Detecting Crises, Jumps, and Changes in Regime

Structural breaksincluding crises, jumps in volatility, and changes in regimeare pervasive in the economy. Detecting breaks can be a major empirical challenge: when and where do they occur, how often, for how long, and of what magnitude? Not detecting actual breaks can deleteriously aect economic analysis, forecasting, and policy. Automated model selection and impulse indicator saturation are two recent methodological innovations that can help detect breaks. Their ability to detect breaks is relevant both in-sample and out-of-sample, so these two tools oer improvements to existing methodology for empirical modeling, forecasting, and policy analysis. This paper illustrates and generalizes these tools by re-analyzing the empirical model of seasonally unadjusted UK narrow money demand in Ericsson, Hendry, and Tran (1994). Both tools demonstrate the robustness of that model to a wide range of feasible alternatives. These tools also yield statistical and economic improvements to that model, and so provide insights into the practical justication of empirical evidence in macro-economics. Combined, these tools permit computer-automated parsimonious detection of structural breaks.

http://www.econometric-institute.org/seminars
Twitter: @MetricsSeminars