Observations and Feedback on Liquidity and Funding in Resolution Implementation

1 Annex Observations and Feedback from the LFIR Test and Review A. Scope and Scenario Design

1. Scope of reporting Participating AIs generally provided the required LFIR information on both a consolidated and solo basis, with breakdowns by material entities and material currencies for LFIR purposes. However, the degree of completeness and granularity varied across submissions.
2. Scenario design All participating AIs constructed fast-moving resolution scenarios, some centred on idiosyncratic stress and others incorporating combined idiosyncratic and market-wide shocks. While entry into resolution occurred within one month in all cases, the duration and nature of the lead-up phase differed markedly: some assumed immediate entry, while others projected several weeks of recovery actions beforehand. A small number of scenarios did not appear to align fully with the HKMA’s preferred resolution strategy for the AI.
3. Differentiation between material LFIR entities Most AIs appropriately identified and reported on material LFIR entities consistent with their resolution strategy. However, the depth and consistency of entity-level breakdowns (projections, collateral pools, management actions) varied. Some AIs developed comprehensive analysis covering the levels of both consolidated group and each material entity, while some focused mainly on the Hong Kong entity. Entity-specific analysis — including tailored collateral eligibility and mobilisation pathways — is key for LFIR. B. Liquidity Projections in Resolution Scenario
4. Key cash flow drivers AIs identified the principal liquidity drivers in their projections, including deposit run-off, restricted or lost access to unsecured wholesale funding, and heightened variation margin / collateral calls arising from market volatility or credit rating downgrades below investment grade. Several AIs also factored in contingency support to subsidiaries / branches outside Hong Kong, additional costs to maintain operational continuity of critical services, one-off staff retention or replacement expenses post-resolution, and restructuring costs in the post-stabilisation phase. On the inflow side, projected sources included committed parental support and effects from other mitigating actions, such as non-renewal of shorter-dated loans and declining margin requirements as derivatives contracts matured.
5. Deposit run-off rates Deposit run-off patterns and magnitudes differed across submissions. Some AIs assumed institutional outflows dominating the first

2 week, followed by broader corporate and retail runs. Others projected the most intense outflows in the initial few days, with rates tapering thereafter. Recent overseas experience — notably the rapid deposit runs in the U.S. in March 2023 — has shown that digitally amplified runs involving high uninsured or sophisticated depositor concentrations can produce extremely front-loaded outflows (daily rates of 15–30% or higher in the first 1–2 days). Assumptions with aggressive early peaks therefore appear plausible in certain conditions for certain business models, whereas more gradual or moderate profiles may benefit from further calibration to reflect modern run dynamics. Cumulative retail CASA run-off assumptions in the lead-up to resolution ranged from approximately 13% to over 50%, with peak and average daily rates between roughly 1% to over 20%, and 0.4% to 17.5% respectively. More granular approaches (distinguishing insured vs uninsured, retail vs operational / non-operational / bank deposits, or customer segments) tended to yield more differentiated outcomes. Broader, single-rate assumptions for non-bank customer deposits risk understate behavioural differences in a resolution context. 6. Deposit behaviour post-entry into resolution AIs’ post-resolution run-off assumptions varied: some anticipated sharp stabilisation (e.g. via parental support, public statements or central bank backstops), while others allowed for continued (albeit tapering) outflows among uninsured or sophisticated depositors. In the March 2023 U.S. cases, deposit outflows were intensely front-loaded but did not invariably cease upon regulatory intervention or entry into resolution / bridge arrangements. In light of this experience, AIs are encouraged to conduct sensitivity analysis of more persistent run-off post entry into resolution — particularly for uninsured or concentrated deposit pools — to ensure prudent estimation of liquidity needs. 7. Incorporation of rating downgrade and derivative / FMI-related outflows Several AIs modelled elevated collateral and variation margin calls, as well as potential restricted access to payment / settlement systems or FMIs, following multi-notch downgrades. However, quantification of these contingent outflows was often high-level or incomplete, reflecting uncertainty over counterparties’ exact triggers and the velocity of margin movements in resolution. More detailed analysis of these second-order liquidity impacts — including potential replacement costs for terminated contracts — is recommended. 8. Time bands and horizon of reporting AIs were generally able to project daily cash inflows and outflows during the lead-up to resolution and for at least 90 days post-entry. Some demonstrated capability for broader time bands over extended horizons (up to 5+ years). The ability to project liquidity positions over longer horizons (beyond the initial stabilisation period) is particularly valuable in resolution. Post-stabilisation restructuring — which may involve orderly wind-down of certain business lines, transfer / sale of assets, recapitalisation, or implementation of an exit strategy — can extend over

3 multiple years, depending on the preferred resolution approach, complexity of the group structure, market conditions, and legal / regulatory processes. Extended projections help evaluate the sustainability of liquidity position throughout this phase, identify potential residual funding gaps that could arise later in the restructuring timeline, and support more informed decision-making on the overall feasibility and cost of the resolution strategy. C. Management Actions and Central Bank Facilities 9. Management actions Common liquidity options included loan book reductions, monetisation of HQLA and non-HQLA (via repo or outright sale), issuance of certificates of deposit or commercial paper, and parental funding support. Due to the compressed execution windows in fast-moving scenarios, longer-lead￾time actions (e.g. disposal of properties, fixed assets or non-core businesses) were frequently excluded. 10. Central bank facilities and collateral Most AIs assumed access to relevant HKMA liquidity facilities shortly before or upon entry into resolution, and modelled implementation times of 2-3 days. It should be emphasised that the availability and timing of any liquidity support would depend on the specific circumstances prevailing at the time (for example, the type, quality and quantity of collateral available, the size of the liquidity need, and broader market and systemic conditions) and would be determined by the HKMA on a case-by-case basis. Assumptions about access and implementation timelines should be treated with appropriate caution in resolution planning. The potential eligible collateral identified by the AIs mainly include Exchange Fund Bills and Notes, high quality liquid securities in foreign currencies such as US Treasuries, other investment-grade securities, residential mortgage loans, and other loans. AIs (including their non-Hong Kong branches and subsidiaries) reported that progress have been made to enhance readiness to access central banks’ facilities through engagement with authorities to understand collateral information required, and in some cases, to conduct end-to-end operational testing. Readiness is relatively advanced for high-quality securities and residential mortgages. D. Key Challenges 11. Modelling resolution related costs Many AIs highlighted difficulties in modelling contingent items and resolvability-related costs (e.g. elevated collateral and variation margin calls, heightened requirements for maintaining access to payment / settlement systems, replacement or re-contracting costs for critical shared services, resolution restructuring, and potential one-off expenses for staff retention).

4 These challenges often stem from the inherent uncertainty in resolution scenarios — such as the timing and scale of FMI access restrictions, counterparty behaviour under stays, or the duration and cost of maintaining critical functions while restructuring progresses — combined with limited historical precedents for certain resolvability barriers in a Hong Kong or cross￾border context. Cross-departmental coordination (including Treasury / Markets, Risk, Legal / Compliance, Operations, Business Units and other key functions) and reference to real-life experiences are essential for improving the coherence and realism of the projections. 12. Estimation of key liquidity options benefits While participating AIs identified the principal liquidity options available in a fast-moving resolution scenario, estimating realistic and credible liquidity benefits proved challenging — particularly for actions involving less liquid or illiquid assets. Key difficulties included (i) valuation uncertainty in a resolution context, where market dislocations, forced-sale discounts, credit rating downgrades, or loss of going-concern value can materially depress asset prices compared to normal conditions; (ii) execution timing and feasibility for actions such as outright sales of less liquid assets (e.g. syndicated loans, corporate term loans, or specialised portfolios), property disposals, or residual recovery actions in the post￾stabilisation phase (e.g. structured wind-down, business-line transfers, or asset carve-outs); and (iii) limited historical data specific to resolution scenarios, making it hard to calibrate haircut assumptions or timing delays realistically. AIs are encouraged to continue refining these estimates through scenario sensitivity analysis, reference to actual stress events, and consultation with specialists where appropriate. 13. Data granularity and readiness Data limitations were widespread across participating AIs. Shortfalls in underlying data granularity, timeliness of extraction, and heavy reliance on manual consolidation processes hindered the ability to produce accurate and timely LFIR projections at the required level of detail and time band (including breakdowns of protected deposits and identification of significant funding providers by type of funding instrument) as well as collateral information. Meeting expectations for rapid generation of key LFIR metrics (ideally T+1 or near-real-time) in a live crisis or early resolution phase will necessitate further investment in data architecture, automation of collateral workflows, and minimisation of manual reconciliation. 14. Reporting and visualisation capabilities Challenges also arose in transforming the available data into decision-useful formats. Many AIs were unable to present key LFIR outputs effectively through visual aids such as stylised balance sheets at different points in the resolution timeline, cash flow

5 trajectory charts, or dashboards illustrating the evolution of net liquidity positions. Such visual representations would significantly enhance senior management and resolution authority understanding and decision-making during a live crisis or early resolution phase. 15. Collateral reporting and mobilisation AIs generally noted that further considerations need to be given to less liquid / non-standard collateral to enhance operational readiness for accessing central banks’ facilities, including facilities outside Hong Kong. Key areas include (i) substantial manual effort and time to extract, validate and summarise eligible pools; (ii) legal and operational requirements (e.g. governing law, assignment restrictions, and legal due diligence); and (iii) the need to maintain ongoing engagement and perform testing with the relevant authorities. E. Way Forward 16. All relevant AIs are expected to progress implementation to ensure they can meet liquidity and funding requirements in resolution in line with LFIR-1 expectations. 17. Participating AIs have outlined constructive enhancement plans, including refinement of modelling assumptions for material LFIR entities, automation of securities monetisation processes, improved deposit and collateral classification / parameter application, leveraging forthcoming supervisory return enhancements and aligning LFIR improvements with parallel resolvability workstreams. 18. The HKMA Resolution Office will continue to engage individual AIs bilaterally on LFIR matters as part of the resolution planning programme, and will undertake additional testing as AIs’ capabilities advance.