Policy Diffusion and Policy Transfer in Comparative Welfare State Research

Abstract

For many years, comparative welfare state research has followed a ‘methodological nationalism’ in the sense that countries were treated as independent units. Yet the recent ‘spatial turn’ in comparative politics has also influenced welfare state research. For some years now, the field has been witnessing a growing interest in questions about interdependencies and policy diffusion between countries. In this article, we provide a structured overview of the state of the art in the policy diffusion and transfer literature that deals specifically with social policy. We present and critically evaluate existing theoretical concepts and quantitative and qualitative methodological approaches that enable the analysis of interdependencies between countries. Moreover, we summarize the empirical findings of quantitative and qualitative studies on the diffusion and transfer of social policy, from some pioneering studies to the latest findings. Against this background we point out what we believe to be promising avenues for future research. We focus on five areas: theoretical work on the mechanisms underlying diffusion and transfer; methodological approaches; the impact of domestic institutions and policy characteristics on social policy diffusion and transfer; programme-specific dynamics; and the systematic combination of horizontal and vertical interdependencies.

Introduction

The welfare state is intimately tied to the nation-state. The nation-state was and still is the main spatial reference point of solidarity, redistribution and social regulation, while the welfare state, in turn, has strengthened the integration and cohesion of the nation-state and enhanced its legitimacy. The traditionally strong focus of comparative welfare state research on the nation-state is therefore hardly astonishing. However, this analytical perspective is increasingly subject to criticism and blamed for its ‘methodological nationalism’ (Zürn 2005). It is argued that contemporary public policy cannot be explained only by socio-economic and political developments within nation-states, but is also shaped by inter- and supranational influences as well as relations between nation-states. Apart from the vertical interdependencies in the European multi-level polity and the growing efforts of international organizations in setting international regulatory standards, the intensity of economic and political interdependencies has also increased horizontally, i.e. between nation-states. For example, countries strategically respond to policies adopted by other countries (e.g. tax competition between states), emulate policies that turned out successful abroad or react to external pressure to adopt a particular policy. In a nutshell, nation-states do not act independently of each other. Manifold interdependencies exist between countries which, in conjunction with domestic factors, account for policy outcomes. Hence, today a relational approach in comparative welfare state research is needed more than ever.

The notion that states and societies are not independent units of observations, however, is far from new. Already in 1889, the British polymath Sir Francis Galton pointed to this problem in a review of an anthropology paper. Nowadays it is therefore referred to as ‘Galton's Problem’ in the methodological literature (cf. Ross and Homer 1976; Goldthorpe 1997). Galton's problem was acknowledged in comparative welfare state research against the backdrop of ever increasing economic globalization and the deepening of European integration (Jahn 2006). Both processes not only lead to higher economic linkages between countries but also intensify information exchange and migration (Duina and Nedergaard 2010; Castles and Schierup 2010).

Interdependencies between nation-states and international influences on national social policy are already well-documented for the formative phase of the modern welfare state. For example, a close exchange of information among experts, labour organizations and governments existed in the 19th century that even stretched across continents (Rodgers 1998). Moreover, the rivalry among nations with a view to outperforming other nations in terms of socio-economic modernization during the age of ‘social imperialism’ (J. A. Schumpeter, cf. Porter 1994: 158) and economic competition shaped social policy from the outset. The foundation of the International Labour Organization (ILO) in 1919 was a milestone in terms of international co-operation in social affairs. By promoting international social standards, the ILO was supposed not only to contribute to a long-lasting peace after the end of the Great War but also to help to avert ruinous competition between the economically highly interrelated industrial economies of that time (Rodgers et al. 2009). After the catastrophe of the Second World War, international collaboration was eventually established on a permanent basis. Various special agencies of the United Nations, the Organisation for Economic Co-operation and Development (OECD), as well as the European Community, strengthened intergovernmental collaboration and promoted the cross-border exchange of information and experience in social policy (e.g. Open Method of Coordination). Moreover, the welfare state featured prominently in the regime competition that took place between the Western democracies and the Soviet bloc during the Cold War (Obinger and Schmitt 2011; Petersen 2012).

All the above examples substantiate the existence of various kinds of political and economic interdependencies between nation-states and their early embeddedness in international and transnational social policy networks. The recent globalization wave and the deepening of European integration since the 1980s have further reinforced these relations and, consequently, encouraged a growing interest in comparative research to study the impact of such interdependencies on public policy-making. In the pertinent literature the repercussions of these interactions on public policy are conceptualized as policy diffusion (Dobbin et al. 2007) and policy transfer (Dolowitz and Marsh 1996, 2000).

This article summarizes the essential theoretical and methodical foundations of the literature dealing with policy diffusion and policy transfer, and provides an overview of important empirical studies in comparative welfare state research. Lastly, we sketch a research agenda for analyzing cross-national interrelations in social policy. Due to limitations on space, we only focus on horizontal relations between nation-states.1

Policy Diffusion and Policy Transfer: Concepts, Causal Mechanisms and Conditional Factors

Comparative welfare state research uses different concepts for the analysis of international interdependencies. The most important concepts are policy diffusion and policy transfer. Policy diffusion is the process by which ‘policy choices in one country affect the policy choices in other countries’ (Meseguer and Gilardi 2009: 528). Policy transfer denotes political actions that use ‘knowledge about policies, administrative arrangements, institutions and ideas in one political system (past or present) […] in the development of policies, administrative arrangements, institutions and ideas in another political system’ (Dolowitz and Marsh 2000: 5). The policy transfer literature builds ‘on previous work concerning lesson-drawing’ (Greener 2002: 162; see, for lesson drawing, Rose 1991, 1993) while the diffusion literature often refers to the quantitative literature on the diffusion of innovations and programme adoption (Walker 1969; Rogers 1995; Collier and Messick 1975; Coleman et al. 1966). Both policy diffusion and policy transfer refer to interdependencies among political systems in the policy-making process. The main difference is the relevance of knowledge and the role of intentional processes (agency) that are emphasized in the policy transfer literature. In contrast, diffusion often includes structural, interest-based and non-intentional processes. A further difference refers to the methodological approach employed. While the concept of policy transfer is dominant in case study oriented welfare state research, policy diffusion is used more frequently in the quantitative research literature. However, the differences between transfer and diffusion are marginal and mostly founded in their affiliation to different research traditions (Marsh and Sharman 2009). This should not obscure the fact that both literatures aim to describe and explain that policies are the result of interdependent decisions. It is important to stress that policy diffusion and policy transfer denote processes which, under certain circumstances, might lead to policy convergence,2 i.e. growing similarities in outcomes over time (Knill 2005: 767).

When summarizing the literature, four different causal mechanisms can be distinguished which might lead to policy diffusion and transfer: learning, emulation, competition and coercion (cf. Elkins and Simmons 2005; Dobbin et al. 2007; Meseguer and Gilardi 2009).

The first mechanism focuses on transnational learning processes. When choosing policy strategies, national policymakers who are confronted with certain problem pressures are geared to successful policies approved abroad (‘best practices’) to reduce the uncertainty of the policy consequences. Learning processes influence the quality of information political actors have about policy instruments and their efficacy. Other countries' experiences may, therefore, influence the expectations of political actors regarding the costs and benefits of implementing a certain policy reform. In consequence, policies may be partly or completely adopted (positive learning) or avoided (negative learning). In general, rational and bounded rational learning processes can be distinguished (see Kahneman 2003 on bounded rationality). Governments acting in a boundedly rational manner do not necessarily evaluate all available information but, instead, use cognitive shortcuts to reduce complexity; they can do this, for example, by orienting themselves to the experiences of early adopters or neighbouring countries that have applied similar policies. Learning mechanisms are linked to the concept of lesson-drawing (Rose 1991, 1993). ‘Lesson-drawing draws upon empirical evidence of programmes in effect elsewhere to create a new programme for adoption at home’ (Rose 1991: 21). It might simply mean to copy a programme ‘using practice elsewhere literally as a blueprint’. But it also implies sophisticated information processing of foreign experiences to design a programme suitable to the specific domestic setting.

Closely related to learning and lesson-drawing are emulation processes. While learning generally implies an improved understanding of the mechanisms that cause policy outcomes, problem solving is of less relevance in emulation processes. Emulation refers to the ambition of political actors to conform to international trends and to belong to an international norm-based community. In the extreme, when a policy is not concerned with concrete problem pressures and efficacy of policies, emulation becomes purely symbolic. The related concept of isomorphism denotes the increasing similarity of policies across the globe caused by an increasing diffusion of ideas that has caused a ‘world society’ to emerge (Meyer et al. 1997).

In contrast to learning and emulation mechanisms, competition is based on the strategic interactions of governments. In times of strong international competition, it is likely that governments have to take other countries' policies into account to realize their competitive advantage or to avoid incurring disadvantages against their competitors. For example, social policy reforms that aim to reduce labour costs may put competitive countries under pressure to adopt similar social policy reforms in order to avoid comparative disadvantages due to differences in factor prices (e.g. negative employment effects in certain segments of the labour market).

The fourth mechanism emphasizes that nation-states are increasingly subject to coercion. On the one hand, countries have to increasingly implement supranational regulations. This vertical mechanism is discussed in the literature on international organizations and European integration, and is not pursued here (e.g. Leibfried and Pierson 1995; Falkner et al. 2005; Strang and Chang 1993). On the other hand, in the case of extensive power asymmetries among states, coercion may also be of particular relevance. An often mentioned case of coercive diffusion is financial aid linked to certain domestic reforms defined by donor countries and international institutions such as the International Monetary Fund or the World Bank. While this form of coercion is rare within developed democracies, recent EU pressure on Greece during the European debt crisis to implement a harsh austerity policy shows, however, that this mechanism nevertheless exists.

In general, it is assumed that the importance of these mechanisms depends on different conditional factors (Holzinger and Knill 2005). This means that the strength and the nature of bilateral relationships influence how states learn from each other or compete with each other. One conditional factor that commonly expresses the intensity of interdependencies is the geographical location of the countries. Exchange of information, as well as competition, should be more frequent between neighbours than between distant countries. A common language or a common cultural background should also positively influence transnational communication. Furthermore, the extent of competition might vary with the intensity of bilateral economic relations. However, competition may also exist between countries with weak direct economic relations when these countries compete in the same markets. Further potential conditional factors include similarities regarding political institutions (political regime type, legal system), ideological positions of the central political actors, the economic system (type of capitalism) or existing policies (type of welfare state).

Methodological Approaches to Spatial Interdependencies

The analysis of diffusion entails central methodological challenges for comparative welfare state research. Whenever cases are not statistically independent, these interdependencies have to be modelled. However, Galton's problem is not a specific problem of the quantitative social policy literature but also applies to the qualitative literature, as Goldthorpe (1997: 9–12) has pointed out. Meanwhile, there are quantitative as well as qualitative approaches to analyze policy transfer and policy diffusion processes empirically.

Quantitative approaches

International influences on national social policy have been measured for a long time via additional independent variables such as trade openness or foreign direct investment (see Busemeyer 2009). These indicators, however, are not appropriate when analyzing mutual interdependencies, since countries are still treated as independent units. Spatial econometrics offers a comprehensive toolbox that enables the systematic quantitative analysis of interdependent data (Franzese and Hays 2007, 2008). It is important to stress that ignoring spatial interdependencies can lead to biased estimates of the influence of the other variables included in the model.

In general, in order to explain the public policy of a focal country, spatial models take into account the information of other units assumed to influence the focal country. The nature of transnational interdependencies is defined via the weighting or connectivity matrix. The weighting matrix is therefore of particular relevance since it models the assumptions about the strength and nature of the bilateral interdependencies. A relatively simple weighting matrix is the contiguity matrix. It is based on the assumption that only countries with a common border influence each other (Anselin 1988). However, particularly in social sciences, interdependencies that are not based on geographical distances may be important. Countries may also influence each other due to bilateral economic relations, cultural similarities and a common language (Beck et al. 2006).

Spatial interdependencies are captured via spatially lagged dependent or independent variables.3 When exogenous variables are spatially lagged, it is assumed that the independent variables do not only influence the policy outputs of the focal country but also the policy outputs of related countries. For example, an increase in German unemployment may not only influence German social policy but also social policy change in Austria, which is an important neighbour and trading partner of Germany.

The most common procedure is, however, to model spatially lagged dependent variables. Here, the spatial lag is a weighted average of the dependent variable of countries related to the focal country. If the dependent variable of the model is social expenditure the spatial lag would be a weighted average of the social expenditure in all other countries. An increase or decrease of the social expenditure in one country would be explained by the changes in social expenditure in nations related to the focal country. Formally, the spatial lag model can be expressed as follows:

• 

with y as the dependent variable, rho as the spatial autoregressive coefficient and w*y as the weighted average of the dependent variable (spatial lag) (Lee and Strang 2006).4

One challenge in the analysis of policy diffusion is to distinguish interdependencies from other potential sources of spatial patterns. Possible causes may be common shocks that provoke similar national reactions. For example, the 2008 financial crisis caused short-term stimulus packages in many countries and therefore produced similar reactions. Spatial patterns in the dependent variable may also be caused by similar preconditions. Similar social cleavages, for example, can lead to similar social policy developments without countries necessarily influencing each other. The only possibility to distinguish diffusion from alternative sources of spatial patterns is to include appropriate independent variables in the model (Plümper and Neumayer 2010).

In addition to the regression approach, further quantitative methods can be applied to analyze spatial interdependencies, including Bayesian (Meseguer 2009) or dyadic approaches (Plümper and Neumayer 2010).5 However, these approaches are of less relevance in comparative welfare state research and are not, as a result, described here in greater detail.

Qualitative approaches

The qualitative methods literature has so far treated policy diffusion and transfer in a rather cursory fashion. In contrast to quantitative research, however, the exact methodological approach of qualitative transfer or diffusion studies often remains implicit. This does not mean that qualitative methods cannot deal with interdependencies. Large-N spatial regression methods, however, cannot be applied to qualitative designs which are usually based on a small number of units. Moreover, some of the ‘classic’ research designs, particularly Mill's methods (Mill 1974 [1872]: 388–96) and their successors (e.g. the ‘most-similar systems design’ by Przeworski and Teune 1970) do not seem especially promising. Apart from a number of general problems (Lieberson 1991), it is not possible to model the interdependence of cases in a Mill's methods-type design. For example, the most important of these designs, the ‘method of difference’, following a simple experimental design, is based on a selection of cases that differ only with regard to their outcome (e.g reform vs. non-reform) and on one single independent variables (i.e. ‘treatment’). Not only is such a case selection extremely unrealistic, it also cannot take into account the situation where the difference in outcome of one case depends on the outcomes of other cases.6 What might be done is to model the dependence of an outcome on the outcome of a single exogenous case, but not the mutual impact of several cases.

Due to the limitations outlined, existing qualitative studies usually rely on a design that combines cross-case analysis and within-case analysis.

Qualitative cross-case analyses of policy transfer are almost always based on a non-random selection of a few so-called ‘diverse cases’ (Seawright and Gerring 2008). Cases are selected so as to maximize variation on theoretically relevant dimensions – for example, the conditioning factors of policy transfer and diffusion. It makes sense, for example, to study both neighbouring and remote countries and both culturally similar and dissimilar countries at the same time. The temporal and spatial patterns that can thus be detected across the countries under study provide important clues as to which causal mechanisms may be at work. Often, these patterns are depicted with the help of diffusion curves or maps.

Of course, while such patterns may suggest processes of diffusion and transfer, they do not provide proof. A ‘wave of reforms’ may just as well be the product of randomness in independent cases or the result of a common exogenous shock. An in-depth analysis of the cases is therefore recommended.

An in-depth analysis of the cases could be based, for example, on expert interviews and document analysis and examine whether an intensive exchange of ideas occurred between governments and whether there is at least evidence of mutual observation of reforms and reform outcomes. In other words, the techniques of qualitative process tracing (George and Bennett 2005) can be applied to processes of diffusion and transfer – especially if precise theoretical expectations can be formulated.

Some studies – albeit so far very few – assess the plausibility of the conclusions inferred from cross-case and within-case analysis through counterfactual thought experiments (Tetlock and Belkin 1996). One may ask whether a particular policy innovation would have been introduced if there had not been any change in other countries. Was a particular reform already on the agenda and did it stand a good chance of being introduced? Whenever there are only a few available or relevant cases, counterfactual analysis may often be the only way to draw inferences about policy diffusion.

The State of the Art: Policy Diffusion and Policy Transfer in Comparative Welfare State Research

The allegation that public and social policy research tends to follow a ‘methodological nationalism’ is largely correct but also somewhat overdrawn as the early literature did not entirely neglect policy diffusion and transfer. By differentiating between various thematic fields of research, this section provides an overview of the state of the art with regard to the analysis of policy diffusion and transfer in comparative social policy.7 For this literature review, we have tried to be as broad as possible in order to include the best (qualitative and quantitative) research dealing with cross-national policy diffusion and transfer of social policy reform. For reasons of space, we decided to exclude studies of subnational processes of diffusion and transfer as well as vertical processes (e.g. between the international or supranational and the national level).

The timing of welfare state consolidation

Almost 40 years ago, Collier and Messick examined whether the adoption of the first national social legislation can, in addition to domestic factors, be attributed to international policy diffusion. They show that out of a sample of 59 countries, the laggards introduced their first programme at a lower level of economic development than the pioneers because they could build on the experiences made by the pioneering nations (Collier and Messick 1975: 1300). However, their empirical analysis is based on simple correlations and scatter plots. A few years later, Jens Alber (1982: 134) investigated whether the German Bismarckian social legislation had demonstrative effects on other nations that may have facilitated its proliferation in other Western European countries. He concludes that German social insurance can hardly be considered a successful export product (Alber 1982: 143). In a contemporaneous qualitative-historical account, Stein Kuhnle (1982, 2011) examines possible spill-over effects of German social insurance to Nordic countries and came up with very similar findings, since only Norway had adopted some principles of Bismarckian social insurance. In a comparison of German and British social policy, however, Hennock (2007) finds evidence for policy imitation and policy transfer of some Bismarckian social insurance principles in the wake of David Lloyd George's trip through Germany in 1908. Abbott and DeViney (1992) use sophisticated statistical methods to analyze sequences in the adaption of several welfare programmes in 18 rich industrial states. They find that the introduction of accident insurance, unemployment compensation and family benefits followed, unlike old age and health insurance, a specific sequential pattern which, however, cannot be attributed to particular linkages between states.

Numerous historical inquiries (e.g. Rodgers 1998) have shed light on the mutual influences of social-reformist ideas during the formative phase of the welfare state. Coleman, for example, argues that (social) reforms in New Zealand in the late 19th century influenced the Progressive Movement in the USA and, consequently, influenced the Social Security Act (Coleman 1987). In his historical inquiry on the introduction and development of the welfare state in Great Britain and Sweden, Hugh Heclo emphasizes the role of learning from past reforms but also points to the importance of policy diffusion caused by cross-border learning (Heclo 2010 [1974]: 310–11). Lastly, Petersen (2011) convincingly demonstrates that social policy in the Nordic states was already shaped by intensive mutual learning in the 19th century. In the course of the 20th century this exchange was gradually institutionalized by the establishment of common expert committees and other networks of collaboration.

Social spending

Quantitative studies, which employ methods of spatial econometrics outlined in the previous section, are of a relatively recent vintage. Jahn (2006) examines the impact of economic interdependencies on social spending in 16 OECD countries from 1980 to 2001. His findings are indicative of an increasing policy diffusion between trade partners which, in the 1990s, even gave rise to a ‘race to the bottom’ in terms of social spending (Jahn 2006: 427). Franzese and Hays (2006) examine 15 EU countries over the period from 1987 to 1998 to consider whether EU member states strategically interact in the field of active labour market policy. They conclude that governments often abandon their intentions to increase their expenditure in this policy field if neighbouring countries increase spending for active labour market policy. The reason is that states can benefit from the spending decisions of their neighbours which, in turn, creates incentives for free-riding. This may propel a downward spiral in spending levels and result in decreasing expenditure across EU member states. Kemmerling (2007) also analyzes policy diffusion in the field of labour market spending. In line with Franzese and Hays' findings, he reports a negative effect between neighbouring countries for a sample of 23 OECD countries studied over the period between 1980 and 2001. Again this finding is indicative of positive cross-border externalities since domestic employees may benefit from high spending levels in neighbouring countries. In order to reap these externalities, states strategically deviate from the policy course adopted by their neighbours. Obinger and Schmitt (2011), using spending data from 16 Western democracies and five Council for Mutual Economic Assistance (COMECON) countries from 1961 to 1989, provide empirical evidence that the welfare state was subject to political regime competition during the Cold War. The expansion of social protection in the Eastern bloc was used as a strategy aimed at increasing mass compliance and stability of authoritarian political regimes and demarcating these regimes from the capitalist West. This effect was strongest in times of military detente during the 1970s, when regime competition increasingly centred on policy outcomes.

Social rights and benefit generosity

Only a limited number of studies to date have examined policy diffusion in relation to the generosity of welfare states. In contrast to spending devoted to labour market policy, Kemmerling (2007) finds no evidence that the net replacement rate offered by unemployment insurance is shaped by spatial interdependencies. The reason is a lack of positive externalities in passive labour market policy, as employees cannot benefit from unemployment benefits offered by other countries. Free-riding is therefore precluded. Gilardi (2010) also focuses on the recent development of unemployment benefits in 18 OECD member states and shows that policy learning is contingent on the partisan complexion of government and the beliefs of policymakers with regard to the electoral consequences of policy reform. Conservative cabinets are more committed to the general trend of benefit retrenchment because, in contrast to left leftist parties, they are less likely to be punished by voters than Social Democrats when it comes to benefit cutbacks. Left-wing parties only comply with foreign retrenchment policies if benefit cutbacks lead to a decline in the rate of unemployment abroad. In contrast to Kemmerling (2007) and Gilardi (2010), Schmitt and Obinger (2013, forthcoming) focus on the long-term development of replacement rates in the post-war period (1955–2000) and examine the impact of various forms of cross-national interdependencies on benefit generosity related to pensions, sick pay and unemployment. Surprisingly, the impact of diffusion on benefit generosity was more pronounced during the ‘Golden Age’ when compared to recent decades. In addition, the relevance of diffusion and the underlying mechanisms vary by programme. While emulation and learning within groups of culturally related countries seem to play a role with regard to pensions, the replacement rates offered by unemployment and sickness insurance are influenced by international competition.

Welfare state reform

The literature addressing the transfer of specific welfare state reforms is dominated by qualitative accounts and heterogeneous. Only a small number of comparative studies have to date examined the international proliferation of particular social policy reforms with statistical methods. Gilardi et al. (2009) show, in their analysis of 19 OECD countries over a period stretching from 1980 to 2005, that the introduction of diagnosis related groups (DRGs) in the hospital sector is a result of policy diffusion.8 Countries where policies prove particularly ineffective are particularly willing to analyze what has occurred in other jurisdictions. This is especially the case where the adoption of patient classification systems has led to lower healthcare costs in other countries. Based on a sample of 71 nations, Brooks (2007) demonstrates that the diffusion of fully funded pensions is influenced by the costs of implementation and the national level of economic affluence. Middle-income countries are most likely to rely on the experiences of other states, and expensive implementation costs make policy learning more likely to occur.

A number of qualitative studies – mostly based on a small number of cases – analyze the cross-border transfer of social reforms. Similar to the situation in macro-quantitative research, systematic qualitative inquiries of social policy diffusion are still rare. Moreover, the existing studies are either rather descriptive or focus on particular policies and countries. Dolowitz (1997; Dolowitz and Marsh 2000) shows that the reform orientation of the British government in the 1980s was to a large extent influenced by the reforms of the labour market policy enacted in the USA. Daguerre and Taylor-Gooby (2004) provide evidence that New Labour's Welfare to Work agenda (the so-called New Deal) was influenced by American ideas and policies, while European ideas were widely neglected. They offer three explanations for this asymmetric policy transfer, namely: similar cultural affinities and welfare state structures; well-established Anglo-American policy networks at the ministerial and official level; and the success of Temporary Assistance for Needy Families (TANF) in terms of lowering the welfare rolls (Daguerre and Taylor-Gooby 2004: 33–6). Seeleib-Kaiser and Fleckenstein (2007) investigate whether learning from the British experience accounts for the recent change in German labour market policy. Relying on a discourse analysis, they show that the departure of German labour market policy from its traditional conservative path was stimulated by the British Third Way debate. Lesson-drawing from the UK and policy emulation eventually led to the liberalization of the German labour market.

One of the most important qualitative studies in comparative welfare state research which fully makes use of the qualitative methodological repertoire described in the previous section is Kurt Weyland's Bounded Rationality and Policy Diffusion: Social Sector Reform in Latin America (Weyland 2006). Based on cross-case comparisons of pension and healthcare reforms and relying on thorough process tracing, he shows that the privatization of pensions in Chile triggered similar pension reforms in Latin America. By contrast, the healthcare sector was not affected by policy diffusion. He concludes that a process of bounded-rational learning accounts for the diffusion in pension policy.9

Several contributions in a 2009 special issue of the International Social Security Review (Volume 62, Number 4) devoted to processes of learning in social policy corroborate this finding. Learning is only related to rational patterns in a limited number of cases. Even though ideas and policies are frequently imported from other countries, they are typically adjusted to fit certain national requirements and, in consequence, extensively changed.

Future Avenues for Research

According to this survey of the state of the art, comparative welfare state research has only recently begun to systematically study relational policy processes. The quantitative and qualitative literature mentioned can, in many cases, demonstrate with some plausibility that policy innovation is based on foreign examples and that reforms have spread across countries. However, this is often not combined with an empirical analysis of the exact conditions and mechanisms of diffusion and transfer. Moreover, the potential of existing methods is often not fully exploited. Against this background, we argue that the following areas should be promising avenues for future research.

First, there is a need for theoretical and conceptual work that further analytically differentiates the various concepts and causes of policy diffusion and transfer in social policy. Learning and competition are often closely intertwined, since competing countries are forced to learn from each other's successes. Countries competing for skilled labour will pursue labour market policies in order to train, attract and integrate the required foreign manpower. They will possibly learn from each other to avoid comparative disadvantage. What is more, the distinction between processes of learning and imitation can be problematic. Among other things, the effectiveness of social policy measures is often difficult to judge. Do governments follow the example of a particular group of countries only to be part of a ‘club of modernizers’ or because they are aware of the effectiveness of their policies? The central reform concepts of the 1990s and 2000s – activation, flexicurity, multi-pillar pensions – have become policy buzzwords. Supposedly, processes of learning and imitation intermingle in such cases. What are, then, the political preconditions for rational learning based on actual policy effects? How can we recognize imitation when we see it? Even the mechanisms of coercion and learning are closely related. In the 1990s, as part of the social dimension of the EU, the Open Method of Coordination was institutionalized in several social policy areas, a learning mechanism in which all member states took and continue to take part. Does this imply ‘coerced learning'? It will be a task of comparative welfare state research to better differentiate the various mechanisms from each other and also to better relate them to one another.

Second, there is room for improvement in terms of methods. Quantitative research, mainly based on spatial regression, allows one to address whether social policy diffuses or not. Why social policy is diffused, which actors are involved and why they are involved, are questions that cannot be answered in a fully satisfactory manner with these methods. Network analysis, as well as qualitative approaches, seems promising in this respect. The main strength of network analysis is the detailed mapping of actors and their mutual relationships. Hence, the actors that are relevant for the diffusion and transfer of social policies can be documented. In many social policy areas – for example, in labour market policy – expert commissions and think tanks play a crucial role in policy advice. The very general relationships that exist between countries – in areas such as trade relations that are used in spatial econometrics – do not really help in this respect. Instead, it is necessary to track the information exchange between experts. In order to measure the motives of social policy actors, qualitative methods such as process tracing on the basis of expert interviews are promising (Starke 2012). Overall, the study of the diffusion and the transfer of social policies should be based more on a mix of qualitative and quantitative methods.

Third, the impact of both domestic institutions and policy characteristics on social policy diffusion and transfer should be studied further. It is plausible, for example, that some political systems are more open to foreign influences than others. One could study whether democracies were better able to learn from other democracies than from autocracies during the early years of the welfare state. Perhaps the different welfare regimes are also affected by diffusion to varying degrees. In this context, it is striking that the globalization/race-to-the-bottom debate has been particularly strong in continental Europe, where non-wage labour costs are relatively high (cf. Scharpf 2000). This may be evidence of higher competitive pressure in these countries. In other words, diffusion and transfer may be conditioned by institutional variables. The question as to what extent the characteristics of the policy themselves have an impact on diffusion and transfer has so far almost been ignored in the literature. Is it easier to adopt general principles rather than specific models (Weyland 2006)? Are relatively technical polices such as DRGs more ‘contagious’ than politicized reforms? Systematic studies across different policies are still lacking. The concept of ‘institutional fit’ combines both domestic and policy-specific features. To date, this approach has only been applied in the literature on Europeanization (Börzel 1999; see Martinsen 2005 for an application of this approach to EU social policy), but it is certainly more widely applicable. In principle, certain reforms should diffuse if they exhibit an institutional fit; that is, if they can easily be combined with existing policies. In contrast, an institutional misfit should hamper the diffusion of a policy. Workfare reforms, for example, have tended to spread well among English-speaking countries while continental European countries, in particular, have followed only belatedly or hesitantly. This may well be due to the dominance of social insurance principles in conservative welfare states. Entitlements to contributory and earnings-related benefits do not fit well with the reciprocal logic of workfare, since beneficiaries tend to regard the insurance contributions paid in the past as a sufficient precondition for benefit receipt. In other words, there may be a misfit between the policy idea of workfare and the traditional principles of social insurance.

Fourth, there are significant research gaps in comparative welfare state research, as mentioned above in detail. For example, given the improvements in methods, the question raised by Collier and Messick (1975) regarding possible effects of diffusion and transfer during the earliest phase of welfare state development should be readdressed. Empirical analyses do not necessarily have to be limited to the old welfare states but could just as well extend to other regions in the world where the consolidation of the welfare state has only just entered the agenda (Kangas 2012). At the same time, the dominant focus on transfer programmes should be readjusted and systematically take the introduction of important regulations in labour law into account. When studying the determinants of social expenditure, private social expenditure has so far not been looked at, a type of expenditure that has been growing in recent years – perhaps also due to diffusion. When it comes to the level of welfare state generosity, scholars have placed a lot of emphasis on replacement rates, while aspects such as benefit duration, coverage and eligibility have been underappreciated. Furthermore, the regulatory dimension of the welfare state has received short shrift, although international economic competition should play an important role here. In the area of employment protection legislation, for example, a strong downwards convergence has taken place since the 1980s which cannot be explained by existing theories of comparative policy analysis (Höpner et al. 2011). Lastly, there is room for research about the ‘migration of concepts’ such as ‘activation’ and the ‘social investment state’ as well as about the reforms that were based on these concepts. The same holds for the diffusion and transfer of new policy instruments such as negative income taxes and paid parental leave policies.

Finally, the systematic combination of processes of horizontal spatial interdependencies with vertical relationships between international and supranational institutions and their member states – i.e. a topic that has not been addressed in this article – is perhaps the biggest challenge for future research. For example, (allegedly) successful national policies are supported and promoted by international organizations or ‘uploaded’ by their member states (see, for example, the three-pillar-model of old age security, Leimgruber 2012). Horizontal and vertical interdependencies in the area of social policy have thus far not been systematically combined, neither theoretically and conceptually, nor methodologically and empirically. Nonetheless, given the multiple and deepening economic and political interrelations between nation states, the systematic analysis of policy diffusion and policy transfer will be of central importance in comparative welfare state research in the future.

Acknowledgements

We would like to thank the editor and the anonymous referee for their very helpful comments.

Notes

1. 1

   Vertical interdependencies are the main focus of research about social policy in the European multi-level system. Since the late 1990s, the EU has relied on the Open Method of Coordination to stimulate collective learning processes in which the agreement on common goals, benchmarking and learning from best practices represent important mechanisms of a new soft governance in social policy (Heidenreich and Zeitlin 2009; Zeitlin et al. 2005; Zeitlin 2009). At the global level, international organizations promote, via reports, data collection and debate forums, the exchange of information among their member states. In addition, the IMF and the World Bank rely on conditionality since loans or debt reliefs are closely tied to domestic policy reforms. Moreover, international organizations such as the ILO can enact regulatory standards which, ratification by member states provided, become binding and have in fact influenced national social standards and social expenditure (e.g. Strang and Chang 1993; Usui 1994; Senti 1998; Abu Sharkh 2010). Regarding the OECD and the IMF, see Schäfer (2006) and Armingeon (2004); for an overview of ‘global social policy’, see Deacon (2007) and Yeates (2008).

2. 2

   The growing scholarship on policy diffusion and transfer is paralleled by a mounting literature on policy convergence (Knill 2005). On welfare state convergence, see, for example, Bouget (2003), Starke et al. (2008) or Schmitt and Starke (2012).

3. 3

   Spatial interdependencies can also be modeled in the error term. This implies that spatial patterns are caused by unobserved heterogeneity or measurement errors (Anselin et al. 1998). In this case, no new variable is included in the model and only the error term is decomposed.

4. 4

   Estimating simultaneous interdependencies is associated with several methodological problems. The spatial lag is a weighted average of the dependent variable and therefore endogenous. Due to endogeneity problems a spatial lag model cannot be estimated with Ordinary Least Squares (OLS). OLS estimations would be inconsistent in this case. In some studies the spatial lag is therefore time lagged. This solves the problem only in the absence of serial autocorrelation. In the case of instantaneous independencies Maximum Likelihood estimators perform best (Franzese and Hays 2007, 2008; Hays 2009).

5. 5

   Social Network Analysis is typically applied when aiming to describe interdependencies, but so far has not been applied to analyzing the effects of interdependencies for policy change.

6. 6

   The same reasoning must be applied to the ‘method of agreement’.

7. 7

   Apart from the studies mentioned, there are many inquiries that investigate the diffusion of social policy at the subnational level (e.g. Volden 2006; Karch 2007). Policy diffusion and transfer is of particular relevance in federal countries. The constituent units are considered as ‘laboratories of democracy’. Local policy innovations are sometimes associated with spill-over effects to the federal level or may spread across the constituent units (see Obinger et al. 2005: 340–3) for local social experiments and social policy diffusion in six federal states). However, subnational policy diffusion is not restricted to federal states but also of relevance in decentralized polities (e.g. smoking ban legislation within the UK). Since we are concerned with cross-national comparisons, these contributions are not considered here in more detail.

8. 8

   Schmid and Götze (2009) corroborate this finding in a qualitative study.

9. 9

   Weyland also shows that there is no systematic evidence demonstrating that international organizations such as the World Bank have an impact on social reforms.