The False Promise Of Target-Date Funds

February 14, 2014

For all the simulations, we assume two assets with contrasting risk levels. These are meant to correspond to a typical stock and bond index. For our base case, we assume a stationary return distribution roughly matching the historical performance of U.S. stocks and bonds. We describe the mathematical details of the return assumptions in the section that follows (Methodology), and we explore the impact of varying return assumptions and outcomes further in the Results section.

We compute retirement wealth amounts for many thousands of simulations of market-return histories and investment patterns. Each glide path implies a wealth amount at target date under a simulated history. Glide paths can be appropriately selected to fairly compare simulated target-date wealth characteristics. Since the usual selling point of TDFs is risk control, we match glide paths on risk at target date, with risk measured by standard deviation of retirement wealth.5 We assess the suitability and quality of the match with other statistics calculated from the retirement wealth simulations, including the median as well as the lower and upper 1, 5, 10 and 25 percent quantiles. We require these additional checks because wealth distributions are often highly right-skewed, and no one measure of risk totally captures the investment implications of investment risk policies. To further justify our comparisons of glide paths matched on standard deviation, we explore the target-date wealth distributions of the matched glide paths in the appendix using graphical and numerical summaries. We find that for our capital market assumptions, the observed differences in skewness are not large enough to outweigh the marked differences in other criteria for strategy selection, most notably expected retirement wealth, and that our basic conclusions are generally reliable.6

Within each family of matched glide paths, there is always a winner with regard to either median or mean target-date wealth, but these attributes peak only weakly at the winning glide path, so there is little compelling reason to prefer them. Under an individual history, the winning glide path has a good chance of losing to another glide path in spite of its slightly better expected return. It should be stressed that a good performance in one glide-path simulation does not guarantee a good performance for other glide paths in the same simulated history. Individual histories show a remarkable variation in terms of the best-performing glide path of the family, often with either extreme exhibiting the best performance. We compare performance of glide paths under specific histories in the Results section.

In addition to the base case, we present results for investors with an initial endowment fully invested at the start and no additional contributions. For stationary returns, a fixed investment over the time horizon dominates any glide path in all risk and return measures for the distribution. This is exactly because a consistent exposure to risk over time is the optimal path. Simulation results with the initial endowment contribution pattern are entirely consistent with this theoretical result.

We then test the sensitivity of the winning glide paths to varying market trends by running the simulations under different capital market assumptions. Market return assumptions that favor either stocks or bonds earlier or later in the investment period will tend to have winning glide paths that allocate more portfolio weight to the winning asset per time period. This sensitivity, along with the differences in glide-path performance under individual simulations, means that choosing a particular glide path is equivalent to a bet that the returns will take a particular path.

 

 

Find your next ETF

Reset All