Steve LeCompte writes the CXO Advisory Group blog. I sent Steve an e-mail re The Stock-Return Predictor and Steve sent me an e-mail in return offering his reactions. I asked if it would be okay with him for me to post his comments as a guest blog entry here and he said that that would be fine. I will post a response to Steve’s comments as tomorrow’s blog entry.
Here are Steve’s words:
Rob,
Thanks for your note.
On a personal level, I am largely in sync with your emphasis on saving and conservative investing. Research such as that summarized at http://www.cxoadvisory.com/blog/external/blog2-1-05/ and http://www.cxoadvisory.com/blog/external/blog11-06-06/ supports your equity return expectations and PE-based approach. However, here are some reservations I have regarding your approach:
– Changes in market participation, regulations, financial services industry offerings and communications technology may have substantially modified the distribution of equity returns over the past century (e.g., decreasing volatility of returns). Sample segmentation helps expose any such trends.
– Simplistically, there is always a reasonably diversified subset of the market with below-average PEs. If PE is reasonably prescriptive for future returns, then an investor could expect to outperform your calculator by continuously restricting investment to such a subset (e.g., “value” funds).
– There are other asset classes besides U.S. stocks and bonds. These other classes may provide attractive (even conservative) alternatives with respect to return distributions.
– There is evidence that financial returns follow power law rather than normal distributions (as argued by Mandelbrot and Taleb — see http://www.cxoadvisory.com/blog/reviews/blog12-17-07/). Power law distributions are “wilder” than normal distributions and offer hardly any confidence for regression-type forecasting. I expect the regressions you use assume a normal distribution for equity market returns.
Best regards,
Steve LeCompte
CXO Advisory Group LLC
Today’s Passion: Karthik Narayanaswamy said of the Predictor: “What’s nice about the Passion Saving website is that it uses a much simpler way to bring the valuation issue to focus.”
John Walter Russell says
There is evidence that financial returns follow power law rather than normal distributions (as argued by Mandelbrot and Taleb
That is fine. It has no effect on the regression equations. It does not affect the Stock Returns Predictor.
You already cover such RARE events with words similar to “assuming that stocks behave at least somewhat as they have in the past.”
We already allow for an error of 5% on the upside and 5% on the downside (a total of 10%). That is where rare events happen. The central 90% falls under the influence of the CENTRAL LIMIT THEOREM. Still, we make allowances for non-symmetrical distributions.
Have fun.
John Walter Russell
Rob says
Thanks very much for those helpful comments, John.
Rob
John Walter Russell says
Simplistically, there is always a reasonably diversified subset of the market with below-average PEs. If PE is reasonably prescriptive for future returns, then an investor could expect to outperform your calculator by continuously restricting investment to such a subset (e.g., “value” funds).
Yes. This is true. It works.
[Be careful about expenses.]
Have fun.
John Walter Russell
John Walter Russell says
From Orders of Magnitude
http://www.early-retirement-planning-insights.com/orders-of-magnitude.html
Summary
Rebalancing has little to offer except, as we have discovered, when valuations are high and you have no way to discern value.
Individual market slices produce significantly different returns, of the order of 2% to 3% (annualized).
Valuations, as measured by P/E10, have a huge effect. Returning to normal valuations would increase 10 year returns by 5% (annualized). Returning to bargain levels would increase 10 year returns by 13% (annualized).
Have fun.
John Walter Russell
P.S. Various changes in the stock market may add as much a 1% to the annualized return. [This estimate requires additional assumptions, which make it more precise but less reliable.]