Introduction

Innovation has been pointed out as an important determinant of technical progress, which in turn is a main driver of international competitiveness and economic growth in the long run. Thus, the question of how to foster innovation has always been a central concern for policy-makers to ensure a good pace of income growth and welfare for the economy.

In the field of comparative capitalisms, the varieties of capitalism (VoC) approach has made a great contribution to the topic of the institutional determinants of innovation. Nonetheless, since the publication of the Hall and Soskice’s seminal work (2001), a controversial and still unresolved debate was triggered (see, e.g., Amable 2003; Taylor 2004; Boyer 2005).

The core argument of the VoC approach is that country-specific institutions not only determine innovation, but also explain the innovation patterns of a country and endow firms with comparative advantages in certain industries. As in other political economy approaches, complementarity and institutional coherence are the key. The former concept embraces the idea that the joint presence or combination of a particular set of institutions improve the efficiency of the whole system or variety of capitalism (Whitley 1999; Hall and Soskice 2001; Amable 2003), while the latter while the latter points to the congruence between the same types of institutions (Kenworthy 2006). Therefore, the existence of institutional coherence boosts institutional complementarities. Hall and Soskice see two kinds of complementarities, the Liberal Market Economy (LME) type and the Coordinated Market Economy (CME) type—an additional model of capitalism is conceptualized, the Mixed Market Economy (MMEs) (Molina and Rhodes 2007), but it is defined by its lack of institutional coherence and by underperformance compared to the other two types. While the LME institutional setting boosts radical innovation due to the mixture of fluid labour and capital markets, general educational system and competitive inter-firm relationships, CMEs encourage incremental innovation, thanks to the combination of cooperative industrial relations and a tight labour market, patient bank-based financial system, industry-specific education and collaborative inter-firm relations. Therefore, liberal institutions support productive specialization in industries characterized by radical innovation, while coordinated institutions encourage specialization in activities, where incremental innovation is required to compete.

The discussion on VoC claims is partly explained, because a strict interpretation of the theory would lead to three contested corollaries. First, VoC scholars assume that firms’ strategies are endogenous to the institutional framework (Allen 2004). National institutions are the main determinant of corporate strategies, and firms are mere passive institutional absorbers with little room for manoeuvre to undertake their own policies. Due to corporate strategies varying across countries, if a firm would like to perform well, it needs to be located inside a particular institutional framework. Consequently, the VoC approach assumes that firms within a given national economy are homogeneous. Second, incremental and radical innovation corporate strategies are fully dependent on the institutions of the variety of capitalism, and institutions that are helpful to incrementally innovate, discourage radical innovation and vice versa. Third, stemming from the latter idea, the distinction between industries characterized by radical innovation and those more prone to incremental innovation makes sense, but one might think that both types of innovative activity can be found in every industry.

Empirical research shows, at best, partial evidence about these corollaries. For instance, some studies have shown that different firm profiles exist within a national economy (Kirchner 2016). Furthermore, economic activities that apparently do not fit well in CMEs are indeed well performed in these countries, and, thus, corporate strategies do not fully rely on national institutions but on the firms’ own capabilities (Herrmann 2008; Herrmann and Peine 2011). At the same time, it is clear that the institutional structure of CMEs such as Germany is only present in the manufacturing sector, due both to its historical origins (Streeck and Yamamura 2001) and to the features of the service industries (Thelen 2014).

Regarding the German case, political economists have detailed the deep process of institutional change experienced by the Modell Deutschland and have pointed out that only large manufacturing firms are embedded in the traditional coordinated institutional framework (Palier and Thelen 2010; Hassel 2014). Likewise, the international business literature underlines other structural characteristics of the firm as determinants of innovation, such as the economic branch in which it operates, its exporting activity or its size (Cohen and Kepler 1996). Therefore, one can ask whether institutions are important for innovation or, on the contrary, if innovation is solely explained by other structural factors. In other words, it might be the case that coordinated institutions are only an option for the most advanced and productive firms, and the link between them and innovation may well be spurious.

Taking the German manufacturing sector as a case study, this paper explores the link between coordinated institutions and innovation highlighted by VoC scholars. Based on the IAB Establishment Panel, we empirically explore the relationships between institutions and innovation at the firm-level. In this way, we try to contribute to the existing literature by measuring the effect of specific institutions on different and apparently contradictory types of innovation, as well as by capturing the effect of other non-institutional variables that have not been studied by VoC scholars but are central for the business literature. Therefore, we try to build a bridge between both strands of literature.

More precisely, two objectives are addressed. The first one (1) is to measure the joint impact of the abovementioned set of institutions on four types of innovation, namely, incremental, radical, process and imitation. It is expected that those firms that make full use of all coordinated institutional arenas—thus expressing coherence—will be more prone to perform incremental and process innovation; on the other side, if the theory is correct, the effect of the same coordinated institutions on radical innovation should be negative. Its effect on imitative innovation is undefined. Our second goal (2) is to capture whether the innovation performance varies when taking into account the structural variables of the firm, i.e., the industry, its exporting activity or its size. In addition, we measure the effect of institutions on the most advanced firms in the economy; hence, potential biases are properly controlled and the role of coordinated institutions is better assessed. We expect that the importance of institutions would be minor for advanced firms. Our methodology consists of logistic modelling.

The reason for selecting Germany as a case-study is threefold. It is an illustrative example of an innovative economy, with an export-oriented and highly competitive manufacturing sector. The country is also the main representative of continental CMEs, and its institutional setting, though it has been transformed in the last 30 years (Streeck 2009), continues to be unique. The third reason is practical, the IAB Establishment Panel offers high quality information about the innovative behaviour of the firm and its business policy.

The structure of the paper is as follows. Section “Theory and findings on the institutional determinants of innovation” contains a comprehensive literature review on the nexus between institutions and innovation. Methodological aspects are presented in Section “Data set and variable definitions”. A sample description and a set of preliminary results are reported in Section “Sample description and preliminary analysis”. The fifth section provides the results of the econometric analysis. Section “Concluding remarks” concludes.

Theory and findings on the institutional determinants of innovation

The role of institutions in the innovation process: the VoC approach

The VoC approach (Hall and Soskice 2001) sustains two interesting proposals: the firm-centered approach to institutions, which are considered as resources or tools for coordination; and the idea that national institutions, when fully coherent with each other, generate environments, where firms develop capabilities to innovate and comparative advantages in some industries. Taking into account both proposals, national economies are clustered in two groups: liberal-market economies (LMEs), exemplified by the US, and coordinated-market economies (CMEs), whose archetype is Germany.

Firms are considered the key players in political economies. They are conceptualized as coalitions of economic actors—shareholders, managers, workers, financers—with relational and dynamic capabilities (Hall and Soskice 2001: 6; Jackson and Deeg 2008: 549). It is only by the interaction among these actors that firms’ core competencies can be developed and exploited. Yet, coordination problems in the form of moral hazard constantly emerge, and national institutions are seen as resources to solve them and ensure coordination.

In economies, such as the US, companies perform in fluid, deregulated and competitive markets, in which prices and individual formal contracts are the main instruments for coordination. Other standard features of these economies are weak collective actors and hierarchical decision-making processes within the firm. In this manner, national institutions encourage investment in transferable assets, allowing companies to make quick decisions and reconfigure their competitive strategies. On the other hand, in economies, such as Germany, firms are embedded into an institutional setting composed of strong and corporatist collective actors that efficiently foster forms of coordination based on strategic interactions, in which the market plays a secondary role. It provides mechanisms for exchanging information, and monitoring and sanctioning others’ behaviour, so agents are able to make credible long-term commitments among themselves. These non-market coordination practices support investments in non-transferable or co-specific assets, which are those whose returns depend upon long-term cooperation among agents.

What connects the first with second theoretical proposal is that corporate strategies are shaped by national institutions. Firms are well-disposed to use country-specific institutions, because they endow them with technological capabilities and comparative advantages in certain industries; i.e., firms’ strategic preferences are endogenous to the national institutions (Allen 2004). Those economic activities that require fast—and sometimes risky—adaptations of contractual arrangements are better supported by LME institutions, such as flexible labour markets, general education, short-term oriented company finance and competitive relations among firms. On the other hand, those activities that involve long-term commitment and cooperation among actors benefit from CME institutions, such as tight labour market, industry-specific education, long-term bank based finance and inter-firm cooperative relations. In this way, LMEs provide comparative advantages and support productive specialization in industries characterized by radical innovation—the creation of a totally new product or a major change in the production process—whereas CMEs provide advantages and encourage specialization in industries characterized by incremental innovation—defined as continuous improvements to a product or production process that already exists.

It is well known that VoC literature highlights four main institutional dimensions in which firms need to coordinate their endeavours with other agents: industrial relations, the financial system and corporate governance, the education system and inter-firm relationships. However, the whole national institutional framework is more than the sum of its parts. What really matters is the coherence between these four spheres, and not each of them individually. Institutional complementarity is a common concept in political economics (Jackson and Deeg 2008). It takes up the idea that the returns of an individual institution increase when another institution exists, and that a particular form of coordination in one institutional arena tends to foster analogous types in others. More precisely, the effects of a coordinated institution on economic performance will improve when interacting with other coordinated institutions. What provides firms with innovation capabilities and advantages in particular activities is the whole institutional structure of an economy (Whitley 1999; Amable 2003; Hall and Gingerich 2009).

Turning to Germany, those firms in the country that make full use of non-market modes of coordination—a non-market corporate strategy—are expected to efficiently achieve the economic results predicted by Hall and Soskice (2001). Focusing on innovation, the reasoning may be summarized as follows. German institutional framework supports incremental innovation. This innovation type is normally based on a skilled workforce with industrial and firm specific skills that perfectly understands the functioning of the production process and the product line in which they work. At the same time, the workforce needs to be sure that technological progress will not affect their job situation, so long-term employment contracts, control and participation mechanisms in firms’ decision-taking are necessary for its cooperation. Wages and working conditions are set in sectoral collective agreements. This is because firms are interested in settling the distributional conflict with labour away from workplaces, so cooperation with the workforce is easily achieved. In addition, by promoting the same labour standards and low wage dispersion for the entire industry, sectoral agreements discourage the poaching of skilled workers by other companies. On the other side, close links between companies, suppliers and research institutions are needed to foster information exchange, and to detect and define areas of improvement to work jointly on them. Finally, the German stakeholder model of corporate governance—whose principle is that the company should not be controlled by any of its constituent parts—encourages incremental innovation too. It discourages high-risk practices that could damage employment security, such as the entry into new markets, and focuses on the increase in market share by the ongoing improvement of existing product lines and production processes. This is also helped by a traditional bank-based financial system that supplies firms with patient capital, and a cross-shareholding network, which protect firms against hostile takeovers so they can focus on productive issues (Vitols 2001).

Nonetheless, business literature has highlighted that some structural features of the firm satisfactorily explain its innovative activity. Three of them are particularly relevant due to their influence on the rate of investment in R&D, namely, the industry, the export status and the firm size. In this respect, in some industries firms require higher levels of investment in R&D and innovation than in others to remain competitive (Shefer and Frenkel 2005). In the same way, exporting firms tend to be more innovative due to the harsh competition in international markets (Rogers 2004; Pla-Barber and Alegre 2007). Concerning the effect of firm size, there are two positions (Cohen and Kepler 1996; Chandy and Tellis 2000). Arguments in favour of superior large firms’ ability to innovate are their greater resource availability, i.e., financing capability, sales volume and productivity—which reduce the fixed costs of innovation—and the wider range of skills of their workforce. On the other side, advantages of small and medium firms are their flexibility and capacity to make faster decisions, reallocate resources and launch new projects. Although both views have received empirical support, the meta-analytical review made by Damanpour (2010) shows that the former position is more plausible, and innovation is more favoured by large firms.

These insights from business literature present some important challenges for VoC’s innovation theory, particularly when looking at the German case. In addition, some works detected that many firm profiles and corporate strategies exist within the same national economy (Herrmann 2008; Herrmann and Peine 2011; Kirchner 2016). Furthermore, the process of institutional change the German economy has gone through has put the traditional Modell Deutschland in marked retreat. Nowadays, coordinated institutions—and mainly the ones related to industrial relations—are primarily employed by the most advanced manufacturing firms. For instance, trade unions and strong work councils are still firmly grounded in large manufacturing companies located in core manufacturing industries, but their influence on other types of firms has been eroded in the last 30 years (Hassel 2014). However, is this simply because their levels of productivity are high enough to deal with coordinated institutions? Or are coordinated institutions real efficient tools to improve their innovative performance? These queries need to be solved to assess the real impact of institutions on innovation and are dealt with in the present paper.

Review of empirical research: indicators of innovation and main empirical findings

Our literature review draws from two types of studies: those which try to test the VoC hypotheses by comparing the economic outcomes of diverse institutional configurations; and those which, focusing on Germany, estimate the effect of country-specific coordinated institutions—mainly those related to the labour market—on different types of innovation. Our aim in this article is connecting both approaches. As a result, it constitutes an empirical country-case study that uses firm-level data and, at once, takes part in the debate on the institutional determinants of innovation.

It is worth pointing out that contributions regarding the impact of institutions on innovativeness are sometimes difficult to compare. On one side, studies on the topic employ different indicators of innovation and select distinct institutional domains as explanatory variables. On the other side, some of them explore the nexus between institutions and innovation in one particular activity—the pharmaceutical industry—others focus on several countries, and others are country-case studies. Table 1 summarizes the most relevant works on the topic so far.

Table 1 Literature review
Full size table

A short comment on the indicators of innovation might be appropriate here. Three types of them are used in the literature. Some papers assume that medium–high (MHT) and high technology (HT) industries are inherently characterized, respectively, by incremental and radical innovation strategies (e.g., Schneider and Paunescu 2012). Thus, a country’s productive specialization, export performance or revealed comparative advantage (RCA) in these activities, reflect institutional support for one of these types of innovation. Second, most articles employ patents as empirical material and build several indexes with them (see, among others, Taylor 2004). The principal advantages of patents are that they are an objective measure of innovation and patent data are widely available for most countries. On the other hand, its main drawback is that it is difficult to know which type of innovation is contained in the patent. Third, firms’ innovative activity is also collected by surveys. This is the case of this paper and others based on the IAB Establishment Panel (e.g., Addison et al. 2017a), which asks whether any type of innovation has been introduced by the establishment. Nevertheless, this option presents two limitations. There is not an objective measure of innovation, but it is the owner or manager of the establishment who is asked to rank it; therefore, this variable might be affected by subjectivity bias. On the other hand, the researcher only knows whether an innovation has been performed over a time period, but not how many times it has been accomplished. Hence, relevant information is lost and methodological options are more restricted—e.g., it is not possible to use count data models, such as Kraft et al. (2011).

Regarding the empirical findings, papers that studied innovation patterns of national economies from a VoC perspective have obtained inconclusive empirical evidence. In one of the most comprehensive studies to date, Schneider and Paunescu (2012) examine the effects of 26 OECD countries’ institutional frameworks on the export share and the Balassa index of RCA in MHT and HT industries, and find that LMEs have RCA in HT industries, while CMEs have it in MHT industries. Furthermore, they detect signals of liberalization and find that those economies that moved towards the LME type of capitalism increased their export share and RCA in HT activities more than those economies that remained LME since the start of the period. Schneider et al (2010) analyse 19 OECD economies and also find empirical support for the hypothesis that, in general, LMEs hold RCA in HT industries, but CMEs do not. Interestingly, they capture that the combination of extensive university training and a large stock market is a sufficient condition to perform well in HT industries, while lax employment protection and low collective bargaining coverage do not appear to be important requirements. Ultimately, this analysis challenges the complementarity concept, and suggests that hybrid institutional frameworks are able to achieve strong economic performance in these industries.

On the other side, a comprehensive critique against VoC’s theory of innovation is in Taylor (2004). He challenges the view that some industries are inherently more prone to radically innovate than others and criticizes the methodology employed by Hall and Soskice (2001) to test their claims. He finds that national institutions cannot explain innovation performance and that the results are strongly influenced by the inclusion of the US in the regressions. VoC’s innovation thesis is also rejected by Akkermans et al (2009). They address the question whether LMEs innovate more radically than CMEs. To this end, they build a multidimensional index of radicalness based on patents (see Table 1). Although LMEs and CMEs present patterns of specialization similar to those predicted by VoC, they find that radical innovations are not more common in LMEs in all industries if we understand radicalness as a multidimensional concept.

Focusing on the pharmaceutical industry, Herrmann and Peine (2011) explore the nexus between skill specificities and the type of innovation and competitive strategies. By comparing the UK, Germany and Italy, they test whether the skills required to perform radical, incremental or imitative innovations result from firms’ competitive strategies or from the availability of these qualifications thanks to the country-specific educational and training system. Although evidence for one of the VoC’s claims is found—the link between skill types and innovation (general skills boost radical innovation, whereas specific skills push incremental innovation)—their results suggest that the main determinant of firms’ innovation patterns are their competitive strategies and not national institutional frameworks, because companies with the same competitive strategy present similar workforce composition and skill profiles, regardless of location. In an earlier study on the pharma industry in the same three countries, Herrmann (2008) obtains similar results: despite any alleged institutional (dis)advantage, firms are able to pursuit the same strategy in different countries using functional equivalents, i.e., by circumventing the economy’s typical institutions to secure the required skill specificities for their corporate goals. They might import from abroad key employees that cannot be found in the country; or perform contractual improvisation by concluding non-standard contracts. Thus, Herrmann concludes, there is not a sole competitive strategy in a national economy and firms are not mere institutional-takers, but they have the ability to institutionally innovate with the aim of being competitive in any particular industry.

One of the most novel contributions to the topic was recently made by Witt and Jackson (2016). They revisit the complementarity concept and develop and alternative theoretical framework to explain why positive effects on radical innovation might arise from the combination between LME unconstrained market oriented transactions and the beneficial constraints of CME institutions—concretely, the combination of either coordinated employment relations and liberal institutions in other domains or liberal corporate governance with coordinated institutions. Thus, a positive marriage between apparently conflictual institutional logics may be achieved to counterbalance some weaknesses of pure institutional frameworks.

A second group of papers relevant to our purposes has investigated the link between institutions and innovation patterns of German firms. Most of them are based on the IAB Establishment Panel and, interestingly, they all find a positive—or at least non-negative—impact of German coordinated institutions on all types of innovation. Therefore, they implicitly contradict the VoC claim on the harmful effect of non-market institutions on radical innovation.

Addison and his co-authors made two recent contributions to the topic. First, they (Addison et al. 2017a) investigate the nexus between cooperative industrial relations—i.e., the joint presence of collective agreement and work council in an establishment—and the same four types of innovation studied in this paper—radical, incremental, imitation and process—with the aim of challenging the studies made in the US that unanimously point to a negative effect. They study how innovation is affected by firm transitions towards sectoral agreements and vice versa and find that German coordinated institutions positively impact on it. In a second paper, Addison et al. (2017b) quantify the effect of the Pacts for Employment and Competitiveness—the so-called opening clauses, a tool for organized decentralization of labour relations—on wages, employment, productivity, innovation and survivability of the establishment. In this case, innovation is a dummy variable that captures whether any process or product innovation was carried out, without differentiating between them. They report a certain positive effect of these clauses on innovativeness.

Focusing on the German pharmaceutical sector, Allen et al. (2011) seek to challenge the VoC approach. Germany is highly competitive in the pharma industry, characterized by radical innovation patterns, despite the fact that this does not fit with the coordinated features of its institutional framework. They estimated a logistic regression for the year 2007 in which the dependent variable is radical innovation and the predictors are 3 HT and 3 MHT industries, along with sectoral agreements and the presence of work councils. The results suggest that neither collective agreements nor work councils had a negative effect on radical innovation, and that none of the industries included in the regression were less prone to radically innovate than pharma.

Kirchner (2016) conducts a survey of German firms to research the links between national institutions, firm profiles and economic outcomes, distinguishing radical and incremental product innovation. He finds five types of firms, one of which is the ideal–typical VoC firm (the most frequent in the sample). While the author captures a non-significant effect between institutions and economic outcomes, he finds that ideo-typical VoC firms display a high extent of institutional embeddedness and the best performance regarding both incremental and radical innovation.

Finally, Kraft et al. (2011) employ a sample of 148 German stock companies to assess the impact of the co-determination law of 1976 (MitbestG) on innovation. The results show that, contrary to what mainstream economics predicts, codetermination does not harm innovativeness (even a small positive effect is found in some regressions).

In sum, although previous research suggests inconclusive evidence on VoC claims, a general positive effect of coordinated institutions on innovation is found in Germany.

Data set and variable definitions

The analysis is based on the IAB Establishment Panel, a representative data set of the Institute for Employment and Research (IAB) of the Federal Employment Agency (BA). It annually surveys establishments (not firms) from all sectors and sizes using a stratified random sample of all plants that employ at least one worker covered by social insurance on 30 June. The first survey was conducted in 1993 only for the former West Germany and was extended in 1996 to eastern establishments. The sample size has steadily grown, increasing from 4265 establishments in 1993 to more than 15,000 from 2001 onward (for further details, see Fischer et al. 2009, and Ellguth et al. 2014). This data set collects information about a wide range of topics, such as innovation, labour relations, training and other business policies. Thus an extensive amount of high-quality variables can be used to achieve this paper’s investigation goals.

Three years are selected for the analysis—1998, 2007 and 2013—since the required variables are only available in these cross sections.

The relevant variables of the study are measured in the following way (see Table A.1. in the Online Appendix for more details):

  1. (a)

    Innovation variables Establishments can introduce three types of product innovation: (1) incremental, defined as the improvement or further development of a product already manufactured by the establishment; (2) radical, which involves the introduction of an entirely new product for which a new market was created; (3) imitative, when an establishment starts to offer a product that was previously available in the market. On the other hand, the owner or manager of the establishment was asked whether new developed procedures which have improved production processes were introduced, thereby capturing (4) process innovation. Finally, we created a (5) dummy variable, coded as 1 when an establishment has performed either incremental, radical or process innovation (imitation is excluded, because it is the less complex form of innovation and might bias the results).

    However, it has to be noted that there is a break in the innovation measure from 2008 onwards. Prior to 2008 the interviewee was inquired about the innovative activity of the establishment in the last 2 years, whereas, since 2008, the question references the last year, so the results in 2013 should be interpreted with caution. Despite that, we have decided to compare these 3 years, because it allows coverage of a larger period of time and, when looking at the descriptive statistics and the models by year in the Appendix, the variables do not display strong variations. In addition, information about process innovation is only available from 2007 onwards, so only two cross sections are included in its analysis.

  2. (b)

    Institutional variables The four institutional spheres highlighted by VoC literature are operationalized as follows:

    1. (i)

      Industrial relations it is a dummy variable that takes value 1 if the joint presence of collective agreement—either firm or sectoral—and work council is found.

    2. (ii)

      Cooperation among firms and other institutions it is captured by a dummy variable coded as 1 if the establishment declares to carry out research and development (R&D) in cooperation with other establishments, universities, consultants or non-university research institutions.

    3. (iii)

      Vocational training this binary variable takes value 1 if the establishment declares to have offered a permanent position to at least one of the apprentices that completed the vocational training in the present year. This means that the establishment both contributes and benefits from the German dual vocational training system.

    4. (iv)

      Corporate governance there is not a variable in the data set that directly deals with this theme. Nonetheless, respondents are asked whether investment plans are set out in writing, so this question was selected to approximate the topic. We assume that if investment plans are explicitly indicated, more voices in the establishment could veto strategic decisions and condition the course of the company.

    5. (v)

      With these four variables, we create the main explanatory variable of the paper. The variable institutional spheres captures the number of institutions present in a firm. Thus, when it is equal to four, it is capturing the adoption of a non-market corporate strategy by an establishment, i.e., the coherence concept. When this happens, the establishment is what we have termed “varieties of capitalism firm” (VoC firm).

  3. (c)

    Structural variables Following the insights from business literature, we consider three variables that may affect the overall performance of the establishment and, particularly, its innovation pattern. They are industry, exports and establishment size.

    1. (i)

      Industry this variable takes four values, due to manufacturing industries are clustered using the OECD taxonomy of economic activities based on R&D (Galindo-Rueda and Verger 2016): medium–low technology (MLT), medium technology (MT), medium–high technology (MHT) and high technology (HT).

    2. (ii)

      Exports it takes three values: non-exporter, low and medium exporter (establishments which exported between 1% of their sales and the average value of the total sample) and high exporter (those which exported above-average sample sales).

    3. (iii)

      Size three establishment sizes are differentiated in the analysis: small (5–49 employees), medium (50–199 employees) and large (200 or more employees) establishments.

In addition, other controls are included in the estimations and are defined in Table A.1. The main descriptive statistics are reported in Tables A.2 and A.3, in the Online Appendix.

Sample description and preliminary analysis

The 3-year sample is composed by 5718 observations in total. At the outset, Fig. 1 shows that incremental innovation is the most common type in German manufacturing industries. On the other side, as might be expected, radical innovation is the rarest one, performed by only the 20.45% of the sample. Process innovation and imitation are undertaken by slightly more than one third of the establishments.

Fig. 1
figure 1

Source: IAB Establishment Panel, own calculations

Proportion of innovative establishments in the sample.

Full size image

Moreover, German establishments have become less innovative over time. Figure 2 offers a comparison of the proportion of establishments that innovate between the 3 sample years. As can be observed, it has decreased in all sorts of innovation, particularly when referring to the incremental one. Nonetheless, as mentioned above, this has to be read with caution due to the changes in the questionnaire.

Fig. 2
figure 2

Source: IAB Establishment Panel, own calculations

Proportion of innovative establishments by year.

Full size image

The institutional features of the sample are graphed in Fig. 3. The sample partially reflects the strong signs of institutional change displayed by the German political economy. Note that the share of VoC establishments has decreased from 18.23 to 11.97% in 2013, and the share of establishments with three institutional spheres also experienced a sharp decline. On the other side, the category that increased the most is the one with no coordinated institutions. In the Online Appendix the level and evolution of the four individual institutions are displayed (Table A.7). Figure 3 also suggests that, contrary to what Hall and Soskice (2001) claimed, there is not only one possible competitive strategy in a particular economy, but a wide array of them. Furthermore, the supposed optimal strategy in Germany, which is supported by the country’s non-market institutions, is carried out by a minority of establishments.

Fig. 3
figure 3

Source: IAB Establishment Panel, own calculations

Number of institutional spheres by year.

Full size image

Focusing on the relationship between institutions and innovation, preliminary results suggest the importance of coordinated institutions for innovation. Table 1 displays conditional probabilities (P(.)) of a given innovation type, i.e., the proportion of establishments that have both introduced an innovation and present a particular institutional feature as a share of total establishments with that institutional feature. For instance, saying that the conditional probability to introduce an incremental innovation by establishments covered by both collective agreement and work council is 81.16%, is the same as saying that 81.16% of establishments within this subsample (establishments with collective agreement and work council) performed an incremental innovation.Footnote 1

In general, the probability of an establishment innovating is always higher when an institutional feature is present. This is particularly true when looking at incremental and process innovation. On the contrary, institutional features seem to be of lesser importance for imitation and radical innovation.

When looking at the combination among institutions in an establishment, it is clear that the conditional probability to innovate increases with the number of coordinated institutional spheres interacting. VoC establishments are particularly prone to implement incremental innovations; but, at the same time, they are in general more inclined to innovate than others. Although the theory states that coordinated institutions are not the best ones to promote radical innovation, data shows that it is more common in VoC establishments than in others. In other words, it seems that, in Germany, the absence of coordinated institutions does not boost this innovation type.

Nonetheless, Table 2 also shows that the three structural variables matter, where innovation is regarded. A fairly linear relationship can be perceived: the probability to innovate increases with the R&D intensity of the industry, with the exporting activity and with the size of the establishment. The former result challenges one of the VoC’s basic axioms. One might expect that the probability to undertake incremental innovations would be greater in MHT industries than in HT industries, because establishments in the former rely on this type of innovation to compete, since “the problem [in MHT industries] is to maintain the high quality of an established product line, to devise incremental improvements to it that attract consumer loyalty, and to secure continuous improvements in the production process to improve quality control and hold down costs” (Hall and Soskice 2001: 39). However, it is clear that the more advanced the industry is, the greater the probability of an establishment performing any sort of innovation.

Table 2 Conditional probabilities to innovate by institutional feature.
Full size table

Finally, Table 3 shows the tetrachoric correlations (correlation between any pair of binary variables) between innovation and institutions. It can be seen that all institutions are significantly and positively correlated to all innovation types. Again, institutions appear to be especially favourable to incremental innovation, but also to process innovation. The strongest correlations are found between R&D cooperation and the four innovation types. Nonetheless, these results are preliminary, because structural variables and other observables are not controlled.

Table 3 Tetrachoric correlations.
Full size table

Econometric analysis

In consistence with the abovementioned goals, the subsequent hypotheses are tested:

HI.

The combination of the four coordinated institutions in an establishment would be beneficial to incremental and process innovation, but would negatively affect radical innovation.

HII.

The effect of institutions on innovation might be affected by the structural variables of the firm; therefore, institutions may be less important to innovation if the establishment is either a high-exporter, large or operates in MHT or HT industries.

Following the methodology of recent studies based on the same data set (Addison et al. 2017a; Allen et al. 2011), the effect of institutions on innovation is assessed using logit regression models. Concretely, five models are estimated—one for each type of innovation. They are pooled data models, in which the three available cross sections—1998, 2007 and 2013—are introduced at the same time. The relevant explanatory variable is the number of institutional spheres found in each establishment. As indicated above, when the four spheres are present, the establishment is a VoC firm. The model also includes the three structural variables—industry, exports and establishment size—and, along with them, four additional controls: the location of the establishment, the state of equipment, satisfaction with past year profits and the expected business volume.

This approach is formalized in Eq. (1):

$$\begin{aligned} L_{i} \left( {{\text{innovation}} = 1} \right) & = \beta_{0} + \beta_{1} {\text{Institutional spheres}}_{i} + \beta_{2} {\text{Exports}}_{i} + \beta_{3} {\text{Industry}}_{i} \\ & \quad + \beta_{4} {\text{Size}}_{i} + \sum \beta_{5} {\text{Controls}}_{i} + \varepsilon_{i} . \\ \end{aligned}$$
(1)

In the equation, \({L}_{i}\) stands for the logit of innovation, which is a function of the linear combination of the set of variables on the right side of the equation and it is estimated using the maximum likelihood procedure. To facilitate the interpretation, the results are presented using odds ratio, average marginal effects and predicted probabilities. These transformations are explained in more detail below.

Multicollinearity problems between the independent variables were not found. In addition, the robustness of the regression results was examined by running the same five models by year.

Testing HI: results of the model

The results of the regressions serve us to test HI and are presented in Table 4. It is clear that the combination of the four institutional spheres is a key innovation driver. This occurs in all estimated models, but it is especially noteworthy for incremental and process innovation. Aside from this, the second model shows that radical innovation is also favoured by VoC establishments, but to a lesser extent than other innovation types. Finally, imitation, as the less complex sort of innovation, presents positive but comparatively smaller effects, as might have been expected.

Table 4 Logistic regressions.
Full size table

On the other side, it has to be noted that the odds ratios of three structural variables are also statistically significant and positive, but its size is much smaller than that of the institutions. The results of some control variables are rather interesting; for instance, the probability to radically innovate and perform imitations is larger in eastern establishments, whereas process innovation is more likely in the West. The state of the equipment is only relevant for incremental innovation. A satisfactory evaluation of the profit made in the last year is negative for incremental innovation, but positive for the remaining types. Finally, the expectations of getting higher business volume encourage all kinds of innovation.

Reported in the Online Appendix are the models with the individual variables (Table A.8). Two of them—cooperation to undertake R&D and corporate governance—impact positively on all innovation types. The effect of vocational training is only statistically significant to perform imitations or process innovations—although the estimated effect size is small—while the impact of cooperative IIRR, even though negative, is only significant for imitative innovation.

All in all, the results lead us to reject HI: coordinated institutions are the main driver of all types of innovations, and the complete absence of them does not have a positive effect on radical innovation.

Testing HII: average marginal effects

To test the Hypothesis II we use the results of the logit models and calculate the marginal effects on innovation of each category of the institutional spheres variable when all other variables are at their means.Footnote 2 These effects are computed for the three structural variables. Using this technique, problems of spurious relation are controlled and the effect of institutions is better isolated. For the sake of simplicity, in the following three graphs only VoC establishments are represented. In the Online Appendix the coefficients for the rest of institutional spheres are included (Tables A8–A11).

Three general comments can be made to begin with. First, in comparison to the reference category (Inst. Sphere = 0), marginal effects are on average much higher for VoC establishments in all categories of the abovementioned variables. Second, the average marginal effects do vary across categories of these variables. Third, as shown in Tables A8–A11, the more institutions are present in an establishment, the higher the probability to innovate.

When looking at the industry (Fig. 4), the effect of institutions on innovation differs among industry clusters and the type of innovation itself. For instance, the propensity to incrementally innovate of VoC establishments in MLT and MT industries is almost 50 percentage points (pp) greater in comparison to the reference category, while in MHT and HT industries it is, respectively, around 40 pp and 35 pp greater. Nevertheless, when looking at radical innovation, the contrary is the case and the combination of the four coordinated institutions has greater impact in MHT and HT industries. Process innovation and imitation present similar effects across industries.

Fig. 4
figure 4

Source: IAB Establishment Panel, own calculations

Average marginal effects of VoC firms by industry. Note: the figure illustrates the increase in the predicted probability of innovating by type of industry, holding all variables at their averages.

Full size image

Similar results are obtained if the exporting activity of the establishment is considered (Fig. 5). Incremental innovation is benefited relatively more from coordinated institutions in non-exporting than in high-exporting establishments, while the contrary occurs for radical innovation. Again, the coefficients remain at almost the same level across categories for process innovation and imitation. The fact that marginal effects in HT industries and large exporters are lower might be reflecting the importance of these structural factors: the baseline innovation levels required to remain competitive in these categories are much higher than in the others; therefore, the effect of institutions is lower (although large and statistically significant).

Fig. 5
figure 5

Source: IAB Establishment Panel, own calculations

Average marginal effects of VoC firms by export status. Note: the figure illustrates the increase in the predicted probability of innovating by exporting intensity, holding all variables at their averages.

Full size image

The probability to innovate of VoC establishments is much more similar across size categories (Fig. 6). Concretely, large establishments are benefited slightly less from coordinated institutions to incrementally innovate than small and medium ones. On the contrary, the effect on process innovation appears to be larger for them.

Fig. 6
figure 6

Source: IAB Establishment Panel, own calculations

Average marginal effects of VoC firms by establishment size. Note: the figure illustrates the increase in the predicted probability of innovating by establishment size, holding all variables at their averages.

Full size image

Broadly speaking, the results show that coordinated institutions are favourable not only to incremental but to all types of innovation and that being a VoC establishment is worthy in all industries, types of exporting status and sizes. Again, these findings present somewhat of a challenge to the Hall and Soskice’s hypothesis that radical innovation is better promoted by market institutions. It seems that in Germany the main road to any kind of innovation is a non-market corporate strategy.

Testing HII: predicted probabilities to innovate of core establishments in core industries

In this section, we continue testing HII by focussing on German core manufacturing establishments. These establishments usually operate in international export markets, concentrate a large amount of resources and employ many workers. Furthermore, they operate in the most advanced manufacturing industries, i.e., MHT and HT ones. Thus, due to their own structural features, it is expected that core establishments would be highly innovative, so the role of institutions here might be of lesser importance. Ultimately, this is the definitive test for the effect of coordinated institutions. For that purpose, we calculate predicted probabilities to innovate at each value of the institutional spheres variable only taking into consideration the abovementioned core establishments. Results are plotted in Figs. 7 (MHT industries) and 8 (HT industries).

Fig. 7
figure 7

Source: IAB Establishment Panel, own calculations

Predicted probabilities to innovate of large exporting establishments by number of institutional spheres (MHT industries). Note: the figure illustrates the probability of innovating of advanced firms in MHT industries, holding all variables at their averages. The probability increases with number of institutional spheres.

Full size image
Fig. 8
figure 8

Source: IAB Establishment Panel, own calculations

Predicted probabilities to innovate of large exporting establishments by number of institutional spheres (HT industries). Note: the figure illustrates the probability of innovating of advanced firms in HT industries, holding all variables at their averages. The probability increases with number of institutional spheres.

Full size image

Interestingly enough, it can be detected that the probability to innovate increases again with the number of institutional spheres. VoC establishments always present the highest predicted probability to innovate, which is particularly important in incremental—above 97% in both MHT and HT industries—and process innovation—above 70%. Furthermore, this relation remains for radical innovation, challenging again the assumption that coordinated institutions are inefficient tools to undertake it,

The main conclusion here is that, although we find highest marginal increases in the probability to innovate when passing from 0 to 1 institutional sphere, we also obtain important increments when passing from 3 to 4. Institutional coherence pushes the capability of an establishment to perform innovations, and this also happens in the case of large exporting high-technology manufacturing establishments. This means that coordinated institutions are critical for innovation, but also that it is worthwhile even for most advanced establishments to be a VoC establishment. Innovation is not just a matter of industry requirements or the individual features of firm, but of institutional settings. The obtained results support the idea that institutions endow firms with technological capabilities and comparative advantages, but in Germany this happens in all industry clusters and promotes a wide array of innovative activities. Therefore, HII is rejected too. These results also support the rejection of HI.

Summing up, our results not only emphasize the importance of coordinated institutions for innovative performance, but also point to its versatility. Since they drive several forms of innovation, they are able to canalize the interaction among agents to produce different positive economic outcomes. In this sense, more evidence is needed to determine whether German VoC firms are more prone to radically innovate in comparison to ideal Anglo-Saxon firms in LMEs, and whether incremental innovation is also predominantly performed by the latter in these economies.

Concluding remarks

This paper researched the link between institutions and innovation drawing on the VoC paradigm by Hall and Soskice (2001). The theory states that country-specific institutions, when fully complementary with each other, foster certain types of innovation and endow firms with comparative advantages to compete in some industries. Broadly speaking, firms embedded in CMEs are more likely to incrementally innovate and are relatively weak in performing radical innovations, while the contrary occurs in LMEs.

In the context of this debate and focusing on the German manufacturing sector, this investigation has examined some of the VoC’s claims using the IAB Establishment Panel. Concretely, two goals have been addressed: (1) we explored the relationship between CME institutional coherence and the different types of innovation, and (2) estimated the effect of institutions when taking into account other structural determinants of innovation, such us firm size, the export status or the industry (Cohen and Kepler 1996; Chandy and Tellis 2000).

On the basis of logistic regressions two hypotheses have been tested. First (HI), it was expected that incremental and process innovation would be boosted by the combination of the four coordinated institutions, while radical innovation would be harmed by them. Nonetheless, we found that all types of innovation are favoured by coordinated institutions, and that the importance of institutional coherence only works in one direction in Germany: all types of innovation are favoured by the CME coherence; therefore, it is when establishments undertake a “non-market corporate strategy” when the probability to innovate is maximized.

Second, HII stated that the effect of institutions on innovation would be weaker in the presence of certain structural characteristics of a firm. Nonetheless, our results revealed that marginal effects of VoC establishments on innovation are the highest in all categories of the structural variables. Furthermore, when looking at the most advanced establishments in the economy—those that are large-sized, high exporters and located in MHT and HT industries—which are usually innovation leaders, the relationship remained the same. Therefore, the importance of coordinated institutions to undertake innovations is great even for advanced establishments and it is worthwhile for them to be a VoC establishment.

Our findings challenge the main statements of the VoC paradigm and point to its need for reform. First of all, descriptive evidence showed that there is not one single competitive strategy in an economy. Thus, firms are not passive institutional takers, but can select among a wide array of competitive strategies to organize their capabilities and interact with their economic environment. In our sample, less than 20% of establishments pursue a coordinated strategy, while 20% of them adopt a fully market corporate strategy. The relationship found between both types of institutional coherences and innovation patterns is linear: the more coordinated institutional spheres are present in a firm, the greater the probability to innovate. The same institutional combination that encourages incremental and process innovation is also beneficial for radical innovation. On the contrary, a coherent liberal strategy is detrimental to innovativeness. Put differently, the main road to innovation in Germany is a non-market corporate strategy, based on institutional mechanisms that promote stable commitments and trusting relationships among economic actors.

Our results are coherent with Herrmann (2008) and the strand of literature that empirically tested the relationship between labour institutions and innovation using the IAB Establishment Panel (Addison et al. 2017a, b). However, future research might be focused on offering a theoretical explanation of the relationship between radical innovation and coordinated institutions found in this paper, and exploring the channel by which institutional constraints support the development of new products or the application of new production processes. It seems that these institutions are more versatile than VoC expected, and are able to coordinate agents to promote a wide array of positive economic results. Another research line might explore the same relationship in other LMEs. It might be the case that in these economies, a full-market corporate strategy would be the most adequate one to undertake all types of innovation, and, in general, a concrete strategy is always the best one in an economy, no matter the sort of innovation the establishment would like to carry out.

Finally, it is worth mentioning the consequence of this paper’s results for policy-making: to keep up the pace of innovation and aggregate technical progress, social agents and political authorities must preserve and develop institutional tools to promote social partnership and the participation of the economic actors that are part of the firm in the decision-taking process. Nonetheless, descriptive evidence has shown that VoC establishments are, in fact, a minority in the German manufacturing sector, and it is a consensus that their number has been decreasing over time due to the erosion of the Modell Deutschland (Streeck 2009; Thelen 2014). This leads to a somewhat contradictory dynamic in the German economy: the most innovative firms, which are supposed to lead the technical progress and productivity growth, account for a diminishing proportion of the total firm population. This means that, in absence of policies that protect coordinated institutions, to keep up the pace of aggregate technical progress, a greater proportion of it must be undertaken by a smaller amount of firms, unless another efficient institutional combination is found.