By Dan Brown, Jeff Huther and Sharon Whitaker
ABA DataBank
A recent working paper published by the Federal Reserve Bank of New York has received attention because of its conclusions that some large banks are undercapitalized and that their distressed commercial real estate loan portfolios are inhibiting growth of other parts of the banks’ balance sheets. However, given the paper’s definitional and methodological weaknesses, its characterizations should be viewed with skepticism. Even without addressing these concerns, the “sizable” effects that the authors identify are insignificant from a balance sheet perspective.
Core definitional weaknesses
At the core of the analysis in the paper are two variables: “distress” of CRE loans and “undercapitalization” due to unrealized losses on banks’ securities holdings. The authors classify a CRE loan as distressed if net operating income of the property at any point is less than it was at the time of origination. Thus, a borrower that sees a fractional increase in operating costs is thrown in the same bucket as a borrower that sees a large increase in vacancies. Since borrowers’ operating costs are likely to rise over time regardless of vacancy rates, the authors are likely to be overstating the amount of distressed properties. Given this approach, neither the readers nor the authors have a reliable way to interpret the econometric results or how they would relate to bank balance sheet management.
The second main explanatory variable, undercapitalization, is a relative measure that the authors derive by dividing the 22 banks that they study into two groups based on banks’ capital levels, adjusted for unrealized securities losses, relative to regulatory requirements. The authors are working with confidential data, so we do not know how this definition relates to individual banks. Given variations in bank capital requirements, however, it is likely that this approach leads to size biases in the characterization of banks as well or undercapitalized, a point we will return to when we consider the data.
Definitional weaknesses in characterizations
‘Extend and pretend.’ The title of the paper suggests that the authors have identified two actions by banks — extending the original maturities of CRE loans and pretending that the loans will eventually be repaid — that ignore the fundamental relationship between banks and their customers.
The inclusion of contractual loan maturities and extensions is a normal credit risk management tool to address temporary weaknesses in loans. Banks review individual credits at least annually, and in many cases quarterly, proactively assessing higher risk credits due to mature in 18 to 24 months. Full analysis at the loan level of current and future vacancies, maintenance costs, debt service coverage and other factors provides a cumulative view of various asset classes of loans. Maturity dates allow banks to enhance covenants — such as strengthening guarantors, right-sizing the loans and demonstrating cash infusion from global debt coverage sources.
Accounting standards and regulatory guidance require banks to assess each credit to determine if it will receive all its cashflows, including market-based interest payments if there is an extension or the loan reprices. To support these specific risk management practices, banks are also required to have loan review verification and validation of the risk rating of CRE assets. This is a part of an accurate risk ratings process that keeps ahead of problems that may exist on the horizon. Renewals or extensions that come out of these reviews ensure that banks are appropriately managing changes in repayment risk.
Accounting standards do not allow banks to “pretend” and ignore impairment measurement. Each loan is evaluated and, if it is impaired, is written down to its fair market value. A loan may also be put on a watch list if it has been identified as having only a temporary risk (that is, it may still be performing and its debt service coverage may still be above 1:1). The bank will identify triggers for future upgrades or downgrades depending on continued loan performance. In other words, even in cases where loans are modified and extended, but market conditions appear to support repayment expectations, such modifications are nevertheless disclosed.
Maturity walls. Previous ABA analysis has emphasized that the mix of shorter and longer term CRE loans that banks finance naturally leads to a false perception of a “maturity wall”, one that that normally evens out over time. Recent data looks similar to the numbers at the beginning of 2023 which ultimately showed the maturity wall fading away as borrowers repaid mortgages or took out new ones. Figure 1 below shows private sector data similar to the authors’ confidential data. It is evident that maturities are not outsized in any particular year. Rather, loan maturities are fairly spread out over time. The maturity wall shown in the New York Fed paper is a characteristic of long-term loan portfolios with diversified maturity dates rather than a sign of stress.
Although the authors make the claim that “extend and pretend” led to a decline in new CRE originations, the motivation for fewer CRE loans is unclear. On the supply side, larger banks have made a concerted effort to reduce their CRE exposure. In Q3 2024, 22 of the 26 banks that reported office exposure of more than $1 billion reduced their exposure quarter-over-quarter by about 1.8% on average. Figure 2 tracks proportional CRE lending by large banks and highlights the proportional decline in CRE lending over the last few years. The demand side, however, has also been weak — developers and buyers of office CRE have been as fully aware of the shift to working from home as the rest of us. Ascribing the decline in loans to bank balance sheet management misses the broader dynamic of lower prices (requiring smaller loans) and a weakening of borrower interest.
‘Sizable’ econometric effects. The authors characterize their findings (for example, “weak” banks have CRE loans with up to a 0.9 percentage point lower probability of default than “well-capitalized” banks and up to 0.8 percentage point lower CRE originations) as “sizable” effects. The authors’ data, however, shows that CRE loans as a share of assets for the banks in their dataset is 7% and non-performing CRE loans are, on average, 1.3% of total CRE loans. That is, non-performing CRE loans are less than one-tenth of 1% of total assets in the authors’ sample of banks. Sizable is clearly in the eyes of the beholder.
Data weaknesses
REIT equity prices as CRE valuation proxies. The article asserts that current collateral values for CRE assets are reflected in the trading value of real estate investment trusts. CRE REITs, however, are not representative of banks’ CRE portfolios and REIT valuations overstate the narrow investments of CRE REITs.
The theory for using REIT prices is understandable — CRE portfolios are composed of properties with unique characteristics in markets with few transactions. So, instead of relying on infrequent and likely outdated prices, one could use the valuations of equity market participants assessments of the values of the REITs that hold similar properties.
This approach, however, ignores REITs’ leverage which creates a multiplicative effect of the changes in the prices of underlying properties on REIT valuations. Levered positions magnify the changes in the underlying asset. For example, if the properties held by a REIT fall in value by 10% and the equity in the REIT was originally one half of the property value (leverage of 2), the price of the REIT equity should fall by 20%.
Given the wide variations in leverage (see Figure 3), it would be difficult to impute real estate price changes from equity price changes, although we can say with certainty that unadjusted REIT price changes overstate changes in property values. Additionally, as the approach is based on the largest REITs, it further exaggerates potential changes in the value of banks’ CRE portfolios. Large CRE REITs are likely to hold a higher share of vulnerable CRE (large office towers in large cities) than would be the case for typical bank CRE portfolios. The blue line in Figure 3 puts into perspective the size of the office CRE REIT market relative to the overall REIT market’s value of $2 trillion.
Undercapitalization. Another key assumption in the paper is that some of the 22 large banks in the sample had weak capital levels when accounting for the mark to market value of securities. However, large bank capital levels have risen substantially over the past several years, and this holds true for the entire cohort. Figure 4 highlights that large bank tier 1 capital has risen by over 50 percent since the financial crisis. A lot of this increase can be attributed to post-crisis related requirements, but the fact remains that this increase cuts directly against the paper’s assumption that some large banks are undercapitalized.
The authors ignore key reasons why bank capital is not measured using the mark to market value of securities. A decline in market value of securities is not a measure of increased risk in those securities per se, but rather short-term fluctuations in interest rates — normal variations in bank balance sheets.
Crowding out C&I lending. An additional claim in the paper was that banks’ CRE management has “crowded out” commercial and industrial lending. However, while the authors tried to account for demand and macroeconomic factors, the magnitude of change in these factors during the period examined (the first quarter of 2022 through the fourth quarter of 2023) was so large, it would make it difficult to parse out any other reasoning behind changes in C&I loan balances. During the small sample period, the following changes occurred that culminated in a once in a generation shift in C&I lending conditions: The Federal Reserve stopped quantitative easing, the Fed raised the federal funds rate over 500 basis points in roughly 18 months, fiscal stimulus dried up, and Paycheck Protection Program loans, which were counted as C&I loans, were still in the process of being taken off of bank balance sheets.
According to the Federal Reserve Bank of Kansas City, rates on urban fixed-term loans have increased from 5.6% in 2022 Q3 to 7.77% in 2024 Q2, and rates on urban variable loans have increased from 6.41% in 2022 Q3 to 8.88% in 2024 Q2. Considering the increase in funding costs for banks, many banks would probably love to have new loans at higher rates from creditworthy borrowers on their books. However, higher rates have caused demand for C&I loans to crater. Data from the Senior Loan Officer Opinion Survey found that the share of respondents reporting stronger demand for C&I lending in 2023 was the lowest since the Great Recession. Clearly, the Fed’s monetary policy has been an important factor in demand for C&I borrowing.
Conclusion
The New York Fed researchers’ analysis employs measures of distress and undercapitalization that are inconsistent with common definitions and ignore bank-borrower relationships, accounting principles and valuation techniques. Their use of confidential data limits our ability to fully assess their analysis and leaves open questions about the banks in the analysis — is bank size correlated with other variables, are there differences in balance sheet composition between the regional banks and the largest banks or are there other bank characteristics that could help explain the differences the authors claim to see? Furthermore, the size of CRE portfolios strongly suggests that losses are unlikely to have a sizable effect on bank balance sheets.
Dan Brown is a senior director and economist at ABA. Jeff Huther is VP for banking and economic policy research at ABA. Sharon Whitaker is VP for real estate finance policy at ABA. For additional research and analysis from the ABA’s Office of the Chief Economist, please see the OCE website.
