United in Currency, Divided in Growth: Dynamic Effects of Euro Adoption

TL;DR

This paper investigates whether euro adoption influences long-run economic growth. It introduces Causal Forests with Fixed Effects (CFFE) to estimate dynamic, heterogeneous treatment effects in a panel with staggered euro adoption across 20 eurozone and 15 control economies (1970–2023). The main finding is a modest average annual growth reduction of about $0.3$–$0.4$ percentage points, with effects emerging at adoption and persisting for roughly a decade, plus substantial heterogeneity by initial income; periphery countries bear larger losses, while core economies are less affected. Mechanisms point to declines in consumption and productivity, partially offset by improved net exports, and a structural New Keynesian DSGE model with hysteresis supports the view that a one-size-fits-all monetary union can generate larger and more persistent output losses in the periphery. Overall, the study contributes to monetary integration literature by jointly estimating dynamic, heterogeneous effects and linking results to OCA theory and macroeconomic mechanisms, while highlighting substantial uncertainty due to a small treated-sample and complex general equilibrium dynamics.

Abstract

Does euro adoption affect long-run economic growth? Existing evidence is mixed, reflecting limited treated countries, long horizons that challenge inference, and heterogeneity across member states. We estimate causal dynamic and heterogeneous treatment effects using Causal Forests with Fixed Effects (CFFE), a machine-learning approach that combines causal forests with two-way fixed effects. Under a conditional parallel-trends assumption, we find that euro adoption reduced annual GDP growth by 0.3-0.4 percentage points on average. Effects emerge shortly after adoption and stabilize after roughly a decade. Average effects mask substantial heterogeneity. Countries with lower initial GDP per capita experience larger and more persistent growth shortfalls than core economies. Weaker consumption and productivity growth contribute to the overall effect, while improvements in net exports partially offset these declines. A two-country New Keynesian DSGE model with hysteresis generates qualitatively similar patterns: one-size-fits-all monetary policy and scarring mechanisms produce larger output losses under monetary union than under flexible exchange rates. By jointly estimating dynamic and heterogeneous treatment effects, the analysis highlights the importance of country characteristics in assessing the long-run consequences of monetary union.

Figures (15)

• Figure 1: Dynamic Treatment Effects of Euro Adoption (CFFE). Notes: Figure shows estimated treatment effects $\hat{\tau}(k)$ by event time $k$ (years since euro adoption). Shaded area represents 95% confidence intervals based on cluster-robust standard errors at the country level. Sample includes 19 eurozone countries and 15 controls, 1970--2023.
• Figure 2: Comparison of CFFE and Classical Event Study Estimates. Notes: Figure compares treatment effect estimates from CFFE (solid line) and classical two-way fixed effects event study (dashed line). Shaded areas represent 95% confidence intervals. Classical standard errors are clustered at the country level.
• Figure 3: Heterogeneous Effects by Initial Income. Notes: Figure shows estimated treatment effects separately for high-income founders (above-median GDP per capita in 1995: Austria, Belgium, Finland, France, Germany, Luxembourg, Netherlands) and low-income founders (below-median: Greece, Ireland, Italy, Portugal, Spain). Classification based on pre-treatment characteristics to avoid post-treatment bias. Shaded areas represent 95% confidence intervals.
• Figure 4: Heterogeneous Effects: Early vs. Late Adopters. Notes: Figure shows estimated treatment effects separately for 1999 founders (11 countries) and later adopters (8 countries with sufficient post-treatment data). Shaded areas represent 95% confidence intervals. The wide confidence intervals for late adopters reflect small sample sizes and should be interpreted cautiously.
• Figure 5: Mechanism Analysis: Effects on GDP Components and Productivity. Notes: Figure shows estimated treatment effects on consumption growth, investment growth, net exports (as % of GDP), and productivity growth. Shaded areas represent 95% confidence intervals. Investment effects shown separately for full sample and pre-2008 sample.