Complexity and Specialization in Clusters of European Automotive Regions: Mapping the Potential in Electric Transition

Type de publication:

Conference Paper


Gerpisa colloquium, Brussels (2023)


complexity, European automotive regions, patents, Specialisation


The transition to electric vehicles is challenging the European regions specialized in the automotive global value chain. A group of such regions have established an Alliance (EU Automotive regions Alliance) calling for ad hoc policy measures to support their transition, to ensure the employment in the regions and the potential of growth, so far made possible by the development of automotive industry in their regions. Different potential is referred to the R&D centers present in the countries (Mordue and Sweeney 2020; Pavlínek 2022), but so far no analysis has explored the cross regional differences in terms of the technological knowledge in a region. In this paper we investigate the potential of automotive EU regions with respect to the technological knowledge needed for the development of paths of new productions to be embedded in the electric vehicles.
Identifying automotive regions
Initially, automotive regions in the EU-27 and UK are identified using Revealed/relative Comparative Advantage (RCA also known as Balassa Index) based on employment data (Eurostat, 2018). Employment data is considered on NUTS2 level and NACE 29 (henceforth C29) is considered as classification of economic activity related to the automotive industry. Missing values in employment data are set to zero to compute the RCA.
Revealed comparative advantage (RCA)/Balassa Index in region i at time t is the employment (L) share in sector k (here C29) at time t in region i, divided by the employment share in sector k at time t across all regions: Regions are identified as "automotive regions", if their RCA is above 1:
Based on RCA, we identify 75 regions specialised in C29.
Problems. The RCA is more a measure of dependence of automotive industry and specialization than a measure of the presence of a large automotive industry. In this way we would include small regions like Cantabria (about 3000 employees in C29) and would exclude Emilia-Romagna (about 17000 employees in C29 and some other important manufacturing and services specialisations).
Alternative ranking. Number of employees in C29, by region. Regional employment L_(i,t) in C29 higher than 3rd quartile value (8563.5) of regional employment considering all NUTS2 regions in EU27 + UK
Based on this approach, 76 regions are identified as automotive regions for 2018.
We decided to combine the two measures and in this way 89 regions have been identified. Map in Figure 1 combines these two measures and indicates by which one (or both) a region is identified.
Figure 1 – Map of EU regions specialized in C29, based on RCA and 3rd quartile on employment share
62 regions are identified with both measures
89 regions are identified in total
13 only by RCA
14 only by above 3rd Quartile
Table 1 Regions by country
All map for all 89 automotive regions:
Knowledge Complexity of automotive regions
We first compute the knowledge complexity index of the automotive regions by using patent data.
Referring to (Hidalgo and Hausmann 2009; Hausmann et al. 2011), we apply the method of Reflections as indicators of economic complexity (first RCA/Balassa then using network structure of patents and regions to compute the complexity measure). Observed levels of knowledge diversification of a region is provided by the number of patent classes in which a region possesses a RCA; ubiquity of a patent class is computed as the number of regions patenting in that class. We apply this measure of complexity of automotive regions in the period 1990-2019, and by periods of 5 years (patents filed to EPO in each period).
Patent data. Source: OECD REGPAT (February 2022); Patent applications filed to the EPO (1977-2022); CPC classes at 3 digits (version 2023.02); fractional counts by inventor share (location of inventor in NUTS3 region) aggregated to NUTS2; EPO application year due to longer period available. Time frame under analysis: overall period: 1990-2019, and six five years sub periods.
We use relative technology advantage (RTA) in patenting to cluster automotive regions by their knowledge base in the overall period and in the six sub-periods
RTA by NUTS2 region is computed using patents on 3-digit level. Cluster of regions by their RTA in 128 technology fields is computed by implementing the Ward method, with Euclidean distance.
In the present version of the analysis, results refer to the overall period 1990-2019.
Figure 2 – Dendogramm of the clusters of EU-27 & UK automotive regions: 8 clusters
Figure 3 – Map of the 8 clusters of EU-27 & UK automotive regions

Cluster characteristics are summarized with respect to GDP per capita, population (cumulated and average), population density, technology stock (number of patents in region over 30 years and over last period), complexity index.
Figure 4 – Regions by cluster in 1990-2019 vs. sub-period 6 (2015-2019): sankey diagram
First of all, we identify the most characteristic technology classes of each cluster and we observe that a pattern emerges across the 8 clusters of automotive regions. We check for changes occurring in the various sub periods, computing clusters of automotive regions in each region.
Figure 5 The most characteristic technology classes (top ten RTA) of each cluster of automotive regions. Period: 1990-2019
We then focus on the outlook of regions in each cluster. In particular, we compute the relatedness of technological classes of regions with respect to two technological domains as defined by USPTO that are relevant in "climate change mitigation … related to transportation" (CPC class: Y02T), encompassing patents classified in code Y02T 10/00: Road transport of goods or passengers (ICE improvement, hybrid vehicles, aerodynamics optimization etc.); and code Y02T 90/00: Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation (electric vehicles charging stations etc.)

  GIS Gerpisa /
  4 Avenue des Sciences, 91190 Gif-sur-Yvette

Copyright© Gerpisa
Concéption Tommaso Pardi
Administration Juan Sebastian Carbonell, Lorenza MonacoGéry Deffontaines

Créé avec l'aide de Drupal, un système de gestion de contenu "opensource"