Using Dynamic Convergence Clubs, Weak Convergence, and Machine Learning to Address Economic Convergence in the EU

This study employs machine learning-based clustering to analyse economic growth patterns among several EU countries, using Germany as a benchmark. It introduces the concept of “weak convergence”, demonstrating that while economies cluster and shift over time, they maintain a stable long-run equilibrium relationship. This finding is crucial, as it suggests that poorer countries are not necessarily accelerating toward wealthier ones but instead persist in a relatively large gap. Although the disparities between clusters appear to narrow – driven by the slowdown of richer economies and the gradual catch-up of poorer ones – this pattern does not necessarily indicate true convergence. Cointegration analysis reveals that these shifts primarily result from ongoing macroeconomic adjustments influenced by trade flows and external shocks, such as the 2007-08 financial crisis. While these shocks disrupted existing economic relationships, they also led to a reduction in between-cluster variance over time, creating a form of pseudo-sigma convergence. However, this convergence was largely driven by the relative decline of richer economies rather than a substantial acceleration of poorer ones.