Forecasting turning points : Is a two-state characterization of the business cycle appropriate?