Adelheid Wölfl
Democratic governance and health A quantitative comparison between groups of 44 European countries
Date of submission: 1.09.2003 Master of Comparative European Social Studies
"This dissertation is the work of 2025433 and has been completed solely in fulfillment of a dissertation for the MA in C.E. social studies at the Hoogeschool Maastricht."
Acknowledgements
This thesis was written thanks to the generousness of my supervisor Professor Jordi Sancho Salido. I am deeply grateful for his time and patience, and his interest and encouragements. Many thanks to my parents for their financial support and their confidence. Finally I would like to express my gratefulness to Tom who supported me with endurance, flexibility and tenderness.
Abstract 1
1.1 1.2
1 3
3 5
Aims of the research
The hypothesis of the study The quantitative comparison
2
2.1 2.2
The theoretical framework
The general theory – Democratization and Socioeconomic development Relevant literature and theories concerning the relationship between ‘health outcomes’ and ‘democracy, political participation and good governance’ The two concepts of the study: "health" and "democratic governance"– a theoretical conceptualisation The concept of "health" The concept of "democratic governance" Measuring health and democratic governance Measuring health Measuring democratic governance Definition of the variables Definitions of the variables of the "health" concept Definitions of the variables of the "democratic governance" concept
8
8
12 17 18 20 23 24 25 28 28 33
2.3 2.3.1 2.3.2 2.4 2.4.1 2.4.2 2.5 2.5.1 2.5.2
3
3.1 3.2 3.3 3.4
Design of the test procedure
Compressing the variables of the health concept and group building Analyzing the differences between the groups of countries regarding economic strength Compressing the variables of the democratic governance concept Proving the influence of the factor variables of the democratic governance concept on the factor variables of the health concept
37
37 37 38 38
3.5
Measuring the relation between the health concept variables and the democratic governance variables along the group of the principal components analysis by controlling the influence of economic strength
39
4
4.1 4.2 4.3 4.4
Methodology of the quantitative tests
The "Principal components analysis" Student's T-Test The regression The partial correlation
41
41 42 42 42
5
5.1 5.1.1 5.1.2 5.2 5.3
Building groups of countries
Analysis of the principal components analysis of the health concept Description of the two components The definition of the groups Analysis of the T-test concerning the differences between the groups of countries in regard to the health concept variables Analysis of the T-test concerning the differences between the groups of countries in regard to economic strength
45
45 46 47 73 76
6
Testing the relation between the health concept and the democratic governance concept
Analysis of the principal components analysis of the democratic governance concept – Compressing the variables of the democratic governance concept Analysis of the regression – Proving the influence of the factor variables of the democratic governance concept on the factor variables of the health concept Regression of the factor 1 variable of the health concept and the two factor variables of the democratic governance concept Regression of the factor 2 variable of the health concept and the two factor variables of the democratic governance concept
78
6.1
78
6.2
81 81 83
6.2.1 6.2.2
6.3
Partial correlation test – Measuring the relation between the health concept variables and the democratic governance variables along the groups of the principal components analysis by controlling the influence of economic strength 84 Partial correlation test of all countries between the variables of the health concept and the variables of the democratic governance concept by holding GDP constant Partial correlation test within Group1 and Group2 between the variables of the health concept and variables of the democratic governance concept by holding GDP constant
6.3.1 6.3.2
85 87
7
Summary and final conclusions
93 96
Bibliography Annex D – Methodological definition of the variables
D1
Annex R – Requirements for the quantitative tests R1
1 2 3 4 The principal components analysis Student's T-Test The regression The partial correlation R1 R7 R8 R13
Annex A – Analysis of the test procedures
1 2 Analysis of the test procedure of the Principal components analysis of the health concept Analysis of the test procedure of the regression
A1
A1 A4
Annex V – Variable chart Annex G – Group chart
V1 G1
Abstract
This study examines the statistical relation between two concepts and between the single variables of these concepts. The first concept includes measurements of the level of health, the performance of health care systems, the expenditure on health and the equality of the distribution of health in 44 European states. The second concept includes measurements of the level of democracy, of good and democratic governance and the degree of political participation of citizens in 44 European states. It is assumed that the relation between the two concepts is influenced by the overall economic strength of a country. The hypothesis and the theoretical framework as the basis for the examination of the relation between these two concepts refer to the process of modernization including socioeconomic development (resource allocation like for instance the allocation of health) and democratization. Furthermore this study applies to theories of political economy that propose a relation between the degree of political participation and the equalization of the socioeconomic status within societies. By using a principal components analysis the variables of the two concepts are compressed. Furthermore groups along the output of the principal components analysis of the health concept are formed in order to make a comparison between countries. The statistical relation between the concepts is examined by a regression test and a partial correlation test. The output of the tests indicates that there is no statistical relation between political participation and the level of health, the performance of health care systems or the expenditure on health. Furthermore the ••1
democratic governance standards are not related to the degree of state responsibility in the funding of health care. The two groups – one with a relatively high level of health, high health care expenditure and a good health care performance, the other with low levels on the health concept variables – differ concerning the relation between the two concepts. The partial correlation adverts that within the first group the ability of citizens to choose and control their governments is positively correlated with fairness of financial contribution to health systems. Within the second group – the one with the lower health standards – the ability of people to choose their governing power, the degree of political and civil freedom and the control of corruption are positively correlated with the expectancy to live a healthy life. Consequently it is summarized that within the first group the relation between democratic power of citizens and health resource allocation refers aspects of equalisation within society. In contrast to that within the second group democratic governance is related with the general status of health. The two groups differ significantly in terms of economic strength.
Mathematic formulas of the variables and of the quantitative tests, the requirements of the test procedures and the evaluation of these requirements are attached in the annexes. The datasets for the tests and the output of the tests are included in the annex but only in the digital version.
••2
1
Aims of the research
This study has two main aims: First the hypothesis is examined by using a multivariate statistical test procedure. Secondly – as this study is a comparative study - groups of countries are formed that are the objects of the comparison. The comparison includes an analysis of the similarities and differences of the groups of countries along the question of the hypothesis. In this chapter the hypothesis will be formulated and the quantitative approach will be discussed within the conceptualization of the comparison.
1.1
The hypothesis of the study
The hypothesis formulated as a question:
Is there a relation between democratic governance (the level of political participation, the level of democracy and democratic and well-functioning governance) and health (the level of health, the performance of health care, the expenditure on health and the degree of state involvement/governmental responsibility in health care financing)? Is this relation crucially influenced by the economic well-being?
The hypothesis is thus based on the probable influence of a wellfunctioning democracy on health. It is assumed that democratic governance is influencing health. Furthermore it is assumed that ••3
economic well-being crucially influences the level of health, health care spending and health care performance. By taking this influence into account it is examined whether a relation between democratic governance and health exists besides the influence of economic strength.
This study is not a comparison between health care systems, it is not concerned with characteristics of welfare systems of the 44 countries or with a certain health policy aspect but solely with the examination of the relation between certain aspects of health (system performance) and certain aspects of democratic governance. This dissertation neither discusses the historical development of health care systems nor the institutional structures of health care systems in European countries. The comparison focus on differences and similarities of groups of European countries along the examination of the hypothesis.
What kind of welfare analysing tradition does this study follow? Lessenich distinguishes in terms of welfare state studies between the functional (industrialization, modernization, capitalism), the conflict theoretical (democratization, class politics, party democracy) and the institutional (competition of states, nation state building, structures of the states) approach. [see Lessenich: 2000, page 42] This study follows aspects of the historical functional approach as it refers to the overall context of modernization. The establishment of health care systems is understood as part of macro processes of Human Development by which western states have been transformed. This study applies also to the conflict theoretical approach by emphasizing the central role of democracy. Studies of this tradition are based on the "Politics matter" principle and mainly refer to economic theories of politics and the conceptualization of the welfare state of T. H. Marshall. [see Lessenich: 2000, page 46] According to Marshall the welfare state is the result of a democratization process and thus the material redemption of the democratic ideal of equality. Modernization of Western societies is based on the conflict between economic and political interests – the market and the state. ••4
1.2
The quantitative comparison
Questions of relations between two concepts – such as the hypothesis of this study – can be analysed by breaking the concepts into dimensions, indicators and variables and by statistically testing the relation between them. The statistical tests of this study that are computed with the SPSS program require scaled data. To perform a principal component analysis a certain amount of cases must be included.
In this study groups of 44 European countries are compared. The decision to compare as many European countries as possible was taken for mainly two reasons. First – by choosing a theoretical approach that focuses on the overall democratization process – it seemed to be especially interesting to examine countries with a longer history of democracy and others that have only established democratic rule since the 1990ies. Although some of them will join the European Union next year the historical division between East and West, ex-communist and capitalist is still dominant within political discussions and political identity. It has been exciting to design this study along a quite comprehensive definition of Europe including countries like Georgia, Azerbaijan and Russia. The second reason for choosing 44 European countries is simply based on the fact that data have only been available for these European countries.
In this study macro-level characteristics of groups of countries are compared. Gauthier criticizes that such comparisons tend to emphasize the dissimilarities across countries because data are often presented in terms of country ranking. [see Gauthier: 2000, page 12] This comparison does not focus on differences between countries but on differences between groups of countries. These groups are built along the output of the first statistical test – the principal component analysis. In the analysis of this study similarities and differences are described along these groups. The focus on European nation states is thus loosened ••5
although the data for the comparison all refer to the individual state level.
This study is a multivariate multi-national analysis. As one single multivariate model is used it is assumed that it applies to all countries. "Unlike some other analytical strategies (…) the aim is not to test whether findings in one country can be generalized to other countries, but instead to develop a single model based on the pooled data from various countries. Countries are thus used to increase the level of variance in the main variable of interest." [Gauthier: 2000, page 13] Gauthier states that such studies may be too much focused on the analysis of measures of central tendency at the expense of measures of dispersion and tend to neglect the importance of outliers.
However, the comparative design of this study offers opportunities to ask the extent to which the question of relation – the hypothesis – is similar or dissimilar across groups of countries. If the statistical output enhances the acceptance of the hypothesis – i.e. processes of democratisation and the provision of democratic rules are related to health outcomes – it is to look what kind of democratic rules have the strongest relation with variables that measure the level of health. It is also to examine in which group of countries the relationship between the variables of the two concepts is stronger or weaker. Although it would be great to examine a causal relation between the two concepts focus on a simple test of correlation.
Critical and ethical aspects of quantitative research
Although quantitative data are often perceived as unquestionable facts, critics of the traditional analytical school of science emphasize that they should be seen as social constructions that represent approaches towards reality and interpretations of reality but never reality itself. While in the traditional analytical school of science the researcher is seen as impartial the critical schools of science emphasize that the researcher should be ••6
seen as involved in the research process and as crucially influencing it. The choice of the procedure, the concepts and the variables that are chosen by the researcher may have a crucial impact on the output of the study. According to the critical rationalism school of science the acceptance of a hypothesis may can be enhanced by several analyses that do not falsify the hypothesis. Critical rationalism demands that all statements of empirical science have to be principally qualified for being falsified due to empirical experience. [see Kromrey: 2002, page 34]
••7
2
The theoretical framework
In this part of the study the general theoretical framework will be pointed out. Relevant studies and literature are examined to refine the approach of the research. Finally the concepts and the variables will be described.
2.1
The general theory – Democratization and Socioeconomic development
In the study "Human Development as a General Theory of Social Change: A Multi-Level and Cross-Cultural Perspective", Welzel, Inglehart and Klingemann develop a model of Human Development by connecting three processes of modernization: socioeconomic development, cultural modernization and democratic regime performance. According to that socioeconomic development gives people more autonomy resources, cultural modernization mobilizes aspirations and democracy provides an institutional structure that codifies civil and political freedom opportunities. In their empirical analysis they found out that Human Development is shaped by a causal effect from resources and aspirations on opportunities. Furthermore they state that good governance – they call it also "elite integrity" – is a strong exogenous determinant of Human Development. They pointed out that elite integrity, a rational administrative and managerial class may be an effective way to promote socioeconomic resource allocation and democratization. [see Welzel et al: 2001, page 28] In their study a strong correlation between "Human ••8
Development" and "elite integrity" is shown.
Source: [Welzel et al: 2001, page 28]
Socioeconomic development is broadly defined as "productivity growth, improving quality of life in terms of health and life expectancy, increasing material prosperity, expanding education and communication, and increasing social diversification". [see Welzel et al: 2001, page 1] Cultural modernization refers to the disburdening of people from hierarchical and clientelistic ties that restrict human autonomy. By cultural modernization values like individualism, self-determination and tolerance of human diversity are emphasized. As the authors refer to various different concepts (civic cultural attitudes, post materialist values, liberal attitudes, social capital and democratic personalities) their meaning of cultural modernization is quite broad and the demarcation to the other two modernization processes (the economic and the political) become blurred. The third part of the Human Development model is democratization. The authors refer to the "third wave of democratization" ••9
in the past decade and to the improving of democratic institutions within established democracies. How do the authors connect these three strands of Human Development?
"If socioeconomic development, cultural modernization and democratization occur, they tend to do so on coincidently low or high levels." [see Welzel et al: 2001, page 2] The authors point out that the classical modernization theory lacks a general definition of the term that clarifies the common principle of its components. The causal links between the three factors of modernization are examined in all possible ways and directions. Animated by this distracted theoretical puzzle the authors propose that the common denominator of all Human Development processes is "human choice" in the sense that human beings are able to choose the live they want. They thus argue that "socioeconomic development, cultural modernization, and democratic regime performance work together in promoting individual choice". [see Welzel et al: 2001, page 5] It is obvious that the authors emphasize the aspect of "individual freedom" along the liberal school of thoughts by choosing "individual choice" as the main principle underlying processes of Human Development. "Human Development of societies means growing individual choice on a mass level." [Welzel et al: 2001, page 5] A very similar definition of Human Development is given by Haugaard: "The substance of Human Development is the enlargement of choices in the economic, cultural and social spheres of life, leading to a free, long and healthy life." [Haugaard: 2001, page 4] But Haugaard does not only emphasise "choice" but also "equality" as substantial feature of Human Development. "Equal access to the market, education, health, the political process are critical elements of Human Development. A well functioning health or education sector is therefore not enough if it is only accessible for a minority of the population." [Haugaard: 2001, page 4]
••10
Source: [Welzel et al: 2001, page 6]
The causal linkage the authors draw between the different modernization processes is as follows: Political modernization depends on socioeconomic modernization. Furthermore autonomy resources (socioeconomic development) give rise to liberty aspirations (cultural modernization). These aspirations put adjustment pressure on opportunities (democratization). They conclude: "Freedom opportunities need corresponding liberty aspirations but cannot create them. (…) If adjustment pressure is at work, it works from liberty aspirations on freedom opportunities rather then the reverse." [Welzel et al: 2001, page 8]
What does that have to do with the hypothesis of this study? First, one may assume that the level of health status, the quality of the performance of health care systems and the health care expenditure are socioeconomic resources shaped by socioeconomic development. Secondly the quality of governance and the participation of citizens depend on the state of democratization. Both concepts "health" and
••11
"democratic governance" are thus embedded in the theory of "Human Development".
2.2
Relevant literature and theories concerning the relationship between ‘health outcomes’ and ‘democracy, political participation and good governance’
The following studies and theories include theoretical links between the two concepts of this study.
There are some good arguments for the assumption that democratic governance has an impact on health although Welzel et al suggest that socioeconomic development causes democratization and not the other way around.
Let us look at the famous theory of social citizenship of T. H. Marshall. He orders Human Development chronologically. In the 18th century civil rights (individual freedom, liberty of the person, freedom to speech, thought and faith, the right to own property and to conclude valid contracts and the right to justice) have been institutionalised, in the 19th century the emergence of political rights (membership of a body invested with political authority or as an elector of the members of such a body) was the next important step towards more independence and finally the establishment of social rights in the 20th century (the right to a modicum of economic welfare and security and to live the life of a civilised being according to the standards prevailing in the society) has built the basis for independence from the unfairness of the market. In this historical perspective social rights are understood as the perfection of human autonomy because they enable people to practice their civil and political ••12
rights. The welfare state is seen as necessary to raise the relatively poor to a condition in which they can enjoy the citizenly condition of full autonomy, freedom and participation. Marshall' s perspective was criticized for being too much focused on Great Britain, and for ignoring the situation of women. Furthermore in several European countries social rights have emerged before political rights, e.g. in Germany. To state that democratization is chronologically prior to public welfare – as part of the socioeconomic development – is thus highly questionable. But Marshall is also criticized for not putting enough attention on the organisation of the state and the autonomy of the citizens.
Sven Murmann criticizes Marshall for ignoring that the institutional structures that provide social autonomy can only be influenced by people's political autonomy. According to that, political autonomy is principally prior because it provides the power for citizens to reach autonomy in the socioeconomic field of their life. But as social rights are not necessarily connected to political rights the principle of autonomy is probably neglected. The claim for social protection must be therefore linked with the ability of citizens to impinge on the monopoly of the state that guarantees security and welfare. [see Murmann: 2000, page 49] In other words: The political role of citizens as authors of the law is prior to all other roles.
Thus democratic legitimacy is part of the argument to examine the influence of democratic governance on health outcomes and not the other way around. As Habermas points out [see Habermas: 1991] it is the political status of citizens that may change their material status. Getting access to markets, labour, schools, communication means or hospitals is not automatically connected with increasing autonomy, he says. [see Habermas: 1991, page 20] Civil rights and social participation rights may be awarded to people in a paternalistic manner. In contrast political rights lead to self-determination. Habermas concludes that the state of law and the welfare state can be principally established without democracy. ••13
For these reasons of democratic legitimacy this study is based on the assumption that democratic governance may influence health outcomes. This theory also builds the very basis for studies that examine the impact of political participation on socioeconomic well-being.
In his study "Participation, Veto Actors, and Policy Responsiveness in the Evolution and Reform of Health Care in Developed Democracies" Robert J. Franzese stresses that "institutions that fostered democratic participation also enhanced governmental responsiveness to inequality and economic hardship (…), inducing greater effective demand for publichealth-spending growth in some democracies than others." [Franzese: 2002, page 1] Very briefly this theory is based on the assumption that democratic power of people (their electoral resource) and their economic resources are unequally distributed in capitalist societies. This produces poorer median than average citizens, generating popular demand for public services. [see Franzese: 2002, page 8] The median voter desires more public transfers the greater the difference between median and mean income is. Democracies respond to the interests of the median voter because political influence is distributed evenly and because majorities rule. As wealthier relative to poorer citizens participate more in election, the political overweighting of the wealthier, should decline as voter participation increases.
Therefore higher voter participation correlates positively with increases from right (rich) to left (poor) in the proportion of the income distribution that votes. According to Franzese's theory electoral institutions and income distribution interact to determine the demand for public health services. It can be thus assumed that by his political power the median voter may be able to change his socioeconomic status.
[Shin: 2002] has linked the "income inequality" theory with democracy and health. His main assumption is that democracy may temper or fortify the association between income inequality and health. He examines whether income inequality itself has an impact on population health, or ••14
income inequality is a surrogate for other political factors such as levels of democracy. He argues that "within egalitarian and democratic societies, healthcare is arguably more easily obtained and there are more opportunities to improve socio-economic status, which in turn provides greater access to the means by which health can be improved". [Shin: 2002, page 3]
Díaz-Bonilla et al focus on the impact of "governance". They suggest that the allocation of investment and the access to public health services, may be distorted by bad governance, e.g. by the payment of bribes, whose distribution would then mimic a market alocation based on capacity to pay. Corruption in government procurement of medicines and equipment could lead to inflated prices or low quality products and thus diminish the welfare impact of a given budget alocation. [see Díaz-Bonilla et al: 2002, page 14] [Gupta: 2000] finds that countries with high corruption have high child and infant mortality rates. In their study "Public Spending and Outcomes: Does Governance Matter?" Rajkumar and Swaroop investigate if differences in efficacy of public spending can be explained by quality of governance. They found out that "public health spending lowers child and infant mortality rates in countries with good governance". [see Rajkumar & Swaroop: 2002, page 1] Their approach is based on the assumption that well-functioning public institutions are critical for translating public spending into effective services. Indeed, they found out that in very corrupt countries with very ineffective bureaucracy, public health spending is inefficacious. As the level of corruption goes down, public spending on health becomes more effective in lowering child and infant mortalities. [see Rajkumar & Swaroop: 2002, page 20]
Finally Kaufman et al show that governance indicators have a strong direct negative impact on infant mortality.
••15
Source: [Kaufmann et al: 1999, page 24]
The WHO has done further research and found out that both the health and overall efficiency measures are strongly positively correlated with the "index of Government effectiveness". There is also a positive correlation between "Voice and accountability", and two health system efficiency measures. [see Travis et al: 2002, page 12]
••16
This study will among other aspects move along this path and provide further exploration of the relationship between the indicators of the World Bank and measurements of "health performance" of the WHO.
The World Health Organization (WHO) has focused on the impact of "good governance" on the health performance in the World Health Report 2000 by conceptualizing the term "stewardship". The report defines it as "the careful and responsible management of the well-being of the population", and "the very essence of good government". [see Travis et al: 2002, page 1]
The theoretical conceptualization of this study comprises the theory of Human Development referring to the relation between socioeconomic development and democratization. Furthermore this study is based on theories that link political participation with the achievement of equality or general social well-being through the power of the electorate. This paper also includes theoretical considerations and de facto measurements of "good governance" and their impact on health. In general this research is strongly influenced by theories of citizenship that emphasize the priority of political autonomy.
2.3
The two concepts of the study: "health" and "democratic governance"– a theoretical conceptualisation
In this chapter the two concepts and the indicators of this study are described. The terms "health" and "democratic governance" are used as names for the special concepts of this study but not as comprehensive definitions.
••17
2.3.1
The concept of "health"
As already mentioned health status, public health expenditure and health care performance are seen as socioeconomic resources that are embedded in the overall socioeconomic development. Health achievement is linked to economic development because high incomes and a stable economic growth increase the state's capacity to provide health care systems and to promote health. [Carey & Judge: 2001] have found evidence that income and better health mutually reinforce each other. Other authors like [Easterlin: 1995, 1999] and [Shiffman: 2000] advert that the causal link between income and health is quite weak.
In regard to the human capital to invest in people's health is an input into economic growth and development. [see Subramanian et al: 2002, page 293] Whatever perspective on the relationship between economic development and health one takes, it is evident that poverty is strongly correlated with poorer health outcomes. [see Subramanian et al: 2002, page 299]
The connection between health outcomes and economic well-being is crucial and it is thus necessary to build a theoretical model where the influence of economic strength on health is included.
In the next step it is to clarify what aspects of "health" outcomes are included in this study. It is to mention that the concept of health in this study is strongly influenced by the work of the WHO. In the hypothesis it was assumed that "democratic governance" might have an influence on the health status, the performance of the health system, and the level of responsibility of the state concerning health care systems. It is obvious, that better health is the primary goal of a health system and should be therefore taken into account. The WHO refers to two other fundamental objectives of health systems. Health systems should respond to people's expectations and they should provide financial protection against the costs of ill-health. ••18
The reason why financial fairness is that important is that health care can be catastrophically costly. As much of the need for care is unpredictable, it is vital for people to be protected from having to choose between financial ruin and loss of health. The responsiveness of the system is crucial because illness and its consequences can threaten people's dignity. [see Musgrove: 2000, page 24] Responsiveness means thus to reduce the damage to one's dignity and autonomy, and the fear and shame that sickness often brings with it. The claim for autonomy within health systems also includes that patients demand more and more to be perceived as responsible partners who want to choose their medical treatment.
According to the WHO health systems should also try to "reduce inequalities by preferentially improving the health of the worse-off, wherever these inequalities are caused by conditions amenable to intervention". [see Musgrove: 2000, page 26] Thus, the evaluation of the performance of health systems also includes evaluations of the equality of distribution of health.
Of course the level of responsiveness, financial fairness or equality of distribution of health systems are not objective criteria but deeply depend on the political values of a government and society. Nevertheless the WHO developed common criteria in order to compare the system of different states.
This study is influenced by theoretical considerations proposing a relationship between the amount of government expenditure spent for welfare and political participation. Thus it will be looked for variables that measure the level of state responsibility in the funding of health care.
The health concept comprises measurements of the level of health (health status), health system performance (financial fairness, responsiveness and equality of distribution of health), the degree of state intervention within health systems and the expenditure on health. ••19
2.3.2
The concept of "democratic governance"
The concept "democratic governance" refers to several aspects of democracy and democratization processes. Democracy in general is characterized by legal competitions for political power through legitimated elections, inclusive political participation in the selection of leaders and policies, and a high level of protection of civil and political rights. Democratization is defined as "a process by which political power is redistributed to be more egalitarian and is a process that establishes and/or reinforces both democratic norms and institutions." [Ward et al: 1997, page 1]
Political participation is a necessary condition for the persistence of democracy and an essential part of a well-functioning democracy. Peaceful competition generates responsiveness for governments to their electorate. In other words: Democracy creates political incentives for rulers to respond positively to the needs and demands of their citizens. Strong political participation and accountability may assure that government provides the services that the citizens need. [see von Hagen et al: 2003, page 5] Related to that study one may assume that politicians that are elected in free and competitive elections with high voter participation are likely to be more accountable and responsive to ideas of the voters concerning health policies.
The level of democracy is the second aspect that is crucial when measuring democratization processes. It is often expressed as the level of political and civil rights.
Third the functioning of democratic institutions can be judged by looking on the goodness of governance. In the model of Welzel et al "governance" is not included in the measurement of the democratization but seen as a separated component of Human Development. In this study it is included in the concept of democratization. ••20
The United Nations defines "governance" as "the exercise of economic, political, and administrative authority to manage a country's affairs at all levels. It comprises "mechanisms, processes, and institutions through which citizens and groups articulate their interests, exercise their legal rights, meet their legal obligations, and mediate their differences." [Girishankar et al: 2001, page 271]
In the OECD definition governance "denotes the use of political authority and exercise of control in a society in relation to the management of its resources for social and economic development. This broad definition encompasses the role of public authorities in establishing the environment in which economic operators function and in determining the distribution of benefits as well as the relationship between the ruler and the ruled." [Girishankar et al: 2001, page 271]
For the European Union governance is embedded in the "context of a political and institutional environment that upholds human rights, democratic principles, and the Rule of law". Good governance is "the transparent and accountable management of human, natural, economic, and financial resources for equitable and sustainable development. It entails clear decision making procedures at the level of public authorities, transparent and accountable institutions, the primacy of law in managing and distributing resources, and capacity building for elaborating and implementing measures that aim to prevent and combat corruption." [Girishankar et al: 2001, page 271]
The World Bank defines it "as the norms, traditions, and institutions through which a country exercises authority for the common good. It includes the processes for selecting, monitoring, and replacing those in authority; the capacity of government to manage its resources and to implement sound policies; and the respect that citizens and the state have for the institutions that govern economic and social interactions among them." [World Bank: 2000, page 3]
••21
Haugaard strongly relates the quality of governance to a state's capacity and legitimacy, i.e. to ensure that all voices are taken into account and that policies are efficiently implemented.
Administrative quality, especially the functioning of a non-politicised bureaucracy, the Rule of law and the level of corruption are crucial for the capacity and legitimacy of the state. [see Haugaard: 2001, page 8] By referring to Max Weber Haugaard stresses the importance of meritocratic recruitment in bureaucracies and the rewarding of long term careers.
For the World Bank good governance is "epitomized by predictable, open, and enlightened policy making (that is, transparent processes); a bureaucracy imbued with a professional ethos; an executive arm of government accountable for its actions; and a strong civil society participating in public affairs; and all behaving under the Rule of law." [Rajkumar & Swaroop: 2002, page 1] "Rule of law" means that formal rules have to be enforced in a predictable way through transparent mechanisms that are applied equally to all citizens. The legal system must constrain and channel government action, and maintain clear procedures for upholding people's Constitutional rights to guard against abuse of power by the state or other actors. [see World Bank: 2000, page 13]
According to the World Bank democratic governance has to meet three essential conditions [World Bank: 2000, page 8]: • "Meaningful, extensive, regular, and peaceful competition among individuals and organized groups (especially political parties) for most positions of political power. • Highly inclusive participation in the selection of leaders and policies, at least through regular and fair elections, with no major social group excluded.
••22
• Civil and political liberties – freedom of expression, freedom of the press, freedom to form and join organizations – sufficient to ensure the integrity of political competition and participation."
In recent times the importance of good and democratic governance for development outcomes have been stressed. The empirically proved correlation between the level of economic development (per-capita GDP) and good governance is strongly positive, i.e. better institutions are typically paired with higher per-capita incomes. [see von Hagen et al: 2003, page 6]
Because of the strong relation between good governance and economic strength it is necessary to design a theoretical model that takes that relation into account. In this model it is assumed that democratic governance and economic strength mutually reinforce each other. The concept "democratic governance" includes political participation as an essential part of a well-functioning democracy, the level of democratization measured by political freedom and good governance as part of the institutionalization of democracy.
2.4
Measuring health and democratic governance
In this chapter the link between the concept, the indicators and the variables is made. The model to measure the concepts is as follows:
••23
The interrogation mark represents the question of the hypothesis that will be examined. In the next step variables are chosen that represent the concepts of this study. The data used for the study are not collected in the same year but all variables are the most recent available.
2.4.1
Measuring health
As already mentioned the health status, the performance of the health system, health expenditure and the level of governmental responsibility ••24
should be reflected in the variables. In the World Health Report 2000 the WHO offers measurements of the level of health (Healthy life expectancy), the responsiveness of the health system, the equality of the distribution of health (equality of child survival) and a variable that measures the fairness of financial contribution to the health system. Furthermore variables were chosen that represent the degree of governmental responsibility (General government expenditure on health as a share of total health expenditure; General government expenditure on health as a share of total government expenditure). Finally variables are included that measure the per capita expenditure on health. As pointed out above the degree of state responsibility and the amount of public expenditure may be linked to the level of political engagement and the state of democracy.
2.4.2
Measuring democratic governance
In the Welzel study good governance is measured by corruption scores of Transparency International. In this study the "good governance" variables of the World Bank will be used. The World Bank [Kaufmann et al: 2003] offers an extensive database that allows international comparisons. They look at governance in six dimensions and measure Voice and accountability, Political stability, Government effectiveness, Regulatory quality, Rule of law and Control of corruption. These variables are strongly connected to democracy, e.g. Voice and accountability measures if and to what extent the claims of the citizens are respected. The level of democratization is measured by the product of the scores for "civil liberties" and "political rights" called "Freedom house rate" that is published annually by the organisation Freedom House.
Political participation is measured by voter turnout. Voter turnout is just one dimension of political participation but it was not possible to find data
••25
for 44 European countries for other participation indicators like union density or party membership.
Finally economic strength of a state may play a crucial role in the relation between the two concepts. For this study GDP per capita will be used to examine the relationship. The economic development measured in terms of gross domestic product determines the degree to which material and societal needs of the population are met [see Hanzek: 1998, page 52]
The following theoretical model includes the variables chosen for the study:
••26
••27
2.5
Definition of the variables
In this chapter the variables will be defined. The methodological definition of the variables – the mathematical formulas, the description of the selection procedure of the variables and critical remarks concerning their conceptualization – are attached in the "Annex D – Methodological definition of the variables".
2.5.1
Definitions of the variables of the "health" concept
Healthy life expectancy (Estimates for 2001) Healthy life expectancy (HALE) is a modification of conventional life expectancy to account for time lived with disability. It is a summary measure of level of health established by the WHO in order to assess the goal of improving average levels of population health. "It is most easily understood as the equivalent number of years in full health that a newborn can expect to live based on current rates of ill-health and mortality." [Murray & Lopez: 2002, page 172] HALE thus measures the health of population by integrating information on mortality and non-fatal health outcome.
Equality of child survival (Estimates for 1999) Equality of child survival measures the distribution of health within countries, i.e. the level of equality/ inequality of health between individual children (until the age of 2) but not between different socio-economic groups. "The value of 1 can be interpreted as complete equality and zero can be interpreted as a degree of inequality that is worse than has been seen in any
••28
country measured directly or estimated indirectly to date." [Musgrove: 2000, page 147]
Fairness of financial contribution (Estimates for 1997) The normative concept of the WHO concerning Fairness of financial contribution to health system implies that the sacrifice created by contributing to the health system should be equalized across households independent of their health status or utilization of health services. [see WHO draft on health inequalities, page 44] Fairness of financial contribution means "that the risks each household faces due to the costs of the health system are distributed according to ability to pay rather than to the risk of illness: a fairly financed system ensures financial protection for everyone. A health system in which individuals or households are sometimes forced into poverty through their purchase of needed care, or forced to do without it because of the cost, is unfair." [Musgrove: 2000, page 35]
Fairness of financial contribution is based on the assumption that the fraction of disposable income that each household contributes to the health system indicates the extent to which resource generation reflects people's ability to pay. [Murray & Lopez: 2002, page 140] The WHO differentiates between two kind of "unfair" financing mechanisms. The first refers to unexpected high out of pocket payments as a consequence of a lack of prepayment systems. Secondly a health system can be unfair due to regressive payments – as a consequence of financing that is related to risks rather than to the ability to afford the care. [see Musgrove: 2000, page 35]
A health system is financed in a fair way "if the ratio of total health contribution to total non-food spending is identical for all households, independently of their income, their health status or their use of the health system". [see Musgrove: 2000, page 36] ••29
The index "Fairness of financial contributions" runs from zero (extreme inequality) to 1 (perfect equality). Countries with high values of fairness of financial contributions have generally well established prepaid financing systems.
Responsiveness of the health care system (Estimates for 1999) Responsiveness is "the outcome that can be achieved when institutions and institutional relationships are designed in such a way that they are cognisant and respond appropriately to the universally legitimate expectations of individuals." [see de Silva: 2000] The measurement is based on the assumption that the way people are treated when they come in contact with the system can improve or reduce well-being independent of health outcomes. [see WHO draft on health inequalities, page 54] The index thus refers to the non-health enhancing, non-financial aspects of the health system. [see de Silva: 2000] It includes measurements of the degree of dignity, autonomy, confidentiality, prompt attention, access to social support, quality of basic amenities and choice a health system offers the clients. Dignity refers to the right of individuals to be treated as persons in their own right rather than merely as patients. Respect for persons included the concept of autonomy (as self-directing freedom with regard to deciding between alternative treatment, testing and care options, including the decision to refuse treatment) and confidentiality that refers to safeguarding privacy. Prompt attention means to have speedily access and short waiting times for consultation and treatment. Access to social support networks during care refers to provisions that integrate health care activities into community interactions. The qualities of basic amenities are physical attributes of health care units such as cleanliness of the facility, adequacy of furniture and quality of food. Finally "the choice of care provider covers choice between and within health care units, including opportunities for gaining specialist care and second opinions". [see Valentine et al: 2000, page 2] The seven elements were scored from 0 to 10. ••30
General Government expenditure on Health 2000, Share in Total expenditure on Health (Estimates for 2000) This variable represents the percental share of government expenditure on health of total (government and private) expenditure on health. General government expenditure on health is the sum of outlays on health paid for by taxes, social security contributions and external resources (including the expenditure to purchase health goods and services by schemes that are compulsory, under governmental control, and covering a sizeable segment of the population). External resources are e.g. concessional loans and grants for medical care and medical goods channelled through the Ministry of Health or via the Ministry of Finance or Central Bank. Cash benefits for sickness and/or loss of employment are not included in the estimates, as these are classified as income maintenance expenditure. "General government" is defined as the central or federal government, regional/ state/provincial authorities, municipal and local authorities, and autonomous trust funds or boards implementing government policies, principally social protection agencies or social security schemes. [see Murray & Lopez: 2002, page 174]
Private expenditure on health comprises four types of entities: • Those that pool resources in order to purchase medical goods and services or to finance delivery facilities. These prepaid private risk-pooling plans include the outlays of private social insurance schemes, commercial and non-profit (mutual) insurance schemes, health maintenance organizations and other agents managing prepaid medical and paramedical benefits, including the operating costs of these schemes. • The non-financial corporations that provide medical and paramedical goods and services to their employees on top of compulsory social insurance or resource pooling entities. • The nongovernmental organizations and non-profit institutions that use resources to purchase health goods and services that are ••31
not allowed to be a source of income, profit or other financial gain for the units that establish. • Out-of-pocket payments: In many systems they top-up benefits accessible through private pooling. Included are gratuities and payments in-kind made to health practitioners and to suppliers of pharmaceuticals and therapeutic appliances. [see Murray & Lopez: 2002, page 174]
General Government expenditure on Health 2000, Share in Total Government expenditure The variable represents the share of government expenditure as already defined. Total government expenditure corresponds to the consolidated outlays of all levels of government, social security institutions, and extra-budgetary funds, including capital outlays. [see Murray & Lopez: 2002, page 175]
Per capita Total expenditure on Health 2000, at International Dollar rate The variable represents the total expenditure on health as already defined above, calculated per capita at International Dollar rate. The per capita figures are calculated using population data supplied by the UN Population Division and the OECD (for OECD countries). [see Murray & Lopez: 2002, page 174] Per capita is expressed in international dollar estimates, derived by dividing local currency units by an estimate of their purchasing power parity (PPP) compared to US Dollar. Thus the measure eliminates the consequences of differences in price levels that exist between countries. [see Murray & Lopez: 2002, page 175]
Per capita Government expenditure on Health 2000, at International Dollar rate The variable represents the government expenditure as already defined, counted per capita at International Dollar rate.
••32
2.5.2
Definitions of the variables of the "democratic governance" concept
The processes by which those in authority are selected and replaced refer to the variables "Voice and accountability" and "Political stability and Absence of Violence".
Voice and accountability (Estimates for 2002) The variable includes a number of indicators measuring various aspects of the political process, civil liberties and political rights. It refers to the extent to which citizens of a country are able to participate in the selection of governments. The variable also comprises indicators measuring the independence of the media, which serves an important role in holding monitoring those in authority and holding them accountable for their actions. [see Kaufmann et al: 2003, page 3]
Political stability and Absence of Violence (Estimates for 2002) This variable combines several indicators, which "measure perceptions of the likelihood that the government in power will be destabilized or overthrown by possibly unconstitutional and/or violent means, including domestic violence and terrorism. This index captures the idea that the quality of governance in a country is compromised by the likelihood of wrenching changes in government, which not only has a direct effect on the continuity of policies, but also at a deeper level undermines the ability of all citizens to peacefully select and replace those in power." [Kaufmann et al: 2003, page 3]
The next two variables summarize indicators of the ability of the government to formulate and implement sound policies.
••33
Government effectiveness (Estimates for 2002) This variable refers to "the quality of public service provision, the quality of the bureaucracy, the competence of civil servants, the independence of the civil service from political pressures, and the credibility of the government's commitment to policies. The main focus of this index is on 'inputs' required for the government to be able to produce and implement good policies and deliver public goods." [Kaufmann et al: 2003, page 3]
Regulatory quality (Estimates for 2002) This variable "includes measures of the incidence of marketunfriendly policies such as price controls or inadequate bank supervision, as well as perceptions of the burdens imposed by excessive regulation in areas such as foreign trade and business development." [Kaufmann et al: 2003, page 3]
The next two variables represent measurements of the respect of citizens and the state for the institutions, which govern their interactions.
Rule of law (Estimates for 2002) In "Rule of law" the World Bank includes several indicators, which measure the extent to which agents have confidence in and abide by the rules of society. "These include perceptions of the incidence of crime, the effectiveness and predictability of the judiciary, and the enforceability of contracts. Together, these indicators measure the success of a society in developing an environment in which fair and predictable rules form the basis for economic and social interactions, and importantly, the extent to which property rights are protected." [Kaufmann et al: 2003, page 4]
Control of corruption (Estimates for 2002) The Control of corruption "measures perceptions of corruption,
••34
conventionally defined as the exercise of public power for private gain. Despite this straightforward focus, the particular aspect of corruption measured by the various sources differs somewhat, ranging from the frequency of 'additional payments to get things done', to the effects of corruption on the business environment, to measuring 'grand corruption' in the political arena or in the tendency of elite forms to engage in 'state capture'. The presence of corruption is often a manifestation of a lack of respect of both the corrupter (typically a private citizen or firm) and the corrupted (typically a public official or politician) for the rules which govern their interactions, and hence represents a failure of governance according to our definition." [Kaufmann et al: 2003, page 4]
Freedom house rate (Estimates for 2003) The Freedom house rate is based on the following definition of democracy as "a political system in which the people choose their authoritative leaders freely from among competing groups and individuals who were not designated by the government". Freedom – in the Freedom House definition "represents the opportunity to act spontaneously in a variety of fields outside the control of the government and other centres of potential domination". [see Freedom House: 2000] Political rights are defined as enabling "people to participate freely in the political process, which is the system by which the polity chooses authoritative policy makers and attempts to make binding decisions affecting the national, regional, or local community. In a free society, this represents the right of all adults to vote and compete for public office, and for elected representatives to have a decisive vote on public policies." Civil liberties are defined as including "the freedoms to develop views, institutions, and personal autonomy apart from the state". [see Freedom House: 2000] Countries that rate on average 1–2.5 are generally considered "free," 3–5.5 "partly free," and 5.5–7 "not free."
••35
Voter turnout in the 1990ies The variable voter turnout (VOTEREG) refers to "the number of votes divided by the number of names on the voters' register, expressed as a percentage". [see IDEA(a)] It represents the average voter turnout in the 1990ies by accounting both presidential and parliamentary elections.
Voter turnout from 1945 to 1999 The variable voter turnout from 1945 to 1999 (VTRG4599) refers to the average level of voter turnout of parliamentary and presidential elections from 1945 to 1999. The variable is measured by dividing the number of votes by the number of names on the voters' register, expressed as a percentage. As many of the countries had a non-democratic period the average is drawn out from very differing – in terms of quantity – election pools. But still the variable shows if a country has "traditionally" a high or low voter turnout.
The last variable to define is the control variable "GDP per capita".
GDP per capita in Dollar The dates of the GDP measurement differ between countries (the most recent data available were chosen). Some GDP per capita data are estimates for 2000, some for 2001 or 2002.
GDP is the gross domestic product or value of all final goods and services produced within a nation in a given year. The variable shows GDP in Dollar on a purchasing power parity (PPP) basis divided by population as of 1 July for the same year. "The data derived from the PPP method provide the best available starting point for comparisons of economic strength and well-being between countries." [World Factbook: 2002]
••36
3
Design of the test procedure
In this chapter a general overview is presented in order to explain the statistical procedure for the examination of the hypothesis.
3.1
Compressing the variables of the health concept and group building
In this first step the variables of the health concept are compressed in a principal components analysis in order to build and analyze groups of the countries that are used for the comparison. Thus the groups are built along similarities and differences on the health concept variables. With the help of the T-test the differences between the groups are further examined. New variables out of the health concept are formed that will be used for the next tests.
3.2
Analyzing the differences between the groups of countries regarding economic strength
By this step it is examined if the four groups differ significantly in terms of GDP. This is done in order to prove if the differences and similarities of ••37
the groups may be shaped by the overall economic well-being of the states of the groups. The examination is done by a T-test.
3.3
Compressing the variables of the democratic governance concept
In this step the variables of the democratic governance concept are compressed through a principal components analysis in order to simplify the measurement of the concept.
3.4
Proving the influence of the factor variables of the democratic governance concept on the factor variables of the health concept
In this step a regression test is made in order to examine if the factor variables (the result of the principal components analysis) of the democratic governance concept do have an influence on the factor variables (the result of the principal components analysis) of the health concept. This is done by a regression test. According to the hypothesis and the theoretical framework it is assumed that democratic governance may influence health. By looking for a probable relationship between the factor variables it will be also examined if the group building along the factors of the health concept do make sense for the evaluation of the relationship between the concepts. If the regression test indicates that the group building does not make sense groups will be changed.
••38
3.5
Measuring the relation between the health concept variables and the democratic governance variables along the group of the principal components analysis by controlling the influence of economic strength
It is to examine if and what kind of variables of the health concept and the democratic governance concept are correlated. As this relationship may be influenced by the overall economic well-being and strength it is taken into account when measuring the correlation. This can be done by using a partial correlation test that enables to examine correlations by holding a third variable constant. The output of the partial correlation test thus reflects the correlation without the probable influence of the third variable. In this final step of the test procedure the relation between the concepts should also be referred to the objects of the comparison: the groups of countries. Thus the partial correlation test will be computed for each of the groups. The output will be analysed and the final conclusions about the differences and similarities between the groups along the question of the hypothesis will be drawn.
The following graphics refer to the procedural design of the study:
••39
••40
4
Methodology of the quantitative tests
In the following chapter the methodological basis for the tests used in this study is described. The description of the test procedures, the mathematical formulas of the tests and the requirements for the evaluation of the tests are attached in the "Annex R – Requirements for the quantitative tests".
4.1
The "Principal components analysis"
The "Principal components analysis" is a "way of identifying patterns in data, and expressing the data in such a way as to highlight their similarities and differences". [see Smith: 2002, page 12] When the patterns are found the data can be compressed without much loss of information. The principal assumption of this analysis is that some variables are closer connected than others. The strength of the connection can be expressed by correlations. The analysis is based on these correlations and tries to explain the values of the variables by latent factors (also called components).
The aim is to thus find components in a set of variables that represent these variables. [see Schwab: 2002, slide 2] The higher the correlation between the values of the variables the better the factors can be explained. The general idea is to extract a factor – a new variable – that correlates with the observed variables to estimate the original matrix. ••41
4.2
Student's T-Test
The test determines whether the difference between two sample means is large enough to convince that the two corresponding population means must also differ. Thus it measures the probability that two populations are the same with respect to the variable tested.
4.3
The regression
"A regression model is based on the assumption that a dependent variable can be explained by a linear relationship with one or more explanatory variables." [Birkegaard & Porsgaard: 2002, page 56] As in the regression test of this study it is to examine if the factor variables from the democratic governance concept explain the variables of the health concept at least two independent variables are used in the regression (factor 1 and factor 2 of the democratic governance component analysis). Thus the regression test is multivariate.
4.4
The partial correlation
A correlation shows the extent to which two variables, x and y, "go together." [see Gerstman: 2003, page 1] A positive correlation is indicated when high values of x are associated with high values of y. A negative correlation is given when high values of x are associated with low values of y. The correlation in this study is measured by Pearson's correlation test. Pearson's correlation coefficient ‘r’ assumes that each
••42
pair of variables is bivariate normal and it is a measure of linear association.
The absolute value of the correlation coefficient indicates the strength, the closer r is to +1, the stronger the positive correlation. The significance level (or p-value) is the probability of obtaining results as extreme as the one observed. If the significance level is less than 0.05 then the correlation is significant and the two variables are linearly related.
Correlations do not indicate that one variable causes the other. Many correlations are spurious because they are based on a third extraneous variable. In order to examine the influence of a third variable a partial correlation test is appropriate.
The "Partial correlation test" offers an opportunity to examine if a third variable (z) influences the correlation between two other variables (x and y) or even causes it. In the latter case the correlation between x and y is spurious. A correlation is called spurious when the x and y variables have a positive or negative correlation but do not "directly causes changes in the other variable". [see Olson: 2003, page 1]
Common Cause Model [see Baur & Schulze: 2003, page 37]
It could be that the third variable z simultaneously changes x and y and therefore creates the correlation between x and y. The z variable is called an "extraneous" variable. The model is called the Common Cause Model.
••43
Intervention Model [see Baur & Schulze: 2003, page 37]
It could also be that the z variable intervenes in the correlation between x and y so that neither x nor y have a direct effect upon the other. This model is called the Intervention Model: x and y are only correlated due to z. [see Olson: 2003, page 1]
The partial correlation will be used to examine if there is a correlation between ‘health’ variables and democratic governance variables beyond the influence of economic strength (measured by the variable GDP).
In a partial correlation test the z-variable is held constant and can thus not "possibly create a spurious correlation" between x and y or "cause simultaneous changes" in both variables. [see Olson: 2003, page 3] If, after holding the z variable constant, the correlation between x and y disappears, it is clear that the original correlation was spurious due to the influence of the z variable or the correlation was based on an intervention of z.
Generally partial correlations between variables show the correlation one would get between two variables "if all the control variables (…) were held constant at their mean values". [see Olson: 2003, page 5] A partial correlation coefficient ranges from -1 to +1. It offers like the correlation coefficient an indicator for the strength of the relation.
••44
5
Building groups of countries
In this chapter groups will be formed along the output of the principal components analysis of the health concept. Furthermore the groups will be more precisely described with the help of the output of the T-test. The groups of countries are the objects of comparison of this study.
5.1
Analysis of the principal components analysis of the health concept
In this chapter the output of the principal components analysis is described. The analysis of the test procedure is attached as "Annex A – Analysis of the test procedures". Thus this chapter only contains the final conclusions of the test. The dataset for this principal components analysis is attached in the annex in excel and SPSS format called "datasetforhealth.xls" and "dataforhealth.sav". The output of this principal components analysis is also attached in the annex called "outputofhealthprincipalcomponentsanalysis.spo"
This is the last of several principal components analyses of the health concept that were undertaken during the process of the study. The former tests contained much more variables than this latest. The variables have been excluded from the dataset as a result of the test procedure. For reasons of clearness and simplicity the test results of these former tests are not shown or explained. Originally a variable was ••45
included that measures the division between tax based health expenditure and social insurance based health expenditure ("Social Security spending on Health, Share in General Government Expenditure on Health"). This division is crucial for the classification of many welfare theoreticians but as this variable did not fit it is not included in this analysis.
5.1.1
Description of the two components
Variables that load on the first component are: • "Per capita Total expenditure on health, at International Dollar rate" • "Responsiveness-Level of health systems" • "Healthy life expectancy (HALE)" • "Per capita Government expenditure on health, at International Dollar rate" • "Fairness of financial contribution" • "Equality of child survival"
The first component refers to health care performance measurement and the money spend for health care.
The second component contains the variables "General government expenditure on Health as a share in Total expenditure on Health" and "General government expenditure on Health as a share in Total Government expenditure". It reflects government expenditures in relation to non-government expenditures. One expenditure variable refers to health care expenditures in relation to other public expenditures. It reports how much a government is focusing on health care in relation to other public tasks. And the variable "General Government expenditure on Health as share in Total expenditure on Health" reports how much a ••46
government focuses on health care spending in relation to total health care spending.
The second component represents the relative governmental concern and responsibility in health care funding.
5.1.2
The definition of the groups
As shown in the scatter-plot the two factors divide the 44 countries in four groups. As the four groups build the basis for further tests they are classified and characterized along the variables that were used in the principal components analysis.
For the analysis of the features of the four groups the variable chart is used. The variable chart shows the country ranking in a descending way on each of the eight variables that belong to the health concept. The variable chart is attached in the annex called "variablechartforhealthcomp.xls". As the countries are coloured along their membership in the four groups one can see where the four groups are positioned on each variable. Furthermore the average score and the standard deviation of each group are provided in the variable chart. An independent classification that offers information about the features of the group is attached, too. In this classification the countries are divided into three groups.
••47
3 Slovak Republic
2
REGR factor score 2 for analysis
Andorra Iceland Macedonia, FYR Czech Republic Russian Federation Belarus Croatia Norway Slovenia 1 Moldova Luxembourg United Kingdom Lithuania Estonia Ireland Germany Hungary Denmark PortugalSpain France Italy Bulgaria Poland Malta Sweden 0 Belgium Ukraine Finland Bosnia and Herzegovi Turkey Netherlands Latvia Austria Armenia Cyprus Albania Yugoslavia, Fed. Rep Switzerland -1 Greece Romania Azerbaijan
1
-2
-3 Georgia -4 -2 -1 0 1 2
REGR factor score 1 for analysis
1
Scatter-plot of the principal components analysis of the health concept
These groups have been built along the three thirds of the concerned variable values on each variable. The maximum value of a variable minus the minimum value of a variable is divided into three groups. The first group contains the first third of the scores between the maximum and the minimum and is thus called the "high value group", the second contains the second third of the scores and is called the "middle value group" and the third group is called the "low value group". With the help of this classification another country ranking was built. The division is visible by a thin black line between the "high value group" and the "middle value group" and between the "middle value group" and the "low value group" on the variable chart. This new grouping helps to examine the importance of variables for the four groups along the factor 1 and the factor 2. This is done by looking in which of the three new groups the average and the majority of the four factor groups are positioned. It is ••48
also examined which of the four groups has the highest or the lowest standard deviation on one variable. The group charts are shown in the excel sheet "groupcharts" that is attached in the annex.
In the following classification the four groups are first compared along their average scores and standard deviations on the eight variables. A high standard deviation within one group is seen as indicating that the variable is not that important to feature the group. A low standard deviation of a group on a variable indicates more precisely the attitudes of the group. Furthermore the groups are compared along the new grouping (high, middle, low). If all members of one group are e.g. in one of these new value groups one could assume that for describing this group the variable is important because the countries have similar values. Unfortunately this kind of classification is very sensitive to outliers. It is also examined to what extent a variable differentiates one group in terms of the average values from the other.
5.1.2.1
Analysis and definition of Group1
This is the group rating high on factor 1 and low on factor 2: • Austria • Belgium • Cyprus • Finland • Greece • Italy • Malta • Netherlands • Spain • Sweden • Switzerland ••49
Per capita Total expenditure on Health at International Dollar rate The members of the group have the second highest average per capita total expenditure on health. The difference to Group2 (with the highest per capita expenditure) is much lower than to the groups 3+4. The standard deviation of expenditure within the group is the highest of all groups. That fact refers to the impact of the outliers – Switzerland spends four times more than Malta. The three countries with the highest values are Switzerland, Belgium and the Netherlands. The three countries with the lowest scores are Cyprus, Greece and Malta. The majority of the countries of Group1 and the average score of Group1 are positioned in the "middle value group". The variable is important to feature the group but not exclusively important because of the similarities with the Group2 and the high standard deviation.
Per capita Government expenditure on Health at International Dollar rate Group1 has on average the second highest per capita government expenditure on health but the difference to Group2 is clearer than on the former variable. This fact is also expressed by looking on the group charts. The average score and the majority of the countries of Group1 are positioned in the "middle value group" – otherwise than the average score and the majority of the countries of Group2. The strugglers are Greece, Cyprus and Malta. The outriders are Switzerland, Sweden and Belgium. The standard deviation is not the highest and not the lowest. But still in Switzerland the government spend more than three times more than in Malta. As the difference to Group2 is quite high on the average score the variable is important to define the group.
"Healthy life expectancy" (HALE) The "Healthy life expectancy" of this group is on average higher than in all the other groups (70.28 years) although the difference ••50
to Group2 is very small. The standard deviation within this group is in the middle. But the difference between the countries is nevertheless impressing. In Switzerland citizens can expect to live 72.769 years in a healthy state while in Cyprus they can only expect to do so for 66.239 years – a difference of more than six years. Nevertheless all the states of this group are in the "high value group". Belgium, Malta and Cyprus are the strugglers while Switzerland, Sweden and Italy have the highest values. The variable is crucial to differentiate the group because it has the highest average values. In the next test it will be examined if that fact can be further elaborated.
"Equality of child survival" The average level of "Equality of child survival" in this group is very high but a little bit lower than that of Group2. The average score is positioned in the "high value group" as all countries of Group1 belong to the high value group. The standard deviation within this group is the lowest of all groups on this variable. Again Malta and Cyprus lag behind but interestingly also Sweden. Greece, Austria and Italy have the highest scores within Group1. The case of Greece seems to be interesting because the per capita expenditure is relatively low. Generally the variable is important but as the difference to Group2 is quite small the variable can not feature the group exclusively.
"Fairness of financial contribution" to health systems The "Fairness of financial contribution" in Group1 is on average nearly as high as in Group2. The majority of the countries and the average score of the group are positioned in the "high value group". The standard deviation is in the middle of the four groups. The difference between the maximum (Belgium with a value of 0.979) and the minimum (Italy with a value of 0.961) is remarkable. The strugglers are Greece, Malta and Italy. The countries with the highest scores are Belgium, Finland and ••51
Austria. As the scores of this group and Group2 are very similar the variable can not feature Group1 exclusively.
Responsiveness-Level of health systems The level of responsiveness is on average the second best of all groups – the difference to Group2 (with the highest average score) is very small. The standard deviation is in the middle of all groups. The majority of the countries and the average score of the group are positioned in the "high value group". Again Malta and Greece lag behind but also Spain. Switzerland does have the best values, followed by the Netherlands and Sweden. As the countries of this group and the Group2 show again very similar values the variable can not help to differentiate those two groups but is generally important to feature Group1.
General Government expenditure on Health as a Share in Total expenditure on Health The average government share on total health expenditure is quite low in Group1 (on average 67.07 percent). In comparison to Group2 the governments of the countries of Group1 thus spend 13.67 percent less on total Health expenditure. Thus the share of private expenditure of total health expenditure is one third. In comparison with Group2 and Group3 this share of private financing is quite high. The difference to Group4 is lower than to Group2+3. The share of public expenditure of Group4 is even smaller than the one of Group1. The majority of the countries of Group1 are positioned in the "high value group" like the average score. This is due to the fact that there are some countries of Group4 where the involvement of the governments is very small and thus the difference to the other states is very big. The countries that have the highest value of Group1 are Sweden, Finland and Italy. The ones that lag behind are Switzerland, Greece and Cyprus. The variable is crucial to describe the group
••52
because of the difference to Group2 and the similarity with Group3.
General Government expenditure on Health as Share in Total Government expenditure The share of government health expenditure on total government expenditure is on average 11.79 percent – the average score is positioned in the "middle value group". There is a big difference to Group2 and Group4 and a smaller difference to Group3. The variance within the group is the lowest of all groups. Greece (9.2 percent of governmental health expenditure on total governmental expenditure) Finland and Austria are the strugglers of Group1. The countries with the highest levels are Spain and Malta and Italy. The variable is important to define the group because of the difference to Group2 and because of the small variance within group.
Summary
Group1 has high average values on all variables of factor 1. On the variable HALE it has the highest average score. Group1 differs from Group2 on the variables that measure the governmental share of health care spending and on the variable "Per capita governmental health care expenditure". Thus the governments of the countries of Group1 spend generally less money for health care as the governments of Group2. Therefore the financing of health care in the Group1 countries is more based on private spending than in Group2.
Group1 has the lowest within group standard deviation on the variables "Equality of child survival" and government health expenditure as a share of total government expenditure. The latter fact corresponds with the general feature of the group as having a low state emphasis on health expenditure in relation to other public expenditure.
••53
The countries with the lowest values on the variables "Healthy life expectancy", "Equality of child survival", "Fairness of financial contribution" and "Responsiveness of the system" and the two per capita expenditure variables are Malta, Greece and Cyprus. Greece is in a somewhat special position because it has the highest values of all countries of Group1 on the variable "Equality of child survival" although its values are low on the other performance variables. Thus although the per capita expenditure, the "Responsiveness of the system" and the "Fairness of financial contribution" levels are quite low the health is very equally distributed within society. The government of Malta – although the general values on the performance variables are low – emphasises the importance of public health care spending. The share of governmental health care expenditure on total governmental expenditure is quite high in this country.
The countries that have in summary the highest levels are Switzerland, Sweden, Italy and Belgium. In Switzerland the share of governmental spending on total health care financing is very low although the share of governmental health care spending on total governmental spending is very high. That refers to the fact that the private share of welfare spending is generally high in Switzerland. Although Sweden has high values on the other performance variables it has a relatively low value on the variable "Equality of child survival". Sweden is thus somewhat like the opposite of Greece because health is quite unequal distributed although the per capita expenditure, the "Healthy life expectancy" and the "Responsiveness of the system" levels are high. In Italy the share of governmental expenditure on total health expenditure and the share of governmental health expenditure on total governmental expenditure is very high. HALE and the "Equality of child survival" levels are high but the "Fairness of financial contribution" level is low.
Austria seems to be one of the typical countries of this group with high scores on "Equality of child survival", HALE and "Fairness of financial contribution", high or middle per capita expenditure but a small share of ••54
governmental health expenditure on total governmental expenditure. Similar to this case is Finland with especially high values on the "Fairness of financial contribution" variable but a small share of governmental health expenditure on total governmental expenditure. Nevertheless the governmental share of total health expenditure is in Finland very high. Thus the private spending part in Finland is less important than in Austria.
Spain is in a special position. On the one hand the government emphasises the importance of the health care spending. The governmental expenditure on health as a share on total government expenditure is the highest in comparison to the other countries of the group but the level of "Responsiveness of the system" is relatively weak. This may be due to the fact that the Spanish Health care system is tax financed and thus the state spends much more money than in the countries with a social insurance system.
In general the countries of the Group1 do not belong to similar welfare systems according to the Western European welfare classification although all of them have along tradition of capitalism. None of them is an ex-communist state.
Short description of Group1: Good health care performing low state interventionist group with especially high average "Healthy life expectancy".
••55
5.1.2.2
Analysis and definition of Group2
This group includes countries with high factor 1 and high factor 2 values: • Andorra • Denmark • France • Germany • Iceland • Ireland • Luxembourg • Norway • Portugal • Slovenia • United Kingdom
Per capita Total expenditure on Health at International Dollar rate The group has on average the highest per capita total expenditure on health (2140 Dollars). The minimum per capita total expenditure at International Dollar rate is 1462, the maximum 2754. That means that on average for a citizen of Germany 1292 Dollars more are spent for health care than for a citizen in Slovenia. Nevertheless the standard deviation in comparison with the other groups is in the middle. The average score is positioned in the "middle value group" while the majority of the countries is positioned in the "high value group". The countries with the highest values are Germany, Luxembourg and Iceland. The ones with the lowest values are Andorra, Portugal and Slovenia. As Group2 shares high per capita total expenditure with Group1 the variable is not exclusively featuring the group. ••56
Per capita Government expenditure on Health at International Dollar rate The average per capita government expenditure on health is the highest of all groups (1738 International Dollars). The difference to Group1 is quite high. The majority of the countries and the average score are located in the "high value group". But there are quite large differences between the countries. While the Portuguese state spends 1045 Dollars per capita Luxembourgian citizen can expect 2518 Dollars of government health expenditure. Therefore the variance of this group on this variable is the highest in comparison with the other groups. Countries with the highest per capita government expenditure are Luxembourg, Iceland and Germany. The ones with the lowest government per capita expenditure are Andorra, Slovenia and Portugal.
"Healthy life expectancy" (HALE) Although all members of this group have very good values, people living in the country with the highest expectancy – France – can expect to have a "Healthy life expectancy" of 71.254 years, while the ones living in Portugal can expect to live 66.62 years in a healthy state. Nevertheless the standard deviation of Group2 is the lowest of all groups. All countries are positioned in the "high value group".
The countries with the highest values are France, Iceland and interestingly Andorra. The ones with the lowest values are Ireland, Slovenia and Portugal. All in all it is difficult to differentiate Group1 from Group2 along that variable but a high level of HALE still features Group2.
"Equality of child survival" The average level of "Equality of child survival" in the states of Group2 is the highest of all groups. The average score and the ••57
majority of the countries are positioned in the "high value group". The standard deviation is nearly as low as in Group1. On this variable Iceland, Portugal and Andorra again belong to the strugglers. The ones with the best values are Norway, the United Kingdom and France. Because of the similarities with Group1 the variable is not exclusively defining this group.
"Fairness of financial contribution" to health systems The fairness of the financial contribution level is on average the highest within Group2 in comparison to the other groups. The standard deviation is the lowest of all groups. All countries and thus the average score are positioned in the "high value group". The strugglers are again Andorra, Portugal and Slovenia. The outriders are Luxembourg, Denmark, Ireland and Germany. The variable features the group because of the low standard deviation.
Responsiveness-Level of health systems Group2 has the highest on average value on the variable "responsiveness of the health systems". The majority of the countries and the average scores are located in the "high value group". The standard deviation is middle. The outriders are Luxembourg, Denmark and Germany. The countries with the lowest scores are Andorra, Portugal and Slovenia. The high levels on this variable are shared with Group1.
General Government expenditure on Health as a Share in Total expenditure on Health Group2 has the second highest general government expenditure on health as a Share in total expenditure on health of all groups. On average the governments of this group spend 80.74 percent of total health expenditure. Group3 has the highest value on this variable but it is also very close to the average score of Group2. The difference to Group1 is high and the difference to Group4 is ••58
even higher. The standard deviation of Group2 is the lowest of all groups. All countries belong to the "high value group". The countries with the highest values are Luxembourg, Andorra and Norway. The strugglers are Ireland, Germany and Portugal. Thus there are countries like France and Germany with a high per capita expenditure but a relatively low state engagement in the funding of health care. Countries with a "universal" – may they belong to a socialist or liberal welfare model – health system like Norway, Iceland, Denmark and the United Kingdom generally show a higher share of governmental expenditure on total expenditure than the typical Bismarckian countries. The variable is crucial to feature Group2 because of the difference to Group1 and the similarities with Group3. Furthermore the low standard deviation refers to a coherent "behaviour" of Group2 countries.
General Government expenditure on Health as a Share of total Government expenditure Group2 has on average the highest general government expenditure on total government expenditure. That means that in policy making health care plays a large role in these countries. The topic is – at least in terms of the financing – emphasised by governments. The majority of the countries and the average score is positioned in the "high value group” – the standard deviation is middle. Luxembourg, Portugal and Denmark lag behind the other countries of the group. Iceland, Andorra and Germany have the best values.
Summary
Group2 has generally very high values on all eight variables. Except for the variables HALE and "Government expenditure on Health as a share of Total health expenditure" it has the highest scores of all groups on the ••59
variables. The most crucial variables are the "Per capita government expenditure" – because of the difference to Group1 – the "Fairness of financial contribution" – because of the low standard deviation – and the "General government expenditure on health as a share in total government expenditure" – because of the high values. Member states of this group thus are above all characterized by high per capita government health expenditure and a general emphasis of the state responsibility for health care financing.
The countries with the lowest values on the eight variables are Portugal, Andorra and Slovenia. Portugal has a somewhat exceptional position because it has relatively low scores on per capita health care spending (both total and governmental) and relatively low scores on "Equality of child survival", "Responsiveness of the system" and "Fairness of financial contribution". Citizens of Andorra can expect to live for a very long time a healthy life.
The countries with the highest scores on the eight variables are Luxembourg, Iceland and Germany. Luxembourg has besides general very high values on the performance variables a low value on the level of governmental health expenditure as a share of total government expenditure. In this regard it is quite similar to the countries of Group1. Iceland has the highest degree of governmental health care spending in relation to total government expenditure and is thus a more "typical" representative of Group2. But Iceland has a relative low level of "Equality of child survival". Germany has high values on the responsiveness level and high per capita expenditure (both governmental and total) but a relatively low share of government health care expenditure in relation to total health care expenditure. Thus the private financing part of the health system is crucial in Germany. The Danish health care system is very fair in terms of financial contributions and the system is very responsive. Thus the clients can expect to be treated with respect and dignity. But Denmark shows the lowest level in governmental health expenditure as a share of total governmental expenditure. Ireland finally ••60
has a highly fair system in terms of the financing but the governmental expenditure on health as a share of total health expenditure is quite low.
Short description of Group2: Highest health performing and high state interventionist group with especially fairly financed and responsive systems.
5.1.2.3
Analysis and definition of Group3
This is the group having low values on factor 1 and high values on factor 2: • Belarus • Croatia • Czech Republic • Estonia • Hungary • Lithuania • Macedonia • Moldova • Russian Federation • Slovak Republic
Per capita Total expenditure on Health The total capita expenditure on health of this group is on average and in comparison with the first two groups very small. On average 538 Dollars per capita are spent, which is somewhat four times less than in Group2. All countries belong to the "low value group". But there is a quite large standard deviation within the group. The maximum per capita total spending is 1031 Dollars (Czech Republic) and the minimum (Moldova) 64 Dollars. But Group3 spends on average more than twice as much as Group4. ••61
The Czech Republic, Hungary and the Slovak Republic belong to the outriders. Moldova, Russia and Macedonia to the strugglers. The variable is important to describe the group because of the differences to Group1 and Group2. Nevertheless the group is quite similar to Group4.
Per capita Government expenditure on Health The average per capita government health expenditure is on average less than in the first two groups (443 International Dollars) and in general very small. The standard deviation within the group is small. But if one looks at the minimum per capita government spending (53 Dollars in Moldova) and the maximum (942 Dollars in the Czech Republic) the differences are crucial. The majority of the countries and the average score belong to the "low value group". The difference on average per capita spending in comparison to Group4 is remarkable. Countries with the highest values are the Czech Republic, Hungary and the Slovak Republic. The ones with the lowest scores are Russia, Macedonia and Moldova.
"Healthy life expectancy" (HALE) The average "Healthy life expectancy" of this group is much lower on average (61.365 years) than in the countries of the first two groups. Citizens of countries of Group3 can expect to live more than eight years less in a healthy state than the ones of Group1 and Group2. The standard deviation within group is the highest of all groups. This is due to the fact that the difference between countries within this group like the Czech Republic (with HALE of 66.615 years) and countries like the Russian Federation (with HALE of 56.682 years) is very big. There is thus a difference of somewhat 10 years of HALE within this group. The average score is positioned in the "middle value group". The outriders are the Czech Republic, the Slovak Republic and Croatia. The countries with the lowest scores are Belarus, Moldova and the ••62
Russian Federation. Because of the high standard deviation within group the variable does not give specific information about Group3 although the difference to Group1 and Group2 is remarkable.
"Equality of child survival" On average this group has much lower values than the first two groups. Health is thus less equally distributed within society. But the majority of the countries and the average score are positioned in the "high value group". This is due to the fact that the countries of Group4 haven even crucially lower scores than those of Group3. The standard deviation within this group is middle. The "best" countries are the Czech Republic, then Croatia and the Slovak Republic. The ones with the lowest values are Moldova, the Russian Federation and Macedonia. The variable is generally very important because the difference between this group and Group4 is high.
"Fairness of financial contribution" to health systems There is a big difference between this group and Group1 and Group2. The average score of Group3 is positioned together with Group4 in the "middle value group". The differences within group are remarkable. Five countries (the Czech Republic, Belarus, the Slovak Republic, Hungary and Croatia) belong to the "high value group", four countries to the "middle value group" and one country (Russia) to the "low value group". The ones with the highest values are the Czech Republic, Belarus and the Slovak Republic; the ones with the lowest scores are Estonia, Moldova and the Russian Federation. The variable does not give specific information about the group because the similarities with Group4 are obvious.
••63
Responsiveness-Level of health systems There is a huge and clear separation in terms of the level of responsiveness of the health systems between this group and the first two groups. The average score of this group is positioned in the "low value group". But the differences between the maximum and minimum values are quite high within this group – the standard deviation is the highest of all groups. The difference to Group4 is not very big but remarkable. The countries with the highest values are the Czech Republic, the Slovak Republic and Hungary. The ones with the lowest levels are Lithuania, Macedonia and Moldova. In general the variable helps to define the group because of the big differences to the Group1 and Group2 and the remarkable difference to Group4.
General Government expenditure on Health as a share in Total expenditure on Health Group3 has on average the highest share of government expenditure on total health expenditure of all groups. 81.26 percent of health expenditure is public expenditure. Thus only one fifth of the total health expenditure is spent privately. The group is in this regard very similar to Group2. All countries belong to the "high value group". The standard deviation within group is middle. The differences between countries of the group are remarkable. The country with the largest share of government expenditure on total expenditure in this group (Czech Republic with 91.4 percent) and the one with the smallest share (Lithuania with 72.4 percent) differ a lot in health care policy making. Countries with the highest values are the Czech Republic, the Slovak Republic and Croatia. The ones with the lowest levels are Hungary, the Russian Federation and Lithuania. As this variable gives some information about the public/private separation in policy-funding it shows which countries rely more on the markets and which ones have a high state intervention. In the case of Hungary it is interesting that although the country has ••64
relative good values on the performance variables the state is not as much involved as in countries like the Czech Republic with similar high performance values. The variable is important to characterize the group because it differentiates it from Group4.
General Government expenditure on Health as a Share in total Government expenditure The average share of governmental health expenditure on total government expenditure is in the "middle value group". Although its average score is positioned together with the average score of Group1 in the "middle value group" the group has the most similar values with Group2. The difference between Group3 and Group4 is much bigger than the difference to Group1 or Group2. The countries with the highest value are the Slovak Republic with (19.4 percent of total governmental is spent on health), Macedonia, and the Russian Federation. The ones with the lowest values are Estonia, Hungary and Moldova. The variable is important because of the difference to Group4 although the difference to Group1 and Group2 are quite similar.
Summary
The values of this group on the variable "Equality of child survival" are exceptional. They are much lower than those of Group1 and Group2 but much higher than those of Group4. Thus in comparison with the other performance variables health is quite equally distributed in the countries of Group3 although in a general comparison the levels are not high. Furthermore the degree of state intervention (as examined in the variable "Governmental health expenditure as a share of total health expenditure") is important to characterize the group because it has on average on this variable the highest values of all groups although the values are similar like those of Group1.
••65
In General Group3 is characterized by relatively low performance levels and a relatively low per capita expenditure but a high level of state intervention in health care financing. The average value of "Governmental health expenditure as a share of total governmental expenditure" is similar like the one of Group2 but also similar like the one of Group1. In comparison with Group4 one could assume that in those countries with a relatively low per capita expenditure and low performance levels there is a big difference in the degree of the involvement of the state in the funding of health care. But the countries with higher state intervention (Group3) do not crucially differ in terms of "Fairness of financial contribution" from those with less state intervention (Group4). Thus one could assume that in countries with low per capita health expenditure but high level of state intervention (Group3) the expenditure is too low to have an impact on fairness of the financing system. In other words: The attainment of "redistributive justice" seems to depend much more on the general level of expenditure than on the private/public division in the financing system.
The most important difference to Group1 is that it has much better values on the performance levels than Group3. The countries of Group2 have although they have a general lower involvement of the state in the financing system much better values on the performance values and spend much more money per capita on health.
The countries with the highest values on the eight variables are the Slovak Republic, the Czech Republic, Hungary and Croatia. In the case of the Czech Republic and the Slovak Republic the relatively high values on the performance variables are connected with the high degree of state intervention in these state. In the case of Croatia this is true for the variable "Governmental expenditure on health as a share of total health expenditure". The case of Hungary is different: Hungary has a relatively high per capita expenditure and a relatively responsive health system but – relatively to other countries of this group – a low level of state intervention. ••66
The countries with the lowest values on the eight variables are Moldova, Russia and Macedonia. Although the performance and per capita spending in Russia and Moldova are very low the share of governmental health expenditure on total government expenditure is high (this is not the case in Moldova). Thus the governments emphasise the importance of health care funding in relation to other public funded policies but the impact on the performance of the system or the health in general is not visible. In the case of Belarus the very low level of HALE seems to be completely disconnected from the relatively fair system of financial contributions. All in all the countries of Group3 are all ex-communist countries without a long experience in capitalism and democracy in the 20th century. The relatively high state intervention in the funding of health is thus based on the socialised welfare systems during the communist era.
Short description of Group3: Low performing group with high state spending as a shore of total health expenditure and relative equally distributed health.
••67
5.1.2.4
Analysis and definition of Group4
The group has low values on factor 1 and low values on factor 2: • Albania • Armenia • Azerbaijan • Bosnia-Herzegowina • Bulgaria • Georgia • Latvia • Poland • Romania • Turkey • Ukraine • Yugoslavia
Per capita Total expenditure on Health at International Dollar rate In Group4 very few money is spend per capita for health care. The average amount is 248 Dollars. The standard deviation within the group is very small (the smallest of all groups on this variable) – 139 Dollars. The difference between the average values of Group3 and Group4 is not very big (290 Dollars) but still it is twice more than the average spending in Group4. All countries belong to the "low value group". The ones with the highest levels are Poland, Latvia and Turkey. The ones with the lowest levels are Ukraine, Albania and Azerbaijan. The variable is important to define the group but the levels are in a general comparison similar to those of Group3.
••68
Per capita Government expenditure on Health at International Dollar rate The average per capita government expenditure is very small (150 Dollars). The standard deviation within the group is not very large. The difference between Group4 and Group3 is also quite small (293 Dollars). The standard deviation within the group is the smallest of all the groups. Thus all countries of this group show similar values and all countries belong to the "low value group". Poland, Latvia and Turkey have the highest values. Albania, Azerbaijan and Georgia the lowest levels.
"Healthy life expectancy" (HALE) The average level of HALE is very low in this group (59.979 years). The standard deviation within the group is middle (3.056 years). Citizens living in Poland can expect to have on average a "Healthy life expectancy" of 64.35 years. Citizens living in Azerbaijan one of 52.83 years. The difference is thus more than eleven years. The majority of the countries and the average score are positioned in the "middle value group" like in Group3. The difference to Group3 is quite small. The countries with the highest levels are Poland, Bulgaria and Bosnia-Herzegovina. The ones with the lowest levels are Armenia, Ukraine and Azerbaijan. The variable does not crucially differentiate the Group3 and the Group4.
"Equality of child survival" The average level of "Equality of child survival" in the states of this group is very low. The difference towards the other groups is quite high. The average score and the majority of the countries of Group4 are alone positioned in the "middle value group". The "low value group" does only enfold three countries – all of them belong to Group4. The standard deviation within the group is the highest of all groups on this variable. Poland and the Ukraine are for example members of the "high level group". Poland is a strong ••69
outlier on this variable because it shows together with some states of Group2 the highest value. The countries with the best values are Poland, Ukraine and Bulgaria. Azerbaijan, Turkey and Albania are the strugglers. Generally the variable seems to be very important to characterize the group because of the difference to the other groups.
"Fairness of financial contribution" to health systems Group4 has similar average values than Group3 – the values are very low. The standard deviation is the highest of all groups. The majority of the countries and the average score are located in the "middle value group". The countries with the best values are Turkey, Romania and Bosnia-Herzegovina. The ones with the lowest levels are Bulgaria, Albania and Armenia. The variable does not crucially differentiate Group3 and Group4.
Responsiveness-Level of health systems The level of responsiveness of the health system is similarly low in this group like in Group3 but generally even lower. The standard deviation within the group is the highest of all groups. The average score is positioned in the "low value group". Poland, Latvia and Romania have the best values. Albania, Bulgaria and Georgia are the strugglers. Group4 differs not very much in comparison to Group3 on this variable.
General Government expenditure on Health as a Share of total health expenditure The share of state expenditure in total health expenditure is very small in this group (average 57.62 percent). This is on average nearly 10 percent lower than in Group2. The average score is alone positioned in the "middle value group" all other average scores of the other groups are positioned in the "high value group". The standard deviation within the group is the highest of all groups. There are countries like Bulgaria with 77.6 percent ••70
governmental expenditure as a share of total expenditure and countries like Georgia with only 10.5 percent governmental expenditure on total health expenditure. Due to this extreme value Georgia builds alone the "low value group" on this variable. Bulgaria, Turkey and the Ukraine have the highest values of this group. Azerbaijan, Armenia and Georgia have the lowest. Generally the values are in comparison to the other groups very low. The variable is important to characterize the group because it indicates the difference between Group3 and Group4 along the criterion "state intervention in health care funding”. Governments of Group4 do not play a major role in the funding system.
General Government expenditure on Health as a Share in Total Government expenditure The concerned governments do not emphasis health care in terms of their public spending policy. On average only 7.8 percent of public expenditure goes to health care. The standard deviation of this group is the highest of all groups. The average score of this group is positioned alone in the "low value group". The difference between this group and all other groups is very high. The countries with the highest values are Armenia, Yugoslavia, and Poland. The ones with the lowest values are Romania, Azerbaijan and Georgia. The variable is important to characterize the group because of the large difference to the other groups.
Summary
First it is to state that this group has on all variables the lowest values. One of the most important features of this group are the low values on the government health expenditure as a share of total health expenditure and the governmental health expenditure as a share of total government expenditure. The two per capita expenditure variables are important because the within group standard deviation is low. Thus the very low ••71
state intervention and the very low per capita health expenditure are crucial features of the group. Furthermore the group has extremely low values on the variable "Equality of child survival". Therefore health is very unequally distributed in the countries of Group4.
The countries with the highest values are Poland, Turkey and Latvia. Although Turkey has relatively (in comparison to the other countries of this group) high values on the two per capita expenditure variables and on the variables "Fairness of financial contribution" and a relatively high degree of state spending as a share of total health expenditure – health is very unequally distributed within the Turkish society.
The countries with the lowest values are Azerbaijan, Albania and Georgia. There are furthermore some countries with relatively (relative to the other countries) high values on one or two variables and very low values on others. Bulgaria has relatively high values on the variable "Government health expenditure as a share of total health expenditure" and "Equality of child survival" but very low values on "Fairness of financial contribution" and "Responsiveness of the system". The Ukraine has very low values on the "Per capita total expenditure on health" and "Healthy life expectancy" but relatively high values on "Equality of child survival" and "Government health expenditure as a share of total health expenditure". The government of Romania spends very few money on health care in relation to other public expenditure but the responsiveness level and the level of fairness of contribution is relatively high in Romania. Armenia has very low values on "Healthy life expectancy", the level of "Fairness of financial contribution" and the Armenian state pays very small share of the total health expenditure. Nevertheless in relation to other public expenditure a quite high percentage is spent for health care. This fact indicates the assumption that the Armenian state hardly spends any money on health care or for other policies that require public expenditure. All in all nearly all of the states of this group (besides Turkey) are former communist states with very little experience with capitalism and democracy in the second half of the 20th century. ••72
Furthermore a crucial part of the countries still have crucial unsolved political problems – or had them at least in the last decade.
Short description of Group4: Lowest performing with very low state intervention and an especially low score on the equality of distribution of health.
5.2
Analysis of the T-test concerning the differences between the groups of countries in regard to the health concept variables
In the next step the group definition will be refined with the help of the T-test. Thus it will be examined if the differences along the health variables between the groups are significant. If they are the features of the groups can be described more precisely. Although significant differences between all groups on all variables are examined the following questions are crucial:
• Do Group1 and Group2 differ significantly on the variable "Healthy life expectancy", "General government expenditure on health as a share of total expenditure on health", "General government expenditure on health as a share of total government expenditure" and "Per capita governmental health care expenditure"? • Do Group1 and Group4 differ significantly on the variables "General government expenditure on health as a share of total expenditure on health" and "General government expenditure on health as a share of total government expenditure"? • Do Group3 and Group4 differ significantly on the variable "Equality of child survival" and "Fairness of financial contribution"? ••73
• Do Group2 and Group3 differ significantly on the variable "General government expenditure on health as a share of total government expenditure"?
How is a significant difference shown in the T-test?
The p-value measures the significance level of the T-test. If p is greater than 0.05 there is more than a 1 in 20 chance that the two population means are actually the same and that the difference between the two sample means only reflects sampling error. If the p-value is smaller the probability that the sample means differ is significant. The dataset and the T-test are attached in the annex ("healthvariablesalonggroups.xls ", "healthvariablesalonghealthgroups.sav ", "outputofttesthealthvariablesalonghealthgroups.spo") In order to compute the Ttest a new variable has been built that orders the countries into the groups according to the principal components analysis of the health concept.
Let us look on the output:
1. Group1 and Group2 differ significantly on the variables "General government expenditure on health as a share of total expenditure on health", "General government expenditure on health as a share of total government expenditure" and "Per capita governmental health care expenditure" but not on the variable "Healthy life expectancy". Thus the higher values on "Healthy life expectancy" of Group1 in comparison to Group2 cannot be perceived as an important feature of Group1. The only important – because of the significance – difference is the degree of state involvement in the financing of the health care system. 2. Group1 and Group3 differ on all variables significantly. 3. Group1 and Group4 differ on all variables significantly besides on the variable "General government expenditure on health as a share in total expenditure on health". Thus they have similar low ••74
values in regard to the degree a state is involved in the financing of the health care system but they do not have similar values concerning the emphasis a state puts on health care financing in relation to other public expenditures. In this regard Group4 has significantly lower values. 4. Group2 and Group3 differ significantly on all variables besides on the variables "General government expenditure on health as a share in total expenditure on health" and "General government expenditure on health as a share of total government expenditure". Thus a clear difference on all the performance variables and the per capita expenditure variables is given. 5. Group2 and Group4 differ on all variables significantly. 6. Group3 and Group4 differ on the variables "General government expenditure on health as a share in total expenditure on health", "General government expenditure on health as a share in total government expenditure", "Equality of child survival", the level of "Responsiveness of financial contributions" and the "Per capita total health expenditure" and the "Per capita government health expenditure". 7. Group1 and Group2 together (the groups with the highest levels of health care performance and high per capita health expenditure) in comparison with a merged group of Group3 and Group4 (the ones with the low performance levels) differ significantly on the variables "General government expenditure on health as a share in total government expenditure", on all performance variables and the two per capita expenditure variables. 8. Group1 and Group4 together (the ones with the lower state involvement in the financing of health care) in comparison with the group with higher state involvement (Group2 and Group3 together) differ significantly on the variables "General government expenditure on health as a share of total government expenditure", "General government expenditure on health as a
••75
share of total government expenditure" and on the variable "Per capita Government expenditure at International Dollar rate".
5.3
Analysis of the T-test concerning the differences between the groups of countries in regard to economic strength
In the next step it is examined if the groups differ significantly in terms of GDP. According to the hypothesis it is assumed that the economic well-being of a state has a crucial influence on the probable relation between the health concept and the democratic governance concept. In the following test the probable influence of GDP on the differences and similarities of the groups along the health concept is examined. The main assumption is that the difference between the Group1/Group2 and the Group3/Group4 is significant. This assumption is very important because it may indicate that the differences in the health care performance are also due to economic strength (GDP). And thus the GDP must be recognized as an extraneous third variable that must be incorporated in further tests and deliberations. The data and the test are attached in the annex ("GDPalonghealthgroups.sav ", "GDPalonghealthgroups.xls ", "outputofttestGDPalonghealthgroups.spo")
The p-values indicate: • Comparison of Group1 and Group2: not significant • Comparison of Group1 and Group3: significant • Comparison of Group1 and Group4: significant • Comparison of Group2 and Group3: significant • Comparison of Group2 and Group4: significant • Comparison of Group3 and Group4: significant ••76
• Comparison of Group1/Group2 and Group3/Group4: significant • Comparison of Group1/Group4 and Group2/Group3: not significant
Group1 and Group2 do not significantly differ in terms of GDP. Thus the differences in health care performance may not be due to GDP. Group1 and the Group2 differ significantly in terms of GDP in relation to Group3 and Group4. If Group1 and Group2 are put together into one sample and Group3 and Group4 as well the two new groups differ significantly in regard to GDP. Thus the huge differences in health care performance and per capita health care expenditure may be due to the influence of the third variable GDP. The groups with a high state intervention (Group2 and Group3) and the groups with a low state intervention (Group1 and Group4) differ not significantly in terms of GDP. The influence of GDP on health care performance and per capita expenditure will be taken into account in the further tests.
••77
6
Testing the relation between the health concept and the democratic governance concept
In this chapter several quantitative tests are computed in order to examine the hypothesis, i.e. it is to prove if the variables of the concept of democratic governance are related with the variables of the concept of democratic governance. In the first test the variables of the democratic governance concept are compressed through a principal components analysis in order to examine if the compressed variables are related with the compressed variables of the principal components analysis of the health concept.
6.1
Analysis of the principal components analysis of the democratic governance concept – Compressing the variables of the democratic governance concept
This second principal components analysis is not made to build and analyze new groups but only to compress the variables of the concept. The data for the principal components test and the output of the test are digitally attached in the annex ("datafordemocraticgovernance.sav",
••78
"datasetfordemocraticgovernance.xls", "outputofdemocracticggovernanceprincipalcomponentsanalysis.spo")
The two components with higher eigenvalues than 1 explain 93.069 of the variance of the variables. The first component contains the variables "Voice and accountability" (with the highest loading), followed by "Rule of law", "Government effectiveness" and "Regulatory quality", "Control of corruption", "Freedom house rate" and finally "Political stability". All variables have very high loadings (higher than 0.9). Factor1 represents general requirements of democratically ruled and efficient governments. The second component contains the variables "Voters turnout in the 1990ies" and "Voters turnout from 1945–1999" and thus represent a measurement of political participation.
2 WITZERLAND FINLAND NETHERLANDS NORWAY SWEDEN IRELAND UNITED KINGDOMDENMARK LUXEMBOURG ICELAND GERMANY AUSTRIA PORTUGAL FRANCE SPAIN ANDORRA HUNGARY BELGIUM ESTONIA SLOVENIA POLAND MALTA LITHUANIA GREECE ITALY CZECH REPUBLIC CYPRUS LATVIA REPUBLIC SLOVAK CROATIA BULGARIA ROMANIA -1 MACEDONIA ARMENIA MOLDOVA YUGOSLAVIA BOSNIA-HERZEGOVINA RUSSIA UKRAINE GEORGIA BELARUS -2 TURKEY ALBANIA AZERBAIJAN
REGR factor score 1 for analysis
1
1
0
-3 -3
-2
-1
0
1
2
REGR factor score 2 for analysis
1
Scatter-plot of the principal components analysis of the democratic governance concept
••79
If one looks at the scatter-plot four groups can be identified. The first group contains countries that load high on the first component and low on the second. The countries are: Estonia, Finland, France, Hungary, Ireland, Lithuania, Netherlands, Poland, Portugal, Slovenia, Spain, Switzerland and the United Kingdom. These countries thus have high standards of democracy but low voter participation. The second group has high standards of democracy and high voter participation. The countries are: Andorra, Austria, Belgium, Cyprus, the Czech Republic, Denmark, Germany, Greece, Iceland, Italy, Luxembourg, Malta, Norway and Sweden.
The third group has low democracy standards but high voter participation: Albania, Azerbaijan, Latvia, Romania, the Slovak Republic and Turkey. The fourth group has low democracy standards and low voter participation: Armenia, Belarus, Bosnia-Herzegovina, Bulgaria, Croatia, Georgia, Macedonia, Moldova, Russia, the Ukraine and Yugoslavia.
A further analysis – although it was undertaken in the research process – will not be illustrated within this study. But it is obvious that there are more ex-communist countries in the first two groups in comparison with the output of the principal components analysis of the health concept.
In the next step it is to examine if the two new factor variables are related to the factor variables of the health concept in order to prove the hypothesis. This will be done by a regression test.
••80
6.2
Analysis of the regression – Proving the influence of the factor variables of the democratic governance concept on the factor variables of the health concept
The data for the regression test and the output of the tests are attached in the annex ("datafortheregression.sav", "f1+f2health+f1+f2democraticgovernance.xls", "outputoftheregressionhealthfactor1.spo", "outputoftheregressionhealthfactor2.spo")
6.2.1
Regression of the factor 1 variable of the health concept and the two factor variables of the democratic governance concept
In this test it is to prove if it is possible to predict the health (system) performance as expressed in the two variables of the factor analysis of the health concept by the democratic governance and participation variables. Thus the hypothesis is: The independent democratic governance variables cause the dependent health concept variables.
The analysis of the test procedure is attached in Annex A – Analysis of the test procedures
••81
Partial Regression Plot Dependent Variable: factor 1 health
3 Switzer land 2 Franc ether lands Ne Ge rma bourg Luxemny rk De nman Nor Finla Austri away nd Swede Ire Be lgium land d Gr eec e Uniten K ingdom Ital y SpaiIcel and Portugal Andorra Cypr Slove nia us Pola nd Ge orgia MaHunga ry lta Czech Republ ic Croa tia Bosnia -He rzegovina Ro mani a Estonia Lithuani a Yugosl avi Be lar us Ukrai ne a Slova k Republ ic Latvi a T urkey Arm enia a Bulga ria Ma cedoni Azerba ij an Mol dova R ussia Alba nia -3 -2 -1 0 1 2
1
factor 1 health
0
-1
-2
factor 1 d emocratic g overnance
Partial Regression Plot Dependent Variable: factor 1 health
1 .5 Ge orgia 1 .0 Switzer land .5
0 .0
factor 1 health
-.5
Italy AzerbaijBe lgium an Gr eec e Luxem bourg Be lar us Ge rma ny Austri a Cypr us Franc e De nma rk Bosnia -He rzegovina Swede n Ne ther lands Nor way Icel and Spai n YugoslUkrai ne avi a T urkey Ireland Unite dR omani a K ingdom Andorra Ma lta Finla nd Slove Arm enia nia Czech Republ ic Portugal Macedoni a Alba nia Croatia R ussia Mol dova Pola nd Slova k Republ ic Latvi a Hunga ry Lithuania Bulga Estonia ria
-1.0 -1.5 -3 -2
-1
0
1
2
factor 2 d emocratic g overnance
The democratic governance factor 1 seems to predict part of the health factor 1 while the democratic governance factor 2 predicts much less. That means that the democratic governance 1 factor seems to cause – at least partly – the health factor 1. Generally there seems to be a problem with heteroskedacitiy in the model. ••82
6.2.2
Regression of the factor 2 variable of the health concept and the two factor variables of the democratic governance concept
In this step the regression between the independent democratic governance factor variables and the factor 2 variable of the health concept is done. This is the variable that represents the degree of governmental responsibility for the funding of a public health care system. Thus the hypothesis for the procedure is: The independent democratic governance variables cause the dependent factor 2 health variable.
The hypothesis must be rejected. The democratic governance factor variables do not predict the health factor 2 variable. As the health factor 2 variable represents the level of influence and responsibility of the government in the funding of health care one can summarize that there is no relationship between the level of governmental funding responsibility and the democratic governance scores. This result indicates that the comparison of the four groups should be thought over. Why? As there is no effect of the democratic governance variables on health factor 2 variable it is indicated that a comparison of the groups along the latter variable does not make sense. The variables that build that factor (General Government expenditure on Health as a share in Total Government expenditure and General Government expenditure on Health as a Share in Total expenditure on Health) are thus excluded in the next steps of the study. And the comparison between the groups of countries will be thus based on just two groups divided by the health factor 1 variable. I.e. Group1 and Group2 are merged to a new Group1 and Group3 and the Group4 are merged to a new Group2.
The result of the regression analysis indicates that further tests have to be done to examine the relationship between the factor 1 health variable and the democratic governance variables. As the result of the regression of the factor 2 variable of the democratic governance concept and the health factor 1 variable is very weak the voter participation variables may be be excluded from further analysis. In the next step a simple correlation test is done to examine what kind of variables of the two concepts are significantly correlated. As the regression test indicated that the division along the health factor 1 variable makes sense to examine ••83
the relation between the two concepts the correlation test is done along the new Group1 (former Group1 and Group2) and the Group2 (former Group3 and Group4).
In the correlation test another important assumption of the thesis is included. As pointed out in the hypothesis the relation between democratic governance and health may be crucially influenced by the overall economic performance or strength. Thus a partial correlation is computed that allows testing the correlation between democratic governance and health by simultaneously controlling the influence of economic well-being. The output of this partial correlation thus reflects the relation between the variables of the concept of democratic governance and the variables of the concept of health separated from the influence of overall economic well-being.
6.3
Partial correlation test – Measuring the relation between the health concept variables and the democratic governance variables along the groups of the principal components analysis by controlling the influence of economic strength
First a partial correlation of all 44 countries between the variables of the health concept and the variables of the democratic governance concept by holding GDP constant is computed. The test thus shows if there are correlations between the variables of the two concepts besides the influence of GDP. The output is shown in the attached files "outputofpartialcorrelationallcountries.xls" and "outputpartialcorrelationallcountries.spo" in the annex.
••84
6.3.1
Partial correlation test of all countries between the variables of the health concept and the variables of the democratic governance concept by holding GDP constant
There are only a few significant correlations. None of them are really strong. The strongest one is the one between "Freedom house rate" and "Healthy life expectancy".
Partial correlations along their strength: • Freedom house rate/Healthy life expectancy • Voice and accountability/Healthy life expectancy • Voice and accountability/Equality of child survival • Control of corruption/Healthy life expectancy • Political stability/Equality of child survival • Political stability/General government expenditure on health as a Share of Total expenditure on health • Government effectiveness/Healthy life expectancy • Rule of law/Healthy life expectancy • Rule of law/Equality of child survival • Control of corruption/Equality of child survival • Regulatory quality/Healthy life expectancy • Control of corruption/Responsiveness • Government effectiveness/Equality of child survival • Freedom house rate/Equality of child survival • Regulatory quality/Equality of child survival • Government effectiveness/ Responsiveness • Political stability/General Government expenditure on Health as a Share in Total Government expenditure • Rule of law/Responsiveness of the Health system
••85
The democratic governance variables are most frequently correlated with "Healthy life expectancy" and "Equality of child survival". The ones with the highest correlations are "Control of corruption" and "Voice and accountability". There are no significant correlations between health variables and voter turnout. Part of the hypothesis, i.e. that the level of political participation is related to health outcomes must be rejected due to this test. The variables used in this study do not indicate such a relationship. The variables that measure voter turnout will be thus not computed anymore in the final test. The variables that measure the share of governmental health expenditure are only weakly correlated with Political stability. As already mentioned in the analysis of the regression the relationship between democratic governance and the responsibility of a government in the funding of health care seems to be very weak or even does not exist at all. Furthermore the per capita expenditure variables have only weak or no relationship at all with the democratic governance variables. That certifies that this part of the hypothesis must be rejected that proposes a relationship between the level of democratic governance and per capita expenditure on health. As shown in the test "Healthy life expectancy" is correlated with the "Freedom house rate", "Voice and accountability", the "Control of corruption", "Government efficiency", "Rule of law" and "Regulatory quality". That indicates that in countries with high levels of freedom in terms of political rights and civil liberties, where citizens have a high ability to control the governments, where the corruption is limited by an efficient controlling system and the judiciary system is well elaborated, the expectancy to life a healthy life is also high. "Equality of child survival" is correlated with "Political stability", "Voice and accountability", "Rule of law", "Control of corruption", "Government efficiency", the "Freedom house rate" and the "Regulatory quality". Thus in countries with a high Political stability and efficient control of governments, an elaborated judiciary system and generally efficient governance health is relative equally distributed. Finally the level of responsiveness of the health system is very weakly correlated with "Control of corruption", "Government efficiency" and the "Rule of law". That may indicate that in countries with low levels of corruption and a reliable law system health care is more client orientated than in those countries with high corruption and less governmental efficiency.
In the next step another partial correlation test is made to examine if and what kind of correlations between health variables and democratic governance variables are significant within the two groups – called Group1 and Group2 that divide countries along factor 1 according of the health concept. ••86
6.3.2
Partial correlation test within Group1 and Group2 between the variables of the health concept and variables of the democratic governance concept by holding GDP constant
The dataset and the output are digitally attached in the annex ("datasetpartialcorrelationGroup1.sav", "datasetforpartialcorralonggroups.xls", "outputpartialcorrelationGroup1.spo", "datasetpartialcorrelationGroup2.sav", "outputpartialcorrelationGroup2.spo")
Within Group1 there is only one significant correlation (with a p-value <0.05): The variables "Voice and accountability" and "Fairness of financial contribution" are significantly (p-value= 0.012) related with a correlation coefficient of 0.5353.
The scatter plot only shows the simple correlation but not the partial correlation by holding GDP constant. Nevertheless it is informative.
1 .8 De nma rk Finland Nor way Switzer land Sweden Ne ther lands 1 .6 Icel and Ge rmany Unite d K ing Be lgium Andorra Luxem bourg Ire land Portugal Ma lta Austria Franc e Spai n
1 .4
Voice and accountability
1 .2 Slove nia Ital y Gr eec e 1 .0 Cyprus
.8 .91
.92
.93
.94
.95
.96
.97
.98
.99
Fairn ess of financial c ontribution to h ealth systems, estimates for 19 97
••87
The correlation is not very strong. What does this correlation indicate?
In European countries with high per capita expenditure on health, a generally good health care system and a high level of health citizen's ability to participate in the selections of governments is related with the fairness of the financial contribution to the health care system. As fairness of financial contribution measures if households become impoverished by contributing to health care and as this variable also contains measurements of progressivity "Voice and accountability" is correlated with elements of equalising redistribution from the rich to the poor.
Interestingly there is no other significant correlation between the variables of the two concepts. It is to summarize that within this group of countries having a high level of health performance democratic rules have no influence on the general level of health, the equality of health or the client orientation of the system but on the fairness of the financial flows of the system.
Within Group2 there are three significantly correlated variables. The highest correlation is the one between the "Freedom house rate" and "Healthy life expectancy" with a correlation of -0.667 and a significance value of 0.001. A high level of civil liberties and political rights is positively correlated with people's "Healthy life expectancy" in these countries. As the countries of this group are characterized by relatively low levels of health care performance and low per capita expenditure on health the differences between the levels of civil liberties and political rights are much bigger than in Group1. There are thus countries with high level of political rights and civil liberties like the Czech Republic, Estonia, Hungary, Lithuania, the Slovak Republic, Bulgaria and Latvia and ones that can only be described as "partly free" like Azerbaijan and Belarus. These differences correlate with the differences in the amount of years expected to live a healthy life.
••88
7
6 Azerbaijan 5
Belarus
Russian Federation
Freedomhouse Rate2003
4
Ukraine Armenia Georgia Bosnia and Herzegovi Moldova Turkey Macedonia, FYR Yugoslavia, Fed. Rep
3
Albania
2
Romania
Croatia
Latvia Hungary Lithuania Bulgaria Republic Republic Estonia Slovak PolandCzech 1 52
54
56
58
60
62
64
66
68
Healthy life expectancy (HALE), estimates for 2001 [WHO]
Those countries that will join the European Union next year generally show good values on both coordinates. Other countries like Macedonia, Yugoslavia, Romania and Croatia having lower "Freedom House" levels show quite high values on "Healthy life expectancy". Furthermore there is a group of countries comprising Ukraine, Armenia, Georgia, Bosnia, Moldova and Turkey that lag behind on both variables. Finally there are the very strugglers: Azerbaijan, Belarus and Russia.
Generally one can state that political rights and civil liberties are positively correlated with "Healthy life expectancy" within Group2. In contrast to Group1 within Group2 the degree of political and civil freedom is related to "Healthy life expectancy" besides the fact that "Healthy life expectancy" depends on the economic strength of a state.
The next scatter plot shows the second significant correlation between "Voice and accountability" and "Healthy life expectancy".
••89
1 .5 Hunga ry Pola nd Estonia Slova k Republ ic Republ ic Latvi a C zech Lithuani a Bulga ria Croa tia
1 .0
.5
R omani a
Voice and accountability
0 .0
Alba nia Yugosl avi a, Fe d. Rep Bosnia and He rzegovi Ma Mol dova Ge orgia cedoni a, FYR Arm enia urkey R ussian Fe dera tion T Ukrai ne
-.5
A zerba ij an -1.0 Be lar us 52 54 56 58 60 62 64 66 68
-1.5
Healthy life expectancy (HALE) , estimates for 200 1 [WHO]
The correlation coefficient is 0.5587 and the p-value is 0.008. The correlation is thus quite weak but significant. Within the countries of Group2 higher or lower levels of "Voice and accountability" thus correlate with higher or lower levels of "Healthy life expectancy". There are some countries with a relative high level of "Voice and accountability" e.g. Poland, the Slovak Republic, Latvia and the Czech Republic and there are countries with very low levels of "Voice and accountability" like Azerbaijan, Russia and the Ukraine. Generally the levels of "Voice and accountability" and the "Healthy life expectancy" of Group2 are lower than those of the Group1. The correlation between the variables "Freedom house rate" and "Healthy life expectancy" may refer to a process of democratisation. The Freedom house rate represents the degree of political and civil freedom. The higher the Freedom house rate the more democratic power can be established in a country. The variable "Voice and accountability" is from a logic point of view connected with the general political freedom. It measures the ability to select governments and the degree of control over political power by a free press. Together the "Freedom house rate" and the "Voice and accountability" measure the legal and structural framework for the functioning of democratic control and its factual functioning. HALE is related to the ability of citizens to ••90
select the power by which they are governed. The correlation of these variables with "Healthy life expectancy" thus indicates the following assumption: The more citizens are capable to influence their governments, to control them and the more these governments are controlled by a free press exercising its watch dog function the higher is the expectation to live a healthy life. But as a simple correlation does not indicate causality one can only conclude that the process of democratic power control and people's health is related. However, freedom, liberties and political rights – thus the ability to be not subjected to arbitrary powers but to have power oneself – is connected to the wellbeing of people as expressed in the „Healthy life expectancy" variable.
Let us turn to the last significant correlation within Group2. The correlation between "Control of corruption" and "Healthy life expectancy" with a correlation coefficient of 0.5511 and a p-value of 0.01.
1.0 Estonia Hungary .5 Poland Czech Republic Slovak Republic Lithuania Croatia Latvia 0.0 Romania Turkey Bulgaria
Control of corruption
-.5
Azerbaijan -1.0
Armenia Belarus Albania Moldova Russian Federation Ukraine Georgia
Bosnia and Herzegovi Macedonia, FYR Yugoslavia, Fed. Rep
-1.5 52
54
56
58
60
62
64
66
68
Healthy life expectancy (HALE), estimates for 2001 [WHO]
In countries where public power is abused for private gain and where the lack of respect for the rules of governance is dominant healthy life ••91
expectancy is lower than in those countries of Group2 where corruption is effectively controlled. The correlation is very weak and thus not that important.
Within Group2 democratic governance variables that matter in relation to health performance and health are the degree of political rights and civil liberties, the ability of citizens to select their governments and the control of corruption. These variables are only significantly correlated with one variable of the health concept ("Healthy life expectancy") by holding GDP constant.
Most aspects of democratic governance in these countries are not correlated with health care performance, the expenditure on health or the equality of distribution of health but only with the level of health. A possible explanation may be that the expenditure on health in these countries is too limited to have an impact on fairness of financial contribution or other performance variables.
The correlation between aspects of democratic governance and aspects of health differ in Group1 compared with Group2. In Group1 the correlation between voice and accountability and fairness of financial contribution indicates that in the countries with a good health care performance and a high level of health citizens’ ability to select governments, to control the ruling power is linked to the fairness of redistributive justice through health care. The equality aspect of health systems is more emphasized in countries with high "Voice and accountability" than in those with low "Voice and accountability". Fairness in the financing of health care is only linked to "Voice and accountability" in countries where good health care performance and high levels of health are already established. In contrast within Group2, i.e. within societies where the level of health, the system performance and the expenditure on health are low – citizens' ability to select the ruling power, the degree of political freedom and the control of corruption are linked to the level of health. This link has little to do with concepts of equality but more generally with the basis of wellbeing.
••92
7
Summary and final conclusions
In the first part of the analysis groups have been built along the health concept variables. By the division of the 44 countries along the two components of the principal components analysis of the health concept four groups have been formed. The evaluation and definition of these four groups along the differences on the health concept variables and along economic strength offers an interesting and partly surprising picture of health care performance standards, the level of health and the degree of governmental influence in the funding of health care within Europe. In the second step of the study the hypothesis is examined. As the regression indicates that the analysis along the second factor of the principal components analysis of the health concept does not make sense for the examination of the hypothesis only two groups (that are divided along the first factor of the principal components analysis of the health concept) are compared. Some aspects of the hypothesis have to be rejected: As shown in the analysis of the regression test the degree of governmental responsibility in the funding of health care is statistically not related with aspects of democratic governance. Furthermore – as the partial correlation within all countries indicate – there is no statistically significant correlation between political participation, i.e. voter turnout and aspects of the health concept. By including the probable influence of the overall economic strength of a country when measuring the relation between the two concepts in the two final groups very few statistically significant correlations can be found. Within the first group "Voice and accountability" is statistically related with the "Fairness of financial contribution". As a correlation test does not indicate a causal relation ••93
further research may offer more precise information about this link. In this study it is shown that in countries with on average high levels of health, a general good health care performance and a quite high expenditure on health care the degree of political self-determination (the ability to choose and control the government) is positively correlated with the fairness of the financing of health care. As the fairness variable includes measurements of equalization from the rich to the poor the correlation indicates that democratic power and fairness (as redistributive justice) are connected. This relationship only exists within the relatively rich countries, with a quite long tradition of democracy and good health performance levels. Within the second group – the one with the lower levels of health performance – a correlation exists between the degree of political freedom and civil liberties, the "Voice and accountability" variable, "control of corruption" and "Healthy life expectancy". Thus in transition countries in South Eastern Europe aspects of democratic power are positively related with the expectancy to life a healthy life. I.e. in countries of this group with relatively low values on the three mentioned democratic governance variables "Healthy life expectancy" is also relatively low. Within countries with higher levels of political freedom and citizen's ability to control governments "Healthy life expectancy" is higher. As partial correlation does not indicate a causal relationship further research should be undertaken in order to examine a probable causality. Principally this study is based on the assumption that democratic governance may influence health aspects and not the other way around. This perception is influenced by theories of political economy that suggest an enhancement of resource allocation through democratic pressure. Furthermore it is influenced by theories of democratic legitimacy (the republican perspective of citizenship theory) and the historical examination of social rights by T. H. Marshall. As the general theory of modernization proposes a causal influence of socioeconomic development on democratization it is furthermore proposed to examine the probable relationship between democratic governance and health along this direction. Although a relation between political participation and health outcomes is not indicated by this study further studies that examine this relation in a longer perspective may be informative. ••94
Finally it is crucial that more data for both concepts for all European countries should be collected and provided. It was for instance not possible to get data referring to party-membership or union density of all European countries. This lack of reliable information reflects the overall missing perspective of a common Europe besides the European Union and besides the past division in East and West.
••95
Bibliography
All internet addresses have been checked the last time on the 17 of June in 2003.
[Ambos & Penz: 2002] Ambos, Björn & Penz, Elfriede (2002): "Advanced Data Analysis", in: International Marketing and Management, Wirtschaftsuniversität Wien. [Baur & Schulze: 2003] Baur, Nina & Schulze, Gerhard (eds.) (2003): "Bivariate Statistik, Drittvariablenkontrolle und das Ordinalskalenproblem. Eine Einführung in die Kausalanalyse und in den Umgang mit zweidimensionalen Häufigkeitsverteilungen mit SPSS für Windows", 2. korrigierte Auflage, Bamberger Beiträge zur empirischen Sozialforschung, Nr. 9. [Birkegaard & Porsgaard: 2002] Birkegaard, Hans Christian & Porsgaard, Rasmus (2002): "Introduction to SPSS", Description IT-Department, Statistical methods in SPSS. [Blendon et al: 2001] Blendon R.J.; Kim, M.; Benson, J.M. (2001): "The public versus the World Health Organization on health system performance", in: Health Affairs, 2001, 20(3):10-20.
••96
[Bontis: 1998] Bontis, Nick (1998): "Intellectual Capital: An Exploratory Study That Develops Measures And Models", published in: Management Decision, Vol 36, Number 2, 63-76. [Braveman et al: 2001] Braveman, Paula; Starfield, Barbara; Geiger, H. Jack (2001): "World Health Report 2000: how it removes equity from the agenda for public health monitoring and policy", in: British Medical Journal 2001; 323: 67881, , [Carey & Judge: 2001] Carey, J., & Judge, D. (2001): "Life Span Extension in Humans is SelfReinforcing: A General Theory of Longevity", in: Population and Development Review 27(3): 411-36. [Darlington: 1997] Darlington, Richard B.: "Factor Analysis", [de Silva: 2000] de Silva, A. (2000): "A framework for measuring responsiveness", GPE Discussion Paper #32, WHO, Geneva, [Díaz-Bonilla et al: 2002] Díaz-Bonilla, E.; Babinard, J.; Pinstrup-Andersen, P. (2002): "Opportunities and risks for the poor in developing countries", Working Paper #83, Indian Council for Research on International Economic Relations, New Delhi, [Easterlin: 1995] Easterlin, R. (1995): "Industrial Revolution and Mortality Revolution: Two of a Kind?", in: Journal of Evolutionary Economics 5(4): 393-408.
••97
[Easterlin: 1999] Easterlin, R. (1999): "How Beneficent is the Market? A Look at the Modern History of Mortality", in: European Review of Economic History 3: 257-94. [Feilmayr: 1997, page 1] Feilmayr, Wolfgang (1997) "Mathematik und Statistik für Raumplaner: Regressions- und Korrelationsanalyse", [Franzese: 1998] Franzese, Robert J. (1998): "Political Participation, Income Distribution, and Public Transfers in Developed Democracies", prepared for delivery at the 1998 Annual Meeting of the American Political Science Association, Boston, [Franzese: 2002] Franzese, Robert J. (2002); "Participation, Veto Actors, and Policy Responsiveness in the Evolution and Reform of Health Care in Developed Democracies", in: Reconstructing the Welfare State (forthcoming), [Freedom House: 2000] Freedom House (2000): "Survey Methodology", [Gakidou & King: 2002] Gakidou, E. & King, G. (2002): "Measuring total health inequality: adding individual variation to group-level differences", in: International Journal for Equity in Health, 2002; 1 (1): 3, [Gauthier: 2000] Gauthier, Anne Hélène (2000): "The promises of comparative research", Paper prepared for the European Panel Analysis Group, published in: ••98
Schmollers Jahrbuch. Journal of Applied Social Science Studies. 122: 530, [Gerstman: 2003] Gerstman, B. (2003): "StatPrimer", [Girishankar et al: 2001] Girishankar, Navin; Hammergren, Linn; Holmes, Malcolm; Knack, Stephen; Levy, Brian; Litvack, Jennie; Manning, Nicholas; Messick, Richard; Rinne, Jeffrey; Sutch, Helen (2001): "Governance", in: Poverty Reduction Strategy Sourcebook, Volume 1 – Core Techniques and CrossCutting Issues; The World Bank, Washington, [Gupta: 2000] Vijay Gupta (2000): "Regression explained in simple terms", in: SPSS for Beginners, VJBooks, [Habermas: 1991] Habermas, Jürgen (1991): "Staatsbürgerschaft und nationale Identität, Überlegungen zur europäischen Zukunft", Erker-Verlag, St. Gallen. [Hanzek: 1998] Hanzek, Matjaz (ed.) (1998): "Health", in: Human Development Report, Slovenia, [Haugaard: 2001] Haugaard, J. (2001): "Is Good Governance the Key to Human Development?", University of Aarhus,
••99
[Heo & Kolacz: 2003] Heo, Giseon; Kolacz, Henry (2003): "Child health and development study", University of Alberta, Department of Mathematical and Statistical Sciences, , [IDEA(a)] International Institute for Democracy and Electoral Assistance: "Voter Turnout from 1945 To Date", [IDEA(b)] International Institute for Democracy and Electoral Assistance: " Voter Turnout from 1945 to 1997: CIS, Central & Eastern Europe: Federal Republic of Yugoslavia", [Kaufmann et al: 1999] Kaufmann, D., Kraay, A. & Zoido-Lobatón, P. (1999): "Governance Matters", The World Bank, Washington, [Kaufmann et al: 2003] Kaufmann, Daniel; Kraay, Aart; Mastruzzi, Massimo (2003): "Governance Matters III: Governance Indicators for 1996-2002", The World Bank, Washington, [Kromrey: 2002] Kromrey, Helmut (2002): "Empirische Sozialforschung – Modelle und Methoden der standardisierten Datenerhebung und Datenauswertung", Verlag Leske + Budrich, Opladen.
••100
[Lala: 2003] Lala, Vishal (2003): "Multiple Regression", [Lessenich: 2000] Lessenich, Stephan (2000): "Soziologische Erklärungsansätze zu Entstehung und Funktion des Sozialstaates", in: Ludwig-Mayerhofer, Wolfgang: "Soziologie des Sozialstaates", Juventa Verlag, Mannheim & München. [Murmann: 2000] Murmann, Sven (2000): "Demokratische Staatsbürgerschaft im Wandel", Königshausen & Neumann, Würzburg. [Murray & Lopez: 2002] Murray, Christopher J.L. & Lopez, A.D. (eds.) (2002): "The World Health Report 2002 – Reducing Risks, Promoting Healthy Life", WHO, Geneva, [Musgrove: 2000] Musgrove, P. (ed.) (2000): "World Health Report 2000 – Health Systems: Improving Performance", WHO, Geneva. [Norris: 2002] Norris, Pippa (2002): "Shared Datasets" Norris, Pippa (2002): "Skeptical Patients: Performance, Social Capital, and Culture", Harvard University, Cambridge. < http://ksghome.harvard.edu/~.pnorris.shorenstein.ksg/ACROBAT/Skeptic al%20Patients.pdf> [Nosikov & Jardel: 2002] Nosikov, Anatoliy; Jardel, Jean-Paul (2002): "The European Health Report", WHO Regional Publications, European Series, No. 97, Copenhagen, ••101
[Oakman: 2003] Oakman, Jonathan (2003): "Psychological Measurement, Lecture 10, Factor Analysis notes", [Old Dominion University: no date] Old Dominion University (no date): "Useful formulas", [Olson: 2003] Olson, Daniel V. A. (2003): "Guide To SPSS – Statistical Controls and Partial Correlation for Research Methods", [Oswaldo Cruz Foundation: 2000] Oswaldo Cruz Foundation (2000): "Report Of The Workshop 'Health Systems Performance – The World Health Report 2000'", Version 1, Rio de Janeiro, 14 -15 December 2000, Ministry of Health, [Rajkumar & Swaroop: 2002] Rajkumar, Andrew Sunil; Swaroop, Vinaya (2002): "Public Spending and Outcomes: Does Governance Matter?", Working Paper #2840, The World Bank, Washington, [Richardson et al: 2003] Richardson, J., Wildman, W.; Robertson, K. (2003): "A critique of the World Health Organisation's evaluation of health system performance", in: Health Economics 12: 355–366 (2003). [Schwab: 2002] Schwab, James A. (2002): "Principal component analysis", University of Texas at Austin, ••102
[Shiffman: 2000] Shiffman, J. (2000): "Can Poor Countries Surmount High Maternal Mortality?" in: Studies in Family Planning 31(4): 274-89. [Shin: 2002] Shin, Michael E. (2002): "Income inequality, democracy and health: A global portrait", University of California, Los Angeles. [Smith: 2002] Smith, Lindsay I. (2002): "A tutorial on Principal Components Analysis", [Subramanian et al: 2002] Subramanian, S. V.; Belli, Paolo; Kawachi, Ichiro (2002): "The macroeconomics determinants of health", in: Annual Reviews Public Health, Vol 23: 287–302, [Travis et al: 2002] Travis, Phyllida; Egger, Dominique; Davies, Philip; Mechbal, Abdelhay (2002): "Towards better stewardship: concepts and critical issues", Global Programme on Evidence for Health Policy, Discussion Papers #48, WHO, Geneva, [Tucker et al: 1997] Tucker, Ledyard R. & MacMalllum, Robert C. (1997): "Exploratory Factor Analysis", Chapter 7, Introduction to Exploratory Factor Analysis, [Üstün et al: 2001] Üstün, T. Bedirhan ; Chatterji, Somnath; Villanueva, Maria; Bendib, Lydia; Çelik, Can; Sadana, Ritu; Valentine, Nicole; Ortiz, Juan; Tandon, Ajay; Salomon, Joshua; Cao, Yang; Wan Jun, Xie; Özaltin, Emre; ••103
Mathers, Colin; Murray, Christopher J.L. (2001): "WHO Multi-country Survey Study on Health and Responsiveness 2000–2001", GPE Discussion Paper #37, WHO, Geneva, [Valentine et al: 2000] Valentine Nicole B.; de Silva, Amala; Murray, Christopher J.L.: "Estimating Responsiveness Level and Distribution for 191 Countries: Methods and Results", World Health Organization, GPE Discussion Paper #22, WHO, Geneva, [von Hagen & Traistaru: 2003] von Hagen, Jürgen.; Traistaru, Iulia (2003): "The South-East Europe Review 2002–2003", World Economic Forum, Geneva, [Ward et al: 1997] Ward, Michael D.; O'Loughlin, John; Shin, Michael; Lofdahl, Corey L.; Gleditsch, Kristian S.; Cohen, Jordin S. (1997): "The spatial and temporal diffusion of democracy 1946–1994", Program on political and economic change, Institute of Behavioral Science, University of Colorado, Boulder, Colorado, [Welzel et al: 2001] Welzel, Christian; Inglehart, Ronald; Klingemann, Hans-Dieter (2001): "Human Development as a General Theory of Social Change: A MultiLevel and Cross-Cultural Perspective", Discussion Paper FS III 01-201, Wissenschaftszentrum Berlin für Sozialforschung (WZB), [WHO proposal on health performance: 2001] WHO (2001): "Proposed strategies for health systems performance assessment", in: Background Documentation for Scientific Peer Review Group Meeting, WHO, Geneva, 7-8 December 2001, ••104
[WHO draft on health inequalities: 2001] WHO (2001): "Report on WHO Technical consultation on the measurement of health inequalities", Draft, WHO, Geneva, Switzerland, November 7-8, 2001, [World Bank: 2000] World Bank (2000): "World Development Report", Chapter 3: 'Making State Institutions Pro Poor', draft, in: "Attacking Poverty". [World Factbook: 2002] World Factbook (2002): "Notes and Definitions",
••105
Annex D – Methodological definition of the variables
Annex D – Methodological definition of the variables
Healthy life expectancy
HALE is build up upon the WHO measure of "life expectancy at birth" which is based on the mortality rates but includes an adjustment for time spent in poor health. The "time spent in poor health" is measured by combining condition-specific estimates from the Global Burden of Disease 2000 study with estimates of the prevalence of different health states derived from health surveys. It thus refers to self-reported health status (results from the surveys) in combination with analyses of disability prevalence in the Global Burden of Disease project, which draws on a wide range of epidemiological and demographic data. [see Nosikov & Jardel: 2002, page 15f] In the Global Burden of Disease (GBD) disability adjusted life years lost by age, sex and cause are estimated – including regional estimates of prevalence of different non-fatal health outcomes. [see WHO proposal on health performance: 2001, page 34]
The basic survey for the self-reported health status data is the WHO Household Survey Study carried out in 2000 and 2001. In the study new health status instrument based on the International Classification of Functioning, Disability and Health, which seeks information from a ••D1
Annex D – Methodological definition of the variables
representative sample of respondents on their current states of health according to six core domains was used. [see Murray & Lopez: 2002, page 173] In the survey instrument performance tests and vignettes to calibrate self-reported health on selected domains such as cognition, mobility and vision were included. The vignettes were developed for the purpose of establishing cross-population comparability.
Due to the calibrated responses the prevalence of different states of health by age and sex could be estimated. In order to correct biases in self-reported health statistical methods were used based on the hierarchical ordered probit (HOPIT) model. [see Murray & Lopez: 2002, page 173]
Equality of child survival
The data for this measurement come from the Demographic and Health Surveys (DHS) program and small area vital registration data on child mortality. The initial aim of the WHO was to measure the distribution of Healthy life expectancy. But due to the limitations of data and methods, health inequality was measured as the distribution of the probability of survival across children.
In order to estimate the distribution of the probability of surviving to age 2 a parametric model was used. The method was based on maximum likelihood estimation of the extended beta binomial distribution to distinguish between variation across mothers in the number of children who have died due to chance and that due to differences in the underlying risks of death. [see Musgrove: 2000, page 146] In order to calculate equality of child survival, child mortality distributions have been transformed into distributions of expected survival time. The distributions of survival time have been summarized by the following formula: ••D2
Annex D – Methodological definition of the variables
where x is the survival time of a given child and
is the mean survival
time across children. [see Musgrove: 2000, page 147]
Each child is compared to every other child in the population. Because all of the WHO goal achievement measures are intended to be positive measures equality in child survival was simply estimated as one minus the inequality index. [see WHO draft on health inequalities: 2001, page 2]
Critical remarks
The WHO concept of measuring health inequality by equality of child survival was largely criticized and discussed in several health care journals and publications. The WHO adopted policies in order to reduce health inequalities between countries and between groups within countries by improving the level of health of disadvantaged nations and groups already in the mid 1980ies. The critics address the WHO's choice to measure health inequality by inequality of child survival not only because that does not include measurements of inequality for adults but mostly because there are many alternatives to measure inequality that were not considered by the WHO. The main critics focus on the fact that the measurement does not include evaluations of differential distribution of health among specific population groups. The indicator is thus criticized as strongly affected by the extent of inequalities in socioeconomic status within the population. [see Oswaldo Cruz Foundation: 2000, page 15] The point of discussion is whether inequalities in health should be measured by summarizing the entire distribution of health in a population or by focusing on differences in average levels of health across different socio-economic sub-groups – ••D3
Annex D – Methodological definition of the variables
defined by income, education, race or social class. [see Braveman et al: 2001] Furthermore the WHO measurement was criticized because the sources of data refer to surveys ranged from 1987 to 1997 and are thus quite old. [see Oswaldo Cruz Foundation: 2000, page 15] The Oswaldo Cruz Foundation also reviews that the methodology suffers from mathematical inconsistencies and procedural unclearness. It is to mention that Gakidou and King who developed the "equality of child survival" measurement established a wider approach that measures " total health inequality" by the variation in health status across individuals within a country as a whole but also within any subgroup within a country. [see Gakidou & King: 2002] But this approach is not included in the data available for this study.
Fairness of financial contribution
Data were drawn from income and expenditure surveys, from tax, social security and health insurance schedules, and from national health accounts. [see MURRAY & LOPEZ: 2002, page 140] More precisely the distribution of households' financial contribution is calculated using household survey data, which includes information on income and household expenditure (by goods and services including health). The calculations are furthermore based on government tax documents (income tax, sales tax, and property tax), national health accounts, national accounts, and government budgets. For countries where such data were not available the distribution of health financing contribution has been estimated using indirect methods and information on important covariates. [see Musgrove: 2000, page 148]
The WHO measurement of "Fairness of financial contribution" is an ex post measurement. It therefore refers to what households actually contribute rather than to their ex ante risks of needing health care. [see ••D4
Annex D – Methodological definition of the variables
Musgrove: 2000, page 38] The index represents the total contributions of a household to health (income taxes, value-added tax, excise tax, social security contributions, private voluntary insurance, and out-of-pocket payments) divided by total expenditure plus tax payments not included in total expenditure (as a proxy for income) minus subsistence. [see MURRAY & LOPEZ: 2002, page 140] Subsistence is taken by measuring food expenditure as an approximation to expenditure on basic needs. Total non-food expenditure is thus an approximation of the household's relatively permanent income that measures what a household can afford to spend on health and other non-food needs. [see Musgrove: 2000, page 36] By using this procedure a measurement of progressivity was included in the index "because richer households spend a lower proportion of their total expenditure on food than poorer households". [see MURRAY & LOPEZ: 2002, page 144]
The formula used for measuring fairness of financial contribution is
where HFC is the financial contribution of a given household and 2000, page 148]
is
the average financial contribution across households. [see Musgrove:
Critical remarks
The WHO index measuring Fairness of financial contribution was criticized as possibly penalizing countries with very progressive payment systems because in these countries the rich may pay a larger share of their capacity-to-pay than the poor. Furthermore it was reviewed that food consumption may be an inadequate proxy for subsistence expenditure because much of the food consumption of the rich is not for subsistence needs. [see WHO draft on health inequalities, page 45] There are also ••D5
Annex D – Methodological definition of the variables
some authors that criticize that the "equity of finance should not be a criterion for the evaluation of a health system" at all. [see Richardson et al: 2003]
Responsiveness of the health care system
The index is based on a survey of nearly two thousand key informants – people identified on the basis of their being knowledgeable about the health system in 35 countries. [see Valentine et al: 2000, page 3] The informants were asked to evaluate the performance of their health system regarding dignity, respect of persons, prompt attention, quality of basic amenities, access to social support networks during care and choice of care provider. [see Musgrove: 2000, page 147]
In the first step of adjustment of the surveys an evaluation of the mean scores across subgroups of individuals within countries was undertaken. It revealed that gender and place of work had a statistically significant impact on the mean reported scores. Women reported lower scores than men, people working for the government reported higher scores than those not working for the government and the mean scores for countries with less political and social freedom were higher at the confidence level. [see Valentine et al: 2000, page 5]
The next step was to adjust the outcomes by calculating country means for each element. These scores were "then regressed against a number of exogenous variables and regression models relating to each element were compared with the aim of finding the regression model with the highest explanatory power". [see Valentine et al: 2000, page 3] The standard to which all element scores were adjusted was to a population of respondents who lived in a totally free society, who did not work for
••D6
Annex D – Methodological definition of the variables
the government and whose sex composition was 50% female and 50% male. [see Valentine et al: 2000, page 5] The output of that procedure – 1000 estimates for each element in each country – were weighted according to weights derived from a web values survey. This web value survey measures preferences for health system performance assessment. [see Musgrove: 2000, page 147] The equation for the estimation of the responsiveness level scores was:
where v is the element score
For unsurveyed countries, the responsiveness level scores were calculated by summing the weighted means of all the elements. [see Valentine et al: 2000, page 13]
Critical remarks
Basically it was criticized that key informant surveys were used to measure the responsiveness level – because key informants may not necessarily reflect the experiences of the general population with their health systems. [see WHO draft on health inequalities, page 54] As the index should measure the responsiveness of the systems towards the clients the critics that not one patient was interviewed seem crucial. Blendon, Kim and Benson developed an other approach by undertaking a survey asking citizens but not key informants. "The results show little relationship between WHO rankings and the satisfaction of the citizens who experience these health systems. The health systems of some top WHO performers are rated poorly by their citizens, including the lowincome and elderly." [Blendon et al: 2001, page 10] It is to mention that the WHO too developed further measurements for examining the responsiveness level, above all a huge survey called the "WHO Multicountry Survey Study on Health and Responsiveness 2000–2001". [see Üstün et al: 2001]
••D7
Annex D – Methodological definition of the variables
General Government expenditure on Health 2000, Share in Total expenditure on Health and General Government expenditure on Health 2000, Share in Total Government expenditure
In general it is to mention that a lot of the initial data come from national governmental institutions but the estimates are made by the WHO. For each country and health system several hundred rows of statistical data and calculations were used. The WHO measurement is based on calculations of expenditure of two groups of entities: the general government and private agents. Data were drawn out from national accounts or from the International Monetary Fund (IMF) Government finance statistics yearbook 2001. The IMF has pioneered in releasing a "functional" breakdown of central government expenditure, which has served as a pilot to track government expenditure. In OECD Member countries the OECD Health data 2002 served as a reference, requiring a few extrapolations to the year 2000. Other sources were the United Nations national accounts, the World Bank Development indicators, household surveys, WHO secretariat estimates and correspondence with officials in Member States. [see Murray & Lopez: 2002, page 175]
Per capita Total expenditure on Health 2000, at International Dollar rate
The international dollars have been estimated by WHO using methods similar to those used by the World Bank. "PPPs are based on price comparison studies for 1996 where they exist. For other countries they are estimated using the GDP per capita in US Dollar, inflation trends, and various dummy variables accounting for regional differences. Forward ••D8
Annex D – Methodological definition of the variables
projections to 2000 are made using the real GDP growth rate adjusted by the relative rate of inflation between the country in question and the United States. The System of health accounts methodology pioneered by the OECD has served as an overall guideline to compile the estimates (…) and to mould hundreds of heterogeneous sources of information into a rigorous and comparative format." [see Murray & Lopez: 2002, page 176]
Voice and accountability
The original data come from: State Failure Task Force State Capacity Survey, The Economist Intelligence Unit, Freedom House, Human Rights Database, Political Risk Services, reporters without borders, World Markets Online, Afrobarometer, Gallup International Citizens Surveys, World Economic Forum, Latinobarometro, Institute for Management Development
Political stability and Absence of Violence
The original data come from: State Failure Task Force State Capacity Survey, Global insight's DRI/McGraw-Hill, The Economist Intelligence Unit, Human Rights Database, Political Risk Services, World Markets Online, Business Environment Risk Intelligence, World Economic Forum, Latinobarometro, Institute for Management Development
••D9
Annex D – Methodological definition of the variables
Government effectiveness
The original data come from: State Failure Task Force State Capacity Survey, Global insight's DRI/McGraw-Hill, The Economist Intelligence Unit, Political Risk Services, World Markets Online, Afrobarometer, Country Policy and Institutional Assessment, Business Environment Risk Intelligence, Country Policy and Institutional Assessment, Freedom House, World Economic Forum, Latinobarometro, Institute for Management Development
Regulatory quality
The original data come from: World Economic Forum, Institute for Management Development, Global insight's DRI/McGraw-Hill, The Economist Intelligence Unit, Heritage Foundation / Wall Street Journal, Political Risk Services, World Markets Online, Country Policy and Institutional Assessment, Country Policy and Institutional Assessment, European Bank for Reconstruction and Development
Rule of law
The original data come from: State Failure Task Force State Capacity Survey, Global insight's DRI/McGraw-Hill, The Economist Intelligence Unit, Heritage Foundation / Wall Street Journal, Human Rights Database, Political Risk Services, ••D10
Annex D – Methodological definition of the variables
Business Environment Risk Intelligence, World Markets Online, Country Policy and Institutional Assessment, Business Environment Risk Intelligence, Country Policy and Institutional Assessment, Freedom House, Gallup International Citizens Surveys, World Economic Forum, Institute for Management Development
Control of corruption
The original data come from: State Failure Task Force State Capacity Survey, Global insight's DRI/McGraw-Hill, The Economist Intelligence Unit, Political Risk Services, Business Environment Risk Intelligence, World Markets Online, Afrobarometer, Country Policy and Institutional Assessment, Business Environment Risk Intelligence, Country Policy and Institutional Assessment, Freedom House, World Economic Forum, Latinobarometro, Institute for Management Development
All six "good governance" (Voice and accountability, Political stability and Absence of Violence, Government effectiveness, Regulatory quality, Rule of law, Control of corruption) variables of the World Bank are based on several hundred individual variables measuring perceptions of governance, drawn from 25 separate data sources constructed by 18 different organizations. Out of this data set six aggregate governance indicators were built. These indicators are based on the World Bank definition of governance "as the traditions and institutions by which authority in a country is exercised. This includes the process by which governments are selected, monitored and replaced, the capacity of the government to effectively formulate and implement sound policies, and the respect of citizens and the state for the institutions that govern economic and social interactions among them." [Kaufmann et al: 2003, page 2] ••D11
Annex D – Methodological definition of the variables
The methodological procedure is the same for all six variables:
The main assumption of the World Bank is that the aggregate indicators (the variables above) have an underlying basic concept of governance. The aggregate indicators are built by using a standard unobserved components model which expresses the observed data as a linear function of the unobserved common component of governance, plus a disturbance term capturing perception errors and/or sampling variation in each indicator. [Kaufmann et al: 2003, page 7] The observed score of country j on indicator k, y(j,k) is written as a linear function of unobserved governance, g(j), and a disturbance term, e(j,k) as follows:
where a(k) and b(k) are unknown parameters which map unobserved governance g(j) into the observed data y(j,k). As a choice of units, it is assumed that g(j) is a random variable with mean zero and variance one. Furthermore it is assumed that the error term has zero mean and a variance is the same across countries, but differs across indicators, i.e. . Errors are assumed to be independent across sources, i.e. the only reason why two sources might be correlated with each other is because they are both measuring the same underlying unobserved governance dimension. [Kaufmann et al: 2003, page 8]
After estimating the parameters a(k), bs(k), and s(k) estimates of governance for each country and measures of the precision of these estimates can be generated. The estimate of governance for a country is the mean of the distribution of unobserved governance conditional on the K(j) observed data points for that country. This conditional mean is the following weighted average of appropriately rescaled scores of each of the component indicators: E[g( j) | y( j,1),..., y( j,K( j))] =
K ( j) k =1
 w(k ) ⋅
y( j, k ) - a(k ) b(k )
••D12
Annex D – Methodological definition of the variables
where the weights applied to each source k, w(k ) =
s e (k ) -2 1+
K ( j) k =1
,
-2
 s e (k )
are inversely proportional to the variance of the error term of that source. [see Kaufmann et al: 2003, page 9]
Freedom house rate
The data come from international sources of information, including both foreign and domestic news reports, NGO publications, think tank and academic analyses, and individual professional contacts.
The rating is based on two checklists one for questions regarding political rights and one for civil liberties. The political rights and civil liberties ratings are then averaged rated on a seven-category scale, 1 representing the most free and 7 the least free. A country is assigned to a particular numerical category based on responses to the checklist and the judgments of the Survey team at Freedom House. According to the methodology, initial ratings to countries are assigned by awarding from 0 to 4 raw points per checklist item. [see Freedom House: 2000]
Voter turnout in the 1990ies
The variable is drawn out from the "Shared Global Database" of Pippa Norris. [see Norris: 2002] The data from Norris are originally from the International Institute for Democracy and Electoral Assistance (IDEA).
••D13
Annex D – Methodological definition of the variables
Voter turnout from 1945 to 1999
The data were drawn out from the "Shared Global Database" of Pippa Norris with one exception. The average voter turnout for Yugoslavia was accounted by myself due to the fact that it was not included in the database. Thus it was built by summarizing all voter turnouts (based on the voters registration counting) from parliamentary elections from 1945 to 1999 and by dividing them by the number of elections. The data for this procedure were drawn out from the IDEA database. [see IDEA(b)]
GDP per capita in Dollar
The data are drawn out from the CIA "World Factbook 2002". "The PPP method involves the use of standardized international dollar price weights, which are applied to the quantities of final goods and services produced in a given economy. The division of a GDP estimate in domestic currency by the corresponding PPP estimate in dollars gives the PPP conversion rate. Whereas PPP estimates for OECD countries are quite reliable, PPP estimates for developing countries are often rough approximations. Most of the GDP estimates are based on extrapolation of PPP numbers published by the UN International Comparison Program (UNICP) and by Professors Robert Summers and Alan Heston of the University of Pennsylvania and their colleagues." [World Factbook: 2002]
••D14
Annex R – Requirements for the quantitative tests
Annex R – Requirements for the quantitative tests
1
The principal components analysis
The estimated correlation between any two observed variables is given by the formula [Oakman: 2003]:
rij = estimated correlation between item i and item j fi1 = correlation between item i and factor 1 fj1 = correlation between item j and factor 1 fi2 = correlation between item i and factor 2 fj2 = correlation between item j and factor 2
In the principal components analysis a linear combination of variables is searched in a way that the maximum variance is extracted from the variables. Then the variance is removed and a second linear combination, which explains the maximum proportion of the remaining variance, is looked for.
••R1
Annex R – Requirements for the quantitative tests
The method is thus designed to transfer a set of interrelated variables into a new set of uncorrelated components, which account for all the variance in the original variables. A group of variables "whose variance can be represented more parsimoniously by a smaller set of factors, or components" has to be identified. [see Schwab: 2002, slide 2]
To obtain a factor solution the principal components analysis has to be repeated a number of times to reach a satisfactory solution. In this process variables are excluded that do not fit in the factor solution.
A principal components factor analysis requires [see Schwab: 2002, slide 3 f]: • Variables on a metric level or dichotomous (dummy-coded) nominal level • The ratio of cases to variables must be 5 to 1 or larger • The correlation matrix for the variables must contain 2 or more correlations of 0.3 or greater • Variables with measures of sampling adequacy less than 0.5 must be removed • The overall measure of sampling adequacy must be 0.5 or higher • The Bartlett test of sphericity has to be statistically significant. • The derived components must explain 50% or more of the variance in each of the variables. Thus they must have a communality greater than 0.5 • None of the variables should have loadings, or correlations, of 0.4 or higher for more than one component. Thus they should not have complex structure • The components should have more than one variable in it
If these requirements are not fulfilled the factor analysis is not appropriate. If they are fulfilled the components can be substituted for the variables in further analyses.
••R2
Annex R – Requirements for the quantitative tests
To compute the analysis several boxes must be marked in the SPSS program: • The Initial solution checkbox provides the statistics needed to determine the number of factors to extract. • The Coefficients checkbox orders the correlation matrix to assess the appropriateness of factor analysis for the variables. • The KMO and Bartlett's test of sphericity and the Anti-Image checkbox provide also information about the appropriateness of factor analysis for the variables. • The most common method as a type of rotation in the analysis is Varimax that was also chosen in this case. Varimax rotation focuses on simplifying the columns of the factor matrix. The criterion for the rotation is to maximize the variance of the factor, while minimizing the variance around the factor. "With the Varimax rotational approach, there tends to be some high loadings (i.e. closer to 1) and some loadings near 0 in each column of the matrix. The logic is that interpretation is easiest when the variable-factor correlations are either closer to 1 thus indicating a clear association between the variable and the factor or 0 indicating a clear lack of association." [Bontis: 1998, page 10]
The following steps represent the necessary diagnostics of the output boxes in order to prove if the factor building was appropriate.
1.1
Strength of Correlations
The first table shows the correlation matrix that reports the correlations among the variables in the analysis. According to the requirements not too many correlations should approach zero. Furthermore it is required
••R3
Annex R – Requirements for the quantitative tests
that some correlations are greater than 0.3. [see Schwab: 2002, slide 20]
1.2
Significance of correlations
The table annexed to the correlation matrix shows the level of significance of the p-values. P-values that are close to zero refer to a high significance. To reach significance p-values should be smaller than 0.05.
Bartlett's Test of Sphericity shows whether a correlation matrix is an identity matrix, that can indicate that variables are unrelated. In the row that refers to the level of significance again values less than 0.05 indicate significant relationships.
1.3
Measures of Sampling Adequacy
In the next step the Measures of Sampling Adequacy are examined on the diagonal of the Anti-image correlation matrix. They can indicate correlations that are not due to the common factors. Small values – the ones close to zero – are relatively free of unexplained correlation. Measures of Sampling Adequacy or a variable, which are less than 0.5, indicate that the variable is to be removed from the analysis.
The Kaiser-Meyer-Olkin (KMO) test measures also Sampling Adequacy. It is a statistic that indicates the proportion of variance in the variables which is common variance. It does thus show the variance that might be caused by underlying factors. The Kaiser-Meyer-Olkin Measure of ••R4
Annex R – Requirements for the quantitative tests
Sampling Adequacy must be greater than 0.50. The output is described as marvelous if it is 0.9 or greater, meritorious if it is in the 0.80's, middling if in the 0.70's, mediocre if in the in the 0.60's, miserable if in the 0.50's, and unacceptable if below 0.5.
1.4
Strength of eigenvalues
An eigenvalue is a ratio between the common variance and the specific variance explained by a specific factor extracted. Each component's eigenvector must be defined as the column of weights used to form it from the x-variables. If the original matrix is a correlation matrix, each component's eigenvalue must be defined as its sum of squared correlations with the x-variables. If the original matrix is a covariance matrix the eigenvalue must be defined as "a weighted sum of squared correlations, with each correlation weighted by the variance of the corresponding x-variable". [see Darlington: 1997] By examining the eigenvalues it is to decide how many factors should be extracted. The criterion is that the total initial eigenvalues of the components should be greater than 1.0. The eigenvalues are shown in the table "Total variance explained".
When the components with lower eigenvalues are ignored some information is lost but if the eigenvalues are small this loose is small. [see Smith: 2002, page 16] In the Scree Plot the eigenvalues for initial components are shown, too. The scree plot helps to determine the optimal number of components to retain in the solution. A good factor analysis Scree plot looks like the intersection of two lines.
••R5
Annex R – Requirements for the quantitative tests
1.5
The proportion of variance
Variance is a measure of the spread of data in a data set. The formula for it is [Smith: 2002, page 4]:
The "% of Variance" column shows the percent of variance accounted for by each specific factor or component, relative to the total variance in all the variables. A good factor analysis contains few factors that explain a lot of the variance. The rest of the factors should only explain small amounts of the variance. "Inclusion of a number of very small factors in a study results in an unmanageably large dimensionality of the common factor space." [Tucker et al: 1997, page 146] The cumulative proportion of variance should explain 60 percent or more of the total variance. [see Schwab: 2002, slide 31]
1.6
Estimates of the variance
Communalities – shown in the box "Communalities" – indicate the amount of variance in each variable that is accounted for. In a principal components analysis the initial communalities (as estimates of the variance in each variable accounted for by all components or factors) are always equal to 1.0. The extraction communalities that are shown in the right row of the SPSS output are estimates of the variance in each variable accounted for by the factors in the factor solution. Small values (the ones near to zero) do not fit well with the factor solution and should be probably dropped from the analysis. The factor solution should explain ••R6
Annex R – Requirements for the quantitative tests
at least half of each original variable's variance. The communality value for each variable should be 0.5 or higher. [see Schwab: 2002, slide 32] Otherwise the variable must be removed from the test.
1.7
Complexity of the structure
Whether variables have complex structure or not can be examined by looking on the Rotated component Matrix. "Complex structure occurs when one variable has high loadings or correlations (0.4 or greater) on more than one component." [Schwab: 2002, slide 41] If a variable has complex structure, it should be removed from the analysis.
1.8
Amount of variables loading on a component
Components should have more than one variable loading on them. Otherwise the variable must be removed.
2
Student's T-Test
This is the formula for a T-test comparing two means with independent samples and non-equal variances [see Old Dominion University, page 2]:
••R7
Annex R – Requirements for the quantitative tests
is the mean of the first sample is the mean of the second sample is the population mean value of the first population is the population mean value of the second population is the standard deviation of the first sample is the standard deviation of the second sample and are the numbers of test samples
3
The regression
In order to make a regression test all variables have to be metric. [see Feilmayr: 1997, page 1] The basic model is:
Y = bo+b1X1+b2X2+...+bnXn
In the model, Y is the dependent variable, X1, X2 etc. are independent variables, b0 is intercept, and b1, b2 etc. are the coefficients for the independent variables [see Lala: 2003]
According to Gupta the reliability and trueness of regression results "hinges on diagnostic checking for the breakdown of classical assumptions". [see Gupta: 2000, page 1] A breakdown indicates that the estimation is unreliable. The testing includes informal and formal testing like the F- and T-test. ••R8
Annex R – Requirements for the quantitative tests
Although the linear regression can be used for many different types of models it is based on common foundations of proven statistical theorems. "If the specific regression model is in concordance with the certain assumptions required for the use of these properties/theorems, then the generic regression results can be inferred. The classical assumptions constitute these requirements." [Gupta: 2000, page 1] If not the result is not acceptable. Why? One can only infer strong statements ( i.e. – "an increase of values of the variable x causes and increase in the value of variable y") if "the presumption that the variables used in a regression, and the residuals from the regression, satisfy certain statistical properties". [Gupta: 2000, page 1] Thus one has to examine if the properties of the distribution of the residuals are satisfied in order to interpret the results. If any breakdowns are examined the model must be corrected and the full range of diagnostic checks must be made again. This process will continue until there is no longer a serious breakdown problem, or the limitations of data indicate to stop. [Gupta: 2000, page 1]
The following diagnostics are used to examine if the zero hypothesis can be rejected and the hypothesis that the dependent variable can be explained by the independent variable can be inferred. According to the literature this procedure should be undertaken step by step. The different steps are therefore explained along the procedure of analysis that should be undertaken.
Gupta suggests to first look on the model fit ("ANOVA"). The ANOVA offers information about the • significance of the model – it thus answers the question if the model explains the deviations in the dependent variable and proves the "goodness of fit". The significance level must be below 0.05 to state that the model fits. If the significance level is below 0.01 the model is significant at 99 percent. If the significance level is below 0.05 the model is significant at 95 percent. In the next step it is to look on the ••R9
Annex R – Requirements for the quantitative tests
• Total Sum of Squares that is "the total deviations in the dependent variable". [see Gupta: 2000, page 3] In the SPSS table they are outlined in the last row of the "Sum of Squares" column. To explain these deviations the best betas (betas indicate the effect a 1 standard deviation change in x on y) – that can minimize the sum of the squares of these deviations – must be found. [see Gupta: 2000, page 3] Above the row with the Total Sum of Squares the • Residual Sum of Squares shows the amount of deviations that could not be explained. • And in the first row referring to "Regression" the amount of Explained Sum of Squares are outlined. • The R-square is shown in the table "Model Summary" in the SPSS program. It shows the percent of deviation from the mean in the dependent variable that could be explained by the model. [see Gupta: 2000, page 3] More concrete it measures the proportion of the variation in the dependent variable that was explained by variations in the independent variables. The higher the R-square is the better the model fits. • The Adjusted R-square measures the proportion of the variance in the dependent variable that was explained by variations in the independent variables. The Adjusted R-square is superior to Rsquare because it is sensitive to the addition of irrelevant variables. • The significance level in the table "Coefficients" gives information about the confidence with which one can support the estimates in the column "B" that outlines the "unstandardized coefficients". If the significance level is less than 0.05 the estimate in column "B" can be stated as true with a 95 percent level of confidence. If the significance level is less than 0.01 the estimate in column "B" can be stated as true with a 99 percent level of confidence. "If this value is more than 0.1 then the coefficient estimate is not reliable because it has 'too' much dispersion/variance." [Gupta: 2000, page 4] ••R10
Annex R – Requirements for the quantitative tests
• In the box "Coefficients" the beta values are outlined. The beta values are equivalent to standardized independent variables before doing the regression. They indicate the effect a 1 standard deviation change in x on y. Betas offer information which variables "matter most".
In order to get a scatter plot of residual versus predicted dependent variable one has to click on the SPSS "Regression test" on "Plots" and then to click on "ZPRED" and "ZRESID". The plot shows the "Regression Standardized Predicted Value" and the "Regression Standardized Residual".
• It indicates the presence of mis-specification (e.g. an incorrect functional form, an omitted variable, or a mis-measured independent variable) and/or heteroskedasticity. "Heteroskedasticity implies that the variances (i.e. the dispersion around the expected mean of zero) of the residuals are not constant, but that they are different for different observations." [Gupta: 2000, page 4 of the Diagnostics annex] If these variances are unequal the relative reliability of each observation is also unequal. How is "heteroskedasticity" examined? One has to look for discernible pattern on the plot. If there are such patterns that indicates "heteroskedasticity". • After proving heterosekdasticity in the plots one has to examine if the residual is normally distributed. One can investigate that by looking on the P-P plot and the histogram. If the residual follow roughly the line on the probability plot and the idealized normal curve they are probably normally distributed. The formula for the probability plot is [see Birkegaard: 2002, page 60]:
• In the next step the collinearities are checked. Collinearity means that two or more of the independent/explanatory variables in a regression have a linear relationship. They represent a problem if their degree is high enough to bias the estimates, ••R11
Annex R – Requirements for the quantitative tests
because "then the estimated regression coefficients and Tstatistics may not be able to properly isolate the unique effect/role of each variable and the confidence with which we can presume these effects to be true." [Gupta: 2000, page 1 of the Diagnostics annex] Collinearity is indicated "if the R-square is high (greater than 0.7514) and only a few T-values are significant". [Gupta: 2000, page 2 of the Diagnostics annex]
In the "collinearity diagnostic" box more useful diagnostics are provided to check the "collinearity". The Eigenvalues represent an indication of how many distinct dimensions there are among the independent variables. Eigenvalues that are close to zero indicate that the variables are highly intercorrelated.
The Condition indices (the square roots of the ratios of the largest eigenvalue to each successive eigenvalue) should not be greater than 15. Tolerance values (the amount of variability of the selected independent variable not explained by the other independent variables) that approach zero indicate a high collinearity. The tolerance level is measured by making each independent variable a dependent variable and regressing it against the remaining independent variables. [see Heo & Kolacz: 2003]
In order to examine collinearity all condition indices above the threshold value of 30 must be identified. For all condition indices that exceed this threshold one has to look for variables with variance proportions above 0.9. "A collinearity problem is indicated when a condition index identified as above the threshold value accounts for a substantial proportion of variance (0.9 or above) for two or more coefficients." [Heo & Kolacz: 2003] In the regression of this study "Model fit", "R squared change", "Descriptives", "Part and partial correlations" and "Collinearity diagnostics" are computed. Concerning "Plots" a scatter plot "ZPRED" versus "ZRESID" is computed and the Histogram and the Normal Probability Plot. ••R12
Annex R – Requirements for the quantitative tests
4
The partial correlation
The correlation coefficient r is [Gerstman: 2003, page 4]:
while
is the sum of squares of the variable x is the sum of squares of the variable y is the sum of cross products
When all points of a scatter plot fall directly on a line with an upward incline, r = +1. When all points fall of a scatter plot fall directly on a downward incline, r= -1. [see Gerstman: 2003, page 3]
This is the partial correlation coefficient of x and y holding z constant:
Source: [Ambos & Penz: 2002, page 33]
A partial correlation coefficient ranges from -1 to +1. It offers like the correlation coefficient an indicator for the strength of the relation.
••R13
Annex A – Analysis of the test procedures
Annex A – Analysis of the test procedures
1
Analysis of the test procedure of the Principal components analysis of the health concept
The test is analyzed with the help of the SPSS tutorial. Let us first look on pre-test requirements for a "Principal components analysis". The ratio of cases to variables must be to 1 or larger. In this case – as there are 44 cases (countries) and eight variables – the ratio is 5.5. Thus this requirement is fulfilled.
1.1
Strength of correlations
None of the correlations approach zero. Thus the first requirement is fulfilled. Another requirement is that there should be some correlations higher than 0.3 in a principal component analysis. By looking at the matrix one can state that this requirement is fulfilled. ••A1
Annex A – Analysis of the test procedures
1.2
Significance of correlations
All of the p-values in the annexed box of the correlations matrix are smaller than 0.05. Thus the correlations are all significant. As shown in the outcome of the Bartlett's test the significance value is 0.000. That shows also that the variables are significantly related and a factor analysis is very suitable.
1.3
Measures of Sampling Adequacy
The values on the diagonal of the anti-image correlation matrix show the Measure of Sampling Adequacy for the respective item. Values less than 0.5 indicate variables that do not seem to fit with the structure of the other variables. In the concerned test all values are higher than 0.5 and thus no variable must be dropped out. As the KMO shows a value of 0.838 the factor analysis can be perceived as very useful. The level of KMO is meritorious.
1.4
Strength of eigenvalues and the cumulative proportion of variance
In the concerned table two eigenvalues are greater than 1.0. Thus this criterion is also fulfilled. As one can see in the test Scree plot the two components are on the steep slope. The ones on the shallow slope can be excluded.
••A2
Annex A – Analysis of the test procedures
1.5
The proportion of variance
As shown in the column "Cumulative %" the two components explain 84.526 percent of the variance. As they explain more than 60 Percent of the total variance the solution is appropriate.
1.6
Estimates of the variance
As shown in the box "Communalities" the estimates of the variance on each variable are higher than 0.5. Thus the factor solutions explain more than half of each original variable's variance. The littlest value is 0.694 (Fairness of financial contribution to health systems) and the largest is 0.944 (Per capita total expenditure on health).
1.7
Complexity of the structure
Let us turn to the Rotated Component Matrix. As none of the variables have higher loadings than 0.4 there is no problem with complex structure.
••A3
Annex A – Analysis of the test procedures
1.8
Amount of variables loading on a component
None of the two components have only one variable loading on them. Thus the requirement is fulfilled.
In order to know what kind of variables load on which component one has to look on the Rotated Component Matrix.
2
Analysis of the test procedure of the regression
2.1
Regression of the factor 1 variable of the health concept and the two factor variables of the democratic governance concept
In the analysis of the output the steps explained in the methodology chapter are undertaken:
• The significance level is at 0.000, thus the model fits at 99 percent. • The Total Sum of Squares – the total deviations in the dependent variable – is 43.0 • The Residual Sum of Squares – the amount of deviations that could not be explained – is 12.154
••A4
Annex A – Analysis of the test procedures
• The Explained Sum of Squares is 30.84. Thus somewhat three quarters of the total deviations in the dependent variable are explained. • The R-square (the percent of deviation from the mean in the dependent variable that could be explained by the model) in the table "Model Summary" is 0.717. As the R-square is quite high the model seems to fit well. • The Adjusted R-square (the proportion of the variance in the dependent variable that was explained by variations in the independent variables) is 0.704 and thus also indicates a high level of "goodness of fit". • The significance level in the table "Coefficients" for the Constant estimate is 1.0 – thus it is not significant. The level for the democratic governance1 factor is 0.0 and thus significant a 99 percent level of confidence. The level for the democratic governance 2 factor is 0.025 and thus significant at the 95 percent level of confidence. Therefore the estimates of column "B" can be supported for the democratic governance1 factor (democratic governance variables) and the democratic governance2 factor (voter turnout) – although the confidence is lower – but not for the Constant estimate. The coefficient estimate for the Constant estimate is not reliable because it has too much variance.
The regression line can be estimated as: Health factor 1= 5.E–08 + 0.825*democratic governance factor 1 + 0.193*democratic governance factor 2
• The beta value for the democratic governance factor 1 is 0.825, the beta value for the democratic governance factor 2 is 0.193. Thus the effect of the variable democratic governance factor 1 is much more important than the effect of the democratic governance factor 2.
••A5
Annex A – Analysis of the test procedures
• The heteroskedasticity is examined in the plots "Regression Standardized Predicted Value" versus "Regression Standardized Residual", the Partial Regression Plot "Health factor 1" versus "democratic governance factor 1" and the Partial Regression Plot "Health factor 1" versus "democratic governance factor 2". The plots show somewhat discernible patterns and therefore heteroskedasticity may be indicated. • By looking on the P-P plot and the histogram the residual seem to be generally normally distributed because it follow roughly the curve. • The R-square is high (0.717) but not greater than 0.7514.The Eigenvalues are not close to zero. The Condition indices are not greater than 15. The Tolerance values do not approach zero. Therefore collinearity is not indicated.
2.2
Regression of the factor 2 variable of the health concept and the two factor variables of the democratic governance concept
As the significance level is at 0.113 the model does not fit. Further diagnostics are not necessary because the output of the regression can be stated as unreliable if the significance level is higher than 0.05.
••A6
Annex V – Variable chart
Annex V – Variable chart
This table refers to the groupbuilding along the output of the principal components analysis of the health concept. It provides the ranking of the countries along the eigth variables of the health concept. Group1 (high on factor1 and low on factor2) is coloured orange. Group2 (high on factor1 and high on factor2) is coloured yellow. Group3 (low on factor1 and high on factor2) is coloured green. Group4 (low on both factors) is coloured blue. Furthermore it offers information about the average values of the groups on the specific variable, the maximum and the minimum values and the standard deviation. The thin black line indicates the grouping along the high/middle/low value groups.
••V1
Annex G – Group chart
Annex G – Group chart
The tables offer information about the ranking of the four groups (that have been formed along the output of the principal components analysis) within the high/middle/low value groups. These new groups have just been built in order to examine the define the four groups. The tables thus indicate if the majoriy, the average and the variance of the four groups have special values within the "help" high/middle/low groups.
••G1