1 Introduction

Innovation is not a development of one individual firm, but an interactive process which interconnects the market view of companies to the scientific development of universities (West et al., 2014). For that reason, universities play a crucial role in society as producers of state-of-art scientific knowledge (^{D’este and Patel, 2007}).

The innovative economies use cooperation with universities as a source of state-of-art knowledge and technology, and many of the world's most economically advanced regions possess top-tier academic and research institutions (^{Etzkowitz et al., 2019}; ^{Henton & Held, 2013}). In contrast, universities in emerging economies are more likely to engage in testing, adapting, and diffusion of foreign technology to local demands, supporting the national industry (Fu & Li, 2016). However, as shown by some recent studies (i.e. ^{Schaeffer; Dullius; Rodrigues; Zawislak, 2017}), in Brazil the most common University-Industry (UI) interactions are development-oriented, therefore, many firms do seek university not just to improve existing products or processes, but to achieve higher level innovativeness.

There are some empirical studies that analyzed factors that influence the innovativeness of the firms that interacted with universities (^{Bishop, D’Este, Neely, 2011}; ^{Kobarg, Stumpf-Wollersheim, Welpe, 2018}; Li, Li & Wu, 2019). However, in general, the number of studies that checked the relations between firms’ innovation performance and UI interaction per se is surprisingly low. Some studies, for instance, only checked whether the presence of cooperation, regardless of its aim, affects innovative performance (^{Arant, Fornahl, Grashof, Hesse, Sollner, 2019}). Also, studies tend to measure firm innovativeness in the number of patents (i. e. Arant et al. 2019; ^{Baba, Shichijo, Sedita, 2009}). However, patent applications do not necessarily translate into the creation of new products or processes and vice versa, particularly in low-tech sectors (^{Reichert, Zawislak, Arundel, 2016}; Smith, 2005).

While measuring the impact of UI interaction on innovativeness, the previous studies did not ask the companies whether they consider UI interaction as an important source for firms’ innovation activities (^{Baba et al., 2009}). Moreover, the knowledge on which type of UI interaction or configuration of UI interaction does lead to high innovative performance remains unknown, at least from an empirical point of view.

In this context, we propose the following research question: which are, whether exist, the best configurations of UI interactions that firms should engage to achieve high innovativeness? The method used to conduct the study had two steps. First, a crisp-set qualitative-comparative method (csQCA) was applied to a sample of manufacturing firms to determine the required types of UI relations that will lead the firm to high innovativeness. Second, the cases covered by csQCA solutions were analyzed.

The results showed that research-oriented interactions (RE) and development-oriented interactions (DE) are the only interactions with universities that alone can allow the firm to achieve high innovativeness. Moreover, these results apply to firms from different sectors, different sizes, and different R&D intensity. In contrast, the diffusion-oriented interactions (DI), which is less complex than science-oriented and development-oriented (^{Schaeffer et al., 2017}) showed no contribution to the firm's high innovativeness.

We highlight that among firms that interact with universities those that engage in more complex relations tend to achieve a higher level of innovativeness than those that do not. These results suggest that it is important for the firm to engage in more complex, long-term UI interactions to achieve high innovativeness. In turn, this brings important theoretical, empirical and policy implications.

The paper is structured as follows. After introduction, university-industry interaction literature is presented. Section three includes the method description. The results and discussion are presented in section four. The conclusion is made in section five. Finally, references are listed.

2 University-industry interaction

In recent decades, it has been observed that a greater number of universities, in the most diverse countries, have been increasingly involved in innovation and industrial development, the so-called third mission (^{Wang, Hu, Li, Li, & Li, 2015}). According to ^{Etzkowitz (1998}), universities were established with the primary objective of teaching (first mission). In the 19th century, research was integrated as one of the main activities of universities (second mission). After the 1980s, another turning point in the evolution of studies on the roles of universities in society was represented by the emergence of the third mission. Since then, the academy is no longer seen only as a producer and disseminator of knowledge, but also has the function of creating relationships, which can be equally important (^{Andersson & Berggren, 2016}). Thus, researching the context of universities has great relevance as they are increasingly playing a central role in advancing sustainability on a local, regional and national scale through intersectoral collaborations (^{Trencher, Bai, Evans, McCormick, & Yarim, 2014}).

The interaction between scientific production and technological production plays an important role in innovation systems (^{Mazzoleni & Nelson, 2007}). It is possible to identify the existence of positive feedback circuits between these two dimensions, where there are flows of information and knowledge in both directions (^{Cohen, Nelson & Walsh, 2002}). One of the important roles of universities for the development of countries is the advance on the frontier of knowledge with a view to its applicability in the productive sector. Universities are general sources of knowledge necessary for basic research activities (^{Nelson, 1990}), as well as sources of specialized knowledge related to technologies applied in companies (^{Klevorick, Levin, Nelson, & Winter, 1995}), in addition to being responsible for the education and training of scientists and engineers able to solve problems related to companies' innovative process (^{Rosenberg & Nelson, 1994}).

Companies, on the other hand, are increasingly focused on cooperation with universities (^{Mascarenhas, Ferreira & Marques, 2018}), this being a solution adopted mainly by those that are intensive in R&D (^{Lam, 2007}). The emergence of this relationship between university-industry (UI) as the main actors in the development of the innovation system has cultivated the links between science and industry, generating various forms of collaboration (^{Yoon, 2015}), promoting economic progress, innovation and competitiveness (^{Skute, Zalewska-Kurek, Hatak, & de Weerd-Nederhof, 2019}). This approximation between UI is developed in different ways and several researchers have placed emphasis on the ways in which relationships are established. ^{Etzkowitz and Zhou (2018}), cite technology parks and attribute to them an interactive model that generates entrepreneurial activity through various means, including incubator projects and academic programs. ^{Clark (2005}), on the other hand, emphasizes the accumulation of business groups within universities - these groups extend from disciplinary departments and interdisciplinary and transdisciplinary research centers at the base, to colleges and schools at intermediate levels, for the entire university. In these places, it is possible to find teachers and managers intertwined at all levels, involved in all issues.

The relationship of academics with industry is largely compatible with the continuous generation of scientific results (^{D’este & Perkmann, 2011}). According to ^{Elia, Secundo & Passiante (2017}), the collaboration between science and technology is intense, usually with a focus within the university in a supporting role of the industry. This industry knowledge transfer is an important source of creation for all partners (^{Alexander, Martin, Manolchev & Miller, 2018}), however, the academy's ability to effectively transfer knowledge to the industry is essential for universities to fulfill their mission and entrepreneurial ambition (^{Miller, Alexander, Cunningham & Albats, 2018}). Within this context, relevant policies may be needed to facilitate collaboration in improving innovation efficiency at different stages (^{Shi, Wu, & Fu, 2020}). Recently, policymakers and industry strategists in developing countries have started to devote more attention to initiatives that promote collaboration between universities and industries. The main objective of these initiatives is to improve the capacities and efficiency of innovation systems, leveraging the role of universities as generators and disseminators of valuable knowledge (^{Fischer, Schaeffer, Vonortas & Queiroz, 2018}).

There are many types of UI relations. Among them we can highlight consultancy (^{D’Este & Patel, 2007}; ^{Meyer-Krahmer & Schmoch, 1998}), contract research (D’Este & Patel, 2007; Meyer-Krahmer & Schmoch, 1998; ^{Santoro, 2000}), education of personnel (Meyer-Krahmer & Schmoch, 1998), technology transfer (Santoro, 2000), formal agreements without clearly defined goals (^{Bonaccorsi & Piccaluga, 1994}), use of universities’ facilities and informal meetings (^{Bishop et al., 2011}; ^{Puffal, Ruffoni, Schaeffer, 2012}).

Recently, ^{Schaeffer et al. (2017}) compiled different typologies according to aspects and dimensions of UI relations previously identified by the academic scholars, which are: duration of the relationship, the direction of information flow, level of knowledge involved; the degree of formality, the complexity of interaction, the absorptive capacity of the actors.

Thus, ^{Schaeffer et al. (2017}) suggested that there are five types of UI cooperation: training-oriented, service-oriented, diffusion-oriented, development-oriented, and science-oriented. Table 1 shows the description of each relation type.

It’s important to add that some previous studies showed that knowledge inflows from universities, that is, science-based flows, tend positively impact on firm innovativeness (^{Bishop et al., 2011}; ^{Murovec & Prondan, 2009}), including firms from emerging markets (^{Puffal & Teixeira, 2014}). ^{Arza (2010}) also argues that long-term interactions may impact more on firm innovation output than less complex, service, and consultancy-based interactions. However, there are also studies that showed that UI interaction per se does not have a significant influence on innovation (Eom & Lee, 2010), at least in some cases (^{Baba et al., 2009}). Moreover, in empirical terms, it is still not clear which type of UI relationships are the best in terms of allowing the firm to achieve high-level innovation output. Therefore, the present study aims to check which are, if exist, combinations of UI relationships that do lead to high-level innovation output.

Considering the Brazilian national context, we can observe the results of the national innovation survey - PINTEC. It indicates that Brazilian companies have shown certain stability in the rate of product and/or process innovation. Since the beginning of the survey, in 1998, the average innovation incidence has been around 35%. During the years 2009-2011, the incidence was 35.6%, between 2012-2014, it was 36.4%, and in the three years 2015-2017 the innovation incidence dropped to 33.6%, it shows a decrease compared to the previous, which suggests a recent phase of a greater increase in the difficulties faced by companies to carry out innovation (^{IBGE, 2020}). The database used in this article, called BR Survey, comes from a survey conducted with 325 companies that had interactions with universities. The data indicate that in this group of companies the incidence of innovation was 91.4%. Despite the highly innovative performance of this group, it is not possible to say that there are a cause and effect relationship with the university-industry interaction. However, there is a positive correlation between the university-company interaction and the innovative performance of the companies, indicating that the interaction can be an important element for innovation when compared with the PINTEC data.

Several articles were produced from this database that analyzed issues as the motivation for the interactions, the funding influence, the types of relationships, among others. ^{Pinho (2018}) stands out that among many relevant results found from this database, the knowledge transfer channels, such as university publications and informal exchange information are more important than usual technology transfer mechanisms, like incubators, technology licensing and spin-offs, results also pointed out by ^{Puffal and Teixeira (2014}).

^{Pinho (2018}), as well as ^{Schaeffer et al. (2017}), indicate that there is evidence to affirm that Brazilian companies value the contribution of universities to their innovative effort and that the relations between universities and companies, in Brazil, not only exist but are relatively intense. However, the relationship is not usually focused on innovations of greater scope, based on the forefront of scientific knowledge, which can be justified by the technological demand of companies.

3 Method

This paper draws on a secondary data retrieved from a national databank of UI relations survey conducted in 2008-2009 by the main Brazilian universities within a project called “University-industry cooperation in Brazil”, also known as “BR Survey”, and CNPQ Database 2004, a Brazilian census of universities research groups.

The UI relations survey was based on the Carnegie Mellon Survey on Industrial R&D (^{Cohen et al., 2002}) and the Yale Survey on Industrial R&D (^{Klevorick et al., 1995}). Currently, it is a largest innovation database on UI cooperation in Brazil. The UI relations survey database was constructed as follows:

Out of CNPQ Database 2004, only those groups that declared to have cooperated with private companies were selected. The procedures resulted in 2.151 research groups, which declared to have cooperated with 1.688 companies with different pay tax numbers.

The 1.688 companies were contacted by phone call and to those that accepted to take part in a survey, it was sent an online questionnaire. In total, 19.3% of the companies accepted to take part in a survey, resulting in a final sample of 325 firms, including 196 manufacturing firms¹.

As shown in Table 2, manufacturing companies that took part in the survey were firms of different sizes, R&D intensity rates, and from different economic sectors.

Figure 1 shows the procedures for the final sample selection. In order to maximize the association between engagement in UI relations and innovation outcomes, out of 196 manufacturing firms with different taxpayer numbers, researchers selected those firms which consider university as a very important element to at least one aspect of its innovation activities. Thus, based on 4-level semantic differential scale, firms that consider a university as “very important” for at least one of the following: (a) suggesting new innovation projects, (b) finishing undergoing innovation projects, (c) research results, (d) prototypes, (e) new techniques and instruments, (f) laboratories were included into final sample. The procedure resulted in 105 firms.

Figure 1 Procedures of sample selection for innovativeness analysis with csQCA 

Out of 105 firms, researchers selected only innovative firms that implemented (1) new process or product innovation during the last 3 years and that one that stated that cooperation with the university research succeeded. The procedures resulted in 79 innovative manufacturing firms.

The UI relationships in which 79 remaining firms engaged were mapped through CNPQ Database 2004 and categorized according to ^{Schaeffer’s et al. (2017}) typology. Previously Schaeffer’s et al. (2017) typology had been applied to analyze the UI cooperation from CNPQ Database 2010, therefore, the typology best fit the Brazilian context and available data.

The study employs a crisp-set qualitative-comparative method (using software fsQCA.com) to identify search for UI relations configurations that result in firms’ high innovativeness.

The crisp-set qualitative-comparative method (csQCA) was first created by Ragin and uses Boolean Algebra (logic of 1 - present and 0 - absent) to identify the combination, that is, the solution formula of given factors that lead to a specific outcome (^{Grofman & Schneider, 2009}). The advantage of csQCA over statistical methods such as, for instance, linear regression, is that it allows researchers to suggest what is the combination of factors/conditions that will lead to a given outcome. Likewise, unlike most statistical methods, the csQCA was designed to work with medium-N samples (Ragin, 2014), which is the case of the sample used in the present study (N=79).

The process of QCA analysis starts with definitions of conditions of the outcome, that is, with the definition of factors that may lead to a given outcome based on the extant theoretical knowledge (Ordanini, Parasuraman, Rubera, 2014). In the present study, we suggest a combinations of 5 types of ^{Schaeffer’s et al. (2017}) interaction types. Therefore, the study tests the following csQCA model: HInovOut = (TR, SE, DI, DE, RE).

Once created the QCA model, the next step is calibration (^{Reichert et al., 2016}; ^{Rihoux & Ragin, 2008}). In this study, we defined high innovativeness as those innovations that exceed the “novelty to the firm” frontier. Thus, only firms that implemented “new to the country” or “new to the world” product or process innovation in the last 3 years were considered as highly innovative. Concerning the ^{Schaeffer’s et al. (2017}) interaction types, for each interaction type of each observation it was set 1 when present and 0 when absent.

Crisp-set QCA uses a truth table function to generate configurations of conditions (TR, SE, DI, DE, RE) that are enough for achieving high innovation output (HighInnovOutput). The truth table for all logically possible combinations is 2 elevated on a few conditions (5), that is, 32 rows. To reduce the number of columns, a selection of consistency level is required. ^{Rihoux and Ragin (2008}), also ^{Fiss (2011}) suggest a consistency cutoff of at least 0.75. In the present study, it was decided to raise the consistency cutoff to 0.8.

The output of csQCA includes three types of solution: complex, intermediate, and parsimonious. ^{Ragin (2014}) suggests using an intermediate solution, as it represents the balance within complexity and parsimony by using procedures like the practice of conventional case-oriented comparative research. In the present study, however, it was decided to use a parsimonious solution. This solution allows to reduce the number of conditions in each individual solution, as well as, in some cases, the number of individual solutions. In turn, using only parsimonious solutions allows to reduce the total number of analyzed elements when working with more than one frequency cutoff. It’s important to add the that together with solution consistency, the “quality” of csQCA solution is measured by solution coverage, which assesses the degree to which solution “accounts for” of an outcome condition, that is, coverage is analogous to R2 in statistical analysis (Ragin, 2014)

After obtaining csQCA solutions, a descriptive analysis of cases covered by the solutions took place.

4. Results

The present section is divided into two parts. First, results of csQCA analysis are presented. Then, a descriptive analysis of cases covered by solutions is performed.

5. Conclusion

The csQCA analysis showed that development-oriented interaction and research-oriented interaction, which are the most complex UI interactions (^{Schaeffer et al., 2017}) are the only UI interactions that alone lead the firm to high innovativeness. In contrast, service-oriented and training-oriented relations, which usually comprise low levels of complexity (Schaeffer et al., 2017), lead the firm to innovate beyond “new-to-the-firm” borders only when complemented by development-oriented or research-oriented interactions. Diffusion-oriented interaction, which also presents a lower level of complexity than development-oriented and research-oriented interactions, showed no contribution to high innovativeness, as it does not appear in any of the solution’s configurations.

The study also revealed that highly innovative firms covered by csQCA solutions have relatively low R&D investment rates, even when they are from high and medium-high technological intensity sectors according to OECD’s (2011) Isic 3 classification. This finding allows us to infer that firms that are engaged in development-oriented interaction or research-oriented interaction with universities can effectively achieve high innovativeness even in the absence of high R&D investments. These finding stresses, even more, the importance of engaging in complex and long-term interactions with universities for the firms that want to become highly innovative.

The present study contributes to the literature in three ways. First, this is the first study to apply crisp-set QCA to identify configurations of UI interactions that lead to the firm’s high innovativeness. Second, unlike previous investigations, our study determines which are the necessary conditions that allow firms engaged in UI interaction to achieve high innovativeness and which are not. In addition, the method applied, and the findings of the present paper empirically validate the ^{Schaeffer’s et al. (2017}) UI interaction typology. It is important to highlight that, unlike many previous studies, the present article analyzes the sample composed exclusively by firms that consider university as an important agent to its innovative activities and has cooperation success. In turn, this sample composition enhances the qualitative validity of the used factors and outcomes.

We expect that the results will help managers to better formulate strategies of engaging in UI interactions and to choose a suitable configuration of UI interaction for the results they want to achieve. Finally, the finding of the present study may motivate the government and policymakers to formulate strategies that promote policies that aim to engage the firms and universities into more complex and long-term interactions

The study has some limitations. First, the present study uses data from the first decade of the century and therefore it would be important to perform the same study with more recent data. Second, the present study measures the firm's innovativeness through one variable, which is also the only output variable available in the BR survey database and therefore it is important to incorporate other innovativeness measures in further investigations.

For future studies, it would be interesting to expand the present research to sectors beyond manufacturing firms. The replication of the study in the context of advanced innovative economies seems to be an interesting opportunity also.

It should be noted that due to this being the first study on the results of university-industry interaction which applies the csQCA, it is likely that there are possibilities for improvement in the adjustments made in the csQCA configuration.