Limits of CreditGrades for Bank CDS and Alternatives for Default Risk
Summary
The discussion considers whether CreditGrades can estimate a one-year CDS spread and survival probability for a large Canadian bank when its CDS market is illiquid. The model is described as representing a firm through stochastic asset value and a debt threshold, with default when assets fall below that level. One answer argues that this simplified balance-sheet picture poorly fits banks: their assets differ in character from the model’s assumptions, while their liabilities and capital structure can change. It therefore cautions against relying on an unmodified CreditGrades estimate for a financial institution.
The answers also identify possible directions for further analysis. One emphasizes the potential importance of government support or bailout expectations in assessing a major bank’s credit risk; another points to research that adapts a Merton-style distance-to-default framework for financial firms. These are suggestions, not a validated estimate for the named bank. The thread provides no calibration, survival probability, or evidence quantifying bailout support, and does not establish that the proposed alternative resolves the illiquidity problem.
Key ideas
- CreditGrades models default as the firm’s assets falling below a debt threshold.
- The simple asset-and-debt representation may fit financial firms poorly.
- Banks’ asset composition and changing liabilities complicate distance-to-default modeling.
- Government support expectations may matter when assessing a major bank’s credit risk.
- A modified Merton framework for financial institutions is suggested, but no estimate is calculated.
Tags
Full text
# Can CreditGrades CDS Pricing Model be used for financial firms? # Can CreditGrades CDS Pricing Model be used for financial firms? For Canadian banks, the CDS market is very illiquid and inactively traded. I want to get an estimate for the spread for a one year CDS on the Bank of Montreal. I was going to estimate this using the CreditGrades model ( http://www.creditrisk.ru/publications/files_attached/cgtechdoc.pdf ). However, they mention that some algorithms (debt per share) may not hold for financial firms. Can CreditGrades be adapted to work for financial firms? Does it even make sense to get a spread estimate in this way? My goal is to find the one year survival probability for BMO. ## Answer by Brian B (score 1, accepted) https://quant.stackexchange.com/a/21249 The capital structure of financial firms, especially one like Bank of Montreal, is indeed quite unlike the simple debts-and-assets model of CreditGrades. For context, the model is basically one that says there are some assets $A$ subject to a stochastic process, and a debt level $L$ (unknown at the present time). If ever $A<L$ the company defaults. They provide some calibration techniques and so on to make all of this work. In any case, the picture of assets, where their value but not character varies, is completely wrong for a bank. The debt also is in flux. Thus, the model is essentially worthless. I personally would price CDS for a big bank almost any other way. My top consideration would be government bailout probability should they get in trouble. For BMo, I rate that probability quite close to 1.0. ## Answer by e.mal (score 2) https://quant.stackexchange.com/a/28027 (Although the question was asked long time ago it may be of help for others as well) You may want to have a look at Nagel and Purnanandam (2015) Bank Risk Dynamics and Distance to Default (https://www.bundesbank.de/Redaktion/EN/Downloads/Bundesbank/Research_Centre/Conferences/2016/2016_06_10_eltville_08_paper_nagel.pdf?__blob=publicationFile) The authors propose a modification of the Merton's model specifically for financial institutions
Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)
This summary was written by Stratmill's research agent from the original; it is not a copy of the source.