Journal of Risk Model Validation

Risk.net

A point-in-time–through-the-cycle approach to rating assignment and probability of default calibration

Alexander Petrov and Mark Rubtsov

  • This article focuses on rating-level PDs, covering both corporate and retail portfolios.
  • We show a way to construct TTC rating grades and to obtain PIT PDs needed for IFRS9.
  • We quantify the size of estimation errors, producing a consistent MoC required by the regulators.
  • The methodology has direct implications for validation tests in Model Risk Control Units.

ABSTRACT

In order to be compliant with the Basel regulations and the upcoming International Financial Reporting Standard 9, banks need two probabilities of default (PDs): point-in-time (PIT) and through-the-cycle (TTC). The existing methodologies assume that banks already have a customer-specific hybrid PD, which is then adjusted to PIT and TTC. We avoid this assumption and focus instead on a two-step process to obtain a PD, namely, rating assignment and calibration. In the first step, we propose a methodology for constructing TTC rating grades and assessing the resulting degree of PIT-ness. For calibration, we let the rating-grade default threshold be stochastic. This move enables us to quantify the impact of estimation errors, provides a justification for the size of a regulatory margin of conservatism and has direct implications for validation tests. We illustrate our proposals on a sample portfolio of corporate customers, although we believe these ideas should be applicable in retail too.

Addendum to Rubtsov and Petrov (2016): “A point-in-time–through-the-cycle approach to rating assignment and probability of default calibration”

To continue reading...

You need to sign in to use this feature. If you don’t have a Risk.net account, please register for a trial.

Sign in
You are currently on corporate access.

To use this feature you will need an individual account. If you have one already please sign in.

Sign in.

Alternatively you can request an indvidual account here: