The Last Taxi: LCR Buffers and Bank Liquidity Provision

 The sources examine the Liquidity Coverage Ratio (LCR) through the lens of the "last taxi" problem, exploring whether regulatory liquidity requirements actually enable banks to provide credit during periods of financial stress.

The Role and Mechanism of the LCR

Introduced as part of the Basel III reforms, the LCR requires large banks to maintain enough high-quality liquid assets (HQLA) to cover projected net cash outflows over a 30-day stress period. The formula is expressed as: $$LCR = \frac{High-Quality Liquid Assets (HQLA)}{Total Net Cash Outflows over 30 days}$$ The primary regulatory intent is to prevent disruptive bank runs by ensuring banks are prepared for short-term liquidity shocks.

The "Last Taxi" Problem

The sources highlight a fundamental tension in liquidity regulation: buffers designed for use during stress may become "frozen" when they are needed most. This is illustrated by the "last taxi" parable, where a weary traveler cannot take the only taxi at a station because local bylaws require one taxi to always be standing ready.

In a banking context, if banks treat the regulatory LCR minimum (e.g., 100%) as a hard floor rather than a usable reserve, the liquidity remains visible on balance sheets but effectively unavailable to support lending. Banks may avoid drawing down buffers to prevent signaling weakness to markets or inviting regulatory scrutiny.

LCR Buffers vs. LCR Levels

A critical finding in the sources is the distinction between a bank’s LCR level and its LCR buffer:

• LCR Level: The total ratio of HQLA to outflows (e.g., 120%).
• LCR Buffer: The amount of liquidity held above the regulatory minimum requirement (e.g., a 20 percentage point buffer for a bank with a 100% requirement and a 120% level).

Empirical analysis of the COVID-19 crisis in March 2020 revealed that only buffers, not overall LCR levels, matter for bank liquidity provision. A "horse race" specification showed that while buffers significantly predicted credit provision, overall LCR levels did not, confirming that the regulatory minimum operates as a binding constraint during stress.

Bank Liquidity Provision During Stress

During the acute phase of the COVID-19 pandemic, firms rushed to draw down committed credit lines as a precautionary measure. The sources document that banks with higher LCR buffers provided significantly more credit to these firms:

• Selective Insurance: This liquidity support was concentrated among "prime borrowers"—those with clean credit profiles and substantial undrawn capacity who were not restricted by financial covenants.
• Temporal Nature: The support was temporary and timely. High-buffer banks provided 10.5% more credit during the peak stress of March 2020, but this differential disappeared by mid-2020 as conditions normalized.
• Syndicate Coordination: Most large credit lines are syndicated, and the sources suggest that these arrangements act as co-insurance. Syndicates may implicitly or explicitly steer drawdowns toward member banks with stronger liquidity positions (higher buffers) during stress.

Ultimately, the sources conclude that the LCR provides selective liquidity insurance, but its effectiveness depends entirely on the existence of buffers above the regulatory floor, rather than the mandated minimum itself.

The "Last Taxi" Problem is a central metaphor in the sources used to describe a fundamental tension in liquidity regulation: liquidity buffers designed to be used during financial stress often become "frozen" exactly when they are needed most.

The Parable and its Banking Context

The problem is illustrated by a parable from Charles Goodhart about a weary traveler at a railway station who sees a taxi but is told by the driver that he cannot be taken because local bylaws require one taxi to always be standing ready at the station. In the banking system, the Liquidity Coverage Ratio (LCR) requires banks to hold high-quality liquid assets (HQLA) to cover projected outflows. However, the sources note that if banks treat the regulatory LCR minimum (typically 100%) as a hard floor they cannot breach, that "last taxi" of liquidity remains visible on the balance sheet but effectively unavailable to support lending during a crisis.

LCR Levels vs. LCR Buffers

A critical distinction made in the sources to address this problem is the difference between a bank's LCR level and its LCR buffer.

• LCR Level: The total ratio of assets to projected outflows (e.g., 120%).
• LCR Buffer: The specific amount of liquidity held above the regulatory minimum (e.g., a 20 percentage point buffer for a bank with a 100% requirement and a 120% level).

The sources argue that the "last taxi" constraint means only banks with substantial buffers above the floor can provide meaningful liquidity insurance. This was confirmed by a "horse race" specification where researchers included both levels and buffers; they found that only buffers matter for predicting bank credit provision during stress, while overall LCR levels do not.

Bank Liquidity Provision and Selective Insurance

The sources examine how these buffers functioned during the "dash for cash" in March 2020 at the onset of the COVID-19 pandemic. Their findings include:

• Buffer-Driven Lending: Banks with high LCR buffers provided 10.5% more credit to firms with large undrawn credit lines during the acute phase of the crisis (March 2020).
• Selective Insurance: This liquidity provision was not universal; it was concentrated among "prime borrowers"—high-quality firms with clean credit profiles and no binding financial covenants.
• Rationing for Others: In contrast, firms that were covenant-constrained or had minimal undrawn capacity received no additional support from high-buffer banks, suggesting that banks use their limited usable liquidity to protect their most creditworthy relationships.
• Temporal Nature: The insurance effect was temporary and timely, disappearing by mid-2020 as market conditions normalized.

The Role of Syndicate Coordination

The sources suggest that the mechanism for this liquidity allocation often occurs through bank syndicates. These syndicates act as co-insurance arrangements where lead arrangers may explicitly or implicitly steer borrower drawdowns toward member banks that have more "room" above the regulatory floor (the LCR buffer).

Regulatory Implications

The "Last Taxi" Problem persists because banks face strong disincentives to draw down their mandated buffers, fearing that doing so might signal weakness to markets or invite increased regulatory scrutiny. Consequently, the sources conclude that the LCR as currently implemented functions more as a run-prevention mechanism (by keeping liquidity frozen) rather than a run-absorption mechanism (by allowing it to be deployed during stress).

The sources emphasize that the distinction between a bank's LCR level and its LCR buffer is the critical factor in determining whether a bank can provide liquidity to borrowers during financial stress. While both terms relate to the Liquidity Coverage Ratio (LCR), they represent different economic constraints under the "last taxi" framework.

Defining the Distinction

• LCR Level: This is the total ratio of High-Quality Liquid Assets (HQLA) to projected net cash outflows over a 30-day stress period. For example, a bank might have an LCR level of 120%.
• LCR Buffer: This is the amount of liquidity a bank holds specifically above its regulatory minimum requirement. If a bank with a 120% LCR level faces a 100% regulatory minimum, its buffer is 20 percentage points.

The "Last Taxi" Mechanism

The distinction is vital because of the "last taxi" problem: banks face strong disincentives to draw down their mandated liquidity minimums. Doing so might signal financial weakness to markets or invite increased regulatory scrutiny. Consequently, if a bank treats the regulatory minimum (e.g., 100%) as a hard floor that cannot be breached, the liquidity at that level remains visible on the balance sheet but is effectively "frozen" and unavailable for lending.

In this context, only the buffer—the "room to absorb stress without falling below the requirement"—provides the usable capacity for banks to act as a "lender of first resort" during a crisis.

Empirical Evidence: The "Horse Race"

To test which of these two measures actually drives lending, the authors conducted a "horse race" specification that simultaneously included interactions for both LCR levels and LCR buffers. Their findings were decisive:

• Buffers Matter: The interaction with the LCR buffer was statistically significant, indicating that banks with higher buffers above their specific regulatory minimums provided significantly more credit during the acute stress of March 2020.
• Levels Do Not: The interaction with the overall LCR level was effectively zero and statistically insignificant.

This result provides direct evidence that the regulatory minimum operates as a binding constraint, and overall liquidity levels are misleading indicators of a bank's ability to support the economy during a shock.

Impact on Bank Liquidity Provision

The sources conclude that LCR buffers enable selective, temporary liquidity insurance. During the COVID-19 "dash for cash," banks with high LCR buffers (above the 20% median) provided 10.5% more credit to firms with large undrawn lines compared to banks closer to the regulatory floor.

However, this provision was not universal; it was concentrated among "prime borrowers"—high-quality firms with clean credit profiles and no binding financial covenants. For these borrowers, high-buffer banks provided 16.6% more credit, while they simultaneously rationed credit to more constrained borrowers. This demonstrates that while the buffer allows for liquidity provision, banks use that limited "usable" liquidity to protect their most creditworthy relationships.

During the COVID-19 crisis of March 2020, the Liquidity Coverage Ratio (LCR) functioned as a critical safety mechanism, but the sources reveal that its effectiveness in providing liquidity to the economy was entirely dependent on buffers held above regulatory minimums rather than the mandated levels themselves.

The COVID-19 "Dash for Cash"

The onset of the pandemic created a "natural experiment" for testing the LCR. As uncertainty peaked in March 2020, corporate firms engaged in a "dash for cash," rushing to draw down committed bank credit lines as a precautionary liquidity measure. This simultaneous pressure across the banking system tested whether the LCR could fulfill its dual role as a safety mechanism: preventing bank runs while enabling banks to act as "lenders of first resort".

Buffers as the Active Safety Component

The sources' most significant finding regarding the COVID-19 period is that overall LCR levels were irrelevant to credit provision; only the buffer over the regulatory floor mattered.

• The "Horse Race": Empirical testing—a "horse race" specification—showed that banks with higher buffers above their specific requirement (e.g., 100%, 85%, or 70%) provided significantly more credit, while banks with high total ratios but small buffers did not.
• Quantifiable Support: Banks with LCR buffers above the 20% median provided 10.5% more credit to firms with large undrawn lines during the acute phase of the crisis.

Selective vs. Universal Insurance

While the LCR acted as a safety mechanism, it did so selectively. The liquidity insurance provided by high-buffer banks was concentrated among "prime borrowers"—firms with clean credit profiles and no binding financial covenants.

• Prime borrowers received 16.6% more credit from high-buffer banks.
• In contrast, covenant-constrained firms received no additional support, suggesting that banks used their usable liquidity (the buffer) to protect their most creditworthy established relationships while rationing credit to others.

The "Last Taxi" Paradox

The sources highlight a paradox in how the LCR functions as a safety mechanism. Ideally, a buffer should be "run-absorbing" (deployable during stress). However, the "last taxi" problem suggests that because banks fear the signaling effect of falling below regulatory minimums, they treat these minimums as a "hard floor".

• Run Prevention: Consequently, the LCR is highly effective at run prevention by keeping liquidity "frozen" and visible on the balance sheet to reassure depositors.
• Run Absorption: It is less effective at run absorption because only the extra liquidity held voluntarily by banks (the buffer) is actually available to be lent out during a crisis.

Temporal Specificity

The role of LCR buffers as a safety mechanism was timely and temporary. The differential in credit provision from high-buffer banks was prominent in March 2020 but disappeared by the second quarter of 2020 as market conditions normalized and firms began paying back their precautionary drawdowns. This confirms that LCR buffers provide a surge capacity for acute stress rather than a permanent shift in lending behavior.

The sources characterize the role of banks during financial stress not as universal providers of credit, but as providers of selective liquidity insurance. This insurance is "selective" because banks with high Liquidity Coverage Ratio (LCR) buffers prioritize their most creditworthy relationships while rationing liquidity for others.

Beneficiaries: Prime Borrowers

The primary beneficiaries of this selective insurance are "prime borrowers," defined as firms with substantial undrawn credit capacity and no binding financial covenant restrictions.

• Increased Credit Access: During the acute phase of the COVID-19 crisis (March 2020), banks with high LCR buffers provided 16.6% more credit to these prime borrowers compared to banks with low buffers.
• Relationship Protection: This pattern suggests that banks use their limited "usable" liquidity—the amount held above the regulatory floor—to protect established and high-quality credit relationships.

The Rationing Effect: Non-Prime Borrowers

In contrast, firms that did not meet the "prime" criteria did not receive the same support from high-buffer banks:

• Covenant-Constrained Firms: These firms have high contractual undrawn lines but limited actual capacity due to financial covenants (like debt-to-EBITDA ratios). High-buffer banks provided no additional support to these firms, and in some cases, they received less credit than they did from low-buffer banks.
• Marginal Borrowers: Firms near median undrawn thresholds without covenant issues also saw no significant increase in credit from high-buffer institutions.

The Context of LCR Buffers and the "Last Taxi"

This selectivity is a direct consequence of the "last taxi" problem. Because banks treat the regulatory LCR minimum as a "hard floor" to avoid signaling weakness or inviting regulatory scrutiny, only the buffer held above that floor is actually available to support lending.

• Buffer vs. Level: Empirical "horse race" testing confirmed that only a bank's buffer, not its total LCR level, determines its ability to provide this insurance.
• Limited Capacity: Because the usable buffer is a finite resource, banks must allocate it strategically, leading to the selective insurance of prime borrowers over more constrained firms.

Mechanism and Duration

• Syndicate Coordination: This selective allocation often occurs through bank syndicates, where lead arrangers may implicitly or explicitly steer borrower drawdowns toward member banks with the strongest liquidity positions (highest buffers).
• Temporary Support: This selective insurance was timely and temporary. The increased credit provision from high-buffer banks was concentrated in March 2020 and disappeared by mid-2020 as market conditions stabilized and firms began paying back their precautionary drawdowns.

The sources identify syndicate coordination as the primary allocation mechanism that allows firms with multiple bank relationships to access liquidity from banks with higher buffers during periods of financial stress. This mechanism explains why, during the COVID-19 "dash for cash," credit drawdowns were not distributed evenly but were instead shifted toward lenders with greater regulatory "room" above their Liquidity Coverage Ratio (LCR) floors.

Syndicates as Co-Insurance Arrangements

Most large corporate credit lines are structured as syndicated arrangements where multiple banks share exposure to a single borrower. These syndicates function as active co-insurance arrangements rather than passive credit allocation tools. When a borrower needs to draw on its credit line, the syndicate structure creates a network that pools and redistributes liquidity shocks across member banks.

The Steering Role of Lead Arrangers

The sources highlight that lead arrangers, who manage the syndicate and coordinate interactions between the borrower and the participating banks, play a crucial role in this allocation. During stress periods, lead arrangers may:

• Explicitly or implicitly steer drawdowns toward syndicate members that have stronger liquidity positions (higher LCR buffers).
• Utilize contractual terms within the syndicated agreement that specify allocation rules accounting for individual bank capacity constraints.

Assortative Matching and Sorting

The allocation mechanism also operates through assortative matching in the credit market. Since the implementation of the LCR, firms that rely heavily on credit lines have increasingly sorted themselves toward syndicates composed of higher-liquidity banks. This indicates that both borrowers and lead arrangers internalize the quality of a bank's liquidity—specifically its buffer above the regulatory minimum—when structuring credit relationships.

Empirical Identification of the Mechanism

The research uses Firm×Time fixed effects to isolate this supply-side allocation. By holding constant the total amount a firm wishes to draw across all its banks in a given quarter, the authors identify that firm-level demand is disproportionately accommodated by high-buffer banks.

In the larger context of bank liquidity provision, this mechanism reveals that LCR buffers do not just provide a general safety net; they enable a supply-driven reallocation of credit within existing relationships. Banks with stronger buffers are more willing to accommodate these drawdowns, while banks closer to the regulatory floor (the "last taxi" constraint) may tighten non-price margins to preserve their required reserves.

The sources outline several significant policy and research implications regarding the Liquidity Coverage Ratio (LCR), primarily focusing on the trade-off between bank safety and the provision of credit during economic shocks.

Policy Implications: Run Prevention vs. Run Absorption

The most fundamental policy debate identified in the sources is whether liquidity regulation should be designed to prevent runs or absorb them.

• The Paradox of Frozen Liquidity: Theoretical research suggests that to eliminate the incentive for depositors to run, banks must maintain a permanently "frozen" reserve. Like the "last taxi" at a station, the mere existence of this reserve—which is never intended to be used—can prevent a self-fulfilling panic.
• The Signaling Problem: A major policy challenge is that current regulations do not account for the negative signaling associated with drawing down mandated buffers. Bank treasurers report that they avoid using mandated liquidity because doing so invites increased regulatory scrutiny and signals weakness to the market.
• Distortionary Ratios: The sources point to research suggesting that quantity-based ratios like the LCR may be more distortionary and pro-cyclical than alternative mechanisms, such as Pigovian taxes on short-term liabilities. Buffers may inadvertently be least binding exactly when excess credit incentives are strongest.

Research Implications: Measuring Regulatory Effectiveness

The sources offer several contributions to the empirical study of banking and liquidity:

• Decisive Evidence for the "Last Taxi" Constraint: The research provides a methodology to prove that the LCR operates as a binding constraint during stress. By using a "horse race" specification, the authors demonstrate that only buffers above the regulatory minimum matter for credit provision, while total LCR levels do not.
• Shift to Selective Insurance: The findings shift the research focus from the "extensive margin" (whether a bank lends at all) to the allocation of utilization across existing relationships. This reveals that liquidity insurance is not universal but is a selective mechanism that prioritizes "prime borrowers" with clean credit profiles and no binding covenants.
• Syndicates as Active Networks: The sources highlight that loan syndicates act as active co-insurance arrangements rather than passive allocation tools. This implies that future research should view the banking system as a network where lead arrangers coordinate the redistribution of liquidity shocks toward banks with the most "room" (highest buffers).
• Market Sorting: There is evidence of assortative matching in the post-LCR era, where firms that rely heavily on credit lines are increasingly sorting themselves into syndicates composed of high-liquidity banks.

Conclusion of Implications

Ultimately, the sources suggest that for the LCR to function as a "lender of first resort" mechanism, policy must address the regulatory and market incentives that prevent banks from using their mandated reserves. Without addressing the "last taxi" problem, the LCR remains highly effective at preventing runs by keeping liquidity frozen, but it fails to support broad-based credit provision during a crisis unless banks voluntarily maintain substantial buffers above the regulatory floor.