After the financial crisis of 2007-2009, bank regulators introduced the liquidity coverage ratio (LCR) to enhance the resilience of the banking system through better measurement and management of liquidity risk. Under the LCR rule, large U.S. bank holding companies are required to hold an adequate amount of unencumbered[1] high-quality liquid assets (HQLA) that can be converted into cash in private markets. In addition, these bank holding companies are required to disclose the average of their LCRs and subcomponents, including HQLA amounts, on a quarterly basis. By requiring banks to maintain a liquidity buffer on their balance sheets, the rule has induced significant changes to banks’ balance sheet management.
While the importance of liquidity management has been highlighted, most banks are not required to hold HQLA because they are not subject to the LCR requirement.[2] Hence, policymakers and researchers have been constructing their own estimates of U.S. banks’ HQLA. These estimates help assess the liquidity of the broader banking sector and analyze how banks have managed their balance sheets over time.
In this blog, we estimate U.S. banks’ HQLA amounts using the methodology proposed by Ihrig et al. (2019) and compare these estimates to eligible HQLA amounts in bank holding companies’ public disclosures.[3] By doing so, we measure the gap between estimated and eligible HQLA. In addition, we analyze differences between the two main HQLA components: Level 1 assets, consisting primarily of reserve balances and U.S. Treasury securities, and Level 2A assets, which are mainly comprised primarily of agency mortgage-backed securities (MBS).
Our study finds a sizable gap between estimated and reported HQLA amounts. This is because the HQLA estimates are intended to measure total HQLA, but the HQLA amounts in bank disclosures only include eligible HQLA under the LCR rule. Between 2018 and present, estimated HQLA has been, on average, 18 percent higher than reported HQLA. In addition, our analysis shows contrasting trends across the two subcomponents of HQLA analyzed: estimated Level 1 assets are 7 percent lower than reported values, while estimated Level 2A assets are 340 percent higher.
The differences between estimated and reported HQLA amounts come from three key factors. First, our estimates include total HQLA across all subsidiaries, but the LCR rule prevents “trapped liquidity” from being included as a holding company’s eligible HQLA. Second, our estimates include both encumbered and unencumbered securities, while eligible HQLA only counts unencumbered assets. Lastly, our estimates use point-in-time data from regulatory reports, whereas eligible HQLA is based on quarterly averages of daily amounts of HQLA.
Our study shows the difficulties in estimating the liquidity of the banking sector using publicly available data. It suggests that researchers and policymakers should be careful when they are using the HQLA estimates given the large differences between estimated and reported. Understanding these measurement challenges is crucial to evaluate banking sector resilience and design an appropriate liquidity regulation.
Data Analysis
We construct estimated measures using the methodology proposed by Ihrig et al. (2019) and compare them to reported amounts in public LCR disclosures. We list the data used for the estimates in the Appendix. We obtain eligible HQLA amounts from banks’ public LCR disclosures.
HQLA is divided into three groups based on the assets’ expected ability to serve as liquid assets during a crisis. Level 1 assets are considered the most liquid and require no haircuts in the HQLA calculation. These include reserve balances, U.S. Treasury securities, GNMA securities (which are full-faith-and-credit guaranteed by the federal government) and other agency debt that is also fully guaranteed (“non-GSE agency debt”). Level 2A assets are expected to be somewhat less liquid in a crisis, and their fair value is discounted by 15 percent when calculating their HQLA value. Level 2A assets include GSE debt and GSE mortgage-backed securities (FNMA, FHLMC and FHLB securities). Level 2B assets are deemed to be the least liquid of the assets eligible as HQLA and include municipal and corporate bonds. They are discounted by 50 percent and capped at 15 percent of HQLA. Together, Level 2A and Level 2B HQLA can make up no more than 40 percent of HQLA.
Following Ihrig et al. (2019), we exclude Level 2B assets from our HQLA estimates. Doing so has a minimal impact on the estimation of HQLA because Level 2B assets tend to comprise a much smaller portion of banks’ balance sheets and face both substantial haircuts and the 15 percent cap. For consistency in our comparison, we subtract Level 2B assets from the reported eligible HQLA amounts. We note that this adjustment has a minimal impact since Level 2B assets constitute less than 1 percent of eligible HQLA.
Figure 1: Aggregate Estimated and Reported Amounts of HQLA
Source: LCR disclosures, FFIEC Y-9C, BPI Calculations.
Note: Estimated measures are constructed using the methodology proposed by Ihrig et al. (2019).
Figure 1 compares the estimated versus reported HQLA between 2018Q2 and 2024Q2. Panels A, B and C plot the total HQLA amounts, Level 1 assets and Level 2A assets, respectively. We further divide the sample period into the pre-pandemic (the period prior to 2020Q1) and the pandemic (the period including and following 2020Q1) period.
We find that the gap between estimated and reported HQLA amounts has changed over time. Estimates are 17.9 percent higher than the reported amounts on average during the entire period; however, this relationship showed noticeable changes around the COVID-19 pandemic. Prior to the pandemic, estimated and reported HQLA moved closely, with estimates 1 percent below reported values on average. However, after the pandemic estimated amounts were much higher than the reported values by an average of 26 percent.
Panel B of Figure 1 compares estimated versus reported amounts of Level 1 assets. Interestingly, the difference between estimated and reported amounts moved in the opposite direction compared to total HQLA. Before the pandemic, estimates were lower than reported amounts by 20 percent, but this gap nearly disappeared afterward, with estimates averaging only 0.7 percent below reported values.
Panel C shows that the estimated amounts of Level 2A assets are consistently overestimated throughout the period, though the magnitude varied significantly. Prior to the pandemic, estimates exceeded reported amounts by 94 percent. This disparity widened considerably after the pandemic, with the gap increasing by an additional 350 percentage points.
Figure 2: Aggregate Estimated and Reported Amounts of HQLA over Total Assets
Source: LCR disclosures, FFIEC Y-9C, BPI Calculations.
Note: Estimated measures are constructed using the methodology proposed by Ihrig et al. (2019).
In Figure 2, we compare the share of the estimated and reported HQLA amounts against total assets. While Panel A of Figure 1 shows a substantial increase in HQLA in terms of dollar amounts after the pandemic, Panel B of Figure 2 indicates that eligible HQLA remained relatively stable at 20 percent of total assets throughout the period. This indicates that the increase in HQLA amounts was primarily driven by growth in overall bank balance sheets.
However, the composition of reported HQLA underwent significant changes during this period. The share of Level 1 assets increased by 3 percentage points, from 16 percent to 19 percent of total assets. This increase was offset by a corresponding 2.5 percentage point decrease in Level 2A assets, which declined from 3 percent to 0.5 percent of total assets. These shifts show a clear reallocation within banks’ reported HQLA portfolios before and after the pandemic, substituting away from Level 2A assets towards Level 1 assets.
Potential Factors Driving Differences
The estimated and reported amounts likely differ for several reasons. First, the discrepancy may be driven by trapped liquidity. A bank holding company can only count its subsidiary commercial bank’s HQLA up to the amount of that bank’s projected net cash outflows unless this so-called “excess HQLA” is available for transfer to the holding company without statutory, regulatory, contractual or supervisory restrictions. This restriction is intended to capture the limits on which a commercial bank can provide funding to the nonbank subsidiaries of the holding company. While our HQLA estimates include total HQLA held across all subsidiaries of a banking institution (commercial bank as well as other subsidiaries of the holding company), banks’ LCR disclosures only include eligible HQLA at the bank holding company and exclude trapped liquidity at the subsidiary level.
Second, the proxy HQLA series includes securities that may be encumbered, while reported HQLA only includes unencumbered securities. Banks use Treasury securities for repo transactions and often pledge MBS to a Federal Home Loan Bank (FHLB) for liquidity management purposes. Securities pledged to a Federal Reserve Bank or a Federal Home Loan Bank are only eligible as HQLA if they are not backing a currently outstanding loan. While the LCR rule only allows assets that are unencumbered by liens and other restrictions on transferability to be counted toward HQLA, our HQLA estimates include both encumbered and unencumbered assets due to data limitations. On the other hand, HQLA received in reverse repos are not an on-balance-sheet item but are included in HQLA. The exclusion of Treasury securities received during reverse repo transactions would lead to an underestimation of HQLA.[4]
Third, estimates and reported amounts of HQLA likely differ because of different time aggregation methods. Our HQLA estimates use point-in-time, end-of-quarter data from the regulatory reports. In contrast, eligible HQLA is based on quarterly averages of daily amounts.
Our analysis shows that the gap between the estimated and reported values of HQLA is large and the gap fluctuates over time. This gap originates from the fact that the proxy series and reported amounts measure the amount of liquid assets differently; the estimates measure the amount of available HQLA, whereas the reported amounts indicate eligible HQLA.
It would be important to know which factors contribute to the difference between reported and estimated HQLA. For example, knowing the magnitude of trapped HQLA at bank subsidiary levels would enhance our understanding of banks’ liquidity profiles. This limitation stemming from the restricted movement of liquidity across entities within the same group (due to national laws, supervisory approaches and internal risk management practices) prevents entities with excess liquidity from compensating for the shortfalls of others.
Conclusion
Our analysis reveals large differences between estimated and reported HQLA amounts for large U.S. banks. The estimated HQLA has been 18 percent higher than reported HQLA on average. In addition, this relationship has varied considerably over time, particularly around the COVID-19 pandemic. Interestingly, our HQLA component-level analysis shows contrasting trends; estimated Level 1 assets are 7 percent lower than reported, while estimated Level 2 assets are 340 percent higher than reported.
These findings highlight the challenges in accurately measuring banks’ liquidity buffers using publicly available data. The large discrepancies between estimated and reported HQLA suggest that researchers and policymakers should be careful when they are using the HQLA estimates to assess the liquidity of the banking sector. Understanding these measurement challenges is crucial to evaluate banking sector resilience and design an appropriate policy.
Reference
Jane E. Ihrig, Edward Kim, Cindy M. Vojtech and Gretchen Weinbach, “How Have Banks Been Managing the Composition of High-Quality Liquid Assets?,” Federal Reserve Bank of St. Louis Review, Third Quarter 2019, pp. 177-201.
https://doi.org/10.20955/r.101.177-201
Appendix Table 1
[1] The assets can be pledged to a Federal Reserve Bank or a Federal Home Loan Bank but cannot be backing any outstanding loans.
[2] Some banks that are not subject to the LCR rule may still be required to hold HQLA to meet the internal liquidity stress tests (ILSTs) requirements.
[3] Our analysis covers banks that reported their HQLA totals between 2018Q2 and 2024Q2. These include JPMorgan Chase, PNC, Bank of America, State Street, U.S. Bancorp, Wells Fargo, Northern Trust, Citi, Morgan Stanley, Capital One, Goldman Sachs, and Bank of New York Mellon. Some banks reported their HQLA in 2018Q4 but stopped doing so following the implementation of tailored LCR rules and disclosure requirements. These include M&T Bank, KeyCorp, Huntington, Fifth Third, Citizens, American Express, Ally, Regions, and Discover.
[4] Securities received in reverse repo would be included in “Securities purchased under agreements to resell (HC 3.b, BHCK B989).” Ihrig et al (2019) likely excluded this line item because these securities are not broken down by type.
