From average to extremes: Application of archetypal analysis in economic geography

Abstract

This article introduces Archetypal Analysis (AA) to economic geographers. AA is a novel unsupervised learning method that identifies and analyses outliers in multivariate datasets. Unlike conventional clustering methods focusing on the average, AA highlights extreme cases and represent each data point as convex combinations of extreme points. This method offers a needed analytical tool for recent economic geography research efforts studying the key drivers of success against all odds, like green transition in peripheral regions or poor outcomes like regional left-behindness. The article showcases the applicability of AA by creating a typology of European regions’ technological specializations in clean and dirty technologies. We provide open access to an R script to facilitate the adoption of AA in future economic geography research.