A0497
Title: A nonlinear Gegenbauer process to model the unemployment rate in the G7 countries
Authors: Gilles Dufrenot - Aix-Marseille School of Economics (France) [presenting]
Ulrich Aiounou - Aix-Marseille School of Economics (France)
Abstract: A new model is proposed, combining smooth transition model (STAR) nonlinearities and long memory through Gegenbauer processes. A model is considered where long memory is captured by Gegenbauer polynomials. These processes are widely used in applied econometrics but have never been used in models combining nonlinearity and long memory. Indeed, the long memory property of ARFIMA models is captured, in particular, by exploiting the spectral density in the vicinity of zero. However, long-memory dynamics can also appear if we have long cycles, which requires examining the existence of poles (explosions) of the spectral density, not only at zero but for frequencies close to zero. For an economist modelling the dynamics of the unemployment rate, the difference between ARFIMA and Gegenbauer processes is important in order to differentiate between business cycles (captured by regime-switching models), on the one hand, and long cycles reflecting changes in the structural unemployment rate, on the other. A Lagrange multiplier test and a conditional likelihood-based estimator are proposed. The size and power of the tests are proposed through Monte Carlo simulations.
