Thesis Economic policy reform 4-4 · In this thesis, economic reforms and policy adjustments are...
Transcript of Thesis Economic policy reform 4-4 · In this thesis, economic reforms and policy adjustments are...
-
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