Canonical correlation complexity of European regions

Önder Nomaler & Bart Verspagen


In an earlier paper (Nomaler & Verspagen, 2022) we introduced a 'supervised learning' based alternative to the competing unsupervised learning algorithms (e.g., Hidalgo and Hausmann, 2009 vs. Tacchella et al, 2012) proposed in the so-called 'economic complexity' literature. Similar to the existing ones, our alternative, which we refer to as the "Canonical Correlation Complexity Method (CCCM)", also aims at reducing the high dimensionality in data on the empirical patterns of co-location (be it nations or regions) of specializations in products or technologies, while the ultimate objective is to understand the relationship between specialization, diversification, and economic development. In our alternative method which combines the toolkit of the Canonical Correlation Analysis with that of Principal Component Analysis, the data on trade or technology specializations and multiple dimensions of economic development are processed together from the very beginning in order to identify the patterns of mutual association. This way, we are able to identify the products or technologies that can be associated with the level or the growth rate of per capita GDP, and (un)employment. In this follow up paper, we use the CCCM to analyse the development patterns of European regions in relation to their respective technology specializations. Our findings provide insights for EU's industrial policies, especially those considered under the 'smart specialization' framework.

Keywords: Economic complexity, economic development, supervised learning, Canonical Correlation Analysis, Principal Component Analysis, technological specialization, revealed technological advantage, European regional development, smart specialization

JEL Classification: F14, F63, O11, R11

Download the working paper