Title: Testing for forecast rationality under Markov switching
Authors: Barbara Rossi - Universitat Pompeu Fabra and ICREA (Spain)
Tatevik Sekhposyan - Texas A and M University (United States)
Florens Odendahl - Banque de France (France) [presenting]
Abstract: Novel tests are proposed for forecast rationality robust to the presence of discrete and recurring switches in forecasting ability, such as Markov switching. Existing forecast rationality tests robust to instabilities are based on non-parametric techniques; relative to the latter, our tests perform better in the presence of discrete switches, rather than smooth changes, under the alternative. Monte Carlo simulations suggest that our tests have better power than existing tests in detecting Markov switching deviations from unbiasedness or efficiency. We investigate whether Blue Chip Financial Forecasts for the Federal Funds target rate are rational. Our test finds evidence against forecast unbiasedness and uncovers the fact that forecasters tend to systematically overestimate the future interest rate especially during periods of monetary easing. The size of the systematic bias component is around 25 basis points.