central bank communication, monetary policy, ECB, exchange rates, AI, event study

We ask whether ECB communication outside monetary policy meeting days moves the EUR/USD exchange rate within minutes. We build an event study on one-minute prices and a Reuters-based corpus of 1,868 statements coded as dovish, neutral, or hawkish from 2008 to 2016. Identification combines strict exclusion windows for macro and central bank confounders, time-of-day–matched controls, and Monte Carlo resampling to test sensitivity. We also open four splits that theory suggests may matter: President, conventional versus unconventional topics, Purdah versus outside the pre-meeting window, and the regime before and during the zero lower bound. Across the full sample and every split, price responses are small and short-lived. Cumulative abnormal returns remain within a few basis points by t=+20, and scattered significant minutes are not sequential. Volatility reacts only modestly. After intraday seasonality adjustment in the matched-difference design, dovish items are associated with a brief decline in volatility in the first half hour, while neutral and hawkish items are statistically similar to controls. The contribution is twofold. First, we provide a comprehensive intraday assessment of ECB communication outside meeting days for the EUR/USD market over a consistently coded 2008–2016 window. Second, we deliver a transparent identification template for high-frequency communication research by combining time-of-day–matched controls with systematic resampling. Together, the results indicate that such communication does not generate lasting directional moves; any impact appears as small and short-lived changes in realised volatility.