CPI-U Revised For Software Error, Sept. 2000

A subpage of The U.S. Real Term Structure
by J. Huston McCulloch
Department of Economics, Ohio State University

On Sept. 28, 2000, the BLS announced it was revising the January 2000-August 2000 CPI-U data because of a software error that had been discovered involving the treatment of air conditioning in the housing component. Click here for the BLS statement. The error increases the CPI-U value for these months by 0.1 or 0.2 (1982-4 = 100), and therefore increases estimated inflation relative to the same month in the prior year, but by at most only 0.12%.

The Treasury has not yet issued a statement (that I can find) detailing the implications of the revision for indexed bond payments. According to the Wall St. Journal, however, a Treasury circular states that "If a previously reported CPI is revised, Treasury will continue to use the previously reported CPI in calculating the principal value and interest payment." (9/28/00, p. C23) I take this to mean that the erroneous CPI, as it stands on the day a bond is auctioned or a payment is due, will be treated as final, and no attempt will be made to revise the auction or payment based on the corrected CPI, in either direction.

The error means that the Treasury slightly undercharged investors in the April 2000 auction re-opening the 4/2029 indexed bond. The implications of this are trivial, however, since the error is on the order of 0.1%, for a difference of 10 cents (or 3/32) per $100. This is less than even the languid TIPS market moves every time Alan Greenspan unexpectedly sneezes.

On the other side, the April and July coupon payments would have been slightly deficient. The Treasury could perhaps voluntarily make up the difference, even if it is not required to do so. However, the value of this is even smaller -- we're talking about 0.1% or less of a semiannual coupon, which itself is only 2% of the value of the bond. This adds up to about $20 per million dollars worth of holdings, hardly worth most investors' spending time to collect or worrying about. The CREF indexed bond account was then about $250 Million, so it got shorted maybe $5000 on coupons. However, during the first half of 2000, this fund grew 73.9%, so CREF surely made this up many times over on its purchases of indexed bonds.

People who sold the bonds before maturity got underpaid slightly, but I don't see it's the Treasury's responsibility to compensate them, even aside from the enormous hassle of figuring out who they are. Investors who bought the bonds on the secondary market slightly underpaid for them, and will get a little extra real return to maturity.

The Treasury's Effective CPI table for the month of November bases Nov. 1 on the uncorrected August 2000 figure (172.7 vs 172.8R), and Dec. 1 on the correct Sept. 2000 figure (173.7). This means that investors will get a little extra indexation during November to make up for the slightly deficient indexation up to Nov. 1. The October 15 coupon payments were apparently based on the unrevised figures, even though the revisions were announced September 28.