Using Credit Memo AI to Navigate Adverse Scenarios in Credit Risk Assessment

Macroeconomic & Geopolitical Risk: In an interconnected world, it is important not only to rely on timely macroeconomic forecasts for individual countries based on expert insights but to ground them in a robust modelling framework that factors in trade flows, financial flows, and other inter-country relationships, via a model such as S&P Global’s Global Link Model. The macroeconomic variables can then be used to produce transition matrices of credit scores using tools such as the Macro-Scenario Model from Credit Analytics, which has been developed on the historical relationships between macroeconomic factors and S&P Global Ratings.[1]

Climate Risk: Similarly, a two-step approach is best suited to assess the impact of both transition and physical risks. The first step involves projecting financial data for corporates, residential and commercial mortgages, government bonds, and financial institutions that are stressed due to transition and physical risks. Transition risk can be evaluated using the transition scenarios of major global regulators in conjunction with emission data from sources such as Trucost; physical risk can be assessed by estimating the impact of flooding, drought, tornadoes, and other disasters on physical asset values. The second step involves using projections from credit risk models to generate new PD/scores, utilizing credit risk models tools like Climate Credit Analytics and Climate RiskGauge

Sustainability and Cyber Risk: An overlay approach is best suited to assessing the impact of the Environmental, Social, and Governance (ESG) factors that are relevant to credit risk and cyber risk, given the limited coverage of data. To this purpose, we employ advanced Machine Learning techniques on Ratings Agencies’ research to identify the sustainability considerations that are relevant to credit risk and then apply a Scorecard with ESG components for low default portfolios, or a quantitative overlay for SME portfolios, to adjust the final credit risk score, Credit Assessment Scorecards with ESG factors. The cyber risk dimension is also captured with alternative datasets that assess the robustness of internet, technology and operational processes of relevant firms.

Credit Memo Automation: To enhance the efficiency of credit analysis, S&P Global has harnessed the power of Generative AI. By developing expert agents tuned to AI-ready and high-quality financials, news, and Ratings Research content, analysts can significantly reduce the time spent pre-populating their credit memos. They can also enrich these memos using prompts to integrate other proprietary documents, early warning signals, and, of course, the impact of adverse scenarios, saving valuable time and effort. You can read more about this in our blog, Credit Memo Automation: Where AI Meets Analyst Efficiency.

Let us look at a practical example on Geopolitical risk: Let’s consider the impact of U.S. tariffs on Italy’s trade flows.

We leverage a macroeconomic scenario based on the following assumptions:

• The United States will impose a tariff of 60% on China, 25% on Canada and Mexico, and 20% on the rest of the world.
• Full retaliation will occur from China, Mexico, and Canada, with energy goods from Canada excluded.
• The European Union will retaliate on final consumer goods

The first question is: how important is the trade flow from the United States to Italy? Very important! The image below is from our Global Trade Atlas, which visualizes the data on trade flows between countries. Focusing on Italy, the United States is one of the three main export countries, along with France and Germany.

[1] Lower case nomenclature is used to differentiate the statistical credit scores generated by S&P Global Market Intelligence from the actual ratings produced by S&P Global Ratings.

The next step involves evaluating the quantity and nature of exports by double-clicking into the trade flows, which is necessary for conducting a more granular analysis of sectors or company-level data. From this analysis, we can see that pharmaceuticals and construction materials are among the main categories. 

Impact on credit risk evaluating changes in Probability of Default: using the two-step approach illustrated above, we have leveraged the Global Link Model to generate the macroeconomic variables for this scenario and the Credit Analytics Macro-Scenario Model to generate transition matrices.

We focused on assessing the impact on publicly traded companies across different sectors of the economy.

However, even for public companies, we cannot rely solely on S&P Global Ratings’ issuer credit ratings; for example, in Italy, there are about 450 listed companies, but fewer than ten percent are rated. To complete the analysis, we used our quantitative models from the Credit Analytics suite

As expected, there will be an impact on Italy, despite it not being one of the countries most impacted by the scenario. In terms of sectors, the construction and materials sectors are more impacted, as usual in the case of a partially recessionary scenario. Even the healthcare sector, which is usually very robust, will suffer in this scenario, ranking fifth in terms of most impacted sectors.

In summary, the best practice is to use multiple strategies to model adverse scenarios depending on the factors that need to be assessed. Data and modes to quickly assess the impact, along with tools to automate the process to incorporate the analysis into credit risk assessment (credit memos), are critical to the ability to connect different considerations, such as economic forecasts and trade flows between countries, to obtain robust predictions in an even more interconnected and fast-evolving world.