FISCAL STUDIES, vol. 29, no. 4, pp. 435–465 (2008) 0143-5671
Fiscal Decentralisation and Empowerment:
Evolving Concepts and Alternative Measures*
JAMESON BOEX† and RENATA R. SIMATUPANG‡
†Public Finance & Development Solutions, LLC
(jamieboex@gmail.com)
‡Georgia State University
(rsimatupang@gsu.edu)
Abstract
Decentralisation reforms are among the most common and significant public
sector reforms, particularly in developing and transitional countries around
the world. Despite the importance of the topic to policy practitioners and
academic researchers alike and the extensive empirical research on the topic,
there is consensus in the literature that the measures of decentralisation that
are currently used are unsatisfactory. In response, we propose an alternative
measure of fiscal decentralisation based on the notion that decentralisation is
more than simply the inverse of centralisation. Following Bahl (2005), we
consider fiscal decentralisation as ‘the empowerment of people by the
[fiscal] empowerment of their local governments’. Accordingly, we develop
a measure of fiscal empowerment that allows us to quantify fiscal
decentralisation as the gain in empowerment due to devolution and we
analyse the proposed measures of empowerment and decentralisation for a
cross-section of developing, transitional and industrialised countries.
*Submitted April 2007.
The authors would like to thank Roy Bahl, Reagan Baughman, Eunice Heredia-Ortiz, François
Vaillancourt, Serdar Yilmaz and three anonymous reviewers for their helpful comments on earlier drafts
of this paper.
Keywords: fiscal decentralisation, intergovernmental finance, local government finance, fiscal
empowerment.
JEL classification numbers: H11, H70, H72.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008. Published by Blackwell Publishing Ltd, 9600 Garsington Road,
Oxford, OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA
436
Fiscal Studies
I. Introduction: background and purpose
Fiscal decentralisation reforms are among the most common and significant
fiscal policy reforms engaged in by developing, transitional and
industrialised countries around the world. This assertion is based not only on
the scope of reform that fiscal decentralisation implies for the structure of
public finances and public service delivery in a country, but also on the
number of countries around the world that engage in decentralisation reform:
over the last quarter-century, more than 75 countries have attempted to
transfer fiscal responsibilities of the state to lower tiers of government
(Ahmad et al., 2005).
Despite the importance of fiscal decentralisation to policy practitioners
(particularly in developing and transitional countries) as well as academic
researchers in economics, public administration and political science alike,
the concept has proven extremely difficult to capture quantitatively.
Although there are a number of commonly-used measures of (fiscal)
decentralisation, there seems to be wide consensus in the literature regarding
the inadequacy of these measures. Indeed, the current state of knowledge on
fiscal decentralisation – as well as our current understanding of how
different degrees of decentralisation interact with economic outcomes – is
limited by our inability to properly quantify the phenomenon. For instance,
empirical studies on the determinants of decentralisation (as summarised by
Bahl and Linn (1992), for example) could yield substantially different results
depending on which measure of fiscal decentralisation is used. Likewise,
exploring the impact of decentralisation on economic growth (e.g. Davoodi
and Zou, 1998; Martinez-Vazquez and McNab, 2005), budget balance (de
Mello, 2000) or the relationships between decentralisation and the size of the
public sector (Oates, 1985) all critically depend on whether the chosen
measure of decentralisation captures ‘the right thing’. In fact, Ebel and
Yilmaz (2002) demonstrate that substantially different results are achieved
when different measures of decentralisation are used for the purpose of such
empirical analyses.
In this paper, we provide a brief review of the most common measures of
fiscal decentralisation. In the process, we note the conceptual and empirical
weaknesses of these prevailing decentralisation measures and recognise that
a number of researchers are exploring alternative approaches to the
measurement of decentralisation. These more recent approaches generally
seek to capture variations in the degree of subnational control over
subnational fiscal resources. Building on these recent explorations, we
consider the importance of political institutions and the proximity to the
populace with which fiscal decisions are made as additional factors in
determining the effective degree of fiscal decentralisation (rather than
treating political decentralisation as a separate dimension of the
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
437
decentralisation phenomenon), and accordingly propose an alternative
measure of fiscal decentralisation that incorporates information on political
responsiveness and a country’s subnational government structure.
There are an increasing number of empirical studies that consider the
impact of subnational fiscal autonomy, political decentralisation or local
governance variables (as independent variables) on different aspects of fiscal
decentralisation or outcomes of decentralised service delivery (as the
dependent variable).1 In contrast to these empirical efforts, what we seek to
accomplish in the current paper is to arrive at a single variable that quantifies
the overall degree of empowerment that is achieved through fiscal
decentralisation, thereby facilitating direct comparisons of the degree of
fiscal empowerment and decentralisation between different countries and
over time. Furthermore, consolidating fiscal, political and institutional
factors into a single measure of fiscal decentralisation allows the proposed
measure to be used in simultaneous-equation models without losing the
impact of the interaction of fiscal, political and institutional factors. As such,
our proposed measure of decentralisation might be used in future research to
replicate existing empirical models of decentralisation, as, in line with the
findings by Ebel and Yilmaz (2002), empirical studies on the determinants
and impacts of decentralisation have been shown to yield substantially
different results depending on which measure of fiscal decentralisation is
used. The alternative measure of decentralisation developed here, when used
alongside existing measures, could thus help to enhance our understanding
of fiscal decentralisation, both as a fiscal policy issue and as a governance
phenomenon.
II. The shortcomings of the current measures of fiscal
decentralisation
The two most commonly-used measures of fiscal decentralisation are the
expenditure decentralisation ratio (EDR) and the revenue decentralisation
ratio, which measure the share of subnational governments in consolidated
public expenditures and public revenues, respectively. A third common
measure of fiscal decentralisation is the degree of local revenue autonomy,
which is typically measured as the aggregate level of own subnational
revenue collections expressed as a share of total subnational resources. Since
the preponderance of the public finance research focuses on the impact of
decentralised public spending on public service quality and economic
growth, the EDR is arguably the most common measure of fiscal
decentralisation in the empirical literature. The measure’s relatively limited
1
For instance, see Asfaw et al. (2007), Bardhan and Mookherjee (2006b) and de Mello (2000).
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
438
Fiscal Studies
data requirements and easy interpretation have made it a favourite in
comparing decentralisation across countries and over time.
Despite the measure’s common usage in empirical studies, the EDR is
widely acknowledged to be an inadequate measure of fiscal decentralisation.
Perhaps its biggest shortcoming is that it treats all local spending as equal
regardless of the degree of discretion that subnational governments are given
over subnational expenditures, thereby typically overstating the degree of
subnational control over fiscal resources. For instance, the EDR fails to
distinguish between local spending on exclusive (fully devolved) local
functions over which local governments typically have substantial control
and spending on concurrent or delegated functions over which local
governments may have relatively little or no control. Being able to identify
the degree of subnational discretion or autonomy over subnational spending
is particularly relevant for empirical analyses that study the impact of
decentralisation on the size and composition of subnational spending, since
many of the presumed benefits of decentralisation (optimal size of the public
sector, allocative efficiency of public spending, and increased local
accountability leading to technical efficiency in the delivery of locallyprovided goods) are only able to arise if local governments have control over
their financial resources.
The conceptual disconnect between the total level of subnational
spending (or the EDR) on one hand and the varying degrees of subnational
control over these financial resources on the other hand is not a recent
realisation. Rather than relying on a single measure of vertical resource
allocation, Hunter (1977) proposed quantifying the degree of fiscal
decentralisation through a series of vertical fiscal imbalance measures based
on the progressive degree of control that subnational governments have over
their financial resources (ranging from earmarked grants to unconditional
grants to own-source revenues). More recently, the OECD (1999) pursued
the argument that the percentage of own-source revenues collected by each
level of government (defined as subnational revenue sources over which the
subnational level has some degree of control) provides a more suitable
indicator of the relative fiscal autonomy of each level of government.2 Ebel
and Yilmaz (2002) built upon this work by treating general-purpose grants
2
Although there is a clear consensus in the literature that decentralisation depends on the resources
over which subnational officials have discretion or control (rather than merely the resources spent at the
subnational level), the empirical challenge to quantify revenue or expenditure control is as yet largely
unresolved. In practice, it is virtually impossible to accurately determine the true degree of expenditure
discretion enjoyed by different government levels. Even when functional responsibilities are clearly
assigned to different government levels in law, major gaps frequently arise between the de jure control
assigned to subnational governments and their (often much more limited) de facto control. However, we
should acknowledge that the approach used in the recent literature to circumvent this problem – by
focusing predominantly or exclusively on the degree of revenue autonomy – quite possibly produces a
bias in the opposite direction.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
439
with objective criteria and unconditional specific grants as own-source
revenues.
In fact, Ebel and Yilmaz (2002, p. 19) further reignited the discussion of
the shortcomings of the EDR as a measure of (fiscal) decentralisation by
noting that ‘empirical estimations are sensitive to variable selection, and the
implications of making the wrong choice [in measuring fiscal
decentralisation] may be far-reaching in policy design’. As such, Ebel and
Yilmaz have set the stage for a renewed effort to develop and compute
alternative measures of decentralisation that specifically take into account
the degree of subnational autonomy over subnational finances. The
collection of quantitative and qualitative fiscal data necessary to estimate
such more comprehensive measures of decentralisation (e.g. Stegarescu,
2005; Bell et al., 2006), as well as the use of such measures in empirical
decentralisation models, is now being pursued by a number of researchers
(e.g. Meloche, Vaillancourt and Yilmaz, 2004).
To the extent that the new measures being proposed and estimated focus
on classifying or quantifying the degree of subnational fiscal autonomy (over
revenues as well as expenditures), this direction of study addresses only part
of the problem. While the degree of subnational fiscal autonomy is a valid
and important concern to incorporate in a measure of fiscal decentralisation,
focusing exclusively on this issue does not address some of the other major
weaknesses of the expenditure decentralisation ratio measure. Two issues
that we consider of particular relevance are the size (and diversity) of
subnational jurisdictions and the empowerment of the community over the
subnational level as a result of subnational political systems. In fact, Oates
(1972) already recognised that it is a priori unclear whether a country that
places fiscal powers with large regional governments should be considered
more or less decentralised than a similarly-situated country that assigns a
somewhat smaller fiscal responsibility to the subnational level but places this
responsibility with smaller, local governments.
The EDR simply does not take jurisdiction size and (subnational)
political responsiveness into account. For instance, subnational government
spending in many industrialised countries (particularly in Western Europe
and North America) takes place in relatively small municipalities which are
close to the local community. In accordance with Tiebout’s assumptions
(1956), this allows for greater community control over subnational public
spending and greater responsiveness to differences in local preferences, and
therefore has the potential to achieve greater allocative efficiency. This
intergovernmental fiscal architecture stands in stark contrast to that of a
country such as India, where most subnational public spending decisions
take place at the regional level, with States and Union Territories having an
average population of 30 million. This fact notwithstanding, the EDR
considers India to be much more ‘decentralised’ than an average OECD
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
440
Fiscal Studies
country, since 52.0 per cent of public spending in India occurs at the
subnational level, compared with 31.8 per cent in an average industrialised
country.
III. Foundations for an alternative measure of fiscal
decentralisation
There is little doubt that alternative measures of fiscal decentralisation are
needed, not only to measure the quantitative aspects of decentralisation but
also to capture some of the more qualitative aspects of intergovernmental
fiscal systems and fiscal decentralisation reforms (such as the degree of
subnational fiscal autonomy). At the foundation of the alternative measures
that we propose here is not only the practical failing of the current measures,
but perhaps more fundamentally the observation that a paradigm shift has
taken place in the definition of decentralisation.
During the 1980s and early 1990s, decentralisation reforms in developing
economies were often pursued in the context of civil service reforms, where
the main goal was to deal with an unwieldy and overly centralised public
service (Pritchett and Woolcock, 2002). In other countries, (fiscal)
decentralisation reforms were politically motivated in response to centrifugal
forces or as a result of dramatic shifts in political systems and power
relations within a country. For instance, to a considerable degree,
decentralisation in the transition economies of the former Soviet bloc was
generally pursued with the specific intent of decreasing the power of the
centralised state (Bird, Ebel and Wallich, 1996; Martinez-Vazquez and
Boex, 2001). Likewise, many of the major decentralisation reforms of the
1980s and (early) 1990s – including the transitions in Indonesia, South
Africa and Uganda – sought to reduce the power of previously highly
centralised and often totalitarian regimes. Decentralisation allowed these
countries to achieve a vertical distribution of power across different
government levels, while at the same time dealing with ethnic, linguistic and
other geographically diverse demands on the public sector. Broadly
speaking, similar political factors were behind recent decentralisation
reforms in a number of developed countries, including Belgium, Italy and
Spain.
It could be argued that in these situations, the primary aim was to devolve
fiscal power away from the central government, with perhaps less
consideration given to which level of government should be given this fiscal
power.3 This modus operandi is reflective of the traditional definition of
3
The excessive fragmentation of the subnational government structure that took place in many
decentralising countries during the early 1990s, and the subsequent reconsolidation of local government
jurisdictions in many countries during the late 1990s and early 2000s, are indicative of this trend of ‘anticentralisation’ (Boex, Martinez-Vazquez and Timofeev, 2004).
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
441
decentralisation, where the act of shifting resources away from the centre is
paramount, while the recipient of the fiscal authority and responsibility that
are being decentralised is somewhat secondary. Decentralisation was
traditionally defined broadly not just to include finances that were shifted
away from the centre to subnational governments but also to include
resources moved to ‘subordinate or quasi-independent government
organizations or the private sector’ (Litvack and Seddon, 1999). Whereas
under these decentralisation reforms financial resources were shifted to
budgets at the subnational level, central government officials often indirectly
retained control over these newly ‘decentralised’ expenditures either through
grant conditionalities or through administrative regulations and ‘guidelines’
over subnational spending.
However, the notion of (fiscal) decentralisation is not a static concept.
We believe that the consensus-definition of fiscal decentralisation has
evolved over the past decade, including a clear recognition that fiscal
decentralisation is about more than just shifting power and money away
from the centre. In his assessment of the traditional (so-called ‘firstgeneration’) approach to fiscal decentralisation, Weingast (2006) notes that
the previous approach was deficient because the assumptions postulated in
the traditional model of fiscal federalism were insufficiently congruent with
reality. Accordingly, Weingast argues that policy changes predicated on the
first-generation models of fiscal federalism were incapable of producing the
desired benefits of a fiscally decentralised system, because traditional
models of fiscal federalism and fiscal decentralisation ignored many of the
preconditions for successful decentralisation reforms.
In contrast, the current ‘consensus view’ of fiscal decentralisation (or the
emerging second-generation approach to fiscal federalism) is much more
deliberate in recognising the constraints imposed by political economy and
the institutional environment on intergovernmental fiscal systems. For
instance, whereas political and administrative decentralisation previously
received relatively little attention in the realm of fiscal decentralisation, the
idea that ‘institutions matter’ in order for the benefits of fiscal
decentralisation to arise (in addition to the mere devolution of financial
resources) now seems to be widely accepted. Furthermore, most observers
would now agree that in addition to the devolution of financial resources and
decision-making authority, achieving the benefits of decentralisation reform
further requires substantive public participation in subnational fiscal
decisions and appropriate mechanisms for subnational accountability. For
instance, Bardhan and Mookherjee (2006a) present a careful theoretical
analysis of the impact of (decentralised) accountability on service delivery
patterns. A number of recent contributions to the public economics literature
also provide empirical evidence to suggest how and to what extent different
institutional dimensions (such as political influence over the distribution of
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
442
Fiscal Studies
intergovernmental transfers at the central level, local accountability
mechanisms, and local political institutions or ‘elite capture’ at the local
level) impact the intergovernmental flow of funds and service delivery
outcomes under a decentralised system (Bardhan and Mookherjee, 2006b;
Asfaw et al., 2007; Olken, 2007).
This deeper understanding of fiscal decentralisation is captured well by
Bahl (2005), who offers that, as a working definition, (fiscal)
decentralisation can be defined as ‘the empowerment of people by the
[fiscal] empowerment of their local governments’. Whereas the traditional
definition of decentralisation emphasises the shifting of power away from
the centre, in Bahl’s definition decentralisation revolves around two issues:
first, assuring the fiscal empowerment of subnational governments – among
others, to deliver public services and infrastructure; and second, the
empowerment of people and communities, to direct their subnational
government officials to use their financial resources in accordance with local
needs and preferences.
This alternative definition of decentralisation provides important
guidance for the measurement of fiscal decentralisation, as the new
definition implies that quantifying decentralisation is not necessarily just a
matter of determining ‘how much’ is decentralised in terms of public
expenditures or revenues, nor is it necessarily only a matter of assessing how
much control subnational governments have over their fiscal resources. In
addition, we argue that a critical factor in the measurement of
decentralisation should be the degree to which households and communities
are empowered by (and over) the decentralised provision of public services,
including the proximity of the people to the subnational officials responsible
for delivering their services.
IV. Deriving measures of fiscal empowerment and fiscal
decentralisation
In arriving at measures of fiscal empowerment and decentralisation in line
with Bahl’s definition, these measures should (as opposed to the existing
measures of fiscal decentralisation) capture both the degree of fiscal
empowerment of subnational governments and the degree of empowerment
of the citizen over his or her subnational government(s), based on the notion
that the closer public resources are spent to the citizen, the more empowered
the populace is. Our formulation is consistent with Raich (2005), who
hypothesises that in the context of decentralised local governance,
empowerment is most likely to occur when three conditions prevail: low
costs of participation; large subnational budgets; and a high degree of
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
443
subnational fiscal autonomy (for instance, resulting from a high proportion
of tax revenues collected locally).
In order to capture these conditions in a systematic framework, we start
by defining fiscal empowerment, E, where Ei is a representative individual’s
degree of empowerment over public spending in his or her jurisdiction at
government level i:
(1)
Ei = Vi Fi
and E = ∑Vi Fi
i
so that Ei at each level of government is equal to an individual’s ‘voice’ over
public spending at the relevant government level, Vi, multiplied by the
degree of fiscal control or autonomy exercised by the government unit at that
level of government, Fi. The relationship between Vi and Fi is multiplicative
because fiscal empowerment requires not only that citizens have control over
their public officials but also that governments at each level of government
have control over the fiscal resources under their management. In turn, E is
defined as the sum of the representative individual’s empowerments over
fiscal decisions across all government levels.
1. The fiscal role of subnational governments
In our attempt to develop an appropriate measure of fiscal decentralisation
based on Bahl’s definition, we first consider the degree of fiscal control of
government units at different levels of the public sector (i.e. the degree to
which government jurisdictions themselves are fiscally ‘empowered’). In its
most basic form, the fiscal power, F, of a representative or average
government unit at government level i in a country could be represented as
(2)
Fi =
1
αiG
Ni
where Ni reflects the number of government jurisdictions at government
level i, αi reflects the share of public spending that is controlled at
government level i (so that 0 ≤ αi ≤ 1 and Σαi = 1) and G equals total public
spending across all levels of government.4 If we consider that an amount
equal to g dollars per person is spent by the public sector as a whole, then we
could substitute G = gP at our convenience, where P equals the total
national population.
4
It should be noted that equation (2) defines fiscal empowerment for an average jurisdiction at each
government level. While it would be preferable to exactly measure fiscal empowerment (and voice) for
every subnational jurisdiction, this would require a level of disaggregated subnational data across
countries that is generally not available.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
444
Fiscal Studies
Similarly to the alternative measures proposed by the OECD (1999 and
2001), Ebel and Yilmaz (2002) and Bell et al. (2006), our conceptual
formulation of fiscal empowerment relies on the share of public spending
over which each government level i has control or discretion (i.e. excluding
earmarked or conditional resources that it receives from other government
levels) rather than simply quantifying the share of public spending that takes
place at each government level. Thus, if a subnational government level is
denied substantial autonomy over its budgetary resources, this would be
reflected in a lower degree of fiscal empowerment, Fi.
2. A generalised measure of a citizen’s voice over public decisions
Despite the depth and breadth of the public choice literature, there is no
agreed indicator for the concept of ‘voice’, either in theory or in the
empirical literature (Keefer, 2004; Alsop, Bertelsen and Holland, 2006). As
the basis for our generalised measure of a citizen’s voice over public
decisions, we posit that although we expect elected officials to make
decisions in a self-interested manner whenever possible, an elected official at
any government level is nonetheless interested in maximising the number of
supporters or votes in his or her respective jurisdiction, and therefore will be
equally open to representing the interests of all constituents: an average
individual’s voice in public decision-making processes will be inversely
related to the number of constituents in a jurisdiction, so that the higher the
population of the jurisdiction, the smaller the relative voice of each
individual.
In addition to the relative loss of the individual’s voice to other
constituents as the number of constituents in the jurisdiction increases,
additional losses to constituents’ voice over public spending decisions occur
as the size (in terms of population and geography) and diversity (whether
socio-economic, ethnic, linguistic or religious) of a government jurisdiction
increase. For instance, the information asymmetries and transaction cost for
making collective decisions increase as a jurisdiction’s population size and
diversity increase; not only will it be harder to achieve some form of
consensus in a larger, more diverse jurisdiction, but it is also likely that the
degree of downward accountability is reduced as a jurisdiction’s size
increases. These factors create fiscal space for public officials to deviate
from the preferences of the electorate: the larger the jurisdiction, the less
voice each individual (and the community as a whole) has to assure that his
or her interests are served and the more beholden public decision-making
can become to special interests or the politician’s self-interest. Therefore, as
jurisdiction size increases and the distance between the electorate and public
officials grows, we expect the relative loss of voice to result in an increasing
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
445
gap between the individual’s preferred set of public services and the basket
of public services ultimately made available.
Of course, jurisdiction size (in terms of population and diversity) is only
one set of determinants of empowerment and voice; another critical
determinant of voice is how receptive and responsive political decisionmakers are to the preferences of the communities that they serve. The
presence of (directly or indirectly) elected officials at a government level
may be considered a necessary condition for assuring a minimum level of
responsiveness; politicians would have little or no incentive to serve their
constituents’ needs in the absence of a minimally participatory governance
mechanism. Therefore, we should expect the degree of political
responsiveness of public officials to be a function of the political institutions
that are in place in different countries and at different government levels
within a country.5
Based on these considerations, we suggest that the voice of an individual
constituent over public spending in an average government jurisdiction at
government level i, Vi, could be represented as
(3)
Vi =
1
J
1+ γ i
i
π i so that Ei =
1
J
1+ γ i
i
π i Fi
where Ji is the number of residents of an average jurisdiction at government
level i (equal to P/Ni), γi is a variable that captures the relative loss of voice
as jurisdiction size increases and πi is a measure of the political system’s
receptiveness and responsiveness to feedback from below.
Given the many institutional factors influencing variations in political
responsiveness across different government levels as well as across different
countries, it may prove difficult to objectively capture the degree of a
political system’s receptiveness and responsiveness in a single measure.
Conceptually, however, it suffices for us to assume that πi is bounded
between 0 and 1, where πi = 1 represents a highly participatory political
system in which policymakers at government level i are fully responsive to
voter feedback, whereas πi = 0 indicates the absence of any participatory
processes, so that policymakers are free to ignore any demands or requests
from the population. Similarly, γi, which captures the rate of voice-loss due
to increases in jurisdiction size, could be a function of several factors,
including the degree of population heterogeneity. If γi = 0, then an
individual’s voice would be strictly inversely proportional to the number of
residents in the jurisdiction. This would reflect a situation in which
politicians and public managers are perfectly responsive to residents’ needs
5
Political institutions may in fact vary not only between countries but also between different
government levels within a country and even between different states or regions (especially within federal
and asymmetric systems).
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
446
Fiscal Studies
in aggregate, while no inefficiencies occur either due to information
asymmetries or due to decreases in the level of accountability as a
jurisdiction’s size increases. However, if γi > 0 (as we would expect), then as
a jurisdiction’s population size increases, an individual’s voice would be
reduced by a more-than-proportional amount as information asymmetries
arise and politicians become less accountable to their constituents as a
whole.
3. Decentralisation as empowerment over fiscal resources
Building on equations (1) to (3), the degree of fiscal empowerment – as
reflected by an individual’s overall degree of empowerment over the
allocation of public resources – should simultaneously take into account his
or her voice over public spending at the various levels of government and the
fiscal empowerment of each government level, so that
(4)
E = ∑Vi Fi =∑
i
i
1
J
1+ γ i
i
πi
α i gP
Ni
.
E could be understood as the monetary value of the benefits received by the
representative citizen from public spending at all government levels. If all
public spending were to take place at a subnational government level
extremely close to the people (with the average jurisdiction size tending
towards one resident) or if government jurisdictions were perfectly
homogeneous and fully responsive to their electorate regardless of
jurisdiction size, then individuals would be highly empowered to direct
public sector spending and could – theoretically, at least – achieve perfect
allocative efficiency (E = g). In contrast, if all public spending (especially in
a highly heterogeneous country) were to take place at the central government
level – the government level furthest from the people – then an individual’s
degree of empowerment over public spending would tend towards zero, as
an individual’s voice over national spending decisions is exponentially and
inversely related to the size of the national population. Fiscal empowerment
– thus defined – ranges in the limit from 0 (in which case citizens are not
empowered at all) to g (in which case individuals have full control over the
spending of their financial contribution to the public sector).
After some straightforward mathematical manipulation, we arrive at the
proposed measure of fiscal empowerment in its final form as
(5)
E
1
= ∑ γ i π iα i
g
J
i
i
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
447
where the left-hand side (E/g) represents a representative individual’s
empowerment over each dollar of per-capita public spending.
As the most comprehensive measure of fiscal empowerment, equation (5)
quantifies the degree of an individual’s empowerment over each per-capita
dollar of public spending at all levels of government, which equals the sum
of his or her relative voice at each government level weighted by the share of
public spending that is controlled at each government level. As a normalised
version of empowerment, E, in equation (4), this measure of fiscal
empowerment ranges in the limit from zero (no fiscal empowerment) to
hypothetical unity – which would indicate complete individual
empowerment over public sector resources. The measure of fiscal
empowerment can be interpreted as the total ‘value-for-money’ that an
average citizen receives from the public sector at the different government
levels.
In fact, the expenditure decentralisation ratio (EDR) represents a single
restricted case of our more general formulation of fiscal empowerment,
notably under the limiting assumptions that (i) each level of government has
control over all spending managed at that government level (as noted
repeatedly, an assumption widely challenged in the literature); (ii) πCG = 0
and πi = 1 for all subnational government levels (i.e. central officials are not
responsive, whereas subnational officials are fully responsive); and (iii) γ
must equal 0 (such that voice-loss does not increase either with jurisdiction
size or with jurisdiction diversity).
To the extent that we are interested in measuring the increase in fiscal
empowerment that results from the devolution of fiscal responsibility to
subnational government levels, the current formulation of fiscal
empowerment allows us to compute the gain in empowerment due to fiscal
decentralisation (devolution) by taking the difference between the most
centralised (least empowering) scenario possible for a country (in which αCG
= 1) and the actual degree of fiscal empowerment as defined by the measure
of fiscal empowerment defined above. This formulation comes the closest to
Bahl’s (2005) definition that fiscal decentralisation is the increase in the
empowerment of citizens that results from the devolution of fiscal decisionmaking powers and the fiscal empowerment of subnational governments.
V. Quantification of the proposed measures
In order for the proposed measures to be relevant for international
comparisons and empirical analyses (including the replication of existing
research on decentralisation that relied on the EDR as an imperfect measure
of decentralisation), we have to be able to meaningfully compute these
newly-defined measures across a broad variety of countries in the world.
With this in mind, in this section we compute the proposed measures of
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
448
Fiscal Studies
fiscal empowerment and decentralisation for a cross-section of
industrialised, transitional and developing countries. This allows us to
engage in an initial analysis and comparison of the proposed alternative
measures. It further enables us to uncover the practical limitations associated
with computing the proposed measures with real-world data.
In fact, the computation of the alternative measures of decentralisation for
a cross-section of countries underscores the scarcity of international
comparable data when it comes to decentralised finance and governance. As
a starting point, while expenditure patterns, population size and subnational
government structures may change and evolve over time, data limitations
lead us to compute the proposed measures only for the latest year for which
data are available. Furthermore, whereas subnational fiscal and governance
data are generally available for individual countries (or even groups of
countries in a specific region), consistent and comparable data on fiscal
discretion, political responsiveness and subnational institutional structures
are seldom available internationally across different regions.
While we recognise that expanding the number of countries included in
the analysis will increase the impact of data limitations on the computation
of the proposed measures, we believe that substantial insights can be gained
from comparing the proposed measures of fiscal empowerment and
decentralisation with the EDR and other existing measures of
decentralisation in the context of a broad cross-country analysis. At the same
time, we submit that additional insights with regard to the dynamics of
decentralisation will surely be gained by computing the measures of fiscal
empowerment and decentralisation for specific country groups (such as
OECD countries or groups of countries within a single geographical region)
where more specific fiscal, political and institutional indicators are often
available and for which the unit of observation is more comparable.6
1. Practical issues in the computation of the proposed measure
A number of data sources were used for the computation of the proposed
measures. First, the necessary public expenditure data for different
government levels and population data were taken from the IMF’s
Government Finance Statistics (GFS) for the latest year available. Although
this data source is commonly used for international comparisons of
subnational finances, a major drawback of the GFS is that these data reflect
expenditures made at different government levels without regard for whether
subnational governments actually had control or discretion over the
resources at their level. Regrettably, this is currently the best estimate of
6
It would similarly be fruitful to apply and analyse the proposed measures of fiscal empowerment and
decentralisation to state–local fiscal relations within a federal system, such as the United States. Such
explorations are left for future research.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
449
subnational expenditure control, as there is no consistent international data
source available at this time that allows us to make more accurate crossregional comparisons of subnational expenditure discretion.
Furthermore, the GFS expenditure data generally limit us to considering
only one regional level and one local government level in each country.7
While considering regional and local government levels separately is an
improvement over the EDR (which lumps together all subnational spending),
it would be even more desirable to have detailed fiscal data for every
regional and local government level or tier in a country’s subnational
government structure.
Information on the territorial-administrative structure of countries was
generally extracted from Administrative Subdivisions of Countries (Law,
2006), combined with secondary sources describing systems of
intergovernmental finance (e.g. OECD, 1997 and 2001; Kandeva, 2001;
Munteanu and Popa, 2001). We should recognise that the current data
sources do not necessarily allow us to fully capture the complexities of
subnational government structures in all countries, which include the
possible concurrent or asymmetric assignment of expenditure
responsibilities, the presence of autonomous regions, the existence of singlepurpose or special-purpose districts (such as school districts), and other
anomalies or idiosyncrasies in countries’ subnational government structures.
In contrast to the computation of the EDR, detailed information on the
degree of political responsiveness in each country (and, when available, at
each level or tier of government) is quintessential for the accurate
computation of fiscal decentralisation and empowerment. A practical
challenge in computing the proposed measures is how to capture the
receptiveness and responsiveness of political systems across different
countries and different government levels (π).
2. Defining and measuring political receptiveness and responsiveness, π
Little theoretical or empirical guidance exists on how to quantify the
receptiveness and responsiveness of an officeholder at government level i, πi,
as differences in political institutions and political systems across (and even
within) countries impact political responsiveness in a variety of different
ways. For example, a constituency-based electoral system might be more
responsive to the needs of individual voters than a political system based on
7
A closely related limitation of the IMF’s Government Finance Statistics is that references to
subnational government levels are often vague and/or the statistics combine public spending for different
subnational government levels. For instance, in the case of Russia, all local government finances are
simply attributed to the regional (oblast) level, thereby ignoring the intricacies of the assignment of
expenditure responsibilities and accountability between the regional and local government levels. See
Dabla-Norris (2006) for a detailed discussion of the complexities of the subnational government
structures, expenditure assignments and intergovernmental fiscal relations in transition economies.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
450
Fiscal Studies
proportional (‘party-list’) representation, while the direct election of
executive officials (in addition to the election of legislators and councillors)
may further improve the responsiveness and accountability of public
decision-making. Countless other details of a country’s political system
impact the responsiveness of public officeholders at different levels of
government, including – but not limited to – the selection mechanism of
party candidates (whether through internal party selection or through open
primary elections); the participatory nature of public decision-making
processes; the separation of power among the executive and legislative
bodies at each government level; the existence (or absence) of avenues for
recalling elected officials between elections; and the possibility for direct
citizen involvement in public decision-making on key issues through
referenda or plebiscites. Furthermore, a strong argument could be made that
political systems that operate in more advanced economic environments tend
to be more responsive, as they generally benefit from greater transparency,
stronger accountability mechanisms, and lower cost for accessing
information and participation.
Since the presence of elected representatives at different government
levels is an important factor in assuring receptive and responsive
governance, we originally considered setting π equal to 1 if some form of
elected government was present at a government level and setting π equal to
0 otherwise. There are two reasons why we did not pursue that approach.
First, conceptually, we felt that the mere presence or absence of elections at
any given government level in a country provided only a weak indicator of
the democratic and participatory processes associated with political
responsiveness. Second, empirically, this approach was essentially made
irrelevant by the high level of aggregation of the available subnational fiscal
data: since expenditure data are generally not reported separately for
deconcentrated tiers of government, essentially all government levels for
which fiscal data are available end up having some form of elected
representation.8
For the purposes of the current discussion, an ideal measure of political
responsiveness would distinguish between differences in political
responsiveness across a broad set of countries as well as capture differences
in political responsiveness within these countries at different government
levels. Although several indices of democracy and civil liberties were
considered, we ultimately chose to use the voter turnout rate as the most
suitable, available indicator of political responsiveness in each country (π).9
8
In this case, we would have had little choice but to set π equal to unity for essentially all countries and
all government levels contained in our data set.
9
See Munck and Verkuilen (2002) and O’Donnell, Iazetta and Vargas (2003) for discussions on
available measures of the participatory or democratic nature of government. For the purpose at hand, a
conceptual weakness of such measures is that they are perhaps as much measures of voice as of political
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
451
Data on voter turnout for parliamentary elections for each country were
drawn from the International Institute for Democracy and Electoral
Assistance (IIDEA).10 Although the level of voter turnout provides more
information on the degree of public participation than the mere presence of
elections alone, we recognise that this is an imperfect measure of political
responsiveness as well. In particular, the use of the voter turnout rate in
parliamentary elections is potentially biased for countries with obligatory
voting requirements or in state-controlled (non-democratic) elections.
3. Defining and measuring voice-loss, γ
Another practical issue to be addressed is the selection of a value of γ, which
quantifies the link between voice and jurisdiction size, in the computation of
the proposed decentralisation measures. If all residents in a jurisdiction are
identical and have identical preferences, little voice-loss might occur even as
the jurisdiction’s size increases; in this case, γ would tend closer to zero. In
contrast, in the presence of a highly diverse and heterogeneous population,
the voice of any individual would dissipate more rapidly as population size
increased (in which case γ would be larger).11
Again, the empirical and theoretical literature provides limited guidance
here; how little or much loss of voice occurs as a jurisdiction’s population
size increases or becomes more heterogeneous is difficult to measure, and it
is likely to vary from country to country and even within a country (Borck,
2002). While it is likely that factors such as geographic size, political culture
and the degree of development of transportation and communication
networks matter in determining the exact value of γ in a country,
unfortunately the character of population heterogeneity differs significantly
between different countries. For instance, whereas racial heterogeneity or
income heterogeneity would be the main factors impeding voice and
responsiveness, while by construction they are unable to distinguish between political responsiveness at
different government levels. Nonetheless, the inclusion of more sophisticated measures of π, perhaps in
the context of a regional analysis of decentralisation, would be an interesting future extension of the
current analysis.
10
Voter turnout data for subnational elections were generally not consistently available. The only
country for which we do not have information on direct parliamentary election turnout is the People’s
Republic of China; our estimate of voter turnout for China’s indirect elections is based on Guan and
Green (2006).
11
Whereas the proposed formulation relies on an average level of heterogeneity within a country or at a
level of government, in reality population heterogeneity is seldom constant across subnational
jurisdictions in a country. It should therefore be recognised that the formulation developed here is not
well-suited to capture asymmetries caused by a spatially concentrated, heterogeneous population. For
instance, consider a situation in which an ethnic or linguistic minority perceives no voice-loss, regardless
of jurisdiction size, as long as the subnational jurisdiction is led by the same ethnic or linguistic group. In
order to consistently model such asymmetries and spatial heterogeneity, detailed data on the degree of
fiscal empowerment as well as data on the size and demographic composition of every subnational
jurisdiction would have to be available. We thank François Vaillancourt for pointing out this issue.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
452
Fiscal Studies
subnational decision-making in the United States (even leading to the
fragmentation of local government structures – see Nelson (1990)), linguistic
differences would most likely serve as the distinguishing demographic factor
in determining heterogeneity in Canada. Likewise, different ethnic, cultural
or religious distinctions most likely serve as the predominant causes of
population heterogeneity in other countries around the world. While it might
be possible to measure population heterogeneity in a comparable manner for
selected groups of countries (for instance, within a single region), the
international scope of the analysis pursued here makes it virtually impossible
to isolate the impact of population heterogeneity on the attrition rate of
voice, γ. Therefore, we are forced in the current analysis to treat γ as a
constant, rather than as a variable that varies with population heterogeneity.
In exploring the appropriate value for γ, the conceptual framework
considers ‘voice’ to be closely related to the degree to which a voter is able
to assure that he or she gets value-for-money from each level of government
for every dollar contributed to the public sector. Based on the functional
form of equation (3), we are therefore able to simulate the impact of different
values of γ (the rate of attrition of voice as jurisdiction size increases) on the
benefit that households perceive from public spending in jurisdictions of
different sizes (assuming πi = 1). Unless we believe that the public sector is
so inefficient or unresponsive as to provide a value-for-money ratio of less
than 50 per cent (at relevant ranges of jurisdiction size), the range of
simulated options presented in Table 1 suggests that γ most likely falls
within the narrow range from 0 to 0.05.
Although it would be interesting to determine the appropriate value of γ
more precisely, the choice of γ within this range is not of tremendous
importance for determining the relative voice at each government level, as
the correlation of E(i|γ=0.01) and E(i|γ=0.10) over relevant ranges of
jurisdiction sizes consistently results in a correlation coefficient
TABLE 1
Relationship between voice (dollars) and jurisdiction size (J)
under various assumptions for parameter γ
Jurisdiction size
5,000
10,000
50,000
100,000
500,000
1,000,000
5,000,000
10,000,000
50,000,000
γ = 0.01
0.91835
0.91201
0.89745
0.89125
0.87702
0.87096
0.85706
0.85114
0.83755
γ = 0.025
0.80821
0.79433
0.76300
0.74989
0.72032
0.70795
0.68003
0.66834
0.64199
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
γ = 0.05
0.65321
0.63096
0.58217
0.56234
0.51886
0.50119
0.46244
0.44668
0.41215
γ = 0.1
0.42668
0.39811
0.33892
0.31623
0.26922
0.25119
0.21385
0.19953
0.16986
γ = 0.2
0.18206
0.15849
0.11487
0.10000
0.07248
0.06310
0.04573
0.03981
0.02885
γ = 0.5
0.01414
0.01000
0.00447
0.00316
0.00141
0.00100
0.00045
0.00032
0.00014
Fiscal decentralisation and empowerment
453
exceeding 0.99. For the purpose of computing our measure of
decentralisation, therefore, we settle on a mid-range value of γ = 0.025.
A final concern that should be kept in mind in interpreting the proposed
measures of fiscal decentralisation and empowerment is that, in reality,
empowerment is not always enhanced when public functions are performed
‘closer to the people’. The subsidiarity principle suggests that the benefits of
proximity should be balanced against certain efficiency concerns. For
instance, recipients of local public services would not be effectively
empowered if the public services were delivered by excessively small
jurisdictions that failed to meet the necessary minimum efficient scale. As
another example, residents of poor subnational jurisdictions would not be
empowered if – in violation of the subsidiarity principle – the responsibility
for financing redistributive poverty reduction activities were devolved to the
local government level. We should recognise that the proposed measure of
decentralisation measures the benefits of decentralisation while assuming
that – consistent with the subsidiarity principle – no additional costs are
associated with decentralisation. Of course, in practice, fiscal
decentralisation reforms should always be subject to the evaluation of the
qualitative design aspects of the system of intergovernmental fiscal relations.
VI. An initial analysis of the proposed measures
The results of our computations for the measures of fiscal empowerment and
fiscal decentralisation are summarised in Table 2, which presents the
relevant descriptive statistics as well as the correlation matrix. These
statistics are analysed and discussed in some detail in this section, while
details used for the computation of the measures of fiscal decentralisation
and empowerment are contained in the appendix for the full set of countries
included in the analysis.
While it is difficult to make any pronouncements with respect to the
absolute values of the fiscal empowerment measure (particularly since the
nominal values of the empowerment measure are sensitive to the exact
selection and measurement of α, π and γ), our measure of fiscal
empowerment suggests that the benefit (or ‘empowerment’) felt from every
dollar of public spending ranges from 19 to 67 cents, with an average of 48
cents per dollar. Subject to the same caveats noted above, the descriptive
statistics suggest that the empowerment gains achieved through
decentralisation (such as the gains in allocative efficiency) are relatively
small, falling in the range from 0 to 8 cents per dollar of public spending.
Furthermore, it should be kept in mind that the descriptive statistics and
correlation coefficients included in the table are a function of the exact
sample of countries included in the initial analysis.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
454
Fiscal Studies
TABLE 2
Alternative measures of fiscal decentralisation and empowerment:
descriptive statistics and correlations
EDR
Mean
Standard deviation
Minimum
Maximum
0.2679
0.1654
0.0330
0.7866
Fiscal
empowerment
0.4791
0.1045
0.1930
0.6711
Correlations
EDR
Fiscal empowerment
Gain in empowerment
Population
1.0000
0.2432
0.8159
0.4869
1.0000
0.3835
–0.0694
Gain in
empowerment
0.0171
0.0136
0.0013
0.0787
Population
(million)
75.99
215.79
0.29
1,288.40
1.0000
0.4925
1.0000
1. Analysis of overall fiscal empowerment
In the existing empirical literature, it is not uncommon for studies to
incorporate the expenditure decentralisation ratio in an empirical model
either implicitly or explicitly as a measure of the degree of empowerment
that people have over public spending or as a measure of the proximity of
the public sector to the people, rather than as a measure of the degree of
expenditure decentralisation per se. In such instances, the proposed measure
of fiscal empowerment (as defined in equation (5)) is a much more direct
measure of this proximity or empowerment. Table 2 shows that only a
relatively weak (but positive) correlation exists between the overall fiscal
empowerment measure and the EDR.
The weak correlation of the proposed fiscal empowerment measure with
the traditional measure of decentralisation (r = 0.24) suggests that the two
measures have substantially different patterns across countries. Closer
inspection of the data shows that the measure of fiscal empowerment
indicates that two types of countries generally tend to be more empowered
vis-à-vis the EDR. First, the citizens of smaller (less populous) countries
tend to be more empowered since they are closer to their central decisionmakers; and second, countries that devolve substantial expenditure
responsibility to relatively small subnational jurisdictions are considered to
be more highly empowered (for instance, South Africa). While large and
federal countries tend to be considered highly ‘decentralised’ or
‘empowered’ by the EDR, the proposed measure of fiscal empowerment
identifies Luxembourg, Belgium and Iceland as countries where the citizenry
is highly empowered over public spending. In contrast, large countries
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
455
(including federal countries such as Canada, India and the United States)
show a large drop in their degree of fiscal empowerment vis-à-vis the EDR.
In fact, when considering the proposed formulation of fiscal
empowerment, India falls from being one of the most decentralised countries
in the world based on its EDR (ranked number 6) to one of the least fiscally
empowered nations when considering the proposed overall fiscal
empowerment measure (ranked 54 out of 60 countries). The alternative
formulation thus resolves our earlier concern that the EDR inappropriately
equates the benefits resulting from ‘decentralised’ public spending at the
state level in India (with 10 states in excess of 50 million people) with the
benefits of decentralisation reform in other countries which devolve
spending to government levels much closer to the people.
Based on the formulation of the measures, we expect neither the degree
of fiscal empowerment nor the gain from fiscal decentralisation to be
independent of the population size of a country. Indeed, whereas Table 2
shows that the EDR suffers from a positive population bias as expected
(more populous countries tend to have a higher EDR), the table further
shows that, also as expected, the proposed measure of fiscal empowerment is
negatively correlated with population (albeit only slightly), since – all else
equal – the degree of public empowerment over central government
spending decreases as the national population increases.
2. Analysis of the gain in fiscal empowerment due to decentralisation
As noted in Section IV, we believe that the gain in fiscal empowerment due
to decentralised public spending represents a conceptually superior measure
of fiscal decentralisation compared with the EDR. This alternative measure
of fiscal decentralisation shows a positive and relatively strong correlation
with the EDR (r = 0.82), thereby giving some support to the use of the EDR
as an appropriate measure of fiscal decentralisation. However, squaring the
Pearson’s correlation coefficient suggests that the EDR is only able to
explain about two-thirds of the variation in the gain in empowerment due to
decentralisation (r2 = 0.67). Thus, even if we set aside the fact that we are
unable in the current analysis to properly measure the degree of control or
discretion that subnational governments have over the subnational
expenditures, we can unequivocally say that one-third of the variation in the
gains from fiscal decentralisation cannot be explained by the mere scope of
fiscal devolution (as measured by the EDR), and is determined by political
and institutional factors instead. Given this divergence between the new
measure of fiscal decentralisation and the EDR, it would be a fruitful
exercise to pursue re-estimating existing empirical models of fiscal
decentralisation based on the newly-proposed decentralisation measure
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
456
Fiscal Studies
rather than the EDR, as this might lead to different outcomes and
conclusions from those previously reached in the empirical literature.
Based on the chosen value for γ, the descriptive statistics presented in
Table 2 suggest that the average gain in empowerment due to fiscal
decentralisation in the selected sample is close to 0.02 (ranging from almost
0 to approximately 0.08), which constitutes about 3.5 per cent increase in the
average degree of fiscal empowerment. This amount could be thought of as
an indication of the ‘empowerment gain’ of decentralised public service
provision as a result of gains in allocative efficiency as well as increased
technical efficiency due to improved accountability. While the total
estimated gains from decentralisation could amount to a substantial
improvement in the efficient allocation of public resources in a country, to
be fair this gain would have to be balanced against any potential costs and
inefficiencies associated with decentralisation reforms. Furthermore, it is
unclear whether the average estimated gain in empowerment of 2 cents per
dollar of public spending is biased either up or down, as the exact size of the
gain in empowerment due to decentralisation depends on the selection of γ,
with smaller values of γ resulting in smaller gains in empowerment from
decentralisation. Although the selection of γ = 0.025 is sufficient to indicate
the relative degree of empowerment among different countries in the current
analysis, more careful consideration should be given in future work to
quantifying the exact relationship between jurisdiction size, voice,
heterogeneity and empowerment, as well as to the functional form of the
relationship between these variables.
Similarly to the positive correlation between national population and the
EDR, Table 2 suggests that the gains that are achieved from decentralisation
are systematically and positively correlated with the population size of a
country. This is consistent with the notion that more populous countries
decentralise more, as they have a bigger potential for receiving gains from
devolving expenditures, whereas there is less of a benefit for smaller
countries from decentralising expenditure functions to the local government
level.
3. Fiscal empowerment and decentralisation by type of country
A final consideration in the analysis of the proposed measures of
decentralisation and empowerment is how these measures break down by
type of country. Descriptive statistics for the various measures are presented
in Table 3 for selected OECD countries, transition countries and developing
countries.
There is both theoretical and empirical evidence to suggest that the level
of economic development in a country and the quality of governance are
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
457
TABLE 3
Alternative measures of fiscal decentralisation and empowerment:
descriptive statistics and correlations by type of country
EDR
OECD countries
Mean
Standard deviation
Minimum
Maximum
Fiscal
empowerment
Gain in
empowerment
Population
(million)
0.3179
0.1671
0.0371
0.6209
0.5284
0.0946
0.3279
0.6711
0.0233
0.0122
0.0046
0.0515
33.97
62.62
0.29
293.66
Correlations
EDR
Fiscal empowerment
Gain in empowerment
Population
1.0000
–0.0958
0.7823
0.2646
1.0000
0.1117
–0.2634
1.0000
0.3908
Transition countries
Mean
Standard deviation
Minimum
Maximum
0.2228
0.0905
0.0670
0.3971
0.4267
0.0753
0.2717
0.6134
0.0114
0.0060
0.0048
0.0267
Correlations
EDR
Fiscal empowerment
Gain in empowerment
Population
1.0000
0.3444
0.4093
0.5046
1.0000
0.3601
–0.2433
1.0000
0.1703
1.0000
Developing countries
Mean
Standard deviation
Minimum
Maximum
0.2552
0.2087
0.0330
0.7866
0.4744
0.1174
0.1930
0.6641
0.0156
0.0176
0.0013
0.0787
182.75
359.47
4.25
1,288.40
Correlations
EDR
Fiscal empowerment
Gain in empowerment
Population
1.0000
0.3115
0.8841
0.7129
1.0000
0.3739
–0.0677
1.0000
0.7087
1.0000
1.0000
17.89
32.94
1.36
143.85
important determinants of the degree of fiscal decentralisation (e.g. Panizza,
1999). Among other reasons, developed economies tend to have a lower
minimum efficient scale for delivering public services and often have a
stronger tradition of democratic governance; they are therefore able to
efficiently deliver more public services ‘closer to the people’. The numbers
for the EDR in Table 3 confirm this trend: OECD countries are the most
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
458
Fiscal Studies
decentralised (typically around 30 per cent), while transition countries and
developing countries are typically less decentralised (20–25 per cent). Of
course, as already noted, the exact statistics will fluctuate depending on the
composition of the countries included in the sample. Sample selection issues
may explain the fact that Table 3 indicates that developing countries have a
slightly higher EDR than transition countries.12
Consistent with the higher-than-expected EDR for developing economies
in the current sample, it is interesting to note that the newly-proposed
measures follow the same pattern between decentralisation and development:
all proposed measures of decentralisation and empowerment concur that
developed economies are consistently among the most fiscally decentralised
and empowered, while – in the current sample, at least – transition
economies are among the least fiscally empowered and decentralised.
Two explanations may be offered for this pattern. First, transition
economies have a decidedly lower voter turnout rate (62 per cent) than the
industrialised and developing countries in the sample (76 per cent and 72 per
cent respectively). Therefore, in addition to a relatively low level of fiscal
devolution, a lower level of political responsiveness seems to result in a
lower degree of fiscal empowerment in transition economies. Second, as
suggested by the population statistics, many transition countries are
relatively small, and therefore there is relatively little to gain from
decentralisation reforms, explaining the absence of substantial gains in fiscal
empowerment as a result of decentralisation.
In contrast, the developing economies included in the sample tend to be
relatively populous (compared with industrialised economies, but especially
when compared with transition economies), seem to have relatively
substantive political participation, and spend a reasonable share of their
public resources at the subnational level. For those developing countries
included in the sample, the level of fiscal empowerment and the gains from
decentralisation exceed those in transition countries and approach the levels
realised in industrialised economies.
VII. Concluding considerations and remarks
What we sought to accomplish in the current paper was to arrive at a single
variable that quantifies the overall degree of fiscal empowerment that is
achieved through fiscal and political decentralisation, thereby facilitating
direct comparisons of the degree of fiscal empowerment and decentralisation
between different countries and over time.
12
Developing countries that were shown to be fully centralised (with an EDR of 0) were excluded from
the current sample. As a result, the developing countries included in the sample tend to be more populous
and/or more decentralised than an ‘average’ developing country (for example, compare with Bahl and
Wallace (2003)).
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
459
The initial computation and empirical analysis of the proposed measures
underscore the constraints in fulfilling this ambitious purpose. Nonetheless,
based on an evolving and deeper understanding of fiscal decentralisation and
empowerment, the alternative measures of fiscal decentralisation and
empowerment proposed in this paper are a relevant and valuable addition to
the analysis and measurement of fiscal decentralisation. The proposed
measures of fiscal decentralisation are empirically relevant in a number of
different ways, and complement existing discussions of the topic (e.g. Ebel
and Yilmaz, 2002; Treisman, 2002) as well as adding to ongoing efforts to
improve the measurement and understanding of fiscal decentralisation
(Stegarescu, 2005; Bell et al., 2006).
To the extent that the expenditure decentralisation ratio (EDR) is
currently used as a proxy for the degree of empowerment that individuals
have over public finances decisions in their country, we believe that the
proposed measures of fiscal empowerment and decentralisation offer
conceptually superior measures. However, the adequacy of the different
decentralisation measures depends on the specific purpose of investigation.
For instance, if one is interested in quantifying the amount of public goods
consumed at each level of government, then the EDR may provide a fairly
good approximation. On the other hand, if one is interested in the benefits of
decentralisation in terms of improved allocative efficiency, greater fiscal
competition and stronger public sector accountability, then the newlyproposed measures of decentralisation accounting for subnational fiscal
autonomy, jurisdiction size and political responsiveness are much more
relevant.
However, quantifying fiscal empowerment and decentralisation in
accordance with Bahl’s (2005) definition of fiscal decentralisation as done in
this paper should only be considered as the first step in exploring the
relationship between voice, fiscal empowerment and the structure of the
public sector. At least three logical avenues for further exploration should be
pursued based on the measures proposed in the current paper.
The first issue left for future work is the more careful consideration of the
measurement and determinants of political responsiveness (π) as well as a
more in-depth analysis of the relationship between voice and jurisdiction
size, including the voice-loss parameter used in the proposed measures (γ).
For instance, how does voice-loss differ from country to country based on
different aspects of population heterogeneity?
Second, one of the main shortcomings identified in the initial empirical
analysis is the absence of data on subnational expenditure discretion at each
government level (αi). Instead, we were forced to use the IMF’s Government
Finance Statistics (GFS) expenditure data, which do not reflect the degree of
control or discretion that each government level has over its resources. As
more detailed data on intergovernmental finances become available, the
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
460
Fiscal Studies
estimates should be updated to properly reflect the share of public spending
controlled by each government level rather than relying on the share of total
spending that occurs at each level of government.
Finally, the initial empirical analysis indicates that at a minimum, onethird of the variation in fiscal empowerment is determined by variations in
political responsiveness and subnational government structures. Given the
different patterns of fiscal empowerment and decentralisation revealed by
the new measures, a final logical next step would be to replicate and reestimate selected existing empirical models of decentralisation using the new
measures of fiscal decentralisation. Such empirical analysis could shed
additional light on the determinants of fiscal decentralisation and
empowerment, as well as on the relationships between fiscal decentralisation
and empowerment on one hand and economic processes (such as economic
growth, the size of the public sector and the efficiency of public service
delivery) on the other hand.
Appendix
Notes to Table A1
Ni is the number of jurisdictions, whereas αi is the share of public expenditure at government level i; C =
central; R = regional; L = local. Data sources and variables are defined in the text; fiscal empowerment is
defined in equation (5). When αR = 0, regional jurisdictions are administrative tiers of the central
government.
Notes to Table A2
a
Country-specific fiscal data were relied on rather than IMF’s Government Finance Statistics.
Notes: Ni is the number of jurisdictions, whereas αi is the share of public expenditure at government level
i; C = central; R = regional; L = local. Data sources and variables are defined in the text; fiscal
empowerment is defined in equation (5). When NR = NL, we presume that the highest subnational
government level has the greatest degree of control over subnational spending. When αR = 0, regional
jurisdictions are administrative tiers of the central government.
Notes to Table A3
a
Country-specific fiscal data were relied on rather than IMF’s Government Finance Statistics.
Notes: Ni is the number of jurisdictions, whereas αi is the share of public expenditure at government level
i; C = central; R = regional; L = local. Data sources and variables are defined in the text; fiscal
empowerment is defined in equation (5). When αR = 0, regional jurisdictions are generally administrative
tiers of the central government.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
TABLE A1
Alternative measures of fiscal decentralisation and empowerment: selected OECD countries
Country
Year
Australia
Austria
Belgium
Canada
Denmark
Finland
France
Germany
Greece
Iceland
Israel
Italy
Luxembourg
Netherlands
New Zealand
Norway
Portugal
Spain
Sweden
Switzerland
UK
United States
2004
2003
2003
2004
2004
2004
2004
2004
2000
2002
2004
2003
2004
2004
2004
2004
2002
2003
2003
2002
2004
2002
Pop.
(m)
20.11
8.12
10.38
31.97
5.40
5.23
60.38
82.52
10.92
0.29
6.80
57.65
0.45
16.28
4.06
4.59
10.37
42.00
8.96
7.28
59.87
293.66
Voter
turnout
0.94
0.84
0.96
0.61
0.85
0.67
0.60
0.78
0.77
0.88
0.68
0.84
0.92
0.80
0.80
0.77
0.64
0.76
0.80
0.45
0.61
0.69
NR
NL
αC
αR
αL
EDR
8
9
11
13
13
6
96
16
52
8
6
107
3
12
17
19
20
17
21
26
4
51
664
2,359
589
4,565
270
430
36,772
16,095
5,828
105
268
8,101
118
504
74
434
308
8,108
290
2,869
434
38,967
0.5536
0.7011
0.6458
0.3791
0.4022
0.6261
0.8147
0.6274
0.9629
0.7318
0.8926
0.7053
0.8932
0.6528
0.9140
0.7037
0.8769
0.7086
0.5581
0.4452
0.7214
0.4896
0.3850
0.1611
0.2243
0.4564
0.0000
0.0000
0.0000
0.2215
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.1825
0.0000
0.3358
0.0000
0.2716
0.0615
0.1378
0.1299
0.1645
0.5978
0.3739
0.1853
0.1510
0.0371
0.2682
0.1074
0.2947
0.1068
0.3472
0.0860
0.2963
0.1231
0.1089
0.4419
0.2190
0.2786
0.2388
0.4464
0.2989
0.3542
0.6209
0.5978
0.3739
0.1853
0.3726
0.0371
0.2682
0.1074
0.2947
0.1068
0.3472
0.0860
0.2963
0.1231
0.2914
0.4419
0.5548
0.2786
0.5104
Fiscal
empow.
0.6388
0.5882
0.6664
0.4226
0.6250
0.4808
0.4068
0.5208
0.5154
0.6617
0.4649
0.5746
0.6711
0.5591
0.5543
0.5531
0.4375
0.5081
0.5729
0.3279
0.4103
0.4643
Gain in
empow.
0.0194
0.0219
0.0234
0.0272
0.0515
0.0277
0.0215
0.0282
0.0046
0.0212
0.0074
0.0398
0.0090
0.0309
0.0054
0.0256
0.0081
0.0200
0.0361
0.0234
0.0179
0.0423
Rank
EDR
8
20
18
2
3
14
41
15
58
26
50
22
51
19
54
21
45
23
9
5
25
7
Rank
FE
5
10
2
44
6
28
48
20
22
4
35
12
1
14
15
16
41
24
13
55
46
36
Rank
gain
23
16
15
11
2
10
17
9
56
18
43
4
39
7
53
13
42
21
5
14
25
3
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
TABLE A2
Alternative measures of fiscal decentralisation and empowerment: selected transition countries
(Commonwealth of Independent States & Eastern Europe)
Country
Year
Pop.
(m)
Voter
turnout
NR
NL
αC
αR
αL
EDR
Fiscal
empow.
Gain in
empow.
Rank
EDR
Rank
FE
Rank
gain
Armenia
Belarus
Bulgaria
Croatia
Czech Rep.
Estonia
Georgiaa
Hungary
Kazakhstan
Kyrgyz Rep.
Latvia
Lithuania
Moldova
Poland
Romania
Russian Fed.
Slovak Rep.
Slovenia
Ukraine
2004
2004
2004
2004
2004
2001
2004
2003
2004
2001
2004
2004
2004
2003
2002
2004
2003
2004
2004
3.03
9.82
7.76
4.44
10.22
1.36
4.52
10.13
14.99
4.95
2.31
3.44
4.22
38.20
21.80
143.85
5.38
2.00
47.45
0.55
0.90
0.56
0.62
0.58
0.58
0.64
0.74
0.68
0.64
0.71
0.46
0.65
0.41
0.59
0.56
0.55
0.61
0.72
11
7
45
144
77
15
12
42
16
8
33
10
12
16
42
89
8
12
27
972
7
345
426
6,200
247
966
3,177
16
8
545
60
12
373
42
89
79
193
27
0.9330
0.6355
0.8508
0.9090
0.7784
0.7570
0.8841
0.7562
0.6029
0.7563
0.7519
0.7937
0.7639
0.7781
0.8139
0.6084
0.8308
0.8216
0.7409
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0670
0.3645
0.1492
0.0910
0.2216
0.2430
0.1159
0.2438
0.3971
0.2437
0.2481
0.2063
0.2361
0.2219
0.1861
0.3916
0.1692
0.1784
0.2591
0.0670
0.3645
0.1492
0.0910
0.2216
0.2430
0.1159
0.2438
0.3971
0.2437
0.2481
0.2063
0.2361
0.2219
0.1861
0.3916
0.1692
0.1784
0.2591
0.3849
0.6134
0.3841
0.4271
0.4077
0.4191
0.4451
0.5178
0.4654
0.4437
0.5145
0.3235
0.4492
0.2717
0.3904
0.3645
0.3785
0.4322
0.4745
0.0048
0.0109
0.0088
0.0063
0.0209
0.0145
0.0095
0.0267
0.0129
0.0057
0.0209
0.0070
0.0067
0.0093
0.0070
0.0162
0.0073
0.0106
0.0103
55
17
44
53
36
33
48
31
12
32
30
37
34
35
40
13
43
42
29
50
7
51
43
47
45
39
21
34
40
23
56
38
58
49
53
52
42
31
55
34
40
50
19
30
37
12
32
52
20
46
49
38
47
28
44
35
36
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
TABLE A3
Alternative measures of fiscal decentralisation and empowerment: selected developing countries
Country
Year
Pop.
(m)
Nigeriaa
South Africa
Tanzaniaa
Uganda
1999
2004
2004
2003
Argentina
Bolivia
Brazil
Chile
Costa Rica
El Salvador
Mexico
Peru
Bangladesha
China, mainland
Indiaa
Indonesiaa
Iran, Islam. Rep.
Malaysia
Thailand
αC
αR
αL
EDR
Fiscal
empow.
Gain in
empow.
Rank
EDR
Rank
FE
Rank
gain
774
284
122
160
0.8120
0.4447
0.8818
0.7354
0.1400
0.3716
0.0000
0.0000
0.0480
0.1837
0.1182
0.2646
0.1880
0.5553
0.1182
0.2646
0.3170
0.6641
0.4768
0.4593
0.0068
0.0309
0.0071
0.0159
39
4
47
28
57
3
30
37
48
6
45
29
24
9
27
51
7
14
32
26
1,617
314
5,563
345
81
262
2,454
1,821
0.5915
0.7200
0.5835
0.8787
0.9670
0.9551
0.8090
0.7341
0.3301
0.1983
0.2775
0.0000
0.0000
0.0000
0.1485
0.1819
0.0784
0.0817
0.1390
0.1213
0.0330
0.0449
0.0425
0.0841
0.4085
0.2800
0.4165
0.1213
0.0330
0.0449
0.1910
0.2659
0.4780
0.5797
0.5481
0.5902
0.4716
0.1930
0.2692
0.5969
0.0198
0.0135
0.0297
0.0110
0.0018
0.0013
0.0059
0.0190
11
24
10
46
60
56
38
27
29
11
17
9
33
60
59
8
22
31
8
33
59
60
51
24
64
31
35
33
30
16
76
4,900
3,143
585
440
314
141
876
0.9600
0.2134
0.4798
0.6323
0.9634
0.8898
0.9006
0.0000
0.0000
0.5202
0.3677
0.0000
0.0711
0.0000
0.0400
0.7866
0.0000
0.0000
0.0366
0.0391
0.0994
0.0400
0.7866
0.5202
0.3677
0.0366
0.1102
0.0994
0.4738
0.5274
0.3596
0.5379
0.4948
0.4879
0.4880
0.0044
0.0787
0.0166
0.0175
0.0028
0.0050
0.0088
57
1
6
16
59
49
52
32
19
54
18
25
27
26
57
1
27
26
58
54
41
Voter
turnout
NR
NL
111.26
45.51
34.57
26.87
0.49
0.98
0.73
0.68
37
9
21
70
2004
2004
1998
2004
2004
2004
2000
2004
38.37
9.01
173.86
16.12
4.25
6.76
97.97
27.56
0.71
0.85
0.83
0.88
0.69
0.28
0.42
0.89
2004
2003
2004
2004
2004
2003
2004
139.21
1,288.40
1,079.72
217.59
67.01
24.44
63.69
0.75
0.76
0.58
0.84
0.77
0.74
0.75
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
464
Fiscal Studies
References
Ahmad, J., Devarajan, S., Khemani, S. and Shah, S. (2005), ‘Decentralization and service
delivery’, World Bank, Working Paper no. 3603.
Alsop, R., Bertelsen, M. F. and Holland, J. (2006), Empowerment in Practice: From Analysis
to Implementation, Washington, DC: World Bank.
Asfaw, A., Frohberg, K., James, K. S. and Jütting, J. (2007), ‘Fiscal decentralization and
infant mortality: empirical evidence from rural India’, Journal of Developing Areas, vol.
41, pp. 17–35.
Bahl, R. W. (2005), Fiscal Decentralization 101, Atlanta, GA: Andrew Young School of
Policy Studies, Georgia State University.
— and Linn, J. (1992), Public Finance in Developing Countries, Washington, DC: Oxford
University Press.
— and Wallace, S. (2003), ‘Fiscal decentralization, the provincial-local dimension’, in J.
Martinez-Vazquez and J. Alm (eds), Public Finance in Developing and Transitional
Countries: Essays in Honor of Richard Bird, Cheltenham: Edward Elgar Press.
Bardhan, P. and Mookherjee, D. (2006a), ‘Decentralisation and accountability in
infrastructure delivery in developing countries’, Economic Journal, vol. 116, pp. 101–27.
— and — (2006b), ‘Pro-poor targeting and accountability of local governments in West
Bengal’, Journal of Development Economics, vol. 79, pp. 303–27.
Bell, M. E., Ebel, R. D., Kaiser, K. and Rojchaichaninthorn, J. (2006), ‘Measuring fiscal
decentralization: a new perspective’, World Bank, Working Paper (draft 25 August 2006).
Bird, R. M., Ebel, R. D. and Wallich, C. I. (eds) (1996), Decentralization of the Socialist
State: Intergovernmental Finance in Transition Economies, Washington, DC: World
Bank Regional and Sectoral Studies.
Boex, J., Martinez-Vazquez, J. and Timofeev, A. (2004), ‘Subnational government structure
and intergovernmental fiscal relations’, Georgia State University, Andrew Young School
of Policy Studies, International Studies Program, Working Paper no. 04-01.
Borck, R. (2002), ‘Jurisdiction size, political participation, and the allocation of resources’,
Public Choice, vol. 113, pp. 251–63.
Dabla-Norris, E. (2006), ‘The challenge of fiscal decentralisation in transition countries’,
Comparative Economic Studies, vol. 48, pp. 100–31.
Davoodi, H. and Zou, H-F. (1998), ‘Fiscal decentralization and economic growth: a crosscountry study’, Journal of Urban Economics, vol. 43, pp. 244–57.
De Mello, L. R. (2000), ‘Fiscal decentralization and intergovernmental fiscal relations: a
cross-country analysis, World Development, vol. 28, pp. 365–80.
Ebel, R. D. and Yilmaz, S. (2002), ‘On the measurement and impact of fiscal
decentralization’, World Bank, Policy Research Working Paper no. 2809.
Guan, M. and Green, D. P. (2006), ‘Noncoercive mobilization in state-controlled elections: an
experimental study in Beijing’, Comparative Political Studies, vol. 39, pp. 1175–93.
Hunter, J. S. (1977), Federalism and Fiscal Balance, Canberra: Australian National
University Press.
Kandeva, E. (ed.) (2001), Stabilization of Local Governments, Budapest: LGI Books.
Keefer, P. (2004), ‘A review of the political economy of governance: from property rights to
voice’, World Bank, Policy Research Working Paper no. 3315.
Law, G. (2006), Administrative Divisions of Countries (‘Statoids’), http://www.statoids.com.
Litvack, J. and Seddon, J. (eds) (1999), ‘Decentralization briefing notes’, World Bank
Institute, Working Paper no. 19863.
Martinez-Vazquez, J. and Boex, J. (2001), Russia’s Transition to a New Federalism,
Washington, DC: World Bank Institute.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008
Fiscal decentralisation and empowerment
465
— and McNab, R. (2005), ‘Fiscal decentralization, macrostability, and growth’, Georgia
State University, Andrew Young School of Policy Studies, International Studies Program
Working Paper no. 05-6.
Meloche, J-P., Vaillancourt, F. and Yilmaz, S. (2004), ‘Decentralization or fiscal autonomy?
What does really matter? Effects on growth and public sector size in European transition
countries’, World Bank, Policy Research Working Paper no. 3254.
Munck, G. L. and Verkuilen, J. (2002), ‘Conceptualizing and measuring democracy:
evaluating alternative indices’, Comparative Political Studies, vol. 35, pp. 5–34.
Munteanu, I. and Popa, V. (eds) (2001), Developing New Rules in the Old Environment,
Budapest: LGI Books.
Nelson, M. A. (1990), ‘Decentralization of the subnational public sector: an empirical
analysis of the determinants of local government structure in metropolitan areas in the
U.S.’, Southern Economic Journal, vol. 57, pp. 443–57.
Oates, W. E. (1972), Fiscal Federalism, New York: Harcourt-Brace-Janovich.
— (1985), ‘Searching for Leviathan: an empirical study’, American Economic Review, vol.
75, pp. 748–57.
O’Donnell, G., Iazetta, O. and Vargas, J. (2003), Democracia, desarrollo humano y
ciudadanía (Democracy, Human Development and Citizenship), Santa Fe: Homo Sapiens
Ediciones.
OECD (1997), Managing across Different Levels of Government, Paris: Organisation for
Economic Cooperation and Development.
— (1999), Taxing Powers of State and Local Government, OECD Tax Policy Study no. 1,
Paris: Organisation for Economic Cooperation and Development.
— (2001), Fiscal Design across Levels of Government: Year 2000 Surveys, Paris:
Organisation for Economic Cooperation and Development.
Olken, B. (2007), ‘Monitoring corruption: evidence from a field experiment in Indonesia’,
Journal of Political Economy, vol. 115, pp. 200–49.
Panizza, U. (1999), ‘On the determinants of fiscal centralization: theory and evidence’,
Journal of Public Economics, vol. 74, pp. 97–139.
Pritchett, L. and Woolcock, M. (2002), ‘Solutions when the solution is the problem: arraying
the disarray in development’, Center for Global Development, Working Paper no. 10.
Raich, U. (2005), ‘Fiscal determinants of empowerment’, World Bank, Policy Research
Working Paper no. 3705.
Stegarescu, D. (2005), ‘Public sector decentralisation: measurement concepts and recent
international trends’, Fiscal Studies, vol. 26, pp. 301–33.
Tiebout, C. (1956), ‘A pure theory of local expenditures’, Journal of Political Economy, vol.
64, pp. 416–24.
Treisman, D. (2002), Defining and Measuring Decentralization: A Global Perspective, Los
Angeles: Department of Political Science, University of California, Los Angeles.
Weingast, B. (2006), ‘Second generation fiscal federalism: implications for decentralized
democratic governance and economic development’, mimeo.
© 2008 The Authors
Journal compilation © Institute for Fiscal Studies, 2008