This paper analyses the change in the Austrian business cycle over time using data back to 1954. The change in the cyclical pattern is captured using a non-linear univariate structural time series model where the time of the break point is estimated. Results for GDP series suggest a break in the frequency of the cycle and in the parameter covering the variance of the disturbances of the cycle taking place in the mid 1970s and early 1980s, respectively. Using data for GDP components a break in these variables is found too, but the timing of the break differs among the series. In a further step the paper assesses the relevance of these findings for forecasting purposes. It is shown that during certain periods the out-of-sample forecasting performance of GDP does improve when a break in one of the two parameters is explicitly modelled.