This paper examines the econometric performance of regime-switching models for interest rate data from the US, Germany, and the UK. The authors find strong evidence supporting the presence of regime switches but note that univariate models are unlikely to yield consistent estimates of the model parameters. Regime-switching models that incorporate international short rate and term spread information perform better in forecasting, match sample moments better, and classify regimes more accurately than univariate models. The regimes in interest rates correspond reasonably well with business cycles, particularly in the US. The non-linear interest rate dynamics implied by regime-switching models have important implications for the macroeconomic literature on the effects of monetary policy shocks on economic aggregates. The implied volatility and drift functions are similar to those estimated using non-parametric techniques. The paper also introduces a new metric, the Regime Classification Measure (RCM), to evaluate the quality of regime classification. The results suggest that regime-switching models can capture the non-linearities in interest rates and improve forecasting accuracy.This paper examines the econometric performance of regime-switching models for interest rate data from the US, Germany, and the UK. The authors find strong evidence supporting the presence of regime switches but note that univariate models are unlikely to yield consistent estimates of the model parameters. Regime-switching models that incorporate international short rate and term spread information perform better in forecasting, match sample moments better, and classify regimes more accurately than univariate models. The regimes in interest rates correspond reasonably well with business cycles, particularly in the US. The non-linear interest rate dynamics implied by regime-switching models have important implications for the macroeconomic literature on the effects of monetary policy shocks on economic aggregates. The implied volatility and drift functions are similar to those estimated using non-parametric techniques. The paper also introduces a new metric, the Regime Classification Measure (RCM), to evaluate the quality of regime classification. The results suggest that regime-switching models can capture the non-linearities in interest rates and improve forecasting accuracy.