Personal and business networks within Chilean biotech ^§

^§A previous version of this research was part of Chapter 6 of the PhD Thesis ‘Social Networks, Collaborations and High-Tech Cluster Formation in an Emerging Country: The Case of Biotechnology in Chile’ submitted at The University of Oxford.

Abstract

Agglomerations and proximities between firms are positively related to the creation of knowledge networks and the innovation performance of firms. Despite the long history of the study of proximities, only recently have interactions between multiple proximities and multiple networks been studied with respect to the economic performance of clustered firms. The role of personal relations in the creation of business relations between firms needs to be studied further. This study proposes a framework for the creation of business relations that combines the proximity framework, the strategic alliances literature, and multiplexity of social networks. The study uses a novel dataset on the ego-networks of personal relationships between owners or managers and the ego-networks of business relations of biotech firms in Chile. It is shown that the existence of previous personal relations between individuals from two organisations positively affects the probability of creating a business relationship between them.

Keywords:

• Personal relations
• business relations
• multiplexity
• biotech clusters
• multiple proximities

Acknowledgements

The author is grateful to Prof. Xiaolan Fu, the editors of this special issue, and two anonymous reviewers for their helpful comments.

Notes

^§ A previous version of this research was part of Chapter 6 of the PhD Thesis ‘Social Networks, Collaborations and High-Tech Cluster Formation in an Emerging Country: The Case of Biotechnology in Chile’ submitted at The University of Oxford.

¹ Biotechnology is defined as ‘the application of science and technology to living organisms as well as parts, products and models thereof, to alter living or non-living materials for the production of knowledge, goods and services’ (OECD 2009, 9).

² It is not intended to provide a detailed review of this aspect of the literature. For detailed reviews on the subject, see Knoben and Oerlemans (2006) and Balland, Boschma, and Frenken (2015).

³ For example, many owners of the first generation of biotech firms in Chile continued to teach in different universities. Their former students later created their own companies. These interactions in the academic context would have been likely to inform both parties about the reliability of the other.

⁴ A logistic distribution of the errors is assumed in the estimation procedure. Estimations were conducted using STATA 13.0.

⁵ See, for example, data from The Enterprise Survey (http://www.enterprisesurveys.org/), the World Value Survey (http://www.worldvaluessurvey.org/), and the OECD Innovation Scoreboard (http://www.oecd.org/sti/scoreboard.htm).

⁶ Asociación Chilena de Empresas de Biotechnología.

⁷ Some business relations show the direction of the link – supplier and client – yet other relations, such as research partners, could be thought of as undirected. In this study, all types of relations between firms will be considered undirected links.

⁸ This study focuses on DBFs since within this group the data has high nodal and relational response rates. Results will be representative of DBF firms, and their links with any organisation of the biotech sector in Chile for which I have full information will be considered for the analysis.

⁹ In the cases where the distance was zero, it was replaced by a value of 200 m.

¹⁰ Interviewees were asked directly for the level of sales in 2012. In cases where they did not respond, they were presented with nine categories of sales (with 1 the lowest level of sales, identifying micro-enterprises, and 9 the highest, identifying large firms, according to the classification of the National Statistics Institute in Chile). This categorical variable was selected to maximise the number of observations.

¹¹ The different biotechnology areas follow the OECD (2009) classification. They are: Health and Diagnosis; Food and Nutrition; Aquiculture and Marine Biology; Agriculture; Dry Zones Biotechnology; Patents and Publications; Bioindustry; Classic Fermentation and Bioprocesses; Bioinformatics and Nano-biotech.

¹² Alternative variables were tested: a variable for the difference in biotech areas, and a binary variable indicating if the organisations are in the same biotech area. However, to include the areas for both organisations allows testing for different effects between internal and external knowledge bases.

¹³ Different measures of R&D capacities were tested as variables, such as dummies for in-house R&D, natural logarithm of researchers, natural logarithm of R&D expenditures, share of expenditures on R&D over sales, share of researchers, among others. The variable chosen had the least missing values.

¹⁴ It is not possible to test for the different effects of variables for each type of business relation, since there are too few positive cases or effective relations in the sample.

¹⁵ The correlation matrix (available upon request) shows that most ego variables are intercorrelated, as well as alter variables, while correlations between ego and alter variables are low, never higher than 0.47, ruling out a potentially serious problem of multicollinearity in the estimation.

¹⁶ It was not possible to compare estimations between intra-and extra-clusters relations, since the number of positive cases of business and personal relations outside the clusters were low for estimating a differentiated model.

¹⁷ Several specifications were estimated, but only the most relevant models are presented. Other specifications with random slopes of ego variables were estimated, but these results and models did not perform better. The results of these models are available upon request.

¹⁸ The results of a chi-squared test between the multilevel logistic model and a logistic regression (without considering the clustered nature of the data), show that multilevel estimations are preferable given the data structure.

¹⁹ There is no direct way of testing endogeneity in the cases where the dependent variable and the endogenous variable are dichotomic. Following Nichols (2011), a bivariate probit was estimated and testing for endogeneity was approximated by testing for the significance of the covariance of the errors of both equations.