The Emergence of Healthcare Systems

Abstract

Some sort of medical infrastructure has existed in all modern states and dependent territories. However, healthcare systems that provide legal entitlements to medical care at least for specific groups of the population, and that regulate access to and provision of healthcare on a national level only came into existence at the end of the nineteenth century. In our chapter, we trace the global emergence of healthcare systems and test whether their introduction date is related to a diffusion process through networks of trade, culture or health-related cooperation, or whether it is mainly influenced by domestic medical needs, capabilities, and wealth.

Introduction

This chapter is a product of the research conducted in the Collaborative Research Center “Global Dynamics of Social Policy” at the University of Bremen. The center is funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—project number 374666841—SFB 1342.

Healthcare systems address basic and often immediate human needs for medical care in case of illness or injury. Of all welfare policies, health is probably the area where the lack of a functioning system has the most severe and direct negative consequences. Nevertheless, healthcare systems that guarantee access to medical services for relevant parts of the population were, globally, not the norm during most of the twentieth century. State-run or sponsored hospitals go back a long time, but meaningful healthcare systems only started to emerge at the end of the nineteenth century, together with old age pension systems and often in conjunction with work injury protection (Schmitt et al. 2015). Today, there are only a few countries left without at least a rudimentary healthcare system. Unfortunately, this doesn’t mean that access to necessary healthcare services is available to everyone. Often, medical services are inadequate and do not meet the needs of the population.

A healthcare system, therefore, is more than the existence of a certain number of hospitals or medical doctors. In line with the established literature, we understand healthcare systems as the sum of all formal arrangements concerning the financing, regulation and provision of qualified health services within a society dealing specifically with healthcare as an area of social protection (Roemer 1991; Freeman and Frisina 2010; Rothgang 2021). Here, we are only interested in systems in which the state is substantially involved in at least one of the aforementioned dimensions of healthcare. We call this a healthcare system under public responsibility. Such a system is introduced when (a) the first nationwide legislation is passed, (b) entitlements to healthcare benefits are secured by law, and (c) the elements of the healthcare system are integrated (de Carvalho and Fischer 2020). Today, based on this comparatively demanding definition, we find healthcare systems that establish entitlements to healthcare for increasingly larger parts of their population in the vast majority of countries (Fig. 5.1).

Fig. 5.1

Adoption of healthcare systems through time

What has driven the remarkable expansion of healthcare systems from a handful of countries at the end of the nineteenth century to almost all countries at the beginning of the twenty-first century? Was it a logical consequence of industrial development and/or increasing wealth? Did healthcare systems develop as a result of democratization, or was it a slow diffusion process whereby states copied the policy innovation from their neighbors?

In what follows, we start with a literature review on theories that might explain the introduction of healthcare systems. The explanatory model we use in the modeling is informed by these theories as well as by the general framework of diffusion theory as laid out in the introductory chapter of this book. After presenting the results of the statistical analysis we then review the theory and draw some conclusions on how the introduction of healthcare systems can be explained.

Theoretical Background: Factors Influencing the Introduction of Healthcare Systems

Although scholarship on the emergence of healthcare systems is often limited to descriptive case studies, lacking large-scale comparisons and generally neglecting countries of the Global South,¹ there does exist an extensive body of literature dealing with the introduction and reforms of social policies. This section reviews this literature to layout possible explanations for the emergence of healthcare systems. In doing so, we distinguish between domestic factors on the one hand and horizontal as well as vertical interdependencies on the other.

Domestic Factors

The emergence of public protection against the major risks of sickness as well as old age, work accidents, or unemployment have been explained as a result of modernization processes related to industrialization and urbanization (e.g., Wilensky 1974). The claim of the modernizationhypothesis is that these developments have damaged traditional means of social protection, but the resulting economic growth provided the resources to establish public social protection programs, including health insurance as one of the major schemes. In particular, for larger samples of countries at varying stages of economic development, per-capita GDP, or other indicators for the level of industrialization have been found to correspond with earlier adoption of social protection schemes (Collier and Messick 1975; Usui 1994). The level of industrialization has been found to increase the likelihood of introducing health insurance among 43 African nations (Kangas 2012). In a study of 177 territories and independent states over the period 1820–2013, Schmitt et al. (2015) identify a positive effect of GDP on the adoption of health insurance. However, the effect disappears if the sample is reduced to independent states. By contrast, Cutler and Johnson (2004), who study a smaller sample of 20 OECD and Latin American countries, find evidence that higher levels of GDP per capita slow down the implementation of national health insurance defined as compulsory coverage for a broad class of people.

Modernization processes also apply to the medical system. By the second half of the nineteenth century, medical progress helped to establish public health and sanitation measures to control epidemics of infectious or parasitic diseases (Trein 2018). The colonial powers disseminated European medical knowledge globally and, in the inter-war period, also began to promote the education of local medical professions, albeit still focusing on disease control and mother and child health (Bruchhausen 2020). Since the 1930s more sophisticated medical interventions have evolved, enabling the development of effective cures for many diseases (OECD 1987). Health-specific problem pressure, as manifested by recurrent epidemics or indicators of population health status, stimulate public healthcare policies. In particular, as therapies become more sophisticated and expensive, regulation of access and third-party financing are required (Moran 2000). Both health-specific problem pressure and dissemination of medical knowledge, as represented for example by the foundation of medical schools, may pave the way for the formation of a public healthcare system. These two correlations can be reformulated as a capabilities hypothesis—countries with more advanced medical infrastructure and knowledge may introduce healthcare systems earlier—and a problem pressurehypothesis—medical needs may speed up the development of a healthcare system.

Conflict and power resource theories have highlighted the role of democratic representation and the power of left-wing parties and unions in the emergence and evolution of welfare states (Korpi 1983). Based on a sample of 76 nations, Cutright (1965) finds that, on a similar level of economic development, nations with more representative governments introduce social security programs earlier. At the same time, studies highlight the role of monarchies or autocratic governments as early adopters of social policy (Flora and Alber 1981; Mares and Carnes 2009). Early implementation of social policy is explained as a means to appease and control workers, to acquire output legitimacy, and stabilize regimes with weak or without democratic legitimacy. Apart from the political regime type, institutional characteristics of political systems contribute to the explanation of timing and expansion of social policies. Blake and Adolino’s (2001) analysis of 20 advanced economies suggest that federalism and a fragmented executive slow down the introduction of national health insurance. Immergut’s (1992) comparative case study of health policy implementation in France, Sweden, and Switzerland shows that the numbers of veto points in the legislative process provide opportunities for opponents of healthcare reforms to block legislation seeking to implement or expand public health insurance. The literature pertaining to the regime type hypothesis is thus undecided whether democracies or autocracies are more likely to create healthcare systems.

Interdependencies

Over the past few decades, new strands of research have emerged to address the limitations of classical comparative social policy studies, which have not been fully able to systematically capture the transnational context and interventions that may impact social policies (Deacon 2007; Yeates 2008). Nowadays, social policies in every country face similar challenges that may require solutions beyond the nation-state level (e.g., demographic changes, growing inequality, global socioeconomic crises). The global social policy literature and the international interdependencies framework seek to address this shortcoming by accounting for the transnational contexts within which social policymaking is developed (Deacon 2007; Kaasch 2012; Obinger et al. 2013). They emphasize the role of international organizations (IOs) in shaping social policies, especially for the countries of the Global South where the possibilities for both financial and technical resource mobilization are frequently limited or non-existent. Relating to the healthcare field, the literature shows how IOs can be influential for the adoption, reform, and maintenance of healthcare policies (see e.g., Walt and Gilson 1994; Marmor et al. 2009). These organizations operate as (a) financing agents through loans and aid, (b) champions of regulation and rights, (c) sources of ideas and normative standards, (d) facilitators of policy exchange, and (e) disseminators of models and prescriptions (Kaasch 2013). Since the introduction of the first healthcare system in 1883 in Germany, IOs have always been involved in the field, notably the Pan American Health Organization (PAHO), the Office International d’Hygiène Publique (OIHP), the League of Nations Health Organization (LNHO), and the World Health Organization (WHO) in the early period, and the WHO and the World Bank more recently. In what we call the IO hypothesis, we thus assume that the likelihood of implementing a healthcare system increases or accelerates with the creation of policy field-specific IOs.

The most substantial and influential way in which IOs operate within the healthcare field is through loans and/or aid, either by means of direct transfers and interventions or via the support of recipient countries’ domestic policies and institutions (Addison et al. 2015). The literature shows that often the disbursement of loans is tied to conditions aligned with the donors’ agenda, which has a massive impact on domestic social policies (Babb and Carruthers 2008; Clements et al. 2013; Kaasch 2013). Although there is considerable scholarly work on aid and healthcare, this is mainly limited to the last 30 years (see e.g., McCoy et al. 2009; Dodd and Lane 2010), when aid in all its forms exponentially increased: For instance, from 1990 to 2016, aid from donors provided more than US$531 billion to economies of the Global South for financing healthcare (Institute for Health Metrics and Evaluation (IHME) 2017). Existing scholarship, therefore, mainly focuses on how aid shapes and/or influences preexisting systems, but does not address whether aid affects the emergence of healthcare arrangements. In order to fill this gap, we examine the development assistance hypothesis, assuming that the likelihood of a country introducing a healthcare system increases as aid grows.

Another recent strand that attempts to explain the introduction and expansion of social policies is the warfare and welfarehypothesis concerning the linkages between war and the welfare state. This relationship was already studied in the 1950s (Titmuss 1958); more recently, Obinger et al. (2018) have addressed this in a global perspective. According to this strand of theory, military conflicts and war experiences are a driving force behind changes in the welfare state. Demands for redistribution and risk-pooling that emerged in the aftermath of the Second World War and resulted in access expansion to social security in Europe are a classic illustration of the warfare and welfare hypothesis (Dryzek and Goodin 1986; Obinger et al. 2018). Obinger and Schmitt (2018, 2020) argue that war enhances state capacity, encourages social protection demands, and increases social spending. The expansion of the state is interpreted as a consequence of wars. It strengthens the legal system and the assertiveness of jurisdiction and has the capacity to bolster democratization. The horrors of military conflicts heighten demands for social protection in order to provide income, employment, education, and housing for invalids, war victims, and their dependents. Lastly, wars impact social spending levels via newly created social protection schemes and the introduction of income and inheritance taxation. Even though case studies and, to a lesser extent, comparative work on the linkages between social policies implementation and wars are abundant for richer countries (Kasza 2002; Ferrera 2018; Lloyd and Battin 2018; Obinger and Schmitt 2020), this relationship has not been fully examined in non-Western countries.

Finally, we assume that diffusion patterns that have been observed in other areas of social policy (Kangas 2012; Schmitt et al. 2015) may also influence the political decision to establish a healthcare system. This might be policy learning among neighboring states or between states among which strong relationships of political exchange exist. One interdependency that has not yet been addressed sufficiently in comparative welfare state research, but which is prominent in studies on transnational political networks (Maoz 2011) is trade relations. Following the theoretical reasoning in this book (see Mossig et al. 2021, in this volume), we therefore also assume that cultural ties, colonial ties, geographic proximity, and the network of trade relations may influence the introduction of a healthcare system. We assume that these four networks may build the underlying structure for the policy diffusion process in general and for the diffusion of the idea of creating a healthcare system in particular. The networks can be seen as avenues or channels through which communication and information about social policies can travel. Under the heading of the network hypothesis we seek to test whether countries are more likely to introduce healthcare systems if they are closely connected through one or more of these networks to other countries that have already established a healthcare system.

Modeling the Introduction of Healthcare Systems

As previously mentioned, we are interested in determining at what point in time a government takes responsibility for healthcare, and we define healthcare system introduction as (a) passing the first nationwide legislation, (b) legally establishing entitlements to healthcare benefits, and (c) integrating the elements of the healthcare system.² For our analysis we utilize the R package netdiffuseR (Vega Yon and Valente 2021) to model the adoption of healthcare systems over time (Valente 1995). In order to operationalize the dependent variable we created a dataset containing the de jure introduction dates for all countries with more than 500,000 inhabitants in 2017, taking into account the first nationwide legislation that defines the population group receiving benefits and an institution or a set of institutions responsible for healthcare. In order to test the eight hypotheses lined out in the theory section above we then have to operationalize the respective independent variables, once again distinguishing between domestic factorsand interdependencies. This description is followed by a short section on data preparation.

Domestic Factors

To address the assumption of the modernizationhypothesis that the creation of a healthcare system may be easier for wealthy countries and that therefore countries with a higher GDP per capita may introduce a healthcare system earlier than poorer countries, we include data on GDP per capita as introduced in Chapter 1 to measure a country’s wealth. For the analysis we convert the values to units of US$10,000.

Another variable to capture domestic institutional developments that may precede the creation of healthcare systems and can serve as a measure for the capabilities hypothesis is data on founding years of medical schools. The binary variable indicates whether a country has an operational medical school (“1”) or not (“0”). The data is based on the “World Directory of Medical Schools” (www.wdoms.org). We see this as a proxy for higher levels of medical knowledge.

While the introduction of healthcare systems may depend on the availability of resources and financial capability, it may also be a response to urgent healthcare needs. In line with the problem pressurehypothesis we assume that a higher problem pressure may induce countries to introduce a healthcare system earlier than their counterparts with a lesser burden of disease. In the model we use two indicators to measure problem pressure or burden of disease,namely, life expectancyand child mortality. Neither indicator is ideal because they only reflect the effect of poor health on mortality in general and on mortality of one, especially vulnerable population group. But these indicators are the only alternatives available with reasonable accuracy for the whole historical period under observation. Both indicators were obtained from the Gapminder project (Gapminder 2017). Life expectancy is measured as the average number of years a newborn child is expected to live if current mortality patterns were to stay the same.³Child mortality describes the number of children which die below the age of five years per 1000 children born alive.⁴

In line with the other chapters in this book and to address the regime type hypothesis, we include as a measure of regime type the levelof democratization in our analysis. As introduced in Chapter 1, the index ranges from low levels of democracy (closed autocracy = 0) to high levels of democracy (liberal democracy = 9).

Interdependencies

In order to address the IO hypothesis and the development aid hypothesis we also include three variables that capture the health-related institutional landscape before system introduction. These are the role of the World Health Organization and its regional organizations, the availability of medical education and research facilities, and the sum of external health-related funds, namely bilateral development assistance for health (DAH). The inclusion of these variables reflects the possibility that being actively involved in international health politics may speed up the introduction of domestic healthcare systems. We include a binary variable with a default value of “0,” which is set to “1” for a period of 10 years around a country’s involvement in the foundation of four early international health organizations; the Pan American Health Organization (PAHO) in 1902, the Office international d’hygiène publique (OIHP) in 1907, the League of Nations Health Organization (LNHO) in 1920, and finally the foundation of the World Health Organization (WHO) in 1948. Because the foundation of these institutions was preceded by multilateral negotiation, our international organizations variable becomes active for the countries involved in the negotiations already one year before the official founding dates (t − 1).

The Development Assistance for Health indicator is defined by the sum of values of all DAH commitments in constant 2011 US$ received from international donors in a given year. All data points were obtained from the “AidData Core Research Release, v3.1”⁵ data set (Tierney et al. 2011). Incoming dyadic project level commitment amounts have been transformed to yearly country-level aggregates.

The motivation for our decision to include a variable that measures in which years a country was involved in wars lies on the one hand in the argument that wars will likely increase the healthcare needs of affected populations. Involvement in wars can thus be interpreted as another measure of problem pressure. On the other hand, the inclusion of a variable that measures involvement in wars also addresses the recent claim of the welfare and warfare hypothesis that the military played an active role in advancing social policies in general and health services in particular to secure healthier and fitter recruits (Obinger 2020). In this study, we include war as a binary variable, which is coded “1” for countries that are affected by war in a given year and “0” for those who are not. Since we expect that wars increase the problem pressure only after a certain time, we decided to lag the start of the war effect by one year (t + 1). The end of a war is lagged by two years (t + 2), because it is unlikely that war-induced healthcare needs of the population immediately cease to exist with the end of the conflict. The war effect used in our model accounts for both intra-state wars as well as inter-state wars as defined by the “Correlates of War” project (Sarkees and Wayman 2010), from which the original data was obtained.

The operationalization of the networks is described in detail in Chapter 1. The proximity network measures the inverse of the distances between countries’ capitals, the trade network measures the volume of trade between country pairs. The network of cultural proximity results from clustering similarities in religion, gender relations, civil liberties, rule of law, government ideology, language, and colonial relations. And the colonial network represents ties between colonizers and colonies.

Data Preparation

Following Aiken and West (1991), we centered several continuous variables of our models to facilitate a more straightforward and meaningful interpretation of the estimated coefficients. We did so by subtracting the grand mean from every value, so that variable values equal to the mean value of the sample in the respective year are exactly at “0” after this procedure. While centering may help us to better interpret the results of our model, it does not affect the overall meaning of the model or its effect sizes, because the variable values are merely proportionally shifted (Aiken and West 1991). The following continuous variables are centered: life expectancy in years, child mortality per 1000 born, development assistance for health, GDP, and regime type. As proposed in the introductory chapter, we also addressed the issue with non-independent observations by using cluster-robust standard errors (Zeileis et al. 2020).

Findings

Table 5.1 presents the log hazard rates for our models of introduction of healthcare systems. Statistically significant effects with p-values lower than 0.1 are printed in bold to facilitate better readability. Model 1 contains the complete set of variables, in Model 2 we added interaction effects between process-time control variables and GDP per capita, network exposure by geographic proximity, and the trade network.⁶

Table 5.1 Network diffusionof healthcare systems—log-hazard ratios

We defined four time periods as intervals for the piecewise constant step function for our discrete time logistic hazard models. The first period covers the years from 1880, the start of our observation period, until 1913, the last year before the beginning of the First World War. The second period from 1914 is historically shaped by the societal and political interruptions of two world wars, and ends in the year of the capitulation of the axis powers in 1945. Starting in the aftermath of Second World War and the beginning of the Cold War, the third period from 1946 until 1978 is characterized by the bloc conflict between the two superpowers USA and USSR, and their two competing ideologies, and it is at the same time a period in which much of the current system of international organizations was created—a period of increasing economic globalization (Su 2002), and decolonization (Betts 2004). The final time period covers the more recent past and extends from 1979 until the end of our observation period 2010. It is shaped by the rise of globalization, the rise and climax of neoliberalism, the collapse of the socialist bloc in the 1990s and the “so-called golden age of development assistance for health” (Dieleman et al. 2017) in the first decade of the twenty-first century.

Among our explanatory variables, Model 1 identifies significant effects only for child mortality, regime type, and proximity. The negative correlation for the regime-type variable indicates that non-democratic states are likely to introduce healthcare systems at an earlier point in time than democratic states. Concerning the regime type hypothesis, our findings therefore support the strand of literature which highlight the role of autocratic regimes in the emergence of social policies. Here, it has to be considered that the early healthcare systems often cover groups pivotal to maintaining political power such as the military, state employees, or workers in the manufacturing industries. Plausible motives for the introduction of social policies under autocratic regimes include the forging of loyalty, striving for output legitimacy, and the appeasement of workers and socialist movements. Moreover, in autocratic regimes there tend to be less veto points to consider in the legislative process, which reduces the chances of bills granting entitlements to healthcare being blocked or rejected.

A positive statistically significant correlation with the proximity variable shows that the introduction of healthcare systems followed regional patterns with mostly European countries among the early adopters, many South American countries following in the second time period, a large number of Asian and North African countries introducing healthcare systems in the third time period. Among the 28 countries that introduced healthcare systems after 1978 only three were not in Africa.

The effect for child mortality is negative, indicating that countries with lower child mortality rates were more likely to introduce healthcare systems earlier than countries with higher child mortality rates. Contrary to the assumption of the problem pressurehypothesis that higher problem pressure may drive the introduction of healthcare systems, the negative sign of the child mortality variable suggests that it is rather countries with better infant health which are more likely to introduce healthcare systems earlier. Since child mortality is often caused by poverty-related illnesses (Black et al. 2010), improved infant health may thus rather be seen as an indicator for better social conditions that usually accompany higher GDP. However, the model does not show a statistically significant correlation between levels of GDP per capita and the introduction of a healthcare system. This finding seems to contradict the existing knowledge of healthcare system emergence (Schmitt et al. 2015).

Model 2 offers an explanation for the puzzling GDP result. There, we have added interaction effects that capture the time dependence of several of our independent and network exposure variables. The introduction of the interaction effects with time periods implies that for each variable the main effect and the interaction effect must be added up to get the overall effect of the independent variable in the respective period. Including these interaction effects significantly improves the overall model fit and explained variance. By adding interaction terms for GDP and the last three time periods, the main GDP effect which now represents the log hazard rate of GDP per capita in the first time period becomes positive and significant. We thus see a correlation between GDP per capita and the likelihood of a healthcare system being introduced in the first time period (1880–1913). Hence, early adopters were more likely to be found among rich countries than among poor ones. In the earliest period of observation, mostly comparatively affluent European states such as Germany, Austria-Hungary, Norway, Russia, and Luxemburg, or Uruguay and Cuba in South America, introduced health systems. In the remaining time periods healthcare systems were introduced in countries with GDP per-capita levels above and below the average. Seemingly, other factors like nation-building in former colonies gain more importance while the wealth of a nation loses its predictive power. The model therefore does not support a general modernizationhypothesis. Only for the pioneering states did higher levels of economic development strongly correlate with earlier healthcare system introduction.

Looking at the other interaction effects the model shows significant correlations between adoption and geographical proximity with a declining magnitude over time. This indicates that regional patterns structure the introduction of healthcare systems, but the importance of proximity to states that have already introduced a healthcare system decreases over time. This reflects on the one hand a saturation process in which the importance of proximity naturally declines with growing cumulative adoption. On the other hand, it also reflects that early adopters were already widely distributed over the globe and can be found in Europe, South America, Asia, and Oceania.

When looking at trade flows, we see a negative significant correlation for the baseline effect for the first period and positive and significant interaction effects between trade flows and process time compared to the first time period. This indicates that for early adopters, trade relations were clearly not a diffusion channel for healthcare system introduction. Compared to the first period, the negative effect of trade relations seems to be damped in later years, with the result that the overall effect is not stable. The negative and insignificant trade exposure coefficient in Model 1, which accounts for the complete time period, also suggests no clear correlation.

The two other network effects—cultural similarity and colonial ties—are not significant, which means that from the networks included in our models only geographical proximity seems to be a candidate for explaining the introduction of healthcare systems for all time periods and for all countries of the world. As Table 5.2 in the appendix shows, an alternative, non-normalized operationalization of colonial network exposure does not substantially alter these results. Of course, this does not preclude the possibility that they were highly important for specific cases.

None of the other variables show statistically significant effects in either of the two models. Nor do they become significant when controlling for the possibility that they may only affect the hazard rate of health system introduction in specific time periods. Being part of one of the groups of countries that founded the major health-related international organizations has not had a measurable impact on these countries’ politics in terms of creating domestic healthcare systems. Some countries, such as Austria, Hungary, or France, had already established health systems before the first international organization was founded. Some, like Cuba or Uruguay, fitted the expected pattern, and others, like the USA, participated in all institution-building processes but only created a domestic healthcare system much later. Our models thus do not provide support for the IO hypothesis. One reason for this may be that the role of IOs is possibly more relevant for shaping the structure of healthcare systems than for spurring their introduction. But this aspect is outside the scope of this article.

The foundation of medical schools does not seem to be related to the introduction of healthcare systems, either. Both indicators—international organizations and medical schools—may be more closely related to a country’s concern for public health, which does not directly translate into the creation of a healthcare system. We thus find no support for the capabilities hypothesis, but obviously, the existence of medical schools is only a very rough measure of medical knowledge and capabilities in the area of health.

The amount of development assistance for health, for which data was only available for the last two time periods, also does not seem to make a difference regarding the willingness of a country to establish a healthcare system, nor does involvement in wars seem to translate into healthcare system creation. The finding that these indicators and also the life expectancy indicator remains not significant in our models, in combination with the negative significant effect of child mortality, strongly suggests that health-related problem pressure does not seem to induce countries to introduce healthcare systems. Fighting the burden of diseases with public health measures does not necessarily go hand in hand with creating entitlements for medical services.

Conclusion

The first major conclusion we can draw from our analysis is that none of the factors highlighted in the theory section is able to fully explain the timing of healthcare system introduction worldwide. Nevertheless, we clearly see regional diffusion patterns and some domestic factors show significant correlation with the hazard rate of introducing healthcare systems. This observation suggests that full explanations should go beyond the realm of traditional comparative welfare state analysis and incorporate ideas from global social policy research, diffusion research, and global history.

As in other instances (e.g., Rothgang and Schneider 2015) the explanation of change demands a complex framework that takes different strands of theory on board and combines domestic factors and international interdependencies.

Starting our review of hypotheses on domestic factors, our models clearly reject a strict modernizationhypothesis that assumes a universal correlation between economic development and wealth on the one hand and the creation of healthcare systems on the other. We rather come up with a highly interesting result: The hypothesis holds for our first period of observation, i.e., for the introduction of healthcare systems before First World War. In this period the introduction of a health system was more likely in more affluent countries, thus confirming modernization theory. In subsequent periods, however, other factors like nation-building in former colonies have gained importance, leading to a decline in the influence of wealth on the introduction of healthcare systems.

Turning to the capabilities hypothesis, we find no evidence to support it. The absence of a statistical correlation in our model may, however, reflect the relatively crude operationalization of this hypothesis due to a lack of better data for the whole period.

The negative correlation between child mortality seems to strongly refute the problem pressurehypothesis and suggests that problem pressure rather decreases the likelihood of earlier healthcare system introduction. This interpretation is, however, misleading. Based on reports from several case studies we rather assume that typically, public health measures in the area of hygiene, water supply, etc., precede the introduction of a healthcare system. Only after such public health measures have brought on a decline in (child) mortality figures, states introduce healthcare systems under conditions of already lower problem pressure.

The literature on the effects of regime types already suggests that in some instances autocracy increases the likelihood of healthcare systems being introduced. Our findings support the observation that healthcare systems often predate democracy. While democratic representation is not a necessary condition for the introduction of a healthcare system, the likely motives of non-democratic regimes for implementing social protection suggest that these policies are not independent of democratization processes, as they seek to stifle the growth of democratic movements.

According to the IO hypothesis we should see a jump in the number of healthcare systems being introduced after the WHO or its preceding health-related international organizations were founded. Figure 5.2 shows that this is not the case. Healthcare systems are introduced at a relatively constant rate over time, with small peaks especially in years in which empires like the Austro-Hungarian empire (in 1888) or the Russian empire (in 1912) introduced systems. Correspondingly, the diffusion models show no significant effect of the foundation of IOs.

Fig. 5.2

Introduction of healthcare systems over time

Our analysis does not support the warfare and welfarehypothesis at all. While the literature has shown that wars have accelerated and even driven social policy developments in some countries, they do not seem to be drivers of healthcare system introduction, globally and over time.

Finally, we find mixed support for the network hypothesis. Trade networks cannot explain policy diffusion. While gaining some importance over time they do not seem to represent relevant channels of policy diffusion regarding healthcare systems. Nor do the links created through cultural similarity and colonial ties offer a universal explanation of healthcare system introduction. But the introduction of healthcare systems clearly followed a regional pattern with European countries coming first, and (South-)American, Asian, and African countries following roughly in this order. Whether this result reflects actual policy learning or follows mainly other logics, e.g., the timing of decolonization and nation-building, needs to be assessed with other methods. Based on our knowledge of healthcare systems around the world, we actually assume that it is more likely the type than the timing of the system introduction that is influenced through transnational policy diffusion networks.

While our results shed some light on possible factors influencing the emergence of healthcare systems worldwide, our analysis clearly has its limitations. Our model can only explain about 21% of the variance in the data. Clearly, other factors not included in our model and case-specific idiosyncrasies play an important role in the decision of a country to create a healthcare system. The long time frame also severely limits the possibility to operationalize some of our hypotheses, as data on many of the—theoretically interesting—variables is unavailable for most of the countries prior to the 1980s. This is especially true for more qualitative data, e.g., on the strength of progressive political actors, which might have enabled us to operationalize the otherwise promising power resource theory. Data on networks other than the four most basic networks included in our model is especially hard to come by. Therefore, it was not possible to test policy-specific relational aspects.

Nevertheless, our modeling for the first time systematically tests many of the assumptions present in the social policy literature which usually looks at individual countries or smaller samples of countries mostly from the OECD world. It shows that so far, no universal model of healthcare system introduction emerges from these assumptions. In the other hand, it highlights regional diffusion and the time-dependent relevance that some of the factors we have identified nevertheless have in a global social policy perspective.

Appendix

See Table 5.2.

Table 5.2. Network diffusionof health care systems—additive modeling approach