-
University of Groningen
Economic policy reformWiese, Rasmus Holland Thomsen
IMPORTANT NOTE: You are advised to consult the publisher's
version (publisher's PDF) if you wish to cite fromit. Please check
the document version below.
Document VersionPublisher's PDF, also known as Version of
record
Publication date:2016
Link to publication in University of Groningen/UMCG research
database
Citation for published version (APA):Wiese, R. H. T. (2016).
Economic policy reform: measurement, causes and consequences.
[Groningen]:University of Groningen, SOM research school.
CopyrightOther than for strictly personal use, it is not
permitted to download or to forward/distribute the text or part of
it without the consent of theauthor(s) and/or copyright holder(s),
unless the work is under an open content license (like Creative
Commons).
Take-down policyIf you believe that this document breaches
copyright please contact us providing details, and we will remove
access to the work immediatelyand investigate your claim.
Downloaded from the University of Groningen/UMCG research
database (Pure): http://www.rug.nl/research/portal. For technical
reasons thenumber of authors shown on this cover page is limited to
10 maximum.
Download date: 03-06-2020
https://www.rug.nl/research/portal/en/publications/economic-policy-reform(c314756a-94f5-4b52-8195-f959e449327e).htmlhttps://www.rug.nl/research/portal/en/persons/rasmus-wiese(e21ea12b-ca47-401e-b999-2948b844125c).htmlhttps://www.rug.nl/research/portal/en/publications/economic-policy-reform(c314756a-94f5-4b52-8195-f959e449327e).html
-
Economic policy reform: Measurement, causes and consequences
Rasmus Wiese
-
Publisher: University of Groningen, Groningen, The Netherlands
Printed by: Ipskamp Drukkers B.V. ISBN: 978-90-367-8689-8 /
978-90-367-8688-1 (ebook)
2016 Rasmus Wiese All rights reserved. No part of this
publication may be reproduced, stored in a retrieval system of any
nature, or transmitted in any form or by any means, electronic,
mechanical, now known or hereafter invented, including photocopying
or recording, without prior written permission of the
publisher.
-
Economic policy reform Measurement, causes and consequences
PhD thesis
to obtain the degree of PhD at the University of Groningen on
the authority of the
Rector Magnificus Prof. E. Sterken and in accordance with
the decision by the College of Deans.
This thesis will be defended in public on
Thursday 24 March 2016 at 11.00 hours
by
Rasmus Holland Thomsen Wiese
born on 19 October 1978 in Aalborg, Denmark
-
b
Supervisor
Prof. J. de Haan
Co-supervisor
Dr. R.M. Jong-A-Pin
Assessment Committee
Prof. H.W.A. Dietzenbacher
Prof. H. Pitlik
Prof. J.E. Sturm
-
Acknowledgements The last five years of my life have been very
special. My partner-in-life Hilda and I moved to
the Netherlands, we got married, and became parents to the
second love of our lives, Sophia.
On top of that, I had the privilege to study and work in a top
academic environment at the
Faculty of Economics and Business at the University of
Groningen, somewhere in between
GEM and EEF. It has been a life changing experience where I have
learned and practised
research surrounded by professional and motivating teachers and
colleagues. For that I’m
grateful and I want to express my gratitude to all who helped me
at RUG. However some
deserve special recognition.
The first two are my closest colleagues: Jakob de Haan and
Richard Jong-A-Pin, who
also were my supervisors. I call you colleagues because I always
felt that you treated my on
equal terms, meanwhile supporting that I pursued exactly the
research ideas that I found
interesting. A special thank to Jakob for always being there
when needed, ready to provide
guidance. I’m truly impressed with how you manage your
incredible workload while still being
such a motivating mentor. A special thank to Richard for
teaching me the importance of daring
to fail, for sharing your academic network with me, and for
always having time to discuss
problems and ideas over a coffee.
I also want to thank Bart Los for understanding and help during
a very tough first year
of the Research Master. I had a lot of catching-up to do in
courses like macro and micro
economics, which made it one of the most challenging periods of
my life (almost comparable to
working in a hectic Michelin starred restaurant kitchen in
London). In such conditions you form
strong bonds with those around you exposed to the same
challenges. For mutual support and
fun during this period, a particular thank to my fellow boy-band
members: Manolis, Renato and
Lexo.
I also want to express my gratitude to all the RUG people that
notably helped improve
my research: Rob Alessie, Jan Jacobs and Viola Angelini for
econometric support; Jochen
Mierau for comments on the research proposals and chapter 2; The
Gree-lab committee and the
connected people, in particular: Adriaan Soetevent, Te Bao,
Marco Haan and Reinder Dallinga
for help and financial support for the experiment; Maite Laméris
for help with recruiting
participants for the experiment and assistance in the lab;
Bernard van den Berg for valuable
health-economics discussions; and Steffen Eriksen for engaging
collaboration on chapter 5. The
members of the now extinct political economy study/research
group: Shu Yu, Dimitrios Soudis,
-
Acknowledgements
ii
ii
Le Van Ha, Yanping Zhao and Anna Samarina deserve appreciation
for interesting discussions
on related research topics.
I also want to thank two interesting and motivating roommates I
had during the PhD
years, I appreciate our collegial friendship: Anna, I appreciate
your straightforward and direct
style: and Kennan, I appreciate our discussions about
differences between China and Europe.
Moreover I want to thank my cohort PhD coffee club members:
Tadas, Brenda, Marianna, Irina
and Pim for having a forum where we could air common PhD
problems.
Lastly a large thanks to the members of the reading committee,
Erik Dietzenbacher,
Hans Pitlik and Jan-Egbert Sturm for kind comments and useful
ideas for improvements that I
have tried to accommodate the best I could.
Rasmus Wiese
-
Contents 1 Introduction
...............................................................................................................................
1
1.1 Introduction
.........................................................................................................................................
1 1.2 Identifying economic policy reforms
.......................................................................................
5 1.3 What drives successful fiscal adjustments?
...........................................................................
6 1.4 Expressive voting and political ideology
................................................................................
7 1.5 Do healthcare financing privatisations curb total healthcare
expenditures? ........... 9
2 What triggers reforms in OECD countries? Improved reform
measurement and evidence from the healthcare sector
.................................................................................
11
2.1 Introduction
......................................................................................................................................
11 2.2 Identifying reforms
.........................................................................................................................
12 2.3 Reform determinants: Political economics hypotheses
................................................. 18 2.4
Explanatory variables and empirical strategy
...................................................................
21
2.4.1 Explanatory variables
............................................................................................................................
22 2.4.2 Control variables
.....................................................................................................................................
24 2.4.3 Estimation strategy
.................................................................................................................................
26
2.5 Empirical results
.............................................................................................................................
27 2.6 Sensitivity analysis
..........................................................................................................................
29 2.7 Conclusion
.........................................................................................................................................
33
3 Are expenditure cuts the only effective way to achieve
successful fiscal adjustment?
................................................................................................................................
45
3.1 Introduction
......................................................................................................................................
45 3.2 Identifying successful fiscal adjustments
............................................................................
47 3.3 Model and data
................................................................................................................................
52
3.3.1 Model
.........................................................................................................................................................
52 3.3.2 Variables included
..................................................................................................................................
54
3.4 Estimation results
............................................................................................................................
56 3.5 Sensitivity analysis
..........................................................................................................................
64 3.6 Conclusion
.........................................................................................................................................
64
4 Expressive voting and political ideology in a laboratory
democracy ................ 67 4.1 Introduction
......................................................................................................................................
67 4.2 Decision-theoretic model
...............................................................................................................
69 4.3 Experimental design
.......................................................................................................................
75 4.4 Data
.....................................................................................................................................................
79 4.5 Empirical analysis
...........................................................................................................................
83 4.6 Conclusion
.........................................................................................................................................
92
5 Do healthcare financing privatisations curb total healthcare
expenditures? Evidence from OECD countries
............................................................................................
99
5.1. Introduction
.....................................................................................................................................
99 5.2 Identifying HCF privatisations
..................................................................................................101
-
Contents
iv
iv
5.2.1 Structural breaks
..................................................................................................................................
101 5.2.2 Healthcare privatisations
...................................................................................................................
104 5.2.3 Case study
..............................................................................................................................................
106
5.3 Propensity score matching and determinants of HCF
privatisations ..................... 107 5.3.1 Propensity score
matching
................................................................................................................
107 5.3.2 Determinants of HCF privatisations
..........................................................................................
108
5.4 Main results
....................................................................................................................................
112 5.5 Robustness analysis
......................................................................................................................
114 5.6 Conclusion
......................................................................................................................................
118
6 Conclusion
.............................................................................................................................
127 6.1 Main findings
.................................................................................................................................
127 6.2 Policy implications, limitations and future research
.................................................... 130
References
.................................................................................................................................
133
Samenvatting (Dutch summary)
.......................................................................................
143
-
Chapter 1
Introduction 1.1 Introduction The European sovereign debt crisis
and the US fiscal cliff made one thing clear: many
developed economies have been slack in controlling their public
finances. At the time of
writing, the US gross public debt level has surpassed 100% of
GDP, while, according to the
OECD, in Japan gross public debt exceeds 150% of GDP. Debt
figures are no better in Europe.
Several European Union (EU) countries do not meet the rules as
set in the Stability and Growth
Pact and have debt ratios above 60% of GDP and/or annual fiscal
deficits exceeding 3% of
GDP. These figures raise worries about the sustainability of
public finances in the
industrialized world. This has put economic reforms and fiscal
adjustments on top of the policy
agenda, especially in the EU.
In this thesis economic policy reforms are defined as a change
in economic policy that
results in a statistically significant impact on the targeted
economic variable. Economic reforms
may change ‘the rules of the game’ by changing the institutional
framework in which
optimising agents operate (North 1994). However, economic reform
may also be the result of
policy adjustments. It is not always straightforward to
determine when a change in economic
policy is a policy adjustment (i.e. a game played under constant
rules), or when it can be
characterised as an institutional change (i.e. a change of the
rules of the game). A fiscal policy
adjustment, for example, can result from a change in the
budgetary institutions (e.g. the
implementation of a balanced budget rule), but it can also
result from adjustments to economic
policy under an unchanging institutional setup. In this thesis,
economic reforms and policy
adjustments are both captured under the heading of ‘economic
policy reforms’.
Dependent on the specific research question addressed, numerous
definitions of
economic policy reforms exist. Economic reforms identified using
composite indexes, such as
the Economic Freedom Index, are generally defined as a
particular change (over a given period)
of the index that exceeds some threshold (e.g. Dreher et al.
2009, Leibrecht and Pitlik 2015).
Fiscal adjustments are generally defined as: a discretionary
(i.e. cyclically adjusted) significant
positive change in the general government’s financial balance
(e.g. Alesina and Perotti 1995,
Mierau et al. 2007, Tavarez 2005).1 As implied by varying
definitions, the specific empirical
identification method used to identify reform often relies on
‘ad hoc’ criteria lacking any notion 1 Here significant does not
refer to statistical significance.
-
Chapter 1
2
2
of statistical significance. The aim of chapter 2 is to develop
a more general methodology that
overcomes some of the measurement issues concerning economic
policy reform identification.
In addition to measurement, both the triggers and the
consequences of economic policy
reforms are the main topics of this thesis. There is a lack of
empirical knowledge about the
factors that drive economic policy reforms. And, apart from
these triggers, little is known about
the economic consequences of certain economic policy
reforms.
Conventional wisdom and theory has it that economic crises
trigger change. When
pressure on public finances mounts, countries undertake economic
policy reforms.2, 3 Economic
analysis that treats the government as a benevolent social
planner is incomplete when it comes
to understanding the occurrence of economic policy reform. The
reason is that political and
institutional factors restrain policymakers in implementing
policies that would lead to optimal
outcomes. A theoretical foundation that includes the constraints
arising from heterogeneous
political preferences and the institutional context therefore is
warranted (Drazen 2000).
In practice, welfare enhancing economic policy reforms are often
delayed. Based on
neoclassical economic theory such sub-optimal ‘behaviour’ is
hard to explain. Political
economy models, however, prove helpful explaining delays and
occurrences of economic
policy reforms. There are models where political opponents
engage in a so-called ‘war of
attrition’. That is, governments postpone unpopular and
controversial policies, even if they are
optimal from a social welfare perspective, until the ‘political
costs’ of reform have fallen below
the ‘political benefits’ of postponing it (Alesina and Drazen
1991).
Related models focus on asymmetric information between voters
and the incumbent
government with respect to the expected economic outcome of
reform. These studies, for
example, show that if the government has private information
about the expected outcome of a
reform, only political actors who are unlikely to support reform
ex ante can credibly signal to
the electorate that reforms will be beneficial ex post (see
Cukierman and Tommasi 1998). In
contrast, partisan theory suggests that right- and left-wing
parties implement policies in line
with the preferences of their electorates instead of basing
policy on a social welfare analysis
(Hibbs 1977).
A second type of models stresses the role of ex-ante uncertainty
about the economic
outcome of reform in explaining delay (Dewatripont and Roland
1995; Fernandez and Rodrik
1991). Welfare enhancing reforms can be rationally delayed or
blocked if there is ex ante
2 Even though this view is ‘conventional wisdom’ (Drazen and
Easterly 2001), there is only scant empirical evidence. Drazen
(2000) provides an overview of the political economy models that
lead to this prediction. 3 Chapter 2 deals with this crisis
hypothesis.
-
Introduction
3
uncertainty about the distribution of costs of the reform and no
credible commitment is made to
compensate the cost-bearing groups.
More recent research combines elements of different models. For
example: a more
ideologically fractionalised government may be a better reformer
because the parties in
government face high political costs if they reverse reforms
agreed upon. When the reforming
government consists of parties representing diverse
constituencies the promise to compensate
ex post reform losers becomes more credible, especially in
high-trust societies (Leibrecht and
Pitlik 2015). There is more political capital at stake as the
government is likely to break down if
constituencies represented in the government are not
compensated, this signals credibility of the
reform.
Another class of models are about electoral cycles.
Self-interested politicians that care
about re-election are unlikely to push for economic reforms that
may harm re-election chances.
The models in this area predict that governments cut taxes
before elections, but, dependent on
the assumptions made, some models predict that spending is
raised pre-elections whereas others
predict that (wasteful) spending is cut before elections (see
Drazen 2000). This view is tested in
chapter 2 and 3.
In sum, there is a large theoretical literature explaining the
delay, occurrence and
resilience of economic policy reforms. Persson and Tabellini
(2000: 481) argue: “the gap
between theory and evidence is a final weakness of the existing
literature”. Commensurate
empirical evidence has yet to materialise. Several empirical
contributions have attempted to fill
the gap (e.g. Leibrecht and Pitlik 2015, Potrafke 2009, Mierau
et al. 2007, Tavarez 2005).
However this literature builds on questionable measurement and
identification of economic
policy reform. By developing an improved identification
methodology this thesis takes a step in
filling this gap. The aim of chapters 2 and 3 is to test several
of the theories mentioned above.
Fundamentally, governments need a political majority to carry
out reforms or
adjustment programs. Chapter 4 investigates how the ideological
preferences of individual
voters impact the political majority generated at elections.
The dominant hypothesis is that voters’ economic self-interests
govern their choices
(Downs 1957). However, in large-scale elections the economic
benefit of voting becomes
small, as the impact of individual voters on the election
outcome is negligible. This implies
voter participation rates that are close to zero, which is in
stark contrast to reality. The Theory
of Expressive Voting has been invoked to explain voter turnout
at elections. But also to argue
that expressive preferences can tip elections in favour of what
a majority considers the
moral/social optimum, rather than the economic self-interests of
the majority (Feddersen et al.
2009, Tullock 1971). This argument is important because it is
left unexplored how the
-
Chapter 1
4
4
ideological preferences of voters impact their decisions in
political elections. The ideological
profile of a given reform may be particularly important if it
appeals to expressive rather than
economic preferences of a majority of voters.
By employing a laboratory experiment we analyse individuals’
choices in large
elections by comparing it to situations where the individual is
highly likely to impact the
outcome of a collective decision. We test two hypotheses: 1)
Whether ideological preferences
make voters more likely to participate in elections. 2) Whether
individuals are more likely to
vote according to their ideological preference in large
elections.
Fowler (2006) finds evidence that benevolent voters with a
strong party affiliation are
more likely to participate in large-scale elections. Yet there
is not much evidence whether
voters with an ideological preference are more likely to turn
out compared to voters without a
clear ideological preference. Feddersen et al. (2009) find
evidence that voters are more likely to
behave ‘moral’ in large scale-elections, where “moral” refers to
a choice that distributes income
equally while simultaneously maximising aggregate income. The
non-moral/selfish choice
maximises the individuals’ economic self-interest. We argue that
such a situation is rare in
democratic decisions. In a more realistic setting a certain
reform choice has a trade-off in terms
of societal consequences (e.g. redistribution versus higher
aggregate income). There is lack of
evidence whether choices of participating voters in large
election depends on the individuals’
ideological/societal preference or strict economic
self-interests. This is the topic of chapter 4.
Another central issue in the reform literature is the assessment
of the economic
consequences of reforms. Reforms often take place gradually; an
important condition for
reform progress is whether the reform outcome delivers what it
is supposed to, before
additional reforms steps are taken. Additionally, from a
societal perspective it is important to
evaluate the economic effects of reforms. The aim of chapter 5
is to evaluate whether
healthcare financing privatisations curb aggregate healthcare
expenditures.
In summary, the overarching research question of this thesis is:
what causes economic
policy reforms and what are the economic consequences of
economic policy reforms? This
research question is split into the following four
sub-questions, which subsequently are
addressed in chapter 2-5:
1. How can economic policy reforms be identified?
2. What drives economic policy reforms?
3. How does individuals’ voting behaviour impact the likelihood
of economic reforms?
4. What are the economic consequences of economic policy
reforms?
-
Introduction
5
Chapter 2 of this thesis focuses on measurement. Without a
proper identification
methodology the causes and consequences of the investigated
economic policy reforms cannot
be assessed. Chapter 2 uses healthcare financing privatisations
as illustration. Chapter 3 also
uses (parts of) the proposed methodology to test central
theories explaining fiscal adjustments.
Chapter 4 relates to the primary research question by
investigating how the preferences of
individual voters tie to political choices. Chapter 5 uses the
methodology proposed in chapter 2
to identify reforms in order to evaluate the consequences of
healthcare financing privatisations.
Chapter 6 provides an overview of the main findings of the
research and discusses policy
implications and limitations.
The following sub-sections provide an introduction to chapters
2-5 of this thesis.
1.2 Identifying economic policy reforms A major reason for the
lack of evidence on the economic impact of reforms is the
difficulty to
measure reform (Campos and Horváth 2012). Naturally, a similar
logic applies to the factors
that trigger or delay economic reforms. Chapter 2 aims to
develop a methodology to identify
economic reforms more precisely than previous methods.
Structural reforms can have elements of both ‘stroke-of-the-pen’
policies, i.e. the
passing of a law with no measurable economic effect (Easterly
2006), and ‘rigid institutions’,
i.e. an implemented reform only becomes visible after some time
has passed (Acemoglu et al.
2006). Consequently, when measuring reforms, and especially
their timing, the distinction
between policy input and economic outcome is important. In
several commonly used reform
indexes this distinction is not made.4 Our proposed methodology
utilises both policy input data
and economic outcome data without mixing it.
More specifically, gradual changes in the financing of public
activities often take place
without any change in ownership. This is, for example, the case
when health, education and law
enforcement provision are outsourced to private companies,
and/or when the financing of these
sectors is privatised. Such reforms can be identified with
appropriate data measuring the extent
of public and private involvement in a given sector.5 This
should be combined with de jure data
on policy changes to validate whether (gradual) shifts are
policy induced. By employing the Bai
4 This holds true for the Economic Freedom of the World index of
the Fraser institute and the Index of Economic Freedom of the
Heritage foundation, for example. This makes it hard to evaluate
outcomes of reforms if these are also used to score reform efforts
(Rodrik 2005). Furthermore, Campos and Horvath (2012) show that the
Economic Freedom of the World index is subject to changes without
attendant changes in the underlying data. This means that the
algorithm used to calculate the index has changed. There is lack of
information about how this algorithm is constructed. Also, the
index makes use of ill-defined reference points. To obtain the
highest score a country must reach the level of a ‘well-functioning
market economy’. This introduces subjectivity in the scores given
to individual countries. 5 E.g. concerning provision of healthcare:
the private share of total hospital beds. Concerning financing of
healthcare: the private share of total expenditures.
-
Chapter 1
6
6
and Perron (1998; 2003) endogenous structural break filter,
statistically significant changes in
the degree of privatisation can be identified. However, such
changes can also come from
factors unrelated to policy such as exogenous shifts in consumer
preferences and/or prices. To
overcome this problem, the identified structural changes need to
be validated to ensure that they
originate in policy and do not reflect some other factor. The
employment of de jure data on
reforms does exactly this. The development of this methodology
is the main contribution of
chapter 2.
The chapter applies the methodology to identify healthcare
financing privatisations. The
identified reforms are used to test several ‘usual suspects’
from the theoretical reform
trigger/delay literature. This also sets the chapter apart from
previous empirical work on the
determinants of reform. In this literature reforms are commonly
measured across sectors (e.g.
Bortolotti and Pinotti 2008, Pitlik and Wirth 2003). This may
not be appropriate since reforms
of some sectors are more politically controversial than others.
The proposed methodology
makes it possible to test theories without assuming a common
process of reform across sectors.
The first part of chapter 2 deals with the identification of
structural reforms. The second
part of chapter 2 examines the determinants of the identified
privatisations. This is done to
highlight the usefulness of the proposed methodology. First, the
‘usual suspects’ (that are
believed to determine economic policy reforms) are discussed.
Then, a binary outcome random
effects logistic model is used to test these determinants. The
chapter ends with a robustness
analysis.
1.3 What drives successful fiscal adjustments? The aim of this
chapter is to re-assess the economic and political drivers of
successful fiscal
adjustments. The dominant view in this literature is that
adjustments that are expenditure based
are more likely to reduce public debt in the long-term than
adjustments that are tax based
(Alesina et al. 1998; 2012, Alesina and Ardagna 1998; 2010;
2013, Alesina and Perotti 1995).
There are, however, several reasons to doubt this finding.
First, the findings on which
this view is based are mostly built on simple comparisons of the
average change in fiscal
expenditures and revenues during successful and unsuccessful
adjustments. Such an approach
ignores factors that simultaneously are correlated with the
probability that a successful fiscal
adjustment occurs and the choice of expenditure cuts or tax
increases during the adjustment.
We apply conditional fixed-effects models to re-assess the
evidence and explicitly take account
of several control variables.
Second, the standard approach to identify fiscal adjustments
relies on what can be
characterised as ‘ad hoc’ filters. Instead of relying on a
rigorous notion of statistical
-
Introduction
7
significance, rough filters based on simple algorithms are
designed to identify adjustments.
Adjustments are generally defined as: a discretionary (i.e.
cyclically adjusted) significant
positive change in the general government’s financial balance.
The notion ‘significant’ in the
definition does not refer to ‘statistical significance’, but
rather to whatever the applied filter
picks-up.6 These filters are based on a one-size-fits-all
principle, as they do not take into
account that the budgetary process in some countries may lead to
a much more volatile budget
balance than the budgetary process in other countries. A filter
that does not take volatility into
account is prone to identify fiscal adjustments that are the
result of the budgetary institutions in
place, rather than a deliberate attempt of politicians to
improve the budget balance. As such,
empirical analyses that have used these ad hoc filters may
suffer from severe measurement
error. Therefore, in this chapter, different versions of the Bai
and Perron (1998; 2003)
endogenous structural break filter are used to identify the
start of fiscal adjustments
Third, most previous studies have ignored potentially important
political-economy
determinants of successful fiscal adjustments. These
determinants are included in the empirical
models that are estimated in this chapter. To that end, this
chapter relies on our own update of
the World Bank Database of Political Institutions measuring
political characteristics in 21
OECD countries from 1975 to present. To investigate the
robustness of our results different
filters to identify fiscal adjustments and alternative
definitions of what constitutes a successful
fiscal adjustment are applied.
1.4 Expressive voting and political ideology The aim of this
chapter is to test the micro-behaviour of voters in large-scale
elections. This ties
to the basic condition for reform in developed democracies: A
political majority is needed to
induce economic reform.
Little is known about the importance of ideological
considerations relative to pure
economic self-interests of individual voters. The Theory of
Economic Voting stresses that
voters predominantly care about economic considerations when
voting, while non-selfish
concerns, i.e., how does the election affect others, matter
less. However, in large-scale elections
the individuals’ vote is unlikely to affect which candidate,
party or government is chosen.
When the probability that an individuals’ vote will impact the
final outcome is very low, the
discounted expected benefit of voting also becomes very low. If
the discounted benefit is lower
than the cost of voting, the economic voting model predicts that
the individual will abstain from
6 For example, an improvement of the budget balance by at least
0.25% points in the first year, a minimum duration of 2 years and a
total improvement of the budget balance by at least 2% points. This
is a so-called ‘gradual adjustment’.
-
Chapter 1
8
8
participation. However, this result is inconsistent with turnout
rates observed in real elections
(Levine and Palfrey 2007). This is the so-called Paradox of
Voting (Downs 1957).
A common solution to the paradox is based on the argument put
forward by Tullock
(1971). He argues that voters behave uncharitable/selfish in
situations where they are likely to
determine the outcome, but may behave charitable in situations
where they are unlikely to
determine the outcome. This argument is based on theories of
‘internal dissonance reduction’,
‘warm-glow’ or ‘identity confirmation’
The “charity of the uncharitable” argument has two important
implications for how
ideology may impact the political majority generated in
elections. Firstly, voters who can
identify themselves ideologically with a political candidate,
party or outcome may be more
likely to turn out and vote in large-scale elections.7 The
reason is that expressive ideological
preferences are independent of pivotal probability. Secondly,
the actual choice of participating
voters may shift from self-interested (instrumental) behaviour
to expressive behaviour when the
electorate grows large.8
Even though highly relevant for contemporary politics, the
evidence concerning the
second implication has not been linked to political ideology. We
provide this link by tracing-
out the instrumental and expressive incentives of voters. In
particular, we focus on the impact
of the size of the electorate. To that end, we develop a
decision-theoretic model.
Our model is embedded in the Public Choice literature on voter
turnout (e.g. Feddersen
et al. 2009, Levine and Palfrey 2007). In our model, voters have
to make two binary decisions.
First, they have to decide whether to participate in the
election by paying a fixed voting cost.
Second, participating voters have to decide option to vote for.
In our setup: option 1 is a
stylised version of socialism, whereas option 2 is a stylised
version of capitalism. At the same
time voters face individual monetary incentives that may be in
conflict with the outcome they
prefer ideologically. The main predictions of the model are: 1)
that non-centre voters (voters
with a socialist/left or capitalist/right ideological
preference) are more likely to participate as
the electorate grows large and 2) expressive ideological
preferences are determining their
preferred outcome choice. We test the predictions of the model
in a laboratory experiment with
90 subjects. A (separate) survey on the ideological preference
of the subjects is used to measure
ideology. Computer simulated voters are introduced in the
experiment to increase the size of
the electorate such that it is comparable to the size of
large-scale elections. A Heckman
selection model used to analyse the data from the
experiment.
7 Fowler (2006) presents evidence supporting this hypothesis. 8
Feddersen et al. (2009) presents evidence supporting this
hypothesis.
-
Introduction
9
1.5 Do healthcare financing privatisations curb total healthcare
expenditures? The aim of this chapter is to evaluate whether more
private healthcare financing curbs total
Health Care Expenditures (HCE). Healthcare reforms, hereunder
healthcare financing
privatisations, have long been proposed as a tool to contain the
rising healthcare expenditures in
advanced economies (OECD 1987, OECD 1992, Oxley and Macfarlan
1995). However, there
is a gap in the knowledge about the effects of these reforms.
Several studies have tried to
disentangle the consequences of healthcare privatisations, both
in terms of quality, equality and
costs (Colombo and Tapay 2004, Saltman and Figueras 1998, Tuohy
et al. 2004). One common
characteristic of these studies is their reliance on case study
evidence, which is largely
descriptive and does not investigate causal relations
rigorously. This research has not been able
to establish whether healthcare financing privatisations deliver
efficiency increases or cost
savings.
A large literature investigates the determinants of HCE. These
studies generally use a
‘determinants approach’, where HCE is regressed on variables
thought to affect it. This
literature cannot answer whether decreases in the share of
publicly financed HCE curb total
spending, since the common econometric approach suffers from
simultaneity and possibly
spurious regression relationships (De Mateo and De Mateo 1998,
Hansen and King 1996).
Some studies look at whether the share of publicly financed HCE
impacts total HCE. Leu
(1986) finds that a larger public share is associated with
higher total spending on health care.
To the contrary, Hitris and Posnett (1992) find no effect of the
publicly financed share of HCE
on total spending.
Theoretically, more private healthcare financing improves
efficiency. That is, if
healthcare is largely publicly financed, consumers are not cost
conscious as they have
incentives to demand more than the optimum level of health care.
Similarly, healthcare
providers have incentives to accommodate demand. Consequently,
if payment for healthcare
shifts from public to private ‘pockets’, all else being equal,
demand reduces to more optimal
levels and total HCE will decrease.
However, the privatisation literature suggests that
privatisations often happen as a
consequence of ‘special interest politics’ instead of pure
efficiency reasons. For example, right
wing market-oriented governments privatise healthcare financing
to reduce the re-distributional
effects of a public tax financed system. If this rationale is
behind privatisations (rather than an
economic welfare analysis), then it is far from clear that
privatisations deliver efficiency gains
(see Cavaliere and Scabrosetti 2008). In sum, it has not been
established whether healthcare
financing privatisations curb total healthcare expenditures.
-
Chapter 1
10
10
This chapter takes a quantitative approach to analyse the
consequences of healthcare
financing privatisations. It sets the analysis apart from
previous work, which suffers from the
absence of a methodology that can identify reforms consistently
across countries over time.
Ideally, costs should not be considered in isolation from
quality and equality in healthcare. In
order to assess efficiency data on the overall level of quality
and equality of health(care) in
countries are needed. Such data is unavailable at the time of
writing. It is however possible to
assess cost effects of healthcare financing privatisations. Such
analysis is the main contribution
of this chapter.
The methodology to identify reforms developed in chapter 2 is
used to identify health
care financing privatisations. In turn, Propensity Score
Matching (PSM) is used to evaluate the
effect of those reforms. PSM has become a standard and effective
way to assess the effect of
policy interventions in the absence of experimental data (Person
et al. 2001). Using part of the
results of chapter 2, common causes for reform are included in
the model to predict the
propensity scores. In addition, sector specific variables are
included in the analysis. To probe
the robustness of the results, several matching techniques are
used to compare the outcome of
the analysed privatisations with (an) appropriate
counterfactual(s).
-
Chapter 2
What triggers reforms in OECD countries? Improved reform
measurement and evidence from the healthcare sector
2.1 Introduction There is a large theoretical literature in the
field of political economics about the factors that
determine whether reforms occur or are delayed (see Drazen
2000). However, robust empirical
evidence about these theories has not materialised (compare
Alesina and Argdana 2012;
Bortolotti and Pinotti 2008; Lavigne 2011; Roberts and Saeed
2012; Tavares 2004). Beyond the
fact that different economic reforms are investigated, it is
likely that the lack of robust evidence
is caused by inconsistent measurement and identification of
reforms. Babecky and Campos
(2011) show that poor reform measurement is a main cause behind
the non-robust evidence
when it comes to the growth effects of reforms. It is therefore
likely that the non-robust
evidence concerning reform triggers is caused by the same
problem – the absence of a coherent
methodology with which economic reforms are identified. The main
contribution of this paper
is the development of such a methodology. We envisage that it
can be applied to identify a
variety of reforms. Not only healthcare-financing
privatisations, which are used as an
application throughout this paper, can be identified by using
it. Identification of reforms in
sectors where change happens gradually and where economic
outcome data or policy input
data, used alone, is likely to lead to identification error can
benefit from the methodology.
In the literature at least two complementary views exist on
which kind of data that
ideally should be used to measure reform. The first view,
expressed in Campos and Horváth
(2012), holds that economic outcome data and policy input data
never should be mixed. The
reason is that a failure to separate the two is likely to cause
imprecise measurement of both the
timing and the impact of reform. The second view, expressed in
Rodrik (1996), holds that only
policy input should be used to measure reform. The argument is
that factors outside
policymakers’ control will impact economic data. Hence, only
policy input data is a valid proxy
for reform according to this view.
Here, a novel sequential approach is taken that utilises both
economic outcome data and
policy input data without mixing it. First, the Bai and Perron
(1998; 2003) (B&P) structural
break filter is applied to detect statistically significant
changes in economic outcome data.
This chapter is based on Wiese (2014).
-
Chapter 2
12
12
Then, the detected structural breaks are validated using
evidence of de jure reforms. Selecting
one of the two steps alone when identifying reforms can lead to
either: 1) Detection of changes
that are not policy induced, and thus, do not qualify as
reforms. 2) Identification of de jure
reforms that did not have a significant impact, and thus, do not
qualify as ‘successful’ reforms.
The methodology detects economically successful de facto
reforms, that is, statistically
significant policy induced changes in the share of publicly
financed healthcare expenditures.
As an application of this improved identification methodology we
test some of the
usual suspects that are believed to determine whether economic
reforms are initiated or
delayed. This paper offers a quantitative empirical analysis of
the political economy
determinants of healthcare financing privatisations. Something
that is absent in the literature,
probably due to the lack of a coherent methodology to identify
that specific kind of reforms. To
complete such an analysis the random-effects binary outcome
logistic estimator is applied.
Controlling for economic factors and duration dependence several
interesting empirical results
are found. First, the results suggest that severe economic
crises trigger healthcare financing
privatisations. Severe recessions, high levels of unemployment
and high interest rates on
government debt are found to significantly increase the
likelihood of the analysed reforms.
Second, no evidence is found that pure political factors impact
the likelihood of healthcare
financing privatisations. Neither ideology, political
fractionalisation, newly elected
governments nor upcoming elections are found to significantly
affect the likelihood of the
investigated privatisations. These results are robust to
different definitions of healthcare
financing privatisations. Furthermore, the main findings are
robust to different proxies for the
hypothesised determinants. Thus, the second contribution of this
paper is a consistent
quantitative analysis of the determinants of healthcare
financing privatisations in OECD
countries.
Section 2.2 outlines the novel identification methodology.
Section 2.3 develops the
hypotheses while the data and the empirical strategy are
discussed in section 2.4. Section 2.5
presents the baseline empirical results and section 2.6 gives
robustness analyses. The paper
ends with a conclusion in section 2.7.
2.2 Identifying reforms In this section the identification
methodology is explained. This is done through an application
to healthcare financing privatisations. However, the methodology
is general and can easily be
adapted to identify other types of reforms, for example
fiscal-adjustments or -expansions, or
privatisations and nationalisations in various sectors.
-
What triggers reforms in OECD countries?
13
There exist two groups of definitions of privatisations in the
literature, a broad and a
narrow one. The broad definition concerns overall shifts in the
boundary between public and
private involvement in the economic sphere (e.g. Vickers and
Yarrow 1991). The narrow
definition concerns shifts in ownership (e.g. Roberts and Saeed
2012). The drawback of the
broad definition is that it is problematic to operationalize.
The drawback of the narrow
definition is that it fails to capture shifts from the public to
the private domain when no shift in
ownership takes place.1 This is often the case with
healthcare-financing privatisations (see
appendix A2). As shown below we can measure public versus
private sector involvement
instead of ownership. And, as discussed in the introduction, we
need to make sure that shifts in
involvement are both policy induced and have a statistically
significant impact. Therefore, a de
facto healthcare financing privatisation is defined as: A
statistically significant policy induced
shift from public to private sector financing of healthcare
services.
What we would like to measure is public versus private funds
incurred to healthcare -
from which pocket is healthcare expenditure paid, public or
private? The historical ratio yit of
public healthcare expenditure relative to total healthcare
expenditure (public + private) in
country i in year t can be used to identify privatisations. 2
This ratio is calculated as:
. It can be interpreted as the percentage of public financing
of
total spending. Hence, we have a measure of public relative to
private financing of heath care
services.
Table 2.1. Descriptive statistics for variables used to
construct the dependent variable Variable Obs. Mean St.d Min. Max.
Source: Public healthcare expenditure % of GDP 932 5.46 1.70 0.84
9.74 OECD.org Private healthcare expenditure % of GDP 942 2.02 1.39
0.11 9.31 OECD.org Total healthcare expenditure % of GDP, private +
public 931 7.50 2.26 1.49 17.67 Calculated Public relative to total
expenditure, yit 931 0.73 0.13 0.22 0.98 Calculated All available
observations for the 23 OECD countries between 1960-2010 have been
used. See table 2.2 for exact sample periods.
The idea then is to apply structural break testing to identify
significant shifts in yit. A
structural break is the timing of a fundamental change in the
Data Generating Process (DGP),
1 This is for example the case with education, law enforcement
and healthcare reforms, both in terms of provision and financing. 2
Public healthcare expenditure is defined as: “health expenditure
incurred by public funds. Public funds are state, regional and
local Government bodies and social security schemes. Public capital
formation on health includes publicly financed investment in health
facilities plus capital transfers to the private sector for
hospital construction and equipment” (OECD.org). Private healthcare
expenditure is defined as: “Privately funded part of total health
care expenditure. Private sources of funds include out-of-pocket
payments (both over-the-counter and cost-sharing), private
insurance programs, charities and occupational health care”
(OECD.org).
yit =
publicfundsitpublicfundsit + privatefundsit
-
Chapter 2
14
14
for example as a result of an economic reform (Hansen 2001).3
However, a structural break can
be caused by other factors, such as an exogenous shift in
consumer preferences, or relative
price movements. Thus, the detected structural breaks need to be
validated using de jure
evidence of reforms. If a change in policy causes a significant
part of healthcare financing to
shift from (to) the public to (from) the private sector it is a
de facto privatisation
(nationalisation).
Perhaps the most well-known structural break test is the
Chow-test. A noticeable feature
of this test is that it is limited to test the hypothesis of
whether a time series contains a single
structural break. To use the test one has to split the sample at
the point in time where a priori
information leads one to suspect a break and then use F-tests to
determine whether subsample
parameters are significantly different. For the application at
hand, de jure evidence gives a
priori information of several structural breaks in each time
series, up to 25 in some cases (see
HSiT country reports). This means that the time series would
have to be split into a large
number of subsamples on which the Chow-test could be performed.
This is infeasible because
the time series are not long enough when a priori information
leads us to suspect so many
breaks. Furthermore, there is often a time lag before a de jure
reform manifests itself in the
data; institutions are rigid (Acemoglu et al. 2006). This means
that the division of the time
series into subsamples would be arbitrary.
The feasible approach is to start from the economic data and
then use de jure evidence for
validation. Hence, the number and timing of structural breaks
are treated as unknown a priori.
The econometric literature on detection of an unknown number of
structural breaks of unknown
timing is relatively sparse. Liu et al. (1997) suggest a method
for pure structural changes, i.e. all
parameters included are subject to changes. Bai and Perron
(1998; 2003) develop a more
general method for this purpose. Their method also allows the
inclusion of parameters that are
not subject to shifts, of which the pure structural change model
applied here is a special case.
The assumptions of the B&P-filter are less restrictive.
Therefore, the B&P-filter is applied to
identify structural breaks.
In order to define privatisations (and nationalisations) in the
context of the B&P-filter
consider a model with m possible structural breaks in an OLS
regression framework that takes
the form:
yt=δj+ut (t=1,...,T , j=1,…,m+1)
Where yt is the dependent variable, in this case the time series
of public relative to total
healthcare expenditure for each country considered. δj is a
vector of estimated coefficients 3 Seminal examples of structural
breaks are: the unification of Germany, and the introduction of a
common European monetary policy.
-
What triggers reforms in OECD countries?
15
(constants) of which there are m+1, i.e. δj is the mean at the
different segments of the time
series yt. ut is the error term. The segments generate a
stepwise linear route through the times
series yt and give m structural breaks.4 Fig. 2.1 provides a
graphical exposition where the j=6
segments are represented by dashed lines and the m=5 structural
breaks are represented by
dotted lines. A downward (upward) regime shift is detected as a
potential privatisation
(nationalisation), for which validation using de jure evidence
is required. See appendix A2 and
B2 for all time series of countries included.
Fig. 2.1. Structural breaks in healthcare financing source: The
case of Austria.
The data used to generate fig. 2.1 has 51 observations, the
trimming parameter was set at h=5 i.e. at least 5 periods must pass
between consecutive breaks. See main text for an explanation.
The idea underlying the B&P-filter is straightforward.5 It
generates the segmented route
through the series that yields the lowest Sum of Squared
Residuals (SSR) up to a maximum
number of breaks. The maximum number of breaks is restricted by
a trimming parameter h,
which specifies a minimum number of observations that has to
occur between consecutive
breaks.6 In the context of fig. 2.1 the segments can be thought
of as regimes where yt fluctuates
around the constant mean δj. A shift to a new regime is unlikely
to happen by chance,
4 The maximum number of breaks is determined exogenously by
specifying the minimum number of observations that has to occur
between breaks prior to the analysis, see below. 5 The process
underlying the algorithm is straightforward. First, it searches for
all possible sets of breaks up to a maximum number of breaks,
restricted by the trimming parameter chosen, and determines for
each number of breaks the set that minimises the SSR (Sum of
Squared Residuals). Then a series of F-tests determine whether the
improved fit produced by allowing an additional break is
sufficiently large, compared to what can be expected randomly, on
the basis of the asymptotic distribution derived in Bai and Perron
(1998). Starting with a H0 of no breaks, sequential tests of k vs.
k+1 breaks allow one to determine the appropriate number of breaks
in a data series. Alternatively, the BIC (Bayesian Information
Criteria) works well if there is evidence of at least one break.
After determining the appropriate number of breaks the program
extracts the corresponding break dates of the optimal sequential
route. The trimming parameter h is expressed either as a fixed
number of observations, or a percentage of the number of
observations. Autocorrelation, trending time series and
non-constant errors are permitted (Bai and Perron 2003). 6 The
trimming parameter h imposes a risk. If two potential
privatisations are less the h periods apart, the B&P-filter by
construction can only identify one of them.
1960 1970 1980 1990 2000 2010
0.64
0.66
0.68
0.70
0.72
0.74
0.76
Austria, h=5
year
publ
ic re
lativ
e to
tota
l exp
endi
ture
-
Chapter 2
16
16
dependent on the test-size employed. Thus, a regime shift
implies that the underlying DGP has
been altered generating a structural break.
When applying the B&P-filter several test procedures are
possible (see Bai and Perron
1998; 2003 or Zeileis et al. 2003). Here the Bayesian
Information Criterion (BIC) is chosen to
select the optimal number of m+1 segments. Information criteria
are often used for model
selection, which in this case means selection of the m number of
breakpoints. Bai and Perron
(2003) argue that the Akaike Information Criterion usually
overestimates the number of breaks,
but that the BIC is a suitable selection procedure in many
situations. 7 Furthermore, when
applying the B&P-filter a choice has to be made concerning
the size of the trimming parameter
h. In case the times series do not exhibit autocorrelation or
heteroskedaticity any trimming will
work regardless of sample size (Bai and Perron 2003). However,
when finite samples that do
exhibit autocorrelation and/or heteroskedaticity are used the
trimming needs to be increased.
Here a trimming of h=5 is chosen because it generates the best
fit with de jure evidence while
still being econometrically sound. The outcome of using a more
conservative trimming of h=6
can be found in appendix A2. Autocorrelation and potential
heteroskedasticity is modelled non-
parametrically by running the filter using a Heteroskedasticity
and Autocorrelation Consistent
(HAC) estimate of the variance-covariance matrix.8 Antoshin et
al. (2008) show that HAC
errors generally deal with autocorrelation better than
parametric modelling.
The outcome of running the B&P-filter on 23 OECD countries
can be found in column
3 in table 2.2, 1960-2010 is selected as sample period. For some
countries data is unavailable
for the whole period. In that case the longest data period
available is used, see column 2 for
exact sample periods.
It is possible that factors outside control of the policy-maker
move the ratio
significantly and hence look like a reform when it in fact was
not. Therefore, it is checked
whether the detected privatisations are likely to be the result
of planned policy. This was done
by employing the WHO’s and European Observatory on Health
Systems and Policies
“Healthcare Systems in Transition” country report series. These
reports are available for each
country covering the sample period and have descriptions of de
jure reforms introduced
through time in each country. When a report describes a policy
reform that directly or indirectly
could have had the objective to either reduce the public share
of healthcare financing, increase 7 The filter is largely robust to
the use of other information criteria such as the Schwarz and the
Akaike criterion, and the global F-test proposed by Bai and Perron
(2003). But not robust to using the sequential F-test proposed in
Bai and Perron (2003) that generally is more conservative
concerning the number of estimated breaks. 8 Choosing the trimming
to be a fixed number of observations instead of a percentage of the
sample size has the implication that the percentage of the sample
size automatically is increased for shorter samples. Bai and Perron
(2003) argue that shorter samples exhibiting autocorrelation and
heteroskedasticity calls for larger trimmings in percentage of the
sample size.
-
What triggers reforms in OECD countries?
17
the private share of healthcare financing, or both, it is taken
as evidence of a de jure reform. A
time lag is present between the de jure reforms and the detected
economic outcome of those; a
change in policy does not manifest itself immediately as an
economic outcome. That is, there is
evidence of ‘rigid institutions’ (Acemoglu et al. 2006). In most
cases a time lag of one year is
present (see appendix A2). If more than two years passed between
a de jure reform and a
detected structural break, the reform is not coded as a de facto
privatisation.9 See column 4 in
table 2.2 for the outcome of this analysis, and appendix A2 for
a detailed description of the
related de jure reforms.
In sum, the analysis reveals that 22 of the 33 detected
privatisations can be validated.
We are therefore confident that these 22 structural breaks are
the result of planned policy, and
therefore match the definition of a de facto healthcare
financing privatisation. The
privatisations detected by the methodology generate a binary
variable. Years in which a
privatisation is detected are coded as 1, the remaining years as
0. This constitutes the dependent
variable used in the estimations that follow in section 2.5 and
2.6.
Table 2.2. Identified privatisations Country Sample period
Detected Privatisations De facto privatisations Australia 1969-2009
Austria 1960-2010 1968; 1988 1968; 1988 Belgium 1995-2010 Canada
1970-2010 1985; 1993 1985; 1993 Denmark 1971-2010 1983; 1989 1989
Finland 1960-2010 1992 1992 France 1990-2010 2005 2005 Germany
1970-2010 1982; 1997; 2003 1997; 2003 Greece 1987-2010 1993 1993
Iceland 1960-2010 1983; 1991; 1996 1991 Ireland 1960-2010 1984
Italy 1988-2010 1993 1993 Japan 1960-2009 Luxembourg 1975-2009
1981; 1999 1999 Netherlands 1972-2002 1997 1997 New Zealand
1970-2010 1989 1989 Norway 1960-2010 1979; 1988 1988 Portugal
1970-2010 1981 Spain 1960-2010 1988; 1994 1988; 1994 Sweden
1970-2010 1984; 1990; 1995; 2000 1995; 2000 Switzerland 1985-2010
United Kingdom 1960-2010 1981; 1986; 1996 1981; 1986 United States
1960-2010 Data source for detected privatisations: OECD.org,
Economic Outlook nr. 90. West German data is used prior 1990 for
Germany. Data source for validated privatisations: HSiT (Healthcare
Systems in Transition) country reports; see appendix A2. 9 A risk
of the methodology is that the outcome of a de jure reform takes
more than two years to manifest in the data. However, increasing
the time-window above two years increases the risk that other
factors might impact the ratio yit, therefore two years are
chosen.
-
Chapter 2
18
18
Before we proceed with a review of the usual suspects that are
believed to trigger
reforms in general, and privatisations in particular, we end
this section with a case study. This
is done to highlight the benefits of the presented methodology
over existing ones. Without loss
of generality the case of Norway is selected. Had we relied on
identification of healthcare
privations by applying the B&P-filter to economic outcome
data alone, we would falsely have
identified a privatisation in 1979. Johnsen (2006) provides an
overview of healthcare reforms in
Norway through time. The report does not describe any reforms
within the two years preceding
the identified structural break. Had we relied solely on
economic output data to identify the
reforms we would falsely conclude that a privatisation had taken
place. However, using de jure
data to verify the detected reforms prevents such identification
errors.
The other possible method that has been argued for in the
literature is the use of policy
input data alone. However, this approach is at least at
problematic as using economic outcome
data alone. Johnsen (2006: 125) provides an overview of “Major
health care reforms and policy
measures” from 1984 to 2004. In that period more than 10 reforms
occurred that potentially
could have had a significant impact on whether the costs
healthcare would shift from public to
private financing. Thus, without application of outcome data we
could make a substantial
amount of identification errors. Additionally, when relying on
policy input data alone it is
necessary to interpret reform descriptions. What was the primary
intention of a given reform, a
privatisation or a nationalisation, if any? Often, as supported
by the statistical analysis above
and the HSiT reports, the objective of the reforms was to
improve the quality and/or efficiency
of the healthcare system. Thus, the combination of both economic
outcome data and policy
input data offers an objective way to identify healthcare
financing privatisations.
A risk of the methodology is that that the outcome of a de jure
reform can be hidden in
the data by unrelated economic changes, such as exogenous shifts
in consumer preferences or
relative price movements. The opposite can also happen, that a
policy change has no significant
impact on the data, but unrelated economic changes lead us to
conclude that it had.
Nevertheless, our sequential procedure is less prone to
identification error than identification
from either policy input data or economic outcome data
alone.
2.3 Reform determinants: Political economics hypotheses In this
section we review what the political economics literature suggests
as factors determining
the occurrence of healthcare financing privatisations. The
hypotheses are not derived from a
single underlying theoretical model for healthcare reform
because no such model exists. Thus,
the hypotheses are the ‘usual suspects’ that are believed to
trigger or delay economic reforms in
general.
-
What triggers reforms in OECD countries?
19
Perhaps the most common view among scholars is that an economic
crisis is a
necessary and maybe sufficient condition to trigger reform.
Nonetheless, the hypothesis has not
been subjected thorough empirical testing. This lack of research
can be due to several causes.
First, when a hypothesis enters the realm of ‘conventional
wisdom’ empirical testing often
ends (Drazen and Easterly 2001). Second, Rodrik (1996) argues
that the ‘crises induce reform
hypothesis’ is a tautology, which means that there is nothing to
test. “Reform naturally becomes
an issue only when policies are perceived not to be working. A
crisis is just an extreme case of
policy failure. That reform should follow crisis, then, is no
more surprising than smoke
following a fire” (Rodrik 1996: 27). However, economic crises
can take different forms and
differences in the severity of crises can have a different
impact on reform probability. For
example, prolonged recessions may trigger healthcare reforms
while below average economic
growth rates do not. Furthermore, the effect of crises in
sparking reform may also depend
critically on what sort of reform is being considered (Drazen
and Easterly 2001). So, the
hypothesis is falsifiable.
Two common explanations for the ‘economic crises induce reform
hypothesis’ exist.
The first is based on the presumption that when significant
political opposition to reform exists,
an economic crisis may overcome reform resistance by convincing
the opposition that
something needs to be done (Drazen 2000). The second is centred
on the observation that ex-
ante uncertainty about the economic outcome of reform often
generates resistance to reform.
An economic crisis raises the cost of not reforming, effectively
decreasing political opposition
(Fernandez and Rodrik 1991). These arguments lead to the first
hypothesis.
H1: The likelihood of healthcare financing privatisations
increases when the economy
is in a state of severe crisis.
Another common topic in the literature is the effect of
ideology. To what extent do left-
and right-wing politicians or parties provide policies that
reflect the preferences of their
electorates? In political economy, such an effect is referred to
as the ‘partisan politics
hypothesis’. According to the conventional approach right-wing
governments implement
policies that favour the preferences of relatively high-income
voters, whereas left-wing
governments implement policies that favour the preferences of
low-income voters (Hibbs 1977,
Alesina 1987). Conventionally, right-wing market-oriented
governments are thought to be
inclined towards privatisation and liberalisation of public
sector enterprises.
However, the so-called “Nixon goes to China” effect, modelled by
Cukierman and
Tommasi (1998), comes to the opposite conclusion. They show that
governments are prone to
become victims of their own ideology. Only governments of the
unexpected ‘political colour’
can credibly signal that a policy is expected to be beneficial
when they have private information
-
Chapter 2
20
20
about the economic outcome of a given reform, and when voters
consider the reform ‘extreme’.
The asymmetric information is caused, for example, by access to
expert knowledge. Thus,
right-wing governments will find it harder to credibly signal
that right-wing policies will be
beneficial. The same holds for left-wing governments and
left-wing policies.
Empirical studies do not provide robust evidence concerning the
partisan politics
hypothesis and economic reforms in general. The discrepancy in
the empirical evidence can be
caused by measurement issues, but it can also depend critically
on the kind of reform is being
considered, i.e. whether private information is substantial and
whether voters consider the
reform ‘extreme’. However, when reviewing the empirical
literature on the relation between
privatisations and ideology the evidence is largely robust
(Bortolotti et al. 2003, Bortolotti and
Pinotti 2008, Elinder and Jordahl 2013, Potrafke 2009, Roberts
and Saeed 2012). 10 The
evidence is in favour of the conventional partisan effect,
namely that right wing governments
trigger privatisations and deregulations in developed countries.
Therefore, we also expect this
to hold in the case of healthcare financing privatisations.
H2: The likelihood of healthcare financing privatisations
increases when a right-wing
government holds office.
The two previous hypotheses concern factors that trigger
privatisations. However, it is
interesting to investigate whether there are factors that
diminish the probability of reform. It is
often observed that welfare-enhancing policies are delayed. This
can be explained by the ‘war
of attrition hypothesis’, which states that political opponents
play a ‘waiting game’. The basic
insight from the Alesina and Drazen (1991) model is that
governments will postpone unpopular
reforms, even if postponement is sub-optimal for social welfare,
until the ‘political’ costs of
reform has fallen below the ‘political’ benefits of postponing
it. This implies that more
heterogeneous governments will be less keen to reform public
enterprises because they have a
more heterogeneous constituency to please. They will face higher
costs, in terms pleasing their
electorate, and hence, postpone reforms. Thus, large coalition
governments will find it more
difficult to agree on reforms compared to more homogeneous
governments. Small groups can
use their veto power to block reforms if the change in
distribution of economic goods resulting
from the reform will be too costly, in terms of preferences of
their electorates. This argument
also holds in the case of privatisations (Lora 2000). This
explains the third hypothesis.
10 Robert and Saeed (2012) is the exception. They do not find a
significant relation between privatisations and ideology in their
subsample of advanced countries.
-
What triggers reforms in OECD countries?
21
H3: The likelihood of healthcare financing privatisations
decreases when
governments are more fractionalised.11
Another common issue is electoral cycles. Do elections affect
government’ actions?
Lora (2000) points-out that the literature on the timing of
reform offers little explanation of the
apparent importance of one factor that seems to be the simplest
reason for reform, namely a
newly elected government. Based on case studies, Haggard and
Webb (1994) argue that a
window of opportunity opens after elections. This can be
exploited by newly elected
governments, which “typically enjoy a period in which the costs
of adjustment can be traded
against political gains” (Haggard and Webb 1994: 8). Such a
window of opportunity could, in
relation to healthcare reform, be the interpretation made by a
newly elected government that
public healthcare expenditure have, or will become unsustainable
unless significant efforts to
curb the costs are initiated.
Another debated issue is whether electoral cycles exist before
elections. Do upcoming
elections affect governments’ actions? The models in this area
predict that governments cut
taxes before elections, but, dependent on the assumptions made,
some models predict that
spending is raised pre-elections whereas others predict that
(wasteful) spending is cut before
elections (see Drazen 2000). Thus, we only expect that
healthcare financing systematically is
privatised before elections if this spending is perceived
wasteful. Healthcare financing is hardly
seen as wasteful spending, so we expect that upcoming elections
decrease the likelihood of
them. The results in Bortolotti et al. (2003) support this view
concerning privatisations in
general. Concerning public healthcare expenditures in
particular, Potrafke (2010) finds that
these generally are raised in election years. Thus, we have two
predictions concerning electoral
cycles and healthcare financing privatisations.
H4: a) The likelihood of healthcare financing privatisations
increases when a newly
elected government or government executive is in office.
b) The likelihood of healthcare financing privatisations
decreases before elections.
2.4 Explanatory variables and empirical strategy In this section
the independent variables and the econometric technique used in the
estimations
are discussed.
11 A referee correctly pointed out that hypothesis 3 and 4 are
symmetric in the sense that they hold for both privatisations and
nationalisations. And thus, nationalisations identified by the
methodology could be included in an empirical model to test these
two hypotheses. However, since the focus is on privatisations we
abstain from such analysis and leave it for future research.
-
Chapter 2
22
22
2.4.1 Explanatory variables To test hypothesis 1 three commonly
applied indicators of economic crisis are used. A variable
commonly related to economic crisis is the unemployment rate.
The dummy variable job crisis
is used to capture this. It takes the value one when the
unemployment rate in country i at time t,
is above 9.57 = the sample mean plus one standard deviation.
Ideally, one should control for
differences in structural unemployment. However, structural
unemployment varies significantly
over time and no annual estimates for the sample period exist
(see Turner et al. 2001). In most
OECD countries an unemployment rate above 9.5% would be viewed
as an indicator of severe
economic crisis.
Second, if public debt reaches unsustainable levels it is
reasonable to expect that this
positively will impact the pressure to privatise fiscally
burdensome sectors of the economy,
such as healthcare financing. What matters is not the stock of
outstanding public debt as such,
but whether financial markets judge the debt to be sustainable.
This we can call a sovereign
debt crisis effect. The dummy variable debt crisis is used to
capture this effect. It takes the
value one when the interest rate on long-term government debt in
country i at time t, is above
11.42 = the sample mean plus one standard deviation. This
approach is in line with the Money
Market Pressure approach developed by von Hagen and Ho (2007)
where large deviations from
norm in the short-term interest rate is used to identify banking
crises. Sovereign debt crises
could be captured in other ways. For example Laeven and Valencia
(2008) identify sovereign
debt crisis using actual events of defaults and restructuring of
debt. This definition is too strict,
however, as it is reasonable to expect that governments will
react to high interest rates on
government debt with reforms, before actually defaulting on
their debt.
Third, poor macroeconomic performance is another commonly used
measure for
economic crisis. Economic growth rates capture this effect.
Specifically the dummy variable
annual recession is used. It takes the value one in years where
annual accumulated economic
growth was negative. The reason not to use the common definition
of a recession (two or more
consecutive quarters of negative economic growth) is that we
want to capture severe economic
growth crises. A country may experience a recession during a
period of a year, but still end up
with a positive growth rate annually. When the crisis is only
mild and/or transitory it may not
call for strong policy action.
All three indicators for economic crises are expected to
generate positive significant
coefficients. Pitlik and Wirth (2003) also make use of an
indicator variable approach to test the
crisis hypothesis. In the robustness analysis in section 2.6 the
variables underlying the
constructed crisis indicators are included in raw form.
-
What triggers reforms in OECD countries?
23
To test hypothesis 2 the Government ideology index proposed by
Budge et al. 1993 and
updated by Potrafke (2009) is used. It attempts to account for
the relative strength of parties in
government with reference to the left-right economic dimension,
on the basis on the parties’
policy actions. It uses a five-point scale in which the seat
shares of the left, centre and right are
transformed into scores from 1 to 5, representing the degree of
dominance of either party, both
in parliament and government.
The measure equals: 1 if the right-wing share of seats in
government and their
supporting parties in parliament is larger than 2/3. 2 if the
right-wing share of seats in
government and their supporting parties in parliament is between
1/3 and 2/3 each. 3 if it is a
centre party government, or if the left-wing and/or the centre
and/or the right-wing parties form
a coalition government. 4 and 5 is symmetric to 1 and 2, counted
in seats taken by left-wing
politicians in the parliament and cabinet. Years in which the
government has changed are
labelled according to the government that was in office for the
longest time that year. The
measure is consistent over time, but it does not attempt to
capture links between sister parties in
different countries (Potrafke 2009). According to hypothesis 2,
a negative significant
coefficient is expected.
To test hypothesis 3 the Government fractionalization measure
from the World-Bank
database on political institutions is used. It measures the
probability that two randomly chosen
deputies from the government parties will be from different
parties (Beck et al. 2001). That is,
how many different parties are parts of the coalition and how
large are the individual coalition
parties relative to all seats taken by the coalition. When the
measure increases the degree of
government fractionalization increases. According to hypothesis
3, a negative significant
coefficient is expected.
To test hypothesis 4a the dummy variable election year is used.
It takes the value 1 if a
legislative election and/or an executive election was held in
the given year. The variable is
constructed using the World-bank database on political
institutions. According to hypothesis 4a
a positive significant coefficient is expected. To test
hypothesis 4b the dummy variable
upcoming election is used. It takes the value 1 if a legislative
or executive election is going to
take place the following year. The variable is created using the
‘years left in current term’
variable from the World-bank database on political institutions.
When one year is left of the
current term the variable take the value 1, all other years are
coded as 0. According to
hypothesis 4b a negative significant coefficient is expected
-
Chapter 2
24
24
Table 2.3. Hypothesis summary table Hypothesis Variables used
Expected sign 1. The likelihood of healthcare financing
privatisations increases when the economy is in a state of
crisis
Job crisis, debt crisis and annual recession, OECD data
+
2. The likelihood of healthcare financing privatisations
increases when a right-wing government holds office
Government ideology, The Potrafke index
-
3. The likelihood of healthcare privatisations decreases when
governments are more politically fractionalised
Fragmentation of governments, The World-Bank
-
4a. The likelihood of healthcare privatisations increases when a
newly elected government is in office
Legislative and executive elections, The World-Bank
+
4b. The likelihood of healthcare financing privatisations
decreases before elections
Years left in term, The World-Bank -
2.4.2 Control variables To build a trustworthy empirical model
control variables are crucial. Factors impacting the
likelihood of healthcare financing privatisations through costs
and funding, while being
correlated with the main variables of interest must be included.
Shirley and Walsh (2000: 44)
assert: “Instead of maximizing its own rents and power, the
government places a priority on
efficiency. It can be argued that governments that engage in
privatisation are not the ones that
seek only rents and power.” That implies that we need to control
for factors that impact the
efficiency of the health care sector because they correlate with
privatisations and our political
variables of interest (Shirley and Walsh 2000). Specifically,
demand- and supply-side factors
for healthcare are expected to affect the efficiency of the
sector, and hence the likelihood of
privatisations, while being correlated with economic factors
such and economic growth and
public debt.
On the demand side an aging population will affect the level of
spending of the
healthcare sector adversely due to increased demand for
healthcare (De Mateo and De Mateo
1998, Oxley and MacFarlan 1995). Basically a larger share of
elderly leads to increased public
expenditures in a largely publicly financed system because
elderly persons have a higher
propensity of needing healthcare. An aging population may
therefore have an impact on the
likelihood of privatisations, because it is a cost determinant.
The demographic factor is
approximated by the percentage of the population over 65 years
of age.
On the supply side, a good performing healthcare system, in
terms of quality, may
impact the likelihood of privatisation. A good performing system
may be less prone to
significant reform, compared to a system that performs badly
where pressure to improve the
system may be significant. In order to control for such an
effect we include the variable
‘potential years of l