Relative valuation shows HY credit to be as cheap to model as it has been in twelve months.
Bond guys look at balance sheets. Well and good as way to assess whether a bond will be money good or not at time of maturity. Well and good to assess a bond’s credit risk. They are even a feasible way to evaluate a bond’s price. Structural models of credit risk use the balance sheet as a basic building block of assessing credit risk. It is a natural step to move from assessing one bond to assessing a portfolio of bonds. But the process for a bond guy is unchanged: for a portfolio of N bonds, he assesses N balance sheets.
There is a complication. In a portfolio of bonds, there is often a correlation between the bonds held. Bonds across the fixed income spectrum can be impacted by the moves of other bonds. Bonds can even be impacted by the moves of other asset classes, like stocks, and currency moves. In fact, if a portfolio of bonds under consideration becomes sufficiently diverse, then it makes sense to look at these correlations as a way to assess portfolio valuation.
This is more in the vein of what credit guys do. Synthetic bond indexes serve the role of the portfolio. This provides a liquidity advantage over holding a portfolio of cash bonds, but the trade-off is that a divergence can arise between the index price and the summed prices of the underlying constituents of the index, called skew. So it is a little different for credit guys. The indexes lessen the need for detailed balance sheet review for concerns of outright default and asset analysis in recovery, but credit risk is just as important as an influencer of price. Indexes don’t default short of what is called the “Armageddon trade”, so the focus is on relative valuation.
A corporate bond has the blood of a treasury in its veins. It has stock-like attributes. It is influenced by the carry trade and risk appetite. This blend of features is another way to assess price independent of an explicit review of balance sheets. You could think of balance sheets as underlying all of it, parameterizing it all, but not coming into foreground.
In keeping with this scheme, a relative valuation model needs to reflect the term risk of long-term bonds, the carry impact of the yen, and the risk premium of stocks. Some color on the matter. DXY stalled at 94.8 a few weeks ago. EURJPY has dropped like a stone indicating that Mrs. Watanabe has a desire to unwind the carry trade in periphery Europe credit. This is not risk supportive for an important class of risky debt. Risk sentiment also hangs on which way DXY breaks.
The valuation model shows that HY credit is as cheap to model as it has been in twelve months.
Here’s another view of the model delta.
Note what the model does. It explicitly accounts for the correlations between the different asset classes that we specified based on how we defined the make-up of a bond. We let the central limit theorem over a large number of names take care of specific idiosyncratic factors. Everything is nice and simple, linear relationships and model fit over 90%. I have no truck with transformations and non-linear models for this purpose. Such almost always leads to an over-fitted model that will make you look like an idiot. And it also makes it difficult to see when correlation dynamics change in a decisive way and identify the change point.
Change points are the biggest problem for relative valuation. The behavior of bond prices is often subject to certain random events that result in abrupt changes in their dynamics. The random time when such an event occurs is called a change point. For example, an adjustment in interest rates can lessen how “treasury-like” a bond is and make it much more “equity-like”. In the event of the default of a major financial institution, or the invasion of one country by another, all bets are off regarding the impact on the asset price. These events are not caused by nor directly reflected in the balance sheet until a quarter or so after the fact, but their occurrence definitely can change the bond price.