Liquidity Coverage Ratio

The formula

The LCR is structurally simple:

LCR  =  Stock of HQLA   /   Total net cash outflows over 30 calendar days   ≥   100%

HQLA               = unencumbered high-quality liquid assets, tiered:
                       Level 1: 100% of value (cash, reserves, top sovereigns)
                       Level 2A: 85% of value (cap: 40% of HQLA)
                       Level 2B: 50-75% of value (cap: 15% of HQLA)

30-day outflows    = Σ (funding source_i × run-off rate_i) − min(75% × inflows, inflow cap)

Two design choices distinguish the LCR from a simpler "cash on hand" rule. First, the denominator is not literal scheduled cash flows — it is a stressed 30-day scenario with prescribed run-off rates that approximate what would happen in a market panic. Second, the numerator excludes encumbered assets (pledged as collateral or in repo), excludes anything inside the bank's own ring-fenced subsidiaries, and tiers acceptable assets by historical liquidity — central-bank reserves count fully, AAA covered bonds at 85%, BBB corporates at 50-75% with caps.

Why the LCR exists: 2008's actual cause of failure

The standard public narrative about 2008 — "banks were undercapitalised" — is only half-right. By the day Bear Stearns collapsed in March 2008 and Lehman Brothers collapsed in September 2008, both firms were technically solvent under Basel II capital rules. What killed them was liquidity, not solvency:

• Bear Stearns. Held an $18bn liquidity pool at the start of March 2008. By mid-March, repo counterparties had pulled funding, prime-brokerage clients had withdrawn balances, and the pool was approaching zero. JPMorgan acquired the firm with Fed assistance over the weekend of March 14-16. Bear was technically solvent on the merger date; it was incapable of funding itself the next business morning.
• Lehman Brothers. Started September 2008 with a $40bn liquidity pool. After Treasury Secretary Paulson signalled in early September that there would be no government rescue, repo counterparties demanded higher haircuts, and clients withdrew balances. The pool ran out by September 14. Lehman filed Chapter 11 on September 15. Like Bear, Lehman was technically solvent on the bankruptcy date — its assets, at carrying value, exceeded its liabilities. It simply could not fund the next morning.
• Northern Rock. Earlier (September 2007), the UK building society had also been technically solvent but lost £25bn of wholesale funding in three weeks as confidence collapsed. The UK government nationalised it in February 2008.

The pattern was identical: solvent banks killed by funding evaporation. The Basel II framework had no liquidity rule. Basel III addresses the gap with two ratios — the LCR for acute (30-day) stress and the Net Stable Funding Ratio for structural (1-year) mismatches. The LCR was published in December 2010, refined in January 2013, and phased in starting January 2015 at 60%, rising 10 percentage points per year to full 100% in January 2019.

Worked example: computing LCR for a mid-size bank

Consider a stylised commercial bank with the following 30-day stressed cash-flow profile and asset book:

Numerator: HQLA stock	$bn	Multiplier	Weighted
Central-bank reserves (Level 1)	$15	100%	$15.0
US Treasuries, unencumbered (Level 1)	$8	100%	$8.0
AAA covered bonds (Level 2A)	$4	85%	$3.4
Investment-grade corporate (Level 2B)	$2	50%	$1.0
Total HQLA (after caps)			$27.4

Denominator: 30-day net outflows	$bn	Run-off rate	Weighted outflow
Insured stable retail deposits	$80	3%	$2.4
Insured less-stable retail deposits	$40	10%	$4.0
Non-financial corporate operational deposits	$30	25%	$7.5
Non-financial wholesale (non-op)	$10	40%	$4.0
Unsecured financial funding (interbank)	$5	100%	$5.0
Inflows (loan repayments)	$3	50%	−$1.5
Total net outflow			$21.4

LCR = $27.4bn / $21.4bn = 128%. The bank passes the 100% floor with a 28-point buffer. If the bank lost $5bn of HQLA value (a 10-year Treasury markdown after a rate spike, for example), its LCR would drop to ($27.4 − $5) / $21.4 = 105% — still passing, but close to the line. This kind of HQLA markdown was precisely what destroyed SVB's effective liquidity in March 2023: rising rates marked their held-to-maturity Treasury portfolio down by tens of billions, and once that loss became public, deposits ran faster than the LCR's run-off assumptions had ever anticipated.

HQLA tiering in detail

Tier	Assets	Haircut	Cap
Level 1	Cash · central-bank reserves · sovereign debt of high-quality issuers (0% RW) · supranational debt (BIS, IMF, ECB)	0% (counts at 100%)	No cap
Level 2A	Sovereign debt of 20% risk-weight issuers · covered bonds (AA- or better) · corporate bonds (AA- or better)	15% (counts at 85%)	40% of total HQLA
Level 2B (corporate)	Corporate bonds (BBB- to A+)	50%	Within 15% sub-cap
Level 2B (equity)	Common shares of non-financial firms in major indices	50%	Within 15% sub-cap
Level 2B (RMBS)	Residential mortgage-backed securities, AA or better, no internal securitisation	25%	Within 15% sub-cap

The vast majority of HQLA at major banks is Level 1: central-bank reserves and Treasury bills. The tiering exists to allow some flexibility but discourages reliance on assets that have historically lost liquidity under stress. The 2008 crisis showed that corporate bonds — even AAA ones — could lose 30% of value in a panic, while Treasuries gained value as a flight-to-quality.

LCR vs other Basel III ratios

Ratio	What it measures	Time horizon	Minimum	2008 lesson
CET1 / Tier 1 / Total capital	Loss-absorbing capital vs risk-weighted assets	Static	7% / 6% / 8%	More and better capital
Leverage ratio	Tier 1 vs total exposure (unweighted)	Static	3% (4-5% G-SIB)	Risk weights can be gamed
LCR	HQLA vs 30-day stressed outflows	30 days	100%	Bear Stearns & Lehman were illiquid, not insolvent
NSFR	Available stable funding vs required	1 year	100%	Don't fund long-dated assets with overnight repo

2023 and the limits of the LCR

The March 2023 collapse of Silicon Valley Bank tested the LCR's design in two ways. First, SVB was tailored out of the full LCR by the 2019 US prudential-tailoring rules — banks below $250bn in consolidated assets faced a "Category IV" reduced LCR with relaxed assumptions, and SVB sat just below that threshold. Second, even if SVB had been subject to full LCR, the actual run that destroyed the bank was much faster than the 30-day assumption: SVB lost $42bn of deposits in a single business day on March 9, 2023 — roughly 25% of total deposits — and another $100bn was queued to leave on March 10 before the FDIC closed the bank. The LCR's run-off rates were calibrated for 30 days; the actual run was 36 hours.

The post-mortem produced three regulatory threads. First, post-SVB proposed rules in the US extended LCR-like requirements down to banks with $100bn or more in assets — the May 2023 SLR FBO/IDI proposals. Second, the Federal Reserve's discount-window apparatus and the newly-created Bank Term Funding Program were positioned as backstops for the LCR's 30-day window — if HQLA could not be sold quickly enough in stress, it could still be pledged to the Fed for cash. Third, calibrating run-off rates for technology-enabled deposit flight (social-media-induced runs, instant deposit transfers via mobile apps) became a recurring topic in BCBS publications through 2024.

Pitfalls and ongoing issues

• Window-dressing. Banks can boost the LCR around reporting dates by selling Level 2 assets, buying Level 1, and reducing wholesale funding maturities to over 30 days. Several supervisors now require intra-period averaging or daily LCR reporting to limit this.
• Operational deposit classification. Treating a deposit as "operational" (25% run-off) instead of "non-operational" (40%) materially improves the LCR. Definitions were tightened in 2014; supervisors continue to challenge bank classifications.
• Cliff effects at 30 days. Funding that matures on day 31 is treated as not running off in the 30-day window. Banks naturally lengthen contractual maturities just past the threshold, even where economic maturity is shorter.
• HQLA concentration. If every bank holds the same Level 1 assets, the assumption that Level 1 remains liquid in stress may not hold — there is no one to buy if every bank simultaneously needs to sell. The 2020 COVID Treasury-market dislocation hinted at this risk.
• Run-off calibration vs digital banking. The 3-10% retail-deposit run-off rate is based on historical experience that pre-dates online and mobile banking. SVB's experience suggests that uninsured, concentrated deposit bases at digitally-native banks can run at 25%+ per day, not per month.
• Cross-border HQLA fragmentation. A US-headquartered G-SIB's HQLA pool is often trapped in subsidiaries by host-country ring-fencing rules. The consolidated LCR can pass while the subsidiary LCR fails — and the subsidiary is what matters for actual liquidity stress.