The challenge posed by Puzzle #10: Erratic Bond Pricing? was to guess what might be the cause of apparently wild swings in market yields for some municipal bonds with the same issuer and maturity:
Early respondents to the puzzle offered these possible causes:
- Different trade types (institutional vs. retail)
- Dealers taking advantage of limited liquidity and transparency
- Different trade sizes
- Some bonds were escrowed with Treasurys and others had significant credit risk
The last of these involved the possibility of more than one kind of bond. In fact, there were two different bonds, but I didn’t make that clear at first, because I wanted to see if anyone would suspect it. After a few days, I posted a hint that there were multiple bonds.
Yesterday, Jeff Pickett of William Blair guessed that the bonds have different call dates. That is the key distinction between the two bonds. Another important distinction is that one bond, with the longer call date, is from a refunding issue that presumably cannot be advance refunded1, and the other bond, with the shorter call date, is part of a new money issue that probably can be advance refunded. There was also a minor difference in coupon.
Now let’s see what really happened with these bonds:
The wild gyrations in yield were just an illusion. What mattered most was which bonds traded on a particular day. When there were trades in the refunding bonds with a ten-year call, the average yield was 20-30 basis points above the benchmark. When there are trades in the new money bonds with a six year call, the yield is far below the benchmark. When both bonds trade, the average yield is in between.
The new money bonds, with their high coupons, seem to benefit from the likelihood that they will be advance refunded. An advance refunding would back the bonds with Treasury collateral while locking in a redemption only six years from today. The refunding bonds probably cannot be refunded again, but perhaps they will be currently refunded2 in ten years.
The spreads shown above follow standard practice. Since all the bonds have remaining maturities of twelve years, their yields are compared to a twelve-year benchmark. But it does not make sense to compare bonds that likely will be redeemed in six years, and bonds that possibly will be redeemed in ten years, to the same twelve-year benchmark.
I used this puzzle at the most recent Municipal Finance Conference to frame my comments on a paper3 that described efforts to use neural networks to develop representative yield curves. The idea is to derive daily yield curves, by rating category, from actual trade data. I wondered if changes in the set of bonds which actually trade might distort the results. That’s what we see in the charts above. The yields seem to bounce around but the individual bond yields are actually fairly stable. It’s like a movie character who puzzles everyone with their strange and sudden changes, when, of course it’s all just a case of mistaken identity (as in Kagemusha or Muppets Most Wanted).
Another question is whether it is adequate to benchmark bonds by maturity when call features can make such a difference for bonds of the same maturity. The SEC has suggested a possible requirement to include a yield spread to a benchmark on every municipal bond confirm. However, it may not be clear in every case which benchmark is most appropriate. A bond that is highly likely to be called in a few years does not trade like a non-callable bond of the same maturity.
Acknowledgement: This puzzle is based on observations made by David Abel of William Blair. He was the skillful moderator of the panel with which I participated. I am indebted to him for sharing his data and his insights.
1An advance refunding allows a municipal issuer to refinance bonds before the first call date arrives. Proceeds of refunding bonds are used to purchase Treasurys to be held in escrow. These securities pay the interest on the old bonds through the call date and also pay to redeem the bonds on the call date. The issuer’s decisions about whether and when to execute an advance refunding are complicated by regulations that prohibit repeated advance refundings, and by the extremely low yields on Treasurys.
2Current refundings, which involve short escrows near the call date, are allowed even when advance refundings are prohibited
3Integrating Big Data, Neuroeconomics, and Learning Networks to Model the US Municipal Bond Term Structure by Gordon Dash, Nina Kajiji and Domenic Vonella.