hmmm...I'll have to agree in general on looking at the general trend rather than a subset
Bigfoot Investments wrote:agip wrote:The right way to interpret such a correlation graph regarding data sets which necessarily have a large random component is to look at the overall trend. Is there an obvious slope? In this case the answer is a resounding 'yes'.
and here is a fascinating bar chart from shiller himself - it pretty much shows that there is little difference in predicitve 3 year returns when the CAPE is between 16 and 24.
In other words, historically, 3 yr real returns have been around the same whether CAPE is 16 or 23. To me that means it is not much of a 3 year predictor.
Maybe it works better over 10 year spans.
To look at some mid-section (between 16 and 24) and say that there is little difference in predictive 3-year returns for this subset is a tempting, but ultimately misleading, method of interpretation. This is analogous to the Global Warming deniers' mantra that there has been no warming for 17 years based on very selectively choosing a convenient starting point.
Look at the overall graph. Is there a meaningful trend? Don't pick out some subsection and say "Oh, but here the trend disappears".
By the way, I point out again that the right question in any such predictive method is not "Are returns likely to be above or below historical averages over the next n years?" but rather, "Are returns likely to be above or below alternative investment opportunities over the next n years?".
first off, this is just 3 years - Shiller doesn't think CAPE predicts 3 year returns.
Charts have tails, and so do investment markets. Meaning the best returns are found when stocks get really cheap. And that is what the bar chart shows. The big returns are from buying when everything is knocked down hard - 1980, 1991, 2002, 2008. And that explains the high returns from the low CAPEs.
the midrange CAPEs...so many factors...interest rates, etc...could be good time to buy, could be a poor time.