Workplace skills as regional capabilities: relatedness, complexity and industrial diversification of regions

ABSTRACT

The literature unanimously agrees that industrial diversification is path dependent: new industries build on the pre-existing capabilities of regions that are partly embodied and reflected in the skills of the region’s workforce. This paper explicitly accounts for regional capabilities as workforce skills to build skill relatedness and skill complexity measures, that is, the skill-spaces, for 107 Italian regions for the period 2013–19. The data-driven techniques we use reveal that skill-spaces form two highly polarised clusters into social–cognitive and technical–physical skills. We find evidence that social–cognitive skills yield the highest probabilities of regional competitive advantage. A comparison with co-location-based measures shows that our measures are able to capture a substantial part of regional capabilities that cannot be measured with previous measures.

KEYWORDS:

• skill relatedness
• economic complexity
• industrial specialisation
• regional capabilities
• regional diversification

JEL:

• J24
• O18
• R10
• R23

ACKNOWLEDGEMENTS

We sincerely thank Cesar A. Hidalgo, Tiziano Razzolini and Juan A. Sanchis Llopis for their valuable comments on the earlier versions of this paper. We are grateful to the editor and two anonymous referees for their time, effort and precious comments.

DISCLOSURE STATEMENT

No potential conflict of interest was reported by the authors.

Notes

1. NUTS-3 regions known as provinces in Italy.

2. See Balland et al. (2022) for an overview.

3. See https://www.inapp.org/it/dati/ICP for more details.

4. See https://www.onetcenter.org for details.

5. The ICP 2007 uses CP 2001, while the ICP 2013 uses CP 2011.

6. Importance question: How important is this competence in carrying out your current profession? Level question: Among those indicated below, at what level is this competence necessary for the development of your current profession? Importance questions are rated on a scale from 1 (not important) to 5 (extremely important), while complexity level questions are rated on a scale from 1 (least complex) to 7 (most complex). Then they are rescaled to be between 0 and 100.

7. The ICP sample does not contain armed forces, we thus excluded these occupational categories. Legislators and senior officials are recoded due to aggregation differences between ICP and ILFS.

8. The standardisation method we used is ‘cosine’, and implemented with Econ Geo R package by Balland (2016). See VanEck and Waltman (2009) for the standardisation methods of co-occurrence data.

9. The network is visualised in Gephi software by using Multi-Scale Force Atlas, which is a force-based algorithm. The two highly polarised skill clusters are robust to different layout algorithm choices.

10. (B–E) Industry, (F) Construction, (G) Wholesale and retail trade; repair of motor vehicles and motorcycles, (I) Accommodation and food service activities, (H) Transportation and storage, (J) Information and communication, (K–N/X) Financial, real estate, scientific and technical, administrative and support service activities, (P/Q) Education; human health and social work activities, and (R–U) Arts, entertainment and recreation; other service activities.

11. We also consider a one year time lag as a robustness check. See the Appendices in the supplemental data online.

12. That is, 2013–16, 2014–17, 2015–18 and 2016–19.

13. (0.0018 × 13.62)/0.062 = 0.395. With respect to the unconditional average probability of specialising in a new industry (entry) observed in our sample, 0.062.

14. (−0.0024 × 14.44)/0.15  =  −0.231. With respect to the unconditional average probability of losing specialisation in an industry (exit) observed in our sample, 0.15.