Because of space limitations, we do not show other experiments that tested the changing risk patterns to allow for more volatility, reducing expected equity and/or bond returns, allowing for a “turbulent” period of several years at the beginning, middle or end of the investment period, or any of several other modifications to the assumptions. Although the optimal glide paths are affected by most of these changes, none of them strongly favors any glide path, and our overall conclusions remain unaffected by these sensitivity analyses.
In general, we prefer simulation to a historical test, since history is only one realization of many possibilities. However, an examination of similar strategies over history may provide reassuring validation of the simulation framework. For this purpose, we use inflation-adjusted stock and bond returns from 1928 to 2012.9 Glide paths pose an additional problem for testing due to the lack of available independent 40-year windows on which to test. International markets don’t help either—few markets have more than one 40-year observation, and international markets are correlated. We chose not to use the 85 observations in a single series due to the asymmetry in the importance of the data—the middle years from 1967 to 1973 would be represented in 40 experiments, whereas the first and last year in the sample would only be used once. To overcome this, we extend the data beyond 2012, reusing the period beginning with 1928, so that the 85 annual observations are used with equal importance. We note that history has been favorable to stocks, and while we believe the example is instructive, we do not believe more equities results in more wealth for all future scenarios.
We apply history to an annual investment starting at $10,000 and increasing by 2 percent per year10 to be consistent with our simulations. We then analyze two static investments, 40 and 60 percent equities, and corresponding glide paths chosen to be superficially risk-equivalent, from 80 percent to 0 percent and from 100 to 20 percent equities with the remainder invested in bonds. This analysis corroborates our simulation experiments—glide paths have no special ability to reduce risk at their target date. If anything, history is harsher than our simulations on extreme glide paths. For example, compared with the 40/60 portfolio, a glide path from 80 to 0 percent equities results in both lower mean/median wealth at 65, as well as greater standard deviation and an inferior worst period. Someone who invested in a 40/60 portfolio from 1941 to 1981—its worst 40-year period—would see real wealth increase only to $675,000, whereas the worst case for the glide path investor is from 1940 to 1980 and results in a real wealth of $632,000. The results for the four cases are summarized in Figure 6.