Resilience, Skill Endowment, and Diversity: Evidence from US Metropolitan Areas

Abstract

This article investigates the role of technological, industrial, and human capital composition in shaping short-term regional resilience in the wake of the Great Recession of 2008. Using data on 295 US Metropolitan Statistical Areas over the period 2008–14, we find that the most resilient regions feature a very diversified industrial structure. At the same time, an excess of technological diversity thwarts the ability to absorb external shocks. Lastly, a high endowment of high-level abstract skills is positively correlated with regional resilience, though the moderating effect of technological diversity is negative.

Key words:

• regional resilience
• human capital
• technological diversity
• industrial diversity

JEL codes:

• O18
• O3
• O51
• R11

Acknowledgments

We are grateful to the editor in chief and two anonymous referees for constructive comments. Fabrizio Fusillo and Francesco Quatraro gratefully acknowledge the funding of the Italian Ministry of University and Research, within the context of the PRIN research project “Innovation for global challenges in a connected world: The role of local resources and socio-economic conditions” [Ref. PRIN 20177J2LS9]. Davide Consoli acknowledges financial support of Ministerio Ciencia e Innovación (PID2020-119096RB-I00/MCIN/ AEI /10.13039/501100011033). The usual caveat applies.

Supplementary material

Supplemental data for this article can be accessed here.

Notes

¹ See OMB report of 2010, https://www.govinfo.gov/content/pkg/FR-2010-06-28/pdf/2010-15,605.pdf.

² We are grateful to Dieter Kogler and the UCD Spatial Dynamic Lab, Dublin, for providing us with the fully regionalized USPTO patent data set.

³ The difference between these two illustrative cases is documented in various studies. On the one hand, Pittsburgh’s rich manufacturing legacy paved the way to the highly specialized manufacturing clusters—some of which exhibit employment concentration more than twice the national average such as automation, industrial machinery, and metal processing (Andes et al. 2017)—that grew significantly in the aftermath of the recession. On the other hand, the Boston metro area stands out as one of the most exposed to the implosion of the housing mortgage bubble of 2008 (Martin 2011; Glaeser and Nathanson 2014) that entailed severe repercussions in terms of diminished access to credit for businesses and higher inequality, both widely believed causes of the sluggish recovery (De Nardi, French, and Benson 2011; Brown et al. 2013).

⁴ It is worth stressing that, even though the decomposition into within—and between—variety (known in the economic geography literature as related variety and unrelated variety [Frenken, Van Oort, and Verburg (2007)]) of the entropy-based diversity index can partly take into account the disparity attribute—by defining two sets with different disparity (Krafft, Quatraro, and Saviotti 2011)—these entropy-based measures better account for the number of distinct categories and the evenness of the distribution (Stirling 2007).

⁵ Information on how patents have been assigned to the MSA and further details are presented in Boschma, Balland, and Kogler (2015).

⁶ The CPC is a new patent classification system, jointly developed by the European Patent Office (EPO) and the USPTO. Based on the European classification system, it is a more detailed version of the International Patent Classification.

⁷ It is worth stressing that, to be consistent with the classification of industrial sectors of the input-output matrices we converted the NAICS industrial codes available in the CBP data to the corresponding input-output sectors using the tables of the US BEA. A total of seventy-one different industrial sectors have been identified.

⁸ The share of employment in manual jobs (MSH_it−1) represents the reference category.

⁹ We calculate the patent stock measure using the permanent inventory method (Bloom and Van Reenen 2002; Peri 2005; Dechezleprêtre, Neumayer, and Perkins 2015): the stock of patent for each MSA i at time t is given by the net amount of patent applications in year t plus the depreciated patent stock at time t − 1. In line with prior literature, we set the obsolescence rate at 15 percent (Keller 2002; Hall, Jaffe, and Trajtenberg 2005; McGahan and Silverman 2006; Nesta 2008, among others).

¹⁰ Entropy, RV, and UV are calculated by applying the information entropy index, and its decomposition, to the local distribution of establishments across two- and three-digit NAICS sectors.