The analysis of heterogeneity within a set of objects, its causes, and consequences is a key task of Data Science in social sciences. This study examines spatial heterogeneity in the relationship between population size and per capita income. We demonstrate that Arctic regions exhibit a negative correlation (per capita income decline with population growth), contrasting with the positive correlation observed in non-Arctic regions. It has been suggested that this may be related to the distinctive features in the GRP sectoral structure of Arctic territories. To examine the GRP structure’s influence, we implemented a novel algorithmic approach based on optimal industry selection for regional clustering. The developed algorithm identifies economic sectors exerting the most substantial impact on the relationship between population and per capita income across regions under study, thereby enabling deeper analysis of the features of heterogeneity hidden by significant variation in observed characteristics. Integrated regression models constructed for identified regional clusters reveal that Arctic regions displaying inverse per capita income-population relationships are characterized by disproportionate dominance of extractive industries, and underdeveloped transportation, information, and financial infrastructure, coupled with limited professional, scientific, and technical activities. These findings suggest that developmental imbalances in these sectors, combined with extreme climatic conditions, create institutional constraints on regional development. The proposed methodology enables identification of key economic sectors that most significantly influence scale effects in Arctic regions, contributing to enhanced understanding of latent heterogeneity masked by substantial variability in observable parameters.