Liquidity early warning indicators: what to measure and where to set thresholds

Why the ratios are not enough#

Liquidity early warning indicators are the quantitative and qualitative signals a bank monitors to detect funding deterioration before it reaches the regulatory ratios.

The reason they matter is structural, not stylistic: the Liquidity Coverage Ratio and the Net Stable Funding Ratio are stock measures, computed periodically against a standardised set of assumed run-off rates. They describe a position, not a trajectory.

A bank losing its most price-sensitive corporate deposits over three weeks, paying steadily more for the same wholesale tenor, and quietly rolling shorter each month can show an entirely comfortable LCR throughout — right up to the reporting date on which it does not.

Early warning indicators exist to observe the trajectory that the stock measure conceals.

Watching deposits move#

Deposit outflow patterns and velocity form the first family, and the discipline is to measure movement rather than balance.

Net outflow by behavioural segment over rolling windows, the rate of change in that outflow, attrition among balances the behavioural model classifies as stable, migration between demand and term products, and the concentration of any single week’s outflow all say something a period-end balance does not.

Velocity deserves its own indicator: an outflow that doubles in pace over a fortnight is a different event from the same cumulative outflow spread over a quarter, and a framework that measures only cumulative amounts will treat them identically.

What the market notices first#

Funding spreads and market access give the earliest external read, because the market usually reprices a name before the name deteriorates. The indicators worth running are:

• the spread paid on new wholesale issuance relative to peers and to the bank’s own recent history
• the tenor actually achievable rather than the tenor sought
• bid participation and the proportion of an intended raise that clears
• secondary-market pricing of the bank’s own paper
• any drift in credit default swap levels where they exist

A widening spread paid to raise the same money is the cleanest signal available that counterparties are re-rating the risk, and it typically appears weeks before a ratio moves.

Three families of vulnerability#

Depositor and funding concentration, collateral capacity and foreign-currency liquidity cover the vulnerabilities that turn an ordinary outflow into a funding event. Concentration indicators track:

• the share of funding held by the largest depositors and counterparties
• dependence on a single product, sector, channel or intermediary
• the maturity clustering that creates refinancing walls

Collateral indicators track:

• unencumbered high-quality liquid assets by currency and location
• realistic monetisation capacity net of haircuts
• the encumbrance ratio and its trend
• pre-positioned collateral at central bank facilities
• the size and frequency of margin calls

Foreign-currency indicators matter because aggregate liquidity can be ample while the currency in which obligations fall due is not: currency-by-currency gaps, dependence on swap markets to convert, and the basis paid to do so all belong on the dashboard, since the assumption that convertibility persists under stress is precisely the assumption stress removes.

The ones that usually get left out#

Intraday behaviour, counterparty signals and qualitative indicators complete the set, and they are the families most often missing.

Intraday indicators — the timing of peak net debit position, reliance on incoming payments to fund outgoing ones, queued or delayed settlements, and the size of intraday credit used — reveal strain that end-of-day balances erase entirely, and payment behaviour is frequently the first place a counterparty notices something.

Counterparty signals include:

• shortened tenors offered
• reduced or withdrawn uncommitted lines
• tighter collateral terms
• the quiet non-renewal that nobody announces

Qualitative indicators belong on the same dashboard with the same escalation treatment: a negative rating outlook, adverse press or social-media attention, a delayed disclosure, a senior departure in finance or treasury, or a supervisory finding all move funding behaviour even though none of them is a number, and a framework that admits only numbers will register them only after they have already cost the bank funding.

When the alarm should sound#

Threshold calibration is where most frameworks fail, and the governing principle is that a threshold must fire while management still has options. Three tests make thresholds defensible:

1. Anchor them in the bank’s own distribution rather than in a peer benchmark: what does this indicator do in a normal quarter, and at what level has it historically sat outside its own range?
2. Calibrate against the bank’s stress results rather than against regulatory minimums — an amber level set at the point where the survival horizon would fall below appetite is informative, whereas an amber level set at the LCR minimum fires only when the option set has already narrowed to fire-sale and forbearance.
3. Prefer rate-of-change triggers alongside level triggers, since deterioration that is fast matters more than deterioration that is deep, and a level trigger alone will always be late.

Where an indicator has no usable history, state the judgement basis and the review date rather than inventing a number and letting it calcify.

Ownership and escalation#

An indicator without a named owner and a decision protocol is decoration. Each indicator needs four attributes recorded alongside its threshold:

• the individual accountable for producing and interpreting it
• the forum that receives a breach
• the timeframe within which that forum must convene
• the specific decision the breach puts on the table

Escalation levels should be few and meaningful — typically a monitoring level that increases reporting frequency and requires an explanation, an alert level that convenes the asset and liability committee and requires a management response with named actions and dates, and a stress level that triggers contingency funding plan activation and crisis governance.

What distinguishes a working framework from a compliant one is that each level changes what somebody does, not merely what somebody reports.

Links in one chain#

The connection to liquidity risk appetite and to the contingency funding plan is what makes the whole apparatus coherent. Risk appetite states the liquidity position the board is willing to run — a minimum survival horizon under defined stress, ceilings on funding concentration and wholesale dependence, limits on encumbrance and on currency mismatch.

Early warning thresholds should be derived from those statements rather than set independently of them, so that a breach is by construction a statement that the bank is approaching or has left its stated appetite.

The contingency funding plan sits at the far end of the same chain: its activation stages should map to the escalation levels, so that reaching the stress level does not require a fresh debate about whether the plan applies.

Where thresholds, appetite and plan activation are calibrated separately, the predictable result is a bank that breaches indicators for months without activating anything, because no document ever said the two were the same event.

Where it goes wrong#

Four failure modes recur often enough to be worth naming:

• Too many indicators is the most common: a dashboard of ninety metrics guarantees that several are always amber, which trains the committee to treat amber as normal and destroys the signal the framework was built to carry — fifteen to twenty-five indicators, each with a reason to exist, outperform ninety.
• Thresholds set at regulatory minimums is the second: it converts an early warning system into a late warning system, because by the time the LCR approaches its floor the useful options are gone.
• No escalation path is the third: indicators reported to a committee that has no defined response produce a paper trail showing the bank knew, which is worse than not measuring.
• Never back-testing is the fourth, and it is the one that quietly invalidates the rest: after any real episode of funding stress, whether at the bank or at a comparable institution, the framework should be replayed to ask which indicators moved first, which moved late, which never moved at all, and which thresholds would have fired in time.

Indicators that failed that test should be recalibrated or retired, and the review itself is evidence of effective challenge that supervisors look for and rarely find.