Spatial determinants of inventive capacity in Brazil: the role of inventor networks

ABSTRACT

This paper investigates the role of inventor networks in regional inventive capacity in Brazil for 558 regions in the period 2000–11. Based on the knowledge-production function (KPF), it focuses on four network measurements: intra- and interregional collaborative inventor links, density of the network, and international collaboration among Brazilian and foreign inventors. The main results show that intra-regional inventive collaborations exert a positive impact, while denser networks negatively influence regional inventive productivity. Interregional collaborative links have a ‘U’-shape relationship with inventive productivity only for South and Southeast. Inventions in the North, Northeast and Centre–West rely on international collaborative links.

KEYWORDS:

• collaborative networks
• Brazilian inventors
• co-patenting activity
• spatial panel data

JEL:

• C31
• O31
• O33
• R1

AKNOWLEDGEMENTS

We thank Pedro Vasconcelos Maia do Amaral for helpful comments on previous version of this paper as well as all comments and suggestions of the two anonymous referees and Professor Paul Elhorst. We are also grateful to INPI team by the patent database.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

ORCID

Eduardo Gonçalves http://orcid.org/0000-0003-2017-3454

Priscila Medeiros de Oliveira http://orcid.org/0000-0002-6691-7503

Eduardo Almeida http://orcid.org/0000-0003-0045-1672

Notes

1 Models 1 and 2 include dummy variables to control the regions without inventive connections (DUM_nlig). Similarly, model 3 controls the regions without international collaborations (DUM_outBRA).

2 A micro-region is a group of municipalities used for statistical purposes by the Brazilian Institute of Geography and Statistics.

3 For the usual limitations and advantages of patent statistics to represent invention or innovation, see Nagaoka, Motohashi, and Goto (2010).

4 Araújo, Cavalcante, and Alves (2009) propose the creation of a variable called ‘technical and scientific personnel’ (PoTec) that has an approximate 90% correlation between aggregate R&D measures and PoTec.

5 The predominance of non-residents in Brazilian patent applications is a feature that distinguishes innovation systems from Brazil and Korea, for example. In Korea, only 22% of deposits were made by non-residents, while in Brazil residents accounted for only 15.78% of total applications at INPI in 2012 (Chiarini, Rapini and Silva, 2017).

6 Descriptive statistics (see Table A2 in Appendix A in the supplemental data online) on more developed regions in Brazil, located in the South and Southeast, confirm higher means for collaboration variables in comparison with regions in the North, Northeast and Centre–West.

7 Defined through the median of the series of total collaborations, that is, the 279 micro-regions with a higher number of total collaborations. The same can be said for micro-regions with a lower number of collaborations.

8 The econometric analyses shown in the following tables only present models of fixed effects with a spatial lag model (SAR-FE) because the Hausman test rejects the null hypothesis that random effects are more appropriate and the CD test of Pesaran (2004) rejects the hypothesis independence of cross-sectional units.

9 To compare the coefficients and measurements that were originally on different scales, the data underwent standardization, that is, all the variables were rescaled to have a mean of 0 and a standard deviation (SD) of 1.

10 The correlation matrix of explanatory variables does not indicate multi-collinearity issues in the regressions (see Table A3 in Appendix A in the supplemental data online).

11 To ensure the reliability of the results, all three models were estimated by changing the inverse-square distance matrix used here for binary first-order contiguity and inverse distance matrices (see Table A4 in Appendix A in the supplemental data online).

12 Table 4 does not show models 1 and 2 with regard to the direct and indirect impacts for South/Southeast and North/Northeast/Centre–West regions. For the regressions, see Table A5 in Appendix A in the supplemental data online.

Additional information

Funding

The authors gratefully acknowledge the financial support from the National Scientific and Technological Development Council – CNPq (306057/2015-8); the Foundation to Support Research for the State of Minas Gerais – FAPEMIG (APQ-03151-16); and Coordination for the Improvement of Higher Education Personnel – CAPES (001).