Small sample properties of forecasts from autoregressive models under structural breaks

Autoregressive models are used routinely in forecasting and often lead to better performance than more complicated models. However, empirical evidence is also suggesting that the autoregressive representations of many macroeconomic and financial time series are likely to be subject to structural breaks. This paper develops a theoretical framework for the analysis of small-sample properties of forecasts from general autoregressive models under a structural break. Our approach is quite general and allows for unit roots both pre- and post-break. We derive finite-sample results for the mean squared forecast error of one-step-ahead forecasts, both conditionally and unconditionally and present numerical results for different types of break specifications. Implications of breaks for the determination of the optimal window size are also discussed.