III ANALYSIS OF REGIME SHIFTS: ROLES OF STOCHASTICITY AND EXPERIMENTS Introduction Statistical methods are powerful for analyzing frequent events, but weak for analyzing

Statistical methods are powerful for analyzing frequent events, but weak for analyzing uncommon events such as regime shifts. The rarity of regime shifts poses deep problems for both understanding and prediction. For any given regime shift, many different models are equally appropriate. There are straightforward statistical methods for sorting among rival models, but these depend on observing many regime shifts. Without extensive data, it may be difficult to decide which model is most appropriate. Consequently, there is considerable doubt about our ability to predict future regime shifts on the basis of any particular model.