confusion with effect of recessions on High yield bonds return

[patriciar]

Hi David,
Trying to get the whole idea of book 5, I found something that sounds weird: (previously have looked through the existing threads and there is no discussion about this)

On reading "When selling becomes viral" (current issues) the author gives 2 ideas :

1st: during the Covid crisis, Investment grade (IG) bonds and High Yield (HY) bonds experienced similar increases (declines) in their spreads (prices), higher than 500bps.

2nd: during the GFC, HY bonds experienced a much higher decline in prices (increase in spreads) that IG bonds.

With this, I understand there could be 2 different scenarios/reactions to a crisis.

The confusion comes when mixing it with the reading of “factors”. Here, it is demonstrated that IG bonds experienced a high increase in returns in recessions (which matches with what happened during GCF and COVID crisis), but…HY bonds does not experience an increase in returns, but a decline, which does not match with the Idea I got from what happened neither in the GFC nor during COVID crisis.

How can this be understood? Maybe I am mixing the spread concept with the return/yield concept? Both are negatively correlated with prices, right? Could you help me to grab the right concept?

Thanks a lot in advance!

 

[David Harper CFA FRM]

Hi @patriciar Candidly I haven't read "When selling becomes viral" (my excellent colleague prepped the notes) so I just started reading it now. I'm not finished (ie, haven't analyzes the data which will take much more time) but I don't see your points yet. So far the main points seem to be, with respect to March 2020 (the beginning of the COVID phase):

• IG bonds: CDS spreads do not widen nearly as much as bond spreads; the author calls this a "dislocation" (or departure). As the CDS-bond basis = CDS spread - bond spread (i.e., in Fig 2, the CDS-bond basis is found roughly by subtracting the blue line from the orange line), this IG dislocation implies a lower basis, or negative basis as shown by the negative green lines in Fig 7)
• HY bonds: both CDS spreads and bond spreads widen nearly in tandem starting in March 2020. If both spreads widen together, the CDS-bond basis is approximately unchanged.

This is summarized thusly:

"This disruption [i.e., switch to negative CDS-bond basis] is most extreme in investment-grade bonds. There, bond spreads experience a large increase while CDS spreads see little change. If investment-grade bond spreads had behaved like CDS spreads, the cumulative return on investment grade bonds would have been only -5% instead of -20%. In contrast, while a basis also opens up in the high-yield market, CDS and bond spreads there increase much more in tandem." -- Page 3

The CDS-bond basis is a relative measure. It's still the case that during initial COVID, both bond and CDS spreads widened for both HY and IG. That's either way characterized by spread widening (how we typically refer to reactions in a recession) or equivalently, price declines or return returns declines.

Then page 13 "Comparing with the GFC 2008-09" the authors also illustrate price declines (aka, spread widening, return declines) but with two differences (from 2020):

In the 2008 episode, high-yield bonds closely track the stock market, while investment grade declines by much less. This stands in contrast to comparable decline in investment-grade and high-yield debt in 2020. Again, this result suggests significant disruption in debt markets, in particular the safer end of the spectrum. Another salient aspect is the extremely high speed at which asset price movements take place in the recent episode." -- page 14

Translation on the first difference: in 2020, both IG and HY cash bond reacted to COVID with price declines (aka, spread widening, asset return drops) about the same, which is to say aggressively and with high relative beta; but in 2008 while both both IG and HY cash bonds reacted to the GFC with price declines, the IG bonds less so and thusly held up their history of lower beta. The authors write "Again, this result suggests significant disruption in debt markets, in particular the safer end of the spectrum" by which they mean (my paraphrase): what's weird is how IG cash bond yields spiked (sending their CDS-bond basis significantly negative) yet CDS spreads didn't really react

I'm tired, it's late, and I'm behind on editing notes, so I will pause there. I've scanned the paper and can't find your points. Perhaps the issue here is the CDS-bond basis? I'll try to pickup on this tomorrow ... Thanks,

 

[patriciar]

Hi David, First of all, I just wanted to say that your answers are always really helpful. Thank you for helping me during late hours (I hope you were able to rest)!.

1st : Maybe I am missunderstanding the mechanics of the spread, you said "reacted to COVID with price declines (aka, spread widening, asset return drops)" when the spread widens does it mean the return has dropped? why price declines when asset return drops? isn't it the opposite?
2nd: When we talk about HY bond spread, what I think is that we are subtracting a benchmark (such as Treasury rates or IG bond rates) from HY bond rates, but, When we refer to IG bond spread, what are we subtracting from IG bond's returns?

3rd: Lastly, regarding to my original question, what I tried to say is that the second reading of book 5 "factors" and the current issue: "when selling becomes viral" gave me 2 very different ideas of what can happen during a crisis:

In the GFC and COVID crisis, the HY bond spread increased by different amounts but in both cases it went up (prices went down). What I can understand in this situation is that in a crisis/recession period, spreads tend to rise but it can do differently, as they did during both crisis for IG and HY bonds. This idea changed when I read "factors" and it showed that during the period 1952-2011, returns increased for government and IG bonds (price decreased I guess), but it DECREASED (just a little [7.4 Vs 7.7], but far from a significant increase as demonstrated during GFC and COVID crisis [>500bps]) for HY bonds. My question is: Are these two ideas compatible? Both are demonstrated with data. It is difficult to predict what can happen to HY bond spreads based on the past, right? as the data observed in some periods, differ from what is observed in others.

I hope I am explaining myself well. Best regards.
Patri

 

[David Harper CFA FRM]

HI @patriciar

I didn't notice anywhere in the appendix where they defined the bond spreads (will look for it when I re-read the paper) but per (eg) Malz in P2.T6 we know it's likely to be (a variation on) the typical yield spread which is the difference between the bond's yield and the government (aka, safe or riskless) bond with similar maturity. So for both the IG and HY, they should be subtracting Treasury yields from bond yields, it's conceivable that's so conventional that they don't bother to explain it actually . The difference is sometimes called a credit spread because it's compensation for credit risk (however, the bond is funded but the CDS is unfunded so the spread is not a pure measure of credit risk, is an advanced point but relevant in this paper).

To illustrate, because I can never get enough of the basics, only by understanding the basics do we have any hope of understanding this paper (and I've only read the paper once and do not yet fully understand it!), say we have:

• $100.00 face, 10-year Treasury STRIP (to make my life easy, no coupon!) with riskfree yield of c.c. 1.0%. Its theoretical price is $100*exp(-1%*10) = $90.48
• Say the 10-year HY bond initially has a (credit) yield spread, s, of 2.0%. It's yield = r + s = 1.0% + 2.0% = 3.0% and its theoretical price is $100*exp[-(1% + 2%)*10] = $100*exp(-3%*10) = $74.08. The HY bond has a lower price due to its credit spread which is compensation (largely) for its default risk.

Now let's imagine the HY experiences some tragedy that makes it even more risky! We typically then observe, or refer to, this increase in risk/default probability as a "widening of the spread." So let's say the spread widens from 2% to 4%, which is a huge 200 bps increase in the spread. The price drops to $100*exp(-5%*10) = $60.65. Don't get me wrong, we could either say "price drops" or "spread widens/increases" as marketplace-wise the price drops first via supply/demand. But for most our purposes, there is no difference between "higher credit spread" and "bond price decline." You can also see how, if you buy the bond at $74.08 and the spread widens, your price drop implies a negative (so far) asset return. In terms of HPR, your simple return is something like (60.65 - 74.08)/60.65. Although i certainly didn't delve into the paper's use of returns, I just automatically accept that higher spreads --> lower bond price and returns.

(I'm omitting the the spread can widen but the price can go up if the riskfree rate drops more than the spread increases, then the total yield can go down. So it's important to keep in mind that the yield here includes two components that can move differently, but I don't think this scenario enters into the paper: I think the COVID/GFC scenarios are only regimes where both components increase a little/lot).

The above is just (HY and IG) bond spreads, the CDS basis definition was the first think i looked for last night: "CDS-bond basis = CDS spread - bond spread". And that's typical, but it's a different animal. If the bond spread is only compensation for default (credit risk), then we have every right to expect this basis to be zero. The paper is largely about why this basis suddenly went to negative, which IMO should not be an automatically intuitive idea.

Let me pause there, and I'll come back tonight/tomorrow to your third point (as mentioned, last night my problem was that I simply could not find the paper to say anything like your third point, sorry). If my explanation of spreads and CDS basis helps, please let me know? Thanks,

 

[patriciar]

HI @patriciar

I didn't notice anywhere in the appendix where they defined the bond spreads (will look for it when I re-read the paper) but per (eg) Malz in P2.T6 we know it's likely to be (a variation on) the typical yield spread which is the difference between the bond's yield and the government (aka, safe or riskless) bond with similar maturity. So for both the IG and HY, they should be subtracting Treasury yields from bond yields, it's conceivable that's so conventional that they don't bother to explain it actually . The difference is sometimes called a credit spread because it's compensation for credit risk (however, the bond is funded but the CDS is unfunded so the spread is not a pure measure of credit risk, is an advanced point but relevant in this paper).

To illustrate, because I can never get enough of the basics, only by understanding the basics do we have any hope of understanding this paper (and I've only read the paper once and do not yet fully understand it!), say we have:

• $100.00 face, 10-year Treasury STRIP (to make my life easy, no coupon!) with riskfree yield of c.c. 1.0%. Its theoretical price is $100*exp(-1%*10) = $90.48
• Say the 10-year HY bond initially has a (credit) yield spread, s, of 2.0%. It's yield = r + s = 1.0% + 2.0% = 3.0% and its theoretical price is $100*exp[-(1% + 2%)*10] = $100*exp(-3%*10) = $74.08. The HY bond has a lower price due to its credit spread which is compensation (largely) for its default risk.

Now let's imagine the HY experiences some tragedy that makes it even more risky! We typically then observe, or refer to, this increase in risk/default probability as a "widening of the spread." So let's say the spread widens from 2% to 4%, which is a huge 200 bps increase in the spread. The price drops to $100*exp(-5%*10) = $60.65. Don't get me wrong, we could either say "price drops" or "spread widens/increases" as marketplace-wise the price drops first via supply/demand. But for most our purposes, there is no difference between "higher credit spread" and "bond price decline." You can also see how, if you buy the bond at $74.08 and the spread widens, your price drop implies a negative (so far) asset return. In terms of HPR, your simple return is something like (60.65 - 74.08)/60.65. Although i certainly didn't delve into the paper's use of returns, I just automatically accept that higher spreads --> lower bond price and returns.

(I'm omitting the the spread can widen but the price can go up if the riskfree rate drops more than the spread increases, then the total yield can go down. So it's important to keep in mind that the yield here includes two components that can move differently, but I don't think this scenario enters into the paper: I think the COVID/GFC scenarios are only regimes where both components increase a little/lot).

The above is just (HY and IG) bond spreads, the CDS basis definition was the first think i looked for last night: "CDS-bond basis = CDS spread - bond spread". And that's typical, but it's a different animal. If the bond spread is only compensation for default (credit risk), then we have every right to expect this basis to be zero. The paper is largely about why this basis suddenly went to negative, which IMO should not be an automatically intuitive idea.

Let me pause there, and I'll come back tonight/tomorrow to your third point (as mentioned, last night my problem was that I simply could not find the paper to say anything like your third point, sorry). If my explanation of spreads and CDS basis helps, please let me know? Thanks,

Yes, David, Your explanation of spreads and CDS basis certainly helps. But it still sounds weird to read that prices decline are related to asset returns drops, I guess we can not say price declines when asset returns drops, but the other way around, and, I guess we are not referring to yields (when we talk about these asset returns), as they would have the same effect as the widening of the spread! Thinks are a little bit clearer now.

I look forward to read your explanation about point 3. I attach the table shown in the reading of “factors” which shows how some factors affect the returns of certain assets: are the percentages shown yields or returns? If they are returns, I guess than when they increase it means prices have increased, right?

If prices of stocks and bonds during GFC and COVID crisis have decreased (return drops), how can this table from 1952 to 2011 show returns increase for low-risk assets( Govt.bonds and IG bonds)? The more I Think about it, the more it confuses me. I mean, It makes sense to me that the less-risky assets are the ones that investors want more in a crisis situation (flight to quality) so their prices go up, but, that was totally the opposite of what happened during both crisis, right? The less risky assets were the ones the investors sold first (specifically in the COVID crisis) in search of liquidity so their prices went down.



Thank you so much!!

 

[David Harper CFA FRM]

Hi @patriciar

Re: "I guess we can not say price declines when asset returns drops, but the other way around" I'm not sure exactly what you mean. Price decline tends to be the more obvious dynamic; e.g., in my illustration above when the spread and yield increases (spread from 2.0% to 4.0%; yield from 3.0% to 5.0%) the bond price declined from 74.08 to 60.65. I do agree, if you are suggesting, that returns are less obvious. Price is observable at a point in time, but returns require at least two prices and beg a timing question, including: historical or expected return. Both of these paper/chapters look to be measuring historical/simultaneous returns: if you own the asset and the price goes down, your return goes down! Right!? But, yea, in P1.T1 we read a lot about expected return. For example, in CAPM, if the risk goes up, the discount rate increases and theoretically the price goes down so that the expected return goes up. But expected return is a theoretical measure.

In this way, there is a sense in which we can say: the following: as demand decreases (or supply increases) such that less buyers "bid" for bond, the bond's price today decreases and it's yield increases (and its credit spread increases, assuming the risk-free rate doesn't change also) such that the realized return of current holders (who are long the bond) drops but the expected return of a new buyer (who buys at the lower price) has a higher expected return consistent with the bond's incrementally higher risk.

Re: comparison to Ang's factor theory: I did not realize you were trying to reconcile this with Andrew Ang's (Chapter 7) factor theory. I was trying to find it in the Appendix of Haddad's paper. That is a potentially complicated comparison, but I would start with: one of the key observations in Haddad's paper is that the the investment grade (IG) bonds were dysfunctional; e.g., "These observations suggest that the investment-grade market in particular, which is the core funding market for US companies and totals over $7 trillion, was dysfunctional; we more sharply characterize this dysfunction in Section 3." You could argue Haddad and Ang are in the same direction w.r.t. IG bonds during recession: Ang shows they have a positive reaction (12.6%) while Haddad says the sudden spread widening (aka, negative CDS basis: the CDS spreads didn't spike) was dysfunctional and not expected. Secondly, it's really important to keep in mind that each of them are making inferences from a certain sample (aka, historical period). Haddad is clearly not generalizing into Factor theory. Ang copied below just for my own reference (emphasis mine)

[Ang, Facctors, Chapter 2] "Government bonds act in the opposite way, generating higher returns at 12.3% during recessions compared to 5.9% during expansions. Investment-grade corporate bonds, which have relatively little credit risk, exhibit similar behavior. In contrast, high-yield bonds are much closer to equity, and their performance is between equity and government bonds; in fact, high-yield bonds do not have any discernable difference in mean returns over recessions and expansions. We can see a similar pattern if we look at periods of low or high growth, as measured by real GOP or consumption growth ... Consistent with the behavior across NBER recessions and expansions, government bonds tend to do relatively well during periods of low growth, averaging 10.0% during periods of low real GDP growth compared to 3.9% during periods of high real GDP growth. All asset returns are much more volatile during recessions or periods of low growth. For example, large stock return volatility is 23.7% during recessions compared to 14.0% during expansions. While government bonds have higher returns during recessions, their returns are also more volatile then, with a volatility of 15.5% during recessions compared to 9.3% during expansions. It is interesting to compare the volatilities of assets over the full sample to the volatilities conditional on recessions and expansions: volatility rends to be very high during bad times" -- “Factor Theory,” “Factors,” “Alpha (and the Low‐Risk Anomaly)” by Andrew Ang, reprinted from Asset Management: A Systematic Approach to Factor Investing (2014).

I hope that helps for now ... thanks!

 

[patriciar]

Hi David, this statement clearly helps me understand the relationship between price-yield-return-expected return and-spread.

In this way, there is a sense in which we can say: the following: as demand decreases (or supply increases) such that less buyers "bid" for bond, the bond's price today decreases and it's yield increases (and its credit spread increases, assuming the risk-free rate doesn't change also) such that the realized return of current holders (who are long the bond) drops but the expected return of a new buyer (who buys at the lower price) has a higher expected return consistent with the bond's incrementally higher risk.

The conclussion I get is that I should not be comparing the study's results of different authors as they both can be true in their respective samples. I hope GARP just does not ask "how a recession affects to prices/yields during a recession?" in a general way, as we could have in mind different studies and theories leading to different conclussions...

Once again, I am really thankful for your help, really saves a lot of time and clears my mind up.
Best Regards

 

[Irishman]

Hi David,

I am bit confused by this statement - "Bond spreads and CDS spreads under normal markets should be equal".
If someone buys a corporate bond to get better returns compared to treasury returns, and, then to mitigate the risk buys a CDS and ends up paying the extra spread as fee. Then by that logic they wouldn't get anything extra?

 

[David Harper CFA FRM]

Hi @Irishman I don't see the context of that statement (what is the source, please?), but there is here a superficial level and a sophisticated level that is instructive. Superficially, yes , your logic is correct. Say the riskfree (US T-bond) yield is 3.0% and a risky bond's yield is 5.0%. In a super simple model, the CDS spread ought to be 2.0% such that, as investor, you'd indifferently get a 3.0% return (5.0% risky yield - 2.% premium paid on the purchased CDS = 3.0% yield) because your default risk is hedged. This is super simple model that says the 2.0% bond spread is

But I'm using the word "model" very deliberately: CDS spread - bond spread = CDS-bond basis. A naïve model says the CDS-bond basis should be zero, and in our example, it is zero because 2% CDS spread - (5% - 3%) = zero. But our simple model assumes the bond's 3.0% spread is entirely and only compensation for default (credit risk). Our model only assumes one fundamental factor, and ignores technical factors too. Hence, in practice, we don't expect a zero basis, and there are several factors. See (eg) https://forum.bionicturtle.com/threads/cds-bond-basis-factors-confusing-impact.10284

The cds-bond basis is one of my favorite examples of what it means to use a model in finance. We can use a model and still not expect the model's output to match the market. Most models use on a few fundamental factors, but we can add factors to most models, too.

I hope that's helpful,