Regime Shift - Detecting Regime Shifts

Detecting Regime Shifts

Using current well known statistical methods such as average standard deviates, principal component analysis, or artificial neural networks one can detect whether a regime shift has occurred. Such analyses require long term data series and that the threshold under study has to be crossed. Hence, the answer will depend on the quality of the data, it is event-driven and only allows one to explore past trends.

Some scholars have argued based on statistical analysis of time series that certain phenomena do not correspond to regime shifts. Nevertheless, the statistical rejection of the hypothesis that a system has multiple attractors does not imply that the null hypothesis is true. In order to do so one has to prove that the system only has one attractor. In other words, evidence that data does not exhibit multiple regimes does not rule out the possibility a system could shift to an alternative regime in the future. Moreover, in management decision making, it can be risky to assume that a system has only one regimes, when plausible alternative regimes have highly negative consequences.

On the other hand, a more relevant question than “has a regime shift occurred?” is to ask “is the system prone to regime shifts”. This question is important because, even if they have shown smooth-like change in the past, their dynamics can potentially become abrupt or discontinuous in the future depending on its parameters configuration. Such question has been explored separately in different disciplines for different systems pushing methods development forward (e.g. climate driven regime shifts in the ocean or the stability of food webs) continues to inspire new research.