(Brad DeLong's Grasping Reality...) Felix Salmon argues, contra Paul Krugman:
The TED Spread and the Flight to Liquidity - Finance Blog - Felix Salmon - Market Movers - Portfolio.com: Well, yes, up to a point. But "liquidity issues" includes a hell of a lot more than just credit risk, otherwise there wouldn't be any spread between on-the-run and off-the-run Treasuries. (Which spread, as I recall, was in large part responsible for the implosion of LTCM)...
Here is what he is talking about:
The on-the-run U.S. Treasury bond is the bond issued at the most recent auction. The off-the-run bonds are the bonds issued at all the other auctions that are still outstanding. The on-the-run bond has a higher price--a lower yield. The yield-to-maturity of the on-the-run bond averages some ten basis points--some one-tenth of a percentage point--lower. If you are thinking of buying-and-holding the ten-year on-the-run Treasury to maturity, you can buy-and-hold the off-the-run and then rollover your money for the few months until the on-the-run matures, and wind up on average some eight-tenths of a percent richer: the on-the-run bond costs an average of eight-tenths of a percent more than it "should."
The magnitude of this on-the-run premium is made even more awkward for those of us who worship at the altar of the efficient market hypothesis by the short duration of the premium. Every three months the on-the-run Treasury is replaced by a new issue at a new auction. This suggests a trading strategy: (1) sell the on-the-run issue short, (2) buy the first off-the-run issue with the proceeds, (3) hold for three months, (4) liquidate, and (5) repeat. This strategy would appear to promise an average pre-expenses return of three percent of your gross position a year with (a) very little initial capital outlay (your short sale pays for the long leg), (b) very little risk (fluctuations in the on-the-run premium are the only source of risk), (c) low administrative costs, and (d) low transaction costs (the bid-ask spread on the ten-year Treasury is typically 1/32 of a percent of value, less than 1/20 of the on-the-run premium.
So why is this money left lying on the sidewalk? What risks and costs shut down this trading strategy that are not obvious to me?
Whenever I talk to people on the Street or to academics (see, for example, Krishnamurthy (2002) I seem to understand their explanations of the on-the-run premium while they are making them, but afterwards I am once again dazed and confused. Explanations in terms of "special repo" costs seem inadequate--that they are not causes but rather consequences of the off-the=run premium.
At the very least the U.S. Treasury should be making a pretty penny off this premium, for it can construct out of its own bond inventory synthetic securities with exactly the same payment characteristics down to the day and the penny as the on-the-run security: it can go short the on-the-run with no risk at all whatsoever.