.The formerrequiressome dataon whether
individualsand firms actuallymove or credibly threatento do so. In much of the
world,ethnic and
linguisticties are
very strongimpedimentsto
mobility.In all but a
small handful of wealthycountries,interjurisdictionalmobility is more likely a des-
peratemove to escape povertythanan expressionof preferencesaboutlocal service
provisionor corruption,andmigrationis often to the slums of the capitalcity where
corruptionand serviceprovision mightbe even worse.
Even so, the theoristscited above focus primarilyon the mobility of capitalrather
than labor.But for capital mobilityto havean effect on governance,subnationalgov-ernmentsmust possess significant regulatoryand especially tax authority.Yet the
datapresentedabove suggest that autonomous subnational ax authority s a fiction
in many countriesthat appearon the surface to be quite decentralized.To examine
the plausibilityof interjurisdictionalompetition,one approachhas been to examinethe size and numberof first tierjurisdictionsunder the assumptionthatmobility is
less costly among smallerjurisdictions.This approach s not very informative,how-
ever,withouta measure of the taxpowersof governments.In fact, simple theoriesof
optimal taxation provide good reasons to suspect a negative correlation between
jurisdictionsize andtax autonomy.In any case, none of the existing empirical studies of corruptionor good gover-
nance constitutesacceptabletests of hypotheses linking intergovernmental ompeti-tion to restraintson subnationalpoliticians.Firstof all, it may not be enough simplyto
identifysubnational ax or
regulatoryautonomyor even note that horizontalcom-
petition takes place. In some countries,the structureof the subnationaltax regime
might encouragetax exportationor corruptdeals between subnationalgovernmentsand manufacturers.Furthermore, hough virtually all of the supposed benefits of
decentralizationhinge on improvedaccountability,very little is known about the
links between the varieties of fiscal and policy decentralization or, more appropri-
ately, overlap) and the ability of voters to use elections to hold local officials
accountablefor their actions. A more promising avenue than cross-country regres-sions is to approachdetailed single country studies from an explicitly comparative
495
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 17/21
ComparativePolitics July2004
perspective.Forinstance,studies of the United Statesshow that votersindependently
punish and reward state politicians for fiscal and macroeconomicoutcomes, while
voters in the Germanstatesdo not.31The next step is to link such cross-countryvari-
ationsexplicitlyto differencesin political and fiscal institutions.
Fiscal Scale and Redistribution A similar set of problems has plagued cross-
nationalempiricalstudies of decentralizationand the size of government.Whether
or not one accepts the implicit antigovernmentbias in these studies, there are goodreasonsto believe thatdecentralization,f it facilitatesintergovernmentalax compe-tition, will lead to a smallerpublic sector.The first generationof empirical studies,
however,didnot attempt o measure decentralized ax autonomy.Oates examinedthe
effects of tax competitionon the size of governmentby using cross-section data on
expenditure decentralization, while Cameron used a federalism dummy.32More
recently,ErnestoStein differentiatedbetween fiscal decentralization undedby localtaxes and transfers, even distinguishing between discretionary and constitutional
transfers,in a cross-section study of governmentsize in Latin America.33Rodden
examined the differential effects of grants and own-source revenue using a large
panel data set and used the OECD tax autonomydata cited above to identify coun-
tries in which tax competitionis most likely.34The findings suggest that decentral-
ization is positively associatedwith the overall size of governmentwhen funded bytransfersbut has a neutralor even negative effect in the rarecases when decentral-
izationis fundedby local taxation.
Related studiessuggest
that federalism inhibits thegrowth
of redistributionand
the welfare state. Thereappears o be unanimousempiricalsupportfor this proposi-tion from a varietyof cross-nationalstudies, but they are limited to aroundfifteen
OECDcountries andrely on a simple federalismdummy.35f limitations on redistri-
butionare drivenby intergovernmental ompetitionthatfavorscapitaloverlabor and
causes local governmentsto fear the welfare magnet phenomenon, federalism is a
very poor proxy.An alternative ederalstory has to do with a statusquo bias amongfederationsowing to multipleveto playersin the era of welfare state expansion, but
it would requirea more refined measurecapturingsome of the concepts from Table
3. Yet anotherattractiveyet heretoforeuntestedtheorymighthave to do with the leg-
islative overrepresentationof fiscally conservative farmers and rural elites at theexpense of urban workers, in which case some variant on the Samuels-Snyder
malapportionmentndexmightbe useful.
Macroeconomic Management Finally,improvementscan be made in conceptual-
izing and using cross-nationaldatato assess the links between decentralizationand
macroeconomicoutcomes like deficits, inflation, and ultimately economic growth.New skeptical studies focus on coordination and collective action problems that
complicatemacroeconomicmanagement,adjustment,and reform when expenditure
496
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 18/21
JonathanRodden
decisions are decentralized.Here, again, simple expendituredecentralizationdata
withoutaccompanying nformationaboutthe institutionalframeworkof subnational
finance are misleading; central governments in countries with very decentralized
expendituresoften place strictlimitationson subnationalborrowing.Combinedwithcentralregulationof subnational axation,these limitations can give the centertools
with which to overcome local government impulses to conduct self-interested but
collectively destructivefiscal policy. It is also importantto examine the incentive
effects of various forms of intergovernmental rants.When the centralgovernmentcan threatento withhold intergovernmental rants, it may have an importantlever
with which to restrain ocal indisciplineand assuage intergovernmental ooperation
problems.On the otherhand,some kinds of intergovernmentalransfersmight also
providesubnationalpoliticians(along with theirvoters andcreditors)with reasons to
expect futurebailouts,undermining heir incentives for fiscal discipline.In the new theoreticaland empiricalstudies that stress macroeconomicdangers,
federalismplays a largerrole than fiscal decentralization.Federalismcompoundscoordinationand cooperationproblemsbecause it entrenchesthe states as veto play-ers in the centralgovernment'sown policymakingprocess. Not only might states or
provinces conduct collectively suboptimal fiscal policy, but constitutionalprotec-tions mightpreventthe centralgovernment rom intervening,or,throughtheirrepre-sentation in the legislatureor intergovernmentalbargainingbodies, the states that
benefit might possess veto powerto preventreform.
However, these aspects of federalism are very poorly capturedwith a dummyvariable.
Perhapsfederal-stylebargainings as
pronouncedin China as in
anyfor-
mal federationand less pronounced n Austria,Pakistan,or Nigeria (duringits peri-ods of militaryrule) thanin unitarycountrieslike Italyor, increasingly, he UK. The
natureof federalbargaininghas been transformed n Indiaand Mexico afterthe fall
of the Congress partyandthe PRI. Case studies of federalismand fiscal indisciplinereveal that small differences in political incentives, executive-legislative relations,and legislative organizationhave importantconsequences.36An importantgoal for
furtherresearch s the developmentof federalismvariables thataddress its constitu-
tionalandrepresentational imensions.Additionally, f federalismaffects macroeco-
nomic stability,the effect is likely to be contingentupon a varietyof otherpolitical
and institutionaldetails, like expenditureand revenue decentralization,subnationalborrowingautonomy,andpartisanharmony.
A final concern with cross-nationalregressions exploring the consequences of
decentralization,even if concepts are clear and measurementsprecise, is the fact
that, as discussed above, the vertical organizationof governmentis hardly exoge-nous. Both the natureof decentralizationor federalismandmacroeconomicvariables
might be codeterminedby othervariables like economic geography, he heterogene-
ity of the population, migration,income distribution,and demands for democracyandredistribution. ncreasedconfidence in theories linking institutionsto outcomes
497
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 19/21
ComparativePolitics July2004
requiresa betterunderstanding f how those institutionsevolve andstabilize, an area
in which theoretically informed country studies are the most promising starting
point.
Conclusion
There have been pitfalls as well as progress in attemptsto understandthe vertical
organizationof governmentusing cross-national data.A first generationof theories
based on welfare economics and public choice has been supplemented with new
approaches hatpay attention o institutionsandincentives,andempiricalstudies are
beginningto follow suit. Early empiricalstudies paid little attentionto the varieties
of fiscal andpolitical decentralization.Yet some of the key findings were useful; fis-
cal decentralization ndfederalismdo not easily translate nto the gains in efficiencyandaccountabilitypredictedby the first generationof theory.The next generationof
empiricalstudies is embracingthe complexity and diversityof decentralizationand
considersthe possibility thatdifferenttypes of decentralizationhave different causes
and effects. The fit between theory and empirical analysis is improving,and there
are severalavenuesfor improveddata collection andanalysis.Cross-nationalempirical analysis also has limits, and the relationship between
quantitativeand case study approachesto comparativepolitics is symbiotic. Some
institutionaldetails, like the incentiveeffects of differenttypes of grants, legislative
organization,and the political relationshipsbetween centraland local officials, mustfirst be analyzed with comparativecase studies before progress can be made in
cross-nationalempiricalanalysis.In general,producersof cross-nationalregressionsshouldbe modest about theirclaims, and consumerscautiousin their interpretation.Cross-nationalmeasurabilityis sometimes conflated with relevance for theory or
even policy, creatingthe dangerthat studies plaguedby poorly conceived or poorlymeasuredconcepts can close minds and move the researchprogram-and perhapseven policy-in unproductivedirections.Cross-nationalregressionsare most useful
whentheybothrespondto andhelp informtruly comparativecase studies.
Finally,as lessons from case studies and cross-national data collection accrue,theories must evolve as well. This evolution is already taking place, as theorists
respondto problemswith accountability,corruption,andmacroeconomicinstabilityin newly decentralizing countries. An importantfurtherstep is to recognize that
decentralizationoverthe last twenty years is not best understoodas a clean transfer
of fixed authorityorresourcesfromhigherto lower level governments,or federalism
as a fixed allocationof spheresof central andprovincial autonomy.A more promis-
ing theoreticalstartingpoint is to analyzethe causes and effects of shared and inter-
twinedfiscal, political,andpolicy authority.
498
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 20/21
Jonathan Rodden
NOTES
1. Ugo Panizza, "On the Determinants of Fiscal Centralization:Theory and Evidence,"Journal of
Public Economics, 74 (October 1999), 97-139; Geoffrey Garrett and JonathanRodden,"Globalization
and Fiscal Decentralization?," n Miles Kahler and David Lake, eds., Globalization and Governance
(Princeton:PrincetonUniversityPress, 2003).2. Hamid Davoodi and Heng-fu Zou, "Fiscal Decentralization and Economic Growth: A Cross-
Country Study," ournal of UrbanEconomics, 43 (March 1998), 244-57; Daniel Treisman,"The Causes
of Corruption: A Cross-National Study,"Journal of Public Economics, 76 (June 2000), 399-457;
Raymond Fisman and Roberta Gatti, "Decentralization and Corruption:Evidence across Countries,"
Journalof Public Economics,83 (March2002), 325-45.
3. Daniel Treisman, "Decentralization and Inflation: Commitment, Collective Action, or
Continuity?," American Political Science Review, 94 (December 2000), 837-57; Erik Wibbels,"Federalismand the Politics of MacroeconomicPolicy and Performance,"AmericanJournal of Political
Science, 44 (October 2000), 687-702; Wallace Oates, "Searchingfor Leviathan:An Empirical Study,"
American Economic Review, 75 (September 1985), 748-57; David Cameron, "The Expansion of thePublic Economy: A ComparativeAnalysis,"American Political Science Review, 72 (December 1978),1243-61.
4. See WallaceOates,Fiscal Federalism New York:HarcourtBraceJovanovich,1972).5. Charles Tiebout, "A Pure Theory of Local Government Expenditures," Journal of Political
Economy(October 1956);Albert BretonandAnthonyScott, TheEconomic Constitutionof Federal States
(Toronto: University of Toronto Press, 1978); Friedrich von Hayek, "The Economic Conditions of
InterstateFederalism,"n Friedrichvon Hayek,Individualismand Economic Order(Chicago: Universityof Chicago Press, 1939); Geoffrey Brennan and James Buchanan, The Power to Tax:AnalyticalFoundationsof a Fiscal Constitution(New York:CambridgeUniversity Press, 1980); Barry Weingast,"The Economic Role of Political Institutions: Market-Preserving Federalism and Economic
Development," ournalof Law, Economics,and
Organization,11 (April 1995),1-31.
6. Treisman,"The Causes of Corruption";Ernesto Stein, "Fiscal Decentralizationand Government
Size in LatinAmerica,"Journal of Applied Economics, 2 (November 1999), 357-91; Wibbels; Davoodi
and Zou.
7. Yingyi QianandBarry Weingast,"Federalismas a Commitment o PreservingMarketIncentives,"
Journalof EconomicPerspectives, 11 (Fall 1997), 83-92.
8. JonathanRodden,"The Dilemma of Fiscal Federalism:Grantsand Fiscal Performancearound the
World,"AmericanJournalof Political Science, 46 (July2002), 670-87.
9. OECD,"TaxingPowersof Stateand Local Government,"OECDTaxPolicy Studies, 1 (1999).10. See JiirgenVon Hagen and Barry Eichengreen, "Federalism, Fiscal Restraints, and European
Monetary Union," American Economic Review, 86 (May 1996); Rodden, "Dilemma of Fiscal
Federalism."
11. VernonHenderson,"TheEffects of Urban Concentrationon Economic Growth,"NBER WorkingPaperW7503.
12. Data codebookavailableat http://econ.pstc.brown.edu/faculty/henderson/.13. WallaceOates, "AnEssay on Fiscal Federalism," ournal of Economic Literature,37 (September
1999), 1124.
14. ChristopherGarman, Stephan Haggard,and Eliza Willis, "Fiscal Decentralization: A Political
Theorywith LatinAmericanCases,"WorldPolitics, 53 (January2001), 205-36.
15. David Samuels, "ConcurrentElections, Discordant Results: Presidentialism, Federalism, and
Governance n Brazil,"ComparativePolitics, 33 (October2000), 1-20.
16. William Riker and Ronald Schaps, "Disharmony n FederalGovernment,"Behavioral Science, 2
499
This content downloaded from 35.8.11.3 on Wed, 24 Jul 2013 10:37:27 AMAll use subject to JSTOR Terms and Conditions
8/13/2019 Rodden 2004
http://slidepdf.com/reader/full/rodden-2004 21/21
ComparativePolitics July2004
(1957); JonathanRodden and ErikWibbels, "Beyondthe Fiction of Federalism,"WorldPolitics, 54 (July
2002), 494-531.
17. WilliamRiker,Federalism:Origin, Operation,Significance (Boston:LittleBrown, 1964).18. Alfred Stepan,"Federalismand Democracy:Beyond the U.S. Model,"Journal of Democracy, 10
(October1999), 19-33.19. Anotherexplanation or malapportionments thatwealthyland and capitalowners seek to insulate
themselves from demandsforredistribution temmingfromurban abor.
20. David Samuels and Richard Snyder, "The Value of a Vote: Malapportionment n Comparative
Perspective,"BritishJournalof Political Science, 31 (October2001), 651-71.
21. See Jacques Cremer and Thomas Palfrey,"Political Confederation,"American Political Science
Review,93 (March1999), 69-93.
22. PrestonKing,FederalismandFederation(London:CroomHelm, 1982), p. 77.
23. Daniel J. Elazar,"From Statism to Federalism:A ParadigmShift," Publius, 25 (Spring 1995),
5-18; Ronald Watts, ComparingFederal Systems (Kingston: Institute of IntergovernmentalRelations,
1999).24.
Oates,Fiscal
Federalism;Panizza.
25. AlbertoAlesina and Enrico Spolaore,"On the Numberand Size of Nations,"QuarterlyJournal of
Economics, 112 (November1997), 1027-56.
26. PatrickBolton and GerardRoland, "The Breakup of Nations: A Political Economy Analysis,"
QuarterlyJournalof Economics, 112 (November1997), 1057-90.27. Fritz Scharpf,Bernd Reissert, and Fritz Schnabel, Politikverflechtung:Theorie und Empirie des
kooperativenFoderalismus n der BundesrepublikKronberg:ScriptorVerlag,1976).28. Andrei Shleifer and RobertVishny, "Corruption,"QuarterlyJournal of Economics, 108 (August
1993), 599-617.
29. Treisman,"TheCauses of Corruption";Fisman andGatti.
30. Daniel Treisman, "Decentralization and the Quality of Government," unpublished paper,
Universityof California,LosAngeles.
31. RobertLowry,JamesAlt, andKarenFerree,"FiscalPolicy Outcomesand ElectoralAccountabilityin the American States,"American Political Science Review, 92 (December 1998), 759-74; Susanne
Lohmann,David Brady,and Douglas Rivers,"PartyIdentification,RetrospectiveVoting,and ModeratingElections in a FederalSystem: West Germany,1961-1989," ComparativePolitical Studies, 30 (August
1997), 420-49.
32. Oates,"Searching or Leviathan";Cameron.
33. Stein.
34. Jonathan Rodden, "Reviving Leviathan: Fiscal Federalism and the Growth of Government,"
InternationalOrganization Fall2003).35. For a review, see Francis G. Castles, Comparative Public Policy: Patterns of Postwar
TransformationCheltenham:EdwardElgar,1999), p. 82.
36. See, for example, JonathanRodden, GunnarEskeland,and Jennie Litvack,eds., Decentralization
and the Challenge of HardBudgetConstraints Cambridge,Mass: MITPress,2003).
500