The Regulatory Floor and the Last Taxi Problem
This chapter focuses on the part of the LCR that a headline percentage hides: the distance between a bank's observed ratio and the regulatory minimum that governs its behavior. The practical goal is to distinguish a compliant level from a usable buffer, then explain why a bank may hesitate to spend that buffer even when a crisis makes liquidity valuable.
The distinction is easy to miss because compliance is usually reported as a ratio. A ratio above the minimum looks like spare capacity, but only the amount above the applicable floor is room that can be consumed before the bank approaches the boundary. The Federal Reserve evidence uses this distinction to interpret bank lending during stress. The BIS evidence uses a related but different concept—a pre-regulation LCR gap—to measure how strongly banks were exposed to the introduction of the rule. Treating these quantities as interchangeable would blur the mechanism each study is trying to identify.
A pass is not the same as room
Suppose the applicable minimum is 100%. A bank with an LCR of 120% has a 20-percentage-point buffer. A bank with an LCR of 105% has a 5-point buffer. Both comply, but they do not have the same distance to the floor. If a shock increases projected outflows or reduces the value recognized in the numerator, the second bank has much less room before the same rule becomes binding.
As a deliberately simplified illustration, imagine that the same shock consumes eight percentage points of regulatory room. The 120% bank would move to 112% and remain above the stated minimum; the 105% bank would move to 97% and cross below it. Real LCR dynamics are more complicated because outflows, inflows, HQLA recognition, and management responses change together. The illustration nevertheless captures the asymmetry that a headline “five points above the minimum” can conceal: the second bank has less capacity to absorb an additional shock without changing behavior, raising funding, or limiting commitments.
This is not a claim that the 120% bank can spend exactly 20 percentage points without consequence. The LCR is a modeled ratio, and its components can move as conditions change. A drawdown may increase cash outflows. A fall in market value or a change in asset eligibility may reduce recognized HQLA. A bank may also face capital, collateral, supervisory, or funding constraints that are not captured by the simple subtraction. The buffer is best understood as regulatory room under the relevant assumptions, not as a separate wallet of cash.
The Federal Reserve paper makes the idea measurable by classifying banks according to their pre-shock buffer. In its 2019Q4 baseline, the median LCR buffer was 20 percentage points above the minimum. Banks above that median had a mean buffer of 45.3%, while banks below it had a mean of 16.0%. These figures describe the study's sample-relative treatment groups; they are not universal prudential categories. Their value is comparative: they allow the researchers to ask whether banks with more room behaved differently when firms tried to draw credit.
The BIS paper asks a different question about the word “gap.” Its LCR gap is the required ratio minus the bank's average realized LCR before the regulation was introduced. A positive gap means that the bank was short of the required level before implementation; a larger positive gap means that the new rule was more binding for that bank. The gap measures pre-treatment exposure to regulatory change. It is not the same as a bank's current post-regulation buffer.
The requirement itself is part of that calculation. In the BIS sample, the required ratio was 100% for banks above the higher size threshold and foreign banks, and 70% for certain banks in the lower covered size range. The paper's gap therefore compares each bank with the standard that applied to its regulatory category; it is not a simple comparison of raw LCR percentages across all banks. This is another reason to keep the floor visible. A ratio has meaning only relative to the rule and the population used to interpret it.
That distinction matters for causal interpretation. In the Federal Reserve study, a larger buffer is an observed condition that may give a bank room to accommodate stress. In the BIS study, a larger pre-regulation gap identifies a bank that had more to adjust when the LCR was introduced. One variable is about room above a floor after the regulatory framework is in place; the other is about the intensity of the regulatory shock before it takes effect. A reader can compare their mechanisms, but should not combine their coefficients or treat “larger gap” as a synonym for “larger usable buffer.”
The same caution applies to the headline level. A high LCR may reflect a large numerator, a small modeled denominator, or both. It can be associated with resilience, but it does not tell us whether the bank has room relative to the requirement that applies to it. Nor does a sample's “high LCR” indicator necessarily identify the same banks as a “high buffer” indicator. The floor is the reference point that turns a percentage into a capacity judgment.
Why the last taxi may stay at the station
The Federal Reserve paper uses the “last taxi” problem to describe a behavioral constraint. Imagine a taxi at a station that could take a passenger home, but a local rule requires one taxi to remain available at the station at all times. The taxi exists, yet its final unit of capacity cannot be used without violating the purpose of the rule. In banking, a bank may hold qualifying liquid assets above the minimum but avoid drawing the buffer down because approaching the floor can invite regulatory scrutiny or signal weakness to creditors.
The mechanism has two sides. First, using the buffer reduces the margin that protects the bank against the next outflow. A bank that meets today's payment may be less prepared for tomorrow's withdrawal, collateral call, or rollover failure. Second, the act of using the buffer can be observable. If counterparties interpret a falling LCR as evidence of funding trouble, they may demand more collateral, shorten maturities, or withdraw funding. The bank then faces a difficult choice: deploy liquidity to meet an immediate obligation, or preserve the visible reserve that helps maintain future confidence.
This is why a buffer can be both insurance and a source of hesitation. Before a run, the existence of a reserve can reassure creditors that withdrawals can be met. During a run, using the reserve may tell those same creditors that the feared state has arrived. The rule's success can therefore depend on a paradox: the buffer may be most valuable when it is believed to be available, while its use can weaken the belief that it remains ample.
The policy tension is not resolved by saying that banks should always use the buffer or never use it. A permanently maintained reserve can help prevent a self-fulfilling run by making withdrawal less attractive. A usable reserve can help the bank honor credit commitments and avoid fire sales when a shock occurs. The first purpose is run prevention; the second is stress absorption. A rule designed for one can frustrate the other if banks believe that crossing below a preferred level will be treated as a failure, even when temporary use would improve the broader outcome.
There is also an allocation consequence. If management treats the floor as inviolable, it does not need to refuse every request to protect the buffer. It can tighten along less visible margins: approve fewer new commitments, direct a draw toward a stronger relationship, limit the size or timing of advances, or conserve liquidity for obligations with clearer regulatory treatment. That possibility matters because a bank may continue to report a compliant LCR while the supply of credit becomes more selective. Compliance can survive while the distribution of usable liquidity changes.
The empirical evidence helps separate this mechanism from a simpler story about bank size. If large or well-connected banks merely lend more during a crisis, a high LCR level might appear to predict support even when the regulatory floor has no special role. The Federal Reserve paper's horse-race specification addresses that possibility by including both LCR level and buffer interactions. Its buffer interaction remains positive and statistically significant, while the interaction using a high LCR level is effectively zero in the reported test. That pattern is consistent with the minimum acting as a binding floor rather than the level alone explaining the allocation of credit.
There is an important qualification. The same paper reports that a continuous LCR ratio enters significantly in a specification that also includes the buffer effect. These results are not identical tests. A continuous ratio can contain information about overall liquidity, while a binary high-level interaction asks a different question about whether being above a sample median predicts a differential response. The most defensible reading is not that the absolute LCR never matters. It is that, once the relevant buffer is considered, the evidence for the stress-related lending differential is concentrated in room above the applicable minimum. That distinction preserves the paper's result without turning it into a stronger universal claim.
The diagnosis also has to allow for banks' internal choices. The regulatory floor may be a formal minimum, while a bank's treasury maintains an internal target above it to absorb uncertainty, protect a rating, or avoid a visible deterioration in liquidity. The studies do not observe every internal target or every conversation with a supervisor. They show a pattern consistent with the floor constraining behavior, not a universal rule that management will always stop at the same percentage. A different crisis could generate a different response if central-bank facilities, deposit flows, collateral values, or supervisory communication change the cost of using the buffer.
That boundary is important for interpreting the word “binding.” It does not require the bank to be literally below the legal minimum before the constraint affects lending. A rule can bind economically when approaching it changes the bank's choices. The relevant evidence is therefore behavioral: do banks with more room accommodate more stress demand, does the effect appear when the shock is acute, and does the result survive a comparison that separates buffer from level? Those questions are more informative than simply ranking banks by their reported LCR.
The last-taxi mechanism also produces a testable prediction about timing. If buffer room matters because a bank needs to absorb an acute shock while protecting a regulatory floor, the effect should be strongest when funding demand surges and weaker after the immediate stress subsides. It should also appear in the allocation of credit among relationships exposed to the shock, not merely as a permanent difference between banks. The next evidence follows that prediction into the March 2020 rush to draw corporate credit lines.
The state path is not a claim that every bank follows one inevitable sequence. A bank may receive new deposits, raise secured funding, or reduce other outflows before reaching the near-floor state. Nor does the diagram prove that market signaling is the only reason a bank preserves a buffer. It identifies the mechanism that the Federal Reserve paper tests: when the minimum is treated as a hard boundary, excess room should matter more for stress lending than the headline level alone.
That mechanism changes how a reader should interpret a compliant ratio. The right follow-up is not simply “How high is it?” but “How high is it relative to the floor, how stable is that distance under the stress scenario, and what would using it communicate?” Those questions move the analysis from a static measure to a behavioral constraint. They also set up the bank-firm evidence: when companies rushed to draw committed credit in March 2020, the relevant issue was not only whether banks possessed HQLA, but which bank relationships had enough room and willingness to accommodate the demand.