Dave,Random Walker wrote: ↑Tue Aug 14, 2018 8:18 amBasically we’re on the same page, but valuations do have some information in them. Why not add that bit of information to the decision process you describe. Knowing that valuations are high, mean expected returns low, and whole dispersion of returns shifted left, the retiring investor might fine tune the decision we are talking about a bit.HomerJ wrote: ↑Mon Aug 13, 2018 10:43 pmBut if he reads this forum, he changes his Asset Allocation to be more conservative as he retires and moves money into bonds, LOCKING IN SOME OF THOSE GAINS.Random Walker wrote: ↑Mon Aug 13, 2018 10:30 pmI think it’s Bernstein or maybe Zwecher who refers to “Murphy’s Law Of Retirement”. An investor profits from a long bull market. As valuations rise, his portfolio swells. Seeing the larger than expected portfolio, the investor decides to retire. But he is retiring into a period of high valuations and low future expected returns.
Someone who saw their stock money double from 1996 to 2000, if they followed the standard advice on this forum, probably did just fine. Their portfolio swelled, they decided to retire, and moved from say 60/40 to 40/60, and got to keep most of that money. Actually ALL of it, since they could live off the bonds while waiting for the stocks to recover.
Not once did they have to look at valuations for this strategy to work.
The key variable is the fact that they retired. Regardless of valuations, they should have changed their allocation to be more conservative the moment they realized retirement was close or at hand.
a couple of questions:
1) What specifically should they do based on that information? If one is already using 4%, or perhaps a bit lower for a longer retirement horizon, what should be the adjustment based on a currently elevated CAPE?
2) How well has CAPE predicted returns AFTER Shiller first proposed it in 1988? If it has been "high" since 1992, perhaps the post data-mining correlation is not as high as the .40 that gets mentioned? i did a little searching, but could not find a reference that calculates the CAPE to returns correlation post-publication. All the correlation discussion i see compute it over all historical data, which is still dominated by the data that was used to develop the metric in the first place. We all know how that can skew actual predictive value of a metric.