This paper studies the spread between the newly issued 30 year Treasury bond and the old 30 year bond. The spread follows a systematic pattern over the auction cycle: it begins high at an auction date, and converges toward zero by the next auction date. I document the profits on establishing a long old bond/ short new bond convergence trade and rolling this over every auction cycle, over a period from 6/95 to 11/99. Despite the systematic convergence, the average profits are close to zero, and profits covary with the stock market in a manner resembling a short put option position. The difference in repo market financing rates between the two bonds is an important component of the costs in carrying the position. I then ask what economic factors account for the level and time variation in the spread. Using the spread between commercial paper and T-Bills to identify changes in investor preference for liquid assets, I establish that aggregate factors affecting liquidity preference plays an important role in the variation of the bond/old-bond spread. The evidence also establishes that new bonds are imperfect substitutes for other bonds, and therefore changes in the supply of new bonds, such as in recent Treasury buybacks, will have important effects on bond market spreads.