Wednesday, August 01, 2007

That's not true...

This story from Bloomberg is highly misleading. Basically it says that major brokerages are trading as though they are junk, according to credit default swaps and Moody's Market Implied Ratings system. The story specifically mentions Bear Stearns, Goldman Sachs, Lehman Brothers, and Merrill Lynch.

I know a little about Moody's Market Implied system, but anyone who knows more is welcome to explain it in the comments section. This system is entirely separate from their actual ratings process. It takes real-life trading spreads and attempts to estimate what kind of rating is "implied" from that spread.

While I understand the spirit of Moody's Implied Rating system, I don't know the actual calculation methodology. Regardless, the concept is simple enough. Take how a bond is currently trading, ignoring the rating. Then look at what other bonds are trading with that kind of spread and see what the ratings are. Alternatively, you could draw a "credit curve" by marking the
spread of a typical Aaa-rated bond, the Aa-rated, etc. You take the trading level of a given bond and see where on the "curve" the bond would fall.

OK, back to the brokers story. The claim that the four brokers mentioned are equal to junk is hard for me to understand. Here are CDS as of 7/31 on the broker group according to Bloomberg.

Bear Stearns: 97
Goldman Sachs: 79
Lehman Brothers: 91
Merrill Lynch: 82

Here are CDS quotes for the top 10 names in the Merrill Lynch High Yield Master index.

Ford: 445
GM: 644
Charter Communications: 958
El Paso: 278
MGM Mirage: 448
Williams Cos.: 184
R.H. Donnelley: 465
Chesapeake Energy: 248
Freeport-McMoran: 117
Clear Channel Communications: 603

Note that the widest brokerage name (Bear) is 20 bps tighter than the tightest high-yield name (Freeport). The story also mentions Xerox, which is quoted at 95. OK, so you found one high-yield name that's trading around where Bear Stearns is. I don't know what that proves, really. Xerox is a relatively small issuer, outside of the top 50 in the index.

We could alternatively look to look at the spread of an index of BB-rated bonds. Currently the asset-swap spread (which is the best comparison to CDS) on the Merrill Lynch BB index is 232.

We could look at the CDX-High Yield index, which is a good comparison because it's a CDS contract but on a basket of high-yield names. That spread on 7/31 was 498.

So I just don't know where Moody's and Bloomberg are getting the notion that a CDS contract trading around 100 is equivalent to junk. Not only is that not true, that's impossible.

30 comments:

  1. I'm not a junk specialist - and especially not a US Junk specialist - but I've been watching along with every other bond guy in the world.

    I've been shown an indication of BSC CDSs at 120/130 (a July 30 number) and have no idea where they may have gapped out to intra-day.

    Anyway, my Bloomberg story of note is Junk Bonds Have Worst Month in Five Years on LBOs, mainly High-yield, high-risk bonds lost 3.1 percent in July, their worst monthly performance since 2002, as concerns about an onslaught of debt to finance leveraged buyouts drove down prices.

    The more I look at that number the less I understand it. Do you have the monthly yield change for those ten names you mentioned? What's the duration of that index?

    Great blog, by the way!

    ReplyDelete
  2. James:

    Thanks for the comment and the kind words on the blog.

    Anyway, Bloomberg has reasonably good prices on CDS. Type in any equity ticker, then RELS. Like BSC Equity RELS. #22 is "Par CDS Spreads" The five-year CDS is the one most commonly quoted and traded.

    Once you select one you can do regular HP/GP functions to see how far something has moved.

    The Merrill HY Index has a duration of 4.7 and the spread widened 112bps during July.

    By the way, I know brokerage paper gapped in on 7/31. But even if you put BSC CDS at 130, its still a stretch to call that spread similar to junk.

    ReplyDelete
  3. well said. this is a typical poorly written story by bloomberg, and traders that don't fully understand it were probably the ones paying 130bps to buy BSC.

    the brokers are too wide, their risk-management is very good and at worst this "crisis" will affect earnings a bit, but the risk of default is way lower than the market is implying.

    ReplyDelete
  4. Hi, is there a generic formula to translate a CDS spread to an ASW spread?

    I understand that for assessing credit risk, the following are best to analyze (in order of preference): CDS, ASW, YTM.
    Is that correct? Thanks!

    ReplyDelete
  5. I was also quite amused when I read this news on bloomberg. One must add also that CDS spread tend to move faster because of the liquidity and the different type of participants, especially a lot of hedgies are using cds for various purposes. CDS spreads do not tend to represent a correct proba of default. By computing the basis between cds and cash bond, you can also get a good idea of the current "exuberence" in the market.

    ReplyDelete
  6. Anon:

    The CDS spread is theoretically equal to what a swap spread should be on a 5-year bond. As Nicolas points out, however, somtimes the CDS spread takes on a life of its own.

    If you click on the "How Does a CDS Work?" on the right hand side of the blog home page, there is some discussion of the relationship between CDS and cash bonds.

    ReplyDelete
  7. Okay, I know this is an inappropriate question, but what do you think of the action in HTR, a closed end fund containing pretty much strictly leveraged MBS stuff?

    It has been treated like dirt until today. Now, it is showing signs of life.

    Does that mean that intelligent people have decided that the MBS blood-bath has been overdone?

    ReplyDelete
  8. A posting above says:
    "By computing the basis between cds and cash bond, you can also get a good idea of the current "exuberence" in the market."

    Wouldn't that basis as calculated be more reflective of credit-specific matters rather than a general market mood/exurberance?

    ReplyDelete
  9. Have you read the Heard on the Street column today?

    This line got my attention

    "The availability of credit has disappeared, and there are $220 billion of [leveraged-buyout] loans" that need to be financed, says J. Kyle Bass, a managing partner at Hayman. "It is going to smoke investment banks. And many more funds will be carried out, feet first."

    That doesn't seem like a huge # to me. Especially spread across all the banks. Does it seem big to you?

    ReplyDelete
  10. Thanks, tddg!

    Now I have one more question ... when I look at the BSC CDSs today, as suggested, I see that there's a 5-10 inversion. Is this as silly as it would be in the cash market, or is there something clever happening in the CDS market I don't understand?

    ReplyDelete
  11. Anon:

    I'll look at HTR. Don't know the stock off hand.

    Fred:

    CDS spread generally yes. But if the CDS is moving wider much faster than cash bonds that can indicate somthing technical is going on.

    Chad:

    I agree. I think anytime the market goes through large vol moves, some hedge funds may get caught with bad trades. I remember a couple hedge funds got hurt in 2005 when Kerkorian bot GM stock causing it to rise at the same time the bonds were still getting killed.

    For whatever reason, the media seems to take pleasure in pain when it comes to the markets. I think it mainly just spreads fear among the general public. But I think if you keep your eyes on fundamentals, you realize that things aren't that bad. You'll also realize that things were never as good as spreads would have made you believe in 2006 and early 2007. This lowly blogger called for HY to falter in 2007 having nothing to do with the housing market.

    So my bet would be that high-yield stabilizes somewhere around here, with a decent probability of a tightening. It will be a very bumpy road, though.

    ReplyDelete
  12. James:

    I suspect the 10-year didn't trade toward the end of the day. 10-year CDS don't trade much. So the 5-year widened all day while the 10-year just didn't have enough prints to show up. That's my guess. I'll see if I hear anything more.

    I can't think of a reason other than very short-term techicals as to why CDS curve would invert. Anyone else?

    ReplyDelete
  13. I just watched the Jim Cramer Video and got admitted to hospital.

    can you explain us what is going on ?

    He says we have an armegeddon in the Fixed income market ?

    ReplyDelete
  14. I think you should stop watching Cramer.

    ReplyDelete
  15. Anon:

    HTR isn't an exchange-traded fund. Its closed-end. This means it can and does deviate from its NAV based on investor sentiment. I'll bet its a pretty good short here.

    ReplyDelete
  16. I am a user of Moody's Market Implied Ratings. It is my understanding that the brokers are trading at Baa/Ba levels versus the names mentioned in the post which are generally much lower rated companies. According to Moody's median credit spreads curve, Bear appeared to be trading very near the cutoff between the Baa3 and Ba1 so the data looks correct.

    ReplyDelete
  17. I'm curious how they build the bands, because if there are a handful of Ba-rated credits that are as tight as Bear Stearns, does it really make sense to say that Bear a Ba credit? I mean, wouldn't it make more sense to say that those ultra tight Ba credits are trading more like Baa or A credits?

    ReplyDelete
  18. Sorry for the delayed response. According to Moody's website, the bands are built using the entire population of Moody's rated entities with CDS. I think there are about 2-3k of such entities.

    ReplyDelete
  19. They might also only measure the band levels once a month or something like that.

    ReplyDelete
  20. They say its published daily

    ReplyDelete
  21. Just finished Choudhry's book on the CDS Basis (Bloomberg Press, 2006).

    Figure 3.3 shows a CDS term structure inversion for Fiat in November 2002, which Choudhry claims is "because a year earlier, Fiat had issued a very large size 'exchangeable' bond that had a July 2004 put date. The basis, previously flat, widened to over 100 basis points due to market makers hedging this bond with convertible bonds of the same name."

    He also claims that very junky CDS will normally be inverted reflecting "the belief that there is a higher probability of default risk right now rather than 5 years from now, because if the company survives the first few years, the risk of default is much lower later on. This gives rise to lower spreads." (Appendix II)

    ReplyDelete
  22. James:

    Interesting... but if I want to speculate on, say, GM going bankrupt and the 1yr CDS is more expensive than the 5yr, why would I buy the 1yr? Because by owning the 5yr I'm still positively exposed to a GM bankruptcy. Maybe the inversions are happening in instances where you have to pay points up front?

    ReplyDelete
  23. Yabbut if you buy the five year then you're taking the risk of having to pay protection money for five years. If you're really all that convinced that it's now-or-never you would buy whichever one had the lowest present value.

    I suppose you could perform a relatively simple calculation with probability of default in any given year (adding up to less than 100%!), estimating your winnings if there is a default and then go for the highest net present value.

    I'm not entirely convinced the procedure makes any sense; you have to have a counter-intuitive view of the incremental probability of default before you don't just jump on an inverted CDS curve. I can sort-of see this for, say, a small pharma company that's got one product in the pipeline and just enough cash to get through trials; but I have greater problems thinking this through for regular companies.

    It seems to me that in most cases, if a company manages to get through its year of 60% default probability, then its next year's one-year default probability is still going to be 60% (give or take ... ).

    In such a case, the probability of default is 60% in year one and 0.4*0.6=24% in year two. So - without having performed any precise calculation - it seems to me that if the one-year's worth Xbp, then the two year is probably worth around X * (84/60) / 2 = 0.7*X bp, because, basically, you're only willing to pay 400bp for the second year of protection.

    I think that means that if the probability of default is constant for every future year, then the CDS curve should be inverted.

    Lemme see ... if the probability of default in any given year is x, then the probability of a default in year two is (1-x)*x = x-x^2. For this to be higher than the probability in year one, then x-x^2 > x, implies -x^2 > 0, which is impossible.

    So to keep the cumulative probability constant, you need to add a constant, so that the probability of default in year 1 is x, and the probability of default in year 2 is x+y. Therefore, the probability measured at time 0 of a default in year two is (1-x)(x+y) = x + y - x^2 - xy.

    We want this to be greater than the probability of default in year 1 alone, so

    x + y - x^2 - xy > x

    and therefore

    -x^2 - xy + y > 0

    and

    x^2 + xy - y < 0

    I'll bet there's some easy software on the web that will graph that. It will hold as long as both x and y are small (which will allow you to toss out the two product terms), but as soon as they start getting bigger it's much harder.

    If x = y (that is to say, the probability of default in year 2 is double the probability of default in year 1) then the condition for a positive slope in the CDS curve is:

    2x^2 - x < 0

    therefore

    2x - 1 < 0

    therefore

    x < 0.5

    which we knew was true anyway because of the condition that the sum of all default probabilities measure at any given time is less than 1

    I don't know. I'll have to do some proper calculations. I only got interested in CDSs recently and am just trying to poke around looking for weird stuff.

    Sorry ... this started out being a short post and then I got to thinking out loud ...

    ReplyDelete
  24. I'll have to check, but I think there's a theorum in there ... if the one-year default probabilility is 20% or more, then there must be an inversion somewhere in the 1-5 year CDS curve, because you cannot increase your default probability monotonically.

    Um, given an assumption of constant recovery.

    ReplyDelete
  25. Yeah... James I thought of that point as I was walking to lunch today. You are 100% right. If its really now or never, then I'd might rather pay more for protection now but have the contract expire if nothing happens.

    I'd attack the math thusly:

    - Say 1yr is 400 and 5 yr is 200.

    - Your view is that within a year, either there will be a default event or the trouble will have passed.

    - So if a default event happens, the one year will cost you 2x as much.

    - But if a default doesn't occur, then you'll want to close your position. The one year will just expire and you'll owe nothing. But the 5-year will have remaining value.

    - You'll have to estimate the change in spread of the 5-year piece one year hence, then calculate the PV of the remaining cash flows. That will be the cost of closing the contract.

    - Since your view is that the trouble has passed, you'd expect to have a loss on the position. If the loss is greater than 200bps, then you're better off in the 1-year CDS.

    - It'd be easy for the loss to be 200bps. A 5-year cash flow is going to have a duration around 4, so 200bps loss would only be a move of about 50bp. If you really believe the trouble will have passed in 1 year, odds are good that the spread will be substantially tighter.

    You could take my math combined with your math and probably come up with an estimate of the market's odds of default. I'm such a gigantic bond geek that this all seems terribly exciting to me.

    ReplyDelete
  26. I'm thinking - vaguely - of coming up with a programme to handle all this calculation stuff.

    I am sick and bloody tired of salesmen telling me the number is such and such and it must be such and such because Bloomberg says so.

    But anyway, one thing I forgot was that if you buy 2-year protection and it does default in one year then you don't have to pay any more. *blush* That doesn't change all the arithmetic, but it changes some of it!

    This stuff is interesting and there's money to be made. I've just been shown Bear Stearns 5-year Maples at +191, with CDS at 135-40. I don't know what the specific terms are on the CDS, but this doesn't seem right to me ...

    ReplyDelete
  27. According to Bloomberg the CDS curve is inverted for both CIT and CFC ... interestingly, the 5-year rate is almost identical for these two issuers, but CIT is 409bp for 1-year, CFC 594bp.

    Screen CDSW is supposed to calculate default probabilities, but can't do it for CFC.

    Hmm...

    ReplyDelete
  28. According to Choudhry, Bloomberg screen CDSW uses Hull & White, 2001 to calculate default probabilities. Now I have something to play with!

    Sorry to keep nattering on like this, but this is fascinating.

    ReplyDelete
  29. James:

    Please keep commenting whenever you feel the urge.

    I've used the Hull/White OAS calculator but I don't know what makes it unique off the top of my head.

    CFC CDS were in the 900 area not too long ago, so its gotten much better.

    ReplyDelete
  30. This comment has been removed by a blog administrator.

    ReplyDelete

Comment rules:
All comments must contribute to the conversation
All comments should be civil
No comment should include any personal attacks, however minor, on the author or other commenter.
Do not hawk your own website unless its a specific reference to the article
If you post anonymously, please give some identifyer
I will delete any comment which doesn't fit this criterea