Abstract:

The main contribution of this research lies in identifying a crucial insight: the slow growth of new firms in local economies may be attributed to a self-sustaining mechanism characterized by volatile influx of new firms. In other words, regions with lower long-term entry rates exhibit higher relative volatility in this aspect. A similar argument can be made for exit rates as well. To categorize spatial units in economics, Machine Learning algorithms can be utilized. In this study, Turkish cities were clustered based on firm dynamics data spanning from 2009 to 2020. Through the implementation of an Unsupervised Learning (k-means) algorithm, four clusters were identified based on entry rates, while six clusters were identified based on exit rates. This approach represents an improvement over traditional methods that often require extensive manual effort to incorporate numerous socioeconomic variables into a criterion. Furthermore, it helps reduce subjectivity inherent in such methods, which heavily rely on qualitative analyses. The proposed method empowers policymakers to obtain groupings that align with their economic objectives and foster policy success.

Keywords: