Wed Apr 27 10:35:31 EDT 2005
Harris, Sabelhaus, and Simpson, "Social Security Benefit Uncertainty Under
Individual Accounts." Contemporary Economic Policy 23(1), January 2005.
Why has this article (written by three CBO analysts) not received more
attention? Is there a problem with it?
The authors estimate the distribution of expected benefits under option I
of the CSSS report. Rather than controlling for the investment risk of the
individual accounts by using a lower, risk-adjusted rate of return, as was
done in CBO's analysis of option II of the CSSS report, the authors use
Monte Carlo simulations to generate sequences of investment returns for the
individual accounts, solve for the benefits payable under each such
sequence, and then analyze the distribution of benefit outcomes across
simulations.
In the Monte Carlo simulations, real equity returns have an expected value
of 6.5 percent (same as assumed by CSSS and in other CBO reports), with a
standard deviation of 19.795 percent and modeled as an independent
white-noise process. (The authors said an alternative approach would be a
mean-reverting process, such as one based on dividend-to-price ratios, but
they said the evidence for such an approach "is statistically weak.") Real
corporate bond yields are integrated with the rest of CBO's macro model.
One important point made in the article is that there is considerable
uncertainty about the real value of current-law scheduled benefits because
of the high variability of real wage growth and inflation; current-law
scheduled benefit replacement rates have much less uncertainty because of
the strong link between benefits and earnings. Introducing individual
accounts only increases uncertainty in benefit levels by one-third over
that of current-law scheduled benefits. However, introducing individual
accounts partially severs the link between earnings and current-law
scheduled benefits and thus enormously increases the uncertainty of benefit
replacement rates compared with current law; and uncertainty in replacement
rates is considerably higher for higher-income workers compared with lower-
and middle-income workers.
Another point made in the article is that while the expected value of
benefits is higher under CSSS option I than under current law, the
difference is not statistically significant. And the odds of negative
gains are substantial (50 percent for workers retiring in 2020; 28% for
workers retiring in 2075).
David N. Beede
Economics and Statistics Administration
U.S. Department of Commerce