“The Old School Studies Have an Extrapolation Problem of Their Own”

Thanks for your frank and helpful e-mail.  I shared the e-mail with John. He sent me the words below, which I thought you might want to take a look at. I hope to be able to get my own response out tomorrow.

Rob,

 

A quick note. I will finish reading this later.

Michael Kitces is focusing on the EXTRAPOLATION problem.

What he does not realize is that the Old School Studies have an extrapolation problem of their own. Roughly speaking, their results apply to the inner two thirds of the data: roughly P/E10=10 to 20. The Old School Studies have the same extrapolation problem as Michael is pointing out when valuations approach extremes.

The Old School Studies have only two or three data points at valuation extremes. This is too few for confidence.

The New School Studies address cause and effect well enough to isolate the effect of valuations. As such, they apply to a wider range of valuations. They use all the data in the historical sequence range. They do not have an extrapolation problem until P/E10 extends beyond the range of 5 through 27. Beyond that, the accuracy of predictions is more “theoretical” than within the historical sequence range. However, they are still well within the range of plausibility.

Another point: Year 2000 valuations have a SAFE withdrawal rate of 2% or so. But the most likely outcome allows a higher withdrawal rate. Just because the Year 2000 sequence may turn out to survive at a higher rate does not mean that it will have been SAFE.

Have fun.

John Walter Russell