Today’s world is more interconnected than ever and while this shift to a global focus has brought a wealth of opportunities and insights, it also poses new and interesting challenges. As such, the tools we use to develop and monitor our portfolios have become more comprehensive and varied. Since 2008, risk models have gained popularity and brought with them their own unique set of questions. Do we use a global or a country-specific model? How can we truly capture the global nature of some companies if we limit ourselves to a country-specific model? How do we account for a global event? Fortunately, we have a suggestion for the last question.
Most single-country risk models only involve market risk, sector/industry-type risk, and style risk, but using geographical revenue data such as FactSet’s Revere Geographic Revenue Exposure (GeoRev) data can capture some of the residual risks and integrate a global perspective into our single-country risk analysis.
Conventional geographic revenue data is difficult to interpret and compare between companies as it is not normalized. FactSet’s GeoRev data answers this challenge by first mapping companies’ revenues to a normalized geographic taxonomy and then applying a proprietary algorithm to estimate percent revenue exposure to countries and regions that are not explicitly disclosed. This enables us to truly understand how other countries impact a company’s bottom line.
For this analysis, we developed two custom risk models. These monthly models were constructed using a stepwise regression methodology with a 60-period variance calculation. The models leverage the FactSet-Northfield Global Equity Risk Model style factors to capture growth, value, momentum, leverage, change in liquidity, quality, volatility, and size. The industry factors were constructed using FactSet Revere Business Industry Classification System (RBICS). The only difference between the two models is an additional regression step for GeoRev data to further decompose residual risk.
With the two models created, we can see the impact (if any) that the addition of GeoRev has on our custom model. Using the Russell 1000 as a proxy for the U.S. market, the two models point to a similar risk profile throughout the end of 2019 and the beginning of 2020 leading up to the spread of COVID-19. Focusing on factor risk, we see that as the global pandemic ramps up, so does the explanatory power of the GeoRev data in our custom risk model.
Starting February 2020, there is a general shift towards more factor risk and away from asset-specific risk in our custom models. The difference comes from the magnitude of this shift with our GeoRev model having an even larger factor risk value. This increase can be directly attributed to the repurposing of residual risk. Now that we know our GeoRev model is providing some additional insight, let’s look at the GeoRev data in more detail. The steep increase in factor risk can set off alarm bells for many and bring up additional questions. To answer these questions, it is important to investigate what individual factors are contributing to overall risk. Given the nature of our modified model, our GeoRev factors bring clarity in this process. Explaining over 5% of the Russell 1000’s total risk at the end of April, some of the top country names will not be surprising.
Unsurprisingly, China leads our GeoRev factors in contribution to total risk with just under 5%.
China’s large contribution to total factor risk is expected being ground zero for the COVID pandemic and their seeming omnipresence in all U.S. companies. What many may not realize or overlook is the impact other factors (specifically in the same region) may have. Japan and Hong Kong are the second and third contributors to total risk coming from our GeoRev factors. Simply recognizing this can help us better understand and protect our portfolio from the volatile nature of markets associated with global uncertainty.
Having analyzed how the new GeoRev factors impact the Russell 1000’s factor risk, another important aspect of factor risk itself should not be overlooked: factor volatility. One question that should be asked is how factor volatility behaves during these global events. Using the three factors from before (China, Japan, and Hong Kong), one can see a spike in factor volatility that coincides with the increase in factor risk.
While these trends are similar, one should note that the increase in volatility from the Japan factor is delayed one month, showing how Japan’s response to the virus provided some level of stability through February.
Without the addition of our GeoRev data, the impact of other countries and the global nature of companies in the U.S. would be completely lost. But what does someone do with this information now that they have it? Proactive risk management, GeoRev risk portfolio optimization, and new risk factor creation hypothesis are just a few ways to apply this insight. Including GeoRev data into our risk model is a pertinent way to increase its efficacy during global events, but it’s just the first step towards preparing and responding to them.