fiscal decentralization and economic growth in the oecd

This article was downloaded by: [Yale University Library]On: 08 October 2013, At: 12:22Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,37-41 Mortimer Street, London W1T 3JH, UK

Applied EconomicsPublication details, including instructions for authors and subscription information:http://www.tandfonline.com/loi/raec20

Fiscal decentralization and economic growth in theOECDPhilip Bodman aa School of Economics, University of Queensland , Brisbane, 4072 QLD, AustraliaPublished online: 07 Oct 2010.

To cite this article: Philip Bodman (2011) Fiscal decentralization and economic growth in the OECD, Applied Economics,43:23, 3021-3035, DOI: 10.1080/00036840903427208

To link to this article: http://dx.doi.org/10.1080/00036840903427208

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) containedin the publications on our platform. However, Taylor & Francis, our agents, and our licensors make norepresentations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of theContent. Any opinions and views expressed in this publication are the opinions and views of the authors, andare not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon andshould be independently verified with primary sources of information. Taylor and Francis shall not be liable forany losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use ofthe Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematicreproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in anyform to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditions

Applied Economics, 2011, 43, 3021–3035

Fiscal decentralization and

economic growth in the OECD

Philip Bodman

School of Economics, University of Queensland, Brisbane, 4072 QLD,

Australia

E-mail: p.bodman@economics.uq.edu.au

What impact, if any, does Fiscal Decentralization (FD) have on economic

growth? Further investigations of the inter-relationships between FD and

economic growth are timely given that government decentralization

remains at the forefront of many Organization for Economic

Cooperation and Development (OECD) policy agendas. This study

incorporates a range of measures of FD to better account for the direct

impact of different levels of subnational fiscal autonomy on economic

growth. The analysis also considers the impact of previously omitted

public sector decentralization variables that provide further indication of

the extent to which Subnational Governments (SNG) are ‘closer to the

people’ and potentially better able to account for local preferences in fiscal

decision-making. Whilst little evidence of a direct relationship between FD

and output growth is found, some evidence is found to suggest that federal

systems tend to have lower growth rates than do unitary states,

independent of their degree of decentralization, and that countries with

more elected tiers of government generally have lower economic growth.

I. Introduction

FISCAL DECENTRALIZATION is in vogue.Wallace E. Oates (1999)

Fiscal decentralization (FD) is once again a hot topic

amongst policy-makers. Out of 75 developing and

transitional countries with populations greater than

five million, all but 12 claim to have embarked on

some form of government decentralization initiative

over the last decade.1 FD has also featured in the

recent policy agendas of many Organization for

Economic Cooperation and Development (OECD)

countries. For example, in the US, the Central

Government (CG) has returned significant portions

of federal authority to the states particularly con-

cerning areas such as welfare, Medicaid, education

and job training. In the UK, under the Blair

government, both Scotland and Wales opted for

their own regional parliaments.2 Australia has

recently experienced vigorous debate over the need

for reforms to its Commonwealth-State fis-

cal arrangements to improve the supply-side of

its economy, with the complex cobweb of

1 See Dillinger (1994).2 Similar responses to demands for greater regional autonomy in countries such as Belgium, Canada, Italy and Spain have ledto a serious push for increased decentralization of government structure (Patsouratis, 1990). In the EU, more generally, threehas been a significant transfer of power from national governments to both supranational authorities as well as to local andregional governments – the so-called ‘Sandwich hypothesis’ (Zimmermann, 1990).

Applied Economics ISSN 0003–6846 print/ISSN 1466–4283 online � 2011 Taylor & Francis 3021http://www.informaworld.com

DOI: 10.1080/00036840903427208

intergovernmental relations accused of dragging backproductivity, creating export bottlenecks, infrastruc-ture shortages and even crisis in its health andeducation systems.3

But what exactly do economists mean by FD? In itsmost basic form, FD refers to the division ofbudgetary responsibilities between different levels ofgovernment. Of course, true decentralization is notsimply a geographical de-concentration of the CG’sbureaucracy or service delivery, rather it is essentiallylinked to the territorial distribution of power. TheCommonwealth Secretariat (1985) defines the decen-tralization of government as ‘the transfer of powerand/or authority to plan, make decisions and/ormanage public functions from a higher level ofgovernment to a lower one’.

Therefore, for the purpose of this article, FD refersto the amount of independent decision-making powerinvolved in subnational expenditure and revenuedecisions. The term ‘subnational’ collectively standsfor the levels of government below the nationalgovernment, both lower level governments (munici-palities, communes or local councils) and intermedi-ate tiers (regions, states, provinces, counties,territories or districts). ‘Decentralization’ is used inthe static sense to describe systems in whichresponsibilities are divided among tiers, rather thanin the dynamic sense of becoming ‘decentralized’.Thus, the extent of FD depends on the ability of lowerlevels of government to make independent revenue andexpenditure decisions regarding the provision of publicgoods and services within a geographic domain, withoutinterference by the CG.

Traditional discussions about the normative designof FD, and analyses of how decentralized systems

perform in practice, were not concerned with theeffects of FD on economic growth. Rather,Musgrave (1959) stated that the three main objectivesof government regarding public finance were effi-ciency, income redistribution and macroeconomicstability. The literature concerning FD and economicgrowth implicitly assumes that FD affects the growththrough its impact on these three factors. Therefore,the issue is whether or not changes in effi-ciency, macroeconomic stability and income redistri-bution resulting from increased (or decreased) FDhave a statistically significant impact on economicgrowth.

In summary, FD may be growth-enhancing if itleads to increased efficiency in the supply of publicgoods by considering local preferences (consumerefficiency)4 or if it leads to subnational innovations,cost reductions and productivity improving intergo-vernmental competition (producer efficiency).5

Alternatively, growth-impeding hypotheses can bepostulated from the suggestion that FD might leadto harmful competition, macroeconomic instability6

and a more unequal distribution of resources.7

Hence, the relationship between FD and economicgrowth is ambiguous.

Whilst there has been a huge amount of theoreticalliterature using endogenous growth models to explainthe significant effects government policies can haveon economic growth, there has been very little in theway of theoretical modelling of the direct relationshipbetween FD and economic growth.8 There has beenlittle effort to formally define the links between FDand growth via the channels of efficiency, macro-economic stability and equity and the modellingthat has been proposed is quite inadequate

3 See, for example, Garnaut (2006) and Twomey and Withers (2007).4 The notion that FD may affect efficiency is underpinned by the seminal work of Oates (1972). See Martinez-Vazquez andMcNab (1997) for a recent discussion.5Martinez-Vazquez and McNab (1997) and Thiessen (2000) assert that FD can foster experimentation and innovation in theproduction and supply of public goods. If such innovation leads to greater producer efficiency, then the higher quantity orquality of public goods could eventually result in increased income and therefore measured economic growth. However, thisnotion remains controversial. Bahl and Linn (1992) point out that greater centralization improves producer efficiency if agiven public service entails economies of scale or scope. Prud’homme (1995) argues that central bureaucracies operate closerto the technical production frontier as they tend to attract more qualified people due to greater career opportunitiesand have more scope to invest in technology, research, development and innovation. More recently, Adam et al. (2008)provide evidence in favour of a positive relationship between expenditure decentralization and the productive efficiencyof governments in the OECD, whilst Barankay and Lockwood (2007) find similar conclusions using a dataset of Swisscantons.6 The links between macroeconomic stability and FD have been highlighted in McLure (1995), Prud’homme (1995), Tanzi(1995, 2008), Sewell (1996) and Martinez-Vasquez and McNab (2005). A related literature has also been developing on‘soft budget constraints’ that explores in considerable depth the ways in which FD can, and potentially has, underminedfiscal performance in certain countries – see, for example, the volume edited by Rodden and Eskeland (2003).7 See Prud’homme (1995).8 Brueckner (2006) is an exception. In this article, an overlapping generations model is used to illustrate how decentralizationcan potentially lead to efficiency gains that lead to greater saving and capital accumulation and ultimately faster economicgrowth.

3022 P. Bodman

(see Martinez-Vasquez and McNab (2003) for a

discussion).9

Of course, one should not necessarily expect amonotonic relationship between FD and growth.

That is, it may not be true that the moredecentralized a country’s fiscal system becomes, the

faster (or slower) that country grows. Rather,Thiessen (2000) and Eller (2004, p. 30) argue that

there is an optimal degree of FD, usually thought tobe some ‘medium degree’, that is less than complete

decentralization. Low levels of FD may not provideenough incentive for Subnational Governments

(SNGs) to improve allocative and productive effi-ciency. In this case, the fixed costs of maintaining

SNGs may outweigh any benefits in terms ofefficiency, hindering economic growth. Too much

FD may lead to macroeconomic instability andinequality, also having a negative impact on eco-

nomic growth in the long run.So given that one of the main stated objectives of

FD in many countries is to promote economic

growth, and given the current policy interest in

fiscal reforms in many developed countries, studies

of the relationship between FD and economic growth

are timely. It remains an open question as to whether

FD actually plays a statistically significant role in

enhancing or inhibiting economic growth. The few

previous estimates of the relationship between FD

and economic growth that are provided in the

literature are largely contradictory, and hence do

not resolve the potential theoretical ambiguity. Of the

cross-country studies, Oates (1995), Yilmaz (1999),

Iimi (2005) and Ding (2007) find a positive relation-

ship between FD and output growth. Davoodi and

Zou (1998) find a negative relationship for developing

countries and none for developed countries.

Similarly, Woller and Phillips (1998) find a negative

correlation for developing countries whilst Carrion-

i-Silvestre et al. (2008) find no effects of revenue

decentralization on growth and significant negative

effects of expenditure decentralization for Spain.

Thiessen’s (2000, 2003) and Eller’s (2004) results

support the hypothesis that a medium degree of FD

tends to best promote economic growth in OECD

countries. Similar evidence for the states of the US is

found by Akai et al. (2007).

II. Methodology

There is no clear theoretical framework to guideempirical work on the relationship between FD andeconomic growth. As Levine and Renelt (1992) pointout, there is no model that completely specifies thefactors that one should hold constant while conduct-ing statistical inference on the relationship betweengrowth and the economic variable of interest. Themost common approach in studies of FD andeconomic growth is to run informal growth regres-sions.10 The basic framework adopted by authorssuch as Davoodi and Zou (1998) and Iimi (2005), anda number of single country studies, uses a simpletheoretical model to justify the inclusion of FD ingrowth regressions. Other control variables arechosen from results in the literature, typicallyincluding those used by Levine and Renelt (1992) –the initial level of real Gross Domestic Product(GDP) per capita, the population growth rate, theinitial secondary school enrolment ratio (as a proxyfor initial human capital) and the investment to GDPratio.

Thiessen’s (2000, 2003) cross-country studies of FDand growth, and Lin and Liu’s (2000) analysis ofChina, use informal growth regressions loosely basedon the Mankiw et al. (1992) economic growth model.As these regressions include the investment ratio andinitial income, the authors insist that they can beinterpreted in terms of the Mankiw–Romer–Weil(MRW) model. However, this extension is not perfectas terms from the theoretical model, such as the initiallevel of efficiency, are omitted from the estimatedequation as they are unobservable or difficult toapproximate. The omitted variable problem meansthat if one or more regressors are correlated withthese terms, the parameter estimates will be biased.For instance, if countries with relatively low levels ofinitial efficiency tend to have low levels of decentral-ization, this correlation may mean that FD takes anegative sign when entered into a growth regression,even if it has no long-run effect on output. Therefore,Temple (1999) asserts that such informal growthregressions neglect important insights in the absenceof a formal theoretical derivation.

This article extends Thiessen’s (2000, 2003) analysisof the relationship between FD and economic

9 The quality of local government officials relative to CG officials and the extent of corruption at the local level also haveimplications for producer and consumer efficiency. However, these factors are more likely to be important for developingcountries than for the sample of OECD countries examined in this article.10An alternative approach, used in only in one previous study of FD and growth, is based on cross-country growthaccounting. This approach has a number of shortcomings. In particular, whilst this model considers the efficiency channelsthrough which FD affects growth, it can still be criticized on the grounds that it does not allow one to distinguish betweenconsumer or producer efficiency or determine FDs impact on growth through the channels of macroeconomic stability orinequality. See Eller (2004) for a discussion.

Fiscal decentralization and economic growth in the OECD 3023

performance in three useful ways. First, this articleuses static panel regressions with fixed time andindividual effects as well as pure cross-sectionregressions.11 Second, the analysis uses the newmeasures of FD developed by Stegarescu (2005),and includes more disaggregated ‘hump-shaped’indicators of FD. Third, the analysis includes anumber of other measures of government decentral-ization, omitted from previous studies of FD andgrowth that potentially provide a better indication ofthe extent to which SNGs are ‘closer to the people’and are therefore better able to account for localpreferences in fiscal decision-making and whichbetter differentiate between ‘administrative’ and‘substantive’ fiscal autonomy.

III. Data and Measurement

There is no single, or simple, measure of FD. FD is‘so multidimensional that specification of a formalhypothesis for statistical testing requires steppingdown from a view of the general picture, to a levelwhich provides only a narrow slice of the panorama’(Guess et al., 1997, p. 1). The first dimension of FDconsidered in this study concerns the formal divisionof expenditures and revenues between levels ofgovernment. The second, and most important dimen-sion, is the extent to which fiscal decision-making isdecentralized.

Measures used in previous analysis

The primary way of measuring FD is the ‘budgetdata’ approach. Previous cross-country studies of FDand growth have used budget data measurementsbased on the Government Financial Statistics (GFS)of the International Monetary Fund (IMF). Moststudies use the subnational share of General

Government (GG) expenditures (EXP) or revenues

(REV) as a proxy for decentralization.12 However,

three main deficiencies of the GFS are identified inthe literature.13 First, the GFS provides a breakdown

of revenues and expenditures by type and function,

but they are reported at the level of government that

receives or operates them, irrespective of whether ithas discretion over them. Thus, local expenditures

that are directed by the CG are included in

subnational expenditure. Second, it does not identifythe sources of revenues, and no distinction is made

between locally determined own taxes, piggybacked

or shared taxes. Third, it does not disclose the

proportion of intergovernmental transfers that areconditional or the criteria (objective or discretionary)

by which transfers are distributed. Therefore,

although GFS data has consistent definitions acrosscountries over time, it ignores the degree of CG

control over local revenues and expenditures. These

measurement errors mean that the degree of FD

tends to be overestimated.More general criticisms of the budget data

measurements include the fact that they do not reflect

restraints on local fiscal autonomy arising from

legislation, regulation, norms, minimum quality stan-dards and other qualitative restrictions imposed by

the CG. SNGs that have the autonomy to decide the

amount and type of tax to collect, and to determine

the allocation of their expenditure, are more decen-tralized than those whose spending and revenue is

determined by national legislation. Further, changes

in budget data measures over time do not necessarilyreflect changes in SNG autonomy. Stegarescu (2005)

claims that the tax bases of national and SNGs

typically have different elasticities. Therefore, busi-

ness cycles cause automatic fluctuations in the reve-nue indicators, even though the assignment of

competencies remains unchanged.Another problem with existing measures of FD is

that they aggregate all SNGs into a single group.

11 The advantages of using panel data or pooled cross-section regressions over pure cross-section regressions are welldocumented – they provide ‘more informative data, more variability, less collinearity among the variables, more degrees offreedom and more efficiency’ (Baltagi, 1996, p. 4). Therefore, it has the potential to produce more reliable parameterestimates. Panel data has the added advantage of allowing one to control for omitted variables that are persistent over time.Including country specific effects allows one to better control for individual heterogeneity. The inclusion of country and timespecific effects may also be necessary to inhibit correlation between the regressors due to contemporaneous time or countryshocks. Without these effects, there may be parameter heterogeneity, which could lead to meaningless estimates. Tests forparameter heterogeneity reject the hypothesis that the intercept can be held constant over countries and time, so this analysis,unlike Thiessen’s (2000, 2003), uses a static (within) panel framework with fixed time and individual effects.12 For example, Oates (1995), Davoodi and Zou (1998), Woller and Phillips (1998), Xie et al. (1998), Thiessen (2000, 2003),Eller (2004) and Iimi (2005). Oates (1995) and Thiessen (2000, 2003) also consider self-reliance ratios, the share of SNGsown revenues in their total revenues. Woller and Phillips (1998) use the share of SNG revenues, less grants, in totalgovernment revenue and construct an expenditure share subtracting defence and social security spending. They argue thatthese provide a better of indication of the revenues and expenditures that could ‘in principle’ be the responsibility of eitherlevel of government.13 For example, Ebel and Yilmaz (2002) and Stegarescu (2005).

3024 P. Bodman

This horizontal aggregation does not take intoaccount the number of participating SNGs and thedifferences in competencies between them. Bahl andLinn (1992) point out that the data does not indicatewhether subnational revenues and expenditures areconcentrated in one or two jurisdictions or evenlydistributed across all areas. The degree of fiscalautonomy may also differ between subnationaljurisdictions.14 A more correct measure of FDmight consider the horizontal disaggregation offiscal data by jurisdiction. The main difficulty withthis involves finding indicators that are comparableacross countries.

Stegarescu’s indicators of FD

Stegarescu (2005) provides six new indicators of taxand revenue decentralization that attempt to capturedifferent levels of subnational autonomy. Theseindicators have not previously been used in a studyof FD and growth.15 The measures are based on theOECD (1999) surveyTaxing Powers of State and LocalGovernment. The survey classifies subnational taxes indecreasing order of fiscal autonomy according to threecriteria: legislative abilities to determine the tax baseand tax rate; the attribution of tax receipts and taxadministration. Based on the four-digit classificationof taxes by tax base reported in the annual OECDRevenue Statistics, the survey classifies each tax foreach country according to the degree ofdecision-making autonomy as presented in Table 1.

In cases (a)–(c), referred to as ‘own taxes’, theSNGs have total or significant control. In the case ofthe revenue sharing categories (d.1)–(d.2), the SNGshave limited influence. For categories (d.3)–(e), theyhave no control.

Stegarescu (2005) uses these classifications, theOECD Revenue Statistics, GFS data and 23 compre-hensive surveys of national financial laws andconstitutions, to create new indicators of tax andrevenue decentralization for 23 countries.16 Heprovides time-series data on ‘own taxes’ (1965–2001)and ‘own revenues’ (1975–2001), adjusting the classi-fication of autonomy for each SNG tax on an annualbasis according to changes in legal provisions.17

These measures indicate a significant trend towards

decentralization amongst most OECD countries, with

the median degree of FD increasing by around 30% in

terms of own (self-financed) expenditures (from

around 25% to over 31%) and by around 70% in

terms of tax autonomy (from around 12–21%).Stegarescu’s (2005) three measures of tax revenue

(tax-only) decentralization are subnational own tax

revenue (TDEC1), subnational own and shared taxes

(TDEC2) and total subnational tax revenue

(TDEC3), all calculated as the share of GG tax

revenue. Own taxes refer to those taxes for which the

SNG can determine the tax rate or tax base or both.Stegarescu (2005) also provides two measures of

expenditure decentralization. These measures are

based on total subnational expenditure and lending,

minus loan repayments, as a percentage of consoli-

dated GG expenditure, without social security and

EU payments. EDEC1 excludes transfers to other

levels of government, whereas EDEC2 includes

transfers to other levels of government net of received

transfers. Again, none of these measures have been

used in previous studies of FD and growth.Unfortunately, it must still be recognized that these

new measures of fiscal (revenue and expenditure)

decentralization do not provide a remedy to all the

limitations of budget data discussed earlier.

The degree of subnational fiscal autonomy is based

on the provisions fixed in legislation, and actual

implementation is not taken into account. Therefore,

the measures only indicate the potential degree of

Table 1. OECD classification of taxes (in decreasing order

of control over revenue sources)

(a) SNG determines tax rate and tax base(b) SNG determines tax rate only(c) SNG determines tax base only(d) Tax sharing:(d.1) SNG determines revenue-split(d.2) Revenue-split only changed with consent

of SNG(d.3) Revenue-split unilaterally changed by CG

(fixed in legislation)(d.4) Revenue-split unilaterally change by CG

(in annual budgetary process)(e) CG determines tax rate and tax base

Source: Stegarescu (2005) and OECD (1999).

14 This is particularly a concern with regions of special status in France, Italy and Spain, and for Scotland, Wales andNorthern Ireland in the UK.15 Stegarescu (2009) does utilize these FD measures in a study of the effects of political integration on FD in the OECD.16 The survey covers Austria, Belgium, the Czech Republic, Denmark, Finland, Germany, Hungary, Iceland, Japan, Mexico,Netherlands, New Zealand, Norway, Poland, Portugal, Spain, Sweden, Switzerland and the UK. Stegarescu (2005) does notcalculate measures of FD for the Czech Republic, Hungary, Mexico or Poland, but adds measures for Australia, Canada, US,France, Greece, Italy, Ireland and Luxembourg.17 There are gaps in the time-series for most countries before 1980. After 1980s there are missing data points for Greece,Iceland, Italy, Japan, New Zealand and Switzerland.

fiscal autonomy and may still overestimate actualFD. The criticisms about horizontal aggregation arestill valid. Further, there is some measurement errorin these indicators for the countries that Stegarescu(2005) adds to the OECD survey.18 Nonetheless,these measures provide a solid step towards betteraccounting for the varying degrees of autonomy thatSNGs have over their taxes and revenues.

Hump-shaped indicators of FD

Using these measures of FD, how does one test thehypothesis that a medium degree of decentralization isbest for economic growth? In previous work, Thiessen(2000, 2003), Eller (2004) and Akai et al. (2007) havefound significant support for this hypothesis using‘hump-shaped’ indicators based on the traditionalbudget data measures (the subnational share of GGrevenues and expenditures). In this study, hump-shaped indicators were also created for each ofStegarescu’s (2005) new decentralization measures.

A number of different approaches were tried in theconstruction of hump-shaped indicators. Followingthe divisions outlined in Thiessen (2000, 2003), threedummy variables were created for high, medium andlow levels of FD. Similarly, dummy variables werecreated following Eller (2004) using equal-sized high,medium and low FD groups. Finally, countries weredivided into five equal-sized groups, denoting verylow, low, medium, high and very high decentralization.Since these indicators proved to be time-invariant formany countries, more disaggregated hump-shapedindicators were constructed by ranking each country,for each time period, from the lowest level ofdecentralization to the highes, and assigning numbersthat increase towards the median value (such that thecountry and time period with the highest and lowestlevels of FD were given the number one, the secondhighest and lowest levels of decentralization wereassigned the number two and so on).19 The latter

approach was used to construct the indicators used inboth the cross-section and panel analyses reported inthis article. 20

Other measures of public sector decentralization

Finally, a number of other measures of governmentdecentralization, omitted from previous studies ofFD and growth, are considered. General decentral-ization of the public sector invokes a number ofnotions, each of which involves vertical decentraliza-tion, that is, the division of government into anumber of tiers. Appointment, electoral and person-nel decentralization refer, respectively, to the level atwhich government officials are appointed and dis-missed, whether there are democratic elections ateach level, and the share of administrative personnelemployed at the subnational level. Bahl (1999)suggests that the true extent of FD may be describedby a number of such factors, ranging from electedlocal councils, locally appointed officers, institutionalprovisions, the size and number of subnationalauthorities and their organizational structures.These more general aspects of public sector decen-tralization provide some indication of the ability ofSNGs to respond to local constituencies in their fiscaldecision-making. Use of these broader measures ofdecentralization further emphasizes the point madeby Thornton (2007) that high sub-national revenueand spending shares do not necessarily mean highlocal autonomy – the differentiation between ‘admin-istrative’ and ‘substantive’ fiscal autonomy.21

Therefore, this article includes a number ofindicators of public sector decentralization. Thenumber of subnational jurisdictions in the intermedi-ate and lower tiers of government is considered. Twocountries may have the same subnational share ofexpenditures or revenues but different numbers ofparticipating SNGs. More participating units,ceteris paribus, would imply more FD.22 Then, an

18Note that there is significant measurement error for Australia. Stegarescu (2005) assigns Australian states completeautonomy over all sources of tax revenue, that is, grants from the Commonwealth Grants Commission are consideredautonomous own taxes. The measures do not reflect Australia’s large vertical fiscal imbalance, with the Commonwealthspending only about a third of the tax revenue it raises. Although most of the gap between subnational expenditure andrevenues is filled by grants from the Commonwealth Grants Commission (which may be considered general purpose grantsbased on objective criteria), purpose-specific grants are not considered. In both cases the subnational governments do notcontrol the tax rates or tax bases, and thus including them in TDEC1 and RDEC1 raises questions about measurement errorfor this country.19 The main problem with this measure, and those based on equal shares, is that what constitutes a ‘medium degree’ of FDdepends on the country sample.20 This approach is similar to that derived in the labour economics literature on centralization of wage bargaining by Calmforsand Driffill (1988). The medium degree of FD was postulated to be a subnational share of 30–45%.21 Thornton (2007) provides evidence that previous studies have used measures that have substantially over-estimated theextent of revenue decentralization in practice, and that when more appropriate measures are used the effect of revenuedecentralization on economic growth is not statistically significant.22 This was pointed out by Bahl and Nath (1986, p. 407) but has not been considered in an empirical study of FD and growth.

3026 P. Bodman

indicator was included to account for electoraldecentralization, taking the value of zero if there areno subnational elections, one if either local orintermediate tiers of government are elected, or twoif both are subject to elections.23 Subnational electionsprovide some indication of the ability of consumers toexpress their preferences to different levels of govern-ment and the incentive for governments to respond tothose preferences. The indicator of constitutionalstructure used in the cross-section analysis was takenfrom Lijphart (1999), provided in Armingeon et al.(2002) dataset.24 This is an index of federalism ona five-point scale; (1) unitary and centralized, (2) uni-tary but decentralized, (3) semi-federal, (4) federal butcentralized and (5) federal and decentralized. Infederal countries, SNGs are more likely to havea permanent right to govern their own affairs. Thesethree variables are largely time-invariant and hencethey are not included in the panel analysis.

Resource decentralization is considered usingthe ratio of SNG employees to CG employees,using data provided by Schiavo-Campo et al.(1997). Unfortunately, this data is not readily avail-able over time. These four general aspects of publicsector decentralization provide some indication of theextent to which SNGs are ‘closer to the people’ andare therefore better able to account for localpreferences in fiscal decision-making.

Other variables

The empirical analysis uses 3-year averages based onannual data (see Appendix for data sources). Growthis measured by the change in the natural log of realGDP per capita in constant local currency units andincome levels are measured using Purchasing PowerParity (PPP) comparisons.25 Both are expressed inper capita terms. The Perpetual Inventory Method(PIM) was used to construct a measure of the capitalstock from Gross Fixed Capital Formation (GFCF),following De la Fuente and Domenech (2000).

In this study, the level of human capital is proxiedby average years of schooling. Unfortunately, thisdata is only available in 5-year increments and thuswas interpolated for the panel analysis. Given this,however, calculating the growth rate of humancapital from this series may be misleading. Since a

more consistent annual time series is available for the

secondary school enrolment ratio, the growth rate of

human capital is proxied by the growth rate of the

secondary school enrolment ratio. It is acknowledged

that there is likely to be some measurement error.

These are typical problems encountered in the

empirical growth literature. Authors such as De la

Fuente and Domenech (2000, p. 1) note that poor

measurement and data quality means that educa-

tional variables frequently turn out to be insignificant

or have the ‘wrong’ sign in growth regressions. Levine

and Renelt (1992, p. 945) acknowledge that incorrect

measures of human capital may induce biased results.

However, they still include the secondary school

enrolment ratio in their growth regressions, arguing

that some measure of human capital is required and

other measures produce similar results. Most previ-

ous studies of FD and growth have included the

secondary school enrolment ratio.To control for macroeconomic disturbances, exter-

nal shocks and structural rigidities that could impact

both FD and growth, the GDP deflator is included.

As Eller (2004) points out, when studying the impact

of FD on economic growth, it is also important to

control for the size of the public sector. Therefore,

GG final consumption expenditure, as a percentage

of GDP, is incorporated into the regressions. Other

control variables used in this article include the

vertical fiscal imbalance (transfers to SNGs as a share

of SNG expenditures), the SD of domestic credit

growth (used as a proxy for the uncertainty of

financial variables) and a measure of political free-

dom (countries allocated a number between 1 and 2.5

are considered ‘free’, between 3 and 5.5 ‘partly free’

and above 5 are considered ‘not free’).26

IV. Empirical Results

Cross-sectional analysis

Formal testing of the hypothesis that FD affects eco-

nomic growth begins with the form in Equation 1, i.e.

D lnRGDPCi ¼ D lnAiþ�D ln kiþ�D ln hi ð1Þ

23 France, Germany, Ireland, Italy, New Zealand, Poland and US were given a value of three as they have three electedsubnational tiers.24 Studies of FD and economic growth have not considered a federalism dummy, although Yilmaz (1999) implements separateregressions for federal and unitary countries.25 This follows Nuxoll’s (1994) argument that one should use PPP comparisons to measure the level of GDP and domesticnational accounts data to calculate and compare growth rates.26 The results for the additional conditioning variables are not reported in the tables below for purposes of brevity but areavailable from the authors by request.

where RGDPC is real GDP per capita, A is efficiency,

k is capital per capita and h is the measure of human

capital. Equation 2, determining the growth rate of

efficiency, implies that efficiency depends on various

measures of public sector decentralization, such that

D lnAi ¼ �A0þ �A1 FDi þ �A2 FUi þ �A3 NSGVTi

þ �A4 ELECTi þ �A5 EMPLOYi ð2Þ

where FU is the federalism dummy, NSGVT is the

number of SNG units, ELECT is the number of

elected subnational tiers of government and

EMPLOY is the subnational to CG employee ratio.Substituting Equation 2 into Equation 1 gives

D lnRGDPCi ¼ �A0þ�A1FDiþ �A2FUiþ�A3NSGVTi

þ �A4ELECTiþ�A5EMPLOYi

þ�1D lnkiþ�2D lnhi ð3Þ

Further, control variables are added to Equation 3

to absorb the effect of the size of the government and

macroeconomic shocks. The estimated equation is

outlined in Equation 4. GCGDP is the government

consumption to the GDP ratio and DEF is the GDP

deflator.

D lnRGDPCi

¼ �A0þ�A1FDiþ �A2FUiþ�A3NSGVTi

þ �A4ELECTiþ�A5EMPLOYiþ�1D lnki

þ�2D lnhiþ�3D lnGCGDPiþ�4D lnDEFiþ "i

Results based on Equation 4 using cross-sectional

data are presented below.27 The number of SNGs, the

subnational employee to central employee ratio and

the GDP deflator are found to be consistently

insignificant and hence are omitted from all of thetables.28 The growth accounting components, FD

and other significant explanatory variables are

reported in Tables 2 and 3 for the linear and hump-

shaped indicators, respectively. No measure of FD,

linear or hump-shaped, is found to be significant.Interestingly, in all cases the federalism indicator is

negatively related to growth. Therefore, in this samplefederal decentralized countries tend to have lower

growth rates than unitary decentralized countries. The

analysis also suggests that countries with more elected

subnational tiers of government generally have lower

economic growth rates. There are many possible

reasons for this result. A higher number of electedtiers of government allows residents greater opportu-

nity to express their preferences. As mentioned earlier,

the level and mix of public goods demanded by local

residents may not necessarily be the level that

maximizes economic growth.29

The growth rate of the capital stock and the growth

rate of government consumption have the expectedsigns. Growth in the secondary school enrolment

ratio, used to proxy for human capital growth, is

consistently insignificant.Therefore, there is no evidence of a direct relation-

ship between FD and economic growth, using cross-

sectional data for this group of 18 OECD countries.

27 To keep the sample constant across the cross-section equations, Japan and Greece were excluded due to missing data for therevenue indicators. Iceland is excluded due to missing data for ELECT.28One empirical issue that was considered before analysing the relationship between FD and economic growth concerned thepotential endogeneity of FD to the growth process. A significant body of empirical literature suggests that the level of incomeis a determinant of FD. Development stimulates demand for variety and quality in the range of public services being providedwhilst increasing the revenue raising capacity of governments, making decentralization affordable. If FD has a high-incomeelasticity, then higher income per capita may allow the constitution of a new level of decentralization. If FD affects economicgrowth, then the new level of decentralization will in turn have an impact on the level of income. This suggests a potentialbi-directional relationship between FD and economic growth. Further, unobservable and omitted variables that tend tosimultaneously affect both decentralization and economic growth may also exist. If this is the case, then simply including FDin a growth regression could lead to simultaneity bias. Bruess and Eller (2004), Eller (2004) and Iimi (2005) acknowledge thispossibility but do not formally test the hypothesis. Hausman tests for simultaneity were conducted for all specifications in theempirical analysis and the null hypothesis that there is no simultaneity associated with the FD variable cannot be rejected.This was the case for both the cross-section and panel regressions.

Additionally, cross-section informal growth regressions necessarily omit the initial level of efficiency. If FD is correlatedwith the initial level of efficiency (which is likely), it will be correlated with the error term in the cross-section regressions. Thegrowth accounting framework, by including variables in their growth rates, has no need for a term in initial efficiency, andinterestingly FD is not found to be correlated with the error term. Nevertheless, the Hausman test results are the same in thepanel framework, where the need for a term in initial efficiency in the informal growth regressions is eliminated by the timedimension. Given that tests do not suggest that simultaneity is a significant problem, the study proceeds with this approach.29Another possible explanation is based in political business cycle theory. Roubini and Sachs (1989) and Alesina and Tabellini(1990) argue that political systems with frequent changes in political power generally have larger government employment,spending, deficits and debt. If more subnational elections lead to a higher turnover of politicians at lower levels, this maydecrease politicians’ concern about the long-term consequences of their actions, leading to overspending, higher deficits andhigher debt. This would tend to lower growth. Further, such electoral decentralization may have a negative impact onmacroeconomic stability, and therefore growth. Nonetheless, such regressions do not allow one to determine why thisnegative relationship exists.

3028 P. Bodman

2.Growth

accounting:OECD

Measure

C0.028***

9.299***

0.0317***

7.107***

0.03192***

0.0313***

0.0317***

0.029***

0.031***

(4.495)

(4.970)

(4.641)

(4.955)

(4.882)

(4.7587)

(4.932)

(4.471)

(4.984)

[0.001]

[0.000]

[0.001]

[0.000]

[0.001]

[0.000]

0.0002

3.95E-05

6.48E-05

1.26E-05

4.17E-05

7.43E-05

5.27E-05

0.0002

0.0001

(1.683)

(0.309)

(1.159)

(0.149)

(0.472)

(1.150)

(0.472)

(1.751)

(1.396)

[0.121]

[0.763]

[0.271]

[0.884]

[0.646]

[0.275]

[0.646]

[0.108]

[0.1903]

�0.004**

�0.002*

�0.002**

�0.003**

�0.003*

�0.003***

(�2.646)

(�1.806)

(�3.020)

(�2.366)

(�2.413)

(�2.995)

(�2.102)

(�3.651)

(�3.348)

[0.023]

[0.098]

[0.012]

[0.037]

[0.034]

[0.012]

[0.059]

[0.004]

[0.007]

�0.005*

�0.004*

�0.004

�0.004*

�0.004

�0.004*

(�1.935)

(�1.839)

(�1.796)

(�1.918)

(�1.903)

(�1.792)

(�1.889)

(�1.847)

(�1.890)

[0.079]

[0.098]

[0.100]

[0.082]

[0.084]

[0.101]

[0.086]

[0.092]

[0.085]

Dln(k)

0.446***

0.349**

0.357***

0.339***

0.357***

0.358***

0.357***

0.365***

0.362***

(3.648)

(3.081)

(3.821)

(3.127)

(3.354)

(3.999)

(3.544)

(4.108)

(4.519)

[0.004]

[0.012]

[0.003]

[0.009]

[0.006]

[0.002]

[0.005]

[0.002]

[0.001]

Dln(h)

�0.038

�0.028

�0.034

�0.021

�0.027

�0.034

�0.028

�0.047

�0.036

(�0.476)

(�0.391)

(�0.531)

(�0.324)

(�0.417)

(�0.546)

(�0.451)

(�0.682)

(�0.601)

[0.643]

[0.703]

[0.606]

[0.752]

[0.685]

[0.596]

[0.661]

[0.509]

[0.560]

C–GDP)

�0.314**

�0.332**

�0.333**

�0.339**

�0.345**

�0.329**

�0.342**

�0.275*

�0.297**

(�2.403)

(�2.483)

(�2.329)

(�2.526)

(�2.535)

(�2.321)

(�2.519)

(�2.102)

�2.368

[0.035]

[0.030]

[0.039]

[0.028]

[0.041]

[0.029]

[0.059]

[0.037]

Adj.R2

F-value

[0.023]

[0.069]

[0.056]

[0.072]

[0.067]

[0.057]

[0.067]

[0.039]

[0.051]

Notes:Ordinary

Squares(O

LS):Whiteheteroscedasticity-consistentSEsandcovariance.t-statisticsare

inparentheses

andp-values

insquare

brackets.

aDlnRGDPC

i¼�A0þ� A

iþ� A

4ELECTiþ�1Dlnkiþ�2Dlnhiþ�3DlnGCGDPiþ" i.

*,**and***indicate

significance

atthe10,5and1%

levels,respectively.

3.Growth

accounting:OECD:hump-shaped

indicators

Measure

HEDEC1

HEDEC2

HTDEC1

HTDEC2

HTDEC3

HRDEC1

HRDEC3

C0.0268***

0.032***

0.0318***

0.029***

0.032***

0.029***

0.028***

0.029***

(3.777)

(4.071)

(4.833)

(4.207)

(4.097)

(4.519)

(4.329)

(4.113)

(3.571)

[0.003]

[0.002]

[0.001]

[0.002]

[0.001]

[0.002]

[0.004]

0.0005

0.0001

0.0005

0.0001

0.0005

0.0006

0.0005

(1.311)

(0.313)

(0.357)

(1.123)

(1.060)

(0.287)

(1.499)

(1.348)

(0.757)

[0.217]

[0.759]

[0.728]

[0.286]

[0.312]

[0.779]

[0.162]

[0.205]

[0.465]

�0.002**

�0.002***

�0.0018***

�0.002*

(�3.030)

(�2.766)

(�3.063)

(�2.953)

(�3.037)

(�2.704)

(�3.305)

(�3.248)

(�2.198)

[0.011]

[0.018]

[0.011]

[0.013]

[0.011]

[0.021]

[0.007]

[0.008]

[0.052]

�0.003

�0.004

�0.004*

(�1.329)

(�1.738)

(�1.915)

(�1.950)

(�1.932)

(�1.882)

(�1.861)

(�1.798)

(�1.847)

[0.211]

[0.110]

[0.082]

[0.077]

[0.079]

[0.087]

[0.089]

[0.099]

[0.092]

Dln(k)

0.321***

0.319***

0.330***

0.3278***

0.304***

0.334***

0.326***

0.338***

0.3486***

(3.688)

(4.218)

(3.869)

(3.934)

(3.980)

(3.868)

(3.964)

(3.877)

(3.688)

[0.004]

[0.001]

[0.003]

[0.002]

[0.003]

[0.002]

[0.003]

[0.004]

Dln(h)

�0.018

�0.015

�0.012

(0.22)

(�0.264)

(�0.227)

(0.104)

(0.186)

(�0.169)

(0.092)

(0.181)

(0.285)

[0.823]

[0.796]

[0.825]

[0.919]

[0.856]

[0.869]

[0.928]

[0.859]

[0.781]

C–GDP)

�0.316*

�0.318**

�0.330**

�0.265**

�0.297**

�0.328**

�0.321**

�0.302**

�0.304**

(�2.197)

(�2.392)

(�2.560)

(�2.566)

(�2.859)

(�2.447)

(�2.523)

(�2.343)

(�2.591)

[0.050]

[0.036]

[0.027]

[0.026]

[0.016]

[0.032]

[0.028]

[0.039]

[0.025]

Adj.R2

F-value

[0.045]

[0.069]

[0.043]

[0.039]

[0.069]

[0.036]

[0.032]

[0.054]

Notes:OLS:Whiteheteroscedasticity-C

ONSISTENT

andcovariance.t-statisticsare

inparentheses

andp-values

insquare

brackets.

aDlnRGDPC

i¼�A0þ� A

iþ� A

4ELECTiþ�1Dlnkiþ�2Dlnhiþ�3DlnGCGDPiþ" i.

*,**and***indicate

significance

atthe10,5and1%

3030 P. Bodman

However, there is some evidence that other aspects ofpublic sector decentralization have a negative impacton economic growth.

Panel data analysis

The panel regressions are based on the OECDcountries for which the four panel indicators of FDare available over the period 1981 to 1998 (Australia,Austria, Belgium, Canada, Switzerland, Germany,Denmark, Spain, Finland, France, the UK, Ireland,Italy, Luxembourg, Norway, Portugal, Sweden andthe US).30 The results are given in Table 4.31 Allregressions include country fixed effects.

Again, few FD measures, linear or hump-shaped,are significantly correlated with growth. The growthrate of human capital is rarely significant and in theone case its sign is counter-intuitive. This is a commonproblem found in the studies of FD and growth, and isprobably due to the measurement error.32 Significantcontrol variables are signed as expected.

V. Conclusions

This article extends the existing empirical work on therelationship between government structure, particu-larly FD, and economic growth by exploring a widerrange of measures of government and FD and bylooking not only at their effect on measures of GDPgrowth, but also on the components of growth,specifically growth in the capital stock, human capitaland Total Factor Productivity (TFP) growth. Theempirical work undertaken suggests that there is littleevidence of a direct relationship between FD andeconomic growth. This supports Davoodi and Zou’s(1998) well-cited study, in which they used thesubnational share of total government spending as aproxy for FD. The analysis further demonstrates thata more diverse range of FD indicators, includingnewly developed measures that capture differentlevels of fiscal autonomy, give similar results.Unlike the studies by Thiessen (2000, 2003),Eller (2004) and Akai et al. (2007), hump-shaped

indicators of FD are not found to be directly related

to the output growth.Nevertheless, other measures of government decen-

tralization, besides FD, do seem to have a directrelationship with the economic growth. In this sample

of OECD countries, federal systems tend to havelower growth rates than do unitary states, indepen-dent of their degree of decentralization. Countries

with more elected tiers of government generally havelower economic growth.

Where should research go from here? Whilst thisarticle incorporates measures of FD disaggregated bydifferent levels of subnational discretion over tax

rates and tax bases, future research may considerdeveloping more disaggregated measures of fiscalautonomy in different areas of expenditure or mea-

sures that are horizontally disaggregated acrosssubnational jurisdictions. Measures of FD that accu-

rately represent changes in FD or capture qualitativerestrictions on subnational autonomy (such as thoseoutlined in the World Bank’s Qualitative

Decentralization Indicators (2001)) may provide fur-ther insights. Better data on human capital, especiallybetter time-series data, is also necessary.

Work on the theoretical front should seek toconnect studies of FD and growth with research

suggesting that income levels are a determinant ofFD. Structural equations specified for both the

determinants of FD and the inclusion of FD ingrowth regressions may allow one to establish morereliable tests for endogeneity or simultaneity in the

relationship. Even if there is no direct relationshipbetween FD and economic growth, further theoret-ical work joining the two areas is warranted. If the

level of FD affects growth in the capital stock, thismay have an indirect impact on long run GDPgrowth, leading to a new level of GDP, which in turn

may have an impact on the level of FD. Similarly, oneof the major problems with the growth literature isthat there is reason to believe that the major

components of the growth equation, capital andhuman capital, are endogenous. With the exceptionof Iimi’s (2005) cross-section analysis, all studies of

FD and growth ignore this issue. Future studiesmay attempt to account for such endogeneity.

30Despite Ireland and Luxembourg’s unusually high growth rates in the late 1990s, their growth over the whole 1981 to 1998period did not seem unusual. Iceland, New Zealand and the Netherlands were excluded as they were missing data onunemployment (Iceland) and fiscal balance (New Zealand and the Netherlands). RDEC1 is missing data for Portugal andItaly. Luxembourg was excluded from the human capital equations due to missing data on average schooling. Politicalfreedom data is missing for Germany prior to 1990.31F-tests indicate significant country fixed effects and Hausman specification tests largely rejected specifications using randomeffects. Other checks for collinearity and outliers were also performed as part of the robustness checks. Results reported werenot found to be affected significantly by these issues. Since Lagrange Multiplier (LM) tests reported serial correlation inthe panel data estimates, the variance–covariance matrix estimates are corrected for heteroscedasticity and serial correlation.32Davoodi and Zou (1998) also have this problem.

4.Panel:fixed

effects(within)estimatesa

Measure

HTDEC1

HRDEC1

C0.036***

0.033***

0.030***

0.0245***

0.024***

0.044***

0.0295***

0.028***

(0.035)

(4.122)

(8.895)

(5.882)

(12.247)

(11.729)

(7.646)

(11.827)

[0.000]

[0.001]

[0.000]

�0.0004

�0.0003**

5.73E-05

8.79E-06

�0.0002***

�4.71E-05

–2.98E-05

(�0.0001)

(�1.206)

(�2.500)

(0.550)

(0.317)

(�6.091)

(–1.234)

(�1.644)

[0.287]

[0.748]

[0.013]

[0.583]

[0.752]

[0.000]

[0.218]

[0.101]

Dln(k)

0.622***

0.518***

0.511***

0.589***

0.516***

0.637***

0.542***

0.589***

(0.576)

(4.437)

(4.285)

(5.145)

(4.474)

(5.274)

(4.737)

(5.294)

[0.000]

[0.0000]

[0.000]

Dln(h)

�0.015

�0.0067

�0.008

�0.006

�0.039*

�0.012

�0.002

(�0.009)

(�0.311)

(�0.351)

(0.385)

(�0.292)

(�1.834)

(�0.571)

(�0.124)

[0.678]

[0.552]

[0.726]

[0.701]

[0.771]

[0.068]

[0.569]

[0.901]

C–GDP)

�0.318***

�0.321***

�0.316***

�0.388***

�0.318***

�0.311***

�0.320***

�0.386***

(�0.319)

(�6.397)

(�6.258)

( �7.498)

(�6.316)

(�5.966)

(�6.381)

(�7.189)

[0.000]

�0.175***

�0.195***

�0.188***

�0.2781**

�0.184***

�0.205***

�0.191***

�0.273***

(�0.193)

(�4.023)

(�4.137)

(�6.299)

(�4.202)

(�5.034)

(�4.449)

(�6.202)

[0.000]

Adj.R2

Countries

11.035

11.012

11.026

13.384

10.966

13.348

11.066

13.477

F-value

[0.000]

Notes:

ey–WestheteroscedasticityandautocorrelationconsistentSEandcovariance

(degrees

offreedom

corrected).Relatedt-statisticsare

inparentheses

andp-values

insquare

brackets.

ARamseyRESET

wasalsoperform

confirm

thespecifiedmodel

wasunbiasedandconsistent,

andinferencesusingLeast

Squares(LS)

werevalid.

aDlnRGDPC

i,t¼� A

0þ� A

i,tþ�A2FU

i,tþ�A4ELECTi,tþ�1Dlnki,tþ�2Dlnhi,tþ�3DlnGCGDPi,tþ" i,t.

*,**and***indicate

significance

atthe10,5and1%

3032 P. Bodman

One possible line of research, suggested by Temple(1999), is to adapt empirical growth models to allowexplicitly for the possibility of regressors that areendogenous to the growth rate or level of income.

Finally, existing intuitive discussion of the poten-tial relationship FD and growth suggests that theconnection may be indirect, that is, decentralizationmay influence economic growth by improving effi-ciency and increasing the quantity and quality ofinputs to the production process. Therefore, futureresearch could also extend the study of the indirectimpact of FD on economic growth, via the compo-nents of the growth equation, to a panel context.33

To conclude, as long as the current interest in thedecentralization of fiscal responsibilities is main-tained, the effect of FD on the overall economy willcontinue to be debated. With this interest, even morereliable data measures and more work on thetheoretical front, the lens through which one studiesthe interrelationships between FD and real economicvariables can continue to be refocused, until a clearerpicture comes into view.

Acknowledgements

This research originated as part of Ms KathrynFord’s Honours thesis supervised by the author at theUniversity of Queensland in 2005. The author wouldsincerely like to thank Ms Ford for her contributionto the research agenda. The author would further liketo thank Robin Boadway, Harry Campbell and twoanonymous referees for comments and suggestions.Financial support is also provided from AustralianResearch Council Discovery Grant DP0877502. Alas,all remaining errors and omissions are attributable tothe author alone.

References

Adam, A., Delis, M. and Kammas, P. (2008) Fiscaldecentralization and public sector efficiency: evidencefrom the OECD, CESifo Working Paper No. 2364,Center for Economic Studies, Ifo Institute.

Akai, N., Nishimura, Y. and Sakata, M. (2007)Complementarity, fiscal decentralization and eco-nomic growth, Economics of Governance, 8, 339–62.

Alesina, A. and Tabellini, G. (1990) A positive theory offiscal deficits and government debt, Review ofEconomic Studies, 57, 407–14.

Armingeon, K., Leimgruber, P., Beyeler, M. andMenegale, S. (2002) Comparative political data set1960–2001. Available at http://www.ipw.unibe.ch/mitarbeiter/ru_armingeon/CPD_Set_en.asp (accessed11 May 2005).

Bahl, R. W. (1999) Implementation rules for fiscal decen-tralization, International Studies Program WorkingPaper Series No. 991(1), Andrew Young School ofPolicy Studies, pp. 1–30.

Bahl, R. W. and Linn, J. F. (1992) Urban Public Finance inDeveloping Countries, Oxford University Press, Oxford.

Bahl, R. W. and Nath, S. (1986) Public expenditure decen-tralization in developing economies, Environment andPlanning C: Government and Policy, 4, 405–18.

Baltagi, B. H. (1996) Econometric Analysis of Panel Data,John Wiley and Sons, West Sussex.

Barankay, I. and Lockwood, B. (2007) Decentralizationand the productive efficiency of government: evidencefrom Swiss Cantons, Journal of Public Economics, 91,1197–218.

Barro, R. J. and Lee, J. W. (2000a) Inequality and growthin a panel of countries, Journal of Economic Growth, 5,87–120.

Barro, R. J. and Lee, J. W. (2000b) International data oneducational attainment: updates and implications.Available at http://www2.cid.harvard.edu/ciddata/(accessed 1 August 2005).

Brueckner, J. K. (2006) Fiscal federalism and economicgrowth, Journal of Public Economics, 90, 2107–20.

Bruess, F. and Eller, M. (2004) Fiscal decentralisation andeconomic growth: is there really a link?, Journal forInstitutional Comparisons, 2, 3–9.

Calmfors, L. and Driffill, J. (1988) Bargaining structure,corporatism and macroeconomic performance,Economic Policy, 3, 13–61.

Carrion-i-Silvestre, J., Espasa, M. and More, T. (2008)Fiscal decentralization and economic growth in Spain,Public Finance Review, 36, 194–218.

Commonwealth Secretariat (1985) Training for decentral-ised systems of government administration: report ofan inter-agency working group, in CommonwealthWorkshop on Decentralised Government, Nairobi.

Davoodi, H. and Zou, H. F. (1998) Fiscal decentralizationand economic growth: a cross-country study, Journalof Urban Economics, 43, 244–57.

De la Fuente, A. and Domenech, R. (2000) Human capitalin growth regressions: how much difference does dataquality make?, CEPR Discussion Paper No. 2466,pp. 1–59.

Dillinger, W. (1994) Decentralization and its implicationsfor urban service delivery, Urban Management andMunicipal Finance Discussion Paper No. 16, WorldBank, Washington, DC.

33 Thiessen’s (2000, 2003) studies use pooled cross-section analysis to investigate the effect of FD on the TFP growth andinvestment growth. He uses the two-step method suggested by Elias (1992) and Fischer (1993) to test the effect of FD on TFP.The first step involves examining the direct relationship between growth and FD. In the second stage, the policy variable isregressed against either the Solow or Mankiw Romer Weil residuals. An earlier version of this article did include an extensionof Thiessen’s analysis looking at the possible effects of FD on the accumulation of physical and human capital and onmeasures of TFP growth. Our results did find some evidence that intermediate levels of FD have a positive effect on factoraccumulation, and to a lesser extent TFP growth. Results are omitted from this version of the article but are available in theworking paper version or from the authors upon request.

Ding, Y. (2007) Fiscal decentralization and economicgrowth in China, 1994–2002, Journal of ChineseEconomic and Business Studies, 5, 243–60.

Ebel, R. D. and Yilmaz, S. (2002) On the measurement andimpact of fiscal decentralization, World Bank PolicyResearch Working Paper No. 2809, pp. 1–26.

Elias, V. (1992) Sources of Growth: A Study of Seven LatinAmerican Economies, San Francisco, ICS Press.

Eller, M. (2004) The determinants of fiscal decentralisationand its impact on economic growth: empirical evidencefrom a panel of OECD countries, UnpublishedDiploma, Vienna University of Economics andBusiness Administration.

Fischer, S. (1993) The role of macroeconomic factors ingrowth, Journal of Monetary Economics, 32, 485–512.

Garnaut, R. (2006) Institutional frameworks to promoteproductive outcomes, in Productive Reform in aFederal System (Eds) R. Garnaut and L. Song,Australian Government Productivity Commission,Canberra, pp. 85–95.

Guess, G. M., Loehr, W. and Martinez-Vasquez, J. (1997)Fiscal decentralization: a methodology for casestudies, CAER consulting assistance on economicreform II, Discussion Paper No. 3, pp. 1–5.

Iimi, A. (2005) Decentralization and economic growthrevisited: an empirical note, Journal of UrbanEconomics, 57, 449–61.

Levine, R. and Renelt, D. (1992) A sensitivity analysis ofcross-country growth regressions, American EconomicReview, 82, 942–63.

Lijphart, A. (1999) Patterns of Democracy: GovernmentForms and Performance of Thirty-Six Countries, YaleUniversity Press, New Haven.

Lin, J. Y. and Liu, Z. (2000) Fiscal decentralization andeconomic growth in China, Economic Development andCultural Change, 49, 1–23.

Mankiw, N. G., Romer, D. and Weil, D. N. (1992) Acontribution to the empirics of economic growth,Quarterly Journal of Economics, 107, 407–37.

Martinez-Vazquez, J. and McNab, R. (1997) Fiscaldecentralization, economic growth, and democraticgovernance, International Studies Program WorkingPaper Series No. 97(7), Andrew Young School ofPolicy Studies, pp. 1–41.

Martinez-Vazquez, J. and McNab, R. (2003) Fiscaldecentralization and economic growth, WorldDevelopment, 31, 1597–616.

Martinez-Vazquez, J. and McNab, R. (2005) Fiscal decen-tralization, macrostability and growth, InternationalStudies Program Working Paper Series No. 06(05),Andrew Young School of Policy Studies, pp. 1–41.

McLure, C. E. (1995) Comment on ‘The dangers ofdecentralization’ by Prud’homme, World BankResearch Observer, 10, 221–6.

Musgrave, R. A. (1959) The Theory of Public Finance,McGraw Hill, New York.

Nuxoll, D. A. (1994) Differences in relative prices andinternational differences in growth rates, AmericanEconomic Review, 84, 1423–36.

Oates, W. E. (1972) Fiscal Federalism, Harcourt BraceJovanovich, New York.

Oates, W. E. (1995) Comment on ‘conflicts and dilemmasof decentralization’ by Rudolf Hommes, in AnnualWorld Bank Conference on Development Economics(Eds) M. Bruno and B. Pleskovic, World Bank,Washington, DC, pp. 351–3.

Oates, W. E. (1999) An essay on fiscal federalism, Journalof Economic Literature, 37, 1120–49.

OECD (1999) Taxing powers of state and local govern-ment, OECD Tax Policy Studies, 1, 1–84.

Patsouratis, V. A. (1990) Fiscal decentralization in the EECcountries, Public Finance, 45, 423–39.

Prud’homme, R. (1995) On the dangers of decentralization,World Bank Research Observer, 10, 201–20.

Rodden, J. and Eskeland, G. S. (2003) FiscalDecentralization and the Challenge of Hard BudgetConstraints, MIT Press, Cambridge, MA.

Roubini, N. and Sachs, J. (1989) Political and economicdeterminants of budget deficits in industrialdemocracies, European Economic Review, 33, 903–38.

Schiavo-Campo, S., De Tommaso, G. and Mukherjee, A.(1997) An international statistical survey of govern-ment employment and wages, World Bank PolicyResearch Working Paper No. 1806, pp. 1–83.

Sewell, D. (1996) The dangers of decentralization accordingto Prud’homme: some further aspects, World BankResearch Observer, 11, 143–50.

Stegarescu, D. (2005) Public sector decentralization:measurement concepts and recent internationaltrends, Fiscal Studies, 26, 301–33.

Stegarescu, D. (2009) The effects of economic and politicalintegration on fiscal decentralisation: evidence from theOECD, Canadian Journal of Economics, 42, 694–718.

Tanzi, V. (1995) Fiscal federalism and decentralization: areview of some efficiency and macroeconomic aspects,in Annual World Bank Conference on DevelopmentEconomics, Washington.

Tanzi, V. (2008) The future of fiscal federalism, EuropeanJournal of Political Economy, 24, 705–12.

Temple, J. (1999) The new growth evidence, Journal ofEconomic Literature, 37, 112–56.

Thiessen, U. (2000) Fiscal federalism in Western Europeand selected other countries: centralization or decen-tralization? What is better for economic growth?,Discussion Paper No. 224, Deutsches Institut furWirtschaftsforschung, pp. 1–49.

Thiessen, U. (2003) Fiscal decentralisation and economicgrowth in high income OECD countries, FiscalStudies, 24, 237–74.

Thornton, J. (2007) Fiscal decentralization and economicgrowth reconsidered, Journal of Urban Economics, 61,64–70.

Twomey, A. and Withers, G. (2007) Federalist paper 1:Australia’s federal future, Report for the Council forthe Australian Federation, April.

Woller, G. M. and Phillips, K. (1998) Fiscal decentraliza-tion and LDC economic growth: an empirical inves-tigation, Journal of Development Studies, 34, 139–48.

World Bank (1999/2000) World Development Report,Oxford University Press, Washington.

World Bank (2001) Fiscal decentralization indicators.Available at http://www1.worldbank.org/publicsector/decentralization/fiscalindicators.htm (accessed 18March 2005).

World Bank (2001) Qualitative decentralization indicators.Available at http://www1.worldbank.org/publicsector/decentralization/qualitativeindicators.htm (accessed 7April 2005).

World Bank (2005a) Global development network growthdatabase. Available at http://www.worldbank.org/research/growth/GDNdata.htm (accessed 27 July 2005).

3034 P. Bodman

World Bank (2005b) World development indicators.Available at http://devdata.worldbank.org.ezproxy.library.uq.edu.au/dataonline/ (accessed 26 July 2005).

Xie, D., Zou, H. F. and Davoodi, H. (1998) Fiscaldecentralization and economic growth in the US,Journal of Urban Economics, 45, 228–39.

Yilmaz, S. (1999) The impact of fiscal decentralization onmacroeconomic performance, in 92nd AnnualConference on Taxation, Atlanta.

Zimmerman, H. (1990) Gewichtsverlagerungen imfoderativen staatsaufbau unter EG-Einfluß,Wirstschaftsdienst, 9, 451–6.

Appendix

Variable Source

Number of subnational jurisdictions World Development Report (1999/2000)Number of elected subnationaltiers of government

World Development Report (1999/2000)

Constitutional structure Armingeon et al. (2002)Subnational and CG employees Schiavo-Campo et al. (1997)GDP and GDP growth World Development Indicators (2005b)GFCF World Bank, World Development Indicators (2005b)Secondary school enrolment ratio World Bank, Database for the Global Development Network;

World Development Indicators (2005a)Average years of schooling Barro and Lee (2000a,b)GDP deflator World Development IndicatorsGG final consumption expenditure World Development IndicatorsVertical fiscal imbalance World Bank, FD Indicators (2001)Domestic credit growth World Development IndicatorsPolitical freedom Freedom House

fiscal decentralization and economic growth in the oecd

Documents

fighting poverty fiscal decentralization usaid

fiscal decentralization in philippines

operational framework for fiscal decentralization

fiscal decentralization and fdi

fiscal decentralization and the efficiency of public ... ·...

fiscal decentralization with distortionary taxation

fiscal decentralization and economic development

modelling azerbaijan’s fiscal decentralization: examing...

fiscal decentralization, regional income inequality, and...

fiscal decentralization in japan

fiscal federalism and decentralization in selected igad ......

decentralization and political...

three years of fiscal decentralization in...

fiscal decentralization: theory as...

fiscal decentralization and regional inequality in … ·...

fiscal decentralization and rural infrastructure

fiscal decentralization reform in cambodia ·...

fiscal decentralization and nigerian macroeconomic

fiscal decentralization and voters’ choices as …fiscal...

fiscal decentralization in indian · pdf file ·...