We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:


Current browse context:


Change to browse by:

References & Citations


(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Physics > Physics and Society

Title: Relatedness in regional development: in search of the right specification

Abstract: A large body of research has documented that the size and growth of an industry in a city or region depends on the local size of related industries. However, there is no consensus on how to best measure, either the relatedness between industries, or the fit of a particular industry with an entire local economy. In this paper, we perform a structured search over tens of thousands of specifications to identify optimal -- in terms of out-of-sample prediction -- ways to construct these quantities, using a dataset that allows us to derive relatedness from co-occurrence patterns of industries in establishments, firms, regions and countries. We find that observing industry combinations at each of these different levels of aggregation leads to different relatedness matrices, each of which can help predict the size and growth of local industries. Finally, we identify specification elements that improve predictive performance, highlight trade-offs between predictive performance and interpretability of relatedness patterns, and offer some guidance for how to use these patterns to support policy-making.
Subjects: Physics and Society (physics.soc-ph); Social and Information Networks (cs.SI)
Cite as: arXiv:2205.02942 [physics.soc-ph]
  (or arXiv:2205.02942v1 [physics.soc-ph] for this version)

Submission history

From: Yang Li [view email]
[v1] Thu, 5 May 2022 21:53:03 GMT (4789kb,D)

Link back to: arXiv, form interface, contact.