Economy of Regions | 2026
Authors: Badykova I.R.
DOI: 10.17059/ekon.reg.2026-1-8
Journal: Economy of Regions
Year: 2026
Publisher: Institute of Economics, Ural Branch of the Russian Academy of Sciences
Document Type: Article
Open Access: All Open Access; Gold Open Access; Green Open Access
Cited by: 0
Spatial inequality among Russian regions remains a key issue of the national economy, driving sustained academic interest in the determinants of gross regional product (GRP). Despite extensive research, traditional econometric methods often fail to fully capture the complex non-linear relationships, time lags, and synergistic effects between growth factors. The aim of this study is to identify and comprehensively analyse the key determinants of per capita GRP in Russian regions by applying advanced machine learning (ML) methods to overcome the limitations of classical approaches. The empirical base comprises panel data for 85 Russian regions from 2013 to 2023. To predict GRP, 10 initial indicators were selected and grouped into thematic blocks: labour resources, investment, and production potential. A critical step was the creation of derivative features (lags and moving averages) and dummy-variables for regions and years. An ensemble of ML algorithms was used to build the predictive model, with the Light Gradient Boosting Machine algorithm showing the highest performance (R² = 0.7345). The results were interpreted using SHAP analysis and elasticity calculations. The results revealed the absolute dominance of population income indicators, particularly their lagged values and moving averages, confirming the hypothesis of cumulative and inertial growth. The second most significant factor was foreign direct investment, which also exhibited a time-lagged effect. The analysis confirmed all three hypotheses: the predominant influence of lags, the importance of synergy (strongest between income and wages), and the enhanced explanatory power of the model with derivative features. The findings may assist regional authorities in prioritizing sustainable income growth and strategic investment policies, accounting for lagged effects. A limitation of the study is its dependence on the quality of official statistics; future research could incorporate qualitative institutional and socio-cultural indicators. © Бадыкова И. Р. Текст. 2026.
foreign direct investment; gradient boosting; gross regional product; machine learning; population income; regional economics; SHAP analysis; time lags