Dynamic recombinant relatedness and its role for regional innovation

ABSTRACT

Previous research has argued that related variety enhances regional innovation as inter-industry knowledge spillovers occur more easily between cognitively similar industries. In this study, we engage with empirical operationalization of what is ‘related’ in related variety. We argue, based on theoretical grounds, that estimating regional knowledge production functions requires related variety measures that capture the recombination of knowledge explicitly. To test this proposition, we develop a set of related variety indicators that account for indirect linkages between industries and allow these linkages to vary over time. Empirically, we estimate the relationship between regional innovation output and regional industry mix in Swedish regions between 1991 and 2010. Our results suggest that related variety measures based on dynamic recombinant relatedness are superior in predicting regional innovation output.

KEYWORDS:

• Related variety
• relatedness
• knowledge recombination
• innovation
• network analysis
• Sweden

JEL CODES:

• L16
• O31
• R11
• R12

Disclosure statement

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

Notes

1 In the remainder of the paper, when we talk about relatedness, we imply industry linkages, while when we mention related variety, we refer to regional industry composition.

2 Since previous studies estimating relationship between innovation and regional industry composition employ related variety measures based on the hierarchical structure of official industry classifications that are not subject to change (Castaldi, Frenken, and Los 2015; Ejdemo and Örtqvist 2020; Miguelez and Moreno 2018; Tavassoli and Carbonara 2014), all variation in related variety comes from the structural change (i.e., shares of industry shares in regional employment) at the regional level. We argue that related variety measures should additionally account for changes in relatedness linkages between industries.

3 Recent literature suggests that the notion of related variety can transcend the knowledge and technological domain. For example, it has been suggested that institutional (Punt et al. 2022) and market (Chang, Eggers, and Keum 2022) relatedness is important for new knowledge generation and innovation. In this paper, we limit related variety to its knowledge component.

4 To our best knowledge, all previous studies analysing the relationship between innovation output and related variety employ this measure.

5 Top five regions here are Hofors, Säffle, Västerås, Laxå, Örnskoldsvik.

6 Technical details of calculating $\widehat{F_{ijt}}$ and determining significance of $S{R}_{ijt}^{Skill}$ are presented in A2 in the online supplement.

7 Notations are similar to those in equation (7).

8 Notations are similar to those in equation (6).

Additional information

Funding

Mikhail Martynovich’s work was supported by Jan Wallander and Tom Hedelius Foundation and Tore Browaldh Foundation (grant number W17-0016). Josef Taalbi’s work was supported by Jan Wallanders and Tom Hedelius Foundation (grant number W15-0445) and Vinnova (grant number 2020-01963). The data used in the paper was supported by Länsförsäkringar Alliance Research Foundation (through the project Regional Growth against All Odds ReGrow) and Vinnova (grant number 2017/01/011).