Error in the real-time identification of breaks

We studied the mistakes that happen in the real-time identification of structural breaks in the selected aggregate-level of U.S. financial data series. We were interested in the real-time identification because of its relevance for forecasting. The level of the noisiness of different datasets and techniques used for the identification of breaks affected the frequency of the mistakes encountered in real-time. We found that mistakes in not finding the true breaks and/or finding the wrong ones in real-time were made more frequently in the case of a noisier financial dataset. Moreover, the techniques for optimal break detection based on the sequential learning of Bai and Perron (2003) were found to make fewer mistakes than those based on the Information Criteria (IC).