Abstract
The business systems approach holds considerable promise for improving our understanding of the relations between societal institutions and technological and economic outcomes. Nonetheless, there have been surprisingly few attempts to validate its proposed typology of business system types. In this paper, I take up this issue and conduct a large-scale empirical assessment of the national business systems typology. I use data on 30 OECD countries from 2000 and 2011 to assess the validity of the typology, and explore its value for comparative institutional analysis through a fuzzy-set analysis of innovation specialization patterns. The findings illustrate that while the national business systems typology needs to be extended, it remains relevant for describing variety in national institutional frameworks. In addition, the detail it adds may provide the nuance needed for exploring more complex relations between institutions and technological and economic outcomes.
A central argument in comparative institutional analysis is that societal institutions affect the organization of economic activities and, thereby, a range of organizational- and country-level outcomes (Hall & Soskice, 2001; Jackson & Deeg, 2008; Morgan, Campbell, Crouch, Pedersen, & Whitley, 2010). For instance, comparative work on innovation posits that institutions and the degree of coordination that they promote encourage firms to pursue certain innovation strategies (Casper, Lehrer, & Soskice, 1999; Hall & Soskice, 2001). Similarly, complementarities between institutions such as between corporate governance systems and labor market institutions may support some competitive strategies more than others (Crouch, 2005; Lehrer, 2001; Sorge & Streeck, 1988; Vitols, 2001). Consequently, comparative institutional analysis suggests that societal institutions offer countries comparative advantages in some activities but not in others, and that those institutions encourage divergent economic and technological specialization patterns (Hall & Soskice, 2001).
While this proposition is intuitively appealing, empirical studies of the links between societal institutions and economic and technological outcomes are often inconclusive (Akkermans, Castaldi, & Los, 2009; Herrmann, 2008; Kenworthy, 2006; Kogut & Ragin, 2006; Taylor, 2004). A likely reason for these inconsistent findings is that the relationship between institutions and economic outcomes is not deterministic. Actors may interpret and use institutions differently, and firms often have considerable leeway to circumvent national institutional constraints (Jackson, 2010; Sorge, 2005). For example, firms may draw on foreign institutional resources (Allen & Whitley, 2012; Herrmann, 2008; Lange, 2009; Schneider, Schulze-Bentrop, & Paunescu, 2010), and national institutional frameworks often encompass institutional niches that favor certain activities (Crouch, 2005; Lane & Wood, 2009, 2012; Lange, 2009; Parker & Tamaschke, 2005; Schneiberg, 2007).
An additional possibility is that the theoretical frameworks used to explain technological and economic outcomes fail to capture sufficient institutional variety. Empirical assessments of the relations between institutions and technological and economic specialization patterns often draw on the insightful but relatively crude distinction between liberal and coordinated market economies (Hall & Soskice, 2001). As a result, more subtle differences between national institutions may remain undetected, and the extent to which different institutional frameworks can promote similar outcomes may be underexplored (see Boyer, 2004). This suggests that a more nuanced comparative institutional approach might be needed to capture institutional variation across countries (Casper, 2010).
The purpose of this paper is to validate one such approach—the national business systems approach (Whitley, 1992, 1999, 2000)—and to explore its value for comparative institutional analysis. Within comparative analyses of capitalism, the business systems approach stands out not only because it makes fine-grained distinctions among alternative forms of non-market coordination but also because it more explicitly focuses on the micro–macro links between institutions and the organization of economic activity (Jackson & Deeg, 2008). Consequently, the business systems approach can differentiate among a greater number of distinctive types of market economies than related approaches, such as the Varieties of Capitalism approach (Hall & Soskice, 2001).
This suggests that the business systems approach may help provide a more detailed understanding of the links between societal institutions and technological and economic outcomes (Jackson & Deeg, 2008; Whitley, 2007). Yet, despite its promise, there have been no attempts to systematically validate its proposed typology of business system types. As a result, several key questions about the accuracy of the business systems typology remain unanswered. For example, we lack knowledge on whether the typology is exhaustive and whether its business system types represent more universal forms of economic organization. The absence of larger-scale analyses also means that it is often unclear how to classify market economies that have so far escaped the attention of business systems scholars.
In this paper, I aim to take up these issues in a first attempt to validate the business systems typology through a large-scale analysis. To this end, I first conduct a two-stage cluster analysis using institutional data on 30 OECD countries from 2000 and 2011. This produces an empirical classification, or taxonomy (Bailey, 1994), of empirically identifiable institutional business system configurations. I then conduct a fuzzy-set QCA analysis of revealed innovation specialization in high-tech sectors. Because innovation patterns form a central theme in comparative institutional analysis (Casper, 2010), this analysis allows me to explore and illustrate the explanatory power of the business systems approach relative to the explanatory power of the Varieties of Capitalism approach (Hall & Soskice, 2001).
The analyses reveal several interesting findings. First, I find that Whitley’s (1999) business system types can be largely reproduced using systematic data from a large sample of developed market economies and that these results are relatively stable over time. Second, the results indicate that the small, open economies of Northern Europe, such as Denmark, Sweden, and Finland, share a coherent pattern of economic organization—a business system—that is not yet included in the business systems typology. Finally, the results of the fuzzy-set analysis illustrate that several institutional configurations are conducive to innovation specialization in high-tech industries and that the characteristics of the configurations with the widest coverage closely correspond to two of the business system types identified in the cluster analyses.
These findings make several contributions to comparative institutional analysis. First, although the results of the cluster analyses provide empirical support for the business systems typology, they also suggest that this typology should be extended. Second, the outcomes of the fuzzy-set QCA analysis contradict suggestions made by the Varieties of Capitalism approach (Hall & Soskice, 2001) and highlight equifinality in high-tech innovation specialization. This complements related analyses (e.g., Boyer, 2004; Schneider & Paunescu, 2012), and may help explain the inconsistent findings regarding links between societal institutions and technological and economic outcomes (e.g., Akkermans et al., 2009; Schneider & Paunescu, 2012). Third, the results of the fuzzy-set QCA analysis illustrate that the national business systems approach complements related approaches to the comparative analysis of capitalism in explaining such complex relations.
In the remainder of this paper, I first provide an overview of the business systems approach and the business systems typology. I subsequently discuss how the strong reliance of business systems scholars on in-depth (but difficult-to-compare) country studies leads to several difficulties, which can be addressed through more systematic taxonomical approaches. I then describe the selection of variables, data sources, and methods, and present the findings of the cluster analyses and the outcomes of the fuzzy-set QCA analysis. In the concluding sections, I reflect on the findings, and discuss their implications for the business systems literature and comparative analyses of capitalism in general.
The National Business Systems Framework as a Typology
The starting point of the business systems approach is that business systems, which are defined as the dominant patterns of economic organization and control (Whitley, 1999), can be characterized and compared on the basis of three key dimensions: the dominant type of ownership coordination, the type of non-ownership coordination, and the type of employment relations. Taken together, these dimensions are reflected in the dominance of certain types of firms that have particular strategies and relations with other economic actors (Whitley, 1999, 2000).
A key notion of the business systems approach is that the dominant mode of economic organization is often internally consistent. That is, the approach recognizes that the characteristics of local patterns of economic coordination and control are often complementary and interdependent. For instance, both high direct ownership and high market control tend to limit cooperation between firms, such as alliances, within and across industries. Similarly, economies with predominantly market-based ownership relations are less likely to support long-term risk-sharing between employers and employees than economies with high levels of inter-firm collaboration. Such interdependencies suggest that the number of viable combinations of business system characteristics is relatively limited, and that established patterns of economic coordination and control are likely to persist over time (Whitley, 1994, 1999).
A second notion of the business systems approach is that the dominant pattern of economic coordination and control is closely linked to the nature and type of societal institutions. As solutions for resolving economic coordination problems, societal institutions help generate and reproduce established coordination and resource-allocation patterns. In this regard, the business systems approach highlights four institutional dimensions: the role assumed by the state, the characteristics of the financial system, the skill-development system, and the norms and values that resonate in work relations (Whitley, 1998, 1999). These institutions are important because they encourage particular types of ownership and economic coordination, as well as particular types of economic actors (Whitley, 1999). In addition, they control access to labor and capital, and govern the nature of these resources, as in the case of the financial and education system.
The business systems framework
The interdependence between the characteristics of business systems and complementary dominant institutions suggests that the actual number of distinct and stable ways of organizing economic activity is relatively limited. Based on this assumption, Whitley (1999, 2000) proposes a conceptual typology of six distinct business system types, each of which is associated with particular institutional configurations and particular types of dominant firms. Fragmented business systems are characterized by high direct control, and by low ownership and alliance coordination (Whitley, 1999). Cooperation among firms is limited and markets are highly competitive. These business systems are often found in environments where trust is low and where financial resources are not readily available. As a result, firms tend to be small and they are typically opportunistic when pursuing new market opportunities (Whitley, 1994). In contrast, coordinated industrial district business systems are dominated by smaller firms but alliance integration is more extensive. Firms tend to be more artisanal and more readily exchange resources and opportunities. In addition, local government agencies often help increase entry and exit barriers and provide a high-quality training system, as is evident in many of Italy’s industrial districts (Best, 1990; Whitley, 1994).
Compartmentalized business systems are characterized by high ownership coordination and low non-ownership coordination, and the firms within them often have dispersed ownership structures. Such business systems develop when strong formal institutions enable extensive market contracting but arms-length state involvement does not encourage long-term commitment among economic actors, as in many Anglo-Saxon economies (e.g., O’Sullivan, 2000; Richardson, 1972; Whitley, 1994). This encourages competition among firms, which tend to be larger and organizationally integrated (Whitley, 2000). In contrast, in state-organized business systems, high ownership coordination is coupled with direct owner control. Such systems have previously been seen in France and South Korea (Hart, 1992; Kim, 1997). In such business systems, states take a more active “dirigiste” role in organizing economic development and are less accepting of intermediary associations. This encourages direct ownership with closer ties among political and economic elites, and limits non-ownership coordination (Whitley, 2000).
The two remaining business system types—collaborative and highly coordinated business systems—are characterized by strong interconnections and risk-sharing among economic actors (Whitley, 2000). Firms in these business systems develop more durable relations with customers, suppliers, and technology providers (Whitley, 2007). Similarly, a stronger reliance on credit encourages closer relations with capital providers. In both systems, the state encourages the establishment of intermediary associations and employer–employee relations are more collaborative, as illustrated by the German system of co-determination (Hancké & Goyer, 2005). In collaborative business systems, however, the integration and coordination of economic activity mostly occur within sectors, a characteristic that is reinforced by the level of wage setting (e.g., Mellahi & Wood, 2004; Swank & Martin, 2001). In contrast, in highly coordinated business systems, collaboration extends across industries and often takes the form of close business networks or alliances, such as those traditionally found in Japan (Gerlach, 1992; Whitley, 2000). In addition, whereas states encourage and facilitate collaboration among economic actors in collaborative business systems, and even delegate considerable social and economic decision making to them or to intermediary associations (Whitley, 2000), in highly coordinated business systems the state takes a more active guiding role, although less so than in state-organized business systems.
From a Typology to a Taxonomy of Business Systems
The discussion above illustrates that the business systems typology offers considerable benefits as a conceptual framework. The constructed business system types provide insight into the specific logics of distinct economic coordination and control patterns, while the typology as a whole provides insight into the more general relations among institutions and economic activity. The business systems typology is also an important classification tool. As with good typologies in general (Bailey, 1994; Hambrick, 1984), it reduces the complexity of political economies to a clear set of analytical dimensions, and offers a yardstick against which individual cases can be studied and compared.
Nonetheless, the relevance of the business systems framework depends on the extent to which the constructed types continue to reflect the dominant logics behind various forms of economic organization. This congruence is hard to assess through in-depth country studies alone. In-depth country studies are excellent tools for assessing how countries compare to established types or for exploring mechanisms of change (see e.g., Kristensen, 2006; Moen & Lilja, 2005; Whitley, 1999). However, as no single case should be expected to fit perfectly with any of these conceptual types (Whitley, 2006), such country studies are less useful for systematically assessing commonalities and universal relations across cases. In other words, reliance on in-depth studies alone leaves certain key questions concerning the business systems typology unanswered.
For instance, the lack of more systematic cross-country comparisons makes it hard to assess whether the business systems typology is exhaustive. The business system types are intended to be used as tools that aid in describing and explaining variation among countries (Morgan, 2007; Morgan, Whitley, & Moen, 2005). However, as the typology is derived inductively, it is hard to assess whether it captures the full variety of economic organization found in developed market economies. Without the systematic analysis of a wider range of market economies, we might overlook more subtle but equally coherent and distinctive ways of economic organization that could enrich the business systems toolbox.
A related issue is whether the individual business system types represent useful classification types, in the sense that they reflect more universal systemic logics and interrelations that can be found across developed market economies (Whitley, 1999). Although the business system types are abstract classifications rather than reflections of particular contexts (Morgan, 2007; Morgan et al., 2005), they are logically derived from studies of a relatively small number of Western and Asian countries (e.g., Hamilton & Biggart, 1988; Kagono, Alonaka, Sakakibara, & Okumara, 1985; Maurice, Sellier, & Silvestre, 1986; Maurice, Sorge, & Warner, 1980; Whitley, 1990). This raises the issue of whether the business system types truly reflect universal patterns of economic organization or merely the characteristics of representative cases, such as Germany, France, South Korea, and Japan.
Finally, the absence of large-scale systematic analyses also makes it difficult to characterize and classify countries that are outside the scope of the traditional focal countries. This means that although the typology of business systems provides useful insights into different types of economic organization, it offers no clear empirical indication of the countries to which these insights can be applied, other than the countries on which its systematic synthesis of country studies is based. As a result, the characterization of the business systems of other countries or regions is often ambiguous and, at worst, based on little more than intuition.
Taxonomies of capitalism
Such concerns are, however, certainly not unique to the business systems approach—they also apply to other classifications of capitalism. For instance, the exhaustiveness of the dichotomous Varieties of Capitalism typology and the homogeneity of its category types are frequently called into question: not only is it difficult to classify all OECD countries along the distinction between coordinated and liberal market economies (Hall & Gingerich, 2009; Hall & Soskice, 2001, p. 21), but the heterogeneity in the group of coordinated market economies also suggests a need for further conceptual differentiation (e.g., Amable, 2003; Schneider & Paunescu, 2012).
There are at least three ways around such typological limitations. The first is to accept a typology’s limitations but recognize its analytical contribution. Different typologies highlight different aspects of capitalist systems (Jackson & Deeg, 2008) and exhaustiveness is not necessarily a typology’s main purpose. Its analytical value may, first and foremost, lie in highlighting and exploring one or more important analytical dimensions, such as the importance of coordination for innovation patterns and macroeconomic performance in the Varieties of Capitalism approach. In this case, the typology serves the analytical purposes of reducing complexity and facilitating the exploration of key dimensions and main relations, without seeking to account for all variety in capitalism.
The second alternative is to inductively reconsider or further refine the dimensions of the typology and/or to adjust the number of criterion types. This might be necessary if a typology fails to account for all cases (see e.g., Schmidt, 2003) or if considerable variation remains between the cases assigned to a particular type. For instance, motivated by the inability of the Varieties of Capitalism typology to account for countries such as France and Spain, Hancké, Rhodes, and Thatcher (2007) propose the role of the state as an additional classification dimension, thereby extending the typology from two types to four. Similarly, the number of types included in the business systems typology has gradually increased from five (Whitley, 1994) to six (Whitley, 1999, 2000) to eight (Whitley, 2007).
The third alternative is to complement typologies and inductively derived classification types with the construction of taxonomies. Whereas typologies are logically derived conceptual classification schemes, which may or may not build inductively on characteristic cases, taxonomies are classifications of empirical cases (Bailey, 1994) that are often numerically derived. Although not all logically derived types may be observed at all points in time, taxonomies are useful tools for exploring and assessing the extent to which existing types can be empirically identified. This may lead to the identification of new types or stimulate the kind of conceptual refinement identified above.
In recent years, such systematic attempts to empirically assess typologies of political economy have been on the rise. For instance, Hall and Gingerich (2009) draw on data from 1990 to 1995 to explore some of the key propositions of the Varieties of Capitalism approach. In line with the predictions of that approach, their results reveal systematic variation in labor relations and corporate-governance practices between coordinated and liberal market economies. In a related but more elaborate analysis of the Varieties of Capitalism approach, Amable (2003) argues that a focus on a restricted number of analytical distinctions, such as coordination or the presence of a welfare state, may lead researchers to overlook a considerable part of the variety among capitalist countries. Based on five institutional dimensions deduced from the political economy literature, Amable (2003) constructs five models of capitalism and tests these models drawing on data from the late 1990s. Principal component and cluster analyses reveal five separate clusters, which Amable labels as market-based capitalism, Asian capitalism, Continental European capitalism, social-democratic capitalism, and Mediterranean capitalism.
Such examples and others (e.g., Allen, Funk, & Tüselmann, 2006; Boyer, 2004; Geffen & Kenyon, 2006; Kenworthy, 2006; Kogut & Ragin, 2006) illustrate that although taxonomies naturally lack the analytical gist of typologies, they may usefully complement qualitative and theoretical work. The fact that such assessments are nonetheless still relatively uncommon can be ascribed to the difficulties of obtaining good indicators, on which the success of the resulting taxonomy is ultimately dependent. For instance, in their quantitative assessment of the varieties of capitalism, Hall and Gingerich (2009) highlight the difficulty of finding cross-country indicators for the degree of coordination, which is the key dimension separating liberal from coordinated market economies. They therefore limit their analysis to one sphere (wage bargaining), and complement their measures of coordination in wage bargaining with indicators on supportive institutional characteristics.
As indicated earlier, I aim to submit the national business systems typology to a similar test. Nonetheless, the large number of clustering dimensions makes assessing the national business systems typology particularly challenging. To overcome some of the difficulties of obtaining useful cross-national data on ownership and coordination patterns, I follow Amable (2003), Hall and Gingerich (2009), and Schneider and Paunescu (2012) in that I focus on the configuration of the national institutional context. While this makes it difficult to capture the diversity of business systems within nations, the continued relevance of nation states and national institutions for regional, national, and transnational economic coordination partially legitimizes such a limitation (Allen & Whitley, 2012; Morgan, 2007; Whitley, 2005).
Methods
To assess the validity of the national business systems typology and explore its explanatory value, I follow a two-step approach and combine two complementary configurational methods. I first construct a taxonomy of national business system types through a two-stage clustering analysis of data from 2000 and 2011. This allows for a comparison with the original business systems framework (Whitley, 1999) and an assessment of the robustness of that framework over time. I subsequently assess the explanatory power of the identified business system types using a fuzzy-set analysis of countries’ comparative institutional advantages in high-tech innovation.
Sample and variable selection
Whitley’s (1999) business system typology generally applies to developed market economies with stable societal institutions. Therefore, I use the 30 OECD member states in 2000 as my sample. This sample includes countries that are frequently covered in the literature, such as Germany, the UK, Denmark, Japan, and South Korea. The sample size is slightly larger than those of related analyses, such as Amable (2003) and Hall and Gingerich (2009), who use data from samples of 21 and 20 OECD countries, respectively.
I combine institutional indicators from several data sources (Table 1). The main data source is the Global Competitiveness Report (World Economic Forum, 2000, 2011), a widely used source of institutional data (e.g., Gaur & Lu, 2007; Rao, Pearce, & Xin, 2005). These data are complemented with financial data and labor statistics obtained from the World Bank and the ICTWSS database. I select 2000 and 2011 as my focal years, and use the most recent data available before 2011 if data for 2011 are not yet available. For interpretive purposes, the scores for state dominance, the burden of regulation, the centralization of bargaining, and paternalism are inverted.
Institutional indicators.
The state
Although states impact economic activity in various ways, three characteristics of the role of the state are particularly important in promoting and sustaining different forms of economic organization. These are the dominance of the state, the state’s tolerance of intermediate associations between the state and firms, and the extent of state involvement in market regulation (Whitley, 1999).
The dominance of the state refers to the strength of the state in relation to special-interest groups, such as social elites, and to the role of the state in the development of economic activity. The strength of the state in relation to special-interest groups is captured with the independence of government policies from special interest groups indicator found in the Global Competitiveness Report (World Economic Forum, 2000, 2011). The role of the state in the development of economic activity is measured using the indicator of the extent to which government subsidies promote fair competition. As this indicator is not included in the 2011–12 edition of the Global Competitiveness Report, I use the effectiveness of anti-monopoly policy indicator from that edition (World Economic Forum, 2011). The assumption is that states that assume a more developmental role are more selective in granting subsidies and pursue less-effective competition policies (Singh, 2002; Westphal, 1990). On the other hand, states that adopt an arm’s-length approach are assumed to have a greater interest in pursuing policies that promote fair competition. The scores for the strength and role of the state are combined into a single measure of state dominance.
The degree of state antagonism to intermediary associations is captured using the pervasiveness of clusters and intermediary associations indicator from the Global Competitiveness Report. I use the burden of regulation that firms experience as the indicator of the degree of formal regulation of markets. The assumption is that the type of regulatory role assumed by the state is reflected in the burden of regulation experienced by firms.
The financial system
Financial systems can differ considerably in terms of how capital is raised. Generally, a distinction is made between financial systems that rely on external capital markets, and financial systems that are based on credit and in which borrowers and lenders are more interlocked. However, it is important to recognize that firms’ dependence on bank loans and the use of the stock market are two distinct characteristics of financial systems (Bencivenga, Smith, & Starr, 1996; Greenwood & Smith, 1997; Monnet & Quintin, 2007). Therefore, the measurement of stock-market capitalization or bank lending alone may fail to reflect the relative importance of either of these sources of finance. To capture the relative importance of capital versus credit, I therefore divide the ratio of market capitalization to GDP by the ratio of private credit extended by deposit money banks to GDP. Scores above 1 reflect financial systems that are more based on the capital market, while scores below 1 reflect financial systems that are more based on credit. The data used for this measure come from the World Bank (see Table 1) and cover 2000 and 2009, with the latter being the most recent year for which data were available at the time of analysis.
The skill-development and control system
The skill-development and control system encompasses the strength of the education system, and the role and organization of trade unions (Whitley, 1999). Strong education systems ensure an equitable distribution of learning opportunities (OECD, 2010). In such systems, student performance is less affected by variation among schools than in less-effective education systems. As a measure of education system strength, I therefore use the intra-class correlation coefficient from the OECD PISA database (2000, 2010). The intra-class correlation coefficient reflects the percentage of total variance in student performance that is accounted for by differences among schools. The data are from 2000 and 2009, with 2009 being the most recent year for which data were available.
To capture the strength of independent trade unions, I follow previous work and use the union-density scale included in the ICTWSS database (see Visser, 2006). The ICTWSS union-density scale measures net union membership as a proportion of employed wage and salary earners. I use the ICTWSS Database 3, which contains data through 2010.
I measure the centralization of bargaining using the wage-setting indicator found in the Global Competitiveness Report. This indicator reflects the extent to which wages are set by individual companies rather than through a more centralized bargaining process. I selected this indicator because it offers greater coverage than the indicator on the centralization of bargaining from the ICTWSS database.
Trust and authority relations
Generalized trust is a relatively homogeneous construct. Therefore, as a proxy for trust in formal institutions, I use the indicator for public trust in politicians from the Global Competitiveness Report. The assumption is that the extent to which formal institutions are able to generate and guarantee trust is associated with the trust people have in the politicians who shape and govern those institutions. Although other factors also affect political trust (Cole, 1973), the association between political trust and general and social trust has considerable empirical support (Brehm & Rahn, 1997; Jackman & Miller, 1998; Levi & Stoker, 2000).
Whitley (1999) characterizes differences in employer–employee relations according to the distinction between paternalistic and formal cultures. Differences between the two essentially pertain to: i) the scope of superior discretion, which is limited to the work context in formal cultures but extends into the personal sphere in paternalistic cultures, and ii) whether subordinates are entrusted with decision-making responsibilities. Two indicators from the Global Competitiveness Report are used: the willingness to delegate authority to subordinates and the extent to which management–worker relations are cooperative. The assumption is that the willingness to delegate is lower and that management–worker relations are less cooperative in more paternalist cultures. In more formal political cultures, trust in subordinates is presumed to be higher, resulting in a greater willingness to delegate tasks and more cooperative superior–subordinate relations. The scores for these indicators are highly related, with Cronbach’s alphas of 0.75 and 0.91 for 2000 and 2011 respectively. I therefore combine them into a single measure of the degree of paternalism, with scores ranging from 1.8 (for Denmark) to 4.8 (for Turkey).
Step 1: Cluster analysis
In the first phase of the analysis, I apply a cluster analysis to construct two empirical taxonomies of business systems: one for 2000 and one for 2011. Cluster analysis is a statistical method used for the classification of empirical data on the basis of pre-defined cluster variables. Clusters are defined such that the within-group variance is minimized and between-group variance is maximized. As this type of analysis relies extensively on a researcher’s judgment, the validity of obtained solutions is susceptible to critique (Ketchen & Shook, 1996). Therefore, to obtain the best results, I follow the recommendations of Punj and Stewart (1983) and Ketchen and Shook (1996) in that I conduct a two-stage clustering procedure, which increases the validity of the cluster solutions (Punj & Stewart, 1983).
In the first stage, I use a hierarchical clustering method to identify the appropriate number of clusters. As the institutional indicators have different scales, z-scores are calculated in order to standardize all variables. I then use SPSS 19 to conduct hierarchical cluster analyses for both 2000 and 2011, with squared Euclidean distance as the interval measure. I use Ward’s minimum variance method as the clustering method because the cluster sizes are expected to be approximately equal (Ketchen & Shook, 1996). In the second stage, I optimize the obtained clustering solution by using the outcomes of the hierarchical cluster analyses as initial cluster seeds in an iterative K-means cluster analysis.
Due to missing PISA intra-class correlation scores, Japan, the Netherlands, Slovakia, and Turkey are not included in the 2000 analysis. For the same reason, France is excluded from the 2011 analysis. In addition, missing data from the World Bank Financial Structure dataset led to the exclusion of Iceland and Norway from the 2011 sample. This gives a sample of 26 OECD countries for 2000 and a sample of 27 countries for 2011.
Step 2: A fuzzy-set illustration of the explanatory power of the taxonomy
While cluster analysis is a useful classification tool, the value of identified clusters is perhaps best illustrated through their effects on an outcome. In the second phase of the analysis, therefore, I assess the predictive power of the business systems framework for understanding innovation specialization in high-tech industries. Innovation specialization in high-tech industries has also been studied from a Varieties of Capitalism perspective (e.g., Akkermans et al., 2009; Hall & Soskice, 2001; Taylor, 2004). This literature can therefore serve as a useful yardstick against which the predictive power of the national business systems framework can be evaluated.
To capture innovation specialization, I construct a Balassa index of revealed innovation specialization (RIS) in biotechnology, nanotechnology, and information and communication technology using patent data from the OECD/EPO patent database. I select these sectors because they depend heavily on radical innovation, in which liberal or Anglo-Saxon regimes are often claimed to have an institutional advantage (Hall & Soskice, 2001; cf., for instance, Casper & Whitley, 2004). As I am interested in innovation activity rather than the value of innovations, I follow previous studies on innovation specialization in the use of patent counts rather than forward citations (Cantwell & Janne, 1999; Dunning, 1994; Hall & Soskice, 2001).
The RIS index as used here reflects the share of patenting of a country in a sector relative to its share of patenting in all industries, and is a frequent measure of innovation specialization (e.g., Dunning, 1994; Hall & Soskice, 2001; Laursen, 2000). The general form of the RIS index is as follows
The nominator reflects EPO patent applications (P) from country j in sector i as a proportion of total EPO patent applications from country j. The denominator reflects the total number of EPO patent applications in sector i as a share of the total number of patent applications from the sample of OECD countries. An RIS value of more than one signifies that a country is relatively specialized in innovation in sector i. For the purpose of this study, I calculated a combined RIS country score for the biotechnology, nanotechnology, and ICT sectors for both 2000 and 2008, the most recent year covered in the OECD patent database at the time of analysis.
I subsequently apply fuzzy-set QCA to analyze countries’ revealed innovative advantage in the three focal sectors. QCA is a set-theoretic approach that applies Boolean logic to identify necessary and sufficient conditions for an outcome (Ragin, 2000, 2008). Two advantages of set-theoretic approaches relative to conventional approaches are that the former allow for studying cases as configurations of conditions and that they allow for the analysis of causal complexity as opposed to studying the net effects of independent variables (Ragin, 2008). These factors make set-theoretic approaches, such as fuzzy-set QCA, highly suitable for analyzing the effects of institutional configurations and complementarities on economic and organizational outcomes (Kogut, 2010). While still a relatively new technique, QCA has therefore increasingly been used in analyses of institutional regimes and their outcomes (e.g., Boyer, 2004; Gjølberg, 2009; Jackson, 2005; Kogut & Ragin, 2006; Kvist, 2007; Pajunen, 2008; Schneider et al., 2010).
Fuzzy-set scores reflect varying degrees of set membership in a conceptual category. To transform variables into fuzzy-set membership scores, I first specified three qualitative anchors for each variable (Ragin, 2008): full membership, full non-membership, and the point of maximum ambiguity at which cases are neither fully in nor fully out of a given set. In line with recommendations (Ragin, 2000, 2008; Rihoux & Ragin, 2009; Schneider & Wagemann, 2010), the qualitative anchors were theoretically informed where possible. For instance, because a score of 1.0 for the financial systems measure indicates a financial system in which market capitalization equals bank lending, this point was assigned maximum ambiguity (a set membership score of 0.5). Where theoretical knowledge offered less guidance on cut-off points, such as with variables from the Global Competitiveness Report, I used substantive case knowledge to set anchors (see Schneider & Wagemann, 2010). For instance, the cut-off point for the centralization of bargaining in 2000, initially set at 3.95, was lowered to incorporate Iceland (Jonsson, 2001) but not Poland (Avdagic, 2005; Orenstein & Hale, 2001).
As cases with membership scores of exactly 0.5 are difficult to analyze (Ragin, 2009), I deducted 0.05 from the identified crossover points of variables on interval scales (cf. Fiss, 2011). For instance, crossover points set at 4.0 were redefined as 3.95. I calibrated the 2000 and 2011 variables separately because the scales in the Global Competitiveness Report are often equivalent but not identical across editions. I then assigned set membership scores with the calibration function in the fs/QCA 2.5 software, which applies the log odds method described in Ragin (2008). This produces a fine-grained calibration of the degree of set membership of cases, with scores ranging from 0.00 to 1.00. I used set intersection (logical AND) to construct set scores for the compound variables of state dominance and paternalism. The anchor points used for calibrating set membership scores are reported in Appendix I.
Results
Outcomes of the cluster analyses
The hierarchical cluster analyses yielded four main clusters for both the 2000 and 2011 data. A K-means analysis was then performed to optimize the results. After varying the number of clusters, Italy emerged as a separate, fifth cluster in 2011 but not in 2000, partially due to the outlier values for Finland. The final cluster centroids and country groupings are displayed in Tables 2 and 3.
Final cluster centers.
Cluster membership.
Ranked according to distance to cluster centre.
Note. Countries that changed cluster membership are in
As Table 3 illustrates, the cluster results are relatively consistent across years. In 2000, cluster 1 contains the Anglo-Saxon economies, Switzerland, Luxembourg, and Spain. The same generally holds true for 2011, although Japan replaces Spain and Ireland is no longer included. Cluster 2 contains Germany, Austria, and Belgium. Italy is included in 2000 but not in 2011, where it forms a separate cluster. Ireland and the Netherlands, which was left out of the 2000 analysis due to missing data, are also part of cluster 2 in 2011. Cluster 3 includes the Mediterranean countries, the Central East European states, Mexico, France, and South Korea. In both years, cluster 4 includes the Nordic economies, while Italy forms a fifth cluster in 2011.
I conducted several robustness tests to assess the sensitivity of the clusters to the selected variables. The removal of single variables revealed that the hierarchical cluster outcomes for 2000 were potentially sensitive to the selected indicator for education system strength. I therefore re-ran the hierarchical cluster analyses for both years with the variance in average student performance that is explained by students’ socio-economic backgrounds as an alternative PISA measure of education system strength. The assumption is that strong education systems limit the effect of differences in socio-economic background on student performance. Finland and Austria are examples of economies that score positively on this measure (with scores of 1.7 and 2.5, respectively), whereas the scores of countries such as the United States (17.4) and the United Kingdom (15.0) suggest that their education systems are characterized by greater differences in the quality of education available to rich and poor pupils. The hierarchical cluster solutions with this alternative measure indicated four dominant clusters, which is consistent with the original results.
A second issue that required attention was the potential impact of the global financial crisis on the financial systems measure. The World Bank data showed a strong decline starting in 2008 in the market capitalization to GDP ratio, which was the numerator of the financial systems measure. To assess the sensitivity of the cluster outcomes to the decline in market capitalization, I re-ran the 2011 hierarchical cluster analysis with a financial systems measure based on the World Bank’s market capitalization values reported for 2007, before the onset of the global financial crisis. The hierarchical cluster results again indicated four clusters, which was consistent with the original results for 2011.
Third, the supplementary analyses showed that Finland (in 2000) and Italy (in both years) were sensitive to variation in the indicators. In Finland’s case, this was partially due to notably high scores for the market capitalization to credit ratio in 2000 (4.7 versus an average of 1.2). Finland’s score was influenced by the high market capitalization of Nokia at the time, which alone constituted up to 70% of the HEX index (Rantapuska, 2008). Nonetheless, the supplementary cluster analyses in which individual variables were removed from the analysis to assess the sensitivity of the cluster outcomes did not suggest a change in cluster membership for Finland. In contrast, Italy’s cluster membership shifted. A re-analysis using the single-linkage cluster method confirmed Italy as a separate cluster within the 2011 sample but not in the 2000 sample. As Figure 1 illustrates, Italy’s defining features include high combined scores for the burden of regulation and the prevalence of clusters.
Z-scores of the burden of regulation and the prevalence of clusters in 2011.
I subsequently re-labeled the final cluster scores to enable comparison with the institutional characterization of Whitley’s (1999) business system types. Cluster scores above 0.5 were labeled “high.” Scores below -0.5 were labeled “low.” Scores between -0.5 and -0.1, -0.1 and 0.1, and between 0.1 and 0.5 were labeled “limited,” “some,” and “considerable” respectively. The labels for the strength of the education system were reversed. I then matched the identified clusters with the business system types from Whitley’s (1999) comparative business systems framework. The results are presented in Table 4, in which the institutional characteristics from the national business systems typology are presented in bold.
Matching clusters and business system types.
As Table 4 illustrates, the results show consistent matches between the cluster characteristics and the institutional characterizations of Whitley’s business system types in both years (Whitley, 1999). For both 2000 and 2011, the institutional characteristics of cluster 1 closely correspond with the characterization of the compartmentalized business system type. Similarly, the features of cluster 2, which contains Germany and Austria, resemble the characteristics of the collaborative business system. Cluster 3, which includes France and South Korea, best resembles the features of the state-organized business system. The characteristics of cluster 5, which consists of Italy, best match the coordinated industrial district business system. Cluster 4, which includes the Nordic countries, is the only cluster that does not closely resemble the institutional characteristics of any existing type of business system. For both 2000 and 2011, cluster 4 combines liberal features, such as low state dominance and low market regulation, with high trust, high union density, and a centralized wage-bargaining system. As expected, given the focus on developed market economies, none of the clusters resembles the fragmented business system.
The results of the fuzzy-set QCA analysis of innovation specialization
The first step in the fuzzy-set QCA analysis was to test for necessary conditions for innovation specialization in biotechnology, nanotechnology, and ICT. Necessary conditions are conditions that are required but not necessarily sufficient for an outcome to occur. None of the individual conditions exceeded the consistency threshold of 0.80 (Appendix III). I then applied the truth-table algorithm (Ragin, 2008) to identify causal combinations of conditions that are sufficient for innovation specialization in the focal sectors. In line with recommendations (Ragin, 2006; Rihoux & Ragin, 2009), the minimum consistency threshold was again set at 0.80 (see Appendix II). The frequency threshold was set at 1 and the selection of prime implicants was guided by the directional expectation that a high regulatory burden impedes market coordination, which in turn hampers radical innovation (Hall & Soskice, 2001). Furthermore, as a result of the onset of the financial crisis, stock market data for 2009 show a significant decline relative to previous years. In order to avoid underestimating the effects of financial systems on innovation activity, I used financial data for 2007.
Table 5 presents the outcomes of the fuzzy-set analysis of sufficient conditions for high-tech innovation specialization, which follows the notation of Ragin and Fiss (2008). The columns represent the configurations of conditions that are associated with innovation specialization in the three high-tech industries and that exceed the consistency threshold of 0.80. These configurations consistently display high correspondence with the outcome. The black circles signal the presence of a condition, while the white circles signal the absence of a condition. In addition, large circles signal core conditions that are part of both the parsimonious and intermediate solutions. Small circles signal conditions that only feature in the intermediate solutions. Relative to parsimonious solutions, intermediate solutions permit fewer simplifying inferences about the absence of causal conditions. Intermediate solutions are therefore more conservative and more complex (see e.g., Ragin, 2008, pp. 160–175).
Configurations for innovation specialization in biotechnology, nanotechnology, and ICT.
Notes. ● = core causal condition present; ⊗ = core causal condition absent; • = complementary causal condition present; ○ = complementary causal condition absent.
Cases not covered by the solution terms: Canada 2000 and 2011; Iceland 2000; Japan 2011; Korea 2000 and 2011; the Netherlands 2011.
Table 5 illustrates that several distinct combinations of conditions, or causal paths, are associated with innovation specialization in the three high-tech sectors. A comparison of the configurations highlights three sets of core conditions. First, configurations 1a, 1b and 1c, which reflect the United States, Australia, and the United Kingdom (in 2000), respectively, share several core conditions: the presence of a capital-market-based financial system, and the absence of burdensome regulations and high trust relations. These configurations also share several complementary causal conditions: the absence of high union density, the absence of centralized wage bargaining, and the absence of paternalism. Their empirical relevance and relatedness are reflected in relatively high raw coverage scores and low unique coverage scores. The characteristics of configurations 1a, 1b, and 1c also have striking similarities with cluster 1, which is the liberal or compartmentalized cluster in the cluster analysis (see Table 4).
The core conditions of configuration 2 combine the absence of burdensome regulations and a dominant state with the prevalence of clusters, high union density, and centralized wage bargaining. High trust and a strong education system are complementary causal conditions for this configuration. The raw and unique coverage scores for configuration 2 demonstrate its unique explanatory power relative to configurations 1a, 1b, and 1c. The scores highlight that configuration 2 reflects a causal path to high-tech innovation specialization that is not captured by the first set of configurations. In terms of its characteristics, configuration 2 closely resembles cluster 4 of the cluster analysis, which is the newly identified business system type covering the Nordic countries.
Configuration 3, which reflects Ireland, combines the absence of burdensome regulations and high trust with the presence of a strong education system and centralized wage bargaining. Peripheral conditions are the presence of strong unions, the absence of a capital-market-based financial system, and the absence of both a dominant state and high paternalism. Configuration 3 displays several similarities to the other configurations. For instance, with configurations 1a, 1b, and 1c, it shares the absence of burdensome regulations and high trust as core conditions. With configuration 2, it shares the presence of strong unions and centralized wage bargaining. Nonetheless, the coverage scores highlight that this configuration constitutes a unique causal path to innovation specialization in the high-tech sectors.
Overall, the solution terms demonstrate considerable consistency with the outcome and uniquely cover 12 cases. Seven cases with innovation specialization in the focal sectors are not covered by the solution terms (Table 5). Canada and Japan (in 2011) are noteworthy because they are included in cluster 1 but are not covered by configurations 1a, 1b, and 1c. An inspection of the data matrix shows that, for Canada, this is due to the presence of a dominant state and high trust (in 2000), and the absence of a capital-market-based financial system (in 2011). Japan (in 2011) is not covered due to the absence of a capital-market-based financial system and its weaker education system.
To assess the impact of limited diversity on the derived solution terms, I constructed two new compound conditions (“state” and “union”) by intersecting the state-related and union-related conditions. I conducted supplementary analyses with eight and seven conditions. As the intersection of conditions creates more narrowly defined sets, this produced solutions with slightly lower coverage. Nevertheless, the solutions were consistent with the original results and produced three solution terms.
Discussion
The results of the cluster analyses support the validity of several of the business system types and illustrate that the business systems typology is relatively robust over time. However, the results also suggest that the typology needs to be extended with a new business system type, that of the open economies in northern Europe. In addition, the results of the fuzzy-set QCA analysis illustrate not only equifinality in innovation specialization in high-tech sectors but also the power of the validated and extended business systems typology in explaining this pattern. These findings are discussed in more detail in this section.
The empirical validation of several business system types
The cluster outcomes provide considerable support for the business systems typology. As Table 4 illustrates, cluster 1 shares several key features with the compartmentalized business system type. The cluster combines the absence of direct state involvement and limited market regulation with a capital-market-based financial system and limited union involvement. This contrasts with cluster 2, in which direct state involvement and market regulation are more prominent, and in which the financial system is credit-based. In addition, cluster 2 is characterized by the incorporation of intermediaries, higher union power, and centralized wage bargaining. Together, these features correspond most closely to the collaborative business system type. Cluster 3 closely reflects the state-organized business system type. It combines relatively high state coordination with limited intermediary associations and union power. Trust in formal institutions is relatively low, and cluster 3 scores characteristically high on paternalism (Whitley, 1990, 1999). Finally, cluster 5 shares many of the features of the coordinated industrial districts business system type, including considerable state involvement and paternalistic authority relations, coupled with high cluster formation and considerable union strength.
While there are many commonalities between the cluster outcomes and Whitley’s (1999) business system types, there are also several differences. For instance, when compared with the compartmentalized ideal type, cluster 1 scores higher on both the strength of the education system and the prevalence of clusters. Cluster 2 displays lower state involvement and less-extensive market regulation than the collaborative business system type would suggest, and scores considerably lower on the strength of the education system. When compared with the state-organized business system type, cluster 3 scores differently on the strength of the education system, albeit considerably higher than would be expected on the basis of the typology. Cluster 5 differs most notably in terms of the level of wage bargaining. Contrary to the characterization of the coordinated industrial districts type, Italy, the sole representative of cluster 5, scores high on this feature. This reflects the trend towards a centralized wage-bargaining system evident in Italy since the 1990s (Pérez, 2000; Thelen, 2001).
A similar picture emerges when we take cluster membership into consideration (Table 3). While there are exceptions, the composition of the clusters generally corresponds to the characterizations in the business systems literature. Cluster 1, which corresponds to the compartmentalized business system, contains most of the Anglo-Saxon economies (Whitley, 1999, 2000), as well as Luxembourg and Switzerland. The inclusion of Germany and Austria in cluster 2, which best corresponds to the collaborative business system type, is also according to expectations (Almond, Edwards, & Clark, 2003; Whitley, 2000). Cluster 3 includes both France and South Korea, which are often used as examples of state-organized economies in the literature (Schmidt, 2003; Whitley, 2000; Yoo & Lee, 2009). Italy, which makes up cluster 5, is a typical example of an economy with an industrial district-based business system (Trigilia, 1990; Whitley, 1998).
There are also some apparent surprises. Spain’s inclusion in the group of liberal countries in 2000 might reflect the country’s extensive economic restructuring and liberalization processes in the 1990s (Guillén, 2001; Molina & Rhodes, 2007). The robustness analyses indicate that the inclusion of Italy in cluster 2 in 2000 is influenced by Finland’s high score on the financial systems indicator, which masks the distinctiveness of Italy as a separate cluster. The 2011 inclusion of Japan in cluster 1 is noteworthy because Japan has often been characterized as a market economy with a highly coordinated business system (Saka, 2004; Whitley, 1994). Japan’s inclusion in this cluster might reflect the changes that occurred in several segments of the Japanese business system after the late 1990s. These included changes in the financial markets and increased labor mobility (Aoki, Jackson, & Miyajima, 2007; Morgan & Kubo, 2005).
Overall, four of the clusters closely correspond to Whitley’s (1999) business system types. Some differences exist between the institutional characterization of the clusters and those of the business system types. However, it must be kept in mind that Whitley characterizes constructed types, whereas the outcomes of the cluster analyses are the central tendencies of empirically identified polythetic clusters. Therefore, some variation is to be expected. Given the relatively broad sample used here, which covers four continents, the outcomes of the two-step cluster analyses and the supplementary robustness tests lend considerable support to the business systems typology.
The robustness of the business system typology over time
A second finding is that the business system framework appears relatively robust over time. The raw data indicate that there have been some changes in, for instance, the degree of market regulation. Nonetheless, the cluster outcomes in Table 4, which are based on standardized scores, suggest that the relative changes in institutional characteristics have been limited. The most substantive changes have occurred in collaborative cluster 2, which includes Germany and Austria, where the results suggest a decline in direct state involvement and market regulation. Nonetheless, other features have remained relatively constant in this cluster, such as the credit-based nature of the financial system. 1
Table 3 also illustrates that, apart from Italy, only two countries have changed cluster membership. In 2011, Spain is included in cluster 3 with the other Southern European countries, such as Greece and Portugal. Also in 2011, Ireland moves to cluster 2, which is the group of countries with a collaborative business system (see e.g., Harcourt & Wood, 2003, for authors who anticipated this shift). Nevertheless, it is important to recognize that the analyses do not capture functional changes in societal institutions. The apparent continuity of the relative differences among the clusters may therefore conceal the fact that some institutions have taken on new functions or have come to serve different purposes (see e.g., Deeg & Jackson, 2007; Goyer, 2003; Hall & Thelen, 2009).
The identification of a Nordic business system
The small, open economies of Northern Europe emerge as a separate group in cluster 4. The results of the cluster analyses suggest that this group of countries represents a distinct, inclusive business system type characterized by low direct state involvement, high trust, a strong public training system, and considerable cooperation and coordination across actors. The clustering of these countries in a separate group is noteworthy, as it provides a counter-argument to suggestions that some of the smaller economies in Northern Europe may be examples of hybrid business systems (e.g., Campbell & Hall, 2006; Whitley, 2000) that combine characteristics of the compartmentalized and collaborative ideal types. While cluster 4 shares characteristics with both types, the results of the cluster analyses illustrate that the institutional structuring of the countries in this cluster is both distinctive and coherent, rather than a combination of the institutional features of ostensibly more distinctive national systems.
The characteristics and composition of cluster 4 are strikingly similar to earlier characterizations. Katzenstein (1985), for instance, sees informal coordination and negotiation among concentrated interest groups as one of the defining features of democratic corporatism, which characterizes the small states in Northern Europe in his analysis. 2 In Amable (2003), Denmark, Finland, and Sweden cluster together in the group of social-democratic countries, while in Boyer’s (2004) Boolean analysis, Denmark, Finland, and Sweden emerge in a group of knowledge-based economies. The identification of a Nordic business system also finds support in recent literature. For instance, Kristensen (2011) highlights the fact that work organization is strikingly similar across the Nordic countries, and that organizations in these countries typically rely on decentralized decision making, skilled labor, and relatively flexible labor markets (see also Campbell & Pedersen, 2007). In addition, most Nordic states actively facilitate the empowerment of and collective risk sharing among economic actors, which encourages considerable firm-level experimentation (Kristensen, 2011).
Equifinality in innovation specialization in high-tech industries
The explanatory value of the business system typology and of the extension proposed above is illustrated by the fuzzy-set QCA analysis of innovation specialization. First, the outcomes of the fuzzy-set analysis show that several institutional configurations are conducive to innovation specialization in high-tech industries. In addition to a liberal configuration, in which radical innovation is promoted through deregulated labor markets and capital-based financing, the results illustrate that high-tech innovation specialization is also facilitated by systems with little regulation that facilitate risk taking and the dispersion of knowledge through an equitable education system, cooperation, and collective risk sharing. This finding, which is similar in spirit to Boyer’s (2004) result, is important for comparative institutional analysis because it demonstrates equifinality in specialization outcomes. In other words, the outcome suggests that different institutional complementarities may produce similar comparative institutional advantages. This finding challenges expectations associated with the Varieties of Capitalism approach (Hall & Soskice, 2001).
The second finding of the fuzzy-set analysis is that the characteristics of the configurations with the greatest coverage closely correspond to the institutional characteristics of two of the business system types identified in the cluster analyses. The first set of configurations (1a, 1b, and 1c) closely resembles the institutional features of cluster 1, which reflects the compartmentalized business system type. In contrast, the second configuration leading to revealed innovation specialization in high-tech industries corresponds closely with the characteristics of the Nordic business system type that is identified in this paper. As Casper and Whitley (2004) illustrate for Sweden, and as Ornston (2012) shows for Finland, the collective risk sharing and collaboration in such business systems is likely to be particularly conducive to innovation in high-tech sectors that require considerable coordination across actors. Thus, the fuzzy-set analysis not only illustrates that a more nuanced approach is needed to capture and explain the links between national institutions and organizational and country-level outcomes (Casper, 2010; Schneider & Paunescu, 2012), but also that the national business systems approach may provide such a framework.
Conclusion
The use of taxonomic approaches to study the diversity of modern capitalism and institutional regimes has risen steadily in the past decade (e.g., Amable, 2003; Hall & Gingerich, 2009; Hicks & Kenworthy, 2003; Schneider & Paunescu, 2012). The purpose of this paper was to subject the national business systems typology to a similar test and to explore its value for comparative institutional analysis. The motivation for the study originated from the fact that the business systems typology is frequently characterized as an important analytical tool for comparative institutionalists (Morgan, 2007; Morgan et al., 2005). However, the lack of large-scale systematic comparisons implies that we know little about whether the business systems typology is valid or exhaustive, or about the range of economies to which its types can be applied. In this regard, the findings presented in this paper advance comparative institutional analysis in three ways, as described below.
The first contribution is that, to the best of my knowledge, this is the first study to provide quantitative support for the distinctiveness of several of the business system types that constitute the business systems framework (Whitley, 1992, 1999). The business systems framework emerged from detailed but relatively particularistic accounts and comparisons of Asian and European market economies (e.g., Hamilton & Biggart, 1988; Redding, 1990). This makes the business systems framework susceptible to the critique that its ideal types reflect the idiosyncrasies of the countries considered—such as Japan and Korea—rather than aggregated business system types, and raises questions about the exhaustiveness of the typology. The results of the cluster analyses, which draw on data from 2000 and 2011, illustrate that the business system types or, at least, their accompanying institutional configurations successfully characterize most of the developed-market economies, and that the typology is relatively robust to changes over time. The findings therefore provide considerable support for the business systems typology as a valuable tool for comparative institutional analysis.
The second contribution of this paper lies in the identification of a distinctive business system type that characterizes the small, open economies of Northern Europe. This group of countries displays relatively low direct state involvement, high levels of trust, the considerable incorporation of intermediaries, and strong education systems. While the institutional characteristics of these economies show similarities with several other business system types, such as the collaborative and compartmentalized ideal types, the results of the cluster analyses suggest that the variation in institutional features is systematic. This implies that the business systems framework may have to be complemented with an additional, inclusive business system type, in which the diverse interests of economic actors are reconciled through inclusion and coordinated on the basis of strong informal institutions.
The third contribution to comparative institutional analysis lies in the outcomes of the fuzzy-set QCA analysis of innovation specialization in high-tech industries. Contrary to expectations of the Varieties of Capitalism approach (Hall & Soskice, 2001), the results of the fuzzy-set analysis suggest that comparative institutional advantages in high-tech industries are not limited to liberal institutional frameworks but are also found in other institutional configurations. While this finding mirrors those of similar analyses (e.g., Amable, 2003; Boyer, 2004; Schneider et al., 2010), the results presented in this paper extend previous contributions in two respects.
First, the results illustrate the potential value of the national business system approach for understanding more complex links between societal institutions and technological and economic outcomes. As their analyses frequently produce hybrid clusters that are difficult to interpret from a Varieties of Capitalism perspective, configurational scholars often call for more nuanced institutional approaches to investigations of the link between institutions and outcomes (e.g., Schneider & Paunescu, 2012; Schneider et al., 2010). In contrast, in this paper, the two most relevant configurations of the fuzzy-set QCA analysis closely correspond to two of the business system types uncovered in the cluster analysis. This might signal that the greater number of institutional dimensions that the business systems framework identifies may be helpful in analyzing and interpreting causal complexity in the link between societal institutions and their technological and economic outcomes.
Second, the results of the fuzzy-set analysis shed additional light on the institutional complementarities that constitute some economies’ relative success in high-tech innovation. Specifically, the results suggest that in addition to their financial and education systems (e.g., Schneider et al., 2010), the Nordic countries may have a comparative advantage in the inclusive nature of their business system. That is, the Nordic countries may succeed in high-tech innovation not only due to their ability to distribute knowledge through education, but also due to the combination of high trust, supportive governments, and collective risk sharing. Such conditions might be particularly conducive to radical innovation in sub-sectors that rely on collaboration and coordination across actors with regard to such elements as technical specifications, standards, or designs (Casper & Whitley, 2004).
An important limitation of the study presented here is that it relies on indicators of national institutions rather than more direct measures of business systems. These indicators were used due to the difficulties associated with directly measuring coordination and control patterns (see e.g., Hall & Gingerich, 2009). As a result, I was unable to capture sub-national variation in business systems (Allen, 2004; Lane & Wood, 2009; Whitley, 2007) or assess the impact of increased internationalization on the diversity of firms within countries (e.g., Allen & Whitley, 2012; Lange, 2009; Whitley, 2012).
A second limitation is that the indicators may not capture all relevant institutional characteristics. One example is the financial systems measure, which captures the relative importance of capital versus credit. Although this measure represents an improvement relative to previous measures, it does not capture other features of national financial systems, such as the relative importance of cross-shareholdings (Whitley, 1999, 2007). Similarly, the education measure captures the effectiveness of the education system in ensuring an equitable distribution of learning opportunities rather than the specificity of the skill-provision system. Although both can be linked to variations in the accumulation of vocational skills, future studies could consider alternative measures, such as measures of the relative importance of secondary and tertiary vocational education (see e.g., Culpepper, 2007; Estevez-Abe, Iversen, & Soskice, 2001).
A third limitation lies in the use of fuzzy-set QCA to assess innovation patterns. The use of fuzzy-set QCA allowed for a configurational assessment of the impact of institutional conditions on innovation specialization. Nonetheless, advanced statistical techniques have advantages on other fronts. For instance, statistical techniques more readily accommodate more detailed measures of specialization outcomes, such as patent-level rather than sector-level indicators of innovation radicality (see e.g., Akkermans et al., 2009). Hence, such techniques might better capture situations in which sub-sectors of the same industry may prosper under different institutional conditions (Casper & Whitley, 2004).
Despite these limitations, the findings illustrate that the business systems typology holds continued relevance for describing national variety in institutional frameworks. The findings also indicate that the detail it offers may provide the nuance needed to explore more complex relations between institutions and outcomes (Casper, 2010), and to uncover new institutional complementarities.
Footnotes
Appendix
Analysis of necessary conditions.
| Condition | Consistency | Coverage |
|---|---|---|
| State dominance | 0.38 | 0.31 |
| Prevalence of clusters | 0.75 | 0.53 |
| Burden of regulation | 0.43 | 0.33 |
| Market capitalization to credit | 0.47 | 0.46 |
| Strength of education system | 0.70 | 0.58 |
| Union density | 0.43 | 0.42 |
| Centralization of bargaining | 0.43 | 0.37 |
| Trust | 0.42 | 0.46 |
| Paternalism | 0.35 | 0.29 |
Acknowledgements
I wish to thank Verena Girschik, Peer Hull Kristensen, Ayse Saka-Helmhout and the three anonymous reviewers for their encouraging and insightful comments and suggestions. I am particularly grateful to Arndt Sorge, who encouraged me to write this paper, and who hosted me at the Wissenschafszentrum Berlin für Sozialforschung (WZB) where I wrote a first full draft. Finally, I thank Christian Asmussen, Hans van Ees, Nicolai Foss and seminar participants at Copenhagen Business School and the WZB for valuable comments on earlier versions of this paper.
Funding
This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.