Module 23: Assessment of credit risk Flashcards
Credit risk may be decomposed into: (2)
- default risk - including, for example, the risk of loss due to a payment due being missed
- credit spread risk - the risk of changes in value due to changes in the credit spread (although this may be alternatively categorised as a component of market risk)
Default risk in a portfolio can be assessed as 3 components
- the probability of default in respect of each counterparty
- the loss on default - a function of the exposure and the likely recoveries in the event of default, both of which could be uncertain
- the level and nature of interactions between the various credit exposures and other (non-credit) risks
Information to assess credit risk may be sought from: (4)
- the credit issuer
- the counterparty
- publicly available data
- proprietary databases
Assessment of credit risk is difficult due to: (4)
- lack of (publicly available) data
- skewness of loss distributions
- complex, and uncertain interdependencies
- model risk
Qualitative credit models
Such models assess both default and credit spread risks using relevant factors, including:
- the nature of the borrower
- financial ratios
- economic indicators
- the nature and level of any security
The subjective nature of qualitative models are their main advantage and disadvantage.
Quantitative credit models
Such models convert financial data into a credit measure, eg probability of default.
Examples include:
- credit-scoring
- structural (or firm-value)
- reduced-form
- credit-portfolio
- credit-exposure
Merton model
Considers equity shares as a call option on the company’s assets.
The Black-Scholes option-pricing formula is used to value the shares. The model produces an estimate for the credit spread for a bond (even if it is unquoted) but the model makes a number of unrealistic assumptions.
KMV model
Estimates the probability of default based upon empirical data on company defaults and how these defaults link to the distance to default (the gap between the current value of the company’s assets and the judged default threshold). It has some advantages over the Merton model, eg it can accommodate more realistic liability structures.
Credit-migration models
Estimate how a credit rating might change over longer periods.
Historical data is used to determine rating-transition probabilities. Matrices of such probabilities are applied (repeatedly) to a company’s current rating to estimate the likelihood of each possible rating state in each future year.
The chance of default by a particular company in a future year is estimated as the assumed probability of default for companies with a particular rating in that year.
The general approach is to assume that the migration process follows a time-homogeneous Markov chain. The time-homogeneous assumption has been criticised using empirical evidence. It also assumes that the likelihood of default (“through the business cycle”) can be determined solely by the company’s credit rating.
Key challenge when modelling the behaviour of a credit PORTFOLIO.
To understand the relationships between the various credit exposures, eg jointly-fat tails.
5 Approaches to modelling credit PORTFOLIOS
- multivariate structural models
- multivariate credit-migration (or financial) models
- econometric or actuarial models
- common shock models
- time-until-default (or survival) models`
multivariate structural models
eg multivariate KMV,
modelling asset values by using correlation matrices or copulas
multivariate credit-migration (or financial) models
eg multivariate CreditMetrics,
which assumes that equity returns can be modelled using country-specific indices and (independent) firm-specific volatility.
econometric or actuarial models
Do not model the asset value going forwards, but estimate the default rate of firms using external or empirical data.
common shock models
determine the probability of no defaults by assuming that each bond defaults in line with a Poisson process, and considering shocks, each of which cause the default of one or more of the bonds in the portfolio
time-until-default (or survival) models`
where survival CDFs (each based on a hazard rate determined from an implied probability of default) are linked by a suitable parameterised copula function, so as to estimate the aggregate default rate for the bond portfolio.
How might recovery be measured?
Recovery might be measured as price after default or ultimate recovery. It is dependent upon many factors including: availability / marketability / liquidity of collateral, seniority of the debt and the rights of other creditors.
Estimated future recovery rates (or stochastic models) may be based upon historical recovery rates (and their volatility).