By Jeff Huther
Over the past year or so, the press, academic papers and policymakers have claimed that bank exposure to commercial real estate may cause hundreds of banks to fail. That view is based on the premise that office occupancies have been permanently reduced by the pandemic and higher interest rates have increased the likelihood that some borrowers will be unable to afford future debt payments. Absent additional shocks, however, the concerns related to banks are overblown. The data used in most analyses (that is, loans underlying commercial mortgage-backed securities deals) are not representative of bank exposures, and data from other researchers confirms previous ABA research showing that banks have applied strong underwriting standards to exposures they do have.
CMBS is a poor proxy for banks’ CRE exposures
CRE is much bigger than offices in business districts of major metro cities. ABA and others have shown that banks have highly diversified CRE holdings and low exposure to office buildings in large cities. Concerns about bank exposure continue to pop up, however, largely from writers looking at data sources that rely heavily on large real estate deals in urban centers. These reports often fail to consider that the lenders include a whole host of nonbank firms, such as insurance companies.
Tying CRE losses to banks is difficult because data on holders of CRE are either too broad (the Federal Reserve’s aggregate statistics) or too narrow (private-sector providers of CRE statistics such as Trepp, Green Street or Newmark that focus on large, and likely localized, transactions). The aggregate statistics tell us that banks have large exposures to CRE, while the private-sector statistics have led to stories of maturity walls, falling prices and returned keys. The data bridge between the macro bank statistics and the large office sector, however, is a rickety one built with academic studies like the frequently cited one above relying on CMBS.
The Fed’s publicly available data classifies CRE as everything from mortgages on ice cream stands to mortgages for computer processing centers. In contrast, the mortgages that get securitized in CMBS are likely to be large loans for financing office towers and hotels, which are strongly biased towards the most vulnerable forms of real estate in the current economic environment.
One indication of the CMBS bias comes from Fed researchers who have access to loan level data for large banks and published a paper carefully comparing banks’ commercial real estate holdings to CMBS exposures for large banks. To make this comparison, the authors excluded bank loans for small CRE deals (those under $1 million), construction and all loans not backed by offices, retail, hotel and industrial mortgages — that is, to make a fair comparison between bank CRE exposures and CMBS exposures, the authors had to exclude the types of loans that are core to bank commercial lending. From the publicly available aggregate data, we know that these exclusions represent more than 40 percent of all CRE exposures of domestic banks.
Looking at CMBS lenders provides another indication of the bias inherent in deriving bank exposures to CRE from CMBS. One source of information on CMBS investors is an SEC requirement that publicly held investment fund managers must disclose their securities investments quarterly (see Form 13F). To get an idea of who holds CMBS, we created a sample of the largest 100 CMBS deals outstanding as of December 31, 2023, which covered $285 billion in underlying loans. Of reporting firms, 818 held positive amounts in the sample, holding almost exactly one third (33.8 percent) of the total in 8,595 positions. Reporters were almost entirely asset managers and insurance companies (see figure 1), clear evidence that CMBS should not be used for representations of bank CRE exposures.
Source: Bloomberg, authors’ calculations
CMBS may have only a few underlying loans (sometimes only one loan) so investors cannot rely on diversification to limit losses. Instead, CMBS issuers provide a lot of information about the underlying loans so potential investors can make informed decisions on borrower risk. This information includes details about the type of real estate backing the loans. In our sample of the largest 100 CMBS deals, office property was the largest category at almost $100 billion but only represented 34 percent of the total exposures in CMBS (see figure 2).
Source: Bloomberg, authors’ calculations
An interesting aspect of CMBS is that deals with office and retail loans are more likely to be diversified across loan types than other deals. CMBS in which office loans are the largest share consist, on average, of only 43 percent office loans, the second most diversified collateral type, shown in figure 3. (Lower values reflect greater diversification.) The comparison to other loan types is likely distorted by the small number of deals, say in mobile home parks or self-service storage, that leads to specialization of investors. Therefore, simply identifying a CMBS deal as “office” is not a good representation of the underlying exposure.
Source: Bloomberg, authors’ calculations
We cannot say whether the investor types from our sample are representative of the entire CMBS universe, but they clearly show that CMBS exposures should not be directly linked to bank exposures. In Morgan Stanley’s data, CMBS maturities are more front-loaded than bank-held loan maturities which appear evenly spread over the next four years, suggesting that predictions of banks hitting a wall of CRE maturities are misplaced. Data from the insurance industry are consistent with the charts above in terms of identifying insurers exposure. Insurers held almost $300 billion of CMBS at the end of 2022, and the Fed’s aggregate data show that insurers held almost $500 billion in total CRE exposures at the end of 2023.
Underwriting standards
Bank underwriting standards for commercial real estate, especially since the global financial crisis, have been very conservative. The Fed paper linked above, using 2012-2017 data, showed average loan-to-value at 56 percent at origination for bank CRE loans (compared to 65 percent for loans in CMBS). A possible word of caution is that the Fed’s analysis includes distribution statistics of LTV that show banks had a much wider range of exposures than CMBS holders although this difference is very likely attributable to differences in loans offered rather than in risk aversion. Nonetheless, the value of underlying loans would have to fall on average by almost half before banks incurred significant losses.
Another Fed paper using confidential loan-level data compared modifications of bank loans to modifications of loans backing CMBS. The authors found that banks more readily modify loans held on their books than loans held in CMBS. The modifications had flow-on effects on later delinquencies: the CMBS loans in the Fed study had slightly lower 90-day delinquency rates than bank-held loans before the pandemic and much higher delinquency rates during the pandemic. The authors concluded that banks’ willingness to work with troubled borrowers generally leads to more modifications than CMBS loans but lower delinquencies during periods of stress. (In the first Fed paper cited above, the authors work through the complicated relationship between banks and borrowers, taking into account the effects of borrowers’ knowledge that bankers will work harder to avoid default than CMBS managers.)
Estimates of the peak-to-trough decline in prices for office real estate vary widely and, in any case, the average will not be reflective of underlying variations. A large CRE service provider, Cushman and Wakefield, uses changes in the prices of REITs to estimate expected losses by property types and concludes that office building prices may ultimately fall to half of peak values. If so, a portion of borrowers will default, and some lenders will experience losses. Most borrowers, however, are reluctant to default even when values fall below their equity stakes. This reluctance may be motivated by non-pecuniary reasons, because the properties are owner-occupied or because borrowers believe that their properties will eventually increase in value.
Conclusion
In contrast to press reports based on CMBS exposures, banks have low exposures to “office” deals and higher underwriting standards. Some CRE lenders, including some banks, will undoubtedly face losses as a consequence of changing work habits. Publicly available data, however, tell us that the losses will be dispersed across borrowers and a wide range of lenders.
