some evidence on hysteresis in unemployment rates

abstract: we suggest an unobserved components model that can be used to evaluate unemployment data for evidence on hysteresis effects. unemployment is decomposed into a natural rate component, assumed to be nonstationary, and a cyclical component, assumed to be stationary. hysteresis effects are modelled by allowing lagged cyclical unemployment to affect the current natural rate. the model is estimated using quarterly canadian, german, and u.s. unemployment series. we find large hysteresis effects in the german unemployment series and to a lesser extent in the canadian series. hysteresis effects in the u.s. unemployment series appear to be small although statistically significant.