- While stress testing investment portfolios is increasingly a focus of climate-risk analysis, understanding the methodology of models such as the MSCI Climate Value-at-Risk (Climate VaR) Model is key to this analysis.
- Linear regression can help overcome the complexity of the input-output variables used in our Climate VaR model and make it easier for investors to understand their primary drivers.
- Using this methodology, we observed a near linear and stable relationship between MSCI Climate VaR results and input variables over time (when there are no changes to the model), which leads to the conclusion that error margins in the input variables were not amplified by the model.
Stress testing investment portfolios under different climate scenarios is a key requirement of the Task Force on Climate-related Financial Disclosures’ (TCFD) framework for climate risk management, as well as for a growing list of industry regulators and supervisors. The MSCI Transition Climate Value-at-Risk (Climate VaR) Model simulates the potential impact of future decarbonization costs and opportunities on companies’ valuation levels for both equity and fixed-income securities, the Transition Climate VaR being the sum of MSCI’s Transition Risk Climate VaR and Technology Opportunity Climate VaR. The Transition Risk Climate VaR, or policy risk, which measures scenario-specific decarbonization costs for companies, is the focus of this blog post.
Insights using regression analysis
MSCI Climate VaR simulations are based on several input variables, including companies’ Scope 1 and 3 emissions, electricity consumption and enterprise values including cash (EVIC). Within these simulations, we make assumptions about how the input variables and additional parameters (e.g., shadow carbon prices) will drive decarbonization costs over the next five decades to quantify Transition Risk Climate VaR.1
The complexity of the input-output variables used in our Climate VaR models may make it difficult for investors to understand the primary drivers. One way to provide transparency is to run a cross-sectional regression of Transition Risk Climate VaR for a given scenario and universe using the above-mentioned variables as inputs into the regression. For our purposes, we used the MSCI ACWI Investable Market Index (IMI).
The regression variables for Transition Risk Climate VaR are Scope 1, 2 and 3 EVIC-based emission intensities and the 11 Global Industry Classification Standard (GICS®)2 sector dummy variables. The EVIC-based emission intensity data was winsorized between 1st and 99th percentile to reduce the effect of outliers in the regression, then standardized.3 We measured the size and stability over time of each of the regression coefficients for each input variable, as detailed in the exhibit below.
Cross-sectional regression coefficient of MSCI Transition Risk Climate VaR model for the MSCI ACWI IMI
The winsor Z function identifies outliers based on Z-score cut-off and replaces with the next most extreme non-outlier value. This involves z-scoring the variable and identifying/replacing any cases beyond the z-score threshold. Data for the period Sept. 30, 2021, to May 30, 2023. Source: MSCI ESG Research
Overall, we observed regression coefficients that behaved as expected and were quite stable over time (September 2021 to May 20234), in between periods of significant model enhancements (e.g., April 2023). We also observed that higher emission intensities in Transition Risk Climate VaR lead to larger simulated losses in valuation.5
Scope 1 emission intensities were clearly the strongest driver of Transition Risk Climate VaR, reflecting that the most significant burden of a transition climate shock for companies is largely driven by their direct emissions. With the majority of transition climate risk being located primarily in Scope 1 emissions (for nonfinancial companies), the reduction of these emissions could become a focus for risk management.
Scope 3 emission intensities were the second strongest driver. Here, we see the importance of avoiding the double counting of Scope 1 and 3 emissions in climate scenario analysis. Transition Risk Climate VaR applies a deduplication factor to correct for Scope 3 emissions being double-counted in Scope 1 emissions within a closed-system environment. Additionally, the relative importance of Scope 3 emissions increased following model enhancements in April 2023 and updates to how value-chain-scenario-specific emissions-reduction requirements are calculated in the model.
R2 of cross-sectional regression coefficient for MSCI’s Transition Risk Climate VaR model
Data for the period Sept. 30, 2021, to May 30, 2023. Source: MSCI ESG Research
The explanatory power (R2) of the regression for Transition Risk Climate VaR remained above 73% over the entire study period. The timing of significant moves in the explanatory power coincided with meaningful model updates and enhancements that occurred in December 2021 and April 2023. The most recent enhancements included the adoption of additional scenario-input parameters from the NGFS, which resulted in an improved explanatory power of around 80%. It also led to a stronger coefficient for Scope 3 emissions, as shown in the previous exhibit.
The table below shows the statistical significance of the regression inputs. The Transition Risk Climate VaR Model was additionally tested for nonlinear effects by introducing each input variable squared in the regression. Simply put, these linear models successfully captured the high-level relationship between primary input variables and Transition Risk Climate VaR valuations over the study period and helped build an intuition of the main drivers of this component of Climate VaR.
Regression statistics for MSCI’s Transition Risk Climate VaR model
Data for the period Sept. 30, 2021, to May 30, 2023. Source: MSCI ESG Research
Conclusion
Although climate stress-testing models like MSCI Climate VaR are complex, the model’s main mechanisms can be explained using linear regression, which can help transform input variables into risk outputs. The main driver for the Transition Risk Climate VaR component — Scope 1 emissions — as well as all other drivers showed a relatively stable relationship to resulting Transition Risk Climate VaR over time.
1Green revenues are associated with alternative energy, green buildings, energy efficiency, sustaianble water and agriculture and pollution prevention. Green patents are representative of companies’ low-carbon research and development capacities. Shadow carbon prices reflect countries’ ambitions and policy actions to mitigate climate change.
2GICS is the global industry-classification standard jointly developed by MSCI and S&P Global Market Intelligence.
3Clean technology and green patents were not winsorized because the large amount of zeros in those datapoints are known not to be outliers. Z-score standardization is then applied.
4September 2021 coincides with the adoption of the 1.5°C Network for Greening the Financial System (NGFS) scenario that is presented here.
5EVICs are inverted since they appear in the denominator of emission intensities.
Further Reading
Foundations of Climate Investing: How Equity Markets Have Priced Climate-Transition Risks