New York Times: Now the Long Run Looks Riskier, Too

From a recent New York Times article: a new academic study shows that equity investment volatility actually increases over time. Although there is evidence that returns are mean reverting, which reduces volatility over longer time horizons, the authors identify several other factors which more than offset this reduction in risk. These factors include uncertainty about future expected returns and estimation risk. Traditional portfolio theory uses sample statistics (e.g. the average return on stocks over the last 100 years) and assumes they represent the true population of possible values going forward. This approach leads you to believe that volatility of returns decreases as your investment horizon increases (i.e. the standard deviation of your portfolio goes down if you invest for 30 years instead of 1 year). In contrast, this study uses Bayesian statistics, which separates predictive variance (which includes parameter estimation error) from true variance (which has no estimation error).

If you estimate future stock market returns using historical averages, that average has some estimation error in it. As your time horizon increases, parameter estimation risk compounds which leads to an increase in volatility (rather than a decrease). To put it another way: traditional theory says that predictors are perfect and parameters are certain. This theory says that predictors are imperfect and parameters are uncertain. Over time, these imprecisions and errors add up to substantially increase the volatility of your portfolio. The authors use 206 years of data to calculate that the volatility of your stock portfolio over 30 years is about 1.5 times greater than it would be over 1 year. Their conclusion is that stocks are less appealing than they would appear to be under the traditional view of historical risk and return.

Consider: if you started investing in 1979 through 2009, would you have done better in 20 year treasury bonds (rolled over each year) or the S&P 500 with reinvested dividends? An upcoming study (Bonds: Why Bother?) shows that regardless of which month you started in, treasuries outperformed stocks over this period. This is further evidence that the conventional wisdom, that stocks are "safe" as long as you're holding them for 20 years, is wrong. In reality, it doesn't matter how long you plan on holding stocks, they're inherently risky. They may still be a good investment, but you are not guaranteed a higher return just because you took more risk.