Basel IRB Credit Risk Models: A Practical Guide in R

Understand risk weighted assets and their risk weights, and how capital adequacy ratio and Basel III buffers, including the 2.5% capital conservation buffer, determine required bank capital.

Calculate Basel II/III minimum capital for credit risk using the standardized approach by applying risk weights to exposures such as sovereign, bank, corporate, mortgage, retail, and securitization.

The Basel A-IRB framework defines probability of default as the one-year performance period default likelihood, with a 90 days past due threshold, through-the-cycle PD, and segments data into homogeneous pools.

Create a custom R function to compute concordance, discordance, and tied pairs from a fitted logistic regression model, outlining how to interpret the rank-based discriminatory power and C statistics.

Validate a probability of default model using real-world credit data in r, assessing discriminatory power with AUC and Gini, and calibrating with the Brier score.

Understand loss given default (LGD), the portion of exposure a bank expects to lose after default, and how recoveries under Basel guidelines influence pricing, capital, and risk management.

Explore downturn LGD in Basel regulatory capital, deriving LGD from recovery rate and considering collateral, seniority, recovery costs, and the economic environment to ensure prudent loss estimates.

Explore how cooling off analysis sets the workout period for LGD modeling by tracking recoveries over time. Apply chain ladder extrapolation to incomplete recovery data for accurate LGD estimates.

Apply the workout method to measure LGD by discounting post-default cash flows and incorporating costs, leading to recovery rate insights and Basel discounting approaches.

Use the chain ladder (runoff triangle) method to estimate LGD by extrapolating missing recoveries from complete and incomplete data. Compute LGD as one minus total recovery over exposure at default.

Explore Basel's rule that LGD stays within 0 to 1, explain why negative LGD and above 100% values are capped, and how the IB approach uses long-run averages under stress.

Segment data by product, collateral, and risk drivers to capture distinct recovery and LGD patterns across mortgages, auto loans, credit cards, and SME loans.

Prepare the LGD data from defaulted accounts in the observation period, use chain and ladder methods to extrapolate recoveries, and consider cure and time value of money.

Explore a mortgage dataset prepared for lgd model development, featuring loan-to-value ratio, recovery rate, lgd values, and transformed variables (log and probit) plus loan purpose and default event indicators.

Learn to load data into RStudio with read_xls, inspect the LGD data frame of 2542 observations and eight variables, and compute descriptive statistics using str, summary, and scheme.

Build a linear regression benchmark for LGD using LTV and loan purpose, note its limitations for 0–1 data, non-normal residuals, and heteroscedasticity, and preview advanced models.

Explore exposure at default (EAD) as the most conservative balance at default under Basel, and learn to model CCF for revolvers and calculate EAD for term loans.

Learn how to model exposure at default for off-balance-sheet exposures using the credit conversion factor. Explore drawn amount, limit, headroom, and Basel guidelines that determine the exposure at default.