A Bayesian Approach for Modelling Recovery Shapes Following a Recession

We consider a two-state Markov-Switching model in which we include bounce-back functions. These bounce-back functions increase the growth rates of real GDP following a recession. We consider ve specications for the bounce-back function. Among these we select the most appropriate for the series of real GDP for the US, the UK and France. A previous similar study used the frequentist approach for this purpose. Their problem was that they had to x three parameter values before they could select the bounce-back function. As we explain their procedure was not valid. In this paper we circumvent this problem by using a Bayesian approach. We select for the US, the UK and France the most general Bounce-Back 'F' (BBF) function which nests the other bounce-back functions. We nd for all three countries that bounce-back eects play an important role in reducing the negative permanent impact of a recession. Our results also show that there is a large amount of uncertainty in the estimated bounce-back eects.