Firm performance and R&D cooperation: what matters?

ABSTRACT

In this paper we analyze whether R&D cooperation improves firms' performance, and which variables would boost R&D cooperation the most. We contribute to the literature linking two different branches: using a sample of Italians firms, we analyze which are the main determinants of R&D cooperation and whether R&D cooperation improves firms' performance. Thus, we conduct a two-step analysis: in the first step we investigate the significant determinants of cooperation, and in the second we employ these determinants to test the effect of R&D cooperation on firm performance, measured by ROA, Ebitda, and value added per capita. The analysis focuses on four relevant disaggregations: (i) small, medium, and large firms; (ii) Pavitt (1984. “Sectoral Patterns of Technical Change: Towards a Taxonomy and a Theory.” Research Policy 13 (6): 343–373) sectors; (iii) location (North, and Center-South); and (iv) public and private partners. Although the determinants of cooperation can vary depending on the type of disaggregation, we find that R&D cooperation always boosts firm performance, that is, on average, profitability and productivity are higher among cooperating firms compared to those that do not cooperate. Identifying the determinants of cooperation allows management and policymakers to understand which strategies should be implemented to stimulate cooperation and, consequently, to improve firms' performance.

KEYWORDS:

• R&D cooperation
• determinants
• performance
• innovation
• panel probit models
• Heckman

JEL CLASSIFICATION:

• C23
• C25
• H32
• O32

Acknowledgments

This work is part of the research project ‘VALERE: VAnviteLli pEr la RicErca’. The authors wish to acknowledge the editor and three anonymous referees for their contribution. This paper is also in memory of our colleague and friend Professor Carlo Capuano (1973–2022), who gave us useful advice in the first phase of the drafting. Any errors are our own.

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The theoretical literature examines the effect of R&D cooperation on investment, profits, social welfare, etc., and in most cases find that R&D cooperation increases these outcomes. The seminal paper is D'Aspremont and Jacquemine (1988); other contributions include De Bondt and Veugelers's (1991), Kamien, Muller, and Zang (1992), Choi (1993), Leahy and Neary (1997) and Goyal and Moraga-Gonzales (2001) more recently, Capuano and Grassi (2019), Cabon-Dhersin and Gibert (2019), Karbowski (2019).

2 See, inter alia, Belderbos, Carree, and Lokshi (2004), Cincera et al. (2003) and Aguiar and Gagnepain (2017).

3 The empirical literature shows that the main determinants of cooperation are subsidies, spillovers, absorptive capacity, public policies, and size. However, these determinants may vary with the partner. The literature includes: Belderbos et al. (2004), Colombo, Grilli, and Piva (2006), Franco and Gussoni (2014), Badillo and Moreno (2016), Cantabene and Grassi (2019, 2020). A survey of the growing empirical literature on the determinants of R&D cooperation can be found in Marinucci (2012).

4 ROA and Ebitda are used to measure firm profitability in various literatures such as corporate social responsibility (e.g. Barnett and Salomon, 2012), and strategy (King and Zeithaml, 2001; McNamara, Luce, and Tompson, 2002; Wan and Hoskisson, 2003; Lavie, 2007; McNamara, Haleblian, and Dykes, 2008). On Ebitda see especially Bouwens, De Kok, and Verriest (2019), Rozenbaum (2019), Stewart (2019), and D’Souza, Ramesh, and Shen (2010) among others. Value added per capita is used to measure firm productivity and is the amount by which the value of an article increases in each production stage, exclusive of initial costs, divided by the number of employees.

5 See Ahuja and Katila (2001), Miotti and Sachwald (2003), Laursen and Salter (2004, 2006) and Lavie and Miller (2008).

6 In Italy, private universities receive huge funding from the State: the Ministry of Education, University and Research (MIUR) allocates the resources of the ordinary funding without distinguishing between public and private universities.

7 In this respect, our paper is coherent with the consolidate literature on university-industry collaborations. See, inter alia: Laursen and Salter (2004), D'Este and Patel (2007), D'Este and Perkmann (2011), Bellini, Piroli, and Pennacchio (2019).

8 For example, the main determinant of cooperation for science-based firms, large firms, and in the case of cooperation with private is private incentives. Public incentives are important for cooperation with public sector organizations, for small firms, and firms in center-south of Italy. Human capital is the main determinant of cooperation with public partner and science-based firms but has little impact on cooperation with private firms. It is important for policy makers and managers keen to encourage cooperation to be aware of its determinants since this will affect the tools and incentives employed for different types of firms. For example, among firms in the center-south of Italy public incentives may have a greater effect on boosting cooperation than investing in human capital.

9 For example, in the disaggregation by size the Ebitda of cooperating firms always increases with respect to the not cooperating one, however the magnitude of such an increase is 32% in the case of small firms and 16% for large firms; the ROA of cooperating firms is 9.2% higher in the Science-Based sector and 6.6% in the Supplier Dominated one, and so on.

10 While the first two have been analyzed in many previous studies, few works provide empirical findings for the link between human capital and R&D partnerships (Colombo, Grilli, and Piva, 2006; Okamuro, Kato, and Honjo, 2011; Cantabene and Grassi, 2020).

11 For an analysis of the variables affecting R&D cooperation see Gussoni (2009).

12 See, inter alia, Branstetter and Sakakibara (1998, 2002), Dekker and Kleinknecht (2008), Czarnitzki, Ebersberger, and Fier (2007) and Nieto and Santamaria (2007, 2010).

13 See, inter alia Benfratello and Sembenelli (2002), Belderbos et al. (2004), Cassiman and Veugelers (2002) and Aguiar and Gagnepain (2017).

14 See, inter alia, Griffith (1999) and Harris and Robinson (2003).

15 A growing literature analyze the Italian case. Among others see: Colombo, Grilli, and Piva (2006), Colombo, Grilli, and Murtinu (2011), Antonelli and Crespi (2013), Carboni (2013), Baraldi, Cantabene, and Perani (2014), Cannone and Ughetto (2014), Magri (2014), Bronzini and Piselli (2016), and Cerulli, Gabriele, and Potì (2016).

16 Details of the survey are included in ISTAT report RS1 available at http://www3.istat.it/salastampa/comunicati/non_calendario/20041201_00/testointegrale.pdf.

17 The $\overline{T}$ of our sample is equal to 2.5. That is, we observe on average every firm for 2.5 years.

18 The appendix provides descriptive statistics and the correlation matrix.

19 Avialable at http://www2.dsu.cnr.it/relazione_ricerca_innovazione/volume/Relazione_sulla_ricerca_e_innovazione_in_Italia_webformat.pdf.

20 Chamber of Deputies is the lower house of the bicameral Parliament of Italy.

21 Avialable at https://www.camera.it/temiap/documentazione/temi/pdf/1105338.pdf?_1591284462151.

22 Avialable at https://publications.jrc.ec.europa.eu/repository/handle/JRC111327.

23 To examine the possibility of collaboration with both a private and a public partner simultaneously, we modify our definition of cooperation in Equations (1(1) $\begin{array}{rl}Coo{p}_{it}& =f(Subsid{y}_{it},PublicOrder{s}_{it},PrivateOrder{s}_{it},HumanCa{p}_{it},\\ & R\mathrm{\&}Dintensit{y}_{it},Siz{e}_{it},Locatio{n}_{it},Int\mathrm{\_}Asse{t}_{it},Cash-Flo{w}_{it},Ris{k}_{it}).\end{array}$(1) ) and (3(3) $Performanc{e}_{it}=const+\alpha Coo{p}_{it}+\theta Coo{p}_{it}\ast Su{b}_{it}+\sum \text{β}{Z}_{it}+\lambda IMR+{ϵ}_{it}$(3) ) by allowing the variable cooperation to take the value 1 for cooperation with both a public and a private entity and 0 otherwise. This does not affect the main results since most of the firms in our dataset already collaborate with both types of organization. The results are available on request.

24 The theoretical literature investigates the role of both public and private knowledge spillovers on R&D cooperation, and shows that spillovers high enough facilitate R&D cooperation. The seminal theoretical contribution is D'Aspremont and Jacquemine (1988). The empirical literature generally confirms this prediction. See, inter alia, Cassiman and Veugelers (2002) and Lopez (2008).

25 Cassiman and Veugelers (2002) show that there is a positive correlation between the cooperation activity and the incoming spillovers.

26 On the relation between size and R&D see, inter alia, Sakakibara (1997), Veugelers (1997), Fritsch and Lukas (2001), Miotti and Sachwald (2003), Franco and Gussoni (2014).

27 The control function (CF) approach relies on the same kinds of identification conditions as the standard IV methods - either two stage least squares (2SLS) or generalized method of moments (GMM). In the standard case where a endogenous explanatory variables appear linearly, the CF approach leads to the usual 2SLS estimator. But there are differences for models nonlinear in endogenous variables even if they are linear in parameters. And, for models nonlinear in parameters, the CF approach offers some distinct advantages. It is more efficient than a direct IV approach and it makes it much easier to test the null for endogeneity of the regressor.

28 We check the relevance and the exogeneity of the instruments using the Angrist-Pischke test (Angrist and Pischke, 2009) for under-identification, the F test for weak instruments (Wooldridge, 2002), and the Sargan tests of overidentifying restrictions. Results available on request

29 See, inter alia, Branstetter and Sakakibara (1998, 2002), Belderbos et al. (2004), Cassiman and Veugelers (2002), Czarnitzki, Ebersberger, and Fier (2007), Nieto and Santamaria (2007, 2010), Benfratello and Sembenelli (2002) and Aguiar and Gagnepain (2017).

30 To rule out the possibility of reverse causality in the R&D cooperation – performance relationship we apply a Granger causality test. We conclude that in the Granger sense, R&D cooperation causes performance. Looking at the other side of the relationship, we cannot reject the null hypotheses: performance does not cause R&D cooperation in the sense of Granger. Although Granger causality differs from causality, we argue that those findings corroborate our main hypothesis that R&D cooperation has an impact on performance. Results available on request. We thank a referee for this comment.

31 For a specific analysis of the determinants of R&D cooperation in Italy see Cantabene and Grassi (2019).

32 Results available on request.

33 Complete results available on request.

34 Includes traditional firms, such as textiles and agriculture which rely on sources of innovation external to the firm.

35 Includes steel and automobile industries; i.e. all the large automotive firms. Sources of innovation are usually internal to the firm.

36 Specialized firms, which sell technology to other firms (e.g. makers of high-tech instruments, machinery and computers).

37 Includes industries such as pharmaceuticals, chemical, and electronics. Firms in this sector invest massively in R&D.

38 See Nooteboom et al. (2007) and Kim and Song (2007).

39 We thank an anonymous refeee for this observation.

40 Between 2001 and 2014 the product per unit of work decreased by 5.8% in Italy compared to an increase for the European Union by 12.5% (Euro area 8.3%).

41 The analysis of causes and effects of this divergence is beyond the scope of the present paper.

42 We compare regression coefficients across the different equations. The null hypothesis is the equality of coefficients. Rejecting the null, we confirm heterogeneity in the impact of R&D cooperation on performance. The test applies only to those firms showing a significant impact of cooperation on firm performance.