An MR reader refers me to this interesting paper (pdf) by Max Gillman, Michal Kejak, and Michal Pakoš, the abstract is here:
Rietz (1988) and Barro (2006) subject consumption and dividends to rare disasters in the growth rate. We extend their framework and extent consumption and dividends to rare disasters in the growth persistence. We model growth persistence by means of two hidden types of growth slowdowns: recessions and lost decades. We estimate the model based on the postwar U.S. data using maximum likelihood and find that it can simultaneously match a wide array of dynamic pricing phenomena in the equity and bond markets. The key intuition for our results stems from the inability to discriminate between the short and the long recessions ex ante.
In essence there is tail uncertainty about the length of the recession.