The politics of health care: examining the role of partisanship in health care

(1)

1

The politics of health care: examining the role of

partisanship in health care

Leiden University

Public administration: Economics and Governance Track Master’s

thesis

Name:

Samir Mustafa Negash

Student id: s1280309

Date:

06/11/2018

Supervisor: Natascha van der Zwan

2nd reader: Alexandre Afonso

(2)

2

Following the golden-age of the welfare state (1960-1980), welfare scholarship has described the welfare state as being in a perpetual state of austerity. According to Paul Pierson (1994), the period after 1980 constitutes a period of permanent austerity and is characterized by welfare retrenchment and restructuring. The period of welfare expansion ended in most mature western democracies as countries were confronted with stagflation, i.e., high unemployment and persistent inflation. Structural changes in the economy exert downward pressures on welfare states to reduce their social wage, e.g., by reducing social spending and cutting taxes and social contributions in order to maintain attractive investment climates. As governments are dependent on business and capital to create jobs and generate revenues for taxes, they are receptive to the demands made by capital, albeit their receptivity to these claims comes in varying degrees. The implications for the welfare state are a potential regulatory and welfare ‘race to the bottom’ (Myles & Quadagno, 2002: p. 43).

A range of domestic and international developments have made the financing and delivery of welfare programs more difficult. On the international front, these changes have been a function of changes associated with globalisation. Internationalisation of the economy has pushed states toward fiscal and monetary discipline. This has been the result of increased trade with low wage countries in the global south, which generates inter-industry competition between developed and developing countries and has led to offshoring and deindustrialisation. Increased capital mobility – enabled by the removal and reductions in restrictions of capital flows – has also incentivised states to be more receptive to market forces (Raess & Pontusson, 2015: p. 3; Burgoon, 2001; Autor, Dorn & Hansen, 2015). Additionally, for European welfare states, membership to the European Union (EU) and European Monetary Union (EMU) has functioned as another means of fiscal and monetary constraint on the welfare state. Membership to these institutions has resulted in reduced autonomy of states to conduct social policy by privileging low deficits and low inflation rates over full employment or risk-sharing (Brady and Lee, 2014; Korpi, 2003: p. 604).

Domestically, the shift from industrial to post-industrial economies with high employment rates in the services industry has resulted in lower economic growth rates (Iversen & Cusack, 2000). This shift toward post-industrial economies has also shifted risks across actors in the economy. This shift has complicated welfare provision by creating additional budgetary pressures on the welfare state (Korpi, 2003: p. 590). Demographic changes such as population aging, a change in the overall makeup of families and households, as well as adjustments in the

(4)

4

constellation of the labour market due to the increased participation of women, has increased pressure on welfare states to adapt. This has resulted in an emphasis on social investment, active labour market policies and a focus on capacity enhancing services such as child care and education, rather than providing passive wage replacement. (Busemeyer et al, 2018: p. 802; Esping-Andersen, 1999: p. 182). Long term unemployment or structural unemployment has also become more prevalent due to technological advances. This has also made funding the welfare state more difficult and poses another major challenge for contemporary welfare states (Pierson, 2001b; Schwartz, 2001: p. 19; Korpi, 2003: p. 603).

These developments have constituted pressures on the welfare state as a whole. Their impact has varied across different policy areas. Health care systems, for example, have experienced cost pressures uniquely related to developments in the policy area of health care. As with other components of the welfare state, health care seems to have been subject to the same general pattern: a golden age of expansion (1960-1980) and a subsequent period of stagnation and restructuring after 1980s (Freeman & Moran, 2000: p. 37). Health care cost have been rising in all OECD members, more than doubling in total expenditures since 1970. It is no wonder then that from the 1980s onward, governments in OECD countries began to raise questions regarding the efficiency of their health care systems. Consequently, mature welfare states started examining the expenditures on health care and re-evaluated their health care systems in an effort to enhance efficiency and restrain costs (Mackenbach et al, 2012: pp. 341-342).

This study aims to contribute to the welfare state scholarship by zooming in on developments in on the field of health care. It seeks to gauge how health care systems of 22 industrialised democracies have adapted to these pressures. Using a Large-N design and studying the period 1981-2014, this project studies whether there is a relationship between partisanship and various measures of health care policy output over time. Previous studies on the importance of partisan makeup of governments for the organizing health care systems have produced mixed results (see

references). It builds on these previous works by expanding the scope of research from the

financing of health care (e.g. Jordan, 2011; Potrafke, 2010) to service provision capacity and regulation. The research question guiding this study is thus as follows: What effects have different

partisan makeups of governments had on degree of retrenchment in the field of health care?

To answer this question, the explanatory power of the power resource theory (Korpi, 2006; 1985; Esping-Andersen, 1990) and the new politics of the welfare state theory (Pierson, 2001;

(5)

5

1996; 1994) will be tested. These two theories make competing claims about the role and importance of partisanship, as well as the likelihood of large scale retrenchment. The power resource theory holds that the distribution of power resources matters for welfare expansion. It posits that organizations that help wage earners engage in collective action, e.g. through unionization or translating their demands into policy via voting for left wing parties, are critical in understanding outcomes of distributive conflicts. Left and right wing parties have different preferences and pursue different policies according to the power resource theory. Left parties favour expansion of the welfare state and pursue higher degrees of redistributions. Right wing parties pursue retrenchment and restructuring of the welfare state.

The new politics approach’s claims differ from this, as the role of partisanship is more circumspect according to this theory. Due to differences in the goals and consequences of welfare expansion and retrenchment, the political logic of retrenchment varies substantially from expansion. The behaviour of left and right parties becomes more similar according to the new politics thesis. That is, left wing parties do not or cannot pursue large expansion, right parties by contrast cannot pursue expansion. The new politics school also holds that large scale retrenchment is unlikely due to the welfare state’s staying power. The existence of the welfare state alters the political landscape and creates powerful constituencies that will come to the aid of social programs.

Governments can choose from a range of options in response to cost pressures in health care. The specific responses to these challenges are the often subject of political and distributive conflicts. Health care can be profoundly political as it concerns an object of collective consumption with a large role for the state in all mature welfare state (Moran, 2000: pp. 141-142). The chosen methods may reflect partisan preferences as some previous studies have indicated (Montanari & Nelson, 2013a; 2013b). Overall, governments have to contend with balancing multiple goals in organizing health care and making decisions about cost containment. These goals include ensuring equity and access; having health systems that are efficient and delivering high quality health care services. (Blank et al, 2007: p. 93). These goals are inherently in conflict as it is impossible to deliver on all fronts. The choices made could then, to a certain degree, reflect the values and priorities of the actors involved. This makes them a relevant area of research for politics researchers.

What is unclear from the existing literature on welfare retrenchment and partisanship is whether or not the shifting of costs, as well as the types of administrative changes in health care

(6)

6

systems systematically vary according to partisan makeup of governments and distribution of other power resources. The types of administrative changes pursued affect the degree to which health care services can readily be accessed and the degree to which health care is provided as a social right. Health care constitutes a crucial element of the social right to citizenship in most mature welfare states. This study expands the scope of analysis to include these dimensions, in line with previous research (Montanari & Nelson, 2013a; 2013b; Bambra, 2005). The analysis will examine how partisan makeup of government affects the range of health care resources available in a western democracy. Since the 1980s, the private share of health care financing has increased in all members of the Organisation of Economic Cooperation and Development (OECD). This suggests some degree of retrenchment in health care. The following sections will argue that financing is not the only relevant dimension of health care retrenchment. Instead, financing is but one dimension of health care in which retrenchment can take place. This study hopes to contribute to the welfare scholarship by examining how partisanship affects health care service provision. It does this by applying two theories often used to explain developments in cash benefits schemes to services such as health care which have different underlying logic. In so doing it tries to expand current scholarship on the determinants of public services as well as the politics underlying services such as health care.

This thesis is structured as follows: in the next section the power resource theory is presented and elaborated on. This is followed by a discussion of the new politics thesis. After a discussion of the power resource theory and the new politics thesis, the politics of health care retrenchment are discussed. Special attention is paid to the differences between services and in-kind benefits versus cash benefits. Following this theoretical discussion, the research design is presented. This study uses time-series data on 21 members of the OECD and performs statistical analysis using linear regression with panel-corrected standard errors to answer the research question. After the research design, the results are presented and analysed. The final chapter contains a discussion of the results and a brief conclusion.

(7)

7 2. Theoretical framework

2.1 Partisan politics and the power resource theory

One of the main approaches used to study the development of the welfare state is the power resource theory (Bradley et al, 2003: p. 193). According to scholars in the power resource school of thought, contemporary welfare states are the product of class-based distributional conflicts. Socioeconomic class is related to the types of life course risks individuals are exposed to. As such, the fault lines of distributive conflicts are likely to reflect these class-based differences in risk exposure. Moreover, socioeconomic class shapes the range of power resources individuals have at their disposal. Consequently, researchers in the power resource school have argued that welfare state development mirrors the class-based distributive conflicts and partisan political distribution of power in different countries (Korpi, 2006: p. 168). Therefore, differences in the distribution of power resources as well as the types of cross-class coalitions present in individual countries accounts for the variation in welfare states (Esping-Andersen, 1990, chapter 1).

Power resources reflect the ability of individuals or groups to sanction or reward other actors (Korpi, 1985: p. 33; Korpi, 2006: p. 172). Their relevance to welfare state development stems from the fact that they enable wage earners to equalize or erase their inherently disadvantaged position vis-à-vis employers and business interests e.g. through unionization or organizing politically in class-based political parties (Korpi, 1985: p. 41; Korpi, 2006: p. 171). The difference between wage earners and capital owners’ position stems from the differences between physical capital and human capital, i.e., the specific means of production owned by different groups in society. Employers’ power flows from their ownership of physical capital. Physical capital lends itself to concentration, is scarcer and is easier to divest or apply as it requires low mobilization cost. Wage earners on the other hand tend to rely on human capital rather than physical capital. Labour power is the primary basis of human capital, this is derived from education and occupational skills for example. This power resource is more difficult to mobilise, is less scarce and, importantly, harder to concentrate and difficult to divest from its owner. It is their advantage in numbers that gives their form of power resource its potency. If wage earners do not pool their resources, then employers’ capital derived power resources can be leveraged to bend distributive conflicts to their will (Bradley et al, 2003: p. 197). For these reasons wage earners engage in collective action to utilise their power resources and settle distributive conflicts in their

(8)

8

favour (Korpi, 1985: p. 34).

The power resources theory is distinctly different from other partisan accounts of welfare state development. What the power resource account has in common with other non-power resource-based partisan approaches to studying welfare development is the emphasis they put on distribution of partisan political power. As such, both the power resource theory and partisan approaches of welfare state development essentially hold that parties represent different social constituencies with different needs and interests, that parties’ social profiles mirror the social preferences of their constituencies and that they are policy-seeking as well office-seeking (Schmidt, 2010: p. 213). Where they differ though, is that power resource accounts of welfare development also emphasises other means wage earners have at their disposal to organise collective action, e.g. unionization, electoral turnout. Moreover, regarding the partisan effects, power resource scholars have correctly identified Christian democratic parties as playing a large role in developing welfare states as well. In many western European countries Christian democrats have been successful in mobilising parts of the working and middle classes by catering to them with their distinct type of social policy (Van Kersbergen, 1995). Therefore, researchers in the power resource school not only take into account strength of left parties, but they also take into consideration the strength of Christian democratic parties (Korpi, 2006: p. 196; Huber & Stephens, 1993; Korpi & Palme, 2003; Esping-Andersen, 1990:).

There is, generally speaking, a substantial amount of support for the power resource account in the comparative welfare states literature. Especially in the old politics era (i.e. from the1960s until roughly 1980), a lot of studies confirm that incumbency of left parties is associated with various indicators of welfare expansion (Korpi and Palme, 2003; Allan & Scruggs, 2004;

reference) For example, Huber and Stephens: (2001: p. 41; 78) find that incumbency of left wing

parties and Christian democratic parties lead to distinctly different outcomes with regards to redistribution, development of social expenditures and transfer payments. They argue that incumbency of left parties that facilitate the translation of preferences, mobilization and organization of the working and lower-middle classes generates generous and redistributive welfare states. This account finds support in other studies as well (Bradley et al, 2003). Social democracy with strong unions has been linked to expansions of the public economy by expanding the public sector, greater degrees of decommodification, redistribution and the provision of goods and services as rights of citizenship (Huber, Ragin and Stephens, 1993: p. 716). Whereas rule by

(9)

9

Christian democratic parties tends to result in overall higher expenditures, primarily through higher transfer payments, Christian democratic governments are less likely to lead spend more on social expenditures. Christian democracy is also associated with lower degrees of decommodification, as Christian democrats pursue social policy that reproduces inequality instead of ameliorating or exterminating it, and Christian democrats are also not associated with expansions of public economy (Huber et al, 1993: pp. 717-718). In the same vein, Brady and Lee (2014) trace the causes of cuts or changes in government spending. They find support for the power resources account as the presence of strong trade unions is associated with lower cuts to government spending.

The main takeaway of these studies seem to be that the different social bases parties represent affects the social policy they pursue in government. These parties’ ability to achieve their policy goals is not only determined by their representation in parliament or government, but is also contingent on workers mobilising outside of parliamentary politics. Moreover, changes or shifts in the distribution of power resources can thus be expected to affect distributive conflicts (Korpi, 1985: p. 39).

2.2 The new politics of the welfare state

The claims advanced by the new politics of the welfare state perspective contradict those of the power resource and partisan theories of welfare development. The new politics approach describes the political logic of welfare development in an age of permanent austerity. According to Pierson (1994; 1996; 2001), the politics of welfare retrenchment differ markedly from the politics of welfare expansion. Welfare retrenchment is not simply the mirror image of expansion and the politics of welfare retrenchment has its own distinctive logic. The differences in goals and political context make the new politics fundamentally different from the “old politics” of the expansion period (Pierson, 1996: p. 144). In contrast to could be expected based on the power resource account, increasing power of the right and declining power of the left and unions does not appear to have enabled large scale retrenchment of social programs (Pierson, 1996: p. 150).

Welfare state expansion often entailed the enactment of popular policies, conversely, retrenchment is an exercise in taking away or reducing the generosity of benefits people receive. As such, politicians enacting new social programs can engage in credit claiming, whereas politicians advocating for and implementing restructuring or retrenching the welfare state will often have to practice blame avoidance (Pierson, 1996: p. 145). Therefore, in so far as politicians

(10)

10

seek to cut back on programs, they will shy away from visible cuts and retrenchment is only possible limitedly and they will only enact them in broad based coalitions (Starke, 2006: p. 106).

Enacting cutbacks to social programs is difficult for a variety of reasons: Enacting retrenchment can be challenging because of social psychological reasons, i.e. taking away is more difficult than granting benefits and engenders weaker responses. Consequently, recipients and citizens in general experience cuts more negatively, regardless of how government intends to compensate citizens. Moreover, another factor that further compounds the challenges inherent to welfare retrenchment is the logic of collective action: retrenchment typically imposes high and concentrated cost on a (comparatively) small group of recipients, while promising diffuse and often uncertain benefits on a larger group of people. Lastly, policy feedback and policy legacies are additional factors that complicate the process of retrenchment. These factors are more institutional in nature. They have bearing on the ways in which the welfare state has become part of the political landscape and how this has altered the basic political calculus of actors in the welfare state. The welfare state’s existence led to the formation of new interest groups and entrenched constituencies who rely on the welfare state, e.g. pensioner’s rights groups and patients’ rights groups. These groups have become the greatest proponents and protectors of welfare programs. In so doing they have replaced trade unions as the interest group of greatest significance, as was the case in the power resource account. Moreover, welfare retrenchment is also likely to engender opposition from existing clientele due to the double payment problem and constrain government’s capacity to change policy due to lock-in effects and pre-commitments and associated problems such as double payment. Moreover, formal institutions such as veto players in the system also make enacting retrenchment more difficult (Pierson, 2001: p. 411-414; Levy, 2010: pp. 554-555).

The implications of the new politics for the relationship between partisanship and welfare state development are clear: in the new politics era, partisanship is likely to matter less as the welfare state has changed actors’ political calculus. For left parties it means they cannot, or are restricted, in their pursuit of their preferred options of expanding existing programs or enacting new programs. This is largely a consequence of the structural changes to the economy that have created budgetary pressures on welfare states. For right parties this means that they are restricted in advancing their preferred policy agenda of welfare retrenchment. As radical retrenchment is not possible and the logic of the new politics pushes them towards maintaining programs, they are

(11)

11

forced to obfuscate their goals or trim social programs around the edges. Therefore, the new politics thesis would lead to the expectation that left and right parties behave more similarly with regards to welfare development. It should be noted here that the new politics account provides explanations for the behaviour of left and right parties. It is worth noting here that this approach does not provide an account of Christian democrats’ preferences. It unclear whether Christian democrats would pursue retrenchment and how they would rearrange risks.

Scholars have sought to corroborate the claims made by the new politics theory of welfare development with overall mixed results. By and large Kittel and Obinger (2003: pp. 33-34) find support for a declining importance of party politics on welfare state development in the new politics era. They conclude that partisan effect decreased overt time, and stress that the importance of partisanship for welfare development declined after 1980s. They claim that a particular type of catch-up dynamic is unfolding. Generous states are in the process of retrenching or scaling down their welfare state, while less generous welfare states on the other hand have expanded. As a results welfare states are converging. Furthermore, they attribute much of the changes in welfare development to the increasing dependency ratio due to aging and rising unemployment. Kwon and Pontusson (2010) also argue that partisan differences seem to have disappeared in how governments compensate domestic workers with welfare in response to globalisation. They argue however that the rate of decline seems to be mediated by the presence of strong labour unions. Brooks and Manza (2006: p. 822) provide additional support for the new politics approach. They argue that welfare state resilience is explained by popular support for the welfare state. This seems to be unrelated to distribution of power resources. They also find no significant association between left power or the continued cabinet participation of left parties and welfare state development in new politics era.

There are studies that find continued support for the power resource account, even in the new politics age. For example, Allan and Scruggs (2004) test the implications of both the power resource theory and new politics thesis and find that in the new politics age, right-wing parties continue to be associated with retrenchment. This seems to disprove the new politics’ account, according to which right wing parties are constrained in pursuing their preferred goal of retrenchment. Korpi and Palme (2003: p. 41) also find that compared to secular and conservative right parties, left-wing parties are more likely to increase, or at least, less likely to reduce the replacement rates of unemployment insurance and sick pay schemes. Swank (2005: p. 191) also

(12)

12

finds that left parties are significantly and positively related to expansions of unemployment insurance, pension and sick pay replacement rates. Her analysis includes data between 1980 and 2010, and the focus of the analysis is thus new politics period.

2.3 Health care retrenchment in a social rights perspective

Thus far the discussion has centred on the potential effects of partisanship on the development of the overall welfare state. Authors across multiple disciplines, e.g. economics, sociology and political science, have honed in on and examined the relationship between partisanship and retrenchment in the domain of health care (see for example Jordan, 2011; Montanari & Nelson, 2013a; Potrafke, 2010; Herwartz & Theilen, 2010; Herwartz, & Theilen, 2014; Béland, 2010; Freeman & Moran, 2000; Fervers, Oster & Picot, 2016; Pavolini & Österle, 2013; Companje, Veraghtert & Widdershoven, 2009). What stands out is the diversity in theoretical perspectives employed, the methods used and the variables of interest in these studies.

The causes attributed to the rising cost of health care are varied and are the result of multiple mechanisms. The primary factors commonly cited as explaining increased health spending are: First of all, increases in per capita gross domestic product (GDP) arguably lead to increases in health care expenses. GDP per capita has long been a powerful explanatory variable in explaining rises in aggregate health care expenditures, as health care appears to be a (somewhat) elastic good (Aron-Dine et al, 2013). This means that consumption of health care increases as national income increases. This increased consumption will then naturally be reflected in increases in aggregate spending levels in health care (Hacker, 2004a: p. 699). Secondly, another commonly cited factor explaining the continued increases in healthcare expenses is Baumol’s disease. Baumol’s disease refers to the limited room for productivity growth in the public sector. The incomes of employees in the services sector have to keep pace with the incomes of employees in the non-service sector. Over time, this combination of stagnant productivity and rising wages leads to higher overall costs in the services sector. Given that health care is a policy field where services make up a large part of the output, stagnant productivity will thus likely be reflected in higher expenditures in health care (Baumol, 1967: p. 417). Thirdly, advancements in medical technology and medicine are another commonly cited factor that could account for rises in health care spending. These new technologies can improve health, but often come at high financial cost. Once these technologies are developed and available, governments are loathe to deny their populations

(13)

13

access to these technologies (Barr, 2010: p. 252; Giaimo, 2001: p. 336). Lastly, increased volume in health care e.g. due to aging population and new treatments being available are other factors that contribute to the increases in overall healthcare expenditures (Mackenbach, Stronks & Anema, 2012: pp. 343-344). Increases in overall healthcare expenditures are not necessarily problematic. The increases in healthcare expenditures could simply reflect overall taste for healthcare and could therefore constitute an entirely legitimate endeavour for governments to spend their money. Increased healthcare expenditures could also reflect increased need for and use of healthcare services due to demographic changes. Increased use of health services shows need for these services and also shows why cutting back on them or otherwise shifting costs could be unpopular (Freeman & Moran, 2000: p. 39).

Increasing expenditures on health care could nevertheless still be problematic in so far as increased health care expenditures also constitute an opportunity cost; governments can invest a euro only once. Spending more on health care essentially walls off, that is, forecloses expenditures on other facets of the economy or the welfare state. Moreover, the ever-increasing expenses on health care could have an impact on the overall economy as well. The effect of increases in health care cost – in instances where employers cover the costs – are potential lower rates of investments. The effects of rising health expenses in systems where employees bear the burden are potentially higher marginal tax rates which act as disincentives for labour (Mackenbach et al, 2012: p. p. 343). This argument finds at least some empirical support, for example by Fervers, Oser and Picot (2016) who extend Burgoon’s (2001) argument that various kinds of trade openness have dissimilar effects in different policy areas. They examine a range of indicators for globalisation and public share of health spending and conclude that increased openness does seem to lead to lower growth in public health care expenses. This indicates that governments are under pressure to reduce public health care spending to remain competitive in the face of increased international completion.

There are good reasons to assume that efforts at retrenchment in health care will differ from efforts in other domains of social policy. Government’s health care cost containment strategies can target the supply side or the demand side. Supply side cost containment strategies could be agreements with preferred providers, restricting the number of doctors by putting restricting access to the medical profession. There are also a number of demand side cost containment strategies. These measures are aimed at making users of health care more conscious of the cost of health care and disincentivizing use via co-payments or deductibles, by making them pay for a share of the

(14)

14

costs. Demand side cost containment can also be achieved through achieving slower rates of rises, stopping cost rises in real terms or in rare instances reducing real costs (Blank, Burau and Kuhlmann, 2007: p. 97). Governments will thus tend to shy away from achieving cost-cutting via imposing costs on health care users. This is also apparent as in practice governments try to pursue cost containment through stimulating competition, putting restrictions on providers and other administrative changes. (Freeman & Moran, 2010: pp. 40-42).

Multiple authors have conducted Small-N comparative studies to research various episodes of health care reform pursued by governments, and explore the politics underlying these reforms (e.g. Hann, 2007; Gingrich, 2011; Giaimo, 2001). Hacker (2004a) for example examines recent health care reforms in a couple of wealthy democracies (US, UK, Netherlands, Canada and Germany). He argues that reform processes have been marked by change without reform and reform without change. Toth (2010) looked at different waves of reform in health care in multiple countries. He finds that each wave of reform is followed by a wave of counter reforms where the effects of the previous reform are (partially) reversed. Interestingly, he notes that ideology seems to matter in dynamic of reform and counter reform. The starting point of these studies is often the cost pressures countries face to contain the growing costs of health care. They, however, rarely examine explicitly the role of partisanship on attempts at implementing cost containment policies, or the degree to which choice of cost containment strategy is tied to partisan and ideological differences. So, while they contribute valuable information on the intentions and preferences of actors in reform episodes, these studies do not allow for systematically quantifying the effects of left, right or Christian democratic parties on health care.

In addition to the case study approach, numerous studies have attempted to map the long term patterns in the development of health care expenses and the extent of retrenchment in health care. These studies have also researched the relationship between partisan distribution of power and changes in health care. Studies of retrenchment politics and welfare retrenchment have often looked at cash benefits programs such as pensions or unemployment insurance (e.g. Esping-Andersen, 1990; Korpi & Palme, 2003; or Allan & Scruggs, 2004). Studies of health care politics have often emphasized the financing aspect by using the public share of health care expenditures as the dependent variable. For example, in mapping health care retrenchment, Jordan (2011) tries to gauge the effects of partisanship on health care retrenchment. His findings suggest that after 1980s, left and right parties do not differ significantly from each other in regards to their effects

(15)

15

on the development of the public share of health spending.

In a similar vein, other authors have examined the relationship between partisan makeup of governments and overall developments of the public share of health spending. These studies have tended to produce mixed results as well with regards to the relationship between partisanship and the public share of healthcare. Some find that while the private share of health spending seems to be increasing in OECD countries (and therefore the public share has to be decreasing), there does not seem to be an ideological basis for this, i.e. left and right parties seem to behave broadly similarly, which would seem to disconfirm the partisan accounts (Fervers et al., 2016; Potrafke, 2010).

However, some authors (e.g. Huber and Stephens, 2001: p. 270; Herwartz, & Theilen, 2014) have found evidence of a relationship between the partisan makeup of government and health care expenditures. The differences in results could in part be explained by differences in how partisanship is operationalized in those studies and different statistical methods applied. Potrafke (2010) for example uses government ideology as an explanatory, a variable which ranges from 1-5 based on the degree of dominance of right wing parties in government. This allows for identifying differences between left and right parties, but neglects the role of Christian democrats in accounting for developments of public health care expenditures. Huber and Stephens (2001: p. 53) on the other hand use cabinet portfolio of left and Christian democratic parties as an explanatory variable. This has the upside of including Christian democrats, however, their approach also fails to include right parties. This is at least questionable; right wing parties are often the actors most likely to pursue retrenchment. Excluding right parties from the analysis could potentially understate the role of partisanship (Alan and Scruggs, 2004: p. 504).

These studies have a number of conceptual limitations. The majority of the studies reviewed thus far have analysed retrenchment in health care by looking at the shifts in financial burdens. That is, they have looked at how the public share of health care spending has changed over time. However, health care as an area of social policy differs significantly from other cash benefits social programs. The numerous actors involved, the set of regulations in place regarding remuneration of providers and coverage, the range of services covered by different insurance plans in addition to the types of health care resources available in a country, are all decisions that could affect the degree and effects of retrenchment in health care. These are all changes that can be relatively independent of the exact financing method, i.e. private or public. Alterations in who pays

(16)

16

for what services are therefore not the only relevant changes governments make with regards to health care. This follows from the fact that healthcare is primarily about services, and not about providing income replacement. This calls for a different approach to studying retrenchment in healthcare. One that takes a broader view of healthcare and does not relegate itself to looking at financing.

There are studies that have reckoned with the difference between cash benefits and services. In this regard, Bambra’s (2005) approach to the study of health care politics tries to elaborate on this distinction. Bambra argues that welfare states differ with regard to how much value they place on service provision or cash benefits. Some welfare states, notably those that belong to the social democratic welfare regime, spend more on services and cash benefits, whereas liberal and conservatives spend comparatively little on services. Unlike other authors (Esping-Andersen, 1990; Ferreira, 1996) who have classified welfare states based on cash benefits schemes, she classifies welfare states based on the types of in-kind benefits they provide. Taking a more indirect approach, some authors have also linked the specific political traditions of countries to differences in health outcomes. In this line of research, the argument goes as follows: Belonging to a specific welfare regime implies presence of a particular political tradition, i.e. social democratic, Christian democratic and liberal traditions. These political traditions in turn structure social outcomes through the types of welfare arrangements that are likely to be in place. Lastly, these social outcomes affect health outcomes (Hurrelmann et al., 2011; Eikemo et al., 2008).

This is something Navarro and Shi (2001) argue as well. Examining the significance of partisanship and policies implemented by parties in government on inequality, and by extension their [political parties’] impact on health inequalities. They conclude that, political parties in government affect levels of inequality though redistributive policies they implement. This in turn affects concrete health outcomes such as infant mortality. They find that social democrats are both most inclined to redistribute and also associated with the highest scores for health outcome indicators. Furthermore, the Christian democratic and liberal tradition also differ with regards to how much they redistribute and the central institutions for delivering and organising health care (Navarro & Shi, 2001). These studies thus suggest that politics seems to matter not only for expenditure levels. Instead, politics also has implications for how services are organised and these choices result in different outcomes. There are thus also qualitative differences between how welfare states organise health care. This preoccupation with qualitative differences mirrors that of

(17)

17

Esping-Andersen (1990, chapter 2), Korpi and Palme (2003), Korpi (1989; 1998) and others in the power resource school who also emphasise the social rights accorded by different welfare states.

These qualitative differences have been studied using a social rights perspective. Social citizenship refers to the rights and duties of individuals in society to benefits and services that allow them to meet social needs enhance their capabilities and guarantee the resources necessary to finance them. Social citizenship is based on three values, namely: reciprocity, inclusion and trust (Taylor-Gooby, 2008: p. 5). Social rights are a critical component of citizenship in most welfare states. As such, social rights are accorded to individuals as part of their rights as citizens. Social rights can, however, also be based on employment as is often the case for example with schemes such as unemployment insurance (Korpi, 1989: p. 310). This is the case with Esping-Andersen’s conception (1990). According to him benefits are provided as social rights, i.e. decommodifying, when they at least allow for an individual to exit the market freely and not experience large drop-off in quality of living (Stephens, 2010: p. 512).

Health care can be commodified as well. Social citizenship in health care includes an entitlement to health care and is fundamentally aimed at eliminating this commodification (Bambra, Fox and Scott-Samuel, 2006: p. 188). It is conceivable that the forces exerting downward pressure on welfare states, also result in a reduction in the degree to which health care can be provided as a social right. To empirically verify this claim, a number of aspects of health care systems can be studied. Health care systems can be classified along three dimensions: service provision, financing and regulation. Service provision entails the main health services, i.e. inpatient and outpatient care, dental care and provision of pharmaceuticals. Financing pertains to method of financing health care, e.g. social insurance, taxes or private and out of pocket expenses. Lastly, regulation pertains to coverage, access of users to provider’s services and remuneration of providers (Böhm et al., 2013: p. 260). These three dimensions can be used to gauge the degree to which health care is commodified, or decommodified.

Some researchers have tried to map these qualitative differences and tried to find distinctions in health care policy output. Montanari and Nelson (2013a), for instance, have examined the role of partisanship in shaping healthcare reform. Approaching the topic via a social citizenship perspective, they focused on service provision. Their approach differs from Bambra’s as they argue that health care can be provided as a social right when critical health care resources – physicians, beds and medical technologies – are available. Bambra (2005: p. 202) argues that

(18)

18

commodification and decommodification in health care is a function of whether services are privately or publicly provided, and whether provision is privately or publicly organized. Publicness indicates collective risk sharing. Bambra equates this with smaller probabilities of exclusion or being denied important services which could be more prevalent in private schemes. But this is at odds with studies that find that retrenchment and restructuring is often more probable and severe in national health systems (NHS) (Giaimo, 2001: pp. 339-340). This effect is likely mediated by private or public financing. Therefore, the difference between private and public coverage is a fruitful way to classify health care systems. It is not the most suitable way to study retrenchment though. Because health care is fundamentally about service provision. Whether services are organised publicly or privately, whether costs are public or private matters of course. However, they need not have any bearing on the degree to which health care can be provided as a social right. Accounting for service provision in research is thus a more promising track.

2.4 Hypotheses

Having argued that studies of retrenchment in health care should incorporate more facets of health care than financing, the specific expectations to be tested are more ambiguous. There are theoretically grounded reasons for why partisanship could matter in health care. What is apparent, though, is that health care is a policy area where partisanship may matter, but in more constrained ways than in other policy areas. Jensen (2011: p. 913) provides a theoretical account of the different preferences of left and right wing parties in health care. He posits that health care politics differ from the politics of other areas of social policy, which have a more direct link to the labour market. This is a function of generic social risks being correlated with income, whereas health risks are only weakly correlated with income, and instead, mainly related to life-course risks. Furthermore, right wing parties practice a form of marketization via compensation. This essentially means that they will – for electoral reasons – spend roughly as much as the left. Right parties will behave this way in order to avoid alienating middle class voters who would otherwise be harmed by pure private provision and financing of health care. However, they will do their best to ensure private or market based health provision, i.e. through supplementary systems of health care. This has two advantages: private provision is likely to be used by high incomes and second may instil in the population a form of market values and loyalties (in some the opposite of broad solidarities generated by universal programs (Jensen, 2011: p. 909). Therefore, political contestation over

(19)

19

healthcare is going to differ to a degree from other policy areas. It will not be based entirely on the distribution of risks.

Gingrich’s (2011) provides a comparable account of policy preferences of left, right and confessional party’s. According to her, left and right wing parties organise markets for public services differently in ways that reflect their own relative ideological predispositions. Accordingly, left parties are more likely to favour linking the fates and risks of individuals in different classes, so as to engender collective solidarity. Right parties, in contrast, favour more private involvement and greater competition, less generous schemes and are more likely to pursue reforms that fragment risks (Gingrich, 2011: pp. 37-38). These changes are often independent of private or

public financing.

Therefore, left wing parties will likely support public provision of health care, public funding of health care and universal and broad coverage. Confessional and right parties will likely be more willing to have a larger share of health care financing privately taken care of, be more willing to introduce and tolerate supplemental systems and more likely to fragment risks where possible. To test for the results of these different preferences on health care, the impact of partisan makeup of government on the three dimensions of health care systems will be tested Based on the following reading, the following hypotheses have been formulated:

The first hypothesis tests the dimension of health care where political contestation should be most similar to that of other social programs. That is because the financing of health care is fundamentally also about how risks are shared. To test for the differences between left, confessional and secular right wing parties on health care financing, the first hypothesis is as follows:

H1a: Left party cabinet seat shares is associated with increases in the public share of HCE. H1b: Confessional party cabinet seat shares is associated with smaller increases in the public share of HCE.

H1c: Secular and conservative right party cabinet seat shares are associated with stabilisation or small increases in the public share of HCE.

Hypothesis two and three will be used to test for the effects of the partisan makeup of government on the availability of health care resources, which are critical for ensuring the social right to health care. The health care resources examined in this study are hospital beds and physician density.

(20)

20

Taking Montanari and Nelson (2013a; 2013b) as starting point, and considering what previous studies have shown about health care regimes, the following hypotheses has been formulated:

H2a: Left and Confessional party participation in cabinets are associated with increases in the number health care resources available.

H2b: Secular and conservative right party cabinet seat share is associated with weaker growth, or even declines in the number of health care resources available.

H 3a: Left parties cabinet seat shares is most strongly associated with increases in employment in health care sector.

H3b: Confessional party cabinet seat share is less so associated with increases in employment in health care sector.

H3c: Secular and conservative right parties are associated with the smallest increase – or potentially even with decreases – in employment in the health care sector.

Hypothesis four tests the effects of partisanship on regulation of access to health care. It extends the argument from hypothesis one. Specifically assuming that different parties have different degrees of preference and commitment vis-à-vis types of health insurance coverage. Even if parties do not differ on how they finance health care, they should care whether or not health insurance is privately organized or publicly organized. As such, hypothesis 4 is as follows:

H4a: Left party cabinet seat share is associated with the largest share of the population covered and the largest share covered in public schemes.

H4b: Confessional party cabinet shares are associated with smaller shares of the population covered and smaller shares covered in public schemes.

This is expected to be a result of Christian democrats’ commitment to the principle of subsidiarity and their preference for the social insurance model (Van Kersbergen, 1995: p. 2).

H4c: Secular and conservative right parties are associated with the smallest shares of the population covered and coverage of a smaller share of the population in public schemes.

Hypothesis 5 tests a claim advanced by the new politics theory. It tests whether or not path dependency explains developments in health care. One of the takeaways from the new politics thesis is that welfare programs are resistant to change and that they are path dependent. The fifth hypothesis is thus as follows:

H5a: The value of public share of health care financing in 1981 is associated with public share of health care financing in the period analysed.

(21)

21 H5b: The value of total health care expenditure in 1981 is associated with public share of health care financing in the period analysed.

H5c: The values of health care resources in 1981 are associated with health care resources in the period analysed.

H5d: The value of coverage in 1981 is associated with coverage in the period analysed.

In accordance with the power resource theory, labour unions are assumed to pursue similar outcomes as left parties. Hence the expected association between trade union density and the various outcome variables is in the same direction as with left parties.

3. Methodology

The research design used for this study will be a Large-N statistical analysis. The aim of this study is to test whether the power resource theory or the new politics theory hold up as explanations for the developments in the health care system of various OECD members. The goal is thus explanatory in nature. Additionally, this study tries to replicate and test claims advanced by previous studies for robustness (Leavy, 2017: p. 89). The politics of health care can be studied using multiple designs, such as Small-N research designs, e.g. by performing single case studies

Table 1: Summary of hypotheses

Variable Expected association

Hypothesis 1

Left cabinet share

Christian democratic cabinet share

+ +

Right cabinet share 0/+

Hypothesis 2

Left cabinet share +

Christian democrat 0/+

Right cabinet share -/0

Hypothesis 3

Left cabinet share

Christian democrat cabinet share

+ +

Right cabinet share -/0

Hypothesis 4

Left cabinet share +

Christian democratic cabinet share 0/+

Right cabinet share 0

Hypothesis 5

(22)

22

or comparative case study designs. These designs lend themselves to accounting for particular outcomes, or identifying the motivations of actors involved in the process. For establishing general patterns, however, Large-N designs are more suitable. This applies to the current study, as the goal is to gauge whether there is a relationship between partisanship and organization of health care. To accomplish this, this study compares of a large number of cases on a limited number of relevant variables across a sufficiently large number of years (Toshkov, 2016: p. 256; Huber & Stephens, 2001: p. 37).

This thesis will analyse pooled data, that is, time-series-cross-sectional (TSCS) data. It includes 33 time periods and 21 units, which yields (in the most optimal scenario) 735 observed values per variable. The specific statistical analysis technique used is linear regression. TSCS have observations on set number of units, in these studies countries are units, for multiple time periods. Ordinary least squares (OLS) models cannot be carried out, or at least yield inaccurate estimates, without regard for the fact that time-series data in TSCS likely violates core assumptions of the OLS models. In particular, the assumptions of homoscedasticity and an absence of correlation of standard errors in units at different time periods is potentially violated as observations are rarely independent of each other (Beck and Katz, 1995: pp. 534-636). The use of panel corrected standard errors (PCSE) helps to account for the heteroscedasticity and autocorrelation in standard errors (Plümper, Troeger & Manow, 2005: p. 330). Using PCSE with a lagged dependent variable can counteract these deficiencies of TSCS. However, the inclusion of the lagged dependent variable as an explanatory variable can also be problematic, as this could artificially reduce the explanatory power of other independent variables. So all the tests will also be conducted by differencing the dependent variables. These additional tests serve as a robustness check. (Janoski & Isaac, 1993: p. 33).

The research design applied in this study has the advantage of being highly reliable. The variables constructed and results produced using this data can be replicated by other scholars. The data used for this study is freely available from authoritative organizations such as the OECD. Regarding validity, the results are only generalizable to a set of industrialised democratic countries. So this study’s external validity is limited to the set of countries studied and other similar countries. Another limitation of TSCS includes the somewhat uncertain validity of inferences across cases. Even if a particular variable produces an effect in one unit, this does not necessarily mean it does

(23)

23

so for other units. This is an innate feature of TCSC and has to be considered when interpreting results.

3.1 Case selection and operationalization

Cases have been selected for their theoretical relevance and availability of data to conduct analysis. The analysis includes 21 members of the OECD from 1981 until 2014. This is an appropriate starting point as the 1970s are often considered the end of the old politics era, signalling the paradigm shift to the new politics era. The following countries included in the analysis: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom and the United States. These countries are highly comparable on the theoretically relevant dimensions. This means that all of these countries are industrial welfare states, and all countries were democracies in the period under analysis. This makes the partisan accounts potentially relevant. Some scholars prefer to exclude countries such as Portugal, Greece and Spain from the analysis; they are not mature democracies and thus the dynamics of democratic politics may differ in those countries (Huber and Stephens, 2001: p. 33). It is conceivable the shift to the new politics context did not arrive in these countries at the same period in time it did for other OECD countries. As such, these cases may inflate the effect of the power resource account. A number of other studies include these countries however. Therefore, in the interest of comparability, the analysis conducted includes Spain, Greece and Portugal. The tests will also be run excluding Spain, Portugal and Greece to see if excluding these countries changes the results substantially.

3.2 Dependent variable

In debates about welfare retrenchment or resilience, the exact operationalization of the dependent variable used could affect the results. Using aggregate or per capita expenditure data, using replacement rates or net replacement rates yields different results (Allan and Scruggs, 2004; Green-Pedersen, 2004; Esping-Andersen, 1990; Starke, 2006: p. 112). An examination of the dependent variable used in studies of retrenchment of health care shows numerous authors using public share of health care as dependent variable (Jordan, 2011; Potrafke, 2010). The public share does not incorporate important features such as tax expenditures, subsidies or other more nuanced differences between health systems. It does, however, say something about how health care is

(24)

24

organized, how costs and risk are distributed and shared (Brady et al, 2016: pp. 356-357). Therefore, to gauge retrenchment in the financing of health care, the public share of health care expenditures is used.

It is worth noting that retrenchment in these studies does not mean reductions in total health care spending. Retrenchment reflects a decline in the public share of total healthcare spending, or a decline in the public share of health spending as a percentage of GDP. This difference between the specific ways of measuring it could have consequences: when the public share is expressed as a share of GDP, the public share of health care may decline or increase irrespective of changes to health care expenditures. This would not be the case when public health care expenses are expressed as a share of overall health care expenses. Therefore, to gauge retrenchment in the financing of health care, the public share of health care expenditures is used for this study. As health care financing is but one dimension that characterizes the performance of health care systems, the other two dimensions, regulation and health care provision, need to be accounted for as well. Montanari & Nelson (2013a: pp. 261-262) operationalized health care provision as an additive index, which is comprised of: health employment (number of nurses and physicians per 1000 inhabitants), hospital beds and medical technology (additive index of five categories of medical technology per 1000 inhabitants). For this study, health care resources are operationalized in the following two things: first, the number of physicians (additive index of nurses and physicians per 1.000 residents). Second, the number of hospital bed. This is also an additive index comprising the number curative and hospital beds per 1.000 residents. The variable health employment is also included to gauge the importance of health care and social employment in the economy. This variable is expressed as health care and social employment as a share of overall civilian employment (OECD, 2016). Lastly, for regulation, this study includes an additive index of

coverage in public scheme and overall coverage. The score ranges from 0 to 200. This variable is

meant to give an indication of how accessible the health care system is. As is customary, this study includes lagged values of all variables to account for time lag between changes in partisan

distribution of power and policy responses (Fervers et al., 2016: p. 365).

3.3 Independent variable

For the main independent variable of interest, partisan power, previous studies have used multiple approaches to quantifying this variable. Some have included the cabinet composition by looking

(25)

25

at cabinet seat shares of left, right and Christian democratic parties. This approach is an appropriate way of studying power resources. This study will use the cabinet composition as the main measure of partisanship. However, it is important to note that there are other approaches to operationalizing this variable. Another method of operationalizing cabinet composition is the

Schmidt index for government ideology. It is a five-point scale indicating dominance of right and

centre parties in government. The value 1 indicates hegemony, 2 indicates dominance of the right, 3 indicates parity, 4 dominance of the left and 5 hegemony of the left. A notable drawback of the Schmidt index is that it does not include an independent measure of Christian democrats in government. The assumption underlying the Schmidt index seems to be that the difference between left and right wing parties is the only relevant one. This makes the Schmidt index an unsatisfactory indicator for this study. However, a number of other studies into health care retrenchment use the Schmidt index. In the interest of comparability, the tests will also be run using Schmidt index to test whether this significantly alters the estimation of partisan effects.

The other main power resource independent variables used will be unionization rate. This is a superior and more relevant approximation of union strength than other measures such as collective bargaining coverage rates, or wage bargaining centralization. It directly relates to union’s strength at that moment. Health insurance is an important fringe benefit in some countries. However, this is often not the case. Using unionization rate should capture the effects of union strength for countries where health care is a fringe benefit and where health insurance is provided by the government. The per capita GDP growth rate is also included to account for health care spending being driven by increases in income and improvements in the overall economy. The net

government deficit is included to account for budgetary pressures which might drive retrenchment.

Total social expenditure is included to capture whether or not potential retrenchment in health care is compensated by increasing social expenditures in other programs. The dependency ratio is included to control for the effects of elderly or underage population as explanatory factors of budgetary pressure. The KOF globalisation index is also included in the analyses to assess whether increased openness and international competition disciplines welfare states to pursue retrenchment in health care.

(26)

26

Additionally, this study has included a dummy variable for single party government (1 if single party majority government, 0 for all other types of governments). This variable is included to account for the fact that single party governments face fewer restraints in pursuing their ideal policy goals than coalition or minority governments do. They do not have to rely on coalition partners or others in parliament to legislate. I have also included an interaction term between single party government and right wing cabinet share, this variable accounts for single party governments formed by the right wing. Additionally, constitutional constraints are included as well. This variable is included to assess whether states with more constraints in policymaking and veto players present in the system exhibit some status quo bias. It hampered expansion in the old politics era, and halted retrenchment in the new politics era (Huber, Ragin and Stephens, 2001: p. 32). The ability of governments to unilaterally impose cost containment is lower in social and private health insurance systems due to various factors, e.g. corporatism and the role of social and private actors in regulating access and provision (Fervers et al, 2016: p. 202; Böhm et al, 2013). Whereas in National Health Service systems, the capacity for governments to act is larger and they can more directly enact cost containment measures. To account for this, this study uses state run health care

system as an additional control variable. The expectation is that countries with state run health care

systems will be successful in imposing cost containments and reducing the availability of health care resources. Lastly, the values of the dependent variable at 1981 are included, in order to assess whether path dependency accounts for outcomes. This test is meant to rule out whether the arrangements in 1981 explain the outcome in 2014.

Table 2: Summary of association with non-power resource independent variables

Control variable Expected association

Social expenditure -

Deficit -

Dependency ratio +

GDP growth +

Single government -/+

Right wing single party government -

Globalization -

Centralization of health care system -

(27)

27 Table 3: Overview of variables used

Operationalization Source

Dependent variable

Health care financing Share of total health care expenditures that is publicly financed OECD (2017c)

Total health care expenditure as a percentage of GDP

Health care regulation Additive index combining share covered by the largest public

scheme and overall share covered with health insurance

OECD (2017c)

Health care resources Hospital beds per 1.000 inhabitants combined with curative care beds

per 1.000 inhabitants

OECD (2017c)

Physicians per 1.000 inhabitants combined with nurses per 1.000 inhabitants

Health and social service employment as a share of civilian employment

Independent variable

Left cabinet Share of cabinet seats held by social democratic parties and parties

to their left

Swank (2013)

Right cabinet Share of cabinet seats held by secular and conservative right wing

parties

Swank (2013)

Christian democratic cabinet

Share of cabinet seats held by Christian democratic parties Swank (2013)

Schmidt index Cabinet composition: (1) hegemony of right-wing (and centre)

parties, (2) dominance of right-wing (and centre) parties, (3) balance of power between left and right, (4) dominance of democratic and other left par-ties, (5) hegemony of social-democratic and other left parties.

Armingeon et al (2017)

Union density Net union membership as a proportion of wage and salary earners in

employment.

Visser (2016)

GDP growth Growth of real GDP, percent change from previous year OECD (2017b)

Dependency ratio Age dependency ratio is the ratio of dependents–people younger than

15 or older than 64–to the working-age population–those ages 15-64.

Jan et al (2018)

Social security expenditures Total public and mandatory private social expenditure as a percentage of GDP.

OECD (2017a)

Government deficit Annual deficit (overall balance / net lending of general government)

as a percent-age of GDP.

(28)

28

KOF index KOF de facto economic globalisation index. This index is

constructed by calculating and weighing trade and financial globalisation. Trade globalisation consists of trade in goods, services and diversification of trade partners. Financial globalisation consists of FDI, portfolio investments, capital investments, international debt, international reserves and international income payments.

Dreher, 2017

Constitutional Constraint Constitutional constraint is an additive index that measures the

presence of institutional constraints of the central government. The values range from 0 to 6, although the data only contains scores from 0 to 5. Higher values indicate more institutional constraints on central governments and lower values suggest fewer institutional constraints. The scores consist of an additive index. This additive index is composed of 6 dummy variables (‘1’ = constraints, ‘0’ = else): (1) EU membership = 1, (2) degree of centralisation of state structure (federalism = 1), (3) difficulty of amending constitutions (very difficult = 1) (4) strong bicameralism = 1 (5) central bank autonomy = 1 (6) frequent referenda =1.

Armingeon et al (2017)

State run health care system Dummy variable with values 1 or 0. The variable is coded 1 if the

health care system is ‘state run’, 0 otherwise. Health care systems are classified as state run if the state was the dominant actor governing and regulating relationships between actors. The following countries were classified as state run: Australia, Canada, Denmark, Finland, Ireland, Italy, New Zealand, Norway, Portugal, Spain, Sweden, and the United Kingdom).

Bohm et al (2013: pp. 264-265)

The politics of health care: examining the role of partisanship in health care