Interplays Between Welfare Regimes Typology and Academic Research Systems in OECD Countries

ARS in liberal regimes

Liberal regimes (e.g., Australia, Canada, New Zealand, the United Kingdom and the United States) are characterized by market-differentiated welfare for most citizens, means-tested assistance for a minority, social guarantees restricted to ‘bad risks’ and residual state intervention. Pechar and Andres’ (2011) analysis of higher education expansion and funding also suggested liberal regimes would tolerate greater inequalities of conditions but would ensure equality of opportunity through high private and medium public investments. Three underlying dimensions thus shape welfare distribution, economic coordination, higher education and, potentially, ARS: market-like behaviors, competition and stratification (Hall and Soskice, 2004).

First, liberal regimes maintain a low level of de-commodification where the market takes care of most good risks (benefits) and the state absorbs bad risks. As research may lead to benefits, liberal states limit their intervention in ARS to protecting private property rights and free markets (Olson and Slaughter, 2014). In the United States, the proportion of federal money to university research has decreased to the benefit of a private sector encouraged by tax credit schemes (Cummings, 2012). In Canada, the government has also established one of the most generous tax incentive programs so that 7% of university research is funded by industry (Sá and Litwin, 2011). For Harman (2010), a similar increase in industry-university collaborations in the United Kingdom has led to more joint scientific publications, accounting for 50% of all industrial scientific output. In brief, markets and private enterprises play a more prominent role in liberal regimes’ welfare mix (Esping-Andersen, 1990).

Second, public research funding remains important but block grants have been gradually replaced by ‘output-oriented core funding with great share of external funding’ (Auranen and Nieminen, 2010, 824). Coordination in liberal regimes is partly achieved through competition (Hall and Soskice, 2004), and even when the state supports research, competitive grants and research contracts encourage market-like behaviors. New Zealand and the United Kingdom use a separate stream of institutional funding for research and education and base that funding on performance measures (Harman, 2010). In Australia, government funding depends on institutions’ research inputs (competitive, public, industry and other funding) and outputs such as publications and graduate degrees awarded (Geuna and Martin, 2003).

Third, market and competition would result in what Benner (2011) names vertical segmentation, where ‘a limited number of universities have reinforced their position in the research system on the basis of aggressive managerial strategies to recruit leading scientists, large endowments, and strong position in the highly stratified funding system’ (14). In addition to revenue generation, international students and scholars are central to universities’ ability to compete for funding and recognition (Avveduto, 2010).

ARS in conservative regimes

Following the Austrian–Germanic corporatist legacy established by Bismarck and von Taffe, conservative regimes (e.g., Austria, Belgium, France, Germany, Italy, the Netherlands and Switzerland) rely more on public than private benefits, but consolidate divisions among wage earners by legislating different programs for different groups, thus upholding status differentials and often granting the highest ranks to civil servants (Esping-Andersen, 1990). In some countries, the Church’s influence has contributed to ‘familialism,’ that is, reliance on families to provide child and elderly care. Esping-Andersen (1999) observes, however, that the conservative group is the least homogenous as familialism is less dominant in France and Belgium, and the Netherlands shares numerous social-democratic features (Esping-Andersen, 1999).

For Pechar and Andres (2011), conservative regimes maintain equality of condition and are the most generous to citizens who do not achieve higher education. HES are characterized by slower expansion and lower levels of support, as well as early-tracking, a high proportion of students in vocational studies and few study grants. Despite the notable heterogeneity among countries, their ARS can be distinguished from other regime types by their stratification.

This stratification manifests firstly as an explicit hierarchy between general education and applied education (Benner, 2011), and secondly with regard to research personnel. In conservative regimes, public research institutes play a dominant role in conducting applied research, whereas higher education institutions train the workforce and conduct basic research (Meyer-Krahmer, 2001). For instance, most German academic research is conducted either in organized research institutes or scientific universities, while polytechnics remain teaching oriented. France, for its part, has a dual system where universities have non-selective entrance policies and limited research infrastructures (Krull, 2013).

Stratification relates thirdly to the prestige of research universities. In conservative regimes, the state supports ARS through channeled competition and resource-stratified systems. In 2005, Germany selected nine universities to become globally visible. The Exzellenzinitiative entailed a €2.7 billion budget to support selected universities, graduate schools, clusters of excellence and institutional strategies (Olson and Slaughter, 2014). In France, the Initiative d’excellence started in 2011 with a €7.7 billion budget aimed at developing 5–10 research clusters (ANR, 2010). ARS in conservative regimes can thus be distinguished by the prominence of the state in the welfare mix and channeled competition that upholds status differentials.

Research governance also seems influenced by existing industrial strongholds, although actual networking between universities and private enterprises based on the private sector’s size and fields of activity. For instance, such interactions are crucial within Austria’s large innovation system (Federal Government of Austria, 2012), while they appear weaker in France (Cour des Comptes, 2013).

Finally, the rising regionalization of Europe influences ARS. The EU’s Horizon 2020 has a budget of €77 billion and promotes intra-European cooperation in research (ERC, 2013). In terms of research production, more than 45% of scientific publications are co-authored with foreign research teams, of which 57% come from the EU (Cour des Comptes, 2013). Regionalization also takes the form of postgraduate mobility, with 31% of post-Ph.D. EU researchers being labeled as ‘internationally mobile’ for at least 3 months of the last 10 years (EC, 2014).

ARS in social-democratic regimes

Social-democratic regimes (e.g., Denmark, Finland, Norway and Sweden) implement comprehensive social policies (socialization of risks), entitlement programs and universal access to quality services (universalism) in view of moderating the influence from both the market and the traditional family. These regimes focus on positive equalizing and freedom of choice for their citizens, and regulate markets through consensus between the state, unions and employers (Esping-Andersen, 1999).According to this model of information sharing, institutions prevent the state from dictating change by participating in the policy-making process and establishing ‘buffer’ organizations (Hall and Soskice, 2004). In their analysis, Pechar and Andres (2011) suggest that social-democratic regimes avoid the trade-off between equality of condition and equality of opportunity by expanding access while providing social benefits for citizens who cannot access higher education.

De-commodification and cooperation with other societal actors seem to be central to this regime type. Following principles of the Reformation, higher education is assumed to serve the public good and higher institutions are either publicly owned or monitored (Kalpazidou Schmidt, 2007). Public research funding takes the form of large block grants, which for instance cover 50% of public-sector research costs in Sweden (Brundenius et al., 2011). These large block grants are increasingly based on the number of degrees awarded, publications, citations and external funding (Eliasson, 2009) and Quality assurance mechanisms seem to be implemented as alternatives to marketization of higher education (Välimaa, 2001). Unlike in conservative regimes, these steering mechanisms would, however, be implemented in a context of mutual trust and consensus (Auranen and Nieminen, 2010).

The social-democratic state strongly supports its institutions but also mirrors liberal regimes by fostering networks between academic and non-academic actors. Networking in social-democratic countries is both consensus and top-down driven,with governments often creating networking agencies (Vinnova, TEKES or VRI) and rewarding institutions that develop partnerships (Herstad et al., 2010). Nordic states also compensate for their small size by means of an internationalized academic research process (e.g., NordForsk).

Regarding research personnel, it should be mentioned that the institute sector is increasingly merged with universities (Benner, 2011). Doctoral students also contribute greatly to the production of research, which is one of the reasons that led Denmark to double its number of Ph.D. students as part of its research policy initiatives (Jensen, 2007). In Denmark, Norway and Sweden, doctoral students have the status of university employees who are paid by the university and, in some cases, are required to work as research/teaching assistants (Jensen, 2007). On average, the number of doctoral degrees conferred by Nordic universities between 2002 and 2011 increased by 32% (Myklebust, 2013).

Data

This study examines if the welfare regime typology can be used as a framework for understanding correspondence between 16 OECD countries and 12 ARS indicators. In his original work, Esping-Andersen (1990) studied 18. Pechar and Andres (2011), for their part, only considered data for 16 countries, leaving out Spain and Portugal in order to have complete data and groups of approximately the same size. For these same reasons and to ensure comparability with Pechar and Andres’ study, we chose to limit our analysis to the same countries spread across the three regime types described above: Austria, Belgium, France, Germany, Italy, the Netherlands, Switzerland; Australia, Canada, New Zealand, the United Kingdom and the United States; and Denmark, Finland, Norway and Sweden.

Focusing on process rather than outputs, we searched OECD databases for ARS indicators and proxy variables for the four ARS features listed above: research personnel, funding, networking with private enterprises and internationalization. These features are crucial to ARS because they give a sense as to which stakeholders within the system are conducting and supporting research.

Had statistics been available for the 16 countries with regard to working conditions, governance or institutional differentiation, a more comprehensive analysis could have be processed. That said, the present attempt nonetheless covers ARS core features.

Firstly, regarding research personnel, the share of research performed by private enterprises, government institutes and HES differ in liberal, conservative and social-democratic regimes (Benner, 2011). We used the percentages of researchers working for governments, businesses and higher education institutions in 2012 as proxies for the relative share of each sector (OECD, 2014b). As the participation of doctoral students to the overall research production is also distinctive in some ARS (Jensen, 2007), we looked at doctoral graduation rates as a percentage of the population in the reference age cohort (OECD, 2014a) as a proxy for the number of doctoral students. We would have preferred to use the variable ‘entry level in advanced research programs’ (OECD, 2014b), but Belgium, Canada and the United States had missing data.

Secondly, regarding funding indicators, the balance of different funding types reflects countries’ priority settings and involvement (Harman, 2010). By dividing aggregated higher education research expenditures in higher education by GDP size (OECD, 2014b), we captured the relative importance of academic actors (Cummings, 2014). We also included countries’ general university funds (GUF) as a percentage of non-military government budget appropriations or outlays for R&D (OECD, 2014b). The United States reports no funding of this type at the federal level while any research funding through GUF that takes place at the state and municipal levels accounts for less than 5% of higher education research funding and less than 1% of total higher education R&D or HERD (Kennedy, 2012; National Science Foundation, 2014). In light of these considerations, the OECD percentage of 0 was maintained.

We also considered direct government research funding (DGRF) as a percentage of public funds (OECD, 2009). The last year for which this indicator was available was 2006, after which the OECD decided to develop a new indicator, namely, project funding in universities. This new indicator has yet to be available for all countries. Such a discrepancy in years of reference presents major issues; however, we conducted the CA with and without this indicator and found that it affected neither the overall stability of the model nor the results.

Thirdly, reference to non-patent literature in patents (OECD, 2013) is as an indicator of the relationship between research and development (Meyer-Krahmer, 2001). The study also considered the percentage of higher education R&D financed by industry (OECD, 2014b) and the percentage of industry-science co-publications in total research publications between 2006 and 2010 (OECD, 2013). Co-publications were not taken as a measure of output as we are not interested in countries’ total number of publications but rather in the involvement of industry in the research process.

Finally, regarding internationalization, we analyzed international co-authorship (2007–2011) as a percentage of total publications (OECD, 2013). Again, we are not interested in the output but rather in the extent of international collaboration in the research process. The last indicator considered was the enrollment of international students in an advanced research program (OECD, 2014b).Yet data for France and Italy refer to foreign students (i.e., students who are not citizens of the country where they are studying) while international students are students who have crossed borders with the intention to study. For France, the proportion of foreign students that are international students was calculated with data from the MESR (2014). Unfortunately, no comparable sources were found for Italy.But as CA does not allow for missing data (Greenacre, 2007), we relied in the case of Italy on the number of foreign students, understanding that this necessarily limits comparability for this indicator.

Analysis

To show the robustness of a well-established model, Esping-Andersen (1999) used logistic regressions based on composite indicators. As the correlation between the different ARS indicators and the countries remains unknown, we opted for an exploratory method that promotes deeper understanding of research production and identifies potential dimensions to be further confirmed. As Pechar and Andres (2011) achieved using variables related to higher education participation and funding, we conducted a CA. A CA is a ‘method of displaying rows and columns of a table as points in a spatial map’ (Greenacre, 2007, 264) in view of interpreting correspondences and distinctions within and between rows and columns.

CA is a valuable tool to compare cases in an intuitive graphical format. In this article, we interpret ‘countries’ as one variable (row) with 16 different categories, and ‘ARS indicators’ as one variable (column) with 12 different categories, thus resulting in a table of 192 cells. These cells create a cloud in a three-dimensional space, and dimensions are projected into the cloud following an ordinary least square method. Cells in the table, shown as points in a cloud, thus affect the orientation of the dimensions, and cells with a higher N and outliers will have greater effect on the orientation (Greenacre, 2007). Cells that contribute most help explain dimensions.

Like Pechar and Andres (2011), we relied on a non-parametric CA where each ARS indicator was ranked in increasing order from 1 to 16. This created a positive pole for each indicator. The data was then doubled and the scale reversed to create a negative pole (see Greenacre, 2007). The coordinates (mirrored) and weights were, therefore, roughly the same for both negative and positive poles. This type of ranking circumvents issues related to the different scales of the ARS indicators. In this way we lose information but are able to produce a stable analysis where no point affects dimensions for more than 45 degrees. We also conducted a multiple CA using indicators as categories, which presented the same trends as the CA presented below.

The CA demonstrates correspondences between country profiles and indicators. Welfare types were kept as supplementary, meaning that they did not affect dimensions but could be used as ellipses to see if the typology had any explaining power. Analyses and factorial planes were made with Coheris Analytics SPAD.

Limitations

Despite a call for international comparisons studying macro-societal phenomena, Teichler (2014) recognizes that the comparative empirical process is tremendously complicated, especially when one considers that even such widely used terms as ‘universities’ and ‘international students’ have different meanings in different countries. Using country data is also quite common, although it implies homogeneity within countries (Bray and Thomas, 1995). In this study, data from Canada, Germany, the United States and Switzerland need to be considered with caution because federated entities play a strong role in regulating and coordinating HES (Cortés and Teichler, 2010). It needs to be emphasized, however, that variables figuring in the present study were selected based on an extensive literature review and their analysis conducted by means of a robust exploratory method.

ARS in 16 OECD countries

As showed in Table 1, social-democratic regimes have higher graduation rates at the doctoral level than liberal regimes, of which only the United Kingdom achieves a rate above 2. The conservative group could be subdivided into two groups, with Belgium, France and Italy counting fewer doctoral graduates than Austria, Germany and Switzerland. Higher doctoral graduation rates are associated with fewer researchers in governmental institutes in Austria, Switzerland, the United Kingdom, Denmark and Sweden. We hypothesized that more researchers worked in the private sector in liberal regimes, although this is only the case for Canada and the United States.

With the exception of Norway, HERD is higher in social-democratic countries and lower in liberal ones. Again, conservative countries form two subgroups: Austria, the Netherlands and Switzerland invest more in HERD than Belgium, France, Germany and Italy. GUF is significantly lower in liberal regimes, while no clear difference is observed between conservative and social-democratic regimes.

Correspondence analysis (CA)

The first dimension is tentatively named ‘Academic Centrality/Peripherality’ and amounts to over 30% of the variance alone. It is the most important dimension of this analysis. ARS indicators that contribute to positive coordinates of Academic centrality are the positive pole of GUF, HERD as a percentage of GDP, percentage of industry-science co-publication, graduation rates at the doctoral level, international co-authorship and the negative pole of DGRF. Countries over the threshold value corresponding to the positive pole include Switzerland, Austria, Sweden and Denmark. Under the threshold value, but still contributing, we find the Finland and Norway. With negative coordinates (referring to Academic Peripherality), we find three liberal countries contributing over the threshold value: the United States, Australia and New Zealand. For conservative regimes, the portrait is more mixed: Switzerland, Austria and Germany (although not over the threshold value) correspond to negative coordinates, while Belgium, France and Italy (while not over the threshold value) correspond to negative coordinates (Figure 1).

Open image in new window

The second dimension, named ‘Academic/Non-Academic Research Workforce’, amounts to over 20% of the variance. Of the four ARS indicators contributing to this second dimension over the threshold value, three relate to the sector in which researchers work. Together, these three indicators account for 35.5% of the contribution from the column variables. Positive coordinates (Academic Research Workforce) correspond to the positive pole of researchers working in higher education institutions, as well as to a high percentage of international students in advanced research programs. At the negative poles we find researchers in business enterprises and governmental institutes. The countries on the positive side of the dimension over threshold value are Switzerland, the United Kingdom and Australia. With negative coordinates (Non-Academic Research Workforce), the opposite poles correspond with the United States and Germany.

The third dimension, named ‘Responsiveness/Non-Responsiveness to Market Forces’, amounts to slightly over 15% of the variance. Positive coordinates (Responsiveness to Market Forces) correspond to the positive poles of patent citing non-patent literature, percentage of HERD financed by industry and, as in the first dimension, DGRF as a percentage of public funds for academic research. Belgium and Canada are the only two countries over threshold value that contribute to this side of the dimension. Meanwhile, negative coordinates (Non-Responsiveness to Market Forces) correspond to the negative poles of the ARS indicators mentioned above and to Italy, whose contribution to this dimension is over 50%. This dimension does not tell us as much as do the first two, the reason being twofold: firstly because of the lower variance, and secondly because of Italy’s strong contribution (Figure 2).

Open image in new window

The factorial planes provide us with a visual map of the findings and allows us to inspect the supplementary ellipses based on welfare typology. Dimension 1 divides the liberal and the social-democratic welfare regimes, although not all countries contribute to this dimension over threshold value. Dimension 2 does not identify an opposition where welfare typology suggests correspondence. Dimension 3 only singles out one liberal and one social-democratic country, namely, Norway and the United Kingdom.

Dimension 1: Academic centrality/peripherality

The first dimension explains most of the variance and differentiates between liberal and social-democratic regimes. Because it is concerned with higher education funding support, doctoral education and co-publications, this dimension can be interpreted as the extent to which ARS sit at the core of and lead research production.

In his description of the ‘academic oligarchy’, Clark (1983) suggested that institutions first seek the autonomy offered by an unfettered lump-sum grant from the government, and secondly a buffer organization that understands the sector. This appears compatible with ARS which have positive coordinates on the first dimension, although we use the word ‘centrality’ instead of ‘oligarchy’ because indicators refer less to authority distribution than to the relative prominence of academic activities in the research process. This might explain why the Italian ARS appears peripheral in our analysis whereas in the context of a weak bureaucracy it was considered by Clark to be an academic oligarchy. In contrast, Clark noted strong state coordination in Sweden while our analysis points out to academics’ relative autonomy and support in conducting research.

Thus, on the positive pole, we find a high proportion of funding in the form of GUF. Despite a general decrease of GUF in OECD countries (Vincent-Lancrin, 2009), these non-oriented block grants given to higher education institutions for research purposes remain important in central ARS (OECD, 2013). Danish and Swedish ARS are less externally funded (Auranen and Nieminen, 2010), and in Switzerland the Confederation provides basic funding, while the binary research system ensures both academic freedom and close partnership with private sectors, particularly for universities of applied sciences (Cortés and Teichler, 2010; OECD, 2011).

Although increasingly performance-based, GUF limit society’s interference in deciding what research projects are to be funded, ensure stable revenues and allow institutions to maintain their research infrastructures (Thorens, 2006; Bonaccorsi, 2007; Sörlin, 2007). Extending the concept of de-commodification to GUF may be a bit of a stretch as research funding and conducting research cannot be considered a ‘right’. That said, the associated concept of ‘socialization of risks’ observed in social-democratic regimes (Yang, 2014) is relevant because depending on the funding formula and institutions’ allocation mechanisms, GUF might imply that a diversity of projects can equally flourish despite their risky nature. This could explain why the four social-democratic countries studied have central ARS, though this relationship needs to be further studied.

Central ARS also have higher graduation rates at the doctoral level and we assume they count more doctoral students. High doctoral enrollment could be the result of interactions between academic traditions and political-economic conditions. Central ARS are influenced by a Humboldtian tradition in which doctoral students often hold a full-time employee status and contribute to the strength of academia (Ahola, 2007). Central ARS also seem to be associated with countries inspired by ‘productivism’ (Esping-Andersen, 1999), that is, the view that welfare states must maximize the productive potential of citizens by guaranteeing work and resources for all people who have the motivation and capacities.

ARS on the left side (negative coordinates) could be qualified as ‘peripheral’ not because of their smaller research output but because of their responsiveness to externally defined priorities. In raw numbers, American and British ARS lead world-science (Royal Society, 2010). In relative terms, however, ‘peripheral’ ARS rely on smaller HERD, are more externally funded through competitive mechanisms and respond to industrial and/or governmental priorities (Musselin, 2007; Auranen and Nieminen, 2010; Kennedy, 2012).

In ‘peripheral’ ARS, graduation rates at the doctoral level are smaller. In the case of liberal regimes, it could be explained by greater recognition of bachelor degrees and the cost of studies (Maslen, 2010). The commodification of graduate education, however, are off set by the higher return-value of PhDs. In contrast, lower graduation rates in conservative regimes can be explained by social stratification and systemic segmentation (Esping-Andersen, 1999; Benner, 2011). In Italy, lower graduation rates are associated with fragmented programs attracting primarily incumbent students (Bonaccorsi, 2007), while in France they can be traced to the prestige of degrees coming from Grandes Écoles and a weak financial support for doctoral studies (Joly, 2005).

Dimension 2: Academic/non-academic research workforce

As stated by Clark (1983, 28), ‘In the beginning, there is work, for if we reduce a knowledge-bearing system to its primordial elements we find first a division of labor, a structure of organized effort within which many people individually and collectively take different actions’. In following this, the second dimension refers to the division of the research workforce, and makes a distinction between welfare regimes. Positive coordinates reflect a higher proportion of researchers working in higher education institutions and, correspondingly, a greater proportion of international students at the graduate level. Negative coordinates, for their part, reflect higher proportions of researchers working in governmental institutes. Despite reforms in the 1980s supporting university research, French, German and Finnish governmental institutes continue to attract top scientists who spend more of their time conducting research than professors and compete for similar funding (Musselin, 2007; Aarrevaara, 2014).

One might wonder the extent to which ARS may consider a lower proportion of researchers working in higher education as ‘academically central’. These ARS co-exist with knowledge-intensive industries (ICT in Finland or pharmaceutics in Sweden) that encourage the production of Ph.D.s and co-publications (Jensen, 2007; Peterson, 2011). Carlsson (2006) describes these ARS as characterized by high levels of knowledge flows and attractive to firms wishing to take advantage of spillover opportunities. Similarly, ARS in the upper-left quadrant in Figure 1 could be characterized as both large and ‘peripheral’. Instead of relying on institutes to conduct research in priority areas, these countries steer and use research conducted in ARS (Benium, 2007).

Dimension 3: Responsiveness/non-responsiveness to market forces

Clark (1983, 162) defines market-based coordination as ‘unregulated exchanges [that] link persons and parts together’ and include consumer, labor and institutional markets. The third dimension here is named ‘responsiveness to market forces’ and includes the indicators DGRF, patents citing non-patent literature and HERD financed by industry, all of which seem to relate to and demand for specific research projects. For Singh (2012), responsive systems have the capacity to meet the needs of the knowledge economy, deliver research, as well as highly trained people and address national needs. Olson and Slaughter (2014) argue that HES are ‘penetrated’ by the new economy, whereas our framework suggests rather that ARS remain distinct entities that may allow more or less external influence to orient their work.

The third dimension explains less variance than the first two but singles out the United Kingdom from the liberal group and Norway from social-democratic group. This dimension provides nuance to the first dimension with its contribution from the indicator related to university-industry collaborations. DGRF also contributes to this dimension — although unlike in the first dimension where it is understood in opposition to GUF, it refers here to networking with businesses. DGRF relies on quasi-market mechanisms and requests increasingly that academics find private partners in order to prove ‘demand’ for their research (Harman, 2010).

In the upper-left quadrant in Figure 2, ARS can be characterized as both ‘peripheral’ and ‘responsive to market forces.’ For example, Canadian research councils provide increasing stimulation for public–private partnerships while the country’s tax credit scheme encourages businesses to contribute to the HERD (Sá and Litwin, 2011). Similarly, in the US governmental support for development was found to outpace investments in basic and applied research (Kennedy, 2012). Belgium’s closeness to Anglo-Saxon countries might be explained by its recent initiatives to promote inter-sectoral mobility in the form of staff exchanges, honorary positions and financial incentives, among others (EC, 2006). On the other hand, Italy appears in the lower-left quadrant because of small business size and investments knowledge-intensive enterprises (Deloitte, 2014).

Interestingly, ARS in the upper-right quadrants appear both ‘central’ and ‘responsive to market forces’: they benefit from DGRF and higher education in such systems appears strongly connected with the private sector. As predicted by Benner (2011), Nordic countries mirror their Anglo-Saxon counterparts (both groups having negative coordinates with the exception of Norway and the United Kingdom) insofar as they foster collaborations with the private sector but prefer top-down strategies implemented by government-supported networking agencies and private foundations (Herstad et al., 2010). At first sight this appears to stand in contradiction with social-democratic public ownership (Esping-Andersen, 1999), buton this point it is worth noting that responsiveness is neither deregulation nor privatization. ARS in social-democratic contexts might, for example, be encouraged to participate in the innovation process (Lepori and Kyvik, 2010) or in evidence-based policymaking (Arter, 2008), but consensus-based decision making, buffer organizations and academic traditions may still protect their ‘central’ character. These relationships should be further tested in future studies.

Another look at the welfare regime typology

The ultimate purpose of this paper has been to observe if Esping-Andersen’s (1999) welfare regime typology explains correspondence between the 16 considered countries and ARS indicators. This study does not attempt to identify a causal relationship, but a correspondence. First, it should be pointed out that indicators related to internationalization and research workforce contribute less to differentiating welfare regimes than indicators related to funding, doctoral education and networking with private actors.

All countries contributing to the first dimension’s negative coordinates above threshold value belong to liberal regimes, and all social-democratic regimes correspond with positive coordinates. ARS in social-democratic countries would appear to be both ‘academically central’ and ‘responsive to market forces’, while their counterparts in liberal regimes can be characterized as ‘academically peripheral’ and ‘responsive to market forces’.

Benner (2011) wonders if there is a distinct Nordic model. This study indeed supports the claim that Nordic countries can be grouped together and that they differ from Anglo-Saxon countries. In their attempts to adapt to a globalized post-industrial era (Esping-Andersen, 1999), HES have expanded in both social-democratic and liberal regimes (Pechar and Andres, 2011). Meanwhile, the welfare mix in each respective case differs because the former relies on public investments while the latter relies on private investments. The liberal regimes let ARS respond to market forces, avoid ‘bad risks’ by funding research through competitive funding and provide needs-based or merit-based funding for doctoral students. Social-democratic states would rather socialize the risks inherent in academic research (Benner, 2011), increase funding for doctoral researchers following a productivist principle, and ensure the centrality of publicly-funded higher education institutions in the academic research production process. Welfare mix, productivism and socialization of risk might thus inform our understanding of the multiple ways in which ARS and welfare regimes interact.

However, in contrast to Pechar and Andres’ (2011) study of HES, the present study shows that conservative countries do not form a coherent group. For Esping-Andersen (1999), the conservative type was the least robust, with France and Belgium appearing distinct from Germany and the Netherlands appearing closer to the social-democratic regime. It is possible that additional indicators related to excellence initiatives and stratification maintain the coherence of the conservative type (Olson and Slaughter, 2014), but no indicator on those types could be found with data for all 16 countries.

Some findings also suggest two potential subsets which future research could use alongside analyses of variance to further differentiate the conservative type. Political economists have already proposed typologies including a fourth type (Ferrera, 1996). More recent studies (e.g., Yang, 2014) observe that conservative regimes respond to ‘new social risks’ quite differently: Belgium, France, Italy and Spain having undergone significant reforms while other Southern European countries have maintained their existing system. In her study of R&D investments, Kim (2013) even distinguishes CMEs from Mediterranean market economies, characterizing the latter by large agrarian sectors, legacies of extensive state intervention and a mix of non-market coordination for corporate finances and liberal arrangements for labor relations.

Subdividing the conservative group is also supported by a cultural analysis of academic traditions. For instance, some ARS in conservative regimes are close to ARS in social-democratic countries because of their shared Humboldtian tradition (Välimaa, 2001). Despite reforms, long-standing values remain perceptible in doctoral degrees’ high social prestige, standardized employment schemes, input-oriented research funding and welfare policy tradition (Auranen and Nieminen, 2010). However, it is not entirely clear that ARS in the second conservative sub-type (which includes Belgium, France and Italy) would be influenced by a common academic tradition.

Findings nonetheless point to a new framework for apprehending differences in countries’ academic research production process by expanding Esping-Andersen’s (1999) theory to a new set of phenomena. Concepts such as productivism, the socialization of risks and welfare mix shed new light on the interactions between countries’ political-economy and the structures, conditions and processes contributing to research production in academic settings.