Historical recovery rate

Rookie_Quant · Post by **Rookie_Quant** » Wed Sep 24, 2014 4:38 am

I emerge from years of silence, now nearing the research portion of my PhD. I am working on structural models of credit, a la Leland (1996) and subsequent papers. One project of mine involves adding in some dynamics to the recovery rate 'process' and I'm looking to test some empirical predictions.

I have been able to find sporadic recovery rate data but am looking for a larger sample, perhaps even one that includes rec rates by class (senior, sub, etc.)

Is this type of data available?

Email is jb60 at "rice" dot (common US educational suffix. )

Cheng · Post by **Cheng** » Wed Sep 24, 2014 10:44 am

I don't have anything at hand right now but maybe a few hints. Edward Altman did a lot of research regarding recovery rates and was also the only one that comes to mind. His papers should be available on defaultrisk.com, look for the sources he qoutes. The rating agencies also have recovery data in their annual default studies but I think you have to look at every single issues. There was also a study about recovery rates, I think from Fitch.

Tl;dr

Yes, this type of data is available but it is tricky to get Smiley .

HTH.

sfca · Post by **sfca** » Thu Sep 25, 2014 5:53 am

Welcome back Rookie Quant. I think Moody's has a free publication that would work for you. Go to their website, create an account, and even for free accounts there are interesting resources. Look under research or some such heading to get "Annual Default Study: Corporate Default and Recovery Rates, 1920-2013 usually published around March for the previous years. Sometimes its not easy to find, but you will find it. Exhibit 20 has recoveries by seniority (you call it class) by year. If you look, there is also a companion excel file usually published in a different month where you can download this as data. I also have not posted here much lately because our IT weasels finally caught up with me and blocked just about everything so I can't post. I can read only. Defaultrisk is also a good resource as Cheng mentioned but for some reason I can't discover, they stopped updating the site about a year ago. Too bad. I really liked it.

Cheng · Post by **Cheng** » Thu Sep 25, 2014 5:45 pm

I also have not posted here much lately because our IT weasels finally caught up with me and blocked just about everything so I can't post. I can read only.

Do you have access to a Bbg ? You could tunnel through that.

polysena · Post by **polysena** » Thu Sep 25, 2014 9:35 pm

Rookie Quant:

I do not know which type of instruments you would be interested in (tranched, simple banking instruments, simple untranched traded instruments). Here some sources, these are no data samples. There are no data samples for free that I know of (all data are from dealogics, internal to a bank, or then for Araten on US banks which he got from either a mandate or being at the FRB). Is there any chance you can do some kind of internship in a bank then you would use the internal data set or the pooling data sets that are being put up in the last years, but then you'd probably not be able to publish your thesis... most banks do not have that many defaults internally to start with: small samples, bad predictability, bimodality of distributions.. for traded instruments, approaches are... basically take some assumption for the mean and then amuse yourself with using a distribution that is sophisticated... hope that helps or maybe not

1. JCR Spring 2012, Frye and Jacobs, Credit Loss and Systematic LGD

2. Altman-Kuehne High-Yield Bond Default and Return Report, February 2012

3.https://www.moodys.com/sites/products/DefaultResearch/2006200000430444.pdf

4.http://michaeljacobsjr.com/JPMC_LGD_Publication_May2004.pdf or http://www.defaultrisk.com/pp_recov_60.htm

5. Asarnow, Elliot, and David Edwards, “Measuring Loss on

Defaulted Bank Loans: A 24-Year Study,” Journal of Commercial Lending, 1995, Vol. 77, No. 7, pp. 11-23.

(4 & 5 remain the most cited studies because there is so little data really public to estimate LGd or recoveries; cites the altman siresti and other studies)

Poly

Rookie_Quant · Post by **Rookie_Quant** » Mon Sep 29, 2014 5:34 am

Greetings and thank you all for the helpful information. I have searched Moody's and have seen some company-level data, but only for the most recent year. Hopefully the resources you mention will have that data going back, even if I have to spend considerable time aggregating.

This may either help or further convolute the issue, but there are 2 general areas I am looking to explore, one involving structured credit and the other just cash bonds and single-name CDS.

In the latter case, I'm attempting to look at some cross-sectional variation in firms which are/are not reference entities for CDS, and whether a Leland-type model of default can be augmented for firms that may suffer from the "empty creditor problem."

In the former case, I am looking at potential pricing and firm leverage effects stemming from structured credit 'networks'. For example, are synthetic CDS a channel for credit pricing (or firm policy) in the following way: firm A and B are both reference entities in a synthetic CDO, does firm A's default impact the price of firm B's credit?

Admittedly, these may be either explored or impossible to get data for (or both), but I continue to trudge on...

unsmt · Post by **unsmt** » Mon Sep 29, 2014 10:12 am

If a structural model is ready to use I am curious how one chooses a boundary which represents default of a company. It should be a general rule which can be applied for any company.

Cheng · Post by **Cheng** » Mon Sep 29, 2014 12:21 pm

@Rookie:

Re empty creditor problem: wasn't this solved with the introduction of the ISDA 2014 protocol ? I think this was one of the reasons the protocol was amended.

Re CDS/CDO: I'm not sure I understand what is cause and what is effect (leave CDOs aside for the moment). Do you say that company A defaulting changes the price of company B's CDS (think of US autos for example...) ? Or do you say that price changes in CDS affect company A's or B's ceditworthiness (because bonds become more expensive to issue, loans harder to get etc) ? And either way, why should pooling within a CDO (or even CDX or iTraxx without tranching) have an impact ?

Cheng · Post by **Cheng** » Mon Sep 29, 2014 12:22 pm

@unsmt:

Usually you can infer something from the PD or from a mixture of long-term and short-term debt (a certain former three letter company is said to do this).

Rookie_Quant · Post by **Rookie_Quant** » Mon Sep 29, 2014 6:12 pm

@unsmt-

I certainly don't have a structural model that is ready to use, but as I begin looking at the literature one of the potential gaps I see is how the recovery rate process is handled. The early models seemed to just have a static boundary and a fixed capital structure. From there some simple dynamics were added but one thing that has always irked me is that models with dynamics allow for refinancing/adding debt and a potentially dynamic optimal default barrier, but don't address the interplay between these two barriers (as I see it).

Putting aside the simplifying assumptions of liquidation protocol and agency issues, it seems to me that the default barrier should at least in part be a function of a dynamic recovery rate expectation. If you think abstractly about a company with a single asset, say a large factory, the recovery rate on that asset in BK likely has a distribution. So look at the expected value, right? Well, I think some of the factors impacting the distribution also affect the firm's prospects in general.

Take liquidity. I don't think it's a stretch to say that if market liquidity dries up, the company may have an optimal default rule based on a recovery assumption that the market can't support. I do to sell my factory and instead of $.40 on the dollar, I get $.27, but if I had known that, my default decision would change, and perhaps my optimal default barrier is higher...

On the flip side, everything is rosy and liquidity is high. Firm prospects are good so my upper boundary condition tells me to add debt at the same time my default boundary is low. In the limit, this looks like a credit bubble, just as the above situation looks like contagion.

Maybe liquidity-based regime switching gets at these dynamics? Maybe both the EBIT process and the asset value process are handled as correlated GBMs? I don't know... but I'm interested in finding out.

==================================

@Cheng-

I need to check into the 2014 protocol. You may very well be right.

On CDS impacts, I think it's intuitive that GM defaulting impacts Ford's CDS prices, but I wonder if after controlling for factors common to firms, if "uncorrelated" firms are still impacted. I hypothesize (pre-data collection) that both effects you mention are positive and significant. Specifically, I wonder if defaults send pricing information through networks, but not the standard supply chain-type of network you think of (GM goes under and their largest OEM gets a revenue shock). I'm imagining a network whereby two firms are only "connected" via inclusion of their debt in CDOs (or CDS in synthetic CDOs). Perhaps the actual market is too small to be a material channel, but my thought experiment sort of goes like this:

I run a firm that is a reference entity in a synthetic CDO. 3 firms in the same pool as me default, but controlling for macro effects, I am economically unaffected. But as the pool has increasing losses, hedging trades cause spreads on my debt to widen in-kind.

It's possible that this effect is non-existent, negligible, or that I just don't understand the real world of synthetic CDOs well enough to be on point, but these are the things I'm looking at.