Thursday, January 06, 2005

CBO and Certainty Equivalents

Brad DeLong today correctly questions the wisdom of a particular CBO report on the Commission Model 2. Brad wonders why CBO would mix a certainty equivalent of a personal account with a distribution of other economic and demographic outcomes. I wondered the same thing a while ago, and posted about it in the context of the Kerry campaign's press release and statements about Social Security reform. The following restates those arguments.

The critical issue is this footnote from the CBO's report:

13 Since the medians are presented here as point estimates, IA payouts are computed assuming risk adjusted returns equal to the Treasury bond rate.

CBO inappropriately uses the phrase "expected retirement benefits" in the text but then clarifies that it is using the Treasury bond rate to accumulate the accounts. The appropriate terminology for CBO would be to say that it is assuming that the portfolios were invested entirely in Treasury bonds. This is the most conservative investment approach, since it chooses to take on zero equity risk. As long as the equity premium is positive, this assumption serves to overstate the reductions in expected benefits that would occur. (As implied in the footnote, we would really want to see the whole distribution of possible benefit levels, not just point estimates.)

This is a very odd assumption to make as a baseline. As far as I know, there is no presumption among people who propose adding personal accounts to Social Security that people who opt for them would then choose to invest them in such a way that eliminated one of the key advantages of having a personal account--the opportunity to obtain higher expected returns in exchange for taking on some financial risk.

This makes the CBO report, unfortunately, much less useful as a guidepost in the discussion of Social Security reform.

Other blogs commenting on this post

1 comment:

brad said...

Note that even though the dashed blue line is computed assuming that the entire private accounts portfolios are invested in Treasury bonds, the 80% confidence intervals reported do *not* make that assumption. They assume substantial investments in equities. That's why the "expected" return is at the 25 percentile of the distribution...

Do you have any insight into why they did it this way? I mean, I understand the desire to use certainty-equivalent rather than expected values to evaluate programs, but I don't like the implicit degree of risk aversion assumed when you use the Treasury bond return as the certainty-equivalent value.