Necessary and sufficient conditions for the absorptive capacity of firms that interact with universities

ABSTRACT

A complex concept, absorptive capacity was first introduced by Cohen and Levinthal (1989, 1990) and is connected to the Penrose theory of growth of firms ([1959] 2009) as it is related to a variety of firms’ internal resources and external aspects of their environment, such as sources of knowledge. Considering firms that interact with universities in the search for external knowledge, this paper aims to identify the necessary and sufficient conditions of these firms to reach certain levels of absorptive capacity. We conducted a survey of firms that interacted with engineering research groups at universities in the state of Rio Grande do Sul (Brazil). We applied the Fuzzy Sets Comparative Qualitative Analysis. The main results were: (1) the necessary and sufficient conditions for high-level absorptive capacity are a combination of high acquisition and assimilation and medium transformation and exploitation; (2) the necessary and sufficient conditions for medium or non-high absorptive capacity are obtained by medium levels of acquisition, assimilation, transformation and exploitation; and c) that there are a variety of paths to improving the absorptive capacity of firms. These findings are important to formulating more diverse, flexible and less costly public policies aimed towards improving firms’ absorptive capacity.

KEYWORDS:

• firm absorptive capacity
• dimensions of absorptive capacity
• university-industry interaction
• qualitative comparative analysis
• fuzzy sets

Acknowledgements

The authors would like to thank the anonymous referees for their helpful comments. We are also grateful for the support received from Luisa Hadres dos Santos and Júlia Driemeier Vieira Rosa, students of UNISINOS. The usual disclaimer applies.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1. For further details, consult the following publications: (Jansen, Van Den Bosch, and Volberda 2005; Torodova and Durisin 2007; Van Den Bosch, Van Wijk, and Volberda 2003; Versiani et al. 2010; Camisón and Fóres 2010; Mowery and Oxley 1995; Kim 1998; Lane and Lubatikin 1998; Dyer and Singh 1998; Van Den Bosch, Volberda, and De Boer 1999; Zahra and George 2002; Lane, Koka, and Pathak 2006).

2. With this population and sample, the margin of error was slightly high by conventional standards. On the other hand, the outcome of the Cronbach´s Alpha test, which assesses the consistency of the questionnaire, is in the acceptable range (Rosa 2013). The sample average is accepted as representative of the population.

3. This consistency formula is applied to the Fuzzy Truth Table Algorithm of the fsQCA (Ragin 2006) used in processing the combinations sufficient to obtain High Absorptive Capacity, Medium Absorptive Capacity, Low Absorptive Capacity and Non-High Absorptive Capacity. Therefore, the consistency and the respective coverage refer to the sufficiency of the explanatory combinations. The consistency and coverage formulas for the necessary conditions are slightly different. Due to limited space, they are not reproduced here and can be found in Ragin (2006).

4. Likert has characteristics of an ordinal scale. There are arguments in favor of the use of summated scales of the Likert scale as an interval scale for partitions of five points or more and continuous constructs (Curado, Vitorino Teles, and Marôco 2013, 449). The sum of scores and averages of Likert scale scores is adopted on an interval scale.

5. Ragin (2006) points out that for consistency values below 0.75 it is difficult to sustain that there is a relationship between the evaluated sets, i.e. that one is a subset of the other.

6. Schneider and Wagemann (2010) recommend great rigor to establish necessary conditions. In this work the diversity in comparison to the original work was improved considerably (Rosa 2013; Rosa and Ruffoni 2014) by proposing twelve antecedent sets. This resulted in 3^12-1 = 531.440 combinations to be tested. To run the necessary algorithm of the fsQCA, the combinations of sets are chosen manually, thereby considerably increasing the risk of making mistakes. Consequently, we adopted a more ‘lax’ procedure and ran the necessity test after the sufficiency test, inserting only sets that are included in the sufficient combinations of antecedent categories. This procedure influenced the consistency index in a way that makes it easier to find necessary conditions. With this in mind, we enhanced the consistency level from 0.8 to 0.95.

Additional information

Funding

This work was supported by the Edital Pró-Publicações Internacionais 051/2019, Universidade Federal de Santa Maria (Pro-International Publications Notice 051/2019, Federal University of Santa Maria), registered under the protocol 043/2019.