Tuesday, September 7, 2010

Do Past Returns Predict Future Returns? Not Very Well - In the Short Term.

An article in this week's Barron's (subscription required) cites a study by Eaton Vance where they analyzed 10-year rolling returns on the S&P 500 back to 1926. The study found that just 4% of 12-month periods experienced negative returns. They also discovered that average annual returns to stocks in the 10 years that followed periods with negative 10-year returns were 9.8%, with a range of 7.2% to 15%.

Unfortunately, the Eaton Vance study fails the sniff test in a number of important ways. First, while 10-year returns have been very poor, the current 10-year period began with market valuations over 4 standard deviations from long-term averages (See chart 1.) Current valuations, with a Shiller PE of 20.6, place stocks within the top quartile of valuations going back to 1870, despite stocks' poor performance over the past decade.


Source: Shiller (2010), Butler|Philbrick & Associates (2010)

In fact, an optimized regression of forward 10-year returns against prevailing Shiller PE provides for a mean expected real return to stocks over the following 10 years of 4.6%, versus a long-term average real-total return of 6.6% (see chart 2.)
Source: Butler|Philbrick & Associates

Further, a simple regression of 10-year returns to 10-year forward returns demonstrates conclusively that the relationship is statistically weak. In fact, the R-square of the regression yields an explanatory factor of just 10%. In other words, prior 10-year returns explain just 10% of future 10-year returns (See chart 3.).

Source: Shiller (2010), Butler|Philbrick & Associates


What we do know is that long-term returns can be quite volatile indeed. The following charts show rolling 10, 20 and 30 year real returns to the S&P since 1870. 



Source: Shiller (2010), Butler|Philbrick & Associates (2010)


Source: Shiller (2010), Butler|Philbrick & Associates (2010)


Source: Shiller (2010), Butler|Philbrick & Associates (2010)