This post may be a little more complex than most. It will also be more theoretical. For those disinclined to wade through the whole thing, skip to the bottom where the conclusions are (assuming that I have any). 😉
Asset Prices are (Mostly) Validated by a Thin Stream of Transactions
One thing that I have been musing about recently is how few transactions exist to validate the pricing of various markets. I’ll start with two obvious ones, and then I will broaden out to some more markets that are less obvious. (Hint: markets that have a high level of transactions relative to the underlying asset value have a lot of speculative “noise traders.”)
Let me start with the market that I know best as far as this topic goes: bonds. Aside from some government and quasi-governmental bonds, very few bonds trade each day — less than a few percent. It’s very difficult to use the small volume of trades to price the whole market, but it can be done.
When I was a bond manager for a semi-major insurance company, I was the only one of the top managers that was a mathematician, and familiar with all of the structures underlying the bonds. I could create my own models of bonds if needed, and I often did for interest rate risk analyses (which was still a responsibility amid bond management). Combined with my knowledge of insurance accounting, it made me ideal to do a certain monthly task: making sure all of the bonds got priced.
The first part of that isn’t hard. The pricing service typically covers 90-98% the bonds in the portfolio. What I would receive on the first day of the month was a list of all the bonds the pricing service could not calculate a price for. I would take that list and compare it to last month’s list of the same bonds and add to it any new bonds we had bought that month, and who the lead dealer was. I would then ask the dealers for their prices on the bonds (which were typically illiquid). I would compare those prices to the prices of the prior month, and maybe ask a question or two about the prices that were out of line. That would usually elicit a comment from my coverage akin to, “The analyst thinks spreads have widened out for that credit because spreads in that industry have widened out, and a less liquid bond would widen out more. The why the price fell more (or rose less).
After that was done, that left me with a small number of utterly illiquid bonds that we had sourced totally privately, or where the dealers who had originated the bonds had ceased to exist. All of those deals lacked options to accelerate or decelerate payment, so it was a question of modeling the cash flows and applying an appropriate yield spread over the Treasury or Swap yield curve. [Note: the swap curve gives the yield rates at which AA-rated banks are willing to trade fixed rate exposures in their own credit for floating rate exposures in their own credit, and vice-versa.]
But what is appropriate and how did the three methods of getting prices differ? The second question is easier. They didn’t differ much at all. The dealers and I were likely doing the same things — just with different sets of bonds. The pricing service, on the other hand, was much more complex, and the other two methods relied on its results.
It was was called “grid” or “matrix” pricing, though it was much more complex than a grid or a matrix. The pricing service models would look at all of the most recent trades that had happened in the bond market, and use all of the prices to estimate yields that were adjusted for the options inherent in the bonds that could accelerate or decelerate payments. From that, they would piece together yield curves that varied by industry and collateral type, credit rating (agency or implied by a model that involved stock prices and equity option prices), individual creditors, etc. Trades on different days were adjusted for market conditions to make the pricing as similar as possible to the end of the month. After that the yield and yield spread curves generated would be applied to the structures of individual bonds with a adjustments for whether the bonds were:
- premium or discount
- large deals that were widely traded or small illiquid deals
- callable or putable
- senior or subordinate or structurally subordinate (a bond of a subsidiary not guaranteed by the parent company)
- secured or unsecured
- bullet or laddered maturities (sinking funds, etc.)
- different currencies
- and more
And there you would have a set of self-consistent prices that would price most of the bond universe. That’s not where transactions would necessarily take place… particularly with illiquid securities, what would matter most is who was more incented to make the trade happen — the buyer or the seller.
Implicitly, I learned a lot of this not just from modeling for risk purposes, but from trading a lot of bonds day by day. How do you make the right adjustments when you compare two bonds to make a swap, and, how much of a margin do you put in as a provision to make sure you are getting a good deal without the other side of the trade walking away? It’s tough, but if you know how all of the tradeoffs work, you can come to a reasonable answer.
One more note before the summary. The less common it is for a bond or group of related bonds to trade, the more effect a trade has on the overall process. It becomes a critical datapoint that can redefine where bonds like it trade. Illiquidity begets volatile prices changes in the grid/matrix as a result. On the bright side, illiquidity is usually associated with small sizes, so it doesn’t affect most of the market. There is an exception to this rule: trades done during a panic or the recovery from a panic tend to be sparse as well. The trades that happen then can temporarily change a wider area of pricing. I remember that vividly from the whipsaw markets 2001-3, especially when the bond market was restarting after 9/11. If that crisis had happened later in the month, the quarterly closing prices might not have been as accurate.
Summary for Part 1 (Bonds)
The bond market is complex, far more complex than the stock market. Pricing the market as a whole is a complex affair, but one for which prices are reasonably calculable. For the average retail investor investing in ETFs, the bonds are liquidi enough that pricing of NAVs is fairly clean. But even for a large ugly insurance company bond portfolio, pricing can be significantly accurate. Next time, I’ll talk about a related market that has its own pricing grid(s) — mortgages and real estate. Till then.