Title: M-estimator based unit root tests in nonlinear dynamic models
Authors: Rickard Sandberg - Stockholm School of Economics (Sweden) [presenting]
Abstract: A new class Smooth Transition (STR) discrete time diffusion process is considered where the drift component is allowed not only to have a regime-switching type of behavior, but also a time-varying behavior. Such nonlinear features are simply motivated by the fact that the drift of the return series (say) depends on the phases of the business cycle (the regime-switching behavior) as well as the business cycle may be subject to structural changes over time (the time-varying behavior). Simple tests to identify such nonlinearities are derived. Moreover, Quasi Maximum Likelihood Estimation (QMLE) techniques to estimate the proposed discrete time STR diffusion models are discussed. In an application to S\&P500 and NYSE data, we find, using our tests, overwhelming evidence in favour of a drift component that is subject to both types of nonlinearities. For an example return series, the QMLE techniques are demonstrated.