Differentiating between coefficient break and volatility break

A new method for differentiating between a coefficient break and a volatility break is proposed. In the proposed method, time series observations are divided into several segments, and an autoregressive model is fitted to each segment. The goodness of fit of the global model composed of these local models is evaluated using the Bayesian information criterion, and the division which minimizes this criterion defines the best model. The proposed method makes a mixture model, such as that with a volatility break in the first break and a coefficient break in the second break, applicable. Simulation results show the efficacy and limitations of the proposed method. Empirical applications to quarterly time series of industrial production for 19 countries provide interesting results.