Incorporating physical climate risks into banks' credit risk models
Summary
Focus
The Group of Central Bank Governors and Heads of Supervision (GHOS), the oversight body of the Basel Committee on Banking Supervision, has agreed to prioritise further analysis on the financial risk implications of extreme weather events and tasked the Basel Committee with analysing the impact of such events on financial risks. However, a major obstacle for banks is the absence of generally accepted industry models of credit risk adjusted for physical risk factors. This raises the question of whether banks should build their own internal models to account for climate-related adjustments to the internal probability of default and loss-given-default.
Contribution
We propose a way to integrate physical risks into credit risk modelling. Our methodology uses an extension of the model which underlies many internal credit risk models at banks, as well as the regulatory internal ratings-based (IRB) approach from Basel II and Basel III. Physical risk is treated as a single stochastic factor that manifests itself in a binary way with an externally defined probability and a jump in the market value of assets of an individual corporate borrower. A portfolio model is then developed to measure how much physical risk contributes to credit risk losses. Beyond typical uses like economic capital allocation for unexpected credit losses and loan loss provisioning for expected losses, the model could also help banks to manage this new factor of credit risk by hedging it with derivatives on climate-induced damage.
Findings
The model could be of interest for both banks and regulators. Its design and choice of parameters accommodate both a directional climate change (ie climate risk) and weather risk due to increased volatility of climate risk factors. It preserves the so-called portfolio invariance property, ie the invariance of the risk measure for a single credit claim to the composition of the loan portfolio to which it is added. This important property is highly desirable from a practical viewpoint to avoid time-consuming full recalculation of a risk measure on a portfolio level, and because of its potential suitability for regulatory purposes. We also find further possible extensions of the model if physical risk manifests itself in more than one state.
Abstract
Over the past few years, physical risks have turned from a niche domain of (re)insurers into a systemic risk factor that may have an impact through various channels on financial markets and financial institutions alike. While physical risks are not a common income-producing or even a sizeable cost-ofbusiness risk factor for most banks, they do affect banks, mostly indirectly, through their loan and trading books. Against this backdrop, standard setting bodies and financial regulators have increasingly called on banks to recognise physical risks as an additional factor in their risk space and internalise it in their risk management policies.
A major obstacle for banks on this way, however, is the absence of generally accepted industry models of credit risk adjusted for physical risk factors. Such models are increasingly needed to account for physical risks in banks' capital requirements, loan loss provisions, pricing of loans and, eventually, derivatives to hedge this risk. This poses the question of building a bank's internal model for climaterelated correction to the internal probability of default and loss given default or using third-party databases on the type of the borrower's assets, their geolocation, exposure to climate factors, statistical description of weather events and damage functions.
This paper proposes a methodology that allows in a relatively simple way the integration of physical risk component into the credit risk modelling, using an extension of the one-factor Vasicek model. The model described by the paper may be of specific interest for both banks and regulators, as it preserves important properties of models currently used while allowing for an informed mitigation of physical risk factor in credit risk. Additionally, the paper discusses further possible extensions of the credit risk model if physical risk manifests itself in more than one state.
JEL classification: C60, G17, P28, Q54
Keywords: climate risk, physical risk, credit risk, risk modelling, Vasicek model