We use non‐parametric procedures to identify breaks in the underlying series of UK household sector money demand functions. Money demand functions are estimated using cointegration techniques and by employing both the Simple Sum and Divisia measures of money. P‐star models are also estimated for out‐of‐sample inflation forecasting. Our findings suggest that the presence of breaks affects both the estimation of cointegrated money demand functions and the inflation forecasts. P‐star forecast models based on Divisia measures appear more accurate at longer horizons and the majority of models with fundamentals perform better than a random walk model.