A Non-Linear Forecasting Model of GDP Growth

We develop a forecasting model of GDP growth that features regime-switching behavior and an error-correction mechanism (ECM). Regime changes are manifested in the behavior of a stochastic regime drift component that moves between expansionary and contractionary phases, thus generating cycles in GDP growth. Regime changes are triggered stochastically by an observable indicator variable we refer to as a tension index. This is constructed as the geometric sum of deviations of actual GDP growth from a corresponding "sustainable rate" (interpretable as the growth rate of potential GDP). At the onset of an expansionary regime, the regime drift component begins to move along an increasing trajectory. As GDP growth follows along this trajectory, the tension index tends to increase; as it does so, the probability that a regime change will be triggered also increases. We model this probability using a logistic specification that is increasing in the absolute value of the index. In the event of a regime change, the process becomes reversed, and the regime drift begins to decline along a newly established path. Linking the behavior of the tension index to GDP growth thus enables us to capture floor and ceiling effects. In the context of the model, the longevity of the economic expansion of the 1990s is attributable to the moderate nature of output growth observed over this period, which generated correspondingly moderate levels of the tension index, and low probabilities of a reversal