The impact of forward-looking assessments provided by the IFRS9 on credit portfolios

The introduction of forward-looking expectations in the credit risk assessment could give rise to an increase of the provisions’ volatility on banking financial statements. The application of alternative economic scenarios to determine losses introduces greater discretionary and variability, implicitly due to the scenarios’ design and probabilities of occurrence. This framework represents an element of discontinuity in relation to the traditional credit risk assessment models subject to Basel regulations. In this work, using econometric modeling techniques, the impact of forward-looking assessments under the IFRS9 accounting standard on credit risk estimates is estimated on a sample of Italian non-financial companies, adopting two alternative hypothetical economic scenarios.
Thanks to Professor Andrea Resti for the article’s review.


The application of the forward-looking approach as set out by the IFRS9 accounting standard has introduced a certain discontinuity compared to the practices already established in the banking system, both regarding stress testing and strategic planning. Indeed, for the first time expectations regarding the economic cycle are introduced directly into the expected loss inference models, and therefore provisioning models, through a more structured method which not only considers an expected scenario (baseline) and a worst case scenario such as in the stress testing framework, but also considers a “range of outcomes”, i.e. a range of alternative forecasts of the expected economic trend.

The methodology assumes that the losses relating to each of the possible alternative scenarios are estimated and that they are weighted by the probability of occurrence. Implicitly, therefore, the new forward-looking view introduces greater volatility in the loss estimations, attributable to the natural and inherent uncertainty about the model's structure and the implicit presence of discretionary elements in these assessments.