See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/312201006

Different forms of decentralization and their impact on government

performance: Micro-level evidence from 113 countries

Article  in  Economic Modelling · January 2017

DOI: 10.1016/j.econmod.2016.12.010

CITATIONS READS

3 3,877

4 authors, including:

Rajeev Goel Ummad Mazhar

Illinois State University Lahore University of Management Sciences
192 PUBLICATIONS   2,568 CITATIONS    28 PUBLICATIONS   52 CITATIONS   

SEE PROFILE SEE PROFILE

Michael A. Nelson
University of Akron
48 PUBLICATIONS   1,487 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Innovation and Internationalization View project

http://www.ummadmazhar.com/ordinary-business-of-life View project

All content following this page was uploaded by Ummad Mazhar on 25 January 2018.

The user has requested enhancement of the downloaded file.

 Economic Modelling 62 (2017) 171–183

Contents lists available at ScienceDirect

Economic Modelling
journal homepage: www.elsevier.com/locate/econmod

Different forms of decentralization and their impact on government MARK

performance: Micro-level evidence from 113 countries☆
⁎
Rajeev K. Goela, Ummad Mazharb, Michael A. Nelsonc, , Rati Rama
a
Illinois State University, USA
b
Forman Christian College, Pakistan
c
University of Akron, USA

A R T I C L E I N F O A BS T RAC T

JEL classification: This study of the effect of decentralization on government performance differs from the vast literature on the
H11 topic in three major ways. First, we compare the effects of four different forms of decentralization, namely, fiscal
H70 decentralization, administrative decentralization, federalism, and aggregate decentralization. Second, we study
K2 the effect on three dimensions of government performance, namely, tax administration, business licensing and
K4
permits, and corruption. Third, the effects are judged in terms of the perceptions of nearly 100,000 business
Keywords: leaders located in 113 countries. Seven statements summarize our findings. First, fiscal and administrative
Fiscal decentralization decentralization improve perceptions of government performance. Second, federalism is associated with worse
Administrative decentralization
perceptions. Third, the effect of aggregate decentralization is ambiguous. Fourth, service industries tend to
Federalism
perceive decentralization more favorably than others. Fifth, large firms perceive decentralization less favorably
Aggregate decentralization
Government performance than other firms. Sixth, the effect of the same form of decentralization varies in magnitude across the three
government activities. Seventh, therefore, decentralization may not be oversold as a policy prescription to
improve government performance; the form of decentralization and its contextualization in terms of the
targeted area of government activity need careful consideration.

1. Introduction
The role that decentralized government structures might play in
promoting good governance has intrigued policymakers and research-
ers for some time. On the one hand, decentralized governments are
closer to the populace and can better understand and deliver services
that are tailored to specific needs. There is also greater transparency of
government actions as citizens can better observe government actions
at the local level. The scope of this has somewhat broadened with e-
government services. On the other hand, there are some drawbacks to
decentralization, notably regarding the (in)ability to coordinate the
provision of services at various levels of government (see Prud’homme
(1995), Shah and Thompson (2004)), the appropriation of benefits due
to spillovers across jurisdictional boundaries (Oates (1972)), and
potentially less media scrutiny of the actions of public officials.
Overall, the trend towards greater decentralization has been promoted
by major international organizations (e.g., International Monetary
Fund (2009), United Nations (2009) and the World Bank (1999))
and has found favor in recent years, as noted by Rodríguez-Pose and
Gill (2003), Stegarescu (2005) and the World Bank (1999).1
Given its numerous possible dimensions, several classifications of
government decentralization have been proposed. For instance, as
noted in a recent survey by Martinez-Vazquez et al. (2016), government
decentralization may be viewed along expenditure/revenue, policy, or
political dimensions (also see Rodden (2004)). While the expenditure/
revenue aspect of decentralization has been most widely studied,
overlapping government jurisdictions and institutional complexities
make clear inferences difficult. Other aspects also come into play,
leading to ambiguities in formal analyses. For instance, a reliable set of
answers or clear policy prescriptions have failed to evolve due to a
number of reasons. First, no two nations (or even regions within a
country) are alike and differ in numerous aspects. This poses chal-
lenges in terms of applying similar policies across jurisdictions (Should
there be x number of government hospitals for every y population in
every jurisdiction? Then, what about tropical regions that are more
prone to certain diseases?). Second, there are many aspects to

☆
An earlier version of this paper was circulated as CESifo working paper #6013 ("Government decentralization and international government performance"). Extremely useful
comments from three anonymous reviewers and the editor are gratefully acknowledged. The usual disclaimer applies.
⁎
Correspondence to: Department of Economics, University of Akron, OH 44325, USA.
E-mail address: nelson2@uakron.edu (M.A. Nelson).
1
See also The Online Sourcebook on Decentralization and Local Development (http://www.ciesin.columbia.edu/decentralization/).

http://dx.doi.org/10.1016/j.econmod.2016.12.010
Received 6 August 2016; Received in revised form 10 December 2016; Accepted 10 December 2016
Available online 10 January 2017
0264-9993/ © 2016 Elsevier B.V. All rights reserved.
R.K. Goel et al.
government decentralization that are qualitatively (in terms of deliver-
ing the services) and operationally (in terms of setup costs and
timelines) different. For instance, geographic and administrative
decentralization are legislatively quite time consuming to alter,
whereas fiscal decentralization is somewhat easily altered. Thus, it is
quite difficult to compare the effects of different forms of government
decentralization. The related measurement issues have been noted by
several scholars, including Blume and Voigt (2011), Brueckner (2003),
Prud’homme (1995), and Stegarescu (2005).2 These issues, coupled
with the level of aggregation available in most data, have led to
ambiguous findings regarding the effects of decentralization on govern-
ment performance (see Martinez-Vazquez et al. (2016)).
The present paper attempts to address the ambiguity in the effects
of decentralization in several ways. First, we analyze the effects of
different types of decentralization on government performance in a
cross-country context. These include fiscal decentralization and feder-
alism, which have been widely used in past literature, and two
relatively new indicators (see Ivanyna and Shah (2014)): “aggregate
decentralization” and “administrative decentralization.” The latter two
measures are designed to assess the independence and actual authority
in the hands of local authorities when it comes to decision making.
Such measures are intended to mitigate the comparability problems
with the more traditionally-used decentralization measures discussed
above, especially within the context of cross-country data sets.
Second, we address the impacts or consequences of decentralization
from business leaders' perceptions of government performance as they
relate to running their businesses. These include business manager
perceptions as to how government licensing and permits, taxation
(specifically, tax administration), and corruption affect their opera-
tions.3 Government performance in the licensing and permits area is
especially relevant for the purpose at hand as it is an activity that is
typically carried out at the local level (e.g., liquor licenses for bars,
zoning permits, etc.). We can assess how greater decision-making
authority in the hands of local officials affects business perceptions of
government performance in this area. Further, corruption is prevalent
in many societies and permeates many levels of government in its
interactions with the public and corruption at the local level can be a
significant issue, perhaps more so than at the national level
(Transparency International (2009)). Bribes may be offered and/or
solicited to speed up government functions or to get undue favors from
government officials. These corrupt transactions might be facilitated at
the local level due to closer proximity between bribe givers and bribe
takers. Our analysis examines business' perceptions of the impact of
corruption on their operations and how such perceptions are affected
by the type of decentralization. Although the decentralization-corrup-
tion relation has been extensively studied in the literature (e.g., Fan
et al. (2009), Fisman and Gatti (2002), Goel and Nelson (2011),
Treisman (2000)), the present analysis is at the micro level and is based
on perceptions of business leaders. This is in contrast to the majority of
the literature that uses aggregate indices of corruption perceptions (see
Lambsdorff (2006), Treisman (2007)).
Third, the underlying data draw upon thousands of individual
responses regarding perceptions of government services across more
than a hundred countries. Thus, we are able to examine how certain
widely prevalent government functions are perceived following differ-
ent types of decentralization. All these aspects enable us to address the
role of decentralization in a deeper (using micro-level data) and wider
(across various types of decentralization and many nations) manner
than previously considered in the literature. Specifically, while some
studies in the literature have examined the effects of decentralization
on government performance in some areas (see, for example, Adam
2 4
Also see Knack and Keefer (1995) for a broader perspective.
3
These aspects of government performance can be seen as tied to the wider discussion
on the quality of government (see La Porta et al. (1999)).
172
Economic Modelling 62 (2017) 171–183
et al. (2014)), the scope of analysis (both in terms of decentralization
types and the government performance aspects considered) and the
consideration of perceptions of government performance by actual
users (i.e., business firms) are new.
Key questions addressed in this research are:
• Are the effects of a government decentralization measure similar
across different dimensions of government performance?
• Do government performance perceptions differ across various
measures of decentralization?
The analysis, based on World Enterprise Surveys of nearly 100,000
individual business owners and top managers in 113 countries, shows
that fiscal decentralization and administrative decentralization are
qualitatively alike in that greater decentralization in each case improves
perceptions of government performance. In contrast, opposite conclu-
sions are reached when decentralization is based on distinctions
between federalist states and unitary states. With regard to tax
administration particularly, fiscal and administrative forms of govern-
ment decentralization result in better outcomes than aggregate decen-
tralization. At a related level, while a larger government results in
worse outcomes with regard to tax administration and business
licensing, it improves perceptions of government performance with
regard to corruption. This finding is consistent with the view that a
larger government machinery involves greater attention to checks and
balances. We also find that relative to the entire sample, business
manager perceptions are different in service industries and at large
firms. Overall, our findings provide useful insights into the effects of
various forms of devolution of powers to subnational government units.
We turn next to a discussion of the underlying theory and related
literature.
2. Literature and theoretical background
The broader literature on the choice by governments to decentralize
operations (and provision of services) can be tied to the work of Besley
and Coate (2003), Oates (1972), and Panizza (1999); also see Faguet
(2014). In a nutshell, government decentralization can have positive as
well as negative effects on public sector performance. On the positive
side, decentralization of government functions leads to greater electoral
control and greater yardstick competition among competing jurisdic-
tions (Salmon (1987)). On the down side, decentralization reduces
scale economies in the provision of government services, creates the
potential for mismatch between the spatial incidence of the benefits for
government services and the political boundaries, empowers local
interest groups to better drive their agendas, and likely makes
attracting qualified applicants for government jobs difficult (see
Adams et al. (2014), Prud’homme (1995)).4 Thus, analyses with greater
size and scope of data are required to discern some of the underlying
linkages - which this paper attempts to do.
The literature on the impacts of decentralization has mostly focused
on the effects of fiscal decentralization on various indicators of
government and macroeconomic performance (see, for example,
Adam et al. (2014), Ebel and Yilmaz (2003), Kyriacou and Roca-
Sagalés (2011a), Lago-Peñas et al. (2011) and Yeung (2009)). The
findings vary and are sensitive to the data and methodology employed.
A few studies examine alternate or multiple forms of decentralization
(see Fan et al. (2009), Goel and Nelson (2011), Goel and Saunoris
(2016), Kyriacou and Roca-Sagalés (2011b), Nelson (2013) and
Rodríguez-Pose and Ezcurra (2011)). However, these broad effects
mask qualitative distinctions across decentralization types and govern-
Bodman et al. (2012) present a theoretical model of the dynamic effects of taxation
and investment on the steady state output level of an economy, with consideration of
multi-level governments.
R.K. Goel et al.
ment services.
One aspect of government performance, namely corruption, has
garnered a fair bit of attention in regard to its nexus with government
decentralization (see Fan et al. (2009), Fisman and Gatti (2002), Goel
and Nelson (2011), Nelson (2013), Treisman (2000); Lambsdorff
(2006) and Treisman (2007) for reviews of the related literature).
Fan et al. (2009) found that in nations with a larger number of
administrative tiers of government and (given local revenues) a larger
number of local public employees, corruption was more prevalent. On
the other hand, when local or central governments received a larger
share of GDP in revenue, corruption was less prevalent. In another take
in a cross-national context, Goel and Saunoris (2016) compare the
relative effects of virtual and physical decentralization in terms of their
mitigating effects on corruption and the shadow economy. They find
virtual decentralization to be relatively more effective. Goel and Nelson
(2011) on the other hand, focused on the effects of decentralization
across states in the United States, distinguishing between general-
purpose and special-purpose governments. They found that while more
general-purpose governments contributed to corruption, the effect of
special-purpose governments was mixed. Similarly, Nelson (2013)
found that more fragmented municipal government structures are
associated with more corruption for a large sample of countries.
Similar mixed results have been found regarding the link between
federal political structures and perceived corruption (see Fan et al.
(2009), Goldsmith (1999), Kunicová and Rose-Ackerman (2005), and
Treisman (2000, 2007)). Generally, this literature has found greater
decentralization to lower corrupt activity. The present research will add
to this discussion by providing evidence on firm's perceptions of
corruption at the local level.
In yet another angle on the impact of decentralization, Kyriacou
and Roca-Sagalés (2011b) study the relation between fiscal and
political decentralization and government quality (measured as the
average of rule of law, regulatory quality, government effectiveness,
and corruption control). The authors find that fiscal decentralization
improves government quality, but not if such decentralization is
accompanied by political decentralization.
Focusing on OECD countries, Rodríguez-Pose and Ezcurra (2011)
consider fiscal, political and administrative measures of decentraliza-
tion and focus on how decentralization impacts economic growth.5
Their findings show a negative relation between fiscal decentralization
and economic growth. However, their analysis reveals that the impact
of political and administrative decentralization on economic growth is
weaker and sensitive to how political decentralization is measured.
There are several important differences between this work and ours.
Specifically, (a) the focus of the present work is on government
performance, rather than economic growth; (b) the sample of countries
covered here is much broader than OECD and includes both developed
and developing nations; (c) the data employed are based on a large
number of micro-level observations of business managers rather than
country-level aggregates; and (d) we also consider a broader range of
aggregate decentralization measures (of which some other decentrali-
zation types (e.g., administrative decentralization, etc.) are sub-com-
ponents).
3. Measuring government performance and decentralization
Since they are central focus of the paper, we provide some details on
how we measure government performance and decentralization in the
analysis that follows. Details about the overall sample and the other
variables used in the analysis are in Section 4. We consider three
aspects of government performance across four forms of government
decentralization.
5
Also see a recent survey by Martinez-Vazquez et al. (2016).
173
Economic Modelling 62 (2017) 171–183
3.1. Government performance
We address the impacts or consequences of decentralization from a
different perspective than has been considered heretofore, specifically
the link between decentralization and business leaders' perceptions of
government performance as it relates to their business.
The data on perceptions of government performance are drawn from
firm-level responses and come from World Enterprise Surveys (WES) of
individual business owners and top managers of 94,000 firms in more
than a hundred countries (http://www.enterprisesurveys.org/). The
WES is a comprehensive annual survey that asks managers about the
various aspects of local business environment.6 Our research focuses on
data from 2006 to 2015, as it is only after 2006 that a standardized
methodology and instrument sets are used for the survey.7 The usable
number of countries in our analysis is 113 due to the need to match the
various variables in the empirical model presented below.
Given the breadth of issues covered in the WES, we focus on the
subset of questions that asks firms about their relationship with the
government. We consider three important factors surrounding govern-
ment-business relations, specifically how tax administration
(TaxAdmin), business licensing and permits (BusLicense), and corrup-
tion (Corrupt) affect the current operation of their establishment.8,9
The three government outcome variables considered in this analysis
- TaxAdmin, BusLicense, Corrupt - are somewhat qualitatively differ-
ent, but all are potentially directly or indirectly affected by the degree of
decentralization of government functions. For instance, greater decen-
tralization would increase transparency and foster greater trust be-
tween the public and government officials, but such proximity might
facilitate corrupt relations or other abuses of government power. Some
taxes are levied and administered by local governments, whereas most
business licenses and permits (e.g., liquor licenses for restaurants) are
conferred by local governments.10
The survey asks respondents or business firms about their first-
hand impressions of various aspects of government performance. For
each dimension of government performance, respondents were asked
to pick one of five levels of the impact of government performance:
whether that dimension of performance was viewed as (i) no obstacle;
(ii) minor obstacle; (iii) moderate obstacle; (iv) major obstacle; or (v)
6
A related measure of the private sector business conditions is World Bank's Doing
Business (DB) Survey, which uses expert opinions and interviews for the purpose. The
DB survey has already acquired wide currency among students of political economy.
However, it is fundamentally different from the WES, with the latter enjoying important
advantages, as shown at length in Hallward-Driemeier and Pritchett (2015). In
particular, the DB Survey, being based on expert opinions, implicitly assumes that firms
are complying with the rules and regulations of their country. This is not the case with
WES. In other words, WES provides the de facto state of the business and legal
environment in which firms are operating, while DB Survey focuses on de jure state.
7
Thus, Enterprises Surveys Indicator Description ESID (2015), August 2015 version
states, “Uniform universe, uniform methodology of implementation, and a core ques-
tionnaire are the basis of the Global methodology under which most Enterprise Surveys
have been implemented since 2006” (p. 2).
8
The reasons for selecting these aspects are their importance in the link between
business environment and governance, and to keep the analysis manageable. Complete
description of indicators covered in the survey is given in ESID. The World Enterprise
Survey Questionnaire Note provides details about the questions posed to managers.
These and the other survey modules related to the methodology of the Enterprise Surveys
are available at http://www.enterprisesurveys.org/methodology). The Business-
Government relations subsection ‘j.30’ covers eight dimensions of firms’ interactions
with the government.
9
Of course, this is not a comprehensive consideration of government performance,
which is multidimensional. For instance, Guccio et al. (2014) examine whether local
governments had better on-time performance in executing public works projects in Italy.
10
Another important distinction is that whereas business licenses may be procured by
potential entrants (new licenses) or by existing firms (recurring licenses), taxes are paid
by existing firms. On the other hand, corruption might be experienced by businesses
throughout their existence, with possible variations across firm size and industry type.
This difference relates to the age of firms and could affect their perceptions of
government performance - existing firms are likely to have a better perspective having
dealt with the full spectrum of government services and their managers are more likely to
have established long-term working relationships with government officials.
R.K. Goel et al.
very severe obstacle. WES coded these responses on a 5-point scale
where 0=no obstacle and 4=very severe obstacle. To facilitate inter-
pretation of the findings, we collapsed these five dimensions into three,
labelled as minor obstacle, moderate obstacle and major obstacle.
The reclassification adopted is based on a simple but intuitive logic.
Specifically, extreme survey responses lying in the neighborhood of
each other like ‘no obstacle’ and ‘minor obstacle’, and ‘major obstacle’
and ‘very severe obstacle’ are not capturing substantially dissimilar
perceptions. Therefore, the differences between these two sets of
responses are likely to be random rather than systematic, given the
inherently imperfect nature of measurement involved in this question.
This reasoning allows us to conclude that we can merge these sets of
responses without affecting the underlying statistical distribution of
responses. The Appendix Table A1 provides additional details on this
classification scheme. The results are generally robust to an analysis of
all categories – see Appendix B.
Our analysis examines the strength of these perceptions for each
form of government decentralization. It investigates whether different
aspects of government performance are equally affected by any given
form of decentralization and whether different forms of decentraliza-
tion affect a given dimension differently. To our knowledge, these
questions and the scope of analysis (in terms of number of nations and
the number of respondents) to address these, have not been previously
considered in the literature.
3.2. Alternate perspectives on decentralization
To address differences across decentralization types, the following
four measures of decentralization are considered:
• Fiscal Decentralization (Fiscal Decent): As noted above, the con-
sequences of decentralization have been most often analyzed in the
literature from the perspective of the degree of fiscal autonomy of
subnational units of government. This has been evaluated in a
number of ways, including (i) local government expenditure share of
overall government expenditures, (ii) subnational tax share, and (iii)
local share of total government employment (Nelson (2013)).
For the purposes of this analysis we selected the former –
expenditure share as our “baseline” decentralization measure. There
is considerable variation among nations with respect to this
measure. For instance, it has the maximum value of 0.55 for
Uzbekistan. The minimum value of the variable is 0 for
Mozambique and Suriname.
• Federalist System of Government (Federalism): In a federalist state,
political power is split in some fashion between the central govern-
ment and subnational (e.g., state, local governments) authorities.
Generally, the country's constitution assigns the duties, rights, and
privileges of each level of government.
A dichotomous indicator is used here that takes a value of one if
the state is classified as "federal" by Elazar (1995) and zero
otherwise. See Nelson (2013) for a summary of the extant literature
that has employed this political decentralization indicator in analyz-
ing public sector corruption.11 Fifteen percent of the countries in our
sample were classified as federalist (Table 1).
• Aggregate Decentralization Index (AggDecent): This is a compre-
hensive measure of the importance and authority of local govern-
ment compiled by Ivanyna and Shah (2014). The index starts with
the fiscal decentralization measure. It then adjusts it for a number of
institutional factors that are important when assessing the actual
authority and independence of decision making by local authorities.
These factors include (1) the security of the existence of local
11
Voigt and Blume (2012) note the shortcomings with capturing federalism via a
dichotomous variable. Birch (1949) provides an early view of federalism. Also, see
Treisman (2006).
174
Economic Modelling 62 (2017) 171–183
governments, (2) local expenditure, tax, and borrowing autonomy,
(3) home rule for self-governance, and (4) local government control
regarding, hiring, firing, and other human resource policies of their
own employees. The value of the index ranges from a high of 20.71
for Sweden and a minimum of 0 for Mozambique, with higher values
implying a larger local sector and greater local authority in decision
making.
• Administrative Decentralization Index (AdminDecent): This index is
one of the components of the Aggregate Decentralization Index that
focuses on local government control regarding hiring, firing, and
other human resource policies of local employees. The construction
of the index considers the local government share of total public-
sector employment and an indicator of local government discretion
related to employment and setting the terms of that employment
(Ivanyna and Shah (2014)).
Relatively speaking, this index captures contract enforcement free-
dom at the local level, which could have interesting outcomes
relative to government performance. Thus, in contrast to, say, fiscal
decentralization, this index considers employment and administra-
tive aspects, but not necessarily freedom regarding related
budget allocations. In our data set, the value of AdminDecent varies
from 0.9 (Sweden) to 0 (Mali, Mozambique, and Suriname), with
larger values implying greater decentralization. The correlation
between this index and the broader aggregate decentralization is
0.67.
The above discussion explains the distinctions across decentraliza-
tion types and government performance measures that will be con-
sidered in this analysis. We now turn to a discussion of the model setup
and estimation strategy to judge the effects of decentralization on
government performance.
4. Model, data and estimation
4.1. Model and data
Based on the above discussion and to focus on the objectives of
this paper, the estimated relation takes government performance to
be a function of decentralization; economic prosperity; a vector (Z)
of institutional and social factors and country size; and indicators
of industry and firm size associated with the survey respondent.
Many of these controls are routinely employed in the related
literature (e.g., Kyriacou and Roca-Sagalés (2011a, 2011b),
Lambsdorff (2006), Treisman (2007)). The main interest, however,
is in the decentralization-performance nexus across the various
dimensions of each. Thus, an equation of the following general
form is estimated (with subscript i denoting a country and j
denoting a survey response):
Perceived government performancei, j, g = f (Government decentralizationi, k ,
Economic Prosperity(GDP),
i Z i, w , Industry typei, j , Firm Sizei, j ) (1)
where
g = TaxAdmin, BusLicense, Corrupt
k = Fiscal Decent, Federalism, AggDecent, AdminDecent
w = GovtSize, RuleLaw, POP, Protestant
The dependent variable is one of the three areas of perceived
government performance drawn from the WES survey (i.e., tax
administration (TaxAdmin), business licensing (BusLicense) and cor-
ruption (Corrupt)). Firms' timing and frequency of interactions with
different government agencies vary and this shapes their perceptions of
government performance. For instance, taxes are ex-post entry but
periodic or repetitive, whereas licenses and corruption could be
sporadic. Thus, our consideration of three performance dimensions
R.K. Goel et al. Economic Modelling 62 (2017) 171–183

Table 1
Variable definitions, summary statistics and sources.

Variable Definition; Mean, SD Source

TaxAdmin An ordered response to the question asking how far tax administration issues affect business World Bank World Enterprise Surveys. World Bank. (www.
operations. The response ranges from 0 (no obstacle) to 3 (very severe obstacle) – see Table A1; enterprisesurveys.org)
1.56, 0.74.
BusLicense An ordered response to the question asking how far business permits and licensing issues affect World Bank World Enterprise Surveys. World Bank. (www.
business operations. The response ranges from 0 (no obstacle) to 4 (very severe obstacle) – see enterprisesurveys.org)
Table A1; 1.47, 0.73.
Corrupt An ordered response to the question asking how far government corruption affects business World Bank World Enterprise Surveys. World Bank. (www.
operations. The response ranges from 0 (no obstacle) to 4 (very severe obstacle) – see Table A1; enterprisesurveys.org)
1.90, 0.91.
Fiscal Decent Fiscal decentralization. Local government expenditures as percentage of general government Ivanyna and Shah (2014)
expenditures. Higher values indicate greater fiscal decentralization; 0.15, 0.13
Federalism Binary variable indicator of a country classified as federalist; (the coding is for unitary (0) and Treisman (2006)
federal (1) states); 0.15, 0.36.
AggDecent Aggregate decentralization. Index incorporates the relative importance of local government Ivanyna and Shah (2014)
(measured by expenditure decentralization), the security of existence of local government, and
fiscal, political decentralization and administrative decentralization indexes; Range: 0–34, with
higher values implying greater aggregate decentralization; 1.53, 2.32.
AdminDecent Administrative decentralization. Index measures the ability of local governments to hire and fire Ivanyna and Shah (2014)
and set terms of employment of local employees as well as regulatory control over own
functions; Range: 0–1, with higher values implying greater administrative decentralization;
0.35, 0.23.
RuleLaw Rule of law index. It captures perceptions of the extent to which agents have confidence in and Kaufmann and Kraay (2014) http://info.worldbank.org/
abide by the rules of society, and in particular the quality of contract enforcement, property governance/wgi/index.aspx#home
rights, the police, and the courts, as well as the likelihood of crime and violence; Range: −2.5 to
+2.5, with higher values denoting stronger institutions; -0.41, 0.67.
Protestant Dummy variable that assumes a value of 1 if the percentage of population professing Protestant La Porta et al. (1999)
faith exceeds 50%, otherwise zero; 0.03, 0.16
GDP Log of GDP per capita in constant 2005 US dollars; 7.45, 1.14. World Development Indicators (http://data.worldbank.org/
indicator/NY.GDP.PCAP.KD?page=2)
GovtSize General government final consumption expenditure (% of GDP); 13.18, 4.87. World Development Indicators http://data.worldbank.org/
indicator/NE.CON.GOVT.ZS
POP Log of country's population; 17.24, 1.94. World Development Indicators (http://data.worldbank.org/
indicator/SP.POP.TOTL)
ServiceIND A categorical variable representing firms belonging to services sector. It includes sectors like World Bank World Enterprise Surveys. World Bank. (www.
construction, hotels and restaurants, IT and IT services, retail, tourism, transport, wholesale enterprisesurveys.org)
etc.; 0.42, 0.49.
LargeFirm A categorical variable representing firms with 100 or more employees; 0.19, 0.39. World Bank World Enterprise Surveys. World Bank. (www.
enterprisesurveys.org)

enables us to address a spectrum of qualitatively different issues.12
As explained above, for each area of government performance, firm
responses in the original WES survey are reclassified into one of three
categories depending upon whether the respondent saw that area of
government activity as a minor, moderate, or major impediment to
their business operations. We report results for “major” and “minor”
responses, while keeping “moderate” as the default (see Table A1 in the
Appendix for the classification used).
The main explanatory variable is one of the four forms of govern-
ment decentralization discussed in the preceding section. However, we
also control for economic prosperity, industry type (by including a
dummy variable for service industry (ServiceIND)),13 and for other
country attributes (denoted by Z in Eq. (1)), including government size
(GovtSize), country size (POP), institutional quality (RuleLaw), and
religion (Protestant – accounting for social influences). More prosper-
ous nations generally have better government machinery and perfor-
mance of government services, while service industries may perceive
government performance differently.
A more direct aspect of institutional quality is considered by
12
Given that the dependent variables in our analysis are at the micro-level, while the
right-hand-side variables are at the country level, the dummy variables identifying
service industry firms and large firms enable us to better exploit the information in our
data (compared to country-level fixed effects). An alternate way to address this might be
multilevel modelling. However, given the nature of survey responses (where each
dimension of performance is considered separately), this approach seems less appro-
priate than the estimation technique used.
13
Our cross-country sample did have information on other industries. Our choice of
controlling only for service industries was dictated by the fact that this set of industries
was qualitatively more distinct from others (e.g., manufacturing, construction, etc.).
including an index for the Rule of Law (RuleLaw), whereby nations
with better institutional quality are likely to perform better, ceteris
paribus (see La Porta et al. (1999)). The inclusion of government size
and country size account for varying constraints and opportunities
nations and governments of different sizes face in providing services.
For instance, larger governments might be associated with greater red
tape on the one hand and with better checks and balances on the other
hand. The former effect will increase corruption and lower government
performance, while the latter will enhance performance (see
Lambsdorff (2006)). The link between government size and corruption
has been addressed empirically in the past, with mixed results (e.g.,
Fan et al. (2009), Fisman and Gatti (2002), Gerring and Thacker
(2005), Goel and Nelson (1998), Kotera et al. (2012)).
Finally, the social dimension is accounted for by controlling for the
Protestant religion. Nations with majority of Protestants are perceived
to have a stronger work ethic. Such nations have largely been found to
be less corrupt (Lambsdorff (2006)). The literature on determinants of
corruption also uses many of the regressors that we use to examine
(aggregate) cross-national corruption (see Lambsdorff (2006) and
Treisman (2007) for surveys of the literature; and Treisman (2000)
for an earlier rather comprehensive study) and we will also investigate
here their possible effects on government performance in other areas.
Given the detail in our data, we provide additional insights by
examining whether service industry firms are different and whether
firm size matters. One could imagine situations where service firms
have different interactions with corrupt bureaucrats (compared to
manufacturing firms) and technologies such as the internet might
enable them to bypass such interactions (think about IT service firms

175
R.K. Goel et al.
able to transmit software over the internet, bypassing governmental
interference - this aspect also relates to bypassing licensing require-
ments). Further, many service industries, such as banking and IT,
might not require pollution permits unlike their manufacturing coun-
terparts, but might offer numerous other services that might require
multiple types of regulations (e.g., the financial services firms). Forty
two percent of firms in our sample belonged to the service sector
(Table 1).
We also consider the role of firm size in perceptions of government
performance (Section 5.5). Broadly speaking, firm size can be seen in
the context of the discussion surrounding the structure of markets -
whether large or dominant firms have different government interac-
tions that affect their views of government performance. Large firms
might have greater interactions with the government because they
might require extra licenses and permits based on size and carry more
political weight within the local community. Further, they might appear
more easily on the radars of government inspectors and thus face
greater scrutiny and rent demands. Conversely, large firms might be
more adept at dealing with the government because of their ability to
hire specialists, although internal bureaucracies might make them
somewhat lethargic (the “X-inefficiency” view).
The data for this study are drawn from several sources in the public
domain and include firm level survey data and country level aggregate
data. There were some challenges with the survey data. The large
survey sample across countries was not completely uniform and we had
to make some adjustments for the analysis. First, the time dimension of
the variables is not entirely uniform. For dependent variables (survey
responses) the time varies from 2006 to 2015, depending upon the
country under consideration. In contrast, for decentralization mea-
sures, taken from Ivanyna and Shah (2014), the time dimension
pertains mostly to 2005 and the values remain unchanged over time.
The control variables are also for 2006 and remain invariant over time.
Importantly, the choice of the period from 2006 to 2015 for the
dependent variables and 2005 or 2006 for the right hand side variables
helps to mitigate the problem of endogeneity in our cross-sectional
analysis.14,15
14
Further, since the dependent variable(s) is based on individual survey responses, it
seems unlikely that reverse feedbacks affect government decentralization, especially in a
short time period.
15
Unlike World Bank's Doing Business Surveys, the Enterprise Surveys (WES) do not
rely on expert opinion. The data in later surveys is based on firm manager's responses to
an extensive questionnaire asking firms about issues related to a broad range of local
business environment. (The topics included in enterprise surveys are: infrastructure,
trade, finance, regulations, taxes and business, licensing, corruption, crime and
informality, finance, innovation, labor, and perceptions about obstacles to doing
business). The initial rounds of WES gathered data under different programs e.g.
Regional Program on Enterprise Development in Africa, Business Environment and
Enterprise Performance in Eastern Europe and Central Asia, and Investment Climate
Surveys. All these were conducted prior to 2005. However, from 2006 onward, the WES
have been standardized and the responses are based on same questionnaire and
methodology. The post-2005 data are available, in a single file containing data of all
the firms surveyed in 135 countries, at www.enterprisesurveys.org to registered users
(and registration is free).
Admittedly, the Enterprise Surveys differ in terms of questions they ask in different
countries. However, since 2005 the data are being offered in two formats (for details see
https://www.enterprisesurveys.org/portal/login.aspx):
a. “Standardized: country data are matched to a standard set of questions. This format
allows cross-country comparisons and analysis but sacrifices those country-specific
survey questions which cannot be matched.
b. Country survey: offers the complete survey information for a particular country's
survey. In this format the question coding is inconsistent across country surveys, even
where the questions are identical, and is useful for single country or limited cross-country
analysis.”
Our paper uses data that falls in format (a) above. It is clear from the above-mentioned
explanation that data on standardized survey instruments are available and that this data
set is provided in this standardized format by the WES and we never merge multiple
surveys to form a single cross country data set. The link “full survey data” is https://www.
enterprisesurveys.org/portal/index.aspx#/library?dataset=Combined. The combined
data set is available for download if one creates free account and log in on the link
above. Anyone interested in replication of our findings can access the data following the
176
Economic Modelling 62 (2017) 171–183
Although the methodology used to cover the universe of the survey
is consistent, the sample of firms surveyed, and the sample of countries
in any year, is not the same over the years. For instance, in the year
2006 the total number of firms surveyed was 14,903 whereas in 2009
the number was 18,784. Similar differences exist for other years. It
implies that we cannot call our data set ‘a single panel of firms over
time’. In other words, the data structure is not suitable for panel data
methods. As explained below, by controlling for the various macro-
economic and other institutional heterogeneities we can manage to
extract reasonable inferences.
The other control variables used in the analysis are from reputed
international sources that are routinely used in the literature. Details
about the variables used along with summary statistics are provided in
Table 1, and Appendix Table A2 provides a list of sample countries.
4.2. Estimation
Given that the firms’ responses follow logical ordering (see Section
3), the ordered logit model is a natural choice for estimation (see
Maddala (1986)). Thus, firm i's response m can assume one of the three
permissible values such as m=1, 2, 3. The responses are ordered such
that higher m values indicate “bad outcome” in terms of business-
government relations. For instance, a firm with m=1 is facing either no
or only minor obstacles in business operations caused by the govern-
ment's rules, regulations, and their implementation, but a firm choos-
ing alternative 3 is facing major or very severe obstacles in firm-
government relations. For a more focused comparison, we report
results for only these two extreme outcomes and do so in the form of
marginal effects for each predictor. The marginal effects tell us the
impact of each of the explanatory variable on the probability that a firm
chooses alternative m under the assumption that the other variables in
the model are at their sample mean values.
The reported standard errors are computed using the Delta method.
The Delta method expands a function of a random variable about its
mean, usually with a one-step Taylor approximation, and then takes
the variance. In logit and probit estimation and the family of models
that belong to maximum likelihood estimators, the Delta method is
used to compute asymptotic variance which makes inference more
reliable. The Delta method allows us to obtain the appropriate standard
errors of any smooth function of the fitted model parameters.16
5. Results
In this section the estimation results are reported for the various
decentralization measures. Of primary interest is the effect that each
decentralization measure has on survey responses regarding how each
dimension of government affects business operations. Three areas are
addressed: tax administration, business licensing and permits, and
corruption. Each of these can be thought of as a relevant area to assess
the quality of government operations from a business person's per-
spective. Tables 2–5 report the main results. In each table the left-hand
columns display marginal effect estimates of how increases in the
decentralization measure or one of the other control variables will
affect the probability that the survey respondent will see the factor
listed in the column heading as either a “minor” or “no obstacle” to
(footnote continued)
links/procedures we have outlined in the text. In addition, the details regarding sources
of data for individual countries in our data can also be seen in the “catalog of all data”, an
excel sheet that is available also at the given link.
16
Technically, it involves applying a Jacobian matrix to the estimated variance matrix
of the fitted model parameters. The Jacobian is a matrix of partial derivatives of the
statistics of interest with respect to each of the fitted model parameters. The STATA's
routine ‘margins’ generates Delta-method standard errors by default for each marginal
effect (linear or nonlinear). Additional details are available at: http://www.statalist.org/
forums/forum/general-stata-discussion/general/98078-delta-method-standard-errors-
for-average-marginal-effects-using-margins-command.
R.K. Goel et al. Economic Modelling 62 (2017) 171–183

Table 2
Effects of fiscal decentralization on government performance.

Dep. variable → (2.1a) (2.1b) (2.2a) (2.2b) (2.3a) (2.3b)

TaxAdmin TaxAdmin BusLicense BusLicense Corrupt Corrupt

Fiscal Decent 0.459*** −0.239*** 0.440*** −0.235*** 0.865*** −0.784***

(0.014) (0.007) (0.013) (0.007) (0.014) (0.013)

RuleLaw 0.095*** −0.050*** 0.105*** −0.056*** 0.176*** −0.159***

(0.003) (0.002) (0.003) (0.002) (0.003) (0.003)

Protestant 0.186*** −0.097*** 0.172*** −0.092*** 0.184*** −0.167***

(0.014) (0.007) (0.013) (0.007) (0.013) (0.012)

GDP −0.043*** 0.022*** −0.039*** 0.021*** -0.049*** 0.044***

(0.002) (0.001) (0.002) (0.001) (0.002) (0.002)

GovtSize −0.003*** 0.002*** −0.005*** 0.003*** 0.003*** −0.003***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

POP −0.011*** 0.006*** −0.013*** 0.007*** −0.019*** 0.017***

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

ServiceIND 0.022*** −0.011*** 0.012*** −0.006*** 0.029*** −0.027***

(0.004) (0.002) (0.003) (0.002) (0.004) (0.003)

Observations 77,360 77,360 82,448 82,448 82,443 82,443

Notes: See Table 1 for variable details. Delta-method standard errors in parentheses; ***p < 0.01, **p < 0.05, *p < 0.1.
Reported coefficients are marginal effects, based on ordered logit regressions (the pseudo-R2s for the underlying regressions are available upon request). All predictors at their average
value. Column header indicates the outcome factor. As explained in the text, only two categories of firms’ responses are considered: the factor in question is a minor obstacle to firms’
business operations, and factor in question is a major obstacle to firms’ business operations. Columns 2.1a, 2.2a and 2.3a reflect the response that (decentralized) government structure
is a minor or no obstacle, while 2.1b, 2.2b and 2.3b reflect the perception that the government is a major or severe obstacle to business operations.

Table 3
Effects of federalism on government performance.

Dep. variable → (3.1a) (3.1b) (3.2a) (3.2b) (3.3a) (3.3b)

TaxAdmin TaxAdmin BusLicense BusLicense Corrupt Corrupt

Federalism −0.144*** 0.076*** −0.116*** 0.062*** −0.098*** 0.090***

(0.006) (0.003) (0.005) (0.003) (0.005) (0.005)

RuleLaw 0.006* −0.003* 0.029*** −0.015*** 0.060*** −0.055***

(0.004) (0.002) (0.003) (0.002) (0.003) (0.003)

Protestant 0.148*** −0.078*** 0.175*** −0.094*** 0.142*** −0.130***

(0.013) (0.007) (0.012) (0.006) (0.011) (0.010)

GDP −0.003 0.002 −0.004** 0.002** −0.001 0.001

(0.002) (0.001) (0.002) (0.001) (0.002) (0.002)

GovtSize −0.003*** 0.002*** −0.005*** 0.002*** 0.006*** −0.006***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

POP 0.006*** −0.003*** 0.003** −0.001** 0.004*** −0.004***

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

ServiceIND 0.034*** −0.018*** 0.024*** −0.013*** 0.051*** −0.046***

(0.004) (0.002) (0.003) (0.002) (0.003) (0.003)

Observations 78,475 78,475 83,603 83,603 83,546 83,546

Notes: See Notes to Table 2. The coding is for unitary (0) and federal (1) states.

their business operations. The right-hand columns pertain to the
likelihood of survey response of either “major” or “very severe obstacle”
impact on business operations.17
17
We also considered the "moderate" obstacle. Results of the marginal effects for the
baseline response option – “moderate” obstacle – are aligned more with "major" or
"severe" obstacles that are reported in Tables 2–5. The results for all categories are
reported in Appendix B.
5.1. Effects of fiscal decentralization (Fiscal Decent)
Table 2 reports the results using the local government's share of
total government expenditures to measure decentralization. As dis-
cussed above, this indicator has often been used in earlier empirical
work analyzing various decentralization topics, but with well-known
deficiencies (Ebel and Yilmaz (2003)). The results in the present case
show that survey respondents are less likely to see any of the three
government areas as a serious detriment to business operation (and
more likely to perceive the government being a minor or no obstacle) in

177
R.K. Goel et al. Economic Modelling 62 (2017) 171–183

Table 4
Effects of aggregate decentralization on government performance.

Dep. variable → (4.1a) (4.1b) (4.2a) (4.2b) (4.3a) (4.3b)

TaxAdmin TaxAdmin BusLicense BusLicense Corrupt Corrupt

AggDecent −0.001 0.001 0.008*** −0.004*** 0.027*** −0.025***

(0.001) (0.001) (0.001) (0.000) (0.001) (0.001)

RuleLaw 0.050*** −0.027*** 0.071*** −0.038*** 0.111*** −0.101***

(0.003) (0.002) (0.003) (0.002) (0.003) (0.003)

Protestant 0.168*** −0.089*** 0.162*** −0.087*** 0.157*** −0.143***

(0.014) (0.007) (0.013) (0.007) (0.013) (0.011)

GDP −0.027*** 0.014*** −0.032*** 0.017*** −0.042*** 0.038***

(0.002) (0.001) (0.002) (0.001) (0.002) (0.002)

GovtSize −0.001* 0.000* −0.005*** 0.002*** 0.002*** −0.002***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

POP −0.001 0.000 −0.009*** 0.005*** -0.016*** 0.014***

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

ServiceIND 0.035*** −0.018*** 0.024*** −0.013*** 0.049*** −0.045***

(0.004) (0.002) (0.003) (0.002) (0.004) (0.003)

Observations 78,208 78,208 83,326 83,326 83,306 83,306

Notes: See Notes to Table 2.

Table 5
Effects of administrative decentralization on government performance.

Dep. variable → (5.1a) (5.1b) (5.2a) (5.2b) (5.3a) (5.3b)

TaxAdmin TaxAdmin BusLicense BusLicense Corrupt Corrupt

AdminDecent 0.048*** −0.025*** 0.106*** −0.057*** 0.141*** −0.129***

(0.008) (0.004) (0.008) (0.004) (0.008) (0.008)

RuleLaw 0.051*** −0.027*** 0.065*** −0.035*** 0.092*** −0.084***

(0.003) (0.002) (0.003) (0.002) (0.003) (0.003)

Protestant 0.175*** −0.093*** 0.186*** −0.100*** 0.192*** −0.176***

(0.014) (0.007) (0.012) (0.007) (0.012) (0.011)

GDP −0.031*** 0.016*** −0.032*** 0.017*** −0.030*** 0.027***

(0.002) (0.001) (0.002) (0.001) (0.002) (0.002)

GovtSize −0.001*** 0.001*** −0.004*** 0.002*** 0.006*** −0.005***

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

POP −0.003** 0.001** −0.008*** 0.004*** −0.006*** 0.006***

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)

ServiceIND 0.034*** −0.018*** 0.024*** −0.013*** 0.051*** −0.047***

(0.004) (0.002) (0.003) (0.002) (0.004) (0.003)

Observations 78,208 78,208 83,326 83,326 83,306 83,306

Notes: See Notes to Table 2.

countries with greater expenditure decentralization. In each case, the
parameter estimates on the Fiscal Decent variable are statistically
significant at the 1 percent level. These results are generally consistent
with much of the literature that has investigated the decentralization –
corruption link (e.g., Fan et al. (2009), Nelson (2013)). Further, the
results are in line with Adam et al. (2014) who used OECD data
between 1970 and 2000 to examine the impact of fiscal decentraliza-
tion on public-sector efficiency in education and health services. Revelli
(2010), however, found that local public expenditures exceeding
central spending standards had a detrimental effect on English
government performance.
Turning briefly to the results for the control variables in the model,
the estimates are generally consistent across all three aspects of
government performance. As might be expected, the probability of
more positive assessment of government outcomes is observed for
countries with a stronger rule of law and Protestant traditions (see
Lambsdorff (2006)). In contrast, larger government size (in terms of
share of GDP) is associated with greater likelihood of a negative
assessment of the impact of government on business operations in
two of the three areas. With regard to corruption (Corrupt), however, a
larger government size in fact improves perceptions of government
performance. This latter finding is in line with the view that a larger
government machinery might not necessarily be adding to greater rent-
seeking. Further, a larger public sector may have stronger institutions
in place to deter such activity (see Goel and Nelson (2005) for a similar
finding with aggregate data; also see Kotera et al. (2012)). In countries
with higher incomes, survey respondents are more likely to have a
more negative assessment of the government outcomes in all cases. The

178
R.K. Goel et al.
effects are, however, modest. For example, with a one standard
deviation increase in per capita GDP (or economic prosperity) from
the sample mean, the probability that a business manager would agree
that business licensing matters are only minor obstacle to business
operations would decrease by only about 4%.
In more populous countries business managers tend to have a more
negative assessment of government performance. This perception
might be formed due to keener competition for resources in more
populous nations, leading to less government attention to individual
issues. Respondents from service industries are more likely to have a
positive assessment of decentralization in all areas. This finding is
consistent with the view that certain regulations or taxes - such as those
related to pollution - are not widely imposed on many service
industries, decreasing their interactions with the government and
thereby enhancing their views of government performance.
5.2. Effects of federalism (Federalism)
Another way to view decentralization is to focus on political rather
than economic decentralization by distinguishing between federalist
and unitary states. In the latter, the central government holds most of
the power and local regional governments - to the extent they exist - are
administrative units of the national government. Fifteen percent of the
countries in this data set are classified as federalist states (the coding is
for unitary (0) and federal (1) states).
Table 3 presents the results using a binary federalist state decen-
tralization indicator along with the same control variables used earlier.
Interestingly, the findings are just the opposite of what was found
relative to the fiscal decentralization measure. In particular, relative to
unitary states, federalist governments perform worse in all areas. For
example, in a federalist state the probability that a business manager
would consider tax administration to be either no obstacle or only a
minor obstacle to business operations decreases by over 14%, other
things equal, based on the estimate reported in column 3.1a. For
business licensing, it decreases by nearly 12% (column 3.2a) and by
10% in the case of corruption (column 3.3a).
These findings are in line with Goldsmith (1999), Kunicová and
Rose-Ackerman (2005), and Treisman (2000) in their analyses of the
effect of federalism on perceived corruption. However, in an analysis of
survey results of business manager's experience with bribery conducted
in the late 1990s, federal structures were not found to be important in
the level of bribery payments surrounding either business licenses or
tax collection (Fan et al. (2009)). Overall, these inconsistent results
may be attributed to the dichotomous nature of the Federalism variable
which may mask other important institutional factors.
As to the control variables, the impact of the Protestant ethic,
government size, and the service industry variables are similar to those
in Table 2. The evidence for the remaining variables in the model
including the rule of law, GDP, and population variables is mixed.18
5.3. Effects of aggregate decentralization (AggDecent)
Table 4 presents the results when the Aggregate Decentralization
Index is used as the decentralization measure. This measure adjusts the
expenditure decentralization indicator (Fiscal Decent) through the
incorporation of several other important institutional factors that are
relevant in assessing the actual autonomy of local governments. These
factors include indicators of expenditure and tax autonomy, the
security of existence of local governments, and variables that measure
political and administrative decentralization within a country.
With these adjustments a somewhat different story emerges from
18
For instance, the effect of affluence on corruption is insignificant in Table 3. The
possible complexity of the relation between corruption and economic development has
been noted in the literature (see, for example, Blackburn (2012)).
179
Economic Modelling 62 (2017) 171–183
what is reported in Table 2 as to how decentralization affects managers'
perceptions of government performance. A positive assessment still
remains regarding business licensing and corruption. The parameter
estimates are statistically significant at the 1% level in each case, but
the magnitudes of the parameter estimates are modest. For example, if
a county currently at the sample mean for AggDecent (=1.5) were to
change its institutions such that its decentralization score increased to
match Brazil (=8.1), the highest index value in the data set, other
things equal, the probability that a business manager would agree that
business licensing matters are at most only a minor obstacle would
increase by just 5%.
Regarding tax administration, the evidence indicates that the
indicator of decentralization has no statistically significant effect on
business managers' perceptions of government performance (columns
4.1a and 4.1b). As noted above, perceptions relative to the impact of
corruption (columns 4.3a and 4.3b) improve with this form of
decentralization.
For the most part, the findings for the control variables are
consistent with what is noted in Table 2. Countries with more effective
rule of law and Protestant-dominated populations tend to have better
perceptions, while nations that are more affluent and with a larger
overall public sector tend to be viewed more negatively. The evidence
for the population variable is mixed here also.
5.4. Effects of administrative decentralization (AdminDecent)
The last decentralization measure considered is more narrowly
focused than the others and is labelled as administrative decentraliza-
tion. It reflects the local government share of all public sector employ-
ment and the degree to which there is local control in hiring policies of
these employees. Greater local control increases accountability and
reduces free riding; however, there is the potential downside of the
possibility of nepotism in hiring. Higher values of AdminDecent imply
that local employees are relatively more important in delivering public
services and that there is more local control over who these employees
are.
Table 5 summarizes the results and shows that administrative
decentralization is statistically significant at conventional levels for all
areas, and greater decentralization is associated with better manager
assessment regarding how government affects their business opera-
tions. The largest impacts are in business licensing (columns 5.2a and
5.2b), activities that are most likely to lie within the purview of the local
public sector. In particular, if a country currently at the sample mean
for AdminDecent in this data set (=0.35) were to change its institutions
such that its decentralization score increased to match the highest
value in the data set (=0.82), other things equal, the probability that a
business manager would agree that business licensing matters are at
most only a minor obstacle would increase by 5%. The finding is
consistent with arguments dating back to Oates (1972) that local
officials are more accountable to the citizens they serve as they are
disciplined by inter-jurisdictional competition for mobile resources.
Moreover, authorities are subject to “yardstick evaluation” (Salmon
(1987)) as their performance can be more easily compared to officials
in similar positions in neighboring jurisdictions. The estimates relative
to corruption also indicate that decentralization improves perceptions
of government performance. This result is consistent with the notion
that greater accountability associated with local control of human
resource policies acts as a deterrent, improving performance. The
findings for the control variables generally mirror earlier results.
5.5. Additional considerations
5.5.1. Effects of firm size
To obtain additional insights and to exploit better the information
available in the survey, we included a dummy variable, (LargeFirm),
identifying respondents from large firms.
R.K. Goel et al.
Table 6
Marginal effects of increase in decentralization on government performance.
Decentralization Dimensions of government performance
measure ↓
TaxAdmin
Fiscal Decent 6.00%
Federalisma −5.20%
b
AggDecent
AdminDecent 1.10%
Notes: See Table 1 for variable details. The marginal effects show the effect of one standard deviation increase in the decentralization measure from the sample mean on the probability
that respondent will see the government activity as no more than a minor obstacle to business operations. The reported effects are significant at 5% level or lower, except as noted.
a
The Federalist government structure variable is included here for consistency with the earlier analysis. Marginal effect results must be viewed with due caution as this is a binary
decentralization measure making the interpretation of a standard deviation change somewhat problematic.
b
Hypothesis that decentralization has no effect cannot be rejected at conventional levels of statistical significance.
The size of firms relates to firm vintage. Large firms are likely to be
old and established (unless they grow big by mergers and acquisitions)
and thus may have a more involved view of government performance
than new firms. In this analysis, large firms are classified as firms with
more than 100 employees and they formed nineteen percent of our
sample (see Table 1).
Re-estimating Tables 2–5 with LargeFirm as an additional regres-
sor provided some interesting results. Unreported estimates indicate
that, relative to other respondents, large firms viewed tax administra-
tion to be a severe impediment to their business operations across the
four dimensions of government decentralization considered.19 This
result is consistent with the view that large firms, either due to their
size or longevity (or both), are likely to have rather involved interac-
tions with tax authorities and they view such interactions unfavorably
(as opposed to small or new firms who might not have enough earnings
to be caught in the tax net). The findings for the decentralization
variables are unchanged by the inclusion of LargeFirm in the model.
5.5.2. Different set of macro controls
As further robustness checks of our findings, we considered three
alternate set of regressors to Tables 2–5 by adding the following: (a) a
dummy variable identifying the OECD countries in our sample; (b) the
degree of trade openness, and (c) adding indices of overall economic
freedom and democracy.20 The OECD dummy enables us to consider
institutional aspects that are not otherwise accounted for in the set of
controls used. Further, democracy and economic freedom (measured
alternately via an index and via trade openness) can affect perceptions
of government performance.
For the OECD dummy and trade openness, the parameter estimates
for the four decentralization measures are similar to Tables 2–5, with
Federalism again having a different impact on government perfor-
mance than the other decentralization types. The OECD dummy
coefficients indicate that OECD nations are less likely to view govern-
ment performance unfavorably across the three dimensions
(TaxAdmin, BusLicense, and Corrupt). This is consistent with the view
that governments in these wealthier nations are run more efficiently.
On the other hand, unfavorable view of government performance is
more likely in more democratic and more economically free nations.21
For instance, survey respondents might be better informed about
government performance in more democratic nations due to freedom
of press, and greater overall economic freedom might shape expecta-
19
The results with regard to corruption were statistically insignificant and showed no
clear pattern with regard to business licensing. Details are available upon request.
20
Specifically, the economic freedom measure is from The Heritage Foundation
(www.heritage.org), trade openness is from the World Development Indicators and the
democracy measure is from Polity IV (www.systemicpeace.org/inscrdata.html).
21
The results mentioned in this section are not reported but are available upon
request.
180
Economic Modelling 62 (2017) 171–183
BusLicense Corrupt
5.70% 11.7%
−4.20% −3.53%
1.80% 6.30%
2.40% 3.22%
tions about specific dimensions of government performance. However,
when openness to trade is used to account for economic freedom, the
findings are opposite of those for the economic freedom index. These
robustness checks support our main findings about the role of
decentralization in perceived government performance.
5.6. Relative magnitudes of impacts
Table 6 presents side-by-side comparisons of the impact of greater
decentralization using the estimates presented in Tables 2–5. For each
decentralization measure the entries in the table show how greater
decentralization affects the probability that business managers will
have a favorable assessment (minor or no obstacle to business
operations) of the government activity in question. In each case greater
decentralization is defined here as a one standard deviation increase in
the decentralization measure above the respective sample mean.
The first takeaway from this comparison is that how decentraliza-
tion is measured is an important consideration when assessing its
influence on government performance. Measured either by local
government expenditure share or a federalist binary variable, relatively
sizable effects from greater decentralization on assessment of govern-
ment performance are observed, but in opposite directions, as can be
seen by comparing the first two rows of Table 6. Of course, neither
measure addresses the institutional details as to how much actual
authority rests with lower tiers of government and this may affect the
results. Our findings suggest that this is an important consideration. In
particular, the aggregate decentralization measure attempts to account
for these details and when this is done (row three of the table), the
evidence reveals that the impact of greater decentralization has a more
modest positive impact on assessment of government performance in
two of the three areas, and no statistically significant effect in tax
administration.22
The second implication of this analysis is that greater adminis-
trative decentralization appears to hold out the most hope to improve
assessment of government performance with regard to tax administra-
tion and business licensing as shown by the results summarized in the
last row of the table. There are improved perceptions in the business
licensing area (2.4%) and also in the case of corruption (3.2%).
Viewed from a somewhat different perspective, 30 countries in the
data set have a zero value for the administrative decentralization index.
If these countries were to adopt policies to take them to the sample
mean in the data set, the improved perception in the business licensing
22
The relative magnitudes of impacts should be interpreted with some caution as
AggDecent and AdminDecent are based on indices, Federalism is a dichotomous
variable, while Fiscal Decent is a continuous variable, based on "hard" data. Indices,
for instance, are based on some aggregation across different aspects and are bounded. So
their impacts are understandably different from those of (unbounded) continuous
variables.
R.K. Goel et al.
area would be nearly 5%. These results should not be surprising as
licensing is typically under the purview of government officials at the
local level. In contrast, corruption reduction is better achieved with
aggregate decentralization. This makes sense in light of varied nature of
interactions between bureaucrats and firms (i.e., potential bribe takers
and bribe givers).
The main takeaway from the above analysis is that the qualitative
differences across decentralization types affect the nature of percep-
tions about government performance. The concluding section follows.
6. Conclusions
The degree and scope of decentralization of government services
continue to intrigue policymakers in their quest to conserve resources
and improve the delivery of government services. The elusive goal of an
“optimal” level of decentralization has kept researchers engaged with
an examination of causes and effects of government decentralization
with varying methodologies and data (see Feld and Schnellenbach
(2011), Martinez-Vazquez et al. (2016) and Yeung (2009)). This paper
adds to the literature by using data on firm's perceptions of government
performance and examining the effect of various forms of decentraliza-
tion on three major dimensions of performance.
The analysis presented in this paper can be distinguished from the
literature on the effects of government decentralization on quality of
governance in several important ways. First, our focus has been
business managers and their perceptions about how various govern-
ment measures in regard to tax administration, business licensing, and
corruption control affect their operations. Second, we have employed a
large micro data set spanning a wide cross-section of countries, thereby
mitigating the potential for two-way link between government perfor-
mance and the extent to which government is close to the citizens.
Third, two of the decentralization measures employed in this analysis
are designed to account for key institutional factors that may affect how
much control local governments actually have relative to the central
government in delivering services.
Tying to the questions posed in the introduction we have found that
how decentralization is measured is important when assessing its
impact on government performance relative to business operations. In
particular, greater decentralization, comprehensively measured to
control for institutional differences across countries, has only a modest
181
Economic Modelling 62 (2017) 171–183
positive effect on business manager assessment of government perfor-
mance. These results hold across different robustness checks (Section
5.5.2). In the context of the related literature, consideration of firm
level perception of government performance across numerous coun-
tries and their nexus with different decentralization measures are key
contributions of this work. The main takeaway from the above analysis
is that the qualitative differences across decentralization types affect
the nature of perceptions about government performance, and a
particular decentralization type may not have the same effect on
different dimensions of performance.
With regard to the control variables, the estimates are generally
consistent across aspects of government performance. Consistent with
intuition, a more positive assessment of government performance is
observed for countries with stronger rule of law and Protestant
traditions (see Lambsdorff (2006)). In contrast, larger government
size is associated with greater likelihood of a negative assessment in
two of the three areas. In wealthier countries, survey respondents seem
to have a more negative assessment of the government performance,
possibly due to a combination of being better informed and having
greater expectations from the government.23 Finally, two dimensions of
size are considered: country size and firm size. Both show negative
perceptions of government performance, especially with regard to tax
administration. This suggests that large firms and more populous
nations might face special challenges in this area, and this is a new
insight that deserves additional consideration in the future.
The policy lessons from this analysis are clear. Decentralization
should not be oversold as a policy prescription to improve government
performance and the details behind any policy initiative to “decentra-
lize” matter. From a business manager perspective, both fiscal and
administrative decentralization can be expected to lower the adverse
impact of corruption on their operations. Beyond that, the strongest
evidence in favor of a positive linkage is in the area of administrative
decentralization relative to business licensing and permits. Delivering
government services and administering such activities with local
employees, subject to employment and human resource policies that
are locally controlled, leads to a moderate, but not inconsequential
improvement in business perception of how well government func-
tions. A large number of developing countries seems to have plenty of
room to make policy changes in this area.
23
However, as Section 5.5.2 notes, the more homogeneous group of OECD nations is
somewhat different.
R.K. Goel et al. Economic Modelling 62 (2017) 171–183

Appendix A

See Tables A1 and A2.

Table A1
Reclassification of firms’ responses in the original WES survey.

WES No obstacle Minor Moderate Major Very

Scheme obstacle obstacle obstacle severe
obstacle
0 1 2 3 4

Modified Minor obstacle Moderate Major obstacle

scheme obstacle
used

Notes: In the case of TaxAdmin the responses ranged from 0–3, so we reclassified them in this case as: 0, 1: Minor obstacle; 2: Moderate obstacle; 3: Major obstacle.

Table A2
List of sample countries.

Albania, Angola, Argentina, Armenia, Azerbaijan, Bahamas, Bahrain, Bangladesh, Belarus, Belize, Benin, Bhutan,
Bolivia, Bosnia and Herzegovina, Botswana, Brazil, Bulgaria, Burkina Faso, Burundi, Cambodia, Cameroon,
Central African Republic, Chad, Chile, China, Colombia, Congo Democratic Republic, Congo Republic, Costa
Rica, Cote d'Ivoire, Croatia, Czech Republic, Dominican Republic, Ecuador, Egypt, El Salvador, Equatorial
Guinea, Eritrea, Estonia, Ethiopia, Fiji, Gabon, Georgia, Ghana, Guinea, Honduras, Hungary, India, Indonesia,
Israel, Jamaica, Jordan, Kazakhstan, Kenya, Kyrgyzstan, Laos PDR, Latvia, Lebanon, Lesotho, Liberia,
Lithuania, Macedonia, Madagascar, Malawi, Malaysia, Mali, Mauritania, Mauritius, Mexico, Moldova,
Mongolia, Morocco, Mozambique, Namibia, Nepal, Nicaragua, Niger, Nigeria, Oman, Pakistan, Panama,
Paraguay, Peru, Philippines, Poland, Qatar, Romania, Russia, Rwanda, Senegal, Serbia, Sierra Leone, Slovakia,
Slovenia, South Africa, Sri Lanka, Suriname, Sweden, Tanzania, Togo, Tonga, Trinidad & Tobago, Tunisia,
Turkey, Uganda, Ukraine, Uruguay, Uzbekistan, Vanuatu, Venezuela, Vietnam, Yemen, Zimbabwe.

Appendix B. Supplementary material

Supplementary data associated with this article, “Different forms of decentralization and their impact on government performance: Micro-level
evidence from 113 countries” can be found in the online version at http://dx.doi.org/10.1016/j.econmod.2016.12.010.

Organ. 59 (1), 233–254.

References
Adam, A., Delis, M.D., Kammas, P., 2014. Fiscal decentralization and public sector
efficiency: evidence from OECD countries. Econ. Gov. 15 (1), 17–49.
Besley, T.J., Coate, S., 2003. Centralized versus decentralized provision of local public
goods: a political economy approach. J. Public Econ. 87 (12), 2611–2637.
Birch, A.H., 1949. Federalism and finance. Manch. Sch. 17 (2), 163–185.
Blackburn, K., 2012. Corruption and development: explaining the evidence. Manch. Sch.
80 (4), 401–428.
Blume, L., Voigt, S., 2011. Federalism and decentralization: a critical survey of frequently
used indicators. Const. Polit. Econ. 22 (3), 238–264.
Bodman, P., Campbell, H., Le, T., 2012. Public investment, taxation, and long-run output
in economies with multi-level governments. Econ. Model. 29 (5), 1603–1611.
Brueckner, J.K., 2003. Strategic interaction among governments: an overview of
empirical studies. Int. Reg. Sci. Rev. 26 (2), 175–188.
Ebel, R.D., Yilmaz, S., 2003. On the measurement and impact of fiscal decentralization.
In: Martinez-Vazquez, J., Alm, J. (Eds.), Public Finance in Developing and
Transitional Countries: Essays in Honor of Richard Bird. Edward Elgar, Cheltenham,
UK, 101–126.
Elazar, D.J., 1995. From statism to federalism: a paradigm shift. Publius 25 (2), 5–18.
Enterprises Surveys Indicators Description (ESID), 2015. Enterprise Surveys Data.
World Bank. http://www.enterprisesurveys.org/data.
Faguet, J.-P., 2014. Decentralization and governance. World Dev. 53, 2–13.
Fan, C.S., Lin, C., Treisman, D., 2009. Political decentralization and corruption: evidence
from around the world. J. Public Econ. 93 (1–2), 14–34.
Feld, L.P., Schnellenbach, J., 2011. Fiscal federalism and long-run macroeconomic
performance: a survey of recent research. Environ. Plan. C: Gov. Policy 29 (2),
224–243.
Fisman, R., Gatti, R., 2002. Decentralization and corruption: evidence across countries.
J. Public Econ. 83 (3), 325–345.
Gerring, J., Thacker, S.C., 2005. Do neoliberal policies deter political corruption? Int.
Goel, R.K., Nelson, M.A., 1998. Corruption and government size: a disaggregated
analysis. Public Choice 97 (1), 107–120.
Goel, R.K., Nelson, M.A., 2005. Economic freedom versus political freedom: cross-
country influences on corruption. Aust. Econ. Pap. 44 (2), 121–133.
Goel, R.K., Nelson, M.A., 2011. Government fragmentation versus fiscal decentralization
and corruption. Public Choice 148 (3), 471–490.
Goel, R.K., Saunoris, J.W., 2016. Virtual versus physical government decentralization:
effects on corruption and the shadow economy. Public Budg. Financ. 36 (4), 68–93.
Goldsmith, A.A., 1999. Slapping the grasping hand: correlates of political corruption in
emerging markets. Am. J. Econ. Sociol. 58 (4), 865–883.
Guccio, C., Pignataro, G., Rizzo, I., 2014. Do local governments do it better? Analysis of
time performance in the execution of public works. Eur. J. Polit. Econ. 34, 237–252.
Hallward-Driemeier, M., Pritchett, L., 2015. How business is done in the developing
world: deals versus rules. J. Econ. Perspect. 29 (3), 121–140.
International Monetary Fund, 2009. Macro Policy Lessons for a Sound Design of Fiscal
Decentralization. International Monetary Fund, Washington, DC.
Ivanyna, M., Shah, A., 2014. How close is your government to its people? Worldwide
indicators on localization and decentralization. Econ.: Open-Access Open-Access. E-
J. 6, 1–61. http://dx.doi.org/10.5018/economics-ejournal.ja.2014-3.
Kaufmann, D., Kraay, A., 2014. The Worldwide Governance Indicators (WGI) Project.
World Bank. 〈http://info.worldbank.org/governance/wgi/index.aspx#home〉.
Knack, S., Keefer, P., 1995. Institutions and economic performance: cross-country tests
using alternative institutional measures. Econ. Polit. 7 (3), 207–227.
Kotera, G., Okada, K., Samreth, S., 2012. Government size, democracy, and corruption:
an empirical investigation. Econ. Model. 29 (6), 2340–2348.
Kunicová, J., Rose-Ackerman, S., 2005. Electoral rules and constitutional structures as
constraints on corruption. Br. J. Polit. Sci. 35 (4), 573–606.
Kyriacou, A.P., Roca-Sagalés, O., 2011a. Fiscal decentralization and government quality
in the OECD. Econ. Lett. 111 (3), 191–193.
Kyriacou, A.P., Roca-Sagalés, O., 2011b. Fiscal and political decentralization and
government quality. Environ. Plan. C: Gov. Policy 29 (2), 204–223.
La Porta, R., Lopez-de-Silanes, F., Shleifer, A., Vishny, R., 1999. The quality of
government. J. Law Econ. Organ. 15 (1), 222–279.
Lago-Peñas, I., Lago-Peñas, S., Martinez-Vazquez, J., 2011. The political and economic

182
R.K. Goel et al.
consequences of decentralization. Environ. Plan. C: Gov. Policy 29 (2), 197–203.
Lambsdorff, J.G., 2006. Causes and consequences of corruption: what do we know from a
cross section of countries. In: Rose-Ackerman, S. (Ed.), International Handbook on
the Economics of Corruption1st ed.. Edward Elgar Publishing, Cheltenham, 3–51.
Maddala, G.S., 1986. Limited-Dependent and Qualitative Variables in Econometrics.
Cambridge University Press, Cambridge.
Martinez-Vazquez, J., Lago-Peñas, S., Sacchi, A., 2016. The impact of fiscal
decentralization: a survey. J. Econ. Surv.. http://dx.doi.org/10.1111/joes.12182.
Nelson, M.A., 2013. Corruption and the size of local governments: are they related? In:
Lago-Peñas, S., Martinez-Vazquez, J. (Eds.), The Challenge of Local Government
Size: Theoretical Perspectives, International Experience, and Policy Reform. Edward
Elgar, Northampton, MA, 87–124.
Oates, W.E., 1972. Fiscal Federalism. Harcourt Brace Jovanovich, New York.
Panizza, U., 1999. On the determinants of fiscal centralization: theory and evidence. J.
Public Econ. 74 (1), 97–139.
Prud’homme, R., 1995. The dangers of decentralization. World Bank Res. Obs. 10 (2),
201–220.
Revelli, F., 2010. Spend more, get more? An inquiry into English local government
performance. Oxf. Econ. Pap. 62 (1), 185–207.
Rodden, J., 2004. Comparative federalism and decentralization: on meaning and
measurement. Comp. Polit. 36 (4), 481–500.
Rodríguez-Pose, A., Gill, N., 2003. The global trend towards devolution and its
implications. Environ. Plan. C: Gov. Policy 21 (3), 333–351.
Rodríguez-Pose, A., Ezcurra, R., 2011. Is fiscal decentralization harmful for economic
growth? Evidence from the OECD countries. J. Econ. Geogr. 11 (4), 619–643.
183
View publication stats
Economic Modelling 62 (2017) 171–183
Salmon, P., 1987. Decentralisation as an incentive scheme. Oxf. Rev. Econ. Policy 3 (2),
24–43.
Shah, A., Thompson, T., 2004. Implementing Decentralized Local Governance: A
Treacherous Road with Potholes, Detours and Road Closures. World Bank Policy
Research Working Paper # 3353, June.
Stegarescu, D., 2005. Public sector decentralization: measurement concepts and recent
international trends. Fisc. Stud. 26 (3), 301–333.
Transparency International, 2009. Corruption and Local Government. Working paper
#05/2009.
Treisman, D., 2000. The causes of corruption: a cross-national study. J. Public Econ. 76
(3), 399–457.
Treisman, D., 2006. Fiscal decentralization, governance, and economic performance: a
reconsideration. Econ. Polit. 18 (2), 219–235.
Treisman, D., 2007. What have we learned about the causes of corruption from ten years
of cross-national empirical research? Annu. Rev. Polit. Sci. 10, 211–244.
United Nations, 2009. International Guidelines on Decentralization and Access to Basic
Services for All. UN-HABITAT, Nairobi.
Voigt, S., Blume, L., 2012. The economic effects of federalism and decentralization: a
cross-country assessment. Public Choice 151 (1), 229–254.
World Bank, 1999. Decentralization: rethinking governmentWorld Development Report
1999/2000: Entering the 21st Century. Oxford University Press, New York,
107–124.
Yeung, R., 2009. The effects of fiscal decentralization on the size of government: a meta-
analysis. Public Budg. Financ. 29 (4), 1–23.