How has COVID-19 Impacted the Credit Derivatives Market?

November 3, 2020

The COVID-19 (C19) pandemic has severely affected global markets, causing economic disruption at unprecedented speed and on a hitherto unknown scale. With the spread of the virus accelerating by mid-March 2020, the US economy has been severely impacted and there are understandable concerns about the damage caused to the worldwide economy. Unemployment continues to shake the workforce, reaching levels not seen since the Great Depression. A number of small businesses have closed, either temporarily or permanently, and even large and well-known companies have declared bankruptcy. 

This blog explores the effects of the pandemic on the credit derivatives market and more specifically, how recent bankruptcies affected North American high yield (NA HY) CDS index trading, including CDX.NA.HY indices and the options on them.

Index HY33

To place the current situation in context, the analysis in this blog concentrates on index CDX.NA.HY.33, which comprises 100 names and was issued a few months before the first C19 cases, on September 20th 2019.  By mid-September of 2020, less than a year after the index issuance, it already has 11 defaults, all of them with recovery auctions. This has put HY33 on a much faster trajectory of defaults than the notorious HY9, which became very popular in structured credit trading during the last financial crisis. HY9, became effective in September 2007, only had six defaults by April 2009 and a total of 20 defaults (out of 100 names) by December 2017, the point at which the longest- traded 10 year maturity expired. 

The list of defaulted companies in HY33 is shown in Table 1. Not all bankruptcies in HY33 were triggered by the pandemic — Dean Foods declared bankruptcy in December 2019. For established retail companies such as JCPenney and Neiman Marcus C19 has been key to their inability to issue more debt. Oil and gas is another industry hit hard by recent global events, with five companies in HY33 already defaulted and others struggling to pay their debts. In addition to C19, oil prices were hit in early 2020 by an excess of supply due to the price war between Russia and Saudi Arabia. At the time of writing, only one travel-related company (Hertz) defaulted. However, other auto rental and airline companies have spent a significant amount of cash over the last several months and, therefore, could also run into financial difficulty in the near future. 

In response to C19, the Federal Reserve and US government have introduced a number of measures to limit the economic damage from the pandemic. The big question is what happens when the government stops programs to support businesses, particularly if there is a second outbreak in the Fall of 2020. Either scenario is likely to trigger another wave of bankruptcies.  

Comparison with S&P

From the outset of the pandemic through June 2020, it is interesting to analyze how the price of HY33 has compared with the broad US market stock index, e.g. SPX (S&P 500). Graph 1 shows the ratio of HY prices and SPX for the period February 18 to June 17, relative to the February 18 price. One can see that although SPX dipped much more drastically around mid-March, to 68 percent against 80 percent for HY33 — both are now close to the pre-C19 level of 93 percent. 

 

Levels of HY33 and SP500 for February 18 to June 17 2020, relative to Feb 18

Graph 1

Implied Credit Spreads

As with all other high-yield indices, HY33 is quoted as a price of a funded note with 100 notional. This equals to par plus the price of an index default swap, the latter from the point of view of the protection seller. As a result, the price is not that sensitive to market changing views on credit. A more sensitive characteristic is an implied CDX spread, which almost linearly depends on the value of CDX swap. Graph 2 plots values of HY credit spreads, implied by HY prices, from February 18 to June 17. 

 

Levels of HY33 spreads in bps for February 18th – June 17th, 2020

Graph 2


One can see that this graph is rather different than the one implied by HY prices. In the midst of the pandemic, the spread more than tripled to 900 bps, and even in mid-June it remained 75 percent higher than its pre-C19 level. Since perception of defaults is roughly proportional to the credit spread, this means that the expectation of losses by the protection seller in June was about 75 percent higher than four months before.

On-the-run Index HY34 and Index Versions

So far, this blog has covered the HY33, index issued in September 2019. The current on-the-run HY index is CDX. NA.HY.34, issued in March 2020. This index already has eight defaults, which is astonishing since it has only existed for five months. Table 1 lists the companies that defaulted and had recovery auctions in HY33 and in HY34 (starting with Frontier on May 13/2020). The auction determines the final price of recovery, which in turn defines the amount to be paid by the protection seller to the protection buyer. Note that the Final price in the table is given in %.

It is worth noting that after a default announcement, the new version of index is issued by Markit, so as of today, HY33 is on Version 12 and HY34 on Version 9. 

 

Auction dates and recovery prices for constituents of HY33 and HY34, as of September 20th 2020.

 

Table 1

Recoveries and Changes in Prices

Table 1 highlights that recoveries set by auctions are extremely low. For nine out of eleven auctions in HY33, the recovery prices are below 10%, and for JC Penney it is only one-eighth of a percent. This is interesting in that in index spread calculations index recovery value is assumed to be 30 percent. However, at the same time, it is consistent with an observation that in times of crisis and concentrated defaults recoveries are significantly lower. Note that this is consistent with a stochastic (aka correlated) recovery model used to calibrate tranches on indices.

The final price of recovery is important, not only for the default payoff incurred by the protection seller, but also because it affects the price of the index after the defaulted reference is removed and the new index version starts being traded. When there are no significant credit movements in the market, we can expect prices to increase around 1-Recovery. Therefore, the lower the recovery, the bigger the jump in price on the day after auction. This is especially true for shorter maturities, which are more liquid. As a reminder, currently traded for HY33 and HY34 are indices with maturities of three, five, seven, and 10 years (with three and five being most liquid).

Challenges for the Trading System

The high number of defaults that have occurred (or are likely to occur in the near future) in indices HY33 and HY34 create a number of challenges for traders of CDX and CDX options. Most importantly, middle-office systems are required to process changes in notionals and cash flows on the dates related to a particular default: announcement of bankruptcy, day of auction, day of new version quoted and recovery settlement. Another challenge is keeping track of payments on defaults, which are fixed after each auction. All of this can be overwhelming, especially for newcomers to HY index trading. As a reminder, since Toys“R”Us defaulted in the Fall of 2017, there were no defaults in NA indices for two years, and now there are several a month. Having stable, sophisticated models that can easily convert price to spread and vice versa is also very important. Another vital component is to be able to replicate an index from its constituents. This is because in times of crisis, the basis between the index level and the level implied by its constituents generally widens.

Adjusting Volatility Surfaces for Defaults

In addition to the aforementioned challenges, there is the issue of volatility skew. HY option volatilities exhibit very pronounced skews. However, dealers only quote volatilities per strike for the current index version. This creates the problem of how to adjust volatilities for older versions. For example, consider someone owning an option on HY33 before December 10: i.e. Version 1. Then starting December 11, the day after the Dean Foods auction, volatilities are quoted for Version 2 only. Applying these volatilities to V1 index would be incorrect. As an example, see below brokers’ quotes for HY33 on and December 10 and 11.


Dealers quote for HY33 before and after the Dean Foods auction

Table 2

Table 2 shows that an auction for Dean Foods on December 10, 2019 set recovery at 9.25 percent. The following day the price jumped from 107.5 to 108.5, which can be explained by the payment on default of 90.75 percent implied by recovery. Now consider a trader who had the V1 ATM option struck at 107.5 on December 10, using the 27.8 percent vol to price it. The following day the trader uses the V2 table, now the only one available, and the corresponding vol is significantly higher at 33.6 percent. At the same time, the new ATM vol corresponding to the strike 108.5 is almost identical to the ATM vol the day before at 27.9 percent. 

As a result, the correct methodology for defining volatility should not be based on the fixed strike, but on the “moneyness”—, i.e. how much strike relates to the ATM level of the current version, which has experienced a price bump due to an auction event for a defaulted credit. A more intuitive approach would be to define that moneyness, not in price, but in spread terms. For this, one has to convert the strike prices in the quote above into spreads. This can be achieved using market conversion convention, although this requires root-finding and can be slow. A faster calculation can be completed using forward ATM spread and DV01 for all strikes. Table 3 compares moneyness and the corresponding vols calculated for V1 and V2 from the quotes above. One can see that from the moneyness point of view, transitioning from V1 to V2 leaves vols almost unchanged, which is the desired behavior.

 

Table 3 Comparison of “moneyness” vols before and after Dean Foods auction

Table 3


Another consideration for pricing CDX options is that each payment on default contributes to the option payoff increasing for the payer and decreasing for the receiver. Therefore, with several defaults the former will be far in the money and the latter far out of the money for both types of options vega will be relatively low.

Conclusion

This blog has analyzed the effect of the pandemic on two North American high-yield indices — 33 and 34 — and the fact that both have an unprecedented number of defaults. Moreover, the current spread indicates that more bankruptcies are on the horizon.

Another point to note is the extremely low level of recovery prices assigned by auctions. However, this is consistent with market views that recoveries are highly negatively correlated with concentration of default. As a result, following each recovery auction, the index price jumps by almost 1 percent. 

The high concentration of defaults creates many challenges for trading systems, given they are required to account for changes in notionals, prices, spreads on dates of default announcement, auction, post-auction, and recovery settlement. For market participants who are relatively new to the HY market and did not experience the high level of defaults during and after the financial crisis, the consequence of these market events can be devastating and long-lasting.

Greater challenges arise for CDX option traders. In addition to monitoring losses, which are part of option payoff, one has to the correct volatility from broker quotes. The problem is that these quotes are only available for the current version of the index, with all defaulted names removed. Consequently, traders who have options on previous versions have to infer vols from available quotes. Leveraging the fixed strike method produces incorrect volatility as it does not take into account a price jump at default. A more reliable methodology is using moneyness, which automatically adjusts for such jumps.

To conclude, with C19 continuing to negatively affect the global economy, trading HY indices in the near future is not for the fainthearted and requires best-in-class models and systems.