The other day I posted an article using a St Louis Fed recession probability chart that was based on the work of Marcelle Chauvet & Jeremy Piger. I used the real-time data to imply that there was perhaps a reliable extrapolation to be made. That was not correct and was a bit hyperbolic of me to imply.
I should have looked into the sources further before citing it because Marcelle Chauvet reached out to me personally to clarify how to use this indicator:
Thanks for your article on the probabilities of recession.
I would like to clarify some points.
My paper with Piger published in 2008 is based on my International Economic Review 1998 paper, which proposes the model used in Chauvet and Piger — I have been running this model in real time since 1993.
The probabilities of recession posted on Fred are ‘smoothed probabilities’, that is, smoothed away from all past kinks that we find in real time.
Real time probabilities are very noisy, and a little bump of 15%, 20% or even 30% does not mean much in terms of signaling recessions.
Please check the graph on real time probabilities of recession on my site:
These are the probabilities we get on a month-to-month basis, without data revisions, and without smoothing. As you can see, the probabilities can even be above 50% and a recession does not follow.
This is why in my paper with Jeremy we set the rule that the probability would have to be above 80% and for a couple of months before one could call a recession.
Thanks for that clarification, professor. This is actually much more in-line with what I am seeing in proprietary indicators because it means the current readings in the indicator are not a clear sign of recession. Sorry for any confusion I might have caused.