IFRS 9: Advanced Credit Risk Modelling in SAS Masterclass

Uncover survival analysis fundamentals—survival function, hazard function, and force of mortality—and see how they model time to event and inform IFRS 9 default and credit risk.

Apply the Kaplan-Meier estimator to handle censored data, estimate survival probabilities, and compute confidence intervals with Greenwood's formula in time-to-default and IFRS 9 contexts.

Discover the Nelson-Aalen estimator for cumulative hazard, a nonparametric method that sums hazard increments to map risk accumulation over time. Apply it to IFRS 9 lifetime modeling and credit risk.

Explore the Cox proportional hazards model, a semi-parametric approach linking covariates to hazard with partial likelihood, enabling censoring-aware lifetime PD modeling for IFRS 9 credit risk.

Build a core time-to-default survival dataset and apply survival analysis to IFRS-9 lifetime PD modeling, handling right censoring, event rules, and monthly data alignment for Cox and parametric models.

Use Nelson-Aalen cumulative hazard to quantify total risk buildup over time in credit risk modeling, a non-parametric, data-driven tool guiding IFRS 9 analysis, modeling, and portfolio diagnostics.

Apply the Cox proportional hazards model to quantify how borrower, loan, and macroeconomic factors shape the instantaneous default risk. Interpret hazard ratios to inform underwriting, pricing, and IFRS 9 modeling.

Leverage time-dependent covariates within the IFRS 9 point-in-time framework to dynamically update hazard rates and produce scenario-driven lifetime PD and PDS curves.

Translate hazards into survival probabilities and PD term structures with a Cox model to derive 12-month, 24-month, and lifetime PD for IFRS 9 provisioning and forecasting.

Validate and backtest time-to-default models for IFRS 9, using Kaplan-Meier survival curves, proportional hazards checks, calibration and discrimination, out-of-sample tests, and regulatory documentation.

Explore how stochastic processes power banking and insurance models, using Markov chains, Hossam processes, survival models, hazard functions, and PD term structures for credit migrations and longevity risk.

Explore transition intensities as the instantaneous rates driving multi-state, survival, and actuarial estimation. Learn how mu_ij(t) converts to probabilities and supports IFRS 9 credit risk modeling.

Explore exposed to risk, the total time at risk for a transition, and central exposure as the denominator in estimating transition intensities.

Introduces maximum likelihood estimation of transition intensities for multi-state credit risk models, building lifetime PD, transition matrices, and IFRS 9 ECL engines.

Study observational plans and waiting time structures for IFRS 9 credit risk, including interval censoring and multi-state transitions, and apply SAS survival models.

Learn to construct multi-state waiting time likelihoods for IFRS 9. Model survival terms, transition intensities, censoring, and paths across stages.

Apply the multi-state likelihood to model credit transitions by combining survival and transition components across stages, default, cure, and recovery to estimate lifetime PD and ECL for IFRS 9 portfolios.

Explore central exposed to risk and its role in IFRS 9 credit risk modeling. Learn how borrower timelines translate into months at risk to estimate transition rates and intensities.

Compare census and exact methods to estimate exposed to risk in transition intensity modeling, and explain when to use each for IFRS 9 credit risk.

Convert exposed-to-risk and transition counts into credible rates and intensities using rate intervals and actuarial estimation for IFRS-9 credit risk models.

Advance through single and multiple decrement models in IFRS 9 credit risk. Learn how competing events like default, cure, and prepayment determine exit routes, intensities, and transition matrices.

Convert transition intensities into monthly transition probabilities to build a transition matrix and enable PIT and lifetime PD modeling for IFRS 9 credit risk.

Construct monthly transition matrices from intensities to depict one-month movements between stages, with default as an absorbing state, to support lifetime PD and IFRS 9 PIT calculations.

Explore the IFRS 9 multi-state Markov framework with three stages, transitions among stage 1, stage 2, stage 3, and default, including absorbing behavior.

Compare stationary and non-stationary transition systems for IFRS 9 credit risk, showing how macro conditions drive time-varying transition matrices and shape PITPD and lifetime PD.

Apply Markov chains to project credit states forward using transition matrices and state vectors. Track stage evolution, calculate PIT, marginal, and lifetime PD under non-stationary matrices.

Propagate credit states month-by-month over multiple periods with non-stationary transition matrices to generate scenario-conditioned 12-month and lifetime PD and the IFRS 9 ECL term structure.

Compute lifetime PD from transition matrices and Markov chains. Generate monthly, scenario-weighted PD curves for IFRS 9 ECL.

Explore stationarity and integration in macro time series, including I0, I1, I2, unit roots, and how differencing informs AR, ARIMA, and VAR models for IFRS 9 forecasting.

Apply the backshift operator, lag polynomials, and filters to build ARMA and ARIMA models, perform differencing and seasonal adjustments, and forecast IFRS 9 macro trends.

Explore AR, MA, ARMA, and ARIMA models for forecasting macroeconomic variables in banking and IFRS 9, producing clean, stable forecasts and supporting scenario-based overlays and lifetime PD estimation.

Explore random walks, drift, and long-run macro behavior in IFRS 9 models, highlighting non-stationarity, differencing, ARIMA, and implications for lifetime PD and ECL forecasting.

Explore co-integration, error correction models, and VAR and VECM to model long-run macro relationships and short-run dynamics for IFRS 9 scenarios and lifetime PD.

Identify the Box-Jenkins steps for ARMA and ARIMA model structure, estimate parameters via maximum likelihood, and diagnose fit with residual analysis, Ljung-Box tests, and normality checks for IFRS 9.

Forecast macroeconomic variables for IFRS 9 using AREMA, VAR, or VAC to generate baseline, downturn, and upside scenarios that drive PITPD, lifetime PD, LGD, EAD, and ECL with interval estimates.

Implement smoothing, seasonal adjustment, and macro overlays to stabilize IFRS 9 inputs, ensuring PD, LGD, and ECL respond to macro shocks while meeting regulatory reporting standards.

Explore foundations of machine learning in actuarial and banking analytics, including the four branches—supervised, unsupervised, semi-supervised, and reinforcement learning—and emphasize interpretability, governance, and regulatory defensibility for credit risk.

Explain how supervised learning with labeled targets like PD, LGD, and EAD builds auditable IFRS 9 models, while unsupervised methods reveal segments to support governance.

Differentiate regression and classification models for credit risk and banking analytics, and apply appropriate methods for PD, LGD, and EAD under IFRS 9 and Basel.

Explore generative and discriminative models, comparing joint versus conditional learning and clarifying their regulatory implications for pd, lgd, and ead in IFRS 9 and Basel risk analytics.

Explore penalized regression techniques—lasso, ridge, and elastic net—and learn how they strengthen credit risk models by improving stability, reducing overfitting, and handling multi-collinearity in IFRS 9, BASL, and insurance pricing.

Decision trees blend interpretability with predictive power for banking and actuarial analytics, capturing nonlinear patterns and interactions, and guiding early warning, segmentation, and risk modeling.

Explore practical machine learning applications across actuarial science and banking analytics within the IFRS 9 lifecycle, including pricing, lapses, mortality, PD/LGD/EAD, segmentation, fraud, and regulatory governance.

Apply the Cox proportional hazards model to create time-to-default IFRS9 PD curves, including 12-month and lifetime PD with macro scenario overlays, and convert hazards to PD term structures for ECL.

Learn loan-level multi-state modeling from stage 1 through stage 2, default, cure, and re-default, to estimate lifetime PD, SICR, and ECL under IFRS 9.

Integrate forward-looking macro forecasts into IFRS 9 transition models to adjust lifetime PD and ECL using ARIMA, VAR, and scenario overlays.

Convert transition-based PD from a multistate model into a smooth, monotonic term structure using actuarial smoothing and mortality projection for IFRS 9, delivering stable PD curves for ECL.

Apply extreme value theory and copulas in SAS to model extreme losses, tail dependence, and downturn LGD for IFRS 9, using gev and gpd concepts in portfolio stress testing.

Explore stochastic processes to model credit, market, and operational risks and produce capital estimates under ICAAP and Basel, using SAS for frequency, severity, and loss simulations.

Master a SAS-based, end-to-end IFRS 9 expected credit loss model, from data foundations and segmentation to PD, LGD, EAD, macro scenarios, and governance.

Explore loss severity distributions for IFRS 9 downturn LGD, modeling skewed, heavy-tailed losses with SAS tools and distribution families to improve lifetime ECL calibration.

Explore how deductibles, excess layers, and proportional loss sharing shape IFRS 9 loss given default and tail behavior, with collateral protection and SAS-based layered modeling.

Explore how deductibles, excess layers, and retention limits shape credit risk and IFRS 9 LGD, revealing nonlinear downturn behavior, loss clustering, and layered loss modeling in SAS.

Evaluate goodness-of-fit for IFRS-9 severity and LGD models using visual diagnostics, QQ plots, and formal tests (KS, AD, CVM) to select robust distributions and assess tail behavior, AIC/BIC tradeoffs.

Apply extreme value theory to model the tail of credit losses and calibrate downturn lgd under IFRS 9, using gev and gpd for robust tail-based scenario overlays.

Explore extreme value distributions that model the tail of loss distributions in IFRS 9, using gev and gpd to quantify downturn loss given default and macro-driven scenario stress in sas.

Compare and rank severity models by tailweight to capture downturn lgd under IFRS 9, selecting light to heavy tails such as exponential to frechet for accurate loss forecasts.

Explore catastrophic losses in credit portfolios from housing market collapses and sovereign downgrades, and how extreme value theory informs IFRS-9 VAR and ES, tail amplification, and SAS-based stress testing.

Examine loss given default (LGD) as the severity of loss in credit risk, its drivers, and its role in IFRS 9 and Basel 3.1, linking PD and EAD to ECL.

Collect and structure a rich LGD data set, define the dependent variable, and establish observation and performance windows to build robust, audit-ready models under IFRS 9 and Basel 3.1.

Master linear regression as a transparent LGD benchmark linking loan characteristics, collateral, and macroeconomic variables to observed losses. Interpret coefficients, assess fit in SAS or Python, and validate for use.

Explore beta regression and fractional response regression to model loss-given default, preserving 0-1 bounds and handling asymmetry, with SAS or Python implementations and predictors like LTV, collateral, and GDP growth.

Inflated beta regression models LGD with zero and one boundaries plus a beta middle, enabling full recoveries, total losses, and partial recoveries for IFRS 9 and Basel.

Explore the mixed effects LGD model that combines fixed and random effects to capture unobserved recovery drivers (e.g., LTV, collateral, GDP growth) for IFRS 9 scenario forecasts.

Compare four LGD models—linear regression, beta regression, inflated beta, and mixed effects—through statistical and qualitative validation, ensuring accuracy, stability, and interpretability under IFRS 9 and Basel III.

Bridge IFRS 9 and Basel frameworks by applying LGD to ECL and RWA, guiding provisions and capital through point-in-time and through-the-cycle calibrations with scenario weighting and governance.

Develop an LGD modelling framework for IFRS 9 and Basel 3.1, using a loan dataset to build regulatory-grade LGD models with four techniques across data prep, estimation, validation, and integration.

Recaps LGD modelling, transitions to ECL, and links LGD to EAD and PD under IFRS 9 and Bazel 3 frameworks, with governance, validation, and macroeconomic considerations.

Learn how to prepare LGD model inputs, apply a logit transformation and inverse transform, and build an end-to-end SAS-based OLS model with train-test validation for IFRS 9 ECL.

Demonstrates beta regression with a logit link in SAS to estimate LGD for bounded 0–1 data, using LTV and other covariates, with train-test split and MAE and RMSE evaluation.

Master loss-given-default estimation with median quantile regression in SAS, handling skewed lgd, using ltv, rate, days past due, and categorical features, with training, testing, and exporting for ecl.

Explore decision tree regression for LGD in SAS with PROCHPSPLIT, pruning by cost complexity, and variable importance highlighting seniority and high LTV as key risk factors.

Apply simple linear regression to estimate LGD from LTV, interest rate, days past due, and the guarantee flag in a 70/30 train-test SAS workflow, clipping predictions to 0–1.

Model loss given default with a fractional logit and logit link to bound LGD, identify LTV and guarantee flag as key drivers, and evaluate MAE, RMSE for IFRS 9 compliance.

Develops a beta regression with random effects for loss-given default, using region-level intercepts and a 70/30 split to predict lgd with ltv, interest rate, and collateral factors for IFRS-9.

Explore the zero-one inflated beta framework for lgd in credit risk, combining multinomial logistic classification with beta regression for the fractional mid range, aligned with IFRS 9.

Explore exposure at default (EAD) as the final input in the credit risk equation, combining current drawn balances with expected future drawdowns under IFRS 9 and Basel.

Learn how credit conversion factors transform undrawn credit into drawn exposure for EAD estimates under IFRS 9 and Basel 3.1, using data-driven modeling, scenarios, and practical examples.

Design a regulator-ready data set for exposure at default modeling by defining observation and default windows, sourcing from loan, collateral, and default systems, preventing look-ahead bias with robust predictors.

Learn linear and log-linear EAD models as transparent baselines, linking exposure to drivers like utilization, rating, collateral, and GDP growth. Implement in SAS and Python, preparing for Tobit models.

Apply Tobit and truncated regression to bound EAD and CCF within zero to one, implementing in SAS and Python for IFRS 9 and Basel 3.1 compliance.

Master beta regression and fractional logit for EAD and CCF to produce bounded 0–1 predictions. Apply IFRS 9 and BASL 3.1 standards for risk modeling, scenario forecasting, and capital planning.

Model credit conversion factor using regression and logistic approaches to estimate drawdowns, then integrate CCF into EAD and ECL under Basel III and IFRS 9.

Validate EAD models and back test predictions across portfolios and cycles, applying IFRS 9 and Basel 3.1 framing, with data integrity, calibration, and governance that pass audits.

Integrate IFRS 9 and Basel 3.1 EAD modeling to align exposure at default with ECL and RWA, capital adequacy, using a common data foundation, PIT versus TTC, and cross-framework governance.

Lead an end-to-end transformation of raw corporate exposure data into a fully IFRS-9 compliant EAD framework, covering data preparation, advanced modeling, validation, and integration into expected credit losses.

Recap exposure at default (EAD), highlight its dynamic nature, and outline how PD and LGD converge into IFRS 9 ECL under Basel 3.1.