Through the Quant Lens: Gauging Direct and Indirect Exposures to the Outbreak of War

• Gauging exposure before crises boil over: Analyzing price reactions in the early days of the Russia-Ukraine crisis offers a real-time case study in determining the relevance of various exposure factors.
• As humans hear noise, machines hear signals: Indirect exposures are neither easy nor practical for human analysts to track, creating an opening for text-processing machines and quantitative scorecards to monitor such dynamics at scale, providing a starting point for their human counterparts.
• Preparing for the next crisis: The methods discussed provide tools for deeper analysis as new concerns arise—whether geopolitical tensions or viral outbreaks or the unknown unknowns that await us in the future.

The ongoing conflict in Ukraine offers lessons for navigating the effects of future crises on equity portfolios. Beyond the more obvious dynamics, we have observed that U.S. stocks have other forms of direct and indirect exposures to this conflict. These include revenues from large customers in the region, supply chain sensitivities, and suppliers located in proximity to the conflict, among others. While much of the war’s impact has been priced into today’s markets, measuring and understanding the exposure for U.S. stocks allows for a deeper analysis of continued downside risk (or potential upside from an eventual resolution). In the interest of gauging exposure before circumstances reach a boiling point, we focus on stocks’ price reactions during the early days of the crisis period, in the immediate aftermath of the invasion.

The fact is, while risk models are a crucial tool, they offer only limited visibility in the form of “country-at-risk” indicators, which mostly reflect where a company is based or listed and its industry exposures. To look beyond risk models, our team has been sourcing alternative data, including:

• Geographic revenue breakdown (i.e., Russia, Ukraine and Belarus)
• Supply chain relationships and indirect regional exposure for customers or suppliers
• Company news stories, looking for co-mentions of Russia or Ukraine
• Company earnings call transcripts, seeking mentions of Russia or Ukraine

While direct revenue exposure is easy to discern from company filings, indirect exposure can be harder to detect and thus slower to be priced in. For instance, one of the hardest-hit stocks in the crisis has been EPAM Systems, which has large fulfillment operations in Ukraine. Its revenue exposure is relatively modest at about 4%, but the emphasis on the current crisis in its recent earnings call, as well as company-related news, was telling.

Another example is PPG Industries, which has even less direct revenue exposure to Russia (about 2%), but whose customers on average derive a sizable amount of revenue (about 9%) from Russia, Ukraine, and Belarus. If the revenues of its chief customers are impacted negatively by the conflict, then demand for PPG’s products may fall as a result.

It is neither easy nor practical for human analysts to track all this. However, text-processing machines and quantitative scorecards can monitor such metrics at scale and provide a starting point for further scrutiny by their human counterparts.

In collaboration with Voya’s Enterprise Data Science team, we implemented a new feature that scans for mentions of specific terms in recent earnings calls—in this case, Russia and Ukraine (see sidebar). Such mentions typically indicated how respective businesses were exposed to the region and therefore the war. For example, Sylvamo Corp. (SLVM) had more than 10 mentions during its earnings call on Feb 11, 2022. As a paper producer, it has a strong Russian operation, with more than 15% of revenues derived from the country. Unsurprisingly, the stock meaningfully underperformed peers that had less exposure to the region.

In addition to NLP analysis of earnings call transcripts, we screened for stocks with:

• High volumes of news stories with co-mentions of Russia or Ukraine
• Strong direct revenue exposures or indirect regional exposures for customers or suppliers
• High historical return sensitivity (i.e., beta) to Russia’s equity market index Our analysis flagged 59 non-energy U.S. stocks for having meaningful Russia exposure. These flagged securities underperformed their respective industries significantly through the early crisis period (Fig. 1).

Finally, we evaluated whether these non-conventional metrics offered statistically relevant insight for predicting companies’ potential price reactions. Most metrics we tracked had some bearing on how companies fared early in the crisis, reflected in statistically significant correlations of factor exposure rankings and industry-relative performance (Fig. 2). The exception was the number of mentions in news articles, perhaps because larger companies tend to get more press. Nonetheless, by grouping all the metrics together, we were able to effectively capture risk exposures to the Russia-Ukraine crisis.

By using the methods discussed, we can evaluate these metrics and others when new concerns arise, whether they be geopolitical tensions, viral outbreaks, or other “unknown unknowns.” This may point portfolio managers to latent exposures in their holdings to be addressed, while identifying “red flags” within a coverage universe for fundamental analysts to conduct further research.

The combination of more informative non-financial data and powerful data science techniques means that the era of solely relying on off-the-shelf models for risk management is nearing an end.