Featured Product

    Qualitative Overlay Factors for CRE Credit Risk Models in the Context of COVID-19

    May 2020

    Qualitative Overlay Factors for CRE Credit Risk Models in the Context of COVID-19

    While we have experienced many economic recessions in the past, the crisis we face today is different, particularly in two key aspects.
    • COVID-19 is first and foremost a public health crisis. To contain the spread of the virus, mandatory social distancing and restrictions on social gathering have become the “new normal,” which requires us to re-evaluate many types of commercial real estate (CRE) spaces within the context of the pandemic.
    • This is the first time countries around the globe have chosen to shut down a substantial part of their economic activities, leading to a “manufactured,” rapid increase in unemployment and decrease in economic production. In response, governments and central banks instituted unprecedented massive fiscal and monetary programs to protect their economies from total collapse.

    Given that we develop our models predominantly using historical observations, and that these new phenomena do not feasibly work well with empirical-based models, we share our thoughts on new factors as possible considerations for qualitative overlays on top of our existing credit risk models for CRE loans. 

    Specifically, in response to the unique nature of this crisis, we also group qualitative factors along the two critical dimensions. On the public health front, we introduce our location-specific Coronavirus Exposure scores and property type-specific CRE risk scores as related to pandemic concerns. On the policy response front, we dissect relevant measures from the CARES Act and the Federal Reserve’s programs and discuss their potential beneficial impact on different commercial real estate sectors. As some of the risk scores are dynamic due to the uneven evolution of the pandemic, we plan to continue updating scores and making them available to clients. 

    Given the rapid escalation of the COVID-19 pandemic, we have updated the Coronavirus Exposure scores at all geographic levels based on the latest confirmed cases on May 29, 2020. In addition, we have also recalculated the other scores at the metro and submarket levels according to Moody’s Analytics REIS market definitions. These scores can serve as timely complements to the quantitative outputs from CRE credit risk models such as Moody’s Analytics CMM™ and CRE Loss Rate Models, which incorporate the upcoming Moody’s Analytics REIS market forecasts. There is no change to the underlying census data given the relatively stable demographic and business compositions of CRE markets over time.

    1. Introduction
    The COVID-19 pandemic continues to ravage countries around the world. As of this writing, early May 2020, the United States has surpassed 1.4 million confirmed cases and the number of fatalities has topped 82,000. With almost all 50 states imposing some form of shelter-in-place orders, the coronavirus has also taken a devastating toll on the US economy. According to the latest estimate from the Bureau of Economic Analysis, Gross Domestic Product has already contracted by an annual rate of 4.8% during the first quarter of 2020.1 The labor market has been hit particularly hard, and very quickly, with more than 33 million Americans filing for unemployment benefits during a seven-week span. Similar economic consequences have hit many countries worldwide. 

    As an integral part of economic and social activities, the commercial real estate (CRE) sector has also fallen victim to the  COVID-19 crisis, and no property type has been spared. Hotels have undoubtedly taken the biggest hit of all property types, as both business and leisure travel has almost entirely diminished. In contrast, other property types generate income through long-term rentals, and these are affected to varying degrees by the inability of tenants to pay rent. According to data from the National Multifamily Housing Council (NMHC), nearly a third of US renters missed their rental payment at the beginning of April.2 However, the NMHC recently found that “91.5 percent of professionally managed apartment households made a full or partial rent payment by April 26 in its survey of 11.5 million units of professionally managed apartment units across the country”3 possibly due to the support of the federal relief funds and flexible options provided by the property operators. Meanwhile, under shelter-in-place orders, all non-essential businesses are required to shut down, including restaurants, clothing stores, and other retailers. Their revenues have plummeted and they face tremendous difficulty paying rent for retail space. The decline in the consumption of non-essential goods has also led to decreasing demand for industrial space. Some manufacturing plants have even been forced to close because of the coronavirus outbreak. The office sector has fared relatively better since office tenants tend to sign longer leases, and many office employees have been able to work from home with relative ease for the time being. However, the prevalence of remote working has placed the future of office space demand in doubt. 

    Given the rapidly evolving situation, CRE lenders must frequently reassess the credit risk of their portfolios based on the latest economic conditions and outlook. However, due to the low frequency of transactions on both leasing and sales fronts, CRE market statistics are typically lagging indicators. In particular, while we have experienced many economic recessions in the past, the crisis we face today is different, particularly in two key aspects: 

    COVID-19 is first and foremost a public health crisis. To contain the spread of the virus, mandatory social distancing and restrictions on social gathering have become the “new normal,” which requires us to re-evaluate many types of CRE spaces within the context of the pandemic. This is the first time countries around the globe have voluntarily chosen to shut down a substantial part of their economic activities, leading to a “manufactured,” rapid increase in unemployment and decrease in economic production. In response, governments and central banks instituted unprecedented massive fiscal and monetary programs to protect their economies from total collapse.

    Given that we develop our models predominantly using historical observations, and that these new phenomena do not feasibly work well with empirical-based models, we share our thoughts on new factors as possible considerations for qualitative overlays on top of our existing credit risk models for CRE loans. 

    Furthermore, it is important to note that the impact of COVID-19 is likely uneven across different regions for various reasons, which makes location consideration even more important at this time. These reasons may be related to the demographic composition or business characteristics of each region, and they are sometimes difficult to quantify. In this paper, we describe location and property type-specific COVID-19 risk scores that can be used as qualitative overlay factors on top of the quantitative outputs produced by Moody’s Analytics CMM™ or other CRE credit risk models. 

    2. Q-Factors Related to Public Health
    To measure the impact of this public health emergency on commercial real estate, we construct COVID-19 CRE risk scores based on the magnitude of the coronavirus outbreak at a granular, regional level. Due to the fundamental differences between multifamily and non-residential CRE market segments, we develop separate scores based on risk factors specific to each property type. 

    In real estate, each metropolitan area is considered a distinct market by industry practitioners. These metropolitan areas mostly coincide with the census definition of metropolitan statistical areas (MSAs). For users’ convenience, and using the availability of various datasets, we construct risk scores for each state and metropolitan area. Because of their size and diversity, the largest MSAs are often divided into several submarkets with their own unique characteristics. Therefore, we also construct submarket-level risk scores wherever applicable. 

    Before elaborating on the various categories of risk scores, we first outline the method used to calculate these metrics. Given that we examined a variety of disparate measures, developing a method of standardizing each quantity was pivotal. To achieve this goal, we used z-scores. A z-score represents the distance of any individual observation from its mean. A score of 1, for example signifies a value that is a single standard deviation above the mean. In other words, the formula for a z-score is as follows: 

    qualitative-overlay-factors_equation 1

    where:
    𝑥 = state, metro area, or submarket being examined 
    𝜇 = arithmetic mean across geographies 
    𝜎 = standard deviation across geographies 

    2.1 Coronavirus Exposure Measures 
    The quintessential CRE risk factor for COVID-19 exposure is the localized incidence rate of coronavirus infections. Whereas the impact of COVID-19 on the aggregate economy has already been felt, there is no doubt that differential (and potentially longer-lasting) effects will present challenges to local areas over the next few years, as locations with greater exposure face increased strain on almost all facets of their economies.  

    To construct our Coronavirus Exposure index, we rely on specific data sources: 

    We calculate the incidence rate of COVID-19 exposure by taking the ratio of confirmed cases to population in a given area. Areas with lower incidence rates have lower disruptions to public life and probably have fewer hurdles to lifting shelter-in-place orders. Conversely, places with higher rates will potentially face longer horizons to return to normal.   

    Figure 1 shows our standardized coronavirus risk factors for major metropolitan areas. Not surprisingly, metropolitan areas in and around New York City all have high exposure factor scores. Other high exposure scores are found in Chicago, Detroit, Miami, and New Orleans — all of which fall in the top decile of exposure scores. Note: the places that dominate the lower end of the exposure risk spectrum are smaller, less-densely populated metros. 

    qualitative-overlay-factors_figure 1

    2.2 Multifamily Risk Scores 
    A multifamily property usually refers to a residential property with more than five rental units. While renters are unlikely to move out of their apartments during this period of uncertainty, many of them failed to pay rent in April and cannot do so in the coming months while the economy remains in turmoil. This subsection describes the demographic and financial factors that affect the performance of multifamily properties in each market.  

    The Multifamily risk score uses data from the following sources: 

    • US Census Bureau: provides statistics on total population as well as various age groups
    • American Housing Survey (AHS): an annual survey conducted by the Census Bureau; it contains information on the number and characteristics of US housing units as well as the households that occupy those units

    Demographics 
    Population density is the most straightforward risk factor for urban dwellers. Residents in extremely dense urban areas have a higher risk of infection because it is more difficult to keep safely distanced from other residents. We calculate population density as the number of residents per square mile at both the state and metro levels (Figure 2). However, this factor is more useful at a more granular level. For example, some of the most urbanized cities — such as Chicago, Boston, and San Francisco — are ranked outside of the top-20. Their respective metros cover large swaths of suburban and exurban areas, which are much more sparsely distributed than inner-city areas. When calculated at metro levels, the population density of these large areas ends up lower than some smaller, more compact metros. To account for the differences between urban clusters and suburban regions within the same metros, we also calculate population density for all multifamily submarkets of the largest metros. As expected, the most urbanized neighborhoods top the list as a result. Section 2.4 explains this factor in more detail. 

    qualitative-overlay-factors_figure 2
    Another demographic factor we construct is the share of population between ages 20 and 39. Residents within this age group are more likely to be renters compared with other age groups. With a looming housing market crash, very few apartment renters are expected to purchase a home before the pandemic is over. Furthermore, many younger renters who share an apartment with roommates will probably look for an individual unit once their leases are up. All these reasons will not result in a booming multifamily market, but perhaps they will at least prevent it from completely collapsing. Not surprisingly, many metros in Florida rank poorly due to their large retirement communities. Given the higher risk of infection, older populations will most likely stay away from apartment buildings, causing the multifamily market in these metros to perform worse than in other metros with younger populations (Figure 3).  
    qualitative-overlay-factors_figure 3

    Apartment Density 
    Finally, we construct a measure of apartment density based on the number of rental units per square mile (Figure 4). This factor affects multifamily markets in very similar ways to population density. The more apartments there are in each area, the higher the risk of transmission among tenants who often share common areas such as elevators and lobbies, as well as neighborhood grocery stores. While renters may try to stay in their current apartments as long as possible for the time being, the trend toward urban living may reverse after the pandemic, as more people favor suburban locations with sparser populations and remote working becomes more prevalent.  

    As expected, apartment density is highly correlated with population density. However, there are some exceptions. For example, the Washington, DC metro ranks among the top-25 in the nation in terms of apartment density, but its population density is below the national average. Again, this includes suburban and exurban areas that are far sparser than the District of Columbia, which is why the submarket-level indexes are more advantageous.  

    qualitative-overlay-factors_figure 4

    2.3 Non-Residential CRE Risk Scores 
    Besides multifamily, the CRE asset class also includes many other types of income-producing properties such as office, retail, industrial, and hotel. The performance of these property types depends crucially on their ability to collect rent and retain tenants, not unlike multifamily properties. The ever-deepening recession means that many businesses are facing even greater challenges in their ability to pay rent, compared to apartment dwellers. Perhaps more alarming is the long and painful road to recovery, as many businesses may choose to forgo their physical spaces entirely as remote working, online shopping, and virtual experiences become increasingly more commonplace. This subsection describes the business factors that affect the performance of each property type.  

    The data source for the Non-Residential CRE risk scores includes the following: 

    • County Business Patterns (CBP): an annual business survey conducted by the Census Bureau; it provides subnational economic data by industry including the number of establishments and employment not only at the county-level but also at the ZIP code-level

    We aggregate the industries into the following six categories based on North American Industry Classification System (NAICS) codes pertinent to each of the major commercial property types: 

    1. Office: includes office-using industries such as financial services, information technology, professional services, and so on
    2. Retail: includes retail-related industries, including many types of retail stores, food services, drinking places, and so on
    3. Manufacturing: an industrial subsector that includes manufacturing industries
    4. Logistics: another industrial subsector that includes warehouse-related industries
    5. Hotel: includes leisure and hospitality industries
    6. Construction: not specific to any property type but useful for differentiating COVID-19’s impact on construction projects across geographies

    Business Density 
    Business density is defined as the number of establishments per square mile for each industry category. Business establishments in dense clusters will have a harder time reopening, because they must coordinate among themselves to ensure that employees can safely return to the business cluster while maintaining safe distances from one another. Table 1 lists the top-10 metros by business density for each industry category. Several metros appear at the top across all industry categories, such as New York, Fort Lauderdale, Orange County, San Francisco, and Los Angeles, all of which have large and diversified economies, with heavy concentrations of various industries due to conglomeration effects. Unfortunately, these characteristics will work against them in their efforts to reopen and allow employees to return. 

    qualitative-overlay-factors_table 1
    Employee Density 
    One way to measure employee density is the number of employees per square mile. This has similar implications to business density, in that business clusters with more employees will not be able to operate with full staffs before the pandemic is completely over. As expected, employee density is highly correlated with business density, since dense business clusters tend to have higher employee concentrations as well. Table 2 lists the top-10 metros by employee density for each industry category. Many metros that appear in this table also top the business density rankings.  
    qualitative-overlay-factors_table 2
    Another way to measure employee density is by using the number of employees per business establishment (Table 3). Unlike the previous measure, this factor concerns the density of employees within each business, rather than over a CRE market’s physical area. It is not so closely correlated with the business density measure. For example, Las Vegas ranks outside the top-10 in terms of hotel business density, but it has the largest average number of employees per hotel establishment. This result is sensible, given the tens of thousands of employees in its many large casinos. While these casinos are well separated from each other physically, it will remain a major challenge separating all the staff and patrons on the premises.  
    qualitative-overlay-factors_table 3

    Industry Reliance 
    Finally, we construct a measure of industry reliance based on the ratio of employment in a specific industrial category to the total employment within a certain geography. As the name suggests, this factor is intended to capture the reliance of a regional economy on a specific industry category. Over-reliance has a negative impact on the recovery of the corresponding CRE subsector because there will likely be a surplus of space supply, with many tenants returning to their spaces very slowly or even not renewing leases.  

    Table 4 lists the top-10 metros by industry reliance for each industrial category. Results are quite intuitive, as we see several metros known for a specific industry ranked near the top of the corresponding category. Those most reliant on office-using industries include some of the largest metros in the nation, such as New York; Dallas; Washington, DC; Atlanta; San Francisco; and Denver. While these metros have well-diversified economies, many of their industries belong to the tertiary sector (such as financial services, information technology, and professional services) and typically occupy considerable office space. By contrast, other industry categories see many smaller metros ranked at the top because of their specialized natures. For example, several metros most reliant on manufacturing industries are well-known industrial towns in the Midwest, while each of the top-10 metros most reliant on hotel-related industries are prominent tourist destinations. 

    qualitative-overlay-factors_table 4

    2.4 Submarket Risk Scores 
    To provide a more spatially granular context for our risk factor scores, we have produced the above-outlined metrics at the submarket level, where defined. Here, we also see differentiation between the Multifamily sector — providing risk factor scores for demographics and apartment density, and the Non-Residential CRE sectors — with business density, employee density, average employee per establishment, and industry reliance scores. We also produce Coronavirus Exposure risk scores at the submarket level. 

    We use the same data sources for the submarket risk scores as those listed previously for their state and metro area counterparts. However, producing the risk indexes at the submarket scale requires more spatially refined input data. We use the finest scale possible for each data item to ensure that we capture as much local variation as possible. 

    We provide two examples to highlight the granularity of our risk scores. First, Figure 5 shows the distribution of our hotel density risk factors across the six submarkets defined for the hotel sector in Boston, Massachusetts. It is clear that the highest concentration of hotels is found in the downtown and Cambridge areas — locations known for their tourist attractions and general commercial concentration (which draws many business travelers to the area). Submarkets in areas further away from Boston’s central business district (CBD) have much lower hotel density, and are thus less exposed to COVID-19’s disruptions. 

    qualitative-overlay-factors_figure 5
    Our second example (Figure 6) highlights localized differences in general office density across office submarkets defined in the Washington, DC metro area. Again, the greatest concentration of office properties is found in the CBD, with decreasing density as distance from the central business district increases. In this era of social distancing, higher-density CBDs will probably have greater challenges completely refilling commercial spaces at the same density as before. 
    qualitative-overlay-factors_figure 6

    2.5 Making Use of the Scores 
    The United States is a vast country, and the pandemic has spread very unevenly across states, metropolitan areas, and even submarkets. Coronavirus Exposure scores can be helpful in qualitatively assessing whether to compensate for the lagging information from the CRE fundamentals’ side. All things being equal, a location particularly affected by the virus has a longer road to recovery, given the public health concerns. As the pandemic situation evolves and the most up-to-date information becomes critically important, we plan to continue updating the scores and making them available.  

    The unique nature of this crisis also makes different impacts on subtypes within a property sector more important than ever. Public health measures — in the forms of shelter-in-place and social distancing — call for re-evaluating the use and viability of different CRE spaces. For example, while the retail property sector is being hit hard in general, the impact is not felt evenly: groceries and pharmacies remain open, while fitness centers, restaurants, and entertainment venues now face significantly more challenges given the immediate lockdowns and expected continued social distancing into the foreseeable future. Lenders should try to identify and analyze tenancy profiles of retail properties to better assess risk. In the office sector, for example, dense markets such as New York and San Francisco may have a more difficult time getting all workers back to CBD offices than their suburban counterparts. The scores we estimate can be used in this context at the state, MSA, and submarket levels. 

    3. Q-Factors Related to Policy Responses
    Given the unprecedented speed and severity of the COVID-19 pandemic, the US government and the Federal Reserve responded with fiscal and monetary stimulus on a massive scale never seen before. Considering the entire stimulus package, we list the most relevant programs for the CRE loans of most financial institutions. Many of these programs are completely new and, as a result, they were not factors captured by the historical datasets used to calibrate our existing models. 

    3.1 Programs Targeted to Help American Workers: Implications for the Multifamily Sector 
    To help American workers get through the crisis, the CARES Act provides immediate help to affected workers and families. These funds will likely help many apartment renters continue paying rents and/or mortgage payments as the program was intended. 

    Unemployment Insurance Provisions 

    • This program extends unemployment insurance by 13 weeks. Eligible individuals may receive unemployment benefits up to a maximum of 39 weeks, whereas previously, many states capped benefits at 26 weeks.
    • Pays an extra $600 a week in addition to the weekly benefit amount an eligible employee otherwise receives under state law.

    Rebates and Other Individual Provisions

    • A one-time tax rebate check of $1,200 per individual and $500 per child, with phase-out conditions.
    • Waives the 10% early withdrawal penalty for distributions up to $100,000 from qualified retirement accounts for coronavirus-related purposes.

    Tenant-Based Rental Assistance

    • Under this program, $1.25 billion is available to preserve Section 8 voucher rental assistance for seniors, the disabled, and low-income working families.

    Implications for Qualitative Overlay for Multifamily Loans 
    While jobless claims have skyrocketed, American workers and families — many of whom are multifamily renters — can possibly continue their rental payments given the generous government assistance. While we do not yet know the extent to which renters can take advantage of the benefits, we should generally expect a positive impact for multifamily property owners and lenders. In our experimentation with the CMM model, we find the effects can be reasonably approximated by adding interest reserves for multifamily loans. 

    3.2 Programs Targeted to Help Businesses: Implications for Non-Residential Properties 
    To help American businesses, particularly those without deep access to financial markets, the CARES Act offers various forms of relief during the shelter-in-place period. With financial help, businesses are in a better position to continue making rental payments to commercial property owners. 

    Paycheck Protection Program (PPP) 
    A key pillar of the CARES Act, this program is targeted to give small businesses temporary relief during the nationwide shutdown. Of the current total $659 billion package (including the initial $349 billion and an additional $310 billion), the potential impact on commercial properties and loans includes: 

    • Loan proceeds that may be used to make mortgage interest, rent, and utility payments.
    • In addition, the loan is forgiven to the extent proceeds are used from March 1−June 30, 2020 for paying mortgage interest and rent.Of all the programs, PPP is likely the most beneficial to commercial tenants and landlords.

    Other Major Programs 
    While smaller in size, other programs outside the PPP also help businesses:

    • Emergency Economic Injury Loan (EIDL) Grants: A total of $70 billion (including an initial $10 billion and an additional $60 billion) is targeted toward small businesses, nonprofits, and sole proprietors, among others.
    • Subsidy for Certain Loan Payments: $17 billion for the Small Business Administration (SBA) to cover all loan payments for existing SBA borrowers, including principal, interest, and fees, for six months.
    • Funding for the Department of Health and Human Services: Signed into law on April 24, 2020 as Phase 3.5 relief, it provides additional funding for hospitals and coronavirus testing. $100 billion is allocated to the Department of Health and Human Services to cover increased hospital expenses and lost revenue ($75 billion), and additional coronavirus testing ($25 billion). We expect this funding will have positive impacts for hospitals and medical offices.
    • Main Street Lending Program: Administrated by the Federal Reserve, it offers four-year loans to companies with up to 15,000 employees and revenues under $5 billion. P&I is deferred for one year. A Main Street Facility will purchase up to $600 billion of these loans, which will be made by banks. Maximum loan size is $25 million.
    • Technical Amendment Regarding Qualified Improvement Property: Enables businesses, especially in the hospitality industry, to immediately write off costs associated with improving facilities instead of having to depreciate those improvements over the 39-year life of the building. Note: this is a correction to an error in the 2017 Tax Cuts and Jobs Act that should help improve the after-tax cash flows of many commercial properties.
    • Project-Based Rental Assistance: A $1 billion program available to owners and sponsors of properties receiving project-based assistance under Section 8.

    In addition to government programs, the private sector has extended many programs to help small businesses get through the crisis. These risk-mitigating programs and factors were certainly not reflected in the historical data or explicitly modeled by our existing models, but they are worth considering as qualitative overlay factors. 

    Implications for Qualitative Overlay for Non-Residential CRE Loans 
    The hardest-hit commercial real estate sectors during the very near term are undoubtedly hotel and retail properties. Many retail property tenants and hotel operators are small businesses that can qualify for PPP loans, through which property owners would be able to continue receiving rental payments.  

    Lenders should consider all these policy response factors and make justifiable qualitative overlays on top of existing CRE credit risk models, such as Moody’s Analytics CMM and CRE loss rate models, by carefully assessing the extent of how much tenants and borrowers will benefit from a range of fiscal and monetary stimulus programs. For example, for the multifamily sector, the effect of combined government support programs could be similar to having additional interest reserve to support non-residential CRE loans.5 

    3.3 Programs Affecting Lenders 
    There are a few programs within the CARES Act that directly affect lenders. 

    • Temporary Relief from Troubled Debt Restructurings (TDR): A financial institution can suspend GAAP under Troubled Debt Restructuring (TDR) and impairments for accounting purposes. It covers March 1 until 60 days after the end of the national emergency or December 31, 2020, whichever is earlier. Lenders can work out temporarily stressed loans more effectively without worrying about the negative impact on financial accounting.
    • Forbearance of Residential Mortgage Loan Payments for Multifamily Properties with Federally Backed Loans: This program involves mandatory forbearance on request for COVID-19 impacts up to 90 days (30-day forbearance on request, extendable for two additional 30-day periods). It applies only to federally insured, guaranteed, supplemented, or assisted mortgages, including mortgages purchased or securitized by government sponsored enterprises (GSEs). There are no evictions or late fees during the period of forbearance. It covers the date of enactment to the end of the national emergency or December 31, 2020, whichever is earlier.
    • Temporary Moratorium on Eviction Filings: This moratorium on evictions applies to properties that participate in federal housing, homelessness, rural programs, or properties financed by federally insured, guaranteed, supplemented, or assisted mortgages, including mortgages purchased or securitized by GSEs. It covers both single-family and multifamily properties up to 120 days after the date of enactment.
    • Optional Temporary Relief from Current Expected Credit Losses (CECL): Commercial banks can now delay CECL implementation until the end of the national emergency or December 31, 2020, whichever is earlier. Furthermore, for banks opting to continue CECL implementation, regulators allow them to delay the incorporation of any adverse effects of CECL into regulatory capital for two years, and then to phase the marginal increase in over the course of the next three years.
    • The Federal Reserve TALF Program: The Federal Reserve also restarted the Term Asset-Backed Loan Facility (TALF) program, which will enable the issuance of asset-backed securities (ABS) backed by student loans, auto loans, credit card loans, loans guaranteed by the SBA, and certain other assets, including agency MBS and triple-A-rated securities from outstanding multi-borrower commercial mortgage-backed security (CMBS) transactions. Already, we have seen a significant narrowing of commercial loan spreads after the announcement, signaling a better functioning capital market for CRE loans.

    All these programs will help lenders ease the pressure from restructuring or selling assets haphazardly. The impact could also put a cap on refinancing rates for CRE loans, for those collateralized up by eligible high-quality properties. This can be particularly helpful for loans set to mature in the near term. 

    3.4 Different Meanings of Economic Measures 
    Given the public health nature of this crisis and unprecedented nationwide lockdowns, traditional measures of economic activities — such as GDP growth rates and unemployment rates — may need to be interpreted differently than in the traditional ways that drove the development of our existing models. For example, in a historical context, an annualized GDP decline of 30-40% and an unemployment rate of 15-20% would imply a devastating level of unprecedented credit losses. However, the nature of this pandemic also clearly indicates that a nontrivial part of the loss in economic activity and employment should be short-lived and will recover once the pandemic lockdown is lifted.6 For example, to think about the expected increase in the unemployment rate of 2Q 2020 — assuming 10% for the sake of explanation — we must break it down into two components:

    • A portion of jobs that is “permanently” lost. In today’s context, we define permanent as those jobs that will not return after we end the lockdown and reopen the economy. In industries such as airlines, hospitality, and dine-in restaurants, it is possible that both demand and capacity will continue to be limited for months, perhaps years. Therefore, a substantial part of the current job losses will be permanent. Within the context of commercial real estate, permanent losses of jobs and businesses lead to increases in vacancy rates, consequently depressing rents and prices.
    • A portion of jobs that is tied to and caused by the government-mandated lockdown only. These temporary job losses will recover soon after the lockdown is over. Continuing the above example using airlines, hospitality, and dine-in restaurants, while a permanent reduction in labor force is more likely than not, we expect rehiring of some employees temporarily laid off. After all, even a half-full hotel or a half-full restaurant will still need workers. Hence, a portion of the unemployed will come back to work shortly after the economy reopens. In the context of commercial real estate, temporary losses of jobs and businesses may not lead to increases in vacancy rates.

    The most difficult question surrounds the breakdown between the permanent and the temporary portions of the unemployment rates. If all the job losses are permanent, then the negative effect from current job losses on the CRE markets can easily surpass anything seen before. On the other hand, if all job losses are temporary, then the effect could be transient and miniscule. The truth is somewhere in between; that is, a 10% increase in the unemployment rate in the current environment will have a less-than-full negative impact, as in previous recessions. Unfortunately, we do not know the exact breakdown of the headline unemployment rates and will not know until we are well past the crisis stage. 

    The same rationale applies to GDP growth rates. It suggests that we ought to apply a haircut to the headline GDP growth rates and unemployment rates to derive more meaningful economic measures, which will then allow our models to correctly estimate credit losses under macroeconomic assumptions that are more comparable to historical norms. Given the lack of data to quantify whether economic measures are effective, one way is to qualitatively vary the inputs to capture the unique nature of this crisis. For example, instead of inputting a 15% unemployment rate into the CMM model, perhaps that number should be 10%, and so on. Today’s high uncertainty also calls for a spectrum of “what-if” analysis by varying a range of effective macroeconomic inputs. This is where management judgement and overlay are highly needed and warranted. 

    4. Applying Q-Factors: Thoughts
    CRE credit risk models can be grouped into loan-level and pool-level models. For instance, Moody’s Analytics CMM is a loan-level model that captures key risk factors across market, property, and loan details. On the other hand, its sister model, the CRE loss rate model, is a pool-level model, sharing a similar model framework. The general principles of applying qualitative factors are the same across both models, and other models for that matter. 

    Using the CMM model as an example, the sequence of our recommended loss estimation process7 would be as follows. 

    • Start by reviewing and adjusting both model configurations and model inputs by asking whether specific “new” elements in today’s environment are adequately incorporated into the existing data and model.
    • Examine the quantitative model predictions against historical episodes and recessions to develop an intuition about how the models make predictions based on quantitative inputs and historical relationships.
    • List and assess the key qualitative factors outside the risk factors already captured by the model. With COVID-19, as discussed in previous sections, there are two broad categories of factors the CMM model does not capture:

    Public health-related factors: This is where the coronavirus pandemic measures and various location and property-type specific risk scores come into play. Here are a couple of examples of qualitative factors:

    • For locations hard hit by the pandemic or places with higher density that may not be conducive to social distancing, the economic recovery may be slow. Hence, the negative impact on the CRE sector might be longer and more severe than what the model predicted.
    • While the retail property sector is severely affected, the impact is not felt evenly: groceries and pharmacies are still open, while fitness centers and entertainment venues now face massive challenges given the immediate lockdowns and social distancing in the foreseeable future. 

    Policy response factors: Of the many fiscal and monetary stimulus programs, it is useful to think along the three dimensions that affect CRE loans through direct or indirect channels:

    • Programs to help workers or the unemployed have the most direct impact on multifamily loans.
    • Programs to help businesses can have a combined positive effect on CRE loans backed by non-residential properties where retail, office, and industrial tenants may be able to withstand temporary business shutdowns with liquidity help.
    • Programs to help the financial markets and lenders can help reduce funding costs and mortgage coupon rates, mitigating the refinancing risks for loans maturing in the near term.

    Finally, do not over-adjust the model prediction unless absolutely justified. For example, all the fiscal and monetary programs are meant to reduce, not eliminate, the pain in the economy. When considering the potential credit losses for CRE loans, we will continue to see increased losses compared to what was expected before the pandemic. It is not necessary to “qualitatively” eliminate the increase in expected losses by citing government programs.

    To summarize, based on the new developments and unknowns in today’s COVID-19 world, it is perfectly justifiable to overlay qualitative factors. We encourage quantitative model users to gain a comprehensive assessment of the new situation, and then make informed judgment overlays along the dimensions where the models may have not fully captured the latest information.

    Please follow the links for state, metro, and submarket factors if you would like more information. Alternatively, please submit your request at www.moodysanalytics.com.  

    Appendix A. Data Structure 
    Our factor scores are provided in spreadsheet format. We produce one file for each geographic scale: state, metro, and submarket. The risk factor scores for specific property sectors are given in individual sheets within each file. Our state and metro files contain one sheet with Coronavirus Exposure scores (Exposure), one with Multifamily risk scores (Multifamily), and six individual sheets with Non-Residential CRE risk scores (Office, Retail, Manufacturing, Logistics, Hotel, and Construction). For example, you will see a column called “Q.Population.Share.20.39” in the state file (Q Factors_State.xlsx) Multifamily sheet. This column contains our Likely Renter Population Share risk factor. Sorting that column in ascending order reveals that the five states with the lowest proportion of population that are likely to be renters include Maine, West Virginia, Vermont, New Hampshire, and Florida. Conversely, Alaska, Utah, North Dakota, Colorado, and California are the states with the greatest percentage of likely renters — although the District of Columbia has by far the highest proportion of all state-level entities.   

    The structure of our submarket file is similar to the state and metro files, with one sheet for Multifamily risk scores (Multifamily) and five sheets for Non-Residential CRE risk scores (Office, Retail, Manufacturing, Logistics, and Hotel). Note that there is no Coronavirus Exposure score sheet within the submarket file, as submarket area definitions differ across property types. Therefore, within each submarket file sheet is a property type-specific Coronavirus Exposure score column to ensure that exposure scores are distinct to the localized submarket area as defined. You can find a Coronavirus Exposure score column (Q.Exposure) in the submarket file (Q Factors_Submarket.xlsx) Office sheet. Sorting in descending order reveals that 13 of the 20 highest exposure score office submarkets are in the New York metro area.

    Appendix B. Sample Q-Factor Table   

    qualitative-overlay-factors_table 5
    Acknowledgements
    We would like to thank Robby Holditch, Mary Hunt, James Hurd, and Masha Muzyka for their help. 
    Footnotes

    1 https://www.bea.gov/news/2020/gross-domestic-product-1st-quarter-2020-advance-estimate

    2 https://www.cnn.com/2020/04/09/business/americans-rent-payment-trnd/index.html

    3 https://www.nmhc.org/research-insight/nmhc-rent-payment-tracker/

    4 To qualify for full forgiveness, employers must use at least 75% of the loan amount to fund payroll costs, with the rest going toward rent, mortgage interest, or utilities. Borrowers must also maintain their pre-crisis full-time headcount. It covers the first eight weeks after receiving the funds. 

    5 Notably, the CMM model provides this feature.

    6 Our assumption is that gradually, the medical crisis will pass. The United States will then reopen and resume most economic activities, despite some lingering economic damage. 

    7 There can be many different variations; we are sharing our ways of thinking as model developers.

    References

    Moody’s Analytics, “Coronavirus (COVID-19): Credit Risk Impact on Commercial Real Estate Loan Portfolios.” March 22, 2020.

    Moody’s Knowledge Portal on COVID-19: https://www.moodys.com/Coronavirus