Climate Trends, Narratives, Shocks and Credit Risk Stress Scenarios
Credit risk measurement still doesn’t adequately capture the effects of climate change. Why is that, and what can be done to improve the modelling?
Thursday, October 26, 2023
By Scott D. Aguais
Climate change has taken center stage on a visceral basis as rising temperatures and unusual weather patterns around the world dominate the front pages of the media. As these observed climate changes become more volatile, substantial focus has been placed on ways to assess the complicated relationship between climate and economics. More narrowly, because the financial system plays a key role in supporting the world’s economic activity and growth, financial regulators have focused their attention on developing early approaches to assess climate impacts on financial risks.
In the credit risk domain, efforts to assess climate impacts are in a very early stage and are complicated by substantial, long-run uncertainty. Climate change impacts are also quite recent in history, therefore, empirical links between climate and measures of credit risk are for the most part not yet observable. In climate stress testing applied to banks, this lack of underlying empirical linkage has motivated the development of longer-run ‘what-if’ climate risk scenarios. Developing these ‘what-if’ scenarios is also the usual approach applied in capital-based stress scenarios due to limited historical experience with economic recessions and related large credit losses.
Scott D. Aguais
In assessing ‘what-if’ climate impacts on credit risk, there are potential lower-risk scenarios where a combination of climate adaptation, advances in green energy technology and a coherent global carbon policy could mitigate most potential future financial risks. Therefore, some commentators have suggested future financial impacts are probably small, and under several future outcomes this suggestion makes sense. However, other commentary has painted a potentially dire view of large, long-run temperature increases that could have significant negative economic and societal impacts. In principle, therefore, climate risk scenarios need to span a quite large set of possible future outcomes.
While a wide range of potential scenarios is required, to-date credit risk scenarios related to climate change have not fully incorporated the primary driver of credit risk (i.e., volatility) and there are problems with how the scenarios are developed. The three key problems will now be highlighted in more detail.
1. Climate and Economic Trends
Recent climate scenario analysis and stress test scenarios have been dominated by the global regulatory efforts summarized in the Network for Greening the Financial System (NGFS) scenarios. These scenarios span a range of potential long-run climate outcomes, primarily founded on ‘top-down’ macroeconomic assumptions, which are modelled using integrated assessment models (IAM). The economic projections in various NGFS scenarios such as ‘Net-Zero’ and ‘Disorderly’ also utilize additional socioeconomic pathways which posit various future changes in population growth, migration etc. Carbon mitigation policies are also integrated in the more detailed application of the NGFS scenarios. In addition, as developed by the ECB, various ‘top-down’ climate scenarios are also applied to company level credit models. The combination of ‘top-down’ NGFS scenarios and firm credit models are meant together to capture combinations of long-run physical and transition climate risk impact on firm-level creditworthiness.
While this overall combined approach dominates current climate risk thinking, there are several key limitations. These include, most notably, the lack of shocks (in the NGFS scenarios) that have historically dominated observed, abrupt increases in credit risk. In contrast to observed credit risk volatility, climate impacts in these firm-level models are driven more narrowly by relatively smooth, increasing climate-related costs that impact firms’ future profitability. In addition, increasing cost impacts are assumed to be only partially passed through to output prices, suggesting credit risk could trend higher. However, there is ample empirical evidence from cost pass-through studies to suggest that the assumptions of partial cost pass-through are probably unrealistic.
So far, these approaches combining smooth ‘top-down’ NGFS scenarios with firm level credit models — that generally exclude volatility — suggest climate impacts on credit risk are relatively small. Although applying mostly unrealistic, partial cost pass-through assumptions raise the suggested risk impacts relative to full cost pass-through, suggested risk impacts are still well below credit models reflecting observed volatility.
2. Climate Scenario Narratives
In response to the standardized NGFS scenario approach, recent commentary on alternative ways to assess climate impacts on financial risk have advocated the development of more detailed scenario narratives. The Real World Climate Scenario (RWCS) project is an example of industry efforts which focus less on ‘top-down’ macroeconomic trends and more on detailed climate narratives. As suggested by the RWCS, current climate risk approaches suffer from a lack of realism (such as volatility as a risk driver) and are hard to link to actionable climate decision-making.
Suggestions to apply detailed climate narratives stems, it seems, from both the lack of transparency and risk in NGFS scenarios as well as the lack of clear, objective, empirical links between climate and credit risk. Consequently, more detailed climate narratives that describe the relationship between climate economics and risk could be a useful alternative to primarily ‘model-based’ approaches.
Climate narratives are also suggested as more appropriate for short-run time horizons, such as to 2030, in comparison to the long-run NGFS scenarios which are applied to 2050. The NGFS seems to agree with this idea, as they have recently solicited input from the industry on ways to develop short-run scenarios.
Because potential future climate risk is complex and not well understood, long (or short) time horizon scenarios require better descriptive economic logic to complement more mechanical macroeconomic projections. More detailed climate narratives can be a key complement to complex models in general, but on their own descriptive narratives fall short of assessing the quantitative impacts of climate on measures of risk. Therefore, they are only part of an improved climate scenario development.
3. Climate and Credit Risk Shocks
Focusing on the usual drivers of credit risk, much regulatory effort has been made over the last 20 years to develop better credit models under the Basel capital regime. At the same time, observed credit risk in banks, measured by loan charge-offs, have varied substantially across economic cycles. Since 1990 the global economy has seen three recessions with substantially rising credit losses culminating in the Great Recession of 2007-2008. The observation of systematic, recurring credit cycles exhibiting high-credit volatility is clear to see. Therefore, credit risk is not understood by measuring average ‘through-the-cycle’ credit losses but by assessing more volatile systematic shocks leading to periodic rapidly rising credit losses.
Correctly-specified credit models therefore need to assess both average credit risk — individual company idiosyncratic risks — and the systematic observed volatility in credit risk. This is usually assessed by applying industry and region systematic credit-factor models.
Incorporating empirically observed credit risk volatility is thus surely required to assess complex future climate impacts that are highly uncertain. As current NGFS scenarios and detailed climate narratives are not able, on their own or together, to incorporate past empirically-observed credit risk variability, a key modeling dimension for improving the development of climate scenarios is currently absent from climate stress testing.
How Could Climate Stress Testing for Banks Be Improved?
Improvements in climate scenarios should focus on better integration of current top-down scenarios, with climate narratives and a more solid, empirical credit risk foundation. Applying more realistic, empirical, credit risk volatility in climate/credit models using detailed industry and region credit-factors could provide a big step forward that is more consistent with past observed credit risks.
Climate change at its core is driven by physical processes across the world that are in flux due to rising greenhouse gas concentrations and, in turn, rising global mean temperatures. These physical processes are extremely complex and diverse and not easily linked to general financial measures, much less the volatility historically observed in credit risk. In addition, IAM models were generally developed to assess broader climate social welfare cost trade-offs, but less so to be used for climate stress testing for credit risk.
Therefore, the focus on IAMs and the physical aspects of climate change have led to the exclusion of more robust systematic credit risk models. Climate stress testing for banks should be focused more on the systematic credit-factor drivers which dominate observed credit risk volatility.
Factor models generally, and in credit risk more narrowly, are models that are based on credit cycle variables — derived from past credit losses — that measure systematic correlation across observed credit defaults. Credit-factor models assessing systematic risk and volatility in specific industries and regions are well established empirically, can be derived from detailed market-based measures of credit risk, and are consistent with broader risk modeling in general. Systematic credit factor models can also be applied using both deterministic scenario drivers and factor simulations to assess future risk and volatility, including assessing climate credit risk ‘tail events.’ Dedicated industry/region credit factor models provide the more solid risk foundation which is currently missing in most climate stress test scenarios. Finally, they can also incorporate key brown/green carbon intensities that drive differential impacts of carbon mitigation shocks.
To substantially extend current climate stress test scenarios for assessing bank credit risk, credit factor models provide the ‘systematic risk glue’ that is currently missing and these models can also be integrated with detailed climate narratives. In addition, credit factor models can be integrated directly with firm-level climate adjusted models. Industry and region systematic credit-factor models are another missing component required for more realistic assessment of future climate-driven credit risk uncertainty over both short and long-term horizons.
Scott D. Aguais is managing director for the Z-Risk Engine solution and has led credit modelling teams for two global investment banks. He has been building and implementing wholesale credit models with his research partner Larry Forest for over 25 years and has published roughly 25 credit and climate risk modelling papers.