Show how Vasicek's model supports the estimation of capital requirements in the internal-ratings-based (IRB) approach of Basel II.
The idea
Vasicek's model assumes each borrower's asset value is driven by one systematic factor and an idiosyncratic factor, with asset correlation ρ, in a large granular portfolio. Conditional on a bad factor outcome, the default rate becomes the worst-case default rate. WCDR = N[(N⁻¹(PD) + √ρ × N⁻¹(0.999)) ÷ √(1 − ρ)]. Here N is the standard normal cumulative distribution function and N⁻¹ is its inverse. The Basel II IRB approach applies it at 99.9 percent over one year. Expected loss, PD × LGD, is handled by provisions, which is why PD is deducted from the worst-case rate. Capital per unit of exposure, before any maturity adjustment, is then calculated as K = LGD × (WCDR − PD). A maturity adjustment multiplies K where the approach requires it. Risk-weighted assets are then given by RWA = K × 12.5 × EAD. Common errors are using ρ instead of its square root, and omitting the deduction of PD.
Formulas
Vasicek worst-case default rate at 99.9%
WCDR = N[(N⁻¹(PD) + √ρ × N⁻¹(0.999)) ÷ √(1 − ρ)]
- N
- the standard normal cumulative distribution function, with N⁻¹ its inverse
- PD
- the borrower's one-year probability of default
- ρ
- the asset correlation with the systematic factor
A large, granular portfolio driven by one factor, at the IRB 99.9% one-year level. The correlation enters as √ρ on top and √(1 − ρ) below; using ρ itself understates the WCDR.
IRB capital per unit of exposure
K = LGD × (WCDR − PD)
- LGD
- loss given default
- WCDR
- the 99.9% worst-case default rate
- PD
- the probability of default
Before the maturity adjustment, which multiplies K where the approach requires it. PD is deducted because expected loss is covered by provisions.
Risk-weighted assets from the IRB capital requirement
RWA = K × 12.5 × EAD
- K
- capital per unit of exposure, after any maturity adjustment
- EAD
- the exposure at default
Converts a capital charge into risk-weighted assets. Leaving out the 12.5 reports the capital itself, and skipping the maturity adjustment understates K.
A practice question
In Vasicek's model as used in the Basel II IRB approach, the worst-case default rate (WCDR) is:
- the long-run average default rate, equal to the PD.Why not: Confuses the worst case with the unconditional average; the WCDR is much larger than PD.
- the one-year default rate that is not exceeded with 99.9% probability given the systematic factor. - correct
- the default rate in the worst year in the bank's own history.Why not: Uses a historical extreme; the WCDR is model-based at a stated confidence level.
- PD multiplied by LGD.Why not: This is expected loss as a fraction of exposure, not a default rate.
Conditional on a bad draw of the systematic factor (a 0.1% tail), the large portfolio's default rate is the WCDR; capital covers WCDR minus PD, scaled by LGD.
A harder one
Using Vasicek's model with PD of 2%, asset correlation of 0.15 and a 99.9% confidence level, the worst-case default rate is closest to (N^-1(0.02) = -2.054, N^-1(0.999) = 3.090):
- 4.2%Why not: Uses rho (0.15) in place of its square root in the numerator.
- 19.6%Why not: Omits the division by sqrt(1 - rho).
- 17.6% - correct
- 10.6%Why not: Uses the 99% quantile (2.326) instead of 99.9%.
WCDR = N[(-2.054 + sqrt(0.15) x 3.090) / sqrt(0.85)] = N[(-2.054 + 1.197) / 0.922] = N(-0.929) = 17.6%. The correlation enters as sqrt(rho) and the denominator is sqrt(1 - rho).
The hardest tier
A bank has EAD of USD 50 million, PD 1%, LGD 45% and a Vasicek WCDR of 14.55%. The maturity adjustment is (1 + (M - 2.5) x b) / (1 - 1.5 x b) with M = 4 and b = 0.1375. Risk-weighted assets are K x 12.5 x EAD. RWA is closest to:
- USD 57.9 million - correct
- USD 38.1 millionWhy not: Leaves out the maturity adjustment of about 1.52.
- USD 4.6 millionWhy not: Reports the capital requirement and omits the 12.5 multiplier.
- USD 46.0 millionWhy not: Applies only the numerator of the maturity adjustment, 1 + 1.5 x 0.1375 = 1.206, and omits the denominator.
K before adjustment = 0.45 x (14.55% - 1%) = 6.10%. MA = 1.20625 / 0.79375 = 1.5197. K adjusted = 9.27%. RWA = 9.27% x 12.5 x 50 = 57.9 million.
In the app: 6 practice questions on this outcome, spaced review that brings it back before you forget it.
Start free →More in Credit Value at Risk
- Compare credit VaR with market risk value at risk (VaR) by how each is defined, the horizon it covers and the tools used to measure it.
- State what Credit VaR is and compute it.
- Outline how rating transition matrices are used to derive credit VaR.
- Explain how each of Vasicek's model, CreditRisk+ and CreditMetrics derives the probability distribution of default losses and models default correlation.
- State what credit spread risk is and weigh its effect on the computation of credit VaR.