In many cases, behavioral changes will remain permanent, or at the very least, lasting to the point at which credit deterioration is experienced to a widely varying degree, across credit segments.
A retrospective of transpired events, beginning with the outbreak of COVID-19 in February and March 2020, highlights that established, well-developed credit models used to evaluate the environment did not work adequately. Internal ratings — an institution’s cornerstone for long-term investment and lending strategies — rely on fundamental, name-level analysis, which cannot be updated consistently or at frequencies required to react to and plan for quickly changing developments. Meanwhile, forward-looking measures used in regulatory stress testing or with CECL/IFRS 9 impairment calculations typically depend on scenarios defined by broad-brushed variables such as unemployment and GDP. These scenarios are not sufficiently differentiated across industries; their performances vary in sensitivity to the sociological reaction to COVID-19 and show potential disparity in their response.
By their very nature, emerging risks and crises reveal behavior incongruent to historic patterns, requiring new and alternative data and analysis. In highly volatile environments, institutions need quantitative and repeatable benchmarks to facilitate current internal rating assessments, as well as projected ratings and loss measures that reflect the environment and anchor to financial institutions’ internal risk measures, allowing them to be relatable and usable.
This paper presents The Cross-Sectional COVID Overlay, which provides a current credit assessment as well as projected ratings and loss measures, anchoring to well-understood starting points and an organization’s traditional forward scenarios, described through GDP and unemployment projections, for example, anchoring to internal credit loss models. Using granular, name-level credit data, along with data proxying for the sociological reaction to COVID, the Cross-Sectional Overlay measures the varying impact COVID has had across a set of 121 industries and dozens of countries and their varying recovery speeds. In addition, the Cross-Sectional Overlay measures the direct and indirect effects of COVID–related stimulus programs targeting individuals, small businesses, corporations, and the airline industry bailout. The Overlay has natural applications for benchmarking internal ratings, Comprehensive Capital Analysis and Review (CCAR)/European Banking Authority (EBA)/European Central Bank (ECB) stress testing, and CECL/IFRS 9 impairment calculations. The Overlay is very useful as a complement to credit portfolio management and capital planning processes.
The sociological reaction to COVID has had material and varying impacts on economic activity worldwide. Predictions of radical change flooded the media at its outset, and the degree, length, and severity of change and recovery in various geographies continue to be a widely debated topic. One certainty is true; emerging risks and crises, by their nature, reveal behavior incongruent to historic patterns. In the context of credit, crises reveal hidden concentration risks and require new data and analytics. The use of established, well-developed models to evaluate rapid changes in the current and the post-COVID-19 environment does not provide adequate risk measures. Internal ratings — a cornerstone to an institution’s long-term investment strategy — rely on fundamental, name-level analysis, and they cannot be updated consistently and at frequencies that allow financial institutions to react and plan when the environment is changing quickly as during the heights of the crisis. Additionally, these models lack a level playing field in assessing cross-segment risks on an ongoing basis. Meanwhile, forward-looking quantitative measures used in regulatory stress testing or with CECL/IFRS 9 impairment typically rely on scenarios defined by broad-brushed variables, such as unemployment, and cannot sufficiently differentiate across industries (for example, Medical Devices, Hotels, or Transportation); their performance varies in sensitivity to COVID-19 itself, and in their response to the direct and indirect protective measures put in place. Models that calibrate the sensitivity of credit losses using the last 20 years of pre-COVID data simply will not pick up on the varying degrees to which different industries are affected by COVID-19.
1. Exploring Alternative Data to Describe COVID’s Varying Impact Across Credit Segments
This section explores the impact of COVID-19 on corporate credit and serves as a backdrop to the approach we take in designing the Overlay detailed in Section 3. The sociological reaction to COVID’s epidemiological progression has varied broadly. Governments have attempted to control the virus’ spread by limiting mobility and bolstering economic activity through financial stimulus. Consumers and businesses have shifted behavior, often drastically, causing demand for services such as air travel and hotel stays to plummet and raising demand for home office furnishings and consumer-driven online shopping. There are no shortages of supply shocks, with illness and lockdowns limiting production. While these dynamics impact credit risk, quantifying their materiality remains very challenging when considering traditional models are calibrated to, say, data over the last 20 years.
By and large, traditional loss forecasting models used for impairment (e.g., CECL or IFRS 9) or stress testing (e.g., CCAR or ECB/EBA) cannot describe COVID’s unprecedented and diverse impact across credit segments. Rather, the crisis requires alternative data and analytics, with proxies that capture sociological reactions, including epidemiological, mobility, and tourism data, as well as detailed information regarding fiscal and monetary programs.
Re-thinking granularity of segmentation
Credit risk models traditionally involve classifying borrowers into several broad sectors. It is not unusual to have fewer than 50 industry categories, for instance. COVID uncovered a number of hidden risk factors, highlighting the need for more granularity, recognizing, for example, its disproportionate impact on business models that rely on the physical proximity of staff or customers, absent traditional methods used for segmentation.
Our methodology begins with Moody’s Analytics one-year name-level EDF™ (Expected Default Frequency) credit measure.1 This measure uses equity market and financial statement information to produce a name-level, forward-looking assessment of default risk for well over 45,000 firms globally, allowing us to explore varying levels of segmentation. Figure 1 provides examples where traditional industry segmentation proved too coarse. The left-hand chart shows the mean EDF value for two segments that were previously included in the Aerospace & Defense segment that contain the likes of Boeing and Lockheed Martin, impacted by the pandemic in dramatically different ways. The fall in air travel has had knock-on effects for Boeing and other Aircraft Manufacturers, leading to reduced orders and expected future revenues, resulting in credit deterioration. On the other hand, Lockheed Martin and other Defense & Space companies rely heavily on government orders that were not cut in meaningful ways during the pandemic, and their credit risk remained largely unaffected. It is worth noting that we did not see such a stark divergence in the performance of these two granular segments during previous crises — Lockheed Martin’s EDF value did materially increase during the Global Financial Crisis, which was accompanied by the efforts of many governments to take control of their fiscal situations, including defense spending. Hence, when modeling the impact of the pandemic in a granular fashion, it would be inappropriate to keep all the aircraft manufacturers and defense contractors in a single Aerospace & Defense category.
Similarly, the right-hand chart in Figure 1 illustrates how more-granular segments, within say Hotels & Restaurants, performed differently during the pandemic. Traveler Accommodations and Dine-In Restaurant’s were impacted more adversely thanks to the inherent physical proximity in their operation, while Fast-Food Restaurants and Coffee Shops (e.g. Starbucks) saw their credit risk increase to a lesser degree, thanks to their readiness to engage with their client base in a socially distant manner.
Investigating these patterns leads to classifying public borrowers into 121 industry segments, far more than the traditional models consider. Similar analyses are required for modeling other emerging risks — separating granular segments more susceptible to supply-chain breakdowns, rolling black-outs, or other disruptions.
Varying performance of industry segments: transient and permanent shocks
To get a sense of COVID-19’s effect on credit risk across industries, and the general level of uncertainty it has generated, Figure 2 highlights the heightened Mean EDF value by industry during February and March as the pandemic began to unfold.2 While EDF values increased across the board, the likes of Automotive, Dine-In Restaurants, and Cruises exhibited the strongest increases, contrasting with Pharmaceuticals. Interestingly, as the implications of the pandemic began to solidify, the Automotive and Consumer Cyclical Products segments recovered to some extent, reaching what appears to be an asymptote in June, while Dine-In Restaurants remained at elevated levels, and the Cruises segment’s credit risk continued to rise into July.
Is it possible to explain the varying reactions of segments to the pandemic by shifts in economic fundamentals underpinning these segments? Figure 3 provides an example for Airlines and Automotive segments. We can see the median EDF values increasing similarly across the two industries with the initial March shock, but stabilizing at very different levels through September. Travel-related industries, including Airlines, continued to suffer financially beyond the initial economy-wide shock, due to continued social distancing policies and practices, as well as travel restrictions. Despite some recovery in the number of TSA-reported airline passengers during the past few months, counts are well below 2019 levels. Segments that depend on physical proximity of consumers, such as Airlines, Leisure & Recreation (Theme Parks), Cruises, Dine-In Restaurants, and others, have endured a more persistent shock, as people make longer-term lifestyle changes to adjust to the pandemic. Behavioral changes of populations, consumer sentiment and business conditions, the overall macroeconomic situation, and government aid have all shifted rather quickly during the pandemic’s various waves within different countries, creating a complicated, multidimensional modeling environment.
Since June, the Automotive segment has diverged in performance when compared with travel-related industries, evident from the trending volume of vehicle sales throughout the pandemic. Sales dropped substantially relative to 2019 in the pandemic’s build-up in March and April, which reflected two effects. First, on the supply side, lockdowns in many countries and U.S. states that mandated closures of plants, along with trade changes, led to supply-chain disruptions. As a result, production declined. Second, consumer demand plummeted in March due to economic uncertainty and looming unemployment. The state of the Automotive industry improved markedly during the summer, when plants reopened and consumer demand picked up. The increase in consumer spending reflects improved consumer sentiment; the extent of the damage became clearer and populations adjusted their lifestyles to its longer-term effects. Moreover, consumers were helped by various government measures. The shock to the Automotive segment in March turned out to be transient — it was linked to the temporary drop in consumer demand, and the adverse shock vanished as demand recovered. This narrative applies not only to the Automotive segment but, for instance, to Furniture Manufacturing or Consumer Cyclical Products. These segments are not impacted by social distancing policies, and, in some cases, benefited from shifts in consumer spending.
Any analytics developed to capture cross-sectional dynamics in credit risk over the course of the pandemic must balance the mixture of permanent and transient shocks responsible for the divergence of responses across segments. It is worth noting that this approach to analyzing credit risk deviates from the models deployed traditionally prior to 2020. Those models often captured the sensitivity of credit segments to the macroeconomic environment by regressing historical credit losses on macroeconomic variables over the course of several recessions. Traditional models on their own are limited in their ability to capture the pandemic’s varying effects for two primary reasons: the pandemic event window is limited to a relatively short period, rendering regressions using quarterly macroeconomic to only a few data points and of limited use; and further, the effects of the pandemic cannot be distilled to unidirectional sensitivities to the macroeconomic environment. As we have seen, the specific characteristics of a segment in relation to the pandemic matter — those segments that require physical proximity were impacted by the initial pandemic shock most, and they have not recovered, despite the improving macroeconomic environment, because populations continue to socially distance and restrict movement. Non-essential consumer businesses experienced a shock at the beginning of the pandemic as well, but they recovered with the increase of consumer confidence. Incorporating such characteristics requires alternative data, such as actual population mobility, given that the data used in the traditional models —credit and macroeconomic time series — cannot alone describe these complex dynamics.
We can also extend this style of analysis to other emerging risks, beyond COVID-19. For instance, the impact of natural disasters, such as volcanic eruptions, is not limited to airlines, but all travel-related segments — Dine-In Restaurants in travel destinations, Leisure & Recreation segments, or Cruises. Climate events, such as typhoons in Southeast Asia, disrupt supply chains and affect those industries around the world that rely heavily upon imports of components or products from the affected area. A large-scale cyberattack would then have a disproportionate impact on segments that cannot operate without safe online connectivity — virtually most of the world. Capturing the cross-sectional effects of such emerging risks requires employing alternative data — from understanding tourist flows to mapping supply chains or measuring dependence on internet connectivity. While financial institutions often think about these risks, they are recognizably difficult to quantify, and they fall outside of the traditional frameworks of most models. We discuss this point further as we describe the Overlay and the underlying analytics that allow us to capture the pandemic’s more nuanced effects.
Epidemiological data is not enough: describing credit risk dynamics with alternative data
This section explores various data that ultimately proxy for the sociological reaction to COVID, and the resulting impact across credit segments. Our objective is to identify industry credit segments, whose performance can be linked to the state of the pandemic and the macroeconomic environment. The natural candidates are epidemiological data, such as the number of new COVID-19 cases per day, ICU utilization capacity, or fatality rate. We investigate the potential for using epidemiological data before delving into further exploration using alternative data to help describe cross-industry country dynamics.
Figure 4 presents the seven-day moving average of new infections for Spain, the UK, Germany, South Korea, and the U.S. Countries such as Spain and the U.S. experienced a strikingly high caseload in March. The first wave of the pandemic was followed by a major respite from the virus in Europe, before the continent entered a second wave, with new infections rapidly increasing again. Throughout, however, the U.S. has not managed to substantially reduce the infection rate since the first wave began, as the virus continued to gradually spread to most corners of the country. These trajectories contrast starkly with that of South Korea, which took almost immediate control over the virus’ spread using rigorous contract tracing protocols and stay-at-home measures.
The difference in COVID management between South Korea and Western countries is reflected in their credit risk patterns — the bottom chart of Figure 4 shows how the Median EDF value relative to its December 2019 level radically increased across all countries in March. The recovery in South Korea to pre-COVID levels once the virus was controlled is perhaps not surprising. However, the median EDF levels for Western countries move almost in sync, with no obvious relation to the state of the epidemic in individual countries. The continued high numbers of new cases in the U.S. did not prevent its EDF level to decline, and the EDF levels for European countries has not reacted — at least not yet — to their second wave.
The difficulty of using the number of new COVID-19 cases to measure the state of the pandemic extends beyond the lack of a stable statistical relationship with credit risk time series. Reported new cases are highly dependent on the availability of testing — as the countries ramped up testing capacities, comparisons of the numbers of cases from March and September lose traction. Second, differences in testing capacities across countries, as well as differences in reporting of COVID-19-related statistics (including fatality rates and hospitalization rates), make epidemiological data even more challenging to process and to interpret.
We explore various alternative datasets, including the Mobility Indexes created by Google.3 These Indexes measure where populations, at the country- or local-level, spend their time, based on Goggle application user location data. The patterns are compared with a pre-pandemic baseline of January 2020. Thus, the Indexes provide a proxy for the reduction in time spent in the workplace or retail establishments compared to pre-pandemic times. Despite the challenges associated with interpreting the Indexes, including varying definitions of workplace or usage of Google applications across countries, we see the Indexes provide a useful indication of how populations and businesses reacted to the pandemic and government measures, such as lockdowns.
Figure 4 compares the dynamics of EDF values with Mobility Indexes for Workplaces and Retail & Recreation locations in the United States. It shows that what seems to matter for credit risk dynamics is not the number of cases itself, but how populations and businesses react to both the pandemic as well as to government measures. Perhaps not surprisingly, the rapid increase in daily cases in March was associated with a decrease in mobility, driven both by health concerns in the population as well as government-mandated lockdowns. The stall in travel and many other daily activities, compounded by fear of the upcoming recession, sharply reduced consumer spending across many segments and sent EDF values, as measures of credit risk, soaring. While the number of new cases, after a brief decline, continued to increase in June, EDF values did not resume their rise. In fact, they remained stable or declined for some segments. This pattern remains consistent with the dynamics in Mobility — after the steep drop in March, Mobility for Workplaces and Retail & Recreation locations began to slowly increase until June, despite the rise in new COVID-19 cases. Even though U.S. sheltering guidelines varied by local area, Mobility seemed to reach an asymptote of sorts in June, which roughly corresponds to the leveling of EDF values across segments starting around that time. The data support a market consensus of a long-term material deterioration in Leisure & Recreation and with the shift to remote work associated with a recovery-albeit partial-in the Automotive and Consumer Cyclical segments.
Exploring the use of alternative data to supplement those used in traditional credit risk models is becoming a necessary step, not limited to incorporating the impact of COVID, but also in the analyses of how other emerging risks affect credit portfolios. For instance, accounting for climate-related hazards such as drought, hurricane, or flood risk requires measuring the likelihood of events, along with detailed trade flow and supply-chain data, to assess the magnitude and structure of disruptions that a climate event can bring. Given the unique nature of the emerging risks, creating a new model or overlay to reflect them in a portfolio analysis requires innovative analytics that collate a range of alternative datasets.
Toward a global model: how to capture cross-country differences in credit risk during the pandemic
Thus far, we have focused the discussion on the varying impact that COVID has had across industries. In Figure 7, we explore cross-country dynamics. The top chart in Figure 7 quantifies the similarity of industry performance ranking across countries. Specifically, for each country, we create a list of industry segments ordered according to their EDF value changes over the course of the pandemic, and we calculate the rank correlation between this list and the global list of industry segments ordered in the same way. The rank correlations are often upward of 70% because the industry segments adversely affected in one country tend to be adversely affected in other countries too. Perhaps this is not surprising, as Airlines, Dine-In Restaurants, Traveler Accommodations, and Leisure & Recreation segments have deteriorated globally, while Food Stores have performed well.
While the ranking of industry segments according to credit deterioration appears to be similar across countries, performance dispersion is not. The bottom chart in Figure 7 shows that the magnitude of credit risk dispersion across industry segments tends to be related to the countries’ sociological reaction to COVID. On the vertical axis, we plot the dispersion in EDF value changes for a given country between February and the specified month. On the horizontal axis, we plot the drop in the Workplace Mobility Index from February to the given month. Countries with the biggest drop in Mobility, including the U.S., the U.K., and Canada, exhibit the highest variance — that is, the adversely affected industry segments in these countries are affected much more than in other countries. On the other end of the spectrum, the Mobility Index in Taiwan, one of the countries that kept the spread of the virus within its borders under control, has returned (or exceeded) pre-pandemic levels. The cross-segment dispersion in Taiwan is fairly low compared to the affected countries. Other countries, such as Germany, fall in between those two extremes.
The empirical analyses presented in this section so far provide us with a set of stylized facts about the effects of the pandemic. First, the granularity of segmentation needs to be reconsidered in light of the pandemic. Second, the credit quality deterioration of industry segments has not replicated the patterns from previous recessions, but rather followed patterns reflective of the nature of the pandemic. Third, the relative credit risk of industry segments has changed throughout the pandemic as some segments’ performance improved, while others remained affected by shocks that appeared to be permanent. To describe the course of the pandemic and measure its effect on credit risk, we need to look for alternative data that represents the reaction of the population to the events and government measures. Finally, rank ordering of industry segments according to the degree of their credit quality deterioration appears similar across countries. But the magnitude of the dispersion of that deterioration tends to be linked to the state of the pandemic in the country. These stylized facts are incorporated into the Cross-Sectional COVID-19 Overlay — the analytics designed to measure the impact of pandemic scenarios on credit risk across countries and a granular set of industries. Section 2 describes the Overlay.
Fiscal and Monetary Response to the Pandemic
We now transition to exploring the uncertainty surrounding fiscal and monetary response to these pronounced events. The past 50 years have seen many significant crises that severely affected different portfolio segments. We have also observed bolstering of targeted segments, including bailouts. Governments have a wide range of mechanisms to help support affected segments during crises.
The way in which COVID-19 is playing out is unique, not only in the effects on different industries, but in the remarkable fiscal and monetary response designed to bolster various sectors. To get a sense of just how remarkable the response has been, Figure 8, shows the level and timing of the funds authorized by the U.S. Congress since the beginning of the coronavirus crisis in red, compared to that during the beginning of the Great Recession in blue.
While we recognize the vast differences between the two crises, it is worth recognizing that Congress authorized more funding during the first few weeks than the amount authorized during a year-and-a-half into the 2008−2009 Global Financial Crisis. It is worth observing that, unlike the previous crises, authorities are not limited by moral hazard concerns as they were during the financial crisis, and they are less apprehensive about supporting segments facing difficulties — though questions remain surrounding the effectiveness of the distributions.
As reported by FRED Economic Data (https://fred.stlouisfed.org/series/USEPUINDXD)4
2. The Cross-Industry COVID-19 and Fiscal & Monetary Overlays
This section describes the Cross-Sectional COVID-19 and Fiscal & Monetary Overlays that serve as the foundation for Current internal rating assessment and projected ratings and loss measures.
Cross-Sectional COVID-19 Overlay Model
The Cross-Sectional COVID-19 Overlay Model captures COVID-19’s varying impact across segments of a credit portfolio, which traditional approaches based on broad-brush macroeconomic scenarios cannot achieve. Figure 10 illustrates this distinction. The top chart shows realized GDP growth for various countries 2020Q1−2020Q3 and projected GDP growth under a 96th percentile downturn scenario beyond that time horizon. GDP Equity Market Indexes, Unemployment Rates, and others are macroeconomic variables used in traditional scenario-based approaches to projecting credit risk, including in CECL and IFRS 9 models or CCAR submission. While GDP provides an overall gauge of economic conditions, any analysis of a credit portfolio must account for the fact that, during the pandemic, these economic conditions impact individual segments of the portfolio in new, unprecedented ways. To that end, the Overlay uses the macroeconomic scenarios as anchors that imply a shock to the overall credit risk, but takes a stance on the industry-level sensitivity to the COVID-19 pandemic, for any given country. The bottom of Figure 10 provides a visual of the Overlay’s outcome. The Overlay translates the macroeconomic scenario into projections for credit risk factors at the level of 121 industry segments for each country.
The Overlay’s analytical framework is designed to reflect the empirical patterns observed in Section 1 —required to capture the impact of the Pandemic. The Overlay recognizes varying sensitivity of industry segments to COVID, but also incorporates the distinction between permanent and transient shocks, leverages the alternative data that describes the state of the pandemic (Mobility Indexes) or consumer sentiment, and reflects cross-country differences in credit risk based on their populations’ reaction to the epidemiological situation, as well as government action. The framework marries statistical analysis with a fundamental review of industry segments’ characteristics. For instance, while the relative performance of some segments, such as Consumer Cyclical Products, depends on the state of consumer sentiment, that of other segments, such as Oil Exploration, Oil Machinery & Equipment, and Pipeline Transportation, are better characterized by Oil Price shocks.
The Overlay was calibrated using a time series of segment-level EDF levels, together with macroeconomic data, and the alternative datasets, over the event window of the pandemic (2020Q1−2020Q3). As Figure 10 indicates, the calibration produced rank-ordering of industry-segment sensitivities in-line with economic intuition — segments such as Dine-In Restaurants react most adversely to the pandemic-induced downturn, while Pharmaceuticals experience only a mild shock. Formally, the calibration of the Overlay follows from relating the quantitative performance of each segment to two components: one representing a persistent component tied to the mobility index, and one representing a transitory component tied to a measure of consumer confidence.
How do we use the Overlay in practice? The segment-level projections are transmitted to name-level credit metrics, such as PDs or ratings, for any borrower in a corporate portfolio. Name-level variation within an industry-country segment is obtained from the variation in the initial PDs and ratings, as well as individual borrowers’ sensitivity to systematic risk, which is name-specific. With projected PDs and ratings in place, the framework facilitates projection of spreads, Risk-Based Capital, or in analyzing the impact on portfolio concentration risk.
Figure 11 demonstrates how materially-different losses calculated using the Overlay (blue) can be compared to loss projection under traditional scenario-based models (orange). While both the blue and orange bars highlight the material deterioration in credit, with expected losses increasing by more than twofold for many industries, the Cross-Industry COVID-19 Overlay recognizes that the most affected industry segments differ from those seen in the past recessions — segments relying upon the physical proximity of clients or employees, such as Cruises, Conventions, or Dine-in Restaurants, will likely see an eightfold increase in expected losses, well exceeding the magnitude of shock they experienced in past recessions.
Fiscal & Monetary Overlay Model
In addition to the direct impact of COVID-19 across industries, Figure 12 highlights the remarkable fiscal and monetary response bolstering segments that must be understood and quantified. As discussed, while markets incorporate expectations of the response and its effectiveness into prices, Figure 9 highlights the material uncertainty around future responses and their effectiveness. The Fiscal & Monetary Overlay Model quantifies the impact of existing programs on credit, providing a point of reference for modeling the impact of unanticipated program effectiveness, as well as the impact of future programs on credit.
To begin, a review of the various programs and ongoing revelation of their details, in conjunction with an assessment of COVID-19’s impact, provides an estimated support level for different corporate segments, whether for small businesses or programs designed for larger corporates. Beyond airlines (receiving a targeted bailout), the Fiscal & Monetary Overlay Model highlights the most-affected segments as those receiving the most quantifiable support in various forms, including deferred interest loans and forgiveness for those meeting various criteria. In Figure 6, the upper-left table represents industries most affected by COVID-19. In addition, a review of individual spending patterns represented at the top-right provides a sense of how individual rebates flow into corporate revenues, estimated to be more than 70%, which are then combined with margin estimates. The total estimated funds relative to industry size are shown at the bottom of the figure, and they can be material, with many segments receiving more than 5%. Quantifying the programs rolled out to date and their cross-industry impacts on mitigating credit risk is the first step in our analysis.
Next, we quantify the impact and effectiveness of various government programs — recognizing there is an assortment of timing and mechanisms by which support is given to targeted segments. Whether referencing various rounds of the CARES Act or the Main Street Lending Program, we must understand that we face uncertainty in the range of fiscal and monetary programs, which will have varying and uncertain timelines. This uncertainty should not be surprising, and it is very much in-line with historical experiences.
A notable example includes the $15 billion airline bailout in September 2001, which took several weeks to fully understand. Figure 13 highlights the behavior of the EDF credit measure, which increased drastically on September 11, 2001. Meanwhile, the bailout announcement on September 23, 2001 took a few weeks to completely affect default probabilities. We see similar patterns with the recent airline bailout reported on April 14, 2020.
On a similar note, a study by the Congressional Budget Office estimates that as much as 50% of the 2009 American Recovery and Reinvestment Act (ARRA) was deployed after 2010, almost two years out.5 During 2008−2009, Congress authorized multiple rounds of funding. We now face wide-ranging fiscal and monetary scenarios, with varying timeline uncertainties and effectiveness levels. We use these observations and studies to construct narrative scenarios related to fiscal actions.
This returns us to a previously observed need addressed by the Fiscal & Monetary Overlay Model — an approach to quantifying the impact of future unexpected actions and the uncertainty surrounding the effectiveness of those actions on credit.
3. Current Internal Rating Assessment and Projected Ratings and Loss Measures
This section describes two applications of the Cross-Industry COVID-19 and Fiscal & Monetary Overlay Models: Current Internal Rating Assessment and Projected Ratings and Loss Measures. We present case studies to provide a sense of how an organization would use these models and a sense of their materiality.
Current Internal Rating Assessment
The Current Internal Rating Assessment anchors to a reasonable, well-understood starting point, December 31, 2019 for example, and uses Moody’s Analytics Cross-Industry COVID-19 Overlay Model to project what has happened up until present. The assessment gives an estimated current credit rating, accounting for granular, name-level, and cross-sectional impacts of the coronavirus across a set of 121 industries and across countries. Figure 14 shows the information needed and how the assessment is conducted. To begin, an organization specifies the most recent, reasonable, and well-understood credit assessment of their portfolio. Then, using the Cross-Industry COVID-19 Overlay Model, we add an assessment of what has happened so far.
We now shift attention to our Projected Ratings and Loss Measures. Moody’s Analytics Cross-Industry COVID-19 and Fiscal & Monetary Overlay Models are anchored to an organization’s current internal rating and use traditional forward economic scenarios to project ratings and loss, as Figure 16 depicts. The projections account for the cross-sectional impacts of COVID-19 across a set of 121 industry segments and across countries, quantifying the direct and indirect effects of COVID-19–related stimulus programs targeted to individuals, small businesses, corporations, and industry bailouts. These projections have natural applications for CCAR/EBA/ECB stress testing and CECL/IFRS 9 impairment calculations. They provide useful complements to credit portfolio management and capital planning processes.
To summarize, Moody’s Analytics Cross-Industry COVID-19 and Fiscal & Monetary Overlay Models anchors to an organization’s current-state internal ratings assessment (possibly from Moody’s Investors Service, or Moody’s Analytics EDF credit measures), along with traditional forward scenarios (for example, GDP and unemployment) to produce granular, name-level projections with natural applications for CCAR/EBA/ECB stress testing and CECL/IFRS 9. Both can complement credit portfolio management and capital planning tools. This paper focuses on two applications of the models. We discuss additional applications in forthcoming papers.
4. Beyond COVID-19
The current situation we are trying to navigate presents an unlimited number of unknown scenarios and outcomes. COVID-19 has challenged many traditional risk assessment methodologies, due to the rapid changes and impacts resulting from the virus. We must re-evaluate common risk factors that we have used for decades. Inevitably, we will continue to develop better insights using tools such as these new overlays. However, despite our attempts to model and project, we may have shifted into a new paradigm where many structural changes and uncertainties radically alter our embedded systems. Adjusting these systems requires unique datasets and analytics that update frequently, perhaps faster than anything we have ever tried.
These datasets must evaluate the current state of credit and a range of economic paths, including fiscal stimulus actions. Because of this unique experience, we can state that all recessions are created equal compared to the current situation. The market is evolving daily. Industry and borrower credit profiles are rapidly deteriorating, with many unknowns and a lack of clarity around when things will return to normal, and which businesses, industries, and so on will survive. Past events do not apply here; we must look forward to accurately gauge the impact of this pandemic.
With this in mind, we should not narrowly focus on the coronavirus and how it is currently affecting credit. Instead, we should recognize the pandemic within the broader context of risks that are becoming increasingly understood to be more relevant and that transcend “basic” risks. Examples include climate and geopolitical risks, which have a common geospatial element, but also risks such as cyber terrorism and grid susceptibility. Emerging concentration risks are shifting our views of geospatial dynamics. As with the Overlay presented in this paper, analyses of other emerging risks require a deeper understanding of the nature of the shocks to a credit portfolio — which of the shocks are due to permanent shifts in business models or consumer behavior (as a reaction to climate change, for example) and which of them reflect temporary disruptions, such as supply-chain breakdown (as a result of a natural disaster or political instability)? As we evaluate frameworks that can help us navigate today’s increasingly complex environment, we have a critical opportunity to think beyond COVID-19 and to plan for risks that will inevitably be present in our future.
1 Pooya Nazeran and Douglas Dwyer, “Credit Risk Modeling of Public Firms: EDF9,” Moody’s Analytics Model Methodology, June 2015.
2 Pooya Nazeran and Douglas Dwyer, “Credit Risk Modeling of Public Firms: EDF9,” Moody’s Analytics Model Methodology, June 2015.
3 For more details on how mobility indexes are created and their interpretation, see https://www.google.com/covid19/mobility/
4 For additional details, see Baker, S., Bloom, N., and Davis, S., "Measuring Economic Policy Uncertainty," October 2015.
5 Congressional Budget Office, “Estimated Impact of the American Recovery and Reinvestment Act on Employment and Economic Output in 2014,” February 2015.