The Congressional Budget Office released a report today claiming that “the federal government paid 16% more in total compensation than it would have if average compensation had been comparable with that in the private sector, after accounting for certain observable characteristics of workers.” Specifically, the report finds that average wages were roughly the same in both sectors, but that the value of benefits was almost 50 percent higher in the public sector (the overall difference comes to 16 percent because the bulk of compensation takes the form of wages).

CBO cautions, however, that “estimates of the costs of benefits are much more uncertain than its estimates of wages, primarily because the cost of defined-benefit pensions that will be paid in the future is more difficult to quantify and because less-detailed data are available about benefits than about wages.” This a serious understatement. In fact, the results appear to hinge entirely on the interest rate used to discount pension benefits, which isn’t even mentioned in the report.

Though the report itself provides no useful information about how the cost of particular benefits was estimated, a working paper by the same author indicates that CBO probably used a 5 percent rate of return to discount the value of future pension benefits (in other words, $105 in pension benefits next year is valued at $100 today). In practice, this means that if the rate of return on pension fund assets turns out to be higher than 5 percent, which is very likely, the present cost of funding future pension benefits will be lower than CBO estimated.

Where does the 5 percent discount rate come from? CBO based this on a 4 percent “risk-free” yield on Treasuries, with a little wiggle room to account for the fact that pension benefits aren’t quite risk-free (“federal pension obligations are not protected by the Constitution, and the pension obligations of private-sector employers are only partially covered by the Pension Benefits Guarantee Corporation”). If you’re wondering why the riskiness of pension benefits matters rather than the expected return on pension fund assets, that’s a very good question.

With compounding, using a nearly risk-free discount rate can double or even triple the cost of pensions compared to using expected returns on actual pension fund assets. Admittedly, expected returns are uncertain, and the uncertainty itself imposes an indirect cost on pension sponsors. But if you’re going to consider indirect costs like uncertainty, you should also factor in indirect advantages for employers, like reduced turnover.

The experts consulted by CBO for this report include my EPI colleague, Heidi Shierholz, as well as two researchers on the other end of the political spectrum, Andrew Biggs of the American Enterprise Institute and Jason Richwine of the Heritage Foundation (Biggs and Richwine were also consulted for the working paper). Though it’s nice that CBO considered a range of opinions, it’s a shame that Biggs, Richwine and others have been successful in convincing CBO and others to use low Treasury yields to estimate the cost of future pension benefits. A paper to be released next week by EPI also shows that Biggs and Richwine use highly unorthodox methods in comparing public- and private-sector pay. Among other things, they ignore enormous differences in educational attainment between teachers and private-sector workers.