We propose an econometric procedure for obtaining de facto exchange rate regime classifications which we apply to study the relationship between exchange rate regimes and economic growth. Our classification method models the de jure regimes as outcomes of a multinomial logit choice problem conditional on the volatility of a country's effective exchange rate, a bilateral exchange rate and international reserves. An `effective' de facto exchange rate regime classification is then obtained by assigning country-year observations to the regime with the highest predictive probability obtained from the estimation problem. An econometric investigation into the relationship between exchange rate regimes and GDP growth finds that growth is higher under stable currency-value regimes. Significant asymmetric effects on country growth from not doing what is said are found for nonindustrialized countries. Countries that exhibit `fear of floating' experience significantly higher growth.