Banks and other financial institutions are advancing their efforts in quantifying climate risk ever since the Task Force on Climate-related Financial Disclosures (TCFD) published a set of recommendations on the disclosure of climate-related risks. To date, however, no prescriptive methods are given and Banks have to work with guardrails and guidelines. Specifics such as methods of analysis, scenario inputs and granularity of disclosures are all left to institutions. Yet, this does not mean that Banks need to start from scratch. Their current toolbox, including scenario analysis, credit and sustainability experts and well-established (credit risk) modelling techniques, can also be employed for quantification of climate risks.
The most specific, freely accessible methodology is currently the output of a working group of sixteen banks convened by the UN Environment Finance Initiative. In the remainder of this blog we will outline the proposed methodology, which is split between physical climate risk and transition climate risk and provide some recommendations, with a focus on credit risk.
Within the financial industry, the main focus has been on transitions risks in the last few years. However, recent events such as the flooding of Limburg in July have shown that physical climate risks can also occur more often here in the Netherlands.
In the report, various banks have developed different methods to assess physical climate risks. The methodology differs per sector. For some sectors, like the agriculture and energy sector, emphasis has been placed on incremental climate changes (e.g. rising temperatures), while for the real estate sector the focus is on extreme weather events (e.g. a flooding).
An overview of the methodology is presented in the figure below:
Figure 1: Methodology for physical risks (source: UN Environmental Finance Initiative)
Although the methodology can differ per sector, the general set-up is similar:
Step 1: The physical climate risk scenarios need to be determined. The scenarios include climate changes, for instance a 2 degrees increase in temperature, but also a horizon when this change becomes effective, say 2040. The scenarios have been investigated to determine per region how they increase the occurrence of extreme weather conditions. An increase in temperature will increase the probability of long droughts, but also of flooding. An important factor in developing the scenarios is the physical location of the clients that are in scope of the portfolio of the bank.
Step 2: Next these scenarios are translated to changes in productivity per sector. For the agricultural sector it is for example investigated what the expected impact of the scenario is on the crop prices and agricultural yields. For the real estate sector the impact on the house prices is estimated. The estimated impact can be based on empirical evidence for those events that have already taken place, e.g. the impact of a storm on the crop production or for utility companies on the downtime of a plant, can be measured based on historical experience. Also (peer reviewed) studies can be used to calculate the impact, where banks are suggested to assume a worst case outcome.
Step 3: The changes in productivity impact the revenues and costs within a sector. These are estimated based on the calculated impact in step 2. For the real estate sector the calculated changes in the value of the properties impact their LTV. Within the calculation of the impact, insurances could be taken into account as mitigating measure.
Step 4: The estimated impacts can in turn be used by banks as input for their credit risk models to evaluate changes in credit risk of their portfolio. This will require banks to change their models to include all these variables and to make calculations on an individual sector basis.
As this methodology includes the investigation of new variables for which limited data might be available, banks can start investigating a representative sample of their portfolio on an individual basis. An important aspect of the above is setting proper and realistic scenarios. This requires collaboration between the climate and economic research community.
When dealing with transition risk, we must remind ourselves that whilst the realizations of physical risks have already been observed (e.g. storms and floods), this is not true for transition risks. Therefore the methodology for transition risks is more heavily dependent on expert input. See the picture below for an overview of the proposed methodology for transition risks:
Figure 2: Methodology for transition risks (source: UN Environmental Finance Initiative)
Step 1: Transition scenarios. There are plenty of climate scenarios with academic levels of research behind them, but not all of them are useful for financial assessments of transition risk. For starters, a consistent baseline scenario is necessary to understand the incremental risk relative to business as usual. Furthermore, the scenarios must provide sufficient granularity and at the same time cover all regions of the world where banks have exposures. Taking this all into account, the IEA World Energy Outlook and Integrated Assessment Models (IAMs) are deemed most appropriate. It is important to note that these IAMs already combine representations of global land-use and energy systems with internally consistent social-demographic and economic projections. From here the next step is to make a selection of economic projections that are relevant for Banks. The working group calls these “risk factor pathways” and developed four of them:
Via the use of relative sensitivities the sector-level risk is brought down to segment level . These sensitivities specify how different segments within a sector relate to the risk factor pathways. For example, whilst the car manufacturers sector as a whole may see a decline, the segment of electric car manufacturers may see an increase in revenue. Banks should customize the segment level to reflect their own view on how transition risks impacts their portfolio. The quantification of the specific magnitude of the risk driver’s impact on the segment, expressed as a calibrated sensitivity, is always identified through borrower-level calibration.
Step 2: Borrower-level calibration. In short, the borrower-level calibration specifies the relationship between climate scenarios and credit outcomes. This module builds on scenario variables, adds expert judgements and existing in-house credit risk tools to assess the changes to the probability of default of borrowers. To manage workload, this assessment is only conducted on a selection of borrowers. The analysis will result in calibration points that provide information for extrapolating borrower-level impact to the rest of the portfolio. It is at this stage that existing credit risk models can be used. Banks can simply estimate the impact of the scenario on rating factors and use the “stressed” inputs to compute a stressed PD.
Step 3: Portfolio impact assessment. In this last step the expected loss is calculated for the entire portfolio. EL is defined as the product of PD, LGD and EAD. The framework only addresses the impact on PD as it is argued that the impact on LGD is sector specific and EAD is assumed to remain constant. To arrive at a PD conditional on the climate scenario, the TTC PD is stressed via a segment and geography specific climate credit quality index. The credit quality index includes all previously mentioned ingredients; risk factor pathways, sensitivity parameters and calibration factors.
One of the major current issues the is lack of sufficient data. Collaboration with climate experts, which was also a vital part of input for the two methodologies, will continue to be of utmost importance. To speed up progress, the sharing of methodologies and data is a needed. Sharing of data and methodologies will also ensure that climate related stress tests across banks remain comparable.
Both physical- and transition risk methodologies make use of existing credit risk models. Although this is certainly a positive feature, it does impose some requirements on the existing models. Obviously the models need to make use of input variables that can be related to the change in economic circumstances that follow from the climate scenarios. Furthermore, the existing models may not be too convoluted with, at times conservative, requirements from the regulator. RiskQuest promotes the development of base models, which are “best estimate” in nature, and use overlays to meet specific requirements, for example IFRS9 and IRB. Such an approach would prepare the bank’s models for future usage for climate risk.