Regime-switching models are well suited to capture the non-linearities in interest rates. This paper examines the econometric performance of regime switiching models for interest rate data from the US, Germany and the UK. There is strong evidence supporting the presence of regime switches but univariate models are unlikely to yield consistent estimates of the model parameters. Regime-switching models incorporating international short rate and term spread information forecast better, match sample moments better, and classify regimes better than univariate models. We show that the regimes in interest rates correspond resonably well with business cycles, at least in the US. This may explain why regime-switching models forecast interest rates better than single regime models. Finally, the non-linear interest rate dynamics implied by regime switching models have potentially important implications for the macro-economic literature documenting the effects of monetary policy shocks on economic aggregates. Moreover, the implied volatility and drift functions are rich enough to resemble those recently estimated using non-parametric techniques.