Is the European Monetary System converging to integration?

The emerging system at the European level can be conceptualized as a pattern of relations among member states that tends to be reproduced despite disturbances in individual trajectories. The Markov property is used as an indicator of systemness in the distribution. The individual trajectories of nations participating in the European Monetary System is assessed using an information theoretical model that is consistent with the Markov property in the multivariate case. Economic and monetary integration are analyzed using independent data sets. Increasing integration can be retrieved in both of these dimensions, notably since the currency crises of 1992 and 1993. However, the dynamics for countries which have strongly coupled their currency to the German Mark are different from those which did not. Additionally, developments in inflation and exchange rates at the European level are assessed in relation to global developments.

💡 Research Summary

The paper investigates whether a genuine European monetary system is emerging from the collection of national economies, using the Markov property and information‑theoretic measures as indicators of “systemness.” The authors treat each country as a node in a network of exchange‑rate relations and ask whether the pattern of these relations persists over time despite disturbances at the individual‑country level. To answer this, they construct two parallel data sets: (1) nominal exchange rates among EMS members (the “political” dimension, reflecting the Maastricht convergence criteria) and (2) real effective exchange rates (REER) together with consumer‑price‑index data (the “economic” dimension, reflecting purchasing‑power‑parity).

Shannon’s entropy and Theil’s dynamic information measure are extended to the multivariate case. For each year the normalized distribution of exchange‑rate or price variables across countries is taken as a probability vector p. The actual distribution in the following year is q. The information gain I = Σ q log(q/p) quantifies how much the observed change deviates from a given prediction. Two competing predictions are generated: (a) a univariate autoregressive forecast for each country taken separately, and (b) a multivariate forecast that assumes the whole system follows a first‑order Markov chain, i.e., the current distribution alone determines the next one. If the multivariate prediction yields a lower expected information (positive I when compared with the univariate case), the system exhibits Markov‑consistent “systemness.”

Applying this framework to data from 1979 to 1996, the authors find that after the currency crises of 1992‑93 the nominal‑exchange‑rate distribution aligns closely with the Markov prediction, indicating a strengthening of systemness. Countries tightly coupled to the German Mark (Germany, the Netherlands, Belgium) form a core sub‑network with very narrow fluctuation bands, while the United Kingdom, Denmark, Sweden and others display wider bands and contribute less to the emerging system. In the economic dimension, REERs gradually converge toward a common mean and inflation differentials shrink, suggesting a parallel but slower process of economic integration.

The study demonstrates that political convergence (fixed nominal rates) and economic convergence (real rates and price levels) can be assessed independently using the same information‑theoretic machinery. It also shows that traditional cointegration or PPP tests, which focus on pairwise relationships, miss the multivariate network dynamics captured by the Markov‑based approach. The authors conclude that the European Monetary System has indeed become more integrated in both dimensions, yet heterogeneity among member states remains. Full monetary union therefore still requires additional structural adjustments and coordinated policy to bring all countries into the same systemic regime.