Boosting IFRS 9 PD with Survival Analysis

IFRS 9 Expected Credit Loss (ECL) calculations require robust estimations, with Probability of Default (PD) being a cornerstone. While seemingly straightforward, accurately forecasting PD, especially on a conditional basis, presents a significant challenge for financial institutions. Moving beyond simple historical averages, we need methods that capture the dynamic nature of credit risk throughout an instrument's lifetime.

The core of conditional PD isn't just about the likelihood of default today; it's about the probability of default *at a future point, given that a customer has survived up to the present*. Traditional approaches often struggle to model this dynamic progression, particularly when considering how external factors or borrower behaviour evolve over time. They might not fully capture the nuance of 'time until default' or appropriately handle customers who haven't defaulted yet.

Introducing Survival Analysis

This is where survival analysis techniques become invaluable. Originating in medical research (think "patient survival rates"), survival analysis is a statistical method designed to model the time until a specific event occurs. In our context, that "event" is default. It allows actuaries and risk professionals to understand not just *if* an event will happen, but *when* it's likely to happen, taking into account various influencing factors.

Survival analysis brings several key advantages to conditional PD estimation under IFRS 9. Firstly, it naturally incorporates time-varying covariates. This means that macroeconomic forecasts, changes in a borrower's financial health, or even shifts in market conditions can be integrated dynamically into the PD model. Secondly, it expertly handles "censored" data – those active accounts that haven't defaulted yet but are still 'at risk'. Ignoring these would lead to biased and typically underestimated PDs.

Modelling the 'Journey to Default'

By employing models like the Cox proportional hazards model or parametric survival models, institutions can estimate hazard rates – the instantaneous probability of default at any given time, conditional on having survived up to that point. This directly translates into more accurate conditional PDs, offering a far more nuanced understanding of credit risk. It allows for the projection of PDs that adjust dynamically as time progresses and as a borrower's credit journey unfolds.

The application of survival analysis provides a more granular and forward-looking view of credit risk, which is exactly what IFRS 9 demands. It moves beyond static historical observations, enabling institutions to produce more robust, sensitive, and explainable conditional PD estimates. This leads to more accurate ECL calculations, better capital allocation, and stronger compliance with regulatory requirements, ultimately enhancing overall risk management capabilities.