The article presents the discussion regarding the influence of taking selected approaches to the classification of regions on estimation results of Solow–Swan growth models (case of convergence processes). As the existing modelling effects, obtained in the area of regional development, show the estimation results of the production function parameters’ assessment, based on Solow–Swan growth model or beta-convergence models, are divided depending on the period and units covered by the study (countries, regions). The article presents modelling results of the development of economies in the European Union regions at NUTS 2 level, depending on the development factors in line with the extended MRW (Mankiw–Romer–Weil) growth model in the selected period for various groups of regions. Each time the econometric analysis was conducted in the groups of NUTS 2 regions, separated in terms of the level of smart specialization. However, every time, along with the change in the grouping method, the dividing boundary of the inclusion of regions into a given group was changing. It allowed the assessment of estimations stability depending on the inclusion of the selected regions into particular groups of smart specialization.