How do institutional innovation systems affect open innovation?

ABSTRACT

It is well known that innovation benefits firms and that openness may enhance these benefits. Yet few studies consider how a firm’s institutional context and different economic systems moderate openness and innovation outcomes in new ventures, which arguably are most exposed to institutional constraints. Comparing data from a liberal (Australia) and central market economy (China), and using search breadth as a measure of openness, we empirically tested the influence of external firm pressures on the relationship between openness and innovation outcomes in new ventures. Our results show that the well-established positive relationship between openness and innovation is market dependent and that, in emerging economies, it can be negative. Our findings demonstrate the importance of institutions and national economic systems in explaining open innovation in different contexts – a point not yet addressed in the open innovation literature.

KEYWORDS:

• Open innovation
• new ventures
• institutions
• business system
• diversity of economic systems

Introduction

The importance of external sources of knowledge to innovation processes is well documented (for example, Allen & Cohen, 1969; Cohen & Levinthal, 1990). More recently, the open innovation literature (Chesbrough, 2006) has explained how firms search for external knowledge to be used within the innovation process (Lee et al., 2010) and disseminate knowledge to external agents (Alexy et al., 2013; West et al., 2014), also known as “inbound and outbound open innovation” (Dahlander & Gann, 2010). While openness – the use of “a wide range of external actors and sources to help [firms] achieve and sustain innovation” (Laursen & Salter, 2006, p. 131) – is key to open innovation, it is affected by the surrounding context of the national innovation system. So far, the open innovation literature has focused on a firm level of analysis, with the assumption that the focal firm is the main agent of managerial decisions and that it interacts with other firms or individuals. Consequently, the literature falls into categories of inter-, intra-, and extraorganizational open innovation (Bogers et al., 2017). Few studies have considered industry-level effects (West & Bogers, 2017), and linking national innovation systems to open innovation in firms is extremely rare.

However, and despite the general lack of recognition in open innovation studies, it is now well established that government policies can alter innovation systems¹ by using mechanisms such as funding for innovative ideas, fair and transparent competition, well-defined and enforceable intellectual property laws, healthy interfirm relations, knowledge transfer between research and development (R&D) centers and industry, and open labor laws (Audretsch & Fritsch, 2002; Storz et al., 2015). Prior research recognizes that institutional pressures that shape national innovation system components do matter, and produce different innovation system arrangements that, in turn, foster alternative innovation patterns and outcomes (Casper & Whitley, 2004). Institutional pressures also shape the extent of the external knowledge search that firms can undertake (Laursen & Salter, 2006) – also known as openness – and how this knowledge can be captured for innovation (Levinthal & March, 1981). However, to our knowledge. the literature has not explored the connection between open innovation and national innovation systems and their institutions.

This disconnect between the open innovation literature and national innovation systems literature leads us, like others (Huizingh, 2011), to question: How do national innovation systems affect the search and capture of external knowledge of the focal firm and how does this influence innovation? To answer this research question, we draw from two fields of study that provide insights into the institutional context that influences the relationship between openness and innovation outcomes. First, institutional economics (Lundvall, 2010; North, 1991; Scott & Davis, 2007) and the diversity of economic systems frameworks (Aoki, 2001; Hall & Soskice, 2001; Whitley, 1999) have established the importance of external constraints for firms. These external forces differ depending on the economic systems in which they are based because the institutional context is country specific (Whitley, 1999). Second, we draw on open innovation literature and use the open innovation framework by focusing on the variance hypothesis and the dependency between diverse sources of information and innovation (Katila & Ahuja, 2002; Nelson, 1982; Powell et al., 1996). Furthermore, we are particularly interested in answering this question in the context of new ventures because of their higher exposure to country institutional pressures, and resource and capability constraints (Sapienza et al., 2006), which lead them to frequently embrace open innovation activities (Lee et al., 2010; Van de Vrande et al., 2009). Indeed, new ventures are able to fully embrace open innovation activities (McDougall et al., 2003) due to their flexibility and adaptability to external ideas when compared with larger firms (Dufour & Son, 2015; Parida et al., 2012) and also due to their need to supplement limited internal resources (Verreynne et al., 2016). Moreover, new ventures are also more exposed to pressures from institutional factors when compared to medium and large enterprises, due to their size and lack of power to influence such institutional pressures (Davidsson et al., 2006).

To understand the relationships between important firm-level components of innovation systems and how they are impacted by institutional factors, a cross-country comparison is required. As China is the largest developing economy and the second-largest economy in the world, we use this economy in our study. Considering that the Chinese economy has its foundations in authoritarian law and a centrally planned economy, it would not be surprising to find that a liberal market economy (LME) (Hall & Soskice, 2001) would have contrasting institutional characteristics (Aoki, 2001). To limit regional effects, we looked to the region for comparison and found that the country that presents such liberal market characteristics is Australia.

By bridging national economic systems with the innovation literature, specifically in new ventures, this article has two aims. First, it aims to highlight how national innovation systems affect the relationship between openness (measured through search breadth) and innovation outcomes at the organizational level. Second, it aims to explore how these external institutional pressures differ in central and liberal market economies. Using data from surveys conducted in China and Australia among new venture companies, we analyze how openness influences innovation outcomes in each country and how different innovation system components – namely, government and private funds, R&D cooperation, and training – moderate this relationship.

We make three contributions. First, on the basis of our data, we propose that openness is not always positively related to innovation outcomes, which contrasts with findings from prior literature that has primarily focused on developed economies and large firms while being silent on new ventures from emerging economies. Second, we introduce an institutional factor to the well-established relationship between openness and innovation outcomes by using the external components of the diversity of economic systems framework (Hall & Soskice, 2001) as moderators. By doing so, we introduce an institutional-level unit of analysis to the open innovation literature that has so far focused on individual, firm, and industry levels while perceiving institutions as a constant. Third, this article empirically explains the impact that different institutional arrangements have on a central and liberal market economic system. Therefore, our results bring clarity to the effect of different economic systems on openness and innovation outcomes more generally.

Theoretical background

The innovation literature (Nelson, 1982), and more recently the open innovation literature (Chesbrough, 2006), recognizes that external knowledge, and in particular the number of different sources involved, can be beneficial to firms’ innovation outcomes (Katila & Ahuja, 2002; Laursen & Salter, 2006; Von Hippel, 1988). Concomitantly, past innovation literature (Cohen & Levinthal, 1990; Edquist, 1997; Lundvall, 1988b) has recognized the interdependence between firms’ innovation processes and the national-level innovation system. However, the open innovation literature assumes the process to be institutional-context free, or, at most, predominantly pays attention to the more generic environmental uncertainty concept (Huizingh, 2011).² Laursen and Salter (2006, p. 134), for example, explain that the firm’s set of different knowledge sources is “partly shaped by the external environment”; however, they do not consider the external environment in their research. We address this shortcoming by drawing on the economic systems literature.

Building on institutional economics, the diversity of economic systems framework explains how institutional contexts build national innovation systems and how, in turn, that impacts on firm-level innovation outcomes (Amable, 2003). A central argument of the economic systems literature relates to the complementarity of institutions within a country and how this complementarity can support firm performance; namely, through innovation outcomes (for example, Hall & Soskice, 2001). The varieties of capitalism literature (Hall & Soskice, 2001) and the business system framework (Whitley, 1999), also known as the “diversity of economic systems” (Amable, 2003; Storz et al., 2015), argue that institutions’ structuring of economic relationships can be systematically clustered based on their institutional complementarities. Moreover, this literature explains that due to institutional contexts, innovation systems vary across countries (Lundvall, 1988b; Whitley, 1999). Accordingly, these systems have different effects on firms’ innovation processes (Casper & Whitley, 2004). For this reason, the diversity of economic system literature offers an opportunity to introduce institutional elements into the firm-level analysis of innovation processes by understanding coordination problems between firms and other economic actors (Storz et al., 2013).

Open innovation and innovation outcomes

The contemporary strategy literature sees a firm’s capacity to innovate as the main source of sustainable advantage, with firms residing within a broader network of knowledge flows (Beinhocker, 2007; Eisenhardt & Martin, 2000; Teece, 2017). The open innovation literature recognizes the importance of these knowledge flows across organizational boundaries (Bogers et al., 2017; Chesbrough, 2006, 2012; Dahlander & Gann, 2010). An important aspect of this use of external knowledge relates to the number of sources from which the firm captures external knowledge (March, 1991; Powell et al., 1996; Von Hippel, 1988), which is referred to in the literature as the variance hypothesis or, more recently, as search breadth (Laursen & Salter, 2006; Terjesen & Patel, 2017) or openness³ (Gentile-Lüdecke et al., 2019; Verreynne et al., 2020). In this article, we are particularly interested in the external search aspect of open innovation (Dahlander & Gann, 2010) because, contrary to outbound forms of open innovation that are controlled by the firm, this type of knowledge is more affected by external pressures.

The extant literature on open innovation has established, particularly in the developed economies context, the positive relationship between openness and innovation outcomes (Katila & Ahuja, 2002; Leiponen & Helfat, 2010). While some have found a curvilinear effect, indicating a potential oversearch (Laursen & Salter, 2006), the positive relationship between openness and innovation outcomes has been shown in various developed economy contexts, such as the United Kingdom, Taiwan, and Korea (for example, Chiang & Hung, 2010; Garriga et al., 2013; Lee et al., 2010).

However, the search for innovation is contingent “on the individuals who stand at the interface of … the firm and the external environment” (Cohen & Levinthal, 1990, p. 132). Accordingly, for firms to innovate, employees need to pay attention to the different sources of external knowledge while still paying attention to internal needs so those external ideas are contextualized and relevant to the focal firm (Dahlander et al., 2016). Search breath requires managing the various external sources of knowledge in parallel with managing the firm’s internal pressures and resource constraints, which implies that staff need to be able to effectively cope with high levels of complexity (Almirall & Casadesus-Masanell, 2010). These arguments indicate that openness may impact innovation outcomes differently in emerging economies because of the different knowledge options available (Kafouros & Forsans, 2012) and due to variance in staff internal capabilities; namely, complexity (Redding, 2002; Redding & Witt, 2009). The education system is frequently referred to as one of the main reasons why in emerging economies in general, and in centrally managed economies in particular, staff have difficulties coping with complexity (Witt & Redding, 2013). Education systems that do not focus on developing divergent thinking skills are associated with lower levels of complexity management (Redding, 2002; Redding & Witt, 2009). Thus, we expect that when Chinese firms engage in broader external search activities, difficulties in assimilating different sources of knowledge will bring disruptions, and thus impact negatively on innovation.

Past research on open innovation in China seems to support this negative impact, as more than 75 percent of firms still innovate using exclusively in-house resources (Fu & Xiong, 2011), and firms consider internal resources as the most important sources for innovation (Xiong et al., 2011). This is in contrast to Australia, where internal and external sources are balanced (Verreynne et al., 2020). Moreover, Gentile-Lüdecke et al. (2019) report that when Chinese firms are successful with outbound open innovation, they tend to use a small number of sources. Accordingly, we hypothesize that:

Hypothesis 1 (H1): Openness is positively related to innovation in Australia, but negatively related to innovation in China.

National innovation systems

The European Union Lisbon Agenda (EU European Council, 2000) and the Technological Change and Innovation policies from the Organisation for Economic Co-operation and Development (OECD, 1994) are examples of policy statements that explain the incentives of a modern innovation system and how it influences the growth of economies. Despite general agreement about the importance of well-functioning innovation systems reflected in policy statements and academic research (for example, Colombelli et al., 2014; Criaco et al., 2014; Grilli & Murtinu, 2014; Rasmussen et al., 2014), different economies follow strategies to resolve coordination difficulties. These strategies align with the nature of labor and capital markets, intellectual property laws, incentives for interfirm and research-industry relations, and the formation of skills (Aoki, 2001; Casper & Whitley, 2004; Whitley, 1999), which vary among different economies and are well studied in economic systems literature.

Hall and Soskice (2001) and Whitley (1999) explain the five institutional domains that resolve coordination difficulties. Of those five, three are external to the firm and are thus influenced by local institutions, while at the same time they influence external sources of knowledge; namely, interfirm relationships (R&D cooperation⁴ ), financial systems, and the labor market.

R&D cooperation and the influence on openness

Formal R&D cooperation activities are a source of external knowledge and link the firm to other sources of information for innovation through a group of common incentives (Becker & Dietz, 2004). Such formal relationships are considered to be a source of explicit (Chen, 2004) and tacit knowledge (Janowicz-Panjaitan & Noorderhaven, 2009). R&D cooperation is an important aspect of any economic system, but the open innovation literature classifies it as “acquiring” because of its formal nature (Dahlander & Gann, 2010). However, past research has established how formal R&D cooperation with competitors (Belderbos, Carree, Diederen et al., 2004), suppliers (Becker & Dietz, 2004), public research organizations (Fontana et al., 2006), and customers (Von Hippel, 1988) influences innovation outcomes – both incremental (Tödtling & Trippl, 2005) and radical (Belderbos, Carree, Lokshin, et al., 2004). Moreover, from the perspective of diverse economic systems, some economic systems incentivize formal relationships with other firms and even competitors, whereas others do not (Whitley, 1999).

From an open innovation perspective, the fact that firms are in formal R&D partnerships leads us to argue that, as a consequence, they have reduced capacity to source and integrate other forms of external knowledge due to managerial attention being focused on formal collaborations (Ford et al., 2017) – known as the “span of attention” (Simon, 1947). In this scenario, the costs associated with sourcing further external information exceed the benefits associated with openness (Ocasio, 1997). Yet the literature indicates that independent associations are one of the most important features that enhance R&D cooperation, German being a case in point (Hall & Soskice, 2001; Whitley, 1999). In many economies, such independent institutions provide important support for R&D cooperation, particularly when formal contracts have low enforceability, as in the case of emerging countries. In China, for example, independent institutions have limited growth opportunities due to the risks they pose to the government; an independent central bank, professional associations, and free press are absent (Redding & Witt, 2007). Further, the lack of trust in contracts, which is a result of poor enforcement of the rule of law (Peerenboom, 2002), does not incentivize R&D cooperation based on formal market relationships (Torres de Oliveira & Figueira, 2018). Thus, economic agents do not trust interfirm strategic interactions because there are no industrial associations or other institutions to support, regulate, and incentivize them (Redding & Witt, 2009). However, Chinese firms do collaborate; they do so by minimizing risk sharing, sharing product or service development, exchanging know-how (Witt & Redding, 2013), and when strong personal connections exist (Park & Luo, 2001). This situation presents an obstacle for economic development because it does not allow for efficient resource allocation and limits the usefulness of external sources of knowledge. Based on the above arguments, we expect that when formal R&D cooperation happens in China, it is based on strong personal connections and not primarily reliant on market rules, which act as an extra inhibitor to the span-of-attention constraint. Accordingly, R&D cooperation negatively moderates the relationship between openness and innovation outcomes in emerging economies such as China.

Conversely, in Australia, enforcement of the rule of law is high (Porta et al., 1999) and, like other LME contexts such as the United States or the UK, the large majority of formal R&D cooperation transactions have a market and contract basis (Hall & Soskice, 2001). For this reason, Australian firms cope better with their span of attention because the R&D cooperation occurs within the boundaries of well-developed institutions. This affords the ability to keep looking to other external sources of knowledge. However, and contrary to central market economies (CMEs) where independent institutions play a critical role in supporting interfirm relationships, the lack of strong industry or worker associations supports strong market competition that is regulated by macroeconomic policies (Amable, 2003). Thus, even if firms are able to expand their focus beyond existing R&D cooperation, it is contradicted by strong market competition. Accordingly, we argue that these forces do not significantly influence the relationship between openness and innovation. Thus, we hypothesize that:

Hypothesis 2 (H2): R&D cooperation negatively moderates the relationship between openness and innovation in China, but does not have a moderating influence in Australia.

External capital and the influence on openness

Research confirms a relationship between externally sourced capital and innovation outcomes (for example, Bertoni & Tykvová, 2015; Block et al., 2018), even in the presence of obstacles such as demand-side factors (Pellegrino & Savona, 2017). External capital can be sourced from government or private institutions. However, government funds are usually provided to correct inefficiencies in the market (Boadway & Tremblay, 2005) and support higher risk-bearing ventures (Arrow, 1962). In contrast, private funds, such as those from venture capitalists, commercial banks, high-net-worth individuals, or other private organizations, are known to involve greater rationality in analyzing risks and returns (Reid & Smith, 2007). Furthermore, private investment usually entails a lower asymmetry of information and a superior influence over the firm’s strategies and operations (Peneder, 2008). This distinction is particularly important when we study CMEs and LMEs given the difference between government and private capital support in the economy in general (Wade, 1990). We suggest that external funding improves firms’ ability to convert external knowledge into innovation, regardless of whether these are derived from private or public sources.

For example, the Chinese government is not known to share risks with the private sector directly and, therefore, prefers to support the state-owned sector (Redding, 2002; Witt & Redding, 2013). Government venture capital is scarce and difficult to obtain, with the only firms able to access it being those with high levels of official connections (Ahlstrom et al., 2007). However, in recent years, the Chinese government has indirectly supported Chinese private firms, particularly new ventures, by funding infrastructure such as university-funded science parks and regional science and technology parks (Zhang, 2014) in China. Another incentive for Chinese firms to associate with technological or university parks relates to the network and ecosystem that are known to foster open innovation (Padilla-Meléndez et al., 2013). Since government funds in China are coupled with a firm’s presence in technological parks (Zhang, 2014), we hypothesize that there is a positive moderation effect of governmental funding in the relationship between external sources of knowledge and innovation for new ventures.

In Australia, government financial support for innovation and new ventures is low – one of the lowest in the OECD (OECD, 2017). Since the 1980s, successive Australian governments appear to have followed financial policies that encourage the financial sector to provide capital to firms (Heyes et al., 2012), which is consistent with liberal economic policies (Lazonick, 2010). Indeed, the LME system assumes that the market and its private economic agents will identify and support feasible new ventures (Hall & Soskice, 2001). However, when firms receive government funds, their pressure to deliver results is lower when compared with private funds (Audretsch et al., 2008; Colombo et al., 2016), and they are therefore more open to external knowledge sources. As such, we argue that government financial support positively moderates the relationship between openness and innovation, particularly for new ventures, hypothesizing that:

Hypothesis 3a (H3a): Government funding positively moderates the relationship between openness and innovation in China and in Australia.

The commercial financial sector in China is underdeveloped (Johnson et al., 2002; Lei, 2012) and not independent because financial institutions are expected to fully cooperate with the central government (Lin, 2011). The system is imperfect (Buckley et al., 2008; Child & Rodrigues, 2005) and known to suffer from a lack of expertise in the analysis of investment projects (Witt & Redding, 2013). However, China’s high rate of savings from firms and the population (Kuijs, 2006) provides Chinese entrepreneurs and private firms access to capital through noninstitutional markets, known as shadow banking.⁵ Such channels are particularly important for private firms and are closely tied to trust, reputation, and relationships. Indeed, some researchers argue that these alternative channels⁶ finance and support the real economy and, therefore, economic growth (Allen et al., 2008). Shadow banking was estimated to represent approximately 50 percent or more of the Chinese gross domestic product in 2012 (Li, 2014). Venture capital in China is still in a developmental phase and not able to guarantee a rational allocation of capital and returns, making it particularly difficult for small private entrepreneurs to access funds (White et al., 2005). Thus, the main source of capital in China remains shadow banking. Because shadow banking is closely tied with trust and relationships (Allen et al., 2019), it involves low levels of risk assessment, which in turn implies that firms do not need to have well-founded ideas of what to develop. Without a well-developed idea underpinning funding, we argue that entrepreneurs are open to different sources of external knowledge, and thus private funds positively impact the relationship between openness and innovation outcomes.

In Australia, the commercial and private financial sector is characterized as one of the four most liberalized sectors among OECD countries next to the United States, the UK, and Canada (Höpner et al., 2009). The changes in the Australian financial sector in the 1970s and 1980s have helped liberalization by ensuring the removal of controls as well as adopting an open market management of the financial sector by the Federal Reserve Bank (Konzelmann et al., 2012). This approach has improved efficiency in competition and accessibility to financial resources, which in turn have positively impacted the country’s economic growth (Thangavelu & Jiunn, 2004). However, unlike the United States and the UK, but similar to Canada, Australia has ensured a balance between liberalization and regulation, and the state has sustained its strong position throughout this development (Konzelmann et al., 2012). Yet Australia’s financial private sector is known for undertaking detailed risk analysis (Baccarini & Archer, 2001). In addition, private funders such as venture capitalists and “angels” are very involved with the firms they fund, again limiting firms’ attention span. Because of their risk aversion, firms need to clearly articulate their R&D projects to be funded, which is contrary to the idea of incorporating knowledge from external sources that the firm is not in control of; thus, we hypothesize that:

Hypothesis 3b (H3b): Private funds positively moderate the relationship between openness and innovation in China and negatively in Australia.

Labor markets and their influence on openness

From an open innovation perspective, a firm’s ability to look for and capture external knowledge will depend on the boundary spanners between the firm and external agents. If those individuals have limited skills or capacity to search for different external sources of ideas and capture them, training can help to enhance such capabilities. However, the incentives for firms to invest in training are different across CMEs and LMEs. Contrary to firms in CMEs, where there is a high incentive to employee trainees due to the low mobility of workers, liberal and product-imitation market firms will not have such an incentive to invest in training workers since this type of investment can easily be lost due to high turnover, and the results will be enjoyed by competitors (Herrmann & Peine, 2011).

Not surprisingly, the Chinese labor market is volatile, with workers changing jobs on a regular basis (Torres de Oliveira & Figueira, 2018). Firms can, to some degree,⁷ easily dismiss workers based on the current rule of law.⁸ Firms expect that workers have the skills required to perform their tasks (Witt & Redding, 2013). In societies like China, the education system is responsible for developing such skills. This means that jobs in China are planned not to depend highly on people’s skills and originality, which leads to a kind of Fordism with the mechanization of tasks, control, and low levels of specialization (Redding & Witt, 2007).

In Australia, similar to the United States, Canada, and the UK (LMEs), the prevalence of traditional apprenticeships has decreased significantly. Students are strongly orientated toward university degrees and not vocational training. This is due to weakening relationships between employers and unions, as well as state regulations (Bosch & Charest, 2008) arising from the liberalization reforms in the 1980s and 1990s (Konzelmann et al., 2012; Thangavelu & Jiunn, 2004). This decline in trades has resulted in more university graduates and less specialized trainees, even though in recent years this trend has been mitigated (Steedman, 2010). Moreover, shortages of tradespeople and workers with vocational skills in Australia (Bolton, 2018) confirm the inefficiency of investments in the education system, which is typical of LMEs (Hall & Soskice, 2001). As such, the Australian labor market is also volatile, with skilled individuals frequently changing employers.

However, training improves firms’ knowledge capital, which is critical to openness (Garriga et al., 2013) because it improves the absorptive capacity of firms (Aldieri et al., 2019, 2018; Cohen & Levinthal, 1989, 1990; Laursen & Salter, 2006). Fluid labor market laws and practices that Australia and China exhibit could also enable training to increase turnover (Herrmann & Peine, 2011), but this only occurs following absorptive capacity and thus results in an improved ability to search for different sources of external knowledge. Against this backdrop, past research explains that having highly skilled staff is particularly important for innovation in new ventures (March, 1991). Furthermore, past research indicates that inadequate management of employees in new ventures is one of the leading reasons for failure (McEvoy, 1984) and, due to new ventures’ smallness and entrepreneurial environment, managers will have the incentives to manage staff adequately and efficiently to decrease turnover (Appelbaum & Kamal, 2000; Messersmith & Guthrie, 2010). We argue that this is even more important in China, where the education system is in an earlier stage of development than in Australia, and thus workers will need to be upskilled to be able to search for and capture external knowledge. Therefore, we argue that:

Hypothesis 4 (H4): Training positively moderates the relationship between openness and innovation, and this holds to be true in both Australia and China.

Data and method

Sample

Since we are interested in understanding how national innovation systems affect firms’ search and capture of external knowledge, we wanted to use distinctive economic systems. Given the increasing importance and the lack of research on open innovation in emerging markets (Fu et al., 2014), we chose China with its authoritarian and centrally planned economy. We then, similar to others (Aoki, 2001), identified an LME to contrast our findings. Australia is the closest LME, in geographical and international trade terms, to China and, thus, served as an ideal candidate in our comparative study.

To test our hypotheses, we used two survey datasets, one collected in Australia and one in China during 2014 and 2015. The Australian survey was distributed randomly to 16,875 firms by mail, with one owner/senior executive (CEO, managing director, general manager, etc.) from each firm asked to complete the questionnaire. Of these, 1,566 firms completed the survey for an overall response rate of 9.3 percent. The survey response rate was similar to other studies on innovation – for example, in Scotland (Freel, 2005, 10.0 to 11.5 percent) and South Africa (Oerlemans et al., 2004, 8.4 percent) – and further tests for bias are presented in the Empirical analysis section of this article.

Data collection in China was accessed through a group of officials who approved the study and introduced it to companies operating in Jiangsu in 2015. This facilitated approach allowed us to improve our response rate and to shorten the data collection period. The surveys were delivered personally to senior managers of 180 firms, randomly selected with agreement from the facilitating officials. The survey was written in English and translated into Chinese. To ensure accuracy, we use a back-translation method (Brislin, 1970), using two different Chinese translators to ensure validity. Out of the 180 distributed surveys, we received 137 usable responses.

Both surveys included questions typically asked in community innovation surveys (CISs), described in the Oslo Manual (OECD, 2005) and widely used in publications (for example, Cosh et al., 2012; Laursen & Salter, 2006). Each survey was adapted to suit the local language and terminology. While the Australian survey covered firms across the age spectrum, the Chinese survey focused on new ventures. We therefore extracted firms six years or younger (McDougall et al., 2003) from the Australian data to match the Chinese sample since we were particularly interested in understanding open innovation in new ventures. After data cleaning and verification, 119 Chinese and 223 Australian firms remained in the database. The sample of 223 firms was representative of the Australian business sector, and we did not find any systematic differences (except for age) between the overall Australian dataset and our extracted one in terms of industries or sales turnover (Australian Bureau of Statistics [ABS], 2018).

Measures

Dependent variable

The dependent variable in the study is innovation outcome.⁹ In the surveys, firms were asked whether they had introduced new or significantly improved products or services during the past three years, following the well-established measure of innovation (for example, Golovko & Valentini, 2011, 2014; Hottenrott & Lopes‐Bento, 2016; Hsieh et al., 2018; Nieto & Rodríguez, 2011).

Independent variables

Openness was measured as search breadth following the procedure suggested by Laursen and Salter (2006). In both surveys, firms were asked about the use of nine external sources of information for their innovation activities over the past three years; namely, suppliers of equipment/materials, customers, competitors, consultant firms, universities/higher education institutes, government agencies, trade fairs, scientific journals/publications, and professional/industry associations. When firms indicated that their use of a knowledge source was important or very important (points 4 and 5 on a 5-point Likert scale) to the introduction of innovation, the source was assigned a value of 1 (otherwise 0) (Laursen & Salter, 2006). The search breadth variable was then calculated by summing these values, creating a value ranging from 0 to 9 (Laursen & Salter, 2006).

Following the diversity of economic systems framework (Hall & Soskice, 2001) and other researchers (for example, Lee et al., 2010), we used four non-firm-level (institutional-level) factors – namely, R&D cooperation, government funds, private funds, and training – as moderators. R&D cooperation was asked as a yes/no question (coded as 1/0) for: sharing R&D, joint purchase materials, and joint development products/services. We summed these responses to create our categorical moderator (ranging from 0 to 3). Government funds, private funds, and training were coded as 1 if used (otherwise 0). All these moderators were measured based on respondents’ answers to questions about R&D cooperation, access to government and private funds, and training provision during the past three years, which aligned with our innovation and openness variables.

Control variables

We included the following variables in the analysis to control for their impact on innovation: firm size, age, industry, ownership, market, and country. We considered firm size because prior research shows a potential positive impact of size on the relationship between openness and innovation (Josefy et al., 2015). We categorized firms into small and medium enterprises (SMEs) when employing fewer than 250 employees, and large when employing 250 employees or more (Rammer & Schubert, 2018). We included firm age in the regressions because older firms are more like to rely on prior experiences, while younger firms are more open to new knowledge (Harison & Koski, 2010). This control variable was measured by the number of years in business (Brunswicker & Vanhaverbeke, 2015). Because we applied a probit model to test our hypotheses, we did not log-transform this continuous variable. We also considered industry because the relationship between openness and innovation differs among industries (West & Bogers, 2014). In our study, industries were classified into manufacturing, service, health care, and information and technology. The three subsectors were then coded as dummy variables, of which manufacturing and service were integrated into the regressions, while “health care, and information and technology” was used as the reference category. We included ownership type in the analyses because family firms are less likely to engage in open innovation than their nonfamily counterparts (Chrisman et al., 2015). Family firms were coded 1 (otherwise 0). We considered the market for sales to be important to control since firms competing in international or larger markets are more likely to have obsolete products when compared with local markets due to the different levels of competition (Kang & Kang, 2009). Markets were local, national, and international, and each was coded as a dummy variable. Finally, we created a country variable to control for the differences between the two countries: 0 for Chinese firms and 1 for Australian firms. All variables are summarized in Table 1.

Table 1. Explanation of the variables used in the analysis.

Variables	References	Explanation
Dependent variable innovation	(OECD, 2005)	Whether a firm introduced a radically new or significantly improved product in the previous three years (1/0)
Independent variables
search_breadth	(Laursen & Salter, 2006)	Number of sources (from 0 to 9) that firms deemed to be important or very important for innovation
R&D cooperation	(Lee et al., 2010)	Number of research and development (R&D) cooperation activities (from 0 to 3)
government_funds	(Lee et al., 2010)	Whether firms receive financial support from government agencies (1/0)
private_funds	(Lee et al., 2010)	Whether firms receive financial support from commercial banks (1/0)
training	(Lee et al., 2010)	Whether firms engage in training activities (1/0)
Control variables small	(Rammer & Schubert, 2018)	If firms have fewer than 250 employees (1/0)
age	(Brunswicker & Vanhaverbeke, 2015)	Number of years in business
manufacturing	(West & Bogers, 2014)	If firms operate in manufacturing industries (1/0)
service	(West & Bogers, 2014)	If firms operate in service industries (1/0)
market_local	(Kang & Kang, 2009)	If firms sell products/services to local markets (1/0)
market_national	(Kang & Kang, 2009)	If firms sell products/services to national markets (1/0)
market_international	(Kang & Kang, 2009)	If firms sell to overseas markets (1/0)
family	(Chrisman et al., 2015)	If firms are family owned (1/0)
country		If firms are located in Australia or China (1/0)

Analytical methods

We applied probit regression with maximum likelihood estimation to test our hypotheses because our innovation dependent variable is binary (Wooldridge, 2010).

We hypothesized that there are differences between the two countries as a result of, for example, the development of business sectors, institutional frameworks, and society that impact on R&D cooperation, access to government and private funds, and training provision. For instance, Chinese firms are more likely to access government funds, and they are also more manufacturing oriented than Australian firms (see more details in Table 2). In addition, we considered that larger, older firms could be more likely to have more sourcing capabilities than their younger and smaller counterparts. Consequently, these situations may bias the estimation of the impact of these factors on the relationships between openness, R&D cooperation, access to government and private funds, and training provision and innovation. Therefore, we applied a four-step procedure to control for these potential issues in our sample.

Table 2. Descriptive statistics.

			Observations	Mean	SD
1	innovation	China	119	0.597	0.493
		Australia	223	0.645	0.477
		Sample	342	0.635	0.482
2	search_breadth	China	119	8.202	1.078
		Australia	223	6.798	2.589
		Sample	342	7.287	2.283
3	R&D cooperation	China	119	0.731	0.909
		Australia	223	0.475	0.924
		Sample	342	0.564	0.926
4	government_funds	China	119	0.672	0.471
		Australia	223	0.099	0.299
		Sample	342	0.298	0.458
5	private_funds	China	119	0.664	0.474
		Australia	223	0.175	0.381
		Sample	342	0.345	0.476
6	training	China	119	0.513	0.502
		Australia	223	0.170	0.377
		Sample	342	0.289	0.454
7	small	China	119	0.975	0.157
		Australia	223	0.839	0.369
		Sample	342	0.886	0.318
8	age	China	119	2.378	2.029
		Australia	223	3.570	1.851
		Sample	342	3.155	1.995
9	manufacturing	China	119	0.361	0.482
		Australia	223	0.215	0.412
		Sample	342	0.266	0.443
10	service	China	119	0.067	0.251
		Australia	223	0.619	0.487
		Sample	342	0.427	0.495
11	market_local	China	119	0.563	0.498
		Australia	223	0.933	0.251
		Sample	342	0.804	0.397
12	market_national	China	119	0.580	0.496
		Australia	223	0.574	0.496
		Sample	342	0.576	0.495
13	market_international	China	119	0.235	0.426
		Australia	223	0.408	0.493
		Sample	342	0.348	0.477
14	family	China	119	0.185	0.390
		Australia	223	0.117	0.322
		Sample	342	0.140	0.348
15	country	Sample	342	0.652	0.477

First, we applied nearest-neighbor matching to estimate the average treatment effect of engaging in R&D cooperation of a firm (using the “teffects nnmatch” command in Stata) (Rosenbaum & Rubin, 1983). The variables included in the estimation were size, age, industry, market, ownership, and country. However, engaging in R&D cooperation is influenced not only by these observed characteristics in Australian and Chinese firms, but also potentially by unobservable factors that might bias our estimation. Therefore, we ran balance tests and the results showed no statistical difference between the treatment (that is, firms having R&D cooperation) and control group (that is, firms not having R&D cooperation) for each covariate (Chang & Shim, 2015). The estimation of engaging in R&D cooperation was then generated following the matching. Second, we created a new search breadth variable that was mean centered, which means that we subtracted the value from the mean value of search breadth. This adjustment allowed us to deal effectively with the interaction between our continuous variable and binary variables (Aiken et al., 1991). Third, we created interaction variables by multiplying the estimated R&D cooperation (from the first step) and the new search breadth variable (from the second step). Fourth, we ran an instrument probit regression to estimate the moderation impact of R&D cooperation on the relationship between external sources of knowledge of firms and innovation (Cameron & Trivedi, 2005; Wooldridge, 2010). We included both search breadth and R&D cooperation in the same instrumental equation, while all control variables were incorporated in the main equation. We applied the same procedure to the remaining moderation tests for government funds, private funds, and training. For the main effect of search breadth on innovation, we executed the second and last steps, including only search breadth as an endogenous regressor in the instrumental variable probit regression.

To test the hypotheses in each country separately, we followed the same procedure, but did not include the country control variables in the regressions. Furthermore, we used average marginal effects to report the moderation impacts.

Empirical analysis

Descriptive analysis

The descriptive analysis in Table 2 shows that 64 percent of firms in our sample had introduced a new product or service in the prior three years. The proportion of Chinese firms that innovated is slightly lower than for Australian firms (60 percent and 65 percent, respectively). While these figures are significantly higher than the average number of innovative firms in Western economies (for example, when compared with the CIS), when we adjusted CIS data to account for new ventures only, the proportion of Australian firms innovating is similar to those reported in our study (ABS, 2017). Furthermore, Chinese firms were likely to access more sources of knowledge, on average 8.2 sources as compared to 6.8 (out of 9) by their Australian counterparts. The majority of the Chinese firms were SMEs and nonfamily enterprises. Approximately 43 percent of firms operated in service sectors, 27 percent in manufacturing, and 20 percent in information and technology/health care businesses. Most firms served local markets and only a third sold products/services to international markets.

As seen in the descriptive statistics (Table 2), there is a significant difference between Chinese and Australian firms in terms of engaging in R&D cooperation activities, access to government and private funds, and training. This was confirmed by the comparative supplementary tests (see the Bias, robustness checks, and supplementary tests subsection). On average, Chinese firms participate in 0.731 (out of 3) R&D cooperation activities, while the figure for their Australian counterparts is only 0.475. This is counterintuitive, but can be explained by Australia’s low scores for R&D cooperation – being 27th among OECD countries (Department of Innovation, Industry and Science [DIIS], 2017). Access to government funds (67 percent for China and 10 percent for Australia), private funds (66 percent for China and 18 percent for Australia), and training (51 percent for China and 17 percent for Australia) also differed, implying that Chinese firms are in a better funding position than those in Australia. The difference in access to government funds between the two countries is understandable, considering the significant support of Chinese government agencies for new ventures (Guan & Yam, 2015). The development of capital-intensive manufacturing industries may explain why Chinese firms also rely heavily on private funds (including commercial banks) for their development, as opposed to their service-oriented Australian counterparts. Chinese firms also tend to provide more training for their employees as a means to fill the gap between tertiary education and industry requirements, which is typical in developing countries (Vaaland & Ishengoma, 2016). This distinctiveness suggests that the impact of R&D cooperation, government and private funds, and training differs between the two countries.

Table 3 shows the significant relationships among innovation, openness, R&D cooperation, government funds, private funds, and training, suggesting possible impacts and interactions. However, there are several high values in the correlation matrix; for instance, R&D cooperation is highly correlated with training in Australian firms, thus raising concerns about multicollinearity. We first assessed this using variance inflation factor (VIF) tests. The results in Table 4 show all VIF values are less than the cutoff value of 5, indicating there are no major issues with multicollinearity (Hair, 2010). We then conducted the Farrar-Glauber multicollinearity chi-square test; the significance of this test indicated there were no major issues with multicollinearity (Farrar & Glauber, 1967).

Table 3. Correlation matrix.

			1	2	3	4	5	6	7	8	9	10	11	12	13	14
1	innovation
2	search_breadth	China	0.231*
		Australia	0.250***
		Sample	0.216***
3	R&D cooperation	China	0.239**	0.399***
		Australia	0.151*	0.085
		Sample	0.166**	0.195***
4	government_funds	China	0.009	0.084	−0.062
		Australia	0.177**	0.078	0.200**
		Sample	0.044	0.178**	0.176**
5	private_funds	China	−0.005	−0.173	−0.230*	0.223*
		Australia	0.185**	0.008	0.213**	0.046
		Sample	0.066	0.044	0.013	0.387***
6	training	China	0.261**	0.478***	0.535***	−0.036	−0.374***
		Australia	0.129	0.125	0.763***	0.210**	0.231**
		Sample	0.150**	0.277***	0.629***	0.274***	0.134*
7	small	China	−0.023	0.086	0.065	−0.112	−0.114	0.165
		Australia	−0.165*	−0.036	0.052	−0.018	−0.087	0.069
		Sample	−0.139*	0.029	0.080	0.093	0.028	0.148**
8	age	China	0.124	0.069	0.127	0.185	−0.061	0.130	0.106
		Australia	−0.018	−0.111	−0.081	−0.037	0.074	−0.029	−0.038
		Sample	0.047	−0.109*	−0.064	−0.116*	−0.120*	−0.067	−0.072
9	manufacturing	China	−0.059	0.031	−0.053	−0.071	−0.168	0.069	0.009	−0.226*
		Australia	0.082	0.114	0.130	−0.027	0.017	0.140*	−0.037	−0.085
		Sample	0.017	0.114*	0.069	0.056	0.022	0.156**	0.008	−0.179***
10	service	China	−0.258**	−0.103	−0.196	0.045	0.049	−0.141	0.043	0.224	−0.202
		Australia	−0.065	−0.109	−0.073	0.074	0.045	−0.086	0.057	0.106	−0.667***
		Sample	−0.057	−0.202***	−0.143**	−0.278***	−0.229***	−0.264***	−0.062	0.249***	−0.520***
11	market_local	China	−0.034	0.112	0.108	−0.038	0.055	−0.046	0.183*	−0.038	−0.043	−0.034
		Australia	0.031	0.094	0.067	0.089	−0.018	0.027	0.077	−0.087	0.054	−0.063
		Sample	0.023	−0.018	0.014	−0.258***	−0.200***	−0.172**	−0.015	0.071	−0.07	0.203***
12	market_national	China	0.098	0.054	0.050	0.204*	−0.029	0.090	0.080	−0.055	0.144	−0.179	−0.407***
		Australia	0.118	0.104	0.022	0.011	0.230**	0.149*	−0.058	0.007	0.098	0.015	−0.087
		Sample	0.111*	0.077	0.031	0.081	0.112*	0.117*	−0.029	−0.017	0.115*	−0.025	−0.215***
13	market_international_ international	China	0.173	0.005	0.114	−0.077	0.059	0.105	−0.164	0.040	−0.087	−0.070	0.090	−0.130
		Australia	0.162*	0.191**	0.077	0.093	0.122	0.085	−0.107	−0.084	0.009	0.051	−0.105	0.512***
		Sample	0.172**	0.099*	0.059***	−0.087	−0.001	0.021	−0.144**	0.009	−0.051	0.114*	0.067	0.304***
14	family	China	−0.050	−0.032	−0.339***	−0.036	−0.074	−0.099	0.077	−0.085	0.047	0.218*	0.114	−0.209*	0.093
		Australia	0.029	−0.087	0.084	−0.027	0.127	0.058	0.083	0.011	−0.02	0.026	0.042	0.059	0.011
		Sample	−0.008	−0.059	−0.065	0.031	0.079	0.021	0.092	−0.050	0.023	0.009	0.03	−0.045	0.023
15	country	Sample	0.057	−0.202***	−0.132*	−0.597***	−0.490***	−0.359***	−0.204***	0.281***	−0.158**	0.531***	0.444***	−0.006	0.173**	−0.094

Spearman correlation, *, **, *** denote p < 0.05, p < 0.01, and p < 0.001, respectively.

Table 4. Variance inflation factor (VIF) values.

	VIF
innovation	1.16
search_breadth	1.23
R&D cooperation	1.78
government_funds	1.64
private_funds	1.41
training	2.04
small	1.11
age	1.14
manufacturing	1.47
service	1.98
market_local	1.39
market_national	1.25
market_international	1.21
family	1.05
country	3.25

Results

Given that we were particularly interested in uncovering how institutional domains moderate the relationship between openness and innovation, we started by testing if openness is indeed related to innovation outcomes in the different settings. Table 5 (Model 1) shows that openness relates significantly to innovation in our sample. A one-unit increase in the number of openness elevates the probability of having new products/processes by 4.5 percent on average. This significant relationship holds for both Chinese and Australian firms (Models 2 and 3), but in different directions. While openness relates positively to the introduction of innovation in Australian firms, in Chinese firms it is negatively related to innovation. Therefore, H1 was accepted.

Table 5. Estimation of the impact of search breadth on innovation.

	Innovation: China	Innovation: Australia	Innovation (combined sample)
	(2)	(3)	(1)
Explanatory variables
search_breadth	−0.799**	0.307**	0.409***
	(0.079)	(0.041)	(0.045)
Control variables
small	0.566	−0.532	−0.419
age	0.086	0.038	0.052
family	0.123	0.238	0.100
manufacturing	−0.110	0.026	−0.164
service	−1.016	−0.021	−0.209
country			0.693
X^2	67.71***	28.15***	69.32***
Observation	119	223	342

Average marginal effects in parentheses. Instrumental variables were: small, manufacturing, service, market_local, market_national, and market_international. *, **, *** denote p < 0.05, p < 0.01, and p < 0.001, respectively.

The results in Tables 6-9 show the moderation impact of R&D cooperation, government funds, private funds, and training, respectively, on the main relationship. As shown in Table 6, engaging more in R&D cooperation activities did not significantly increase the positive impact of openness on innovation (Model 6). This nonsignificant moderation effect holds for Australian firms (Model 5). However, for Chinese firms, participating in more R&D cooperation activities significantly reduces the negative impact of openness on the probability of introducing innovation (Model 4) by 3.3 percent (this reduction is due to the main relationship being negative). Therefore, H2 was accepted.

Table 6. Estimation of the moderation impact of research and development (R&D) cooperation on the relationship between search breadth and innovation.

	Innovation: China	Innovation: Australia	Innovation (combined sample)
	(4)	(5)	(6)
Explanatory variables
search_breadth	1.040***	0.248	0.249
	(0.051)	(0.045)	(0.034)
R&D_cooperation	−0.771	−0.362	−0.717
	(−0.069)	(0.057)	(−0.002)
search_breadth*R&D_cooperation	0.568*	0.099	0.112
	(−0.033)	(0.013)	(−0.011)
Control variables
small	−1.128	−0.668**	−0.682**
age	−0.038	−0.012	−0.019
family	0.152	0.522	0.739
manufacturing	0.009	0.277	0.026
service	0.306	0.248	0.138
country			0.328
X^2	199.88***	23.76**	76.30***
Observation	119	223	342

Average marginal effects in parentheses. Instrumental variables were: small, manufacturing, service, market_local, market_national, and market_international. *, **, *** denote p < 0.05, p < 0.01, and p < 0.001, respectively.

Table 7. Estimation of the moderation impact of government funds on the relationship between search breadth and innovation.

	Innovation: China	Innovation: Australia	Innovation (combined sample)
	(7)	(8)	(9)
Explanatory variables
search_breadth	−1.404***	0.708**	0.934***
	(0.050)	(0.076)	(0.058)
government_funds	0.430	0.033	−0.371
	(0.075)	(0.174)	(0.070)
search_breadth*government_funds	−1.302***	0.663*	0.903***
	(−0.048)	(0.046)	(0.017)
Control variables
small	0.225	−0.403	−0.099
age	−0.002	−0.010	0.010
family	0.054	0.450	0.284
manufacturing	0.073	−0.048	−0.264
service	−0.422	−0.231	−0.490**
country			0.604***
X^2	122.85***	53.05***	148.41***
Observation	119	223	342

Average marginal effects in parentheses. Instrumental variables were: small, manufacturing, service, market_local, market_national, and market_international. *, **, *** denote p < 0.05, p < 0.01, and p < 0.001, respectively.

Table 8. Estimation of the moderation impact of private funds on the relationship between search breadth and innovation.

	Innovation: China	Innovation: Australia	Innovation (combined sample)
	(10)	(11)	(12)
Explanatory variables
search_breadth	−1.157***	0.341	−0.666***
	(0.072)	(0.055)	(0.044)
private_funds	0.165	−0.087	0.797
	(−0.015)	(0.247)	(0.102)
search_breadth*private_funds	−0.898*	0.270	−0.801***
	(−0.015)	(0.022)	(−0.003)
Control variables
small	0.398	−0.635	0.024
age	0.025	0.040	−0.105**
family	0.009	0.389	−0.323
manufacturing	−0.018	0.210	−0.162
service	−0.230	0.100	−0.254
country			−0.241
X^2	140.35***	24.24**	191.21***
Observation	119	223	342

Average marginal effects in parentheses. Instrumental variables were: small, manufacturing, service, market_local, market_national, and market_international. *, **, *** denote p < 0.05, p < 0.01, and p < 0.001, respectively.

Table 9. Estimation of the moderation impact of training on the relationship between search breadth and innovation.

	Innovation: China	Innovation: Australia	Innovation (combined sample)
	(13)	(14)	(15)
Explanatory variables
search_breadth	1.216***	−0.233	0.409
	(0.067)	(0.057)	(0.047)
training	−1.788*	−1.714**	−0.732
	(−0.157)	(0.060)	(−0.059)
search_breadth*training	1.401***	−0.390	0.348
	(0.014)	(0.039)	(0.002)
Control variables
small	−0.888	−0.476	−0.694*
age	−0.009	−0.111*	−0.016
family	0.189	0.158	0.247
manufacturing	0.028	0.323	−0.136
service	0.254	0.142	−0.130
country			0.461*
X^2	125.21***	68.76***	47.32***
Observation	119	223	342

Marginal effects. Instrumental variables were: small, manufacturing, service, market_local, market_national, and market_international. *, **, *** denote p < 0.05, p < 0.01, and p < 0.001, respectively.

Overall, the moderation of access to government funds is positive and significant (Model 9). While being negative and significant for Chinese firms (Model 7), it is positive and significant for Australian firms (Model 8). This finding means that while access to government funds increases the impact of openness on the probability of introducing innovation by 4.6 percent in Australian firms, it reduces the negative impact of openness on innovation in Chinese firms by 4.8 percent. Therefore, H3a was accepted.

Access to private funds significantly lessened the positive impact of openness on innovation (Model 12). In Chinese firms, private funds significantly moderate the relationship between openness and innovation (Model 10). Practically, access to private funds reduces the negative impact of openness on the probability of having new products/business processes by 1.5 percent on average. However, this moderation effect was not significant for Australian firms (Model 11). Thus, H3b was partially accepted.

Regarding the moderation impact of training, as shown in Model 15, training increases the influence of openness on innovation, but not in a statistically significant way. This moderation effect is positive and significant in Chinese firms, but not significant in Australian firms, which means that providing training can increase the negative impact of openness on innovation for Chinese firms by 1.4 percent (see Model 13 and Model 14). Consequently, H4 was rejected.

Bias, robustness checks, and supplementary tests

To avoid the issue of nonresponse bias, selection bias, and endogeneity, we applied different statistical techniques to deal with each. Specifically, we utilized a stratified sample and checked whether differences existed between data from the survey and the population to eliminate any nonresponse bias problem. We applied the statistical technique of matching to avoid sample selection bias. Harman’s single-factor test, the marker variable technique, and two datasets from China and Australia were used to avert any common method and common source biases. Endogeneity problems caused by unobservable factors were treated using three statistical techniques: matching to control for the difference between the two countries, a mean-centered approach, and an instrument variable method. Please refer to the Appendix for more details on each method.

To check the robustness of the results, we applied the same procedure, but replaced the main dependent variable with a new variable that came from a different question in the survey (where we asked whether the firm’s innovation was only new to the firm or new to the market/industry). By survey design, only firms that introduced innovation during the past three years responded to this question. We generated a variable based on this question – if a firm answered the question, it meant that it had innovated, and the new variable was assigned a value of 1 (otherwise 0). We also replaced the industry variables by creating weighted variables for various industries and included them in our regressions. The results displayed the same patterns as those in our main findings, indicating that our results are robust.

As previously mentioned in the descriptive analysis, there is a significant difference between Chinese and Australian firms. We therefore conducted t-tests to check if these differences were statistically significant. The results confirmed that Chinese firms are more likely than Australian firms to engage in R&D cooperation, access government and private funds, and provide training. The results support the differences in the institutions between the two countries, thereby explaining the significant moderation of these factors on the relationship between openness and innovation in Chinese versus Australian new venture firms. Furthermore, our data show that Chinese firms produce more radical innovation than Australian firms, but the opposite is true for incremental innovation (Tables 10, 11, 12, 13, 14, and 15).

Table 10. Comparison of research and development (R&D) cooperation in Chinese and Australian firms.

Group	Obs.	Mean	SE	SD	[95% Conf.	interval]
0	119	0.7311	0.0833	0.9085	0.5662	0.8960
1	223	0.4753	0.0619	0.92422	0.3534	0.5973
combined	342	0.5643	0.0500	0.9255	0.4659	0.6628
diff		0.2558	0.1043		0.0506	0.4609
diff = mean(0) – mean(1)					t = 2.4521
Ho: diff = 0					degrees of freedom = 340
Ha: diff < 0		Ha: diff ! = 0			Ha: diff > 0
Pr(T < t) = 0.9926		Pr(|T| > |t|) = 0.0147			Pr(T > t) = 0.0074

Table 11. Comparison of access to government funds in Chinese and Australian firms.

Group	Obs.	Mean	SE	SD	[95% Conf.	interval]
0	119	0.6723	0.0432	0.4714	0.5867	0.7578
1	223	0.0987	0.02	0.2989	0.0592	0.1381
combined	342	0.2982	0.0248	0.4582	0.2495	0.347
diff		0.5736	0.0418		0.4914	0.6558
diff = mean(0) – mean(1)					t = 13.7299
Ho: diff = 0					degrees of freedom = 340
Ha: diff < 0		Ha: diff ! = 0			Ha: diff > 0
Pr(T < t) = 1.0000		Pr(|T| > |t|) = 0.0000			Pr(T > t) = 0.0000

Table 12. Comparison of access to private funds in Chinese and Australian firms.

Group	Obs.	Mean	SE	SD	[95% Conf.	interval]
0	119	0.6639	0.0435	0.4744	0.5778	0.75
1	223	0.1749	0.0255	0.3807	0.1246	0.2251
combined	342	0.345	0.0257	0.4761	0.2944	0.3957
diff		0.489	0.0472		0.3962	0.5818
diff = mean(0) – mean(1)					t = 10.3633
Ho: diff = 0					degrees of freedom = 340
Ha: diff < 0		Ha: diff ! = 0			Ha: diff > 0
Pr(T < t) = 1.0000		Pr(|T| > |t|) = 0.0000			Pr(T > t) = 0.0000

Table 13. Comparison of training in Chinese and Australian firms.

Group	Obs.	Mean	SE	SD	[95% Conf.	interval]
0	119	0.5126	0.046	0.502	0.4215	0.6037
1	223	0.1704	0.0252	0.3768	0.1207	0.2201
combined	342	0.2895	0.0246	0.4542	0.2412	0.3378
diff		0.3422	0.0482		0.2474	0.437
diff = mean(0) – mean(1)					t = 7.1017
Ho: diff = 0					degrees of freedom = 340
Ha: diff < 0		Ha: diff ! = 0			Ha: diff > 0
Pr(T < t) = 1.0000		Pr(|T| > |t|) = 0.0000			Pr(T > t) = 0.0000

Table 14. Comparison of radical innovation in Chinese and Australian firms.

Group	Obs.	Mean	SE	SD	[95% Conf.	interval]
0	119	0.3782	0.0446	0.487	0.2897	0.4666
1	223	0.2601	0.0294	0.4397	0.2021	0.3181
combined	342	0.3012	0.0248	0.4594	0.2523	0.35
diff		0.1181	0.0518		0.0161	0.22
diff = mean(0) – mean(1)					t = 2.2774
Ho: diff = 0					degrees of freedom = 340
Ha: diff < 0		Ha: diff ! = 0			Ha: diff > 0
Pr(T < t) = 0.9883		Pr(|T| > |t|) = 0.0234			Pr(T > t) = 0.0117

Table 15. Comparison of incremental innovation in Chinese and Australian firms.

Group	Obs.	Mean	SE	SD	[95% Conf.	interval]
0	119	0.3675	0.0448	0.4842	0.2789	0.4562
1	223	0.5243	0.0349	0.5006	0.4555	0.593
combined	342	0.4675	0.0278	0.4997	0.4128	0.5222
diff		−0.1568	0.0573		−0.2694	−0.0441
diff = mean(0) – mean(1)					t = − 2.7368
Ho: diff = 0					degrees of freedom = 340
Ha: diff < 0		Ha: diff ! = 0			Ha: diff > 0
Pr(T < t) = 0.0033		Pr(|T| > |t|) = 0.0065			Pr(T > t) = 0.9967

Discussion and implications

An important finding from this study is that national differences in institutions matter for open innovation. Testing open innovation hypotheses with datasets involving firms from single countries may obscure important intercountry differences in open innovation. The aggregated data analyses revealed mixed findings regarding the role of different moderators in the relationship between openness, as a form of open innovation, and innovation outcomes. The presence of significant moderators when we analyze the countries individually means that open innovation models derived from aggregations of different countries will hide important national variations in open innovation. The absence of national-level analysis and institutional differences in the open innovation literature therefore needs to be revisited. Even the axiomatic assumption that openness is positively correlated with innovation is challenged by the results of the Chinese data.

Theoretical implications

Our results suggest that the attention “budget” (Ocasio, 1997) for external search activities is constrained in Chinese SMEs, and this results in a reduced ability to assimilate external knowledge (Liu & Woywode, 2013) and a lower ability to deal with complexity (Redding, 2002). It seems that the available options for external knowledge in China are not adequate to motivate firms to engage in these external search practices (Savitskaya et al., 2010). These results are consistent with literature that discusses the still underdeveloped national markets for technology in China (Savitskaya et al., 2010) and how open innovation strategies can indeed be problematic in developing economies (Fu & Xiong, 2011). From a transactions cost perspective, openness may be preferred when the markets for external knowledge are more efficient compared to organizing innovation within the firm (Dahlander & Gann, 2010). These results are particularly important to the innovation literature, as we limit – as others recently have done using different limits (Dahlander et al., 2016; Terjesen & Patel, 2017) – the “variance hypothesis” by explaining that absorptive capacity and coping with complexity is crucial but differs across contexts.

When considering the different moderators included in this study, our results support the hypothesis that R&D cooperation in China negatively influences the relationship between openness and innovation, which is aligned with past empirical research (Huang et al., 2015). The resource dependence theory (Pfeffer & Salancik, 1978) and the dynamics of Chinese society, where guanxi or networking is deeply rooted (Park & Luo, 2001), support these results and our arguments. Resource dependence theory states that an organization strategically shapes its cooperative relationships and arranges its resources and capabilities in response to environmental uncertainty (Pfeffer & Salancik, 1978; Van de Ven, 1976). This is particularly important in countries such as China where firms in general, and new ventures in particular, are exposed to high levels of environmental uncertainty due to rapid changes in the economy and society (An et al., 2016). However, the idea that formal collaborations also consume management’s span of attention (Simon, 1947) is observed in the positive moderation result since it worsens the impact of openness on innovation.

Our results relating to the use of government and private funds are particularly interesting when Chinese and Australian samples are compared. While the literature on the relationship between financial support and innovation is quite mature (for a review, see Levine, 2005), researchers are yet to fully understand the relationship between access to funds and open innovation processes (Hall & Lerner, 2010). Our results suggest that when Chinese businesses utilize private or government funds, openness becomes less important for innovation outcomes, possibly due to a substitution effect. A potential explanation can be found in the small business life cycle literature (Greiner, 1972; Scott & Bruce, 1987), which explains that when firms receive private funds they usually have passed the survival test and, thus, will focus on the specific business plan in place and on building internal competencies that are common in the so-called success phase (Lester et al., 2003). They are, therefore, less incentivized to use broader openness, but instead source specific knowledge as the need arises (Zack, 1999). The result with Australian firms and government support having a positive moderation effect on openness and innovation could be due to the more effective administration of these support schemes; however, more research is required.

We found no moderation effect for private funds in Australia. We propose that this may be a result of the lack of availability of funds in general. The well-reported (DIIS, 2017; Gans & Hayes, 2006) limited availability of private and venture capital¹⁰ to firms in Australia, in comparison to other markets, leads us to believe that the number of private funding opportunities is not sufficient to result in a moderation effect on the relationship between sourcing and innovation outcomes. Indeed, our results indicate much lower instances of government and private funds in Australian firms more generally.

When considering the role of training, we did not expect the lack of significance of this moderator in the Australian context. Even if training is not regarded as a strategy itself but rather as a condition to support different strategies (Smith & Hayton, 1999), we expected training to have a significant impact on the relationship between openness and innovation. Critical steps after sourcing include the capabilities to select (Felin & Zenger, 2014), capture (Cassiman & Veugelers, 2002), and build (Teece, 2000) from such knowledge to produce innovation outcomes. Therefore, it goes to reason that specific training would enhance the capability to move from openness to innovation. However, and as we previously explained, the incentive for training in LMEs is low Hall & Soskice, 2001). Indeed, Rogers (2004) found that in Australia, there is no association between training and innovation, suggesting that training may not be aligned with firms’ needs to foster innovation. In addition, the Australian education system reports a mismatch between industry demand and skills in the workforce, coupled with the vocational education sector being poorly resourced (DIIS, 2017). In contrast, the entrepreneurship literature suggests that SMEs favor informal rather than formal specialized training, as the latter enables them to realize returns only in the long run (Hill & Stewart, 2000). The lack of emphasis on formal training is more prevalent among firms that have high levels of vulnerability and uncertainty, such as new ventures, and thus these firms tend to focus on shorter-term commitments (Coviello, 2006). Previous research on training in Australia (Kotey & Folker, 2007) supports these arguments, leading us to believe that the lack of a moderation effect from training may relate to the mismatch of formal and specialized training within Australian new ventures due to the training-skills mismatch (DIIS, 2017). Another argument may be that workers in Australia, particularly within the services industry, are already well qualified and that managers perceive this as adequate and, thus, do not prioritize training.

However, the fact that training negatively impacts the relationship between external knowledge sources and innovation in China is somewhat surprising. This result indicates that more training will be prejudicial for open innovation activities to produce innovation outcomes. This is contrary to our understanding of open innovation (Y. Wang et al., 2012), where education and training are a cornerstone to open innovation practices. It could be the case that training makes staff more attractive to competitors, with the result that trained people leave the business, making it more difficult for these firms to innovate. This needs further exploration.

From a diversity of economic systems literature perspective, our results show that the environment in LMEs influences firms’ innovation ecosystems differently from what is reported in the literature (for example, Casper & Whitley, 2004; Hall & Soskice, 2001). The results challenge, for example, some of the assumptions regarding the influence of R&D cooperation and private financial systems on firms’ innovation outcomes. In our CME, the opposite was found, with all external variables influencing the relationship between openness and innovation outcomes. Based on the diversity of the economic systems literature, one would expect openness to not have a significant relationship with innovation in a poorly institutionalized country, such as China, where high transaction costs are well documented (Carney et al., 2009). However, this is contrary to what the open innovation framework asserts because, independently of the institutional environment, firms would benefit from open innovation activities (Chesbrough, 2006). Indeed, our results show that when R&D cooperation and training involvement is high, a weaker relationship between openness and innovation outcomes occurs. These findings are particularly important because the literature on central economic systems is still limited (Li et al., 2014) and because our results challenge past research (for example, Redding, 2002), especially on how important, and effective, government involvement is in CMEs. Furthermore, our results on the similar effects between government and private funds support past research that explains that central governments have a tight influence on the allocation of financial resources, including commercial banking systems (Witt & Redding, 2013).

From an institutional viewpoint, previous conceptual research (Redding & Witt, 2009) has suggested that centralized economies, such as China, present a lack of institutional support – namely, a lack of effective and efficient rule of law – which is required to develop innovative firms. Our results add to this argument by empirically showing that even if Chinese firms report being innovative, they are not doing it through an openness strategy. On the contrary, openness is an inhibitor of innovation in China. Furthermore, our results show the importance of understanding institutions in a centralized economy by explaining that all the externalities studied have a significant impact on the relationship between openness strategies and innovation capabilities. However, this is not true in LMEs, where only two external factors influence this relationship.

Our results report that Chinese firms produce more radical innovations than those in LMEs. While this finding is contrary to the patterns of innovation described in the diversity of economic systems literature, recent studies (for example, Meelen et al., 2017), even if in a different type of economy, support the finding of a lack of relationship between liberal economies and radical innovation outcomes. This difference can be explained by the entrepreneurship literature, which clearly establishes the importance of government and private funds (Hall & Lerner, 2010) for the development of radical innovation. These features are not present in our LME (Australia), but are present in our CME (China). Also, from an institutional perspective, Redding (2002) analyzed the Chinese business system and concluded that it would not readily support innovation arrangements due to the lack of institutional arrangements. However, some (Torres de Oliveira & Rottig, 2018) have questioned whether the Western institutional lens is sufficient to understand the informal and semiformal arrangements and institutional complementarities that exist within central economies.

In sum, from an open innovation viewpoint, our results establish that the external, national-level pressures influence one of the most preeminent relationships in the open innovation literature – the positive influence of openness on innovation outcomes (Laursen & Salter, 2006). This is particularly important because past research has established that intrafirm capabilities (for example, Teece, 2007) and industry dynamics (efor example, Jacobides et al., 2006) moderate this relationship. Our study goes further to establish the influence of country-level factors on open innovation. By doing so, we demonstrate that beyond the firm or industry level of analysis, institutional complementarities affect open innovation, and that only a multilevel perspective that also contemplates institutional factors, similar to studies in other fields (for example, Peng et al., 2008, in international business), can capture an understanding of the dynamics of open innovation. Furthermore, our results explain how these dynamics vary in different economic systems; namely, by explaining that, in an open innovation context, external constraints are more dominant in CMEs compared to LMEs.

Managerial and policy implications

Our results have important managerial implications for both LMEs and CMEs. In particular, our results suggest that managers from LMEs should focus on openness to capture external knowledge, but also develop internal capabilities and absorptive capacity to counter the lack of influence that external factors have on the firm’s innovation ecosystem. The only exception is the use of government funds, which enhance the relationship between openness and innovation outcomes. In contrast, managers from CMEs need to be aware of the possible impact that external factors have on the translation of openness into innovation. They should also utilize government and private funds to foster open innovation strategies. Within such ecosystems, as the literature (Löfsten & Lindelöf, 2002) suggests, R&D cooperation will improve.

From a policy perspective, our results show that even within the confounds of limited government support (DIIS, 2017), funds impact innovation and how openness translates into innovation in LMEs. Our results challenge the idea that government capital is incapable of effectively building innovation ecosystems (Mason & Brown, 2014). Furthermore, our results show that in a CME, government funds facilitate the relationship between openness and innovation. This finding has important implications for technological parks and university incubators, given the association between government funds and such investments. We also show that in a CME, the influence of institutional factors is high on the firm’s openness ability and how this impacts innovation. Thus, governments should be aware of how different initiatives impact firms’ ability to search and innovate to enhance the relationship between open innovation and innovation outcomes.

Conclusion, limitations, and future research

In this article, we examined how institutional complementarities moderate the relationship between openness and innovation outcomes. We were motivated by the lack of attention to the institutional level of analysis in the open innovation literature. We therefore explored how different economic systems, in our case a central and liberal economic system, influence open innovation.

Using the economic systems literature and institutional theory, the contribution of our article is threefold. First, the article introduces an institutional complementarities moderator to the well-established relationship between openness and innovation, and empirically shows that economic system arrangements impact this well-established relationship differently. In doing so, we use the external components of the diversity of economic systems framework to measure the effect of innovation systems on openness and innovation outcomes. This is particularly important to the open innovation literature given that the influence from national institutions on open innovation is usually treated with skepticism. Our results show that institutional factors have an impact on the open innovation framework.

Second, grounding our work in the diversity of economic system literature, this article studies a liberal and central economic system and uncovers some institutional complementarities to show how different institutional arrangements affect firms in different ways. A particularly important finding, which is contrary to past research based on Western economies, is that, in a CME, openness does not increase the likelihood of innovation outcomes.

Third, the article empirically establishes the influence of a central market economic system on open innovation and compares that influence in an LME. Our results bring clarity to the role of different economic systems in open innovation and innovation outcomes more generally. Specifically, we uncover how firms deal with the different pressures and the implications this has on open innovation.

This article has several limitations. The first relates to the nine-month time difference between the data collection in China and in Australia. This lag was due to different procedures and discussions with Chinese authorities. During this period, however, no major changes occurred in the environment in China or in Australia. As such, this time difference should not influence our results because they are limited to the use of cross-sectional data, similar to previous studies (for example, Laursen & Salter, 2006). A future longitudinal study would help to understand how these environmental changes happen over time. Another limitation relates to the use of binary variables as moderators, even if aligned with past literature (for example, Lee et al., 2010). The use of appropriate data analysis techniques mitigated this limitation. Another limitation is the use of a single respondent from each firm. We addressed this limitation by interviewing senior managers who had sufficient knowledge and a holistic perspective of both organizational processes and practices, and the external business environment (Verreynne et al., 2016). We also acknowledge that the responses in China are from a single province. However, similar to others (Hall & Soskice, 2001; Whitley, 1999), we considered the boundary conditions of capitalist systems at a country level, and thus they are not province dependent.

Looking ahead, our study poses interesting questions on which future research can build. For example, future research can look at other open innovation strategies such as acquiring, selling, and revealing. If we expect that a firm’s external environment will shape all of those different strategies, considering that internal firm capabilities have varied impact, we can expect that those external pressures may pose different influences. Moreover, these external pressures shape not only open innovation knowledge breadth, but depth as well (Laursen & Salter, 2006). Thus, future research can look at how external pressures in general and the specific economic system in particular moderate the relationship between open innovation depth and innovation outcomes. Furthermore, as this study is the first attempt to understand how the external environment influences open innovation, future research can study those pressures exhibited at different levels. From a firm-level and an industry perspective, the extant strategy literature has established how different firm capabilities (Barney, 1991) and specific industry dynamics (Teece, 2000) influence firms differently. Accordingly, one can expect that different economic system settings will incentivize different forms of breadth and depth of search for knowledge. Testing hypotheses such as these but using a multilevel unit of analysis will allow researchers to better understand the open innovation framework.

From a diversity of economic systems perspective, future research should focus on the results relating to our CME; in particular, it should aim to address the lack of adequate explanations in the institutional complementarities literature regarding the mechanisms that pressure firms and the capabilities that firms in such environments need to develop to resolve their coordination problems. The “quasi-experimental” system (Carney et al., 2009) that non-Western economies are sometimes referred to as, might just be a reflection of an understudied – and, thus, poorly understood – coordination system. Ideally, this should comprise in-depth case studies to uncover these constructs and mechanisms, which could then be complemented by econometric analysis.

Acknowledgments

The authors acknowledge Dr. Tam Nguyen, who carefully helped with part of the statistical analysis. We are grateful for the guidance and support of the Associate Editor Dr. Masatoshi Kato and the anonymous reviewers.

Additional information

Funding

This work was supported by the Australian Research Council, Discovery [Project No. DP160100602].

Notes

¹ In this work, we use “innovation system” as the concept of national innovation system as introduced by Lundvall (1988a, 1988b) and widely diffused to the Organisation for Economic Co-operation and Development (OECD), European Commission, United Nations Conference on Trade and Development (UNCTAD), World Bank, International Monetary Fund (IMF), and the US Academy of Science, to name a few.

² A notable exception is Garriga et al. (2013).

³ In this article, we use openness and search breadth interchangeably.

⁴ Differently from the diversity of economic systems literature, which uses the designation of interfirm relations, the innovation literature calls these relationships “R&D cooperation.”

⁵ We would define shadow banking as all the credit not regulated by standard and conventional financial institutions. For a thoughtful discussion, please refer to Harutyunyan et al. (2015).

⁶ Following F. Allen et al. (2008, p. 3), such informal channels are “informal financial intermediaries, internal financing and trade credits, and coalitions of various forms among firms, investors, and local governments.”

⁷ If workers are less than 5 years from retirement, have been with the firm for more than 15 years, or are pregnant, they are protected from unfair dismissals. If not, the market presents a flexible human resources policy based on financial compensations.

⁸ 中华人民共和国劳动合同法 – Labor Contract Law of the Peoples Republic of China (PRC).

⁹ From here on referred to as “innovation.”

¹⁰ In 2011, the percentage of Australian venture capital over gross domestic product was less than 0.01 percent, whereas in the United States it was more than 0.2 percent (Cumming & Johan, 2016), and even in 2016 the expenditure of private and venture capital in Australia was well below the OECD average (DIIS, 2017).

Appendix.

Bias testing

To avoid the issue of nonresponse bias, the Australian survey utilized a stratified sample (based on data from the Australian Bureau of Statistics) to ensure representativeness across states, industries, and firm sizes. We also checked if differences existed between data from the survey and the population. We found that, for instance, the proportion of firms operating in the service and manufacturing sectors was almost the same in the sample and in the population. We also did not find any significant differences among firm sizes in our sample and the population. Thus, we did not consider nonresponse bias to be an issue. In our Chinese data, a 76 percent response rate signaled that nonresponse bias was not a major issue (Armstrong & Overton, 1977).

Although both surveys followed a stratified procedure – which to some extent helped us to avoid sample selection bias (Pannucci & Wilkins, 2010) – there are concerns about the representativeness of our data. For instance, we examined only 223 out of the 1,566 Australian firms in our database (new firms), and firms in our sample include a high proportion of innovators – around 60 percent. However, we note that we limited our sample to new ventures, six years and younger, as prescribed by the literature (McDougall et al., 2003). These firms are more likely to have a higher rate of innovation compared with established counterparts. There could also be a concern that the differences between our two subdatasets might affect the validity of our comparison. For instance, Chinese firms are more likely to be in manufacturing, whereas Australian firms are more service oriented; this is typical of the structures of these two economies (Garnaut et al., 2018). To alleviate these concerns, we applied matching statistics in our first empirical step. Taken together, sample selection bias has been appropriately mitigated.

To check for common method bias, we applied Harman’s single-factor test. The results showed that the first factor explained only 19 percent of the variance. We also applied the marker variable technique to detect common method variance using correlations between the marker variable and variables in the study (Lindell & Whitney, 2001). We chose the years in business as our marker variable. The results indicate there are no substantive correlations between this marker variable and other variables; thus, common method bias did not appear to be a problem (Crowston et al., 2015). Combining the two datasets from China and Australia further averted any common source bias that may arise from exploring only a single data source (Podsakoff et al., 2003).

There could be a range of influences on a firm’s ability to innovate, such as management practices and ties with vital customers, which are not observable. These influences advantage some firms when searching for different external sources of knowledge. R&D cooperation, access to government and/or private funds, and training provision also depend on the firm’s characteristics such as industry, size, and ownership. For instance, larger firms tend to have privileges in accessing government funds in some countries (Colombo et al., 2016; C. Wang et al., 2012). Additionally, search breadth, access to government and private funds, engaging in R&D cooperation activities, and providing employee training differ between Chinese and Australian firms, which may bias our estimation – especially when comparing the different impact between the two countries. Taken together, the effect of search breadth and our four moderators on innovation could be overstated, raising concerns about endogeneity. Therefore, we applied three statistical techniques to deal with this issue.

First, we used matching to control for the difference between two countries, as shown in the first step of our statistical method and, second, we used mean-centered search breadth in the second step. Third, the instrument variable method was applied by executing an instrumental variable probit regression in our fourth step, in which two endogenous variables – namely, search breadth and the moderator (R&D cooperation, government funds, private funds, and training, respectively) – were included in one instrumental equation. Drawing on the literature, firm size (Brunswicker & Vanhaverbeke, 2015), industry (West et al., 2014), and markets (Spithoven et al., 2013) were used as instrumental variables for the endogenous variables. The Wald tests of exogeneity were all significant, implying that the application of instrumental variables is appropriate (Wooldridge, 2010).