Building Credit Risk Models from Financial Ratios
Summary
The discussion outlines a practical workflow for developing credit ratings or default-risk estimates from company fundamentals. First choose a target, such as observed defaults, agency ratings, or expert assessments, then assemble candidate financial predictors. Expert review can remove variables considered irrelevant; standalone analyses such as rank correlation can help identify weak predictors. The remaining factors can be standardized and evaluated in ordinal or ordinary logistic regressions across candidate combinations.
Model selection is described as both quantitative and judgment-based: inspect coefficient directions for economic sense, consider whether factor weights are meaningful, rank candidate models by association with the target, and review leading candidates with experts. The answers also mention published studies using country fundamentals, leverage, peer-group comparisons, CDS spreads, and ratings, alongside Altman’s Z-score as an established ratio-based bankruptcy indicator. A proprietary vendor model is mentioned but its internals are not available. The discussion provides a modeling outline, not a dataset, validation results, or a universal set of good and bad ratio thresholds.
Key ideas
- Choose the target carefully, since default histories, agency ratings, and expert rankings represent different outcomes.
- Screen candidate predictors with expert input and standalone analyses such as rank correlation.
- Standardize retained predictors before evaluating logistic or ordinal logistic regression models.
- Check whether coefficient signs and factor weights make economic sense, then compare candidate models by rank association.
- Ratio-based models require validation, and proprietary implementations may not disclose their methods.
Tags
Full text
# Credit Rating or Probability of Default from Financial Ratios # Credit Rating or Probability of Default from Financial Ratios Does anyone know of any papers about credit rating development or probability of default estimation done based on financial ratios that also include methodology and maybe good/bad criteria? Something like they have some financial ratios and then they have some methodology that reduces it to a few financial ratios and then they make a regression model out of it or something. ## Answer by Kyle Balkissoon (score 4, accepted) https://quant.stackexchange.com/a/16001 Most of the papers concern CDS spreads which you will need to convert to a PD. Paper using country specific fundamentals: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2517018 This paper uses leverage: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2361872 Another one that decomposes them against peer groups: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=2413011 Comparing spreads and ratings: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1551406 ## Answer by teucer (score 3) https://quant.stackexchange.com/a/16026 I am also not aware of any papers in this area. But having developed many such models, I can list the important steps: - Decide on the target variable: usual choices are historical default data, agency ratings and expert rankings - Create a sample containing the possible predictors - Reduce the list with the help of some expert, e.g. exclude all the predictors deemed to be irrelevant - Analyse the predictors standalone, e.g. with a rank correlation - Discuss the poor predictors with the experts and possibly eliminate some. At this stage one usually has about 10 to 20 predictors - Run regression analyses ((ordinal) logistic regression) after standardisation with all the possible combinations (e.g. consider combinations with at most 8 factors and more than 2 predictors). - Check for each combination: the coefficients are intuitive (e.g. higher asset/debt better the rating etc.) and that they are high enough (e.g. more than 5% weight). - List e.g. top 10 models using rank correlation and discuss them with the experts. ## Answer by Kiwiakos (score 2) https://quant.stackexchange.com/a/16046 Bloomberg has a Default Risk model, which is similar to what you are querying. You can see a screenshot in this PDF. There you can also see the kind of variables they use. You can access it by typing DRSK at the CDS screen is Bloomberg. (If the screenshot in the PDF is not clear enough, let me know and I can post one with better resolution from Bbg) This model uses fundamental data, and obviously they have calibrated and backtested it; however it is a proprietary model, therefore you might have a hard time finding the details. You can try googling for it, there might be a white paper on it. ## Answer by Bikenfly (score 1) https://quant.stackexchange.com/a/16035 Altman Z Score - http://en.wikipedia.org/wiki/Altman_Z-score - From Wikipedia - "The Z-score formula for predicting bankruptcy was published in 1968 by Edward I. Altman, who was, at the time, an Assistant Professor of Finance at New York University. The formula may be used to predict the probability that a firm will go into bankruptcy within two years. Z-scores are used to predict corporate defaults and an easy-to-calculate control measure for the financial distress status of companies in academic studies. The Z-score uses multiple corporate income and balance sheet values to measure the financial health of a company.
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.