6 The Deposit-Lending Synergy

Loan Commitments, Natural Hedges, and Crisis Limits

6.1 Overview

The previous chapter asked why banks monitor: why is it efficient to have an intermediary sit between small savers and opaque borrowers? The answer — that a monitored fixed-rate deposit contract minimizes delegation costs — explains the asset side of the balance sheet. But it takes the liability structure largely as given. This chapter asks a different question: why do deposit-taking and lending coexist under one roof?

A finance company can make loans. A money market mutual fund can issue liquid claims. Why should the same institution do both? The question matters because the answer determines whether the bundling is efficient — a genuine synergy — or merely a regulatory artifact that policy should dismantle.

The answer developed across four papers over thirteen years. Kashyap, Rajan, and Stein (2002) identify the synergy: deposits and loan commitments (credit lines) are both forms of liquidity on demand, and combining them lets a bank share the costly buffer of liquid assets required to honor each. Gatev and Strahan (2006) sharpen the mechanism: bank deposit inflows spike precisely when firms want to draw their credit lines, creating a natural hedge that no other intermediary can replicate. Gatev, Schuermann, and Strahan (2009) test the synergy directly in the data, confirming that transactions deposits insulate banks from the risk of liquidity exposure. Acharya and Mora (2015) then confront the limits: during the 2007–2009 financial crisis, the mechanism broke down, because the banking system itself was the source of the distress rather than a bystander to it.

6.1.1 Learning Objectives

By the end of this chapter, you should be able to:

Explain why deposit-taking and loan commitments are two manifestations of the same underlying function — liquidity on demand — and why combining them is efficient.
Solve the KRS model and derive the complementarity between transactions deposits and loan commitments.
Explain the natural hedge hypothesis of Gatev and Strahan (2006): why bank funding costs fall precisely when loan demand rises during market crises.
Describe the GSS (2009) empirical design and interpret what bank stock-return volatility reveals about the deposit-lending synergy.
Identify the conditions under which the bank liquidity provision mechanism breaks down, and explain why government support was essential during 2007–2009.

6.1.2 Roadmap

Part 1: The Puzzle — Why Do Banks Combine Deposits and Credit Lines?

The question Diamond (1984) leaves open
Loan commitments as liquidity on demand
The KRS (2002) model: setup, solution, synergy result, and empirical predictions

Part 2: Banks’ Comparative Advantage in Liquidity Insurance

The CP market and backup lines of credit
The natural hedge: why deposit inflows arrive when loan demand surges
Gatev and Strahan’s (2006) model and evidence
The role of the government safety net

Part 3: Testing the Synergy — Commitment Exposure and Bank Risk

From theory to measurement: bank stock-return volatility
The GSS (2009) finding: risk and transactions deposits
When does the hedge strengthen? Market conditions and the CP spread

Part 4: Crisis and the Limits of Liquidity Provision

The standard story and the 2007–2009 anomaly
What happened: deposit inflows stalled while drawdowns surged
Why the mechanism broke down: banking system as crisis epicenter
Government support as the true backstop
Policy implications

6.2 Part 1: The Puzzle — Why Do Banks Combine Deposits and Credit Lines?

6.2.1 A Question Diamond (1984) Leaves Open

Chapter 5 established Diamond’s (1984) result that a bank holding a diversified loan portfolio can monitor borrowers on behalf of depositors at negligible delegation cost. The model endogenizes both sides of the balance sheet: monitored loans on the asset side and fixed-rate deposits on the liability side. The deposit structure is not arbitrary — it is the contract that disciplines the bank’s monitoring effort through the threat of costly liquidation.

But Diamond’s model says nothing about loan commitments (also called credit lines or lines of credit). A commitment is a promise to lend in the future at a prearranged rate: once signed, it behaves like a demand deposit from the bank’s perspective — the borrower can draw funds at will, and the bank must provide them. If we take Diamond’s model literally, there is no reason the committed lender needs to fund itself with deposits. A finance company that raises funds by issuing short-term commercial paper could just as well extend credit lines to firms.

Yet the empirical pattern is striking: banks that take more transactions deposits also make more loan commitments. Kashyap, Rajan, and Stein (2002) document a robust positive correlation between these two activities across thousands of U.S. commercial banks. This is not an accident of regulation. It is, they argue, a genuine efficiency: the same institution should do both because both require the same expensive input.

6.2.2 Loan Commitments as Liquidity on Demand

Definition: Loan Commitment (Credit Line)

A loan commitment (or credit line) is a contractual promise by a bank to lend a specified amount to a borrower on demand over a specified period, at a prearranged spread over a benchmark rate (e.g., LIBOR). From the bank’s perspective, an undrawn commitment is an off-balance-sheet contingent liability: no cash changes hands at signing, but the bank must fund the loan if the borrower chooses to draw it down.

The critical observation is that a loan commitment, once made, behaves just like a demand deposit. Both require the bank to provide cash on demand with little advance notice. Both subject the bank to random outflows — deposit withdrawals and commitment takedowns — that cannot be perfectly predicted in advance.

Because honoring random demand requires holding a buffer of liquid assets (cash and short-term securities), an institution that provides either service must pay the cost of that buffer. This cost is significant: liquid assets earn below-market returns (the opportunity cost) and, in a corporate form, may generate additional agency and tax costs.

The key insight is that if a single institution provides both services, it can share the buffer between them. The buffer is only depleted when outflows actually occur, not when they are merely possible. If deposit withdrawals and commitment takedowns do not always happen simultaneously, the bank needs only enough liquidity to cover the worst realized combination — not the sum of worst-case withdrawals from each side independently.

This is the diversification synergy: combining liquidity provision on both sides of the balance sheet reduces the total buffer required, lowering average costs and making the institution a more efficient provider of each service.

6.2.3 The KRS Model

6.2.3.1 Setup and Environment

Kashyap, Rajan, and Stein (2002) formalize this intuition in a partial-equilibrium model of a single bank operating over three dates.

Assets. At date 0, the bank makes term loans $L$ that mature at date 2. Loan revenue is $r(L)$, where we assume $r'(L) > 0$ and $r''(L) < 0$ (concave revenue reflecting market power or increasing loan-screening costs). The bank also holds a buffer stock of liquid assets $S_0$ — cash, short-term Treasuries, or central-bank reserves. These earn the market interest rate $i$ per period, but holding them costs $\tau$ per dollar per period (an opportunity cost reflecting foregone returns on higher-yielding illiquid assets, plus potential agency costs of financial slack).

Liabilities. Demand deposits $D_0$ are exogenously determined by the bank’s deposit franchise — its branch network, customer relationships, and local market position. Demand deposits pay no explicit interest (the franchise value is the deposit insurance subsidy and below-market funding cost). The bank can also raise external finance $e_0$ at date 0 at the risk-free rate $2i$ (over the two periods), and $e_1$ at date 1 if needed. External finance at date 1 is costly because of capital market frictions: the total cost is $ie_1 + \frac{\alpha}{2}e_1^2$, where $\alpha > 0$ measures the severity of the frictions (adverse selection in equity markets, as in Myers and Majluf 1984).

Why Is External Finance at Date 1 Costly?

At date 1, the bank may have private information about the quality of its loan portfolio. Outside investors, not knowing whether the bank is in good or bad shape, will demand a premium on any new funding. This adverse-selection cost is increasing in the amount raised: the quadratic cost $\frac{\alpha}{2}e_1^2$ captures the idea that raising more funds is progressively more expensive. The bank can avoid this cost only if it holds enough liquid assets to self-fund any shocks internally.

This is the key friction that makes the liquid-asset buffer valuable: holding $S_0$ now is costly at rate $\tau$, but it saves the expected cost of costly external finance at date 1.

Commitments. At date 0, the bank also sells loan commitments of face value $C$, earning fee revenue $f(C)$ (concave: $f' > 0$, $f'' < 0$). With probability $\frac{1}{2}$, a fraction $z = 1$ of commitments are drawn at date 1; otherwise, $z = 0$. Borrowers who draw their commitments pay interest $i$ on the balance outstanding.

The correlation structure. At date 1, a fraction $\omega$ of depositors withdraw their funds: $\omega = 1$ (all withdraw) with probability $\frac{1}{2}$, $\omega = 0$ otherwise. Commitment takedowns ($z$) and deposit withdrawals ($\omega$) are correlated. Define:

\[\rho \;\equiv\; P(\omega = 1 \mid z = 1)\]

This is the probability that deposits are withdrawn given commitments are drawn down.

Interpreting the Correlation Parameter $\rho$

Value of $\rho$	Meaning
$\rho = 1$	Perfectly positively correlated — when firms draw credit lines, depositors also flee. Both shocks hit simultaneously.
$\rho = \frac{1}{2}$	Independent — commitment takedowns tell you nothing about deposit withdrawals.
$\rho = 0$	Perfectly negatively correlated — when firms draw credit lines, depositors arrive. Ideal natural hedge.

The central hypothesis of KRS is that $\rho < 1$ in practice. Empirical evidence (reviewed below) suggests $\rho$ is often close to 0 or even negative.

6.2.3.2 The Bank’s Optimization Problem

The bank chooses loans $L$, liquid assets $S_0$, and commitments $C$ to maximize expected net income denominated in date-2 dollars:

\[\max_{L,\, C,\, S_0} \;\; E\!\left[rL + fC + izC + (i - \tau)S_0 + iS_1 - 2ie_0 - ie_1 - \frac{\alpha}{2}e_1^2\right] \tag{1}\]

subject to the balance-sheet constraints at each date:

\[L + S_0 = D_0 + e_0 \tag{2}\]

\[L + S_1 + zC = D_0(1 - \omega) + e_0 + e_1 \tag{3}\]

\[S_1 \geq 0 \tag{4}\]

Constraint (2) is the balance sheet at date 0: assets equal liabilities. Constraint (3) is the balance sheet at date 1, after shocks are realized. Constraint (4) says the bank cannot sell liquid assets it doesn’t have — it cannot short-sell its buffer.

Since raising external finance at date 1 is costly, the bank exhausts its liquid assets before going to the market. Combining (2)–(4), the amount of external finance the bank must raise at date 1 is:

\[e_1 = \max\!\left[zC + \omega D_0 - S_0,\; 0\right] \tag{5}\]

Interpretation: if commitment takedowns $zC$ plus deposit withdrawals $\omega D_0$ exceed the buffer $S_0$, the bank must raise the shortfall externally.

6.2.3.3 First-Order Conditions

The first-order conditions with respect to $L$, $C$, and $S_0$ are:

\[\text{(Loans):} \quad r + L r_L = 2i \tag{6}\]

\[\text{(Commitments):} \quad f + C f_C = \frac{\alpha}{2}\, \frac{dE(e_1^2)}{dC} \tag{7}\]

\[\text{(Liquid assets):} \quad \tau = -\frac{\alpha}{2}\, \frac{dE(e_1^2)}{dS_0} \tag{8}\]

Equation (6): The marginal revenue on loans equals the opportunity cost of long-term funds. This pins down $L^*$ independently of deposits and commitments — the loan portfolio is separated from the liquidity management problem.

Equation (7): The bank expands its commitment portfolio until the marginal fee revenue ($f + Cf_C$) equals the marginal expected cost of external finance induced by commitments. A larger $C$ means more potential drawdowns, requiring more external finance when $z = 1$.

Equation (8): The bank holds liquid assets until the marginal cost of the buffer ($\tau$) equals the marginal savings on external finance. Holding more $S_0$ reduces the amount the bank needs to raise at date 1 when shocks hit.

6.2.3.4 Solving the Model: The Synergy Result

Combining constraints (2)–(4) and the expression for $e_1$, we can compute $E(e_1^2)$ in the region of the parameter space where $S_0^* \leq \min(C^*, D_0)$ — that is, the buffer is insufficient to fully absorb any single shock. This is the most interesting region, where both deposits and commitments create genuine liquidity risk.

In this region, using the joint distribution of $(\omega, z)$:

\[E(e_1^2) = \frac{\rho}{2}\,(C + D_0 - S_0)^2 + \frac{1-\rho}{2}\!\left[(D_0 - S_0)^2 + (C - S_0)^2\right] \tag{9}\]

The first term is the contribution from the state where both shocks hit ($\omega = z = 1$), which occurs with probability $\rho/2$; the second term covers the states where only one shock hits ($z=1, \omega=0$ or $z=0, \omega=1$), each with probability $(1-\rho)/2$.

Applying the first-order condition (8) and simplifying:

\[S_0^* \;=\; \frac{C^* + D_0 - 2(\tau/\alpha)}{2 - \rho} \tag{10}\]

Proposition 1 (KRS): The Synergy Result

The optimal liquid-asset buffer $S_0^*$ satisfies:

\[\boxed{S_0^* = \frac{C^* + D_0 - 2(\tau/\alpha)}{2 - \rho}}\]

For any given levels of commitments $C^*$ and deposits $D_0$:

When $\rho = 1$ (perfectly correlated shocks): $S_0^* = C^* + D_0 - 2\tau/\alpha$. The bank needs a large buffer to cover the worst case of both shocks hitting simultaneously.
When $\rho < 1$ (imperfectly correlated): $S_0^* < C^* + D_0 - 2\tau/\alpha$. The bank needs a smaller buffer because both shocks rarely hit at the same time.
When $\rho = 0$ (independent shocks): $S_0^* = \frac{C^* + D_0 - 2\tau/\alpha}{2}$, exactly half as large as in the perfectly correlated case.

The synergy is the cost saving from sharing the buffer: the combined institution needs less total liquidity than two separate institutions, each holding its own buffer.

Economic interpretation. Suppose two separate firms — a finance company that only does commitments, and a mutual fund that only holds deposits — each must hold a buffer to cover their respective liquidity demands. The finance company holds $S_C$ to cover potential takedowns; the fund holds $S_D$ to cover potential withdrawals. Their combined buffer is $S_C + S_D$.

A single bank that does both needs a buffer of size $S_0^* < S_C + S_D$ whenever $\rho < 1$, because the shocks don’t always coincide. The saved resources — $(S_C + S_D) - S_0^*$ — can be deployed in higher-yielding loans. This is the efficiency gain from bundling deposit-taking with lending.

6.2.3.5 The Complementarity: Deposits Drive Commitments

The first-order condition for commitments (equation 7) can be combined with (9) and (10) to yield:

\[f + C^* f_C \;=\; \frac{\alpha}{2}\!\left[\rho D_0 + C^* - S_0^*\right] \tag{11}\]

This equation equates the marginal fee revenue on commitments to their marginal expected financing cost. The key observation: the larger the deposit base $D_0$, the more commitments a bank optimally offers — as long as $\rho < 1$.

Intuition: Why Do Deposits Enable More Commitments?

Suppose the bank has more deposits ($D_0$ rises). The right-hand side of (11) rises — but so does $S_0^*$, the optimal buffer. The net effect: the effective marginal cost of commitments is reduced. Why?

Because deposits provide a partial cushion against commitment takedowns. When $z = 1$ but $\omega = 0$ (firms draw their lines, but depositors don’t flee), the bank can fund the drawdown out of deposits that happen to stay. A larger deposit base makes this scenario more valuable: there are more “idle” deposits that can cross-subsidize commitment takedowns.

Formally, the cross-derivative: \[\frac{\partial^2 E(e_1^2)}{\partial C \,\partial D_0} = \rho > 0\] but the net complementarity arises because $S_0^*$ also adjusts, and the net effect of $D_0$ on the marginal cost of $C$ is negative when $\rho < 1$. A larger deposit base reduces the marginal liquidity cost of the commitment portfolio.

6.2.3.6 Empirical Predictions

The model generates two clear testable implications:

Cross-sectional prediction: Across banks, those with higher ratios of transactions deposits to total deposits should also have higher ratios of loan commitments to loans. The deposit franchise and the commitment business are complements, not substitutes.
Institutional prediction: Banks should do more commitment-based lending than other intermediaries (finance companies, insurance companies) that cannot fund themselves with demandable deposits.

KRS test both predictions using Call Report data for U.S. commercial banks (1992–1996) and Find strong support for both:

Within the banking sector, the correlation between transactions deposits and unused loan commitments is consistently positive across size groups and time periods.
Banks do substantially more commitment-based lending (especially unsecured credit lines) than competing intermediaries such as finance companies.

The evidence is inconsistent with the view that deposit insurance artificially inflates both activities. The synergy appears to be real.

6.3 Part 2: Banks’ Comparative Advantage in Liquidity Insurance

6.3.1 The CP Market and the Demand for Backup Liquidity

The Kashyap-Rajan-Stein framework explains why deposits and credit lines belong together. Gatev and Strahan (2006) extend and sharpen this by focusing on who can most cheaply provide the commitment service — and why that institution is the commercial bank.

The setting is the commercial paper (CP) market, where large, well-rated firms borrow at short maturities (typically 30–90 days) by issuing unsecured notes directly to institutional investors. CP is cheap during normal times: yields are close to Treasury rates, and firms can roll over their paper smoothly as it matures. But the CP market is systemically fragile: when investors lose confidence in any corner of the market, they may refuse to roll over paper across the board, denying credit to firms that have done nothing wrong.

Historical Episodes of CP Market Disruption

1970 — Penn Central: Penn Central Transportation filed for bankruptcy with $82 million in CP outstanding. Investors panicked, refusing to roll paper for other large issuers. The Federal Reserve responded by lending aggressively through the discount window and encouraging banks to extend credit to their large borrowers. This episode established the modern practice of CP issuers maintaining bank backup lines — committed credit lines that allow the firm to borrow from its bank if the CP market closes.

1998 — LTCM and Russian default: The Russian sovereign default and the near- collapse of Long-Term Capital Management triggered a flight to quality. Investors moved funds from CP and corporate bonds into bank deposits and government securities. Firms that relied on CP for short-term funding drew heavily on their bank credit lines.

2002 — Enron and Accounting Scandals: Doubts about financial-statement accuracy caused investors to shun the CP of firms perceived as opaque. The Wall Street Journal reported: “the commercial-paper market has served as the corporate world’s automated teller machine, spitting out a seemingly endless supply of cash… But now, amid financial jitters… that machine is sputtering.”

2007–2009 — ABCP Freeze and Subprime Crisis: The ABCP (asset-backed commercial paper) market froze in August 2007. Unlike previous episodes, however, this time the banking system itself was under stress — a point we return to in Part 4.

Since 1970, virtually all CP issuers have maintained backup lines of credit from commercial banks. These lines allow the firm to “take down” the commitment — to borrow from the bank at a predetermined spread — if market liquidity evaporates and the firm cannot roll its paper. The backup line is essentially insurance against a market-wide liquidity shock.

Why do commercial banks, rather than finance companies or insurance companies, dominate this market? Gatev and Strahan argue the answer lies in a natural hedge that is unique to deposit-funded institutions.

6.3.2 The Natural Hedge

The core argument is elegant. During normal times, CP is cheap, firms issue paper, and funds flow from investors to firms. Banks are relatively quiet: they hold their backup lines undrawn. During periods of market stress — when the CP market tightens — two things happen simultaneously:

Firms want to draw their credit lines — they cannot roll their paper and need bank funding.
Investors seek safe havens — they pull funds from CP, money market instruments, and other risky assets, and deposit them with banks.

These two flows offset each other on the bank’s balance sheet. The increase in loan demand (from drawdowns) is funded by the increase in deposit supply (from flight to safety). The bank can honor its commitments without depleting its liquid assets or going to the capital market.

The Natural Hedge (Gatev-Strahan)

Define: - $\Delta L_t$ = change in bank loan volume at time $t$ (positive when commitments are drawn) - $\Delta D_t$ = change in bank deposit volume at time $t$ (positive when investors deposit funds) - $\text{Spread}_t$ = CP rate minus Treasury bill rate (a measure of CP market stress)

The natural hedge requires: \[\text{Corr}(\Delta L_t,\; \Delta D_t \mid \text{high Spread}_t) < 0\]

When spreads are wide (market stress), deposit inflows arrive just as drawdown demand rises. The two flows are negatively correlated conditionally on market conditions. This is precisely the empirical finding of Gatev and Strahan.

Why can’t a finance company replicate this hedge? A finance company funds itself in the same short-term markets (CP, short-term notes) that are stressed during a crisis. When investors flee CP, the finance company’s cost of funds rises precisely when its commitment customers want to draw. There is no natural hedge — both sides of the financing problem deteriorate together.

The bank’s advantage is that deposit funding is counter-cyclical with respect to market liquidity: it gets cheaper precisely when the bank needs it most. This is not merely luck — it reflects the institutional feature that bank deposits are perceived as safe by investors during periods of market stress. Government guarantees (deposit insurance, implicit bailout expectations, the Fed backstop) are the mechanism that makes bank deposits a safe haven.

6.3.3 A Simple Model of Competitive Advantage

Gatev and Strahan formalize this intuition. Consider a competitive market for backup lines. A firm needs a CP backup line and pays an annual fee $s$ per dollar of commitment. The expected cost to the intermediary of providing the backup line depends on the likelihood of being drawn ($p$) and the cost of funds at the time of the drawdown.

Suppose market liquidity follows a state variable $\ell \in \{H, L\}$ (high or low). In state $L$ (low liquidity, wide CP spreads), the backup line is drawn with probability 1; in state $H$, it is not drawn. The intermediary’s expected cost per dollar of commitment is:

\[\text{Expected cost} = P(L) \cdot \left[\text{cost of funds in state }L\right]\]

For a bank with deposit franchise: \[\text{Cost of funds in state }L = i - \phi \cdot \Delta D\]

where $\phi > 0$ captures the fact that deposit inflows reduce the bank’s marginal cost of funds — more deposits arrive in state $L$ (flight to safety), pushing the bank’s average funding cost down. For a finance company with no deposit franchise: \[\text{Cost of funds in state }L = i + \phi' \cdot \text{Spread}_L\]

where the premium on wholesale funds rises with the CP spread.

Since the bank’s funding cost falls when the commitment is drawn and the finance company’s rises, the bank can price backup lines at a lower fee $s$ and still break even. The bank dominates the market for CP backup lines in competitive equilibrium.

Proposition (Gatev-Strahan): Comparative Advantage in Liquidity Insurance

An intermediary whose funding cost co-varies negatively with market liquidity (i.e., its funding gets cheaper when market liquidity falls) has a comparative advantage in providing liquidity insurance to CP issuers.

Banks have this property — deposit inflows arrive when CP spreads widen — and therefore banks can offer backup lines at lower cost than competing intermediaries whose funding costs move with market conditions.

6.3.4 Empirical Evidence

Gatev and Strahan test three main empirical implications using aggregate U.S. data and bank-level panel data:

Finding 1: Bank assets grow faster when the CP spread is wide.

When the paper-bill spread (CP rate minus T-bill rate) is high, bank total assets grow faster than when it is low. This acceleration occurs not only in the loan portfolio (reflecting drawdowns on credit lines) but also in the holdings of liquid assets (cash and securities). The simultaneous increase in loans and liquid assets is the key signature of the natural hedge: banks are not drawing down their liquid buffer to fund drawdowns; instead, new deposit money is arriving that can fund both.

Finding 2: Transactions deposits increase with the CP spread.

The quantity of assets funded by transactions deposits (demand deposits and other checkable deposits) rises when the paper-bill spread widens. This confirms the flight-to-safety mechanism: investors move funds into bank deposits when market liquidity deteriorates.

Finding 3: Bank CD yields fall relative to finance company yields when the CP spread is wide.

The yield spread between bank-issued certificates of deposit and finance-company- issued paper narrows (and may invert) when CP spreads are high. Banks can borrow more cheaply than finance companies precisely when the demand for backup liquidity is highest. This is the most direct evidence of the comparative advantage.

The Role of the Government Safety Net

An important question is whether the natural hedge reflects a genuine economic synergy or a subsidy from the government safety net (deposit insurance, implicit bailout guarantees).

Gatev and Strahan present evidence consistent with the latter playing a significant role. Using a historical comparison, they note that Pennacchi (2006) finds no evidence of flight-to-safety deposit inflows during periods of tight market liquidity in 1920–1933 — before federal deposit insurance was introduced. Once the FDIC was established in 1934, the pattern emerges.

This suggests that deposit insurance is what makes bank deposits a safe haven. Without the government backstop, depositors would be as wary of bank deposits as of any other claim during a crisis, and the natural hedge would disappear. We return to this implication in Part 4.

6.4 Part 3: Testing the Synergy — Commitment Exposure and Bank Risk

6.4.1 From Theory to Measurement

The KRS model predicts that banks with more transactions deposits can take on more loan commitment exposure without increasing their risk — the deposit inflows hedge the commitment outflows. How can we test this? The challenge is that loan commitments are off-balance-sheet: they don’t appear directly in reported assets and are easy to miss in simple accounting ratios.

Gatev, Schuermann, and Strahan (2009) take an elegant approach. They measure bank risk directly using stock-return volatility — the annualized standard deviation of daily equity returns. This captures total risk as perceived by equity markets, including both on- and off-balance-sheet exposures. They then ask: does stock-return volatility increase with loan commitment exposure, and does having more transactions deposits dampen this relationship?

6.4.2 The GSS Framework

The key variables are:

Commitment ratio: $\text{LC} = \frac{\text{Unused commitments}}{\text{Unused commitments} + \text{Loans}}$. This measures the share of total credit exposure that is off-balance-sheet and contingent. Banks with high LC are more exposed to liquidity demand shocks.
Transactions deposits: $\text{TD} = \frac{\text{Transactions deposits}}{\text{Total deposits}}$. High-TD banks have more of the natural hedge; they are more likely to receive deposit inflows when commitment drawdowns occur.
Risk measure: $\sigma$ = annualized standard deviation of daily stock returns.

The KRS prediction translates into a specific interaction: the slope of $\sigma$ on LC should be positive for low-TD banks (commitment exposure raises risk when the hedge is absent) but flat or negative for high-TD banks (the hedge offsets the exposure).

6.4.3 The Key Finding

Gatev, Schuermann, and Strahan sort banks into a 3×3 grid on (TD, LC) and compute average stock-return volatility in each cell. The result, shown in their Table 1 and Figure 1, is striking:

GSS Main Result

For banks with low transactions deposits (bottom third of TD): Moving from low to high commitment ratios is associated with an increase in stock-return volatility of approximately 8–10 percentage points (annualized). The regression slope of $\sigma$ on LC is 0.28 (t-statistic: 5.55).

For banks with high transactions deposits (top third of TD): The same increase in commitment exposure is associated with essentially no increase in stock-return volatility. The regression slope is $-$0.10 (t-statistic:$-$1.22) — if anything, risk falls slightly.

This reverses the standard narrative of liquidity risk at banks. Transactions deposits — long seen as a source of fragility (because they can trigger runs) — are in fact a stabilizing force that insulates banks from their off-balance- sheet liquidity exposures.

The finding can be summarized in a simple 2×2:

	Low Commitment Exposure	High Commitment Exposure
Low Transactions Deposits	Baseline risk	High risk (no hedge)
High Transactions Deposits	Baseline risk	Low risk (hedge works)

The key cell is bottom-right: banks that combine high commitment exposure with high transactions deposits are not especially risky, because the natural hedge is operative.

What Drives This Result?

Two mechanisms are consistent with the finding:

Selection: Banks with high transactions deposits choose to take on more commitment exposure because they know the hedge will work. This is the pure KRS story: the deposit franchise enables more commitment business.
Insurance: Given a level of commitment exposure, high-TD banks face lower volatility because deposit inflows actually arrive when drawdowns do. This is the GS story: the natural hedge materially reduces the cost of honoring commitments.

GSS are careful to address reverse causality — the concern that lower-risk banks simply prefer both high TD and high LC. Their tests rule this out by showing that the risk-dampening effect of TD strengthens during market stress episodes when causal identification is cleaner (see below).

6.4.4 When Does the Synergy Kick In? Market Conditions

A further prediction of the Gatev-Strahan mechanism is that the hedge should be stronger during episodes of market stress — precisely when the paper-bill spread is high and deposit inflows are largest. GSS test this by interacting the LC measure with the CP-Tbill spread.

The results confirm: the dampening effect of transactions deposits on risk is significantly larger during periods of market tightness (high CP spread). During normal times, the hedge is present but modest; during crises, it becomes a powerful stabilizer.

This time-series variation is important for causal identification. If the pattern were driven by selection (safe banks happen to have both high TD and high LC), we would not expect the hedge to become stronger during stress periods — there is no particular reason a bank’s inherent safety should be more valuable precisely when external markets are disrupted. The time-varying evidence points to the hedging channel as the true mechanism.

The KRS–GS–GSS Research Program in Summary

Paper	Contribution	Method
KRS (2002)	Why deposits and commitments belong together: the shared-buffer synergy	Theory + cross-sectional regression
Gatev-Strahan (2006)	How the synergy works: flight-to-safety deposit inflows as a natural hedge	Time-series aggregate data
GSS (2009)	How strong the synergy is: high-TD banks don’t see risk rise with LC	Bank-level panel, stock-return volatility

Together, these papers establish a coherent theory of bank specialness: the combining of demand deposits with loan commitments is not arbitrary. It is efficient, and it confers a genuine comparative advantage in providing liquidity insurance that no other intermediary can easily replicate.

6.5 Part 4: Crisis and the Limits of Liquidity Provision

6.5.1 The Standard Story and the 2007–2009 Anomaly

The KRS–GS–GSS framework implies that the banking system should be especially helpful during financial market crises. When outside markets seize up, banks serve as a liquidity backstop: firms draw their credit lines, deposit money flows in from investors seeking safety, and the two flows offset each other. Credit continues to flow; the broader disruption is cushioned.

This is the standard story, and it is well supported by evidence from earlier episodes: the bond-market turmoil of fall 1998 (LTCM and Russian default), the 9/11 attacks in 2001, the Enron accounting crisis in 2002. In each case, CP spreads spiked, bank deposit inflows surged, firms drew their credit lines, and banks honored their commitments without apparent strain.

Then came the 2007–2009 financial crisis. Acharya and Mora (2015) document that the standard story broke down. The mechanism that had worked smoothly in past crises failed in the most severe crisis since the Great Depression. Understanding why — and what substituted for it — is the focus of this section.

6.5.2 What Happened: Deposit Inflows Stalled

The ABCP market began to freeze in August 2007. Based on the prior literature, the prediction was clear: banks should see a surge in deposit inflows as investors fled risky short-term instruments for the safety of bank deposits, even as firms drew down their credit lines.

The prediction was wrong.

The Acharya-Mora Finding: The Natural Hedge Broke Down

Deposit inflows were anemic at the onset of the crisis: Core deposits at all U.S. banks grew by only $90 billion from 2007Q3 to 2008Q2 (covering the ABCP freeze through just before the Lehman failure). In the comparable pre-crisis five-year period, the average semi-annual increase was $130 billion. The gap was even more pronounced for small banks.

Lending outpaced deposits: In aggregate, bank lending growth exceeded core deposit funding growth throughout the first year of the crisis (2007Q3–2008Q2), producing a sustained loan-to-deposit shortfall. This is the opposite of what the natural hedge predicts.

Banks with high undrawn commitments were hit hardest: Cross-sectionally, banks with large proportions of undrawn commitments experienced the largest loan-to-deposit shortfalls, forced to scramble for alternative funding. These banks raised deposit rates significantly — 25 to 50 basis points more than low-commitment banks — in an attempt to attract deposits.

Despite higher rates, core deposit growth was lower at commitment-exposed banks. This is the most alarming finding: even after paying more for deposits, these banks attracted fewer core deposits than their low-commitment peers. The flight-to-safety mechanism — the very engine of the natural hedge — was not delivering funds to the banks that needed it most.

6.5.3 Why the Mechanism Broke Down

The key difference between 2007–2009 and earlier crises is the location of the distress. In 1998 and 2001, the crisis originated outside the banking system: a hedge fund failure, a geopolitical shock, an accounting scandal at a non-bank firm. Investors lost confidence in market instruments but remained confident in bank deposits. They moved money from CP and money-market instruments into bank deposits because banks were perceived as safe havens.

In 2007–2009, the banking system itself was at the center of the crisis:

Banks had originated or sponsored the subprime mortgages and CDOs that were experiencing severe losses.
Opaque exposure to structured credit products made it difficult to distinguish safe from distressed institutions.
Over 62% of bank deposits at the onset of the crisis were uninsured (above the $100,000 FDIC limit), so investors seeking explicit safety could not rely on deposit insurance alone.

The flight to safety therefore went not to bank deposits but to instruments with explicit government guarantees: Treasury securities, government-sponsored FHLB discount notes, and money market mutual funds investing exclusively in government securities. As Figure 3 of Acharya and Mora (2015) illustrates, VIX and aggregate deposit flows were uncorrelated or slightly negatively correlated during the first year of the crisis — the opposite of the pattern in 1998.

The Role of Uninsured Deposits

At the onset of the 2007–2009 crisis, more than 62% of deposits in the U.S. banking system were in accounts above the $100,000 FDIC insurance limit. These uninsured depositors behaved like wholesale creditors: they monitored bank balance sheets and moved funds based on perceived solvency risk.

Acharya and Mora show that failed and near-failed banks offered significantly higher CD rates in the run-up to failure (their Figure 1), consistent with these institutions competing for uninsured funds to plug liquidity gaps. The market discipline mechanism — uninsured depositors withdrawing funds from risky banks — was working, but it was working against the flight-to-safety pattern that underpins the natural hedge.

This illustrates the fundamental tension: deposit insurance enables the natural hedge (by making deposits a safe haven), but incomplete deposit insurance means the mechanism can break down precisely when banks themselves are risky.

The Rajan (2005) conjecture, made before the crisis, proved prescient: the flight-to-safety mechanism works only if “banks are not perceived as credit risks themselves.” When uncertainty is about losses located within the banking system, even healthy banks struggle to attract deposits, because investors cannot easily distinguish which institutions are sound.

6.5.4 Government Support as the True Backstop

The striking aspect of the 2007–2009 episode is that banks ultimately did honor most of their credit line commitments — firms that had lines in place drew them down, and (with rare exceptions) banks funded these drawdowns. How?

The answer is not the natural hedge. It is direct government intervention:

Government Substituted for Deposits as the Liquidity Backstop

Acharya and Mora document that FHLB (Federal Home Loan Bank) advances covered approximately 65% of the nondeposit borrowing growth at commitment-exposed banks during the first year of the crisis. The shortfall between loan demand and deposit funding was bridged by:

FHLB advances: Banks pledged mortgage-backed securities and other eligible collateral to borrow from the Federal Home Loan Banks at government-subsidized rates.
Federal Reserve liquidity facilities: The Fed created several emergency facilities (TAF, TSLF, PDCF, AMLF) to provide liquidity directly to banks and other intermediaries.
Expanded deposit insurance: In October 2008, the FDIC raised the insurance limit from $100,000 to $250,000 and offered full insurance for noninterest-bearing transaction accounts. This eventually triggered the deposit inflow surge — but not until Lehman’s failure in September 2008, more than a year into the crisis.

Without these interventions, banks would likely have curtailed credit lines significantly. The sharp deposit inflow surge following Lehman’s failure (roughly $272 billion in deposits in 2008Q3) was driven by the expanded government guarantees, not by the organic flight-to-safety mechanism.

This finding has a sobering implication: the “natural hedge” that makes bank liquidity provision efficient in normal times is not a market mechanism — it is, at its core, a government-supported mechanism. The flight-to-safety into bank deposits works precisely because investors trust that the government stands behind bank deposits. When that trust wavers (because the government’s commitment is uncertain, or because deposit insurance coverage is incomplete), the mechanism breaks down.

Causal Identification: Solvency vs. Liquidity

A concern in interpreting Acharya and Mora’s findings is reverse causality: perhaps commitment-exposed banks were simply insolvent (they had made bad loans and were exposed to real-estate losses), and their deposit problems reflected depositor discipline rather than a breakdown of the liquidity provision mechanism.

Acharya and Mora address this carefully. They control for measures of solvency risk (real-estate exposure, capital ratios, CDS spreads) and show that the commitment-exposure effect on deposit rates and funding shortfalls is significant above and beyond these solvency proxies. Even fundamentally sound banks that happened to have large undrawn commitments experienced funding pressure — because liquidity tensions hit both sides of their balance sheet simultaneously.

This is consistent with the Diamond-Rajan (2005) model in which there exists a range of bank fundamentals for which a solvent bank can be rendered illiquid by a coordination failure. In 2007–2009, even the healthier commitment-exposed banks did not experience deposit inflows and therefore could not reintermediate funds within the banking system.

6.5.5 Policy Implications

The Acharya-Mora findings have direct implications for how we think about bank regulation and systemic risk:

1. The safety net enables the natural hedge, but creates moral hazard.

The flight-to-safety mechanism that makes bank liquidity provision efficient depends on the government guarantee behind deposits. Removing or weakening that guarantee — as narrow-banking proposals would — would also remove the natural hedge. Banks would lose their comparative advantage in providing backup liquidity.

2. Capital requirements interact with liquidity provision.

A well-capitalized bank that is committed to honoring its credit lines can still face liquidity problems if deposit inflows don’t materialize. Capital ratios and liquidity requirements must be managed jointly; solvency alone does not guarantee the ability to provide liquidity in a systemic crisis.

3. The size and structure of undrawn commitments matters for systemic risk.

In the 2007–2009 crisis, the aggregate volume of undrawn commitments — an off-balance-sheet liability rarely stressed in regulatory frameworks — became a major source of fragility. Commitments create pro-cyclical demand for bank funding precisely when funding is scarce.

4. Government liquidity facilities are not costless backstops.

FHLB advances and Fed facilities bridged the gap in 2007–2009, but they required banks to pledge collateral, paid below-market rates, and involved significant government exposure. The private natural hedge that KRS and GS described was replaced by a public backstop — transferring risk from banks to taxpayers.

6.6 Part 5: Extensions and Recent Evidence

The four-paper arc — KRS, Gatev-Strahan, GSS, Acharya-Mora — was completed by 2015. Since then, three episodes have provided new stress tests of the framework: the post-crisis QE era, the COVID-19 shock of March 2020, and the regional bank failures of 2023. Each has generated important evidence that extends, qualifies, or deepens the original results.

6.6.1 The Deposit Franchise as a Pricing Hedge

Gatev and Strahan (2006) showed that bank deposit quantities rise when market liquidity tightens. Drechsler, Savov, and Schnabl (2021) ask a deeper question: why are bank deposits a safe haven in the first place, and what does the deposit franchise imply for the natural hedge beyond quantity flows?

Their answer centers on deposit rate betas — the sensitivity of the deposit rate to changes in market interest rates. Banks systematically pay deposit rates that adjust very slowly to changes in benchmark rates (the fed funds rate, T-bill yield). When market rates rise by 100 basis points, banks typically raise deposit rates by only 30–50 basis points, keeping a large spread between market rates and the cost of deposits.

Drechsler-Savov-Schnabl (2021): The Deposit Franchise Hedge

Let $r^m_t$ be the market interest rate and $r^d_t$ be the deposit rate. Define the deposit spread as $s_t = r^m_t - r^d_t > 0$. The bank captures this spread on its entire deposit base $D$, earning rents $s_t \cdot D$.

When market rates rise, $s_t$ widens because $r^d_t$ is sticky. The bank earns more interest income on its loan portfolio (which reprices at market rates) while its deposit funding cost rises only partially. This creates a natural hedge against interest rate risk: the deposit franchise insulates the bank’s net interest margin from rate movements.

Formally, let bank assets pay a floating rate $r^m_t$ and deposits cost $r^d_t$. The bank’s net interest income per unit of assets is approximately:

\[\text{NII}_t \approx r^m_t - r^d_t = s_t\]

Since $\partial s_t / \partial r^m_t > 0$ (the spread widens when rates rise), bank income is positively correlated with rate increases — the opposite of what a standard maturity-mismatch story would predict.

This result resolves a puzzle in the Gatev-Strahan framework: the natural hedge works not only through quantity (more deposits arrive during stress) but also through pricing (existing deposits become cheaper relative to market rates during rate spikes). Together, these two channels explain why bank funding costs are counter-cyclical in a broader sense — counter to both credit spreads (the GS quantity mechanism) and interest rate movements (the DSS pricing mechanism).

Implications for the 2022–23 Rate Cycle

The DSS framework predicted that the sharp rate hikes of 2022–2023 would widen deposit spreads and benefit bank income — which proved correct for banks that retained their deposits. But it also predicted a fragility: if depositors leave for higher-yielding alternatives (money-market funds, T-bills), the franchise value evaporates. Banks with large shares of uninsured deposits — which are rate-sensitive and not protected by the FDIC guarantee — are most exposed.

This is precisely what happened at Silicon Valley Bank in March 2023: the bank held long-duration assets funded by uninsured, rate-sensitive deposits. When depositors began moving to higher-yielding instruments, unrealized losses were realized, triggering a run that the deposit franchise pricing model explicitly anticipates.

6.6.2 COVID-19: A Second Stress Test of Acharya-Mora

The COVID-19 pandemic of March 2020 provided a sharp natural experiment. As financial markets dislocated violently, firms simultaneously drew down credit lines at record speed and investors sought safe assets. Would the banking system repeat the 2007–2009 breakdown documented by Acharya and Mora, or would the natural hedge work?

What happened in March 2020. Li, Strahan, and Zhang (2020) document that U.S. banks faced the largest single-quarter increase in loan demand ever recorded: C&I loans grew by roughly $480 billion in Q1 2020, almost entirely from drawdowns on pre-committed credit lines. Large, investment-grade firms — those with the largest pre-committed lines — drove the surge. Unlike 2007–2009, however:

Li-Strahan-Zhang (2020): Banks as Lenders of First Resort in COVID-19

The natural hedge worked in 2020 — but why?

Three conditions distinguished COVID from 2007–2009:

Pre-crisis bank capitalization was strong. Banks entered 2020 with capital ratios roughly 3–4 percentage points higher than in 2007, providing a larger buffer to absorb unexpected drawdowns without triggering fire-sale dynamics.
Deposit inflows did arrive. Unlike 2008, flight-to-safety deposit inflows to the banking system coincided with drawdowns. The Federal Reserve’s rapid reserve injection (over $1.5 trillion in two weeks) ensured banks had ample liquidity. Deposit insurance covered most retail deposits, maintaining depositor confidence.
Policy intervention was faster and larger. The Fed’s emergency facilities (MMLF, CPFF, PMCCF/SMCCF) activated within days, providing backstop funding before banks exhausted their liquid-asset buffers.

Conclusion: The natural hedge operated in 2020, but its operation was conditional on government support. Without the Fed’s speed and scale, the outcome could have resembled 2008. COVID thus reinforces rather than contradicts Acharya-Mora: the private natural hedge is fragile; the public backstop is the true stabilizer.

The distributional cost: the credit-line channel. Even though the banking system as a whole accommodated the drawdown surge, Greenwald, Krainer, and Paul (2025) identify an important distributional consequence. Using supervisory loan data covering nearly all U.S. C&I credit, they show that banks absorbing large drawdowns reduced new term lending to smaller firms — by roughly 10–30 cents per dollar drawn. This occurs because drawdowns consume regulatory capital, and banks operating near capital constraints cut new credit to manage their ratios.

The Credit-Line Channel (Greenwald-Krainer-Paul 2025)

The credit-line channel operates through bank balance sheets:

Drawdown: Large firm A draws $1 on its pre-committed credit line. The bank funds this from its liquid buffer or deposit inflows.
Capital encumbrance: The drawn loan carries a risk weight of 100%, consuming roughly $0.08–$0.12 of regulatory capital (at a 8–12% capital ratio).
Crowding out: To maintain its capital ratio, the bank reduces new term loans to other borrowers — typically smaller, relationship-based borrowers with no pre-committed lines. These firms lose access to credit precisely when they need it most.

Implication: The KRS synergy — efficient combined provision of deposits and credit lines — has a distributional shadow. Honored commitments to large firms impose an externality on small firms. The efficiency of the synergy at the bank level coexists with a credit-reallocation externality at the system level.

6.6.3 The 2023 Banking Stress: A Third Failure Mode

The failures of Silicon Valley Bank, Signature Bank, and First Republic in spring 2023 provided a third stress test of the Acharya-Mora framework. This episode is distinct from both 2008 and 2020 in its mechanism.

In 2008, the natural hedge failed because the banking system was the source of credit losses — depositors and investors lost confidence in bank solvency. In 2023, the trigger was different: interest rate risk materialized as unrealized losses on securities portfolios, and uninsured depositors — disproportionately sophisticated tech firms and venture funds — ran before insolvency occurred.

Jiang-Matvos-Piskorski-Seru (2024): Bank Fragility from Rate Risk and Uninsured Deposits

The paper marks to market the entire U.S. bank securities portfolio using Q1 2023 data. Key findings:

Average unrealized loss across all U.S. banks: approximately 10% of asset value, totaling roughly $2 trillion — not reported on balance sheets under hold-to-maturity accounting.
A bank is fragile if its mark-to-market assets fall below its uninsured deposits: even without any fundamental insolvency, uninsured depositors rationally run because they fear being last out. The paper estimates 186 banks were in this fragility zone as of March 2023.
SVB is the limiting case: 94% uninsured deposits plus large unrealized losses on long-duration MBS and Treasuries created a textbook fragility. The FDIC’s announcement that it would fully backstop all deposits (even uninsured) stopped the broader run.

Connection to Acharya-Mora: The 2023 episode generalizes their finding. The natural hedge fails not only when credit losses hit bank assets (2008) but also when rate losses do — because both make depositors doubt bank solvency and trigger flight rather than inflow.

How 2023 differed from 2008. The 2023 stress was concentrated at a small number of outlier banks (high uninsured deposits, long-duration assets) rather than systemic across the industry. Most banks — including those with large undrawn credit-line exposures — did not experience the Acharya-Mora funding breakdown. In fact, the largest banks experienced deposit inflows as funds fled regional banks, precisely the flight-to-safety mechanism Gatev and Strahan document. The 2023 episode thus confirms that the natural hedge is not monolithic: it works for large banks perceived as systemically important, while breaking down for smaller institutions perceived as fragile.

Three Stress Tests of the Natural Hedge: A Unified View

Episode	Did deposit inflows arrive?	Natural hedge operative?	Key mechanism of failure/success
2007–2009	No (at onset)	Failed	Banks were source of credit losses; uninsured deposits fled
2020 (COVID)	Yes	Worked	Rapid Fed intervention; strong pre-crisis capital; deposit insurance intact
2023 (SVB)	Partially (large banks yes, regionals no)	Mixed	Rate losses + uninsured deposits at specific banks; systemic spread stopped by FDIC guarantee expansion

Lesson: The natural hedge is a conditional mechanism, not a structural constant. It operates when the government backstop is credible and the banking system is not itself the source of distress. Policy design — the speed of intervention, the scope of deposit insurance, and capital requirements — determines whether the hedge holds.

6.6.4 The QE Era: Liquidity Dependence and the KRS Synergy

A fourth development post-2015 concerns how unconventional monetary policy (quantitative easing) interacts with the deposit-lending synergy. Acharya, Chauhan, Rajan, and Steffen (2023) document a phenomenon they call liquidity dependence: as the Federal Reserve expanded its balance sheet during 2009–2021, banks responded by issuing more uninsured deposits and extending more credit lines simultaneously.

This is consistent with KRS — banks that have more deposits (supplied by QE-driven reserve creation) optimally offer more commitments. But the authors identify an asymmetry: during quantitative tightening (QT), bank credit lines did not contract proportionally. Banks had built up a stock of long-dated commitments that could not be withdrawn unilaterally, while their deposit bases shrank as reserves left the system.

The result is a QT-induced fragility: the deposit-lending synergy that was built up during QE becomes a source of vulnerability during QT, because the liquidity demands on both sides of the balance sheet no longer shrink in step. This suggests that the macroprudential implications of large central bank balance sheets extend through the bank liquidity provision channel in ways not anticipated by KRS.

6.7 Summary and Connections

This chapter has traced the logic of bank liquidity provision through four core papers and four post-2015 extensions. The table below summarizes the key contributions.

Paper	Question	Answer	Method
KRS (2002)	Why do deposits and credit lines coexist?	Shared liquid-asset buffer: synergy when $\rho < 1$	Theory + cross-section
Gatev-Strahan (2006)	Why can banks provide this service cheaply?	Natural hedge: deposit inflows fund drawdowns during market stress	Model + time series
GSS (2009)	Does the hedge actually reduce bank risk?	Yes, but only for high-transactions-deposit banks	Bank panel, equity volatility
Acharya-Mora (2015)	What happens when the hedge fails?	Crisis of banks as LPs: government support filled the gap in 2007–2009	Crisis event study
Drechsler-Savov-Schnabl (2021)	Why is the deposit franchise itself a hedge?	Sticky deposit rates create a pricing hedge against rate movements	Cross-bank panel, rates
Li-Strahan-Zhang (2020)	Did the hedge work in COVID-19?	Yes — but only because of rapid policy intervention	COVID event study
Greenwald-Krainer-Paul (2025)	What is the distributional cost of honoring commitments?	Credit-line drawdowns crowd out small-firm lending	Supervisory loan data
Jiang-Matvos-Piskorski-Seru (2024)	Can rate risk trigger Acharya-Mora failures?	Yes: unrealized losses + uninsured deposits create a new fragility mode	Mark-to-market analysis

Connection to previous chapters. The liquidity provision framework in this chapter complements the liability-side theory from Chapters 2–3 and the asset-side theory from Chapters 4–5:

In Diamond-Dybvig (Chapter 3), deposits are fragile because of the sequential- service constraint. Here, deposits are stabilizing because their inflows offset commitment drawdowns. The same instrument can be either a source of fragility or a source of stability, depending on the correlation structure of shocks.
In Diamond (1984, Chapter 5), the fixed-rate deposit contract disciplines the bank’s monitoring effort. Here, the deposit contract is valuable because of the liquidity services it provides to the bank itself. The liability structure serves both functions simultaneously.
The breakdown documented by Acharya and Mora connects to the systemic risk and bank-run literatures: when uncertainty is about losses within the banking system, the coordination failures modeled in Chapter 3 can extend to wholesale funding markets, not just retail deposits.

6.8 References

6.8.1 Primary Sources

Kashyap, A. K., Rajan, R., and Stein, J. C. (2002). Banks as Liquidity Providers: An Explanation for the Coexistence of Lending and Deposit-Taking. Journal of Finance, 57(1), 33–73.
Gatev, E., and Strahan, P. E. (2006). Banks’ Advantage in Hedging Liquidity Risk: Theory and Evidence from the Commercial Paper Market. Journal of Finance, 61(2), 867–892.
Gatev, E., Schuermann, T., and Strahan, P. E. (2009). Managing Bank Liquidity Risk: How Deposit-Loan Synergies Vary with Market Conditions. Review of Financial Studies, 22(3), 995–1020.
Acharya, V. V., and Mora, N. (2015). A Crisis of Banks as Liquidity Providers. Journal of Finance, 70(1), 1–43.