Social Security Reform: Responding to the Critics

Studies | Social Security

No. 281
Wednesday, November 16, 2005
by Andrew J. Rettenmaier and Zijun Wang


Results Based on Simulated Data

One of the shortcomings of using historical data is the fact that there are few independent simulation periods. However, there are several techniques that allow researchers to simulate data and in a sense examine more time periods. The Appendix details our simulation results, but they are summarized here briefly.

"Our assumptions are conservative."

The simulation results based on 1,000 independent 44-year investment periods are consistent with the results based on the historical period, but the percentage of simulations in which the portfolios perform worse than the benefit offset threshold is slightly higher. We consider one-year and five-year sampling blocks. The first assumes that the year-to-year ordering of the sampled data is random, while the second attempts to capture the possibility of mean reversion over a longer period. The longer block reduces the rate of portfolios falling below the benefit offset. For the two lifecycle portfolios, the personal account returns fall below the benefit offset about 6 percent to 7 percent of the time when the simulated data is drawn from the unadjusted historical data.

When the simulated data is drawn from data we have adjusted to reflect a possible reduction in the equity premium, the failure rate rises.

However, we caution that a reduction in the equity premium should carry with it a reduction in the variability of the stock returns. Because our simulations in the Appendix and in the previous section did not include a reduction in the variability of stock returns, our estimates using adjusted stock return data probably overestimate the cases where accounts fall short of the benefit offset.


Read Article as PDF