Thesis Economic policy reform 4-4 · In this thesis, economic reforms and policy adjustments are both captured under the heading of ‘economic policy reforms’. Dependent on the

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