Ijcb 24 Q 3 A 6

Bank Risk-Taking and Impaired Monetary
Policy Transmission∗
Philipp J. Koeniga and Eva Schliephakeb

a
Deutsche Bundesbank
b
Lisbon School of Business and Economics
How does risk-taking affect the transmission of interest

rate changes into loan issuance? We study this question in a
banking model with agency frictions. The risk-free rate affects
bank lending via a portfolio adjustment and a loan risk chan-
nel. The former implies that the bank issues more loans when
the risk-free rate falls. The latter implies that the bank may
issue fewer loans because lower risk-free rates lead to higher
risk-taking. Thus, the loan risk channel can counteract the
portfolio adjustment channel. There exists a reversal rate, so
that loan supply even contracts due to higher risk-taking. The
model’s implications square with recent evidence on monetary
transmission.
JEL Codes: G21, E44, E52.
∗
We are grateful to an anonymous referee and the editor, Huberto Ennis, for
their comments and suggestions. We are also grateful to Giovanni Dell’Ariccia,
David Pothier, and Kartik Anand for detailed and insightful feedback. We
thank Diana Bonfim, Olivier Darmouni, Falko Fecht, Thomas Gehrig, Hendrik
Hakenes, Yann Kobi, Jochen Mankart, Yuliyan Mitkov, Steven Ongena, and
Jeremy Stein as well as seminar and conference participants at the Banking
Theory Brown Bag, Catolica Lisbon, Deutsche Bundesbank, and the Danish
Central Bank’s 7th Annual Meeting on New Developments in Business Cycle
Analysis for valuable discussions and comments. Eva Schliephake thankfully
acknowledges support from the Lamfalussy Fellowship Program sponsored by
the European Central Bank, according to clause 8 of the Fellowship Agree-
ment as well as the support from FCT — Portuguese Foundation of Science and
Technology for the project PTDC/EGE-ECO/6041/2020. The views expressed
in this paper are those of the authors and do not necessarily represent those
of the Deutsche Bundesbank, the ECB, or the Eurosystem. Author contact:
Koenig: [email protected]. Schliephake (corresponding author):
[email protected].
257
258 International Journal of Central Banking July 2024
1. Introduction
Central banks in several advanced economies have until recently kept

their policy rates at historically low levels, often close to the zero
lower bound. The extant literature has highlighted two key concerns
on interest rate policies in such an environment. First, lower policy
rates may induce banks to take more risks, which could pose a threat
to financial stability (Borio and Zhu 2012). Second, lower rates could
also depress bank profits to the point where banks respond less
to additional monetary stimulus (Eggertsson et al. 2019) or even
reduce the supply of credit to the economy (Brunnermeier, Abadi,
and Koby 2023). Both phenomena have been studied in separate
theoretical frameworks, but empirical evidence covering the recent
period of low interest rates also suggests a close link between them,
which is difficult to reconcile with existing models: the weakening of
the transmission of policy rates correlates with an increase in riskier
lending (Heider, Saidi, and Schepens 2019; Miller and Wanengkirtyo
2020; Arce et al. 2021).
In the present paper, we ask how such a link between impaired
transmission and risk-taking can arise. We show that if deposit
rates are bounded below and banks hold a sufficiently large share
of fixed-income assets whose return changes with the risk-free rate,
higher risk-taking and the impairment of monetary transmission can
become “two sides of the same coin.” In particular, if interest rates
are at a sufficiently low level, further reductions of interest rates
incentivize banks to increase risk-taking, which, in turn, weakens
the transmission of policy rates into loan rates and loan volumes.
We consider a purposefully simple model of a penniless banker
who uses deposits to fund the issuance of risky loans and the holdings
of safe assets, such as bonds or central bank reserves. The banker
can exert a monitoring effort to reduce the risk of default of her
loan portfolio. Depositors can observe the loan issuance and the
safe asset holdings of the banker. However, the monitoring effort is
not observable (and hence uncontractible), thus creating an agency
problem between the banker and her depositors.
In this setting, we study how changes in the risk-free rate affect
loan rates and loan volumes. The transmission of the risk-free rate
works via two channels, a direct portfolio adjustment channel and
an indirect loan risk channel (see Figure 1).
Vol. 20 No. 3 Bank Risk-Taking and Impaired Monetary Policy 259
Figure 1. Direct Portfolio Adjustment Channel

and Indirect Loan Risk Channel
The portfolio adjustment channel reflects the conventional view

of monetary transmission, which holds that lower risk-free rates are
expansionary and translate into more bank lending. As in standard
banking models, a lower risk-free rate reduces the return on safe
assets and the opportunity cost of loan issuance. The banker, in
turn, optimally issues more loans at lower loan rates (Freixas and
Rochet 1997).
The indirect loan risk channel arises because changes in the risk-
free rate also alter the banker’s monitoring incentives, which, in turn,
affect the banker’s optimal loan issuance.1 In particular, if monitor-
ing incentives improve, loan risk declines and the banker optimally
expands the issuance of loans by lowering the loan rate.
However, the risk-free rate exerts two opposing effects on mon-
itoring incentives, implying that it is a priori not clear whether
the loan risk channel amplifies or counteracts the portfolio adjust-
ment channel. On the one hand, a lower risk-free rate reduces the
profitability of safe assets and depresses expected profits. This safe
asset effect worsens monitoring incentives. On the other hand, if
the banker can pass on a lower risk-free rate to depositors, prof-
its increase. This deposit pass-through effect improves monitoring
1
We use the term “loan risk channel” to refer to the indirect effect of the
risk-free rate on loan issuance via changes in monitoring incentives. Our loan
risk channel should be distinguished from the “risk-taking channel,” which refers
to the direct effect of the risk-free rate on risk-taking incentives (Dell’Ariccia,
Laeven, and Marquez 2014).
incentives. Whenever the safe asset effect dominates the deposit

pass-through effect, the banker’s monitoring incentives worsen, and
her risk-taking increases when the risk-free rate becomes lower (and
vice versa if the deposit pass-through effect dominates).
We show that the interaction between the portfolio adjustment
channel and the loan risk channel can lead to three possible cases
depending on the level of the risk-free rate.
First, if the risk-free rate is sufficiently high, the deposit pass-
through effect dominates the safe asset effect, and the loan risk
channel amplifies the portfolio adjustment channel. Second, for lower
values of the risk-free rate, the safe asset effect dominates the deposit
pass-through effect. The loan risk channel counteracts the portfo-
lio adjustment channel. The banker still increases her loan issuance
when the risk-free rate falls, but the increase in loan risk lessens
her loan issuance. Third, if the risk-free rate is sufficiently low, the
loan risk channel can even dominate the portfolio adjustment chan-
nel. In this case, further reductions in the risk-free rate lead the
banker to reduce lending. The critical value below which the loan
risk channel dominates the portfolio adjustment channel constitutes
a reversal rate, as in Brunnermeier, Abadi, and Koby (2023). Like in
their model, a precondition for the existence of a reversal rate is that
bank profits decrease in the risk-free rate. In contrast to their model,
the reversal rate in our model does not arise due to an exogenous
constraint on future bank profits but stems from the agency friction
between the banker and her depositors. We derive a simple condition
for the occurrence of this “reversal scenario”: the loan risk channel
dominates the portfolio adjustment channel if the banker reduces
her monitoring more than one-for-one in response to a reduction in
the risk-free rate.
To simplify the exposition of the key mechanism behind the inter-
action of risk-taking and monetary policy transmission, we make two
assumptions in our baseline model.
First, there is a lower bound on deposit rates; i.e., there exists
a minimal return that the banker must offer on deposits for agents
to be willing to hold them rather than switch to cash. This assump-
tion reflects the empirical observation that changes in deposit rates
become progressively smaller and approach a lower bound when pol-
icy rates are lowered towards negative territory (Eggertsson et al.
2019).
Second, the banker always holds a non-negligible amount of

assets whose rate of return changes in lockstep with changes in the
policy rate. We assume that these are “safe assets” such as central
bank reserves, government bonds, or senior tranches of mortgage-
backed securities.2 For the safe asset effect to arise, the banker’s
expected profit must react sufficiently to changes in the risk-free
rate. That is, the banker must be somewhat constrained in reducing
her safe assets in order to mitigate (or even to offset completely)
the negative effect of a lower risk-free rate on her profits. There
are various reasons why banks face such constraints in practice. For
example, due to setup and switching costs, deposits are a quasi-
fixed factor of production (Flannery 1982; Sharpe 1997). Once banks
raise deposit funding before deciding on their loan issuance, deposit
and loan volumes are not necessarily completely balanced.3 As a
consequence, deposits in excess of what is required to fund loan
issuance may be held as reserves with the central bank or invested
in short-term fixed-income securities. In addition, banks face reg-
ulatory constraints, such as reserve or liquidity requirements, that
force them to cover a certain share of their deposit liabilities with
safe and liquid assets, like reserves or government bonds. Moreover,
since the financial crisis of 2008/09, central banks have expanded
reserves through asset purchase and lending programs beyond what
is required by the banking sector in aggregate (Ennis and Wolman
2015; European Central Bank 2017; Bechtel et al. 2021). Under such
a regime, individual banks may end up with excess reserves without
being able to instantly dispose of reserves via the interbank market
(Brandao-Marques et al. 2021).
To simplify the exposition in the baseline model, we follow
Acharya and Naqvi (2012) and assume that the banker has a fixed
amount of deposits and cannot adjust the “intensive margin” of her
deposits. Moreover, we fix the deposit amount such that the banker
2
We could also allow the banker to invest in risky fixed-income securities, pro-
vided that their payoffs are uncorrelated with the payoffs from the bank’s loans
and that a no-arbitrage condition ensures that their expected return matches the
risk-free rate.
3
In practice, banks often adjust deposit volumes by changing the rate offered
on deposits. Empirically, in particular at low rates, rate adjustments and, by
extensions, adjustments in deposit volumes, occur rather infrequently (Paraschiv
2013; Jobst and Lin 2016; Döpp, Horovitz, and Szimayer 2022), suggesting an
imperfect adjustment between loans and deposits.
is bound to hold more deposits than what she needs to fund her opti-
mal loan issuance. Any residual deposits are invested in safe assets
whose return moves in lockstep with the risk-free rate. Put differ-
ently, although the banker can trade off loan issuance and safe assets
at the margin, she cannot shrink her balance sheet by issuing fewer
deposits and disposing of safe assets completely.
We consider several extensions of this baseline model to probe
the robustness of its mechanism. First, we analyze the effect of
insured deposits on the possibility of transmission reversal. Deposit
insurance (if not fairly priced) provides an exogenous subsidy to the
banker that increases her profits. As a consequence, deposit insur-
ance mitigates the problem of transmission reversal. In the limit,
when all deposits are insured, the reversal rate ceases to exist. Thus,
ceteris paribus, a transmission reversal constitutes less of a problem
for banks that are funded with a larger share of insured deposits.
Second, we relax the admittedly stark assumption that the
banker cannot adjust the intensive margin of her deposits and the
size of her balance sheet. Instead, we allow the banker to endoge-
nously choose deposits and safe assets. We consider two variants of
the model that both preserve the safe asset effect. In the first, we
assume that the bank faces random inflows or outflows to deposi-
tors’ accounts. These random changes to deposits are matched on
the banker’s balance sheet by inflows and outflows of central bank
reserves. As a consequence, the bank may end up holding excess
reserves with a certain probability. This extension illustrates that the
presence or absence of the safe asset effect depends on the banker’s
ability to optimally adjust her safe asset position. In particular, we
recover the results in the benchmark model if the probability of
a deposit inflow (i.e., ending up with excess reserves) approaches
unity, whereas the safe asset effect disappears if the probability of
random reserve changes goes to zero. In the second variant, instead
of random deposit flows, we assume that the banker faces a liquid-
ity requirement that forces her to hold safe assets equal to a certain
share of her deposits (as in Brunnermeier, Abadi, and Koby 2023). In
this case, the portfolio adjustment and the loan risk channel always
move in the same direction, but both switch signs once the safe asset
effect dominates the deposit pass-through effect. The dominance of
the safe asset effect becomes a necessary and sufficient condition for
the reversal of monetary transmission.
Related Literature. Our paper relates to a large body of lit-

erature that analyzes the transmission of monetary policy through
the banking sector. The traditional view is that a reduction in policy
rates reduces banks’ funding cost and induces greater loan supply
(Bernanke and Blinder 1988; Bernanke and Gertler 1995; Kashyap
and Stein 1995). A variant of this channel is at work in our model,
but we show that it can be weakened or amplified by an (a priori)
ambiguous indirect loan risk channel that arises from the agency
problem between the bank and its depositors.
The loan risk channel connects our paper to the literature on the
risk-taking channel of monetary policy (e.g., Dell’Ariccia, Laeven,
and Marquez 2014; Martinez-Miera and Repullo 2017). The risk-
taking channel refers to the direct effect of interest rate changes on
the bank’s monitoring incentives. We show how the presence of a
lower bound on deposit rates and the presence of safe asset holdings
creates a novel variant of the risk-taking channel. However, the focus
of our model is on the loan risk channel, i.e., the indirect effect of
the risk-free rate on loan issuance via monitoring incentives.
The banks in our model face an agency problem similar to
that of Dell’Ariccia, Laeven, and Marquez (2014). In contrast to
Dell’Ariccia, Laeven, and Marquez (2014), who focus on the effect
of leverage, we adopt the assumption of a fixed deposit volume from
Acharya and Naqvi (2012) to concentrate on the effect of mone-
tary policy on the bank’s endogenous portfolio adjustment between
loans and safe assets. Our model therefore complements Dell’Ariccia,
Laeven, and Marquez (2014) in two ways. Firstly, in contrast to
their results, even a fully leveraged bank can increase risk-taking in
response to a lower policy rate because of the interaction between the
deposit pass-through and the safe asset effect. Secondly, the effect of
the risk-free rate on loan rates depends on the interaction between
the portfolio adjustment and the loan risk channel. This decompo-
sition allows us to show how the transmission of policy rates can
become weaker at low levels of the policy rate.4
The dependency of monetary transmission on the level of the pol-
icy rate connects our paper to the growing literature on monetary
4
In Dell’Ariccia, Laeven, and Marquez (2014), the total effect of interest rates
on loan rates is unambiguously positive, so that a reversal of transmission cannot
arise.
policy transmission in a low interest rate environment. Eggertsson

et al. (2019) argue that the increasing attractiveness of cash impairs
the pass-through to deposit rates when the policy rate approaches
the zero lower bound or becomes negative. Brunnermeier, Abadi,
and Koby (2023) show the existence of the reversal rate below which
further reductions in policy rates lead to an increase in loan rates.
Eggertsson et al. (2019) and Brunnermeier, Abadi, and Koby (2023)
derive their results by imposing an exogenous net worth constraint
that mechanically increases equilibrium loan rates. Darracq Pariès,
Kok, and Rottner (2020) study the reversal rate in a general equilib-
rium model with agency frictions. Their net worth constraint arises
because the banker can abscond with deposits. Our model comple-
ments these papers by showing how a reversal rate can arise from
an agency problem and the bank’s risk-taking incentives.
Several recent papers analyze the effects of excess reserves on the
determination of the price level (Ennis 2018) or the effect of bank
lending (Martin, McAndrews, and Skeie 2016). Our results comple-
ment Martin, McAndrews, and Skeie (2016). They argue that reserve
holdings do not matter for bank lending in a frictionless economy,
but they do so in the presence of balance sheet costs. We show how
excess reserves affect lending in the presence of agency frictions.
The implications of our model are in line with empirical obser-
vations at low levels of the policy rate, such as a positive relation-
ship between bank profits and policy rates (Ampudia and Van den
Heuvel 2022; Wang et al. 2022), or a negative relationship between
mortgage rates and policy rates (Basten and Mariathasan 2020;
Miller and Wanengkirtyo 2020). Our model suggests an explanation
for higher risk-taking at rock-bottom interest rates (Heider, Saidi,
and Schepens 2019; Basten and Mariathasan 2020; Bittner, Bonfim,
et al. 2021), and shows why the pass-through to loans may weaken
specifically for riskier banks (Arce et al. 2021).
2. Model Setup
We consider a bank over two periods, indexed by t = 0, 1. The bank

is run by a penniless risk-neutral bank owner/manager (“banker”).
The banker can obtain deposits from a large number of risk-neutral
depositors. The banker decides on the issuance of loans and on the
monitoring of her loans. Monitoring entails a private cost for the
banker and reduces the riskiness of her loans. Depositors cannot

observe the banker’s monitoring choice and the banker cannot com-
mit to a certain level of monitoring. The main focus of our analysis
is on the transmission of monetary policy to loan rates, the loan
volume, and the loan risk. We take the gross risk-free interest rate
r > 0 as the measure of monetary policy and assume that it can be
perfectly controlled by the central bank.
Bank Liabilities. The banker raises deposits in period 0.
Deposits are uninsured and depositors must be compensated for the
risk that the banker cannot fully repay depositors in period 1.5 Thus,
to attract deposits, the banker must offer a deposit rate, rD , which,
in expected terms, matches the depositors’ outside option, u(r).
Assumption 1. The depositors’ outside option u(r) ≥ r is bounded

below by u, i.e., u(r) = u for all r ≤ u−1 (u). For r > u−1 (u), u(r)
is continuously increasing and convex; moreover, u (r) is continuous
at u−1 (u), i.e., u (r) satisfies limr↓u−1 (u) u (r) = 0.
The lower bound u reflects the idea that depositors would switch
to other assets, such as non-interest-bearing cash holdings, once the
risk-free rate becomes too low. The lower bound is not necessarily
equal to zero, as negative rates could still be compensated for in
the form of non-pecuniary benefits of deposits, such as the safety
and ease of making payments. The lower bound on u(·) is the key
assumption needed for the mechanism of our model, whereas the con-
tinuity and convexity assumptions are made for the sake of technical
tractability and can easily be dropped (see Section 4.3).
To further simplify the exposition of the model, we assume that
the banker cannot adjust the “intensive margin” of her deposits.
That is, she either raises an amount D or no deposits at all. We
relax this assumption in Section 4.2 where we allow the banker to
choose deposits endogenously.
Assumption 2. The amount of deposits, D, is exogenously given.
Bank Assets and Monitoring. The banker is a monopolist in

the local loan market. The demand for loans in period 0 is described
5
Section 4.1 considers the effect when the banker also issues insured deposits.
by a demand curve L(rL ), with L (rL ) < 0 and L (rL ) ≤ 0, where
rL denotes the gross interest rate the banker charges on loans.
Loans are risky and are repaid in period 1 with probability
q ∈ (0, 1). The banker can exert unobservable monitoring effort to
influence the repayment probability of her loans. We assume that
monitoring translates one-to-one into the repayment probability, i.e.,
the banker can choose q directly. Monitoring involves a private cost6
κ 2
c(q) = q , where κ > 0.
2
Alternatively, the banker can invest in a risk-free asset that yields
the gross risk-free return r in period 1. One can think of the risk-free
asset as government bonds or reserves held with the central bank.7
The amount invested in the risk-free asset is denoted by R.
The bank’s funding constraint in period 0 is given by
R + L = D. (1)
The amount invested in the risk-free asset is determined endoge-

nously through the banker’s choice of loans as the residual R =
D − L. Henceforth, we use ρ ≡ D R
= 1− D
L
to denote the share of
deposits held in the risk-free asset.
We simplify the exposition of the model by imposing the follow-
ing assumption on the relationship between loan issuance and the
fixed deposit volume.
Assumption 3. The elasticity of the loan demand function, η(rL ) ≡

− L L(r
(rL )rL
L)
, satisfies
η(L−1 (D)) < 1.
Assumption 3 implies that the banker never exhausts her entire
funding base to issue loans, but always holds a strictly positive
6
For analytical tractability, we assume that monitoring costs do not depend on
the banker’s loan issuance. For example, c(q) may represent setup costs for risk-
management systems that, once in place, can be used to process a large number of
loans. As we show in Appendix A.2, our results remain qualitatively unchanged if
we assume a cost function that scales with the loan volume, c(q, rL ) = κ2 q 2 L(rL ).
7
The asset can be risky as long as its payoffs are not correlated with the bank
loan risk and a no-arbitrage condition holds so that the asset’s expected return
equals the risk-free rate.
Figure 2. Sequence of Events
amount of safe assets. We relax Assumption 3 together with Assump-

tion 2 in Section 4.2.
Sequence of Events and Equilibrium. Figure 2 shows the
sequence of events in the model. An equilibrium of the model is
∗ ∗
given by a loan rate rL and a deposit rate rD , which jointly deter-
∗
mine the bank’s optimal loan supply, L , optimal safe asset holdings,
R∗ , and the monitoring choice, q ∗ . The loan rate rL
∗
and the mon-
∗
itoring choice q maximize the banker’s expected profits given the
∗
funding constraint (1), while the deposit rate rD ensures deposi-
tor participation, given depositors’ rational expectations about the
bank’s monitoring choice.
3. The Portfolio Adjustment and the Loan Risk Channel
Optimal Monitoring Choice. We solve the model backwards by

first considering the banker’s optimal choice of monitoring and then
determining her optimal loan issuance. The banker’s expected prof-
its, for given rL and R, can be written as
κq 2
Π = q (rL L(rL ) + rR − rD D) − . (2)
2
The first-order condition for the optimal monitoring choice
becomes8
rL L(rL ) + rR − rD D − κq̂ = 0. (3)
8
All derivations can be found in the appendix.
Given that depositors rationally anticipate the bank’s optimal

monitoring choice q̂, the interest rate on deposits that ensures depos-
itor participation must satisfy
rR
q̂rD + (1 − q̂) ≥ u(r). (4)
D
Depositors expect to be paid rD with probability q̂. With converse
probability, the bank defaults when loans do not pay out at matu-
rity, and depositors obtain a senior claim over a pro rata share of
the remaining safe assets. The expected repayment to the deposi-
tors must be at least as large as their outside option u(r). Since the
banker’s expected profits are strictly decreasing in rD , condition (4)
binds at the optimum, so we can substitute
u(r) − (1 − q̂) rR
D
rD = (5)
q̂
into condition (3) and solve for the optimal monitoring choice q̂.9
Lemma 1. The banker’s optimal monitoring choice is given by a

function q̂(rL , r) with

L −r
∂ q̂ ≥ 0 if q̂rq̂r ≤ 1
η(rL ) ∂ q̂ ≥ 0 if u (r) ≤ ρ
L
and ,
∂rL < 0 else ∂r < 0 else
where η(rL ) ≡ −L (rL )rL /L(rL ) denotes the loan demand elasticity
and ρ ≡ R/D.
The effects of rL and r on the optimal monitoring level reflect

the effects of these rates on the banker’s expected profits. Whenever
a marginal increase in these rates raises profits, the banker increases
her monitoring and vice versa.
More specifically, a higher loan rate increases monitoring when-
ever the loan rate rL is such that the Lerner index, (q̂rL − r)/q̂r, is
9
The equation that pins down q̂ is quadratic and has two solutions. Follow-
ing Allen, Carletti, and Marquez (2011), we choose the larger of the two roots.
Moreover, as Dell’Ariccia, Laeven, and Marquez (2014), we focus on the interior
solution where q̂ < 1 and abstract from the corner solution where q̂ = 1. There
is a sufficiently large range of values for κ such that the interior solution exists.
lower than the inverse loan demand elasticity, 1/η(rL ), which is the
standard condition for the profits of a monopolistic bank to (locally)
increase in rL (Freixas and Rochet 1997).
Whether a lower risk-free rate increases profits and leads to
higher monitoring depends on the relative magnitude of two effects.
On the one hand, a marginal reduction in the risk-free rate lowers
the value of the depositors’ outside option and thereby reduces the
banker’s expected deposit funding costs. This deposit pass-through
effect increases profits by an amount u (r)D and incentivizes the
banker to increase monitoring. On the other hand, a marginal reduc-
tion in the risk-free rate reduces the banker’s return on safe assets.
This safe asset effect reduces profits by R and induces the banker to
reduce monitoring. Thus, a lower risk-free rate decreases monitoring
if the deposit pass-through effect is smaller than the safe asset effect,
i.e., if
u (r)D < R ⇔ u (r) < ρ. (6)
Optimal Loan Issuance and Reserve Holdings. Substitut-

ing the funding constraint (1), the deposit rate (5), and the banker’s
optimal monitoring choice q̂(rL , r) into (2) allows us to rewrite
expected profits as
Π = q̂(rL , r)rL L(rL ) + r(D − L(rL )) − u(r)D

Expected earnings on loans. Earnings on reserves Cost of funds
κ
− q̂(rL , r)2 . (7)
2

Monitoring cost
The banker’s remaining choice variable is the loan rate rL . The

∗
optimal loan rate, rL , is determined by the standard condition for
loan issuance of a monopolistic bank: the Lerner index equals the
inverse loan demand elasticity
∗ ∗
q̂(rL , r)rL −r 1
∗ ∗ = ∗ ). (8)
q̂(rL , r)rL η(rL
At the optimum point, the elasticity of the loan demand exceeds

∗
unity, η(rL ) > 1. Condition (8) takes this particularly simple form
because the effect of rL on q̂ can also be expressed in terms of the

Lerner index and the inverse demand elasticity (cf. Lemma 1).
Monetary Policy Transmission. Monetary policy actions
that change the risk-free rate affect the banker’s optimal loan rate
(and therefore the loan volume) through a portfolio adjustment chan-
nel and a loan risk channel:
(+) (−) (+)/(−)

∗ ∗ ∗ ∗
drL ∂rL ∂rL ∂ q̂(rL , r)
= + × . (9)
dr ∂r
∂q ∂r

portfolio adjustment loan risk channel
channel
The conventional view of monetary policy transmission holds

that a lower risk-free rate is expansionary because it induces an
increase in bank loan issuance. The portfolio adjustment channel
reflects this conventional transmission of monetary policy. Effec-
tively, the banker solves an optimal portfolio problem by allocat-
ing her funds between two investment opportunities (loans and safe
assets).10 Given q̂, a lower risk-free rate reduces the opportunity cost
of investing in loans rather than safe assets. As a consequence, the
banker optimally reduces the loan rate and increases the amount of
loan issuance.
In contrast to the portfolio adjustment channel, the effect of the
loan risk channel is ambiguous: it can either amplify or dampen the
portfolio adjustment channel. To understand the intuition behind
the workings of the loan risk channel, note that, ceteris paribus,
a lower success probability increases the loan rate and reduces the
∗
amount of loan issuance, i.e., ∂rL /∂q < 0. The reason is that the
bank optimally reacts to a lower success probability by increasing
the loan rate in order to keep the expected marginal benefit from
issuing an additional loan equal to the risk-free rate that it earns
on safe assets (cf. Equation (8)). Thus, whenever the safe asset
10
Since the volume of deposits is fixed, the optimal loan rate is independent
of the costs of deposits as in the textbook version of a monopolistic bank with
separable loan and deposit choices (Freixas and Rochet 1997). In Section 4.2, we
show two variants of the model where the banker can choose the deposit volume.
Figure 3. Required Marginal Deposit Rate, u (r),

and Reserves-Deposit Ratio, ρ
Note: To the right (left) of r, the loan risk channel amplifies (weakens) the port-
folio adjustment channel, as can be seen from the change in the slope of the red
curve at r.
effect dominates, a reduction in the risk-free rate reduces monitor-

ing, ∂ q̂/∂r > 0, and the loan risk channel counteracts the portfolio
adjustment channel, thus weakening monetary transmission.
Proposition 1. For a sufficiently small risk-free rate, the safe asset

effect dominates the deposit pass-through effect and the loan risk
channel weakens the transmission of monetary policy via the portfo-
lio adjustment channel, i.e., there exists r such that
∂ q̂
r<r ⇒ > 0. (10)
∂r
Figure 3 illustrates Proposition 1. Note that the lower bound on
u(r) implies that for r < u−1 (u), the deposit pass-through becomes
fully impaired, i.e., u (r) = 0. At this level of the risk-free rate,
the banker is unable to pass a lower risk-free rate through to her
depositors and she becomes unable to further reduce her expected
funding costs. The dashed curve in Figure 3 shows the marginal
required deposit rate, u (r), which becomes flat below u−1 (u) when
the pass-through is fully impaired. However, by Assumption 3, the
banker always holds a strictly positive level of reserves, even at low

risk-free rates below u−1 (u). Thus, the safe asset effect dominates
the deposit pass-through effect whenever the risk-free rate falls below
r > u−1 (u). The solid curve shows the ratio of safe assets to deposits,
∗
evaluated at the optimal loan rate, ρ(r) = 1 − L(rL (r))/D. For r
above the critical value r, the loan risk channel amplifies the portfo-
lio adjustment channel. Below the critical value r, a lower risk-free
rate reduces the banker’s monitoring incentives, and the loan risk
channel weakens the portfolio adjustment channel.11 The slope of
the ratio of safe assets to deposits becomes less steep when r < r.
The reason is that, due to the counteracting loan risk channel, the
interest rate reduction required to achieve a given reduction in safe
assets (a given increase in loan issuance) becomes larger.
Reversal of Monetary Transmission. The loan risk channel
may not only weaken the portfolio adjustment channel; it can also
dominate it. In this case, a lower risk-free rate leads to an increase
in the loan rate and a reduction in the bank’s loan supply.
Proposition 2. The loan risk channel dominates the portfolio

dr ∗
adjustment channel, i.e., drL < 0, if and only if
∗
∂ q̂(rL , r) r
∗ > 1. (11)
∂r q̂(rL , r)
To understand the intuition behind Proposition 2, recall that,

on the one hand, a lower success probability makes loan issuance
relatively less profitable compared to holding safe assets, implying
that the bank cuts back its loan issuance when q is lower. On the
other hand, a lower risk-free rate reduces the return on safe assets
and makes holding safe assets less profitable. If the reduction in the
risk-free rate lowers the success probability and the profitability of
loans by more than the profitability of reserves, the bank prefers to
hold more safe assets, despite the lower risk-free rate. However, for
the profitability of loans to fall by more than the profitability of safe
11
Observe that condition (10) is only a sufficient condition. It does not rule
out the possibility that the safe asset effect dominates the deposit pass-through
effect at a higher level of the risk-free rate (above r̄). Whether such a case can
arise depends on the other properties of u(r) and L(rL ), such as the curvature
or magnitude of its rate of change.
assets, the banker’s monitoring must react strongly enough, i.e., a

reduction in r must lead to an overproportional reduction in q̂.
Proposition 3. If monitoring costs are sufficiently high, then the

loan risk channel dominates the portfolio adjustment channel if the
risk-free rate becomes sufficiently low: that is, for κ > κ, there exists
a critical value r̂ < r̄ such that
∗
drL
r < r̂ ⇔ < 0.
dr
The critical value r̂ is strictly increasing in the bank’s monitoring
cost κ.
Proposition 3 translates the condition in Proposition 2 into a

critical value for the risk-free rate. In particular, whenever the mon-
itoring costs are sufficiently high and the risk-free rate falls below
the critical rate, the elasticity of q̂ becomes sufficiently large so that
the loan risk channel becomes the dominant transmission channel
of monetary policy. As in Brunnermeier, Abadi, and Koby (2023),
below r̂, reductions in the risk-free rate are contractionary rather
than expansionary and r̂ constitutes a reversal rate.
Figure 4 illustrates Proposition 3. As in Figure 3, it plots the
required marginal deposit rate u (r) (dashed curve) against the ratio
of safe assets to deposits, ρ (solid curve). However, in Figure 4, we
assume that κ > κ, so that the reserves-deposit ratio becomes down-
ward sloping for values of r below the reversal rate r̂. Since the
reversal rate is equal to the point at which the loan risk channel just
offsets the portfolio adjustment channel, it follows that the reversal
rate must be strictly below the threshold r.
Figure 5 summarizes our results by illustrating how the trans-
mission of changes in the risk-free rate in our model depends on the
prevailing level of the risk-free rate.
Our analysis complements Brunnermeier, Abadi, and Koby
(2023) by showing that a reversal rate can arise as a consequence of
banks’ risk-taking behavior. The reversal rate in their model arises
due to a binding exogenous constraint on future profits, whereas the
reversal in our model is a consequence of the banker’s endogenous
risk choice that only exists if the banker increases her risk-taking suf-
ficiently strongly in response to a change in the risk-free rate. The
Figure 4. Reversal of Transmission when κ > κ
Note: For r ∈ (r̂, r), the loan risk channel weakens the portfolio adjustment
mechanism. For r < r̂, the loan risk channel dominates and monetary transmis-
sion reverses.
Figure 5. Monetary Transmission and the Risk-Free Rate
Note: The transmission of changes in the risk-free rate depends on the prevailing
level of the risk-free rate. For r > r, a lower risk-free rate, r, reduces risk-taking
and raises loan issuance. For r ∈ [r̂, r], a lower r raises risk-taking and weakens
loan issuance. For r < r̂, risk-taking is too strong and transmission into loans
reverses.
reversal rate in our model is just the most extreme manifestation of

the more general phenomenon that the loan risk channel weakens
monetary transmission for sufficiently low risk-free rates.
Implications of the Model. We use Propositions 1 and 3 to
derive several testable implications from our model.
Hypothesis 1. An increase in the banker’s exogenous deposit fund-

ing is associated with
• higher safe asset holdings, higher loan rates, and a lower loan
volume;
• higher bank risk-taking;
• weaker monetary policy transmission.
Hypothesis 1 follows because an increase in the deposit vol-

ume strengthens the safe asset effect compared to the deposit pass-
through effect. As a consequence, the threshold r̄ increases, and the
range of policy rates at which the loan risk channel weakens the
transmission via the portfolio channel becomes larger.
Hypothesis 1 is in line with recent empirical findings of
Jimenez et al. (2012), Miller and Wanengkirtyo (2020), and Bittner,
Rodnyansky, et al. (2021). Note first that an increase in deposits
leads to an increase in safe asset holdings, e.g., in the form of excess
reserves with the central bank or government bonds. Jimenez et al.
(2012) show that banks with more liquidity on their balance sheet
expand the issuance of loans less after a rate cut. However, they
do not distinguish between required reserves and excess reserves.
Miller and Wanengkirtyo (2020) show that, following a reduction in
the policy rate, banks with larger excess reserves extend lending to
riskier borrowers. Bittner, Rodnyanski, et al. (2021) further provide
evidence that in the presence of a zero lower bound on deposit rates,
banks that depend more on deposit funding and have greater expo-
sure to large-scale asset purchases lend relatively less and increase
their risk-taking more.
Hypothesis 2. The reversal rate is larger if, ceteris paribus,
• the bank has more deposits;

• the bank is riskier, and its loan portfolio is more costly to
monitor.
Hypothesis 2 follows from the effects of leverage, and the moni-

toring cost parameter, κ, on the reversal rate r̂ (cf. Proposition 3).
Higher deposits (and a larger cost parameter κ) exacerbate the
agency conflict and increase the banker’s risk-taking incentives.
Since higher risk-taking raises the loan rate for any value of r, the
reversal rate (at which the loan risk channel offsets the portfolio
channel) also increases.
Hypothesis 2 is in line with recent evidence by Arce et al. (2021),

who show that a negative correlation between policy rate and loan
rates can be found for banks that are poorly capitalized and whose
lending is riskier. Similarly, Basten and Mariathasan (2020) and
Miller and Wanengkirtyo (2020) find that lower policy rates are neg-
atively correlated with mortgage rates, but not with interest rates
on other types of loan. Hypothesis 2 is consistent with these findings
to the extent that mortgage handling is relatively more costly than
the origination and handling of other types of loans.
4. Extensions and Discussion
4.1 Insured Deposits

In this section, we consider how deposit insurance alters the trans-
mission of monetary policy via portfolio adjustment and loan risk
channels and the possibility of a transmission reversal. Suppose that
a share δ ∈ [0, 1] of deposits is insured at a flat rate normalized to
zero. For simplicity, insured depositors have the same outside option
as uninsured depositors.12
As before, we solve the model backward by first deriving the
banker’s optimal monitoring choice and thereafter the optimal loan
rate. The first-order condition for the monitoring choice is as in
Equation (3), except that we replace the deposit rate rD with the
average deposit rate r̄D which depends on the share of insured
deposits. As uninsured depositors rationally anticipate bank moni-
toring q̂, the average deposit rate is13
(δ q̂ + 1 − δ)u(r) − (1 − δ)(1 − q̂) rR

D
r̄D = .
q̂
Substituting r̄D into Equation (3) implicitly defines the banker’s

optimal monitoring q̂(rL , r, δ). Importantly, the condition for q̂ to
increase in r is the same as in Lemma 1,
12
Our results remain qualitatively unchanged if insured and uninsured
depositors have different outside options. We discuss this case in Appendix A.3.
13
For simplicity, we assume that, after default at maturity, the bank’s cash
flows from reserves are split on a pro rata basis among all depositors, insured
and uninsured.
∂ q̂(rL , r, δ)
> 0 ⇔ u (r) < ρ.
∂r
An increase in δ increases monitoring: ∂∂δq̂ > 0. This “char-
ter value effect” of deposit insurance is described in Cordella,
Dell’Ariccia, and Marquez (2018). Because the deposit rate is
given when the banker chooses her monitoring, a higher share of
insured deposits amounts to a greater implicit subsidy from the
deposit insurance, thereby reducing the repayments to depositors
and increasing the banker’s profits. As a consequence, higher deposit
insurance coverage strengthens monitoring incentives.14
The banker’s expected profit takes the same form as before in
Equation (7) except that the implicit subsidy from funding with a
share δ of insured deposits is added. Substituting the average deposit
rate and the optimal monitoring choice into the expected profits
yields
κq̂(rL , r)2
Π = q̂(rL , r)L(rL ) + rR − (u(r)D) − + S(δ, rL , r, R),
2
where S(δ, rL , r, R) ≡ δ(1 − q̂(rL , r, δ))(u(r)D − rR) is the implicit
subsidy from the deposit insurance. The subsidy is equal to the part
of insured deposit funding costs that has to be covered by the deposit
insurance in case of bank default. As can be seen from the expres-
sion for S(·), an increase in R reduces the implicit subsidy. This is
because the deposit insurance can rely on a larger amount of safe
assets to cover (part of) its liabilities in case the loans fail.
The transmission of monetary policy works as before through
the portfolio adjustment and loan risk channels. Since optimal mon-
itoring increases in the risk-free rate whenever the safe asset effect
dominates the deposit pass-through effect, the condition for the loan
risk channel to weaken monetary transmission remains formally the
same as in the benchmark model with δ = 0.
However, the presence of insured deposits changes the relative
importance of the portfolio adjustment and loan risk channels in
the transmission of monetary policy.
14
Cordella, Dell’Ariccia, and Marquez (2018, Proposition 1) show that the char-
ter value effect occurs if the share of deposit liabilities that are priced “at the
margin” is sufficiently small. This is the case for our specification because we
abstract from such deposits entirely.
Proposition 4. Given a share δ of insured deposits, the loan risk

dr ∗
channel dominates the portfolio adjustment channel, i.e., drL < 0,
if and only if
∗ ∗
∂ q̂(rL , r, δ) r q̂(rL , r, δ)δ
∗ > 1 + .
∂r q̂(rL , r, δ) 1−δ
Comparing Propositions 2 and 4 shows that the condition for the

dominance of the loan risk channel is stronger when the banker is
funded with insured deposits. The reason is that the banker obtains
a larger implicit subsidy from deposit insurance when she holds
fewer safe assets. This asset substitution motive provides an addi-
tional incentive for the banker to increase her loan issuance when the
risk-free rate falls. Thus, deposit insurance strengthens the portfolio
channel and alleviates the problem of transmission reversal. Simply
put, the deposit insurance subsidy mitigates the adverse effect of
lower rates on the bank’s profitability by increasing its profits.
Hypothesis 3. The reversal rate r̂ is smaller for banks that are

funded with a larger share of insured deposits. In the limit for δ → 1,
the reversal rate ceases to exist.
4.2 Endogenous Deposit Choice, Deposit

Shocks, and Liquidity Requirements
In this section, we briefly discuss the consequences of relaxing
Assumptions 2 and 3 for our main results. In the benchmark model,
the fixed amount of deposits (Assumption 2) determines the bank’s
balance sheet length and Assumption 3 implies that the bank holds
a strictly positive amount of safe assets whose rate of return, in con-
trast to the loan rate, cannot be controlled by the banker. These
assumptions ensure that the banker is exposed to the safe asset
effect so that reductions in the risk-free rate can reduce her expected
profits and lead to higher risk-taking and lesser loan issuance.
We now dispense with Assumption 2, i.e., we allow the banker
to endogenously choose the amount of deposits and we consider two
alternatives to Assumption 3. Under both alternatives, the banker
continues to be exposed to the safe asset effect. First, we consider
exogenous liquidity shocks to deposits, i.e., exogenous inflows and
outflows to and from the depositors’ accounts that randomly change

the bank’s end-of-period safe asset holdings. Second, we consider
an exogenously imposed liquidity requirement, akin to the Basel
regulations’ liquidity coverage ratio (LCR).
It is worth pointing out that in the absence of these or other alter-
native assumptions, the banker in our model would have no incen-
tives to hold safe assets. Absent the safe asset effect, the loan risk
channel would always work into the same direction as the portfolio
adjustment channel.
Deposit Shocks. We begin by considering a variant of the
model where the bank can choose its deposits at the beginning of
date 0. However, depositors are subject to a liquidity shock at the
start of date 1, i.e., they face inflows and outflows to and from their
deposit accounts. We assume that the bank cannot invest additional
deposits in t = 1 into loans so that deposit flows must be balanced
by an equivalent change in safe assets.15
To rule out precautionary motives for holding safe assets, we
assume that the bank can access the central bank’s standing deposit
and lending facilities at an interest rate r to deposit excess reserves
or to cover deposit outflows and reserve shortfalls.16
Furthermore, we assume that the bank can also borrow ex ante
from the central bank up to a fraction σ ∈ (0, 1) of its loan issuance,
i.e., we impose R ≥ −σL.
Liquidity shocks, denoted x, are proportional to deposits, D,
and are drawn from a continuous distribution F (·) and with density
f (·) over support [−1, z]. zD is the maximal inflow to an individual
deposit account. We assume that the liquidity shocks realize after
the bank has contracted the deposit rate, set its loan rate, and has
chosen the optimal monitoring effort. Without loss of generality, we
set E[x] = 0.
As before, we solve the model backwards. Given rD , the optimal
monitoring of the bank is still determined by Equation (3). The ran-
dom inflows and outflows to deposit accounts affect the deposit cost
15
Inflows to deposit accounts automatically add to the bank’s central bank
reserves. Outflows from deposit accounts need to be covered by running down
reserves or by additional borrowing from the central bank.
16
Allowing for a symmetric interest rate corridor around the main policy rate
by making the standing borrowing rate higher than the standing deposit rate
would complicate our analysis without altering the main results.
of the bank. In particular, if the bank is solvent, depositors receive

rD on their entire deposit holdings at maturity. With probability
1 − q, the bank defaults. In this case, depositors obtain a pro rata
share of the remaining assets. The expected repayment to depositors
must be equal to their outside option u(r) such that
z
(1−q̂)r max{R+xD,0}dF (x)
u(r) − −1
D
rD = . (12)
q̂
By substituting rD into Equation (3), we can solve for the bank’s

monitoring choice q̂(rL , D; r). The partial effects of rL , r, and D on
the banker’s optimal monitoring q̂ reflect the effects of these vari-
ables on her expected profits. As before, the effects of rL and r are
ambiguous, with the respective conditions now taking into account
expected deposit flows.17 However, the effect of D on q̂ is unambigu-
∂ q̂
ously negative, i.e., ∂D < 0. This is because u(r) ≥ r so deposits
are relatively more expensive than borrowing from the central bank
(cf. Assumption 1).
Lemma 2. The bank chooses a strictly positive loan issuance L∗ (r).

Given Assumption 1, the bank minimizes its funding cost by choosing
R∗ = −σL∗ and D∗ = (1 − σ)L∗ .
Lemma 2 shows that the bank borrows from the central bank on
a permanent basis as long as this is feasible (i.e., if σ > 0). Even
though the bank does not hold a positive level of reserves ex ante,
the possibility that it ends up with a positive reserve balance due to
random deposit inflows implies that the loan risk channel can still
mitigate and even dominate the portfolio channel.
Proposition 5. The loan risk channel dominates the portfolio

adjustment channel, i.e., dr
dr < 0, if and only if
L

∗ σ
∗
∂ q̂(rL , r) r q̂(r L , r)F 1−σ
∗ > 1+ . (13)
∂r q(rL , r) 1 − F 1−σ σ
17
See the appendix for details.
As Proposition 5 shows, the condition for the loan risk channel

to dominate the portfolio adjustment channel is similar to condi-
tion (11) when deposits are exogenous. The difference is that con-
dition (13) depends on the probability that the bank ends up with
safe asset holdings due to inflows into its depositors’ accounts.
Inflows to deposits reflect the amount of safe assets that the bank
cannot adjust optimally ex ante. Therefore, the probability of ending
up with a positive balance can be interpreted as a measure of the ease
with which the banker can adjust her safe assets. At one extreme,
if the probability of an inflow of deposits becomes negligibly small,
F (σ/(1 − σ)) ≈ 1, then condition (13) could never hold. In this case,
the loan risk channel and the portfolio adjustment channel always
go in the same direction and a reversal rate cannot exist. As is the
case for a fully levered bank in Dell’Ariccia, Laeven, and Marquez
(2014, Proposition 3), a lower risk-free rate leads to less risk-taking.
On the contrary, if the bank would almost surely obtain a deposit
inflow, i.e., F (σ/(1 − σ)) → 0, then condition (13) converges to our
benchmark condition (11).
Condition (13) further allows us to illustrate the effect of perma-
nent central bank lending programs on the existence of the reversal
rate.
Hypothesis 4. The reversal rate becomes smaller if the central bank

∂ r̂
is willing to fund a larger share of the banker’s lending, i.e., ∂σ < 0.
In the limit, for σ → 1, the reversal rate ceases to exist.
Consider the extreme case where the bank can finance its entire
loan portfolio by borrowing from the central bank ex ante, i.e.,
lim σ → 1. In this case, a reversal rate would cease to exist.18 This
case is similar to the case with full deposit insurance, δ = 1, in
Section 4.1. The entire risk of the bank’s loan issuance and the
bank’s exposure to interest rate risk would be borne by the cen-
tral bank, and lower policy rates would unambiguously increase the
bank’s profit.19
18
The right-hand side of Equation (13) converges to ∞, while the left-hand
side assumes a finite value, implying that the condition could never be satisfied.
19
We abstract from the possibility that the central bank can risk-adjust its
interest rate when lending to the banker. In practice, central banks are able to
Binding Liquidity Requirement. Next, instead of Assump-

tions 2 and 3, we assume that the banker can endogenously choose
her deposits, but she is required to hold a certain fraction of her
deposits in the form of safe and liquid assets, e.g., reserves with
the central bank or government bonds. This requirement is akin to
the LCR that banks must satisfy under Basel III regulations (see
Brunnermeier, Abadi, and Koby 2023 for a similar assumption). As
we now show, under a binding liquidity requirement, the portfo-
lio adjustment channel and the loan risk channel always move into
the same direction. However, they both switch sign once the safe
asset effect dominates the deposit pass-through effect. The banker’s
liquidity requirement can be written as
R ≥ ρD,
where ρ is now the exogenously given regulatory liquidity ratio. To
the extent that u(r) ≥ r, the liquidity requirement is binding. Since
the expected profits are strictly decreasing in D, the banker mini-
mizes the amount of deposits. Combining the liquidity requirement
and the funding constraint yields
L
= D.
1−ρ
Substitution into the banker’s profits yields
rD − ρr κq 2
Π = q rL − L(rL ) − .
1−ρ 2
Because of the binding liquidity requirement, the direction of
the portfolio adjustment channel also depends on the relationship
between deposit pass-through and safe asset effect (like the loan
risk channel). Thus, compared to Equation (9), the two channels are
perfectly aligned, and we have
∗ ∗ ∗ ∗
drL ∂rL ∂rL ∂ q̂(rL , r) u (r) − ρ
= + × ∝ .
dr ∂r
∂q ∂r 1−ρ

portfolio adjustment loan risk channel
channel
(14)
achieve some degree of risk adjustment by lending against collateral and applying
haircuts to riskier asset classes.
Figure 6. Reversal Rate with a

Binding Liquidity Requirement
Note: For r > r, the portfolio adjustment and the loan risk channel are positive,
while they become negative for r < r.
Proposition 6. Under a binding liquidity requirement, a rever-

sal of monetary transmission occurs whenever the safe asset effect
dominates the deposit pass-through effect. The reversal rate is
r̂ = r̄, where r̄ satisfies u (r) = ρ.
A higher liquidity requirement weakens the transmission of monetary

policy.
Figure 6 illustrates the case of a binding liquidity requirement.

Because ρ is now exogenously determined, the solid curve is flat at
the level set by regulation. For r < r, the safe asset effect dominates
and reverses both the loan risk and the portfolio adjustment chan-
nel, i.e., a further reduction in the risk-free rate leads to a higher
loan rate and a reduced loan issuance.
The case of a binding liquidity requirement allows us to empha-
size a potential cost associated with liquidity requirements. The lit-
erature usually discusses the direct costs of liquidity requirements,
i.e., the opportunity cost of foregoing profitable investments when a
larger share of deposits is held in the form of more liquid but less
profitable assets. Our model reveals another indirect cost of liquid-

ity requirements, namely the costs that arise from the impairment of
the monetary transmission channel. As can be seen in Equation (14),
∗
an increase in ρ reduces the effect of r on rL . Moreover, once ρ is
sufficiently high, monetary transmission is reverted.
4.3 Depositors’ Outside Option

Finally, let us briefly discuss how the results in our model depend
on Assumption 1. The crucial element of Assumption 1 is the lower
bound on the outside option, whereas the additional assumptions
(convexity of u(r) and continuity of u (r)) are technical and imposed
for the sake of tractability. Consider the following example where
depositors, instead of holding deposits, could either invest into a
risk-free bond that pays r at date 1 or hold cash which provides a
per-unit convenience yield θ and requires a per-unit storage cost ζ.
Thus, u(r) = max{1 + θ − ζ, r} and u (r) = 1 − 1[r<1+θ−ζ] . For
this specification of u(r), the convexity and continuity assumptions
(u (r) > 0 and limr↓1+θ−ζ = 0) fail to hold. Because ρ ∈ (0, 1),
it follows that ρ > u (r) if and only if r < 1 + θ − ζ. Since u (r)
jumps discontinuously at 1 + θ − ζ, the critical r at which the
safe asset effect dominates the deposit pass-through effect equals the
lower bound of the outside option. Hence, the safe asset effect can
dominate the deposit pass-through effect for small values of r even
without the continuity and convexity imposed by Assumption 1.20
What about the possibility that deposits themselves provide a
convenience yield so that u(r) < r? Because the main results in the
paper depend on the marginal costs and benefits of deposits versus
reserve assets, allowing for a convenience yield on deposits such that
u(r) < r for r > u−1 (u) would leave the results in Propositions 1–3
unaffected.21
20
The continuity and convexity assumptions ensure that r ≥ u−1 (u) and that
at r = r there is no discontinuity so deposit pass-through and safe asset effect
are balanced at the margin.
21
Changing the ordering between the outside option and the risk-free rate
would, however, change the implications derived in Hypothesis 1. In this case,
an exogenous increase in deposits would increase the banker’s expected profit
and therefore increase her incentives to monitor and lead her to issue more loans.
Note, however, that u(r) < r would allow the bank to make a risk-free profit from
The key element in Assumption 1 is the assumption that the out-

side option of depositors cannot fall below u. Suppose we dispense
with this assumption and set u(r) = r for all values of r, which is a
standard assumption in the corporate finance and banking literature,
e.g., Dell’Ariccia, Laeven, and Marquez (2014) and Martinez-Miera
and Repullo (2017). Because ρ ∈ (0, 1) while u (r) = 1, the deposit
pass-through effect dominates the safe asset effect for all values of r
and a lower interest rate always leads to an increase in monitoring,
∂ q̂/∂r < 0. This mirrors the effect of r on the bank’s risk-taking
incentives for the case of sufficiently high leverage in Dell’Ariccia,
Laeven, and Marquez (2014, Proposition 3). As a consequence, the
loan risk channel always amplifies the portfolio adjustment effect
and lower interest rates unambiguously lead to a lower loan rate and
higher loan issuance. This argument shows that the lower bound on
the depositors’ outside option is a key condition for the weakening
of monetary transmission via the loan risk channel.
5. Conclusion
This paper argues that the empirically observed correlation

between weaker monetary transmission and higher risk-taking in an
environment of low interest rates (e.g., Miller and Wanengkirtyo
2020 or Arce et al. 2021) can be viewed as the consequence of an
agency friction between banks and their depositors.
The main contributions of our paper are two. First, we show
that lower policy rates lead banks to increase risk-taking when
the pass-through to deposit rates is too small to compensate for
the reduction in the profitability of banks’ safe assets. Higher risk-
taking, in turn, leads to a weakening of the monetary transmission
because it induces banks to optimally raise loan rates and issue fewer
loans.
Second, our model complements Brunnermeier, Abadi, and Koby
(2023) by showing an alternative mechanism by which a reversal
of monetary transmission can arise. The existence of a reversal
issuing deposits. To the extent that the bank could control the level of deposits
(as in Section 4.2), the banker would issue as much deposits as possible.
rate depends on banks’ characteristics (insured deposits, monitor-

ing technology, leverage). Our model emphasizes that the reversal
of monetary transmission is only an extreme manifestation of the
more general phenomenon of weakened transmission due to higher
risk-taking incentives in a low interest rate environment. This phe-
nomenon should be of concern to central banks and may require
them to devise policies that address the underlying causes of weaker
transmission.
Our model suggests two policy implications that could help to
alleviate the problem of weaker transmission. First, when operat-
ing in an environment with high excess reserves, central banks could
implement reserve remuneration schemes that boost profits of banks
holding excess reserves. While such schemes redistribute seignior-
age revenues back to banks, they could nevertheless strengthen the
transmission in an environment with protracted excess reserves and
render monetary policy more effective. In this sense, our model pro-
vides a rationale for the two-tiered remuneration for excess reserves
by the Eurosystem, which seeks to mitigate the effect of negative
interest rates on bank profitability.22
Second, even though we abstracted from explicitly considering
the effect of bank equity and capital regulation, our model can also
speak to a recent debate on the importance of bank capitalization
for monetary policy. In the context of our standard agency model, a
capital requirement would weaken the link between loan rates and
monitoring incentives. Put differently, an increase in loan risk would
have a relatively smaller effect on the optimal loan rate if the bank
must satisfy a larger capital requirement. As a consequence, a higher
capital requirement would reduce the relative weight of the loan
risk channel and strengthen monetary transmission via the portfolio
adjustment channel. For banks with a smaller leverage, the reversal
rate would be lower and the range of interest rates where trans-
mission is unimpeded would be larger. Our model, therefore, echoes
arguments by Gambacorta and Shin (2018) or Darracq Pariès, Kok,
and Rottner (2020) who argue that bank capital matters not only
for the central bank’s financial stability but also for its monetary
policy mandate and for the transmission of monetary policy.
22
https://www.ecb.europa.eu/mopo/two-tier/html/index.en.html.
Appendix
A.1 Proofs
Proof of Lemma 1. Maximizing expected profits for a given deposit
rate rD with respect to q yields the first-order condition
rL L − rD D + rR − κq = 0.
By substituting rD from the participation constraint, we can

obtain q̂ as the solution to the following implicitly defined function:
u(r)D − rR
φ(q, rL , r) ≡ rL L − − κq = 0.
q
The latter is quadratic in q. Following Allen, Carletti, and Marquez

(2015), we take the larger of the two roots, such that
∂φ u(r)D − rR
= − κ < 0.
∂q q2
Moreover, we have
∂φ R − u (r)D
=
∂r q
and, using the fact that R = D − L(rL ),
∂φ r
= rL L (rL ) + L(rL ) − L (rL ).
∂rL q
An application of the implicit function theorem yields the expres-

sions for ∂ q̂/∂rL and ∂ q̂/∂r.
Proof of Proposition 1. From Equation (7), the first-order condition
for the optimal loan rate is given by
dΠ
= q̂(rL , r) (rL L (rL ) + L(rL )) − rL (rL )
drL
∂ q̂
+ (rL L(rL ) − κq̂) = 0
∂rL
=(u(r)D−rR)/q
r
= q̂(rL , r) rL L (rL ) + L(rL ) − L (rL )
q̂(rL , r)
u(r)D − rR
× 1− = 0.
u(r)D − rR − q̂ 2 κ
∗
q̂ and the second bracket are positive, so that the optimal rL satisfies
condition (8) in the text.
∗
The second-order condition, evaluated at the critical point rL ,
23
becomes
r L (rL
∗ ∗
)L(rL )
rL L (rL
∗
) + 2L (rL
∗
) − L (rL
∗
)=− ∗ + 2L (rL
∗
) < 0,
q̂ L (rL )
which is satisfied since L(·) is a decreasing and concave function.
∗
Thus, rL maximizes the bank’s profits.
Applying the implicit function theorem to the first-order condi-
∗
tion evaluated at rL yields
∗
drL ∂r∗ ∂r∗ dq̂
= L+ L
dr ∂r ∂q dr
L (rL ∗
+ q̂r2 L (rL
∗ ∂ q̂
)
− q̂ ) ∂r L (rL∗
) ∂ q̂ r
= − L (r∗ )L(r∗ ) ≷0 ⇔ − 1− ≷ 0,
− LL(r∗ ) L + 2L (rL ∗) q̂ ∂r q̂
L
(A.1)
dq̂ ∂ q̂ ∂ q̂
where we replaced dr with ∂r because ∂r L
= 0 when evaluated at
∗
rL = r L .
Equation (A.1) implies that the loan risk channel weakens the
portfolio channel whenever ∂ q̂/∂r > 0, which is equivalent to
u (r) < ρ (cf. Lemma 1).
Next, we show the existence of a value r̄ such that for all r < r̄, we
must have u (r) < ρ. Note that by Assumption 3, for all r < u−1 (u)
∗
we have ρ = R/D = 1 − L(rL (r))/D > 0 = u (r). If r becomes suffi-
ciently large, R converges to a positive and finite value, while u (r)
diverges (because of the strict convexity of u(·) for r > u−1 (u)) so
that we have u (r) > ρ for sufficiently large r. Thus, there exists a
smallest value r such that u (r) = ρ. For all r < r, we have u (r) < ρ.
23
Note that the partial effect of rL on q̂ is irrelevant for determining the sign
∗
of the second-order condition since ∂ q̂/∂rL = 0 when evaluated at rL .
Thus, for r < r, we have u (r) < ρ and, as a consequence of

Lemma 1, ∂ q̂/∂r > 0. From Equation (A.1) follows that the loan
risk channel weakens the transmission via the portfolio channel for
r < r.
Proof of Proposition 2. The proof follows immediately from
Equation (A.1):
∗
drL ∂ q̂ r
<0 ⇔ 1< .
dr ∂r q̂
Proof of Proposition 3. We show the existence of r̂ that satisfies

∗
∂ q̂(rL , r̂) r̂
1= ∗ , r̂) .
∂r q̂(rL
From the proof of Lemma 1 it follows that
∂ q̂ r (u (r)D − R)r
= ≥ 1 ⇔ (u(r) − u (r)r)D ≥ κq̂ 2 .
∂r q̂ u(r)D − rR − q̂ 2 κ
The last inequality requires that u (r)D ≤ R since u(r)D−rR < κq̂ 2
(cf. Lemma 1). Therefore, consider u (r) < ρ. Since u (r) > 0, the
left-hand side of the above inequality is strictly decreasing in r. Since
u (r) = 0 for r < u−1 (u), we have argmaxr {u(r) − u (r)r} = u−1 (u).
Thus, a necessary and sufficient condition for the existence of a
reversal rate is that κ satisfies κq̂ 2 ≤ uD. Note further that
dκq̂(κ)2 q̂ 2
= u(r)D − rR + κq̂ 2 < 0.
dκ u(r)D − rR − q̂ 2 κ
Thus, we can find a value κ such that κq̂(κ)2 = uD and where κ
also satisfies the condition for ∂φ/∂q < 0 in the proof of Lemma 1.
Since (u(r) − u (r)r̂)D − κq̂ 2 is strictly decreasing in r for r ≤ r,
there exists r̂ < r such that for κ ≥ κ
(u(r̂) − u (r̂)r̂)D − κq̂ 2 = 0. (A.2)
For r < r̂, we have
∂ q̂ r
(u(r) − u (r)r)D > κq̂ 2 ⇔ > 1.
∂r q̂
Proof of Hypothesis 1. We consider an exogenous increase in the

deposit volume and show that this leads to an increase in excess
reserves, a higher reserves-deposit ratio, and less lending.
The equilibrium effect of an increased D follows by applying the

implicit function theorem to the two equilibrium conditions
u(r)D − r(D − L(rL ))

rL L(rL ) − − κq = 0,
q
rL (rL )
rL L (rL ) + L(rL ) − = 0.
q
Let J ∗ denote the Jacobian of the above system of two equations

evaluated at the optimum. From the proofs of Lemma 1 and Propo-
sition 1 follows that J ∗ < 0 (when the variable vector is (rL , q)).
Note further that the second equation is independent of D. Thus,
by the implicit function theorem
dq ∗ u(r) − r ∗
drL u(r) − r
∝− <0 and ∝ > 0.
dD q∗ dD q∗
Since rL increases in D, a higher D leads to less lending and

higher excess reserves
dL∗ ∗ drL
∗ ∗
drL
= L (rL ) < 0 and dR = dD − L (rL
∗
) > 0.
dD dD dD
Finally, note that the latter implies also a higher reserves-deposit
∗
ratio ρ because ρ < 1 and L (rL
∗ drL
) dD < 0 such that we obtain
∗
dρ 1 drL
= 1 − ρ − L (rL
∗
) > 0.
dD D dD
Proof of Hypothesis 2. Applying the implicit function theorem to

Equation (A.2) yields
dκq̂ 2
∂ r̂ dκ
= >0 and
∂κ −ru (r)D − 2κq̂ ∂r
∂ q̂
∂ r̂ −(u(r) − u (r)r) + 2q̂κ ∂D

∂ q̂
= > 0.
∂D −ru (r)D − 2κq̂ ∂r
∂ q̂
Proof of Proposition 4. q̂ is given by the solution to the following

implicit function:
1−δ
φ(q, rL , δ, r) ≡ rL L(rL ) − δ + (u(r)D − rR) − κq = 0,
q
with
∂φ 1−δ
= (u(r)D − rR) − κ < 0,
∂q q2
∂φ (R − u (r)D)(qδ + (1 − δ))
= > 0 ⇔ ρ > u (r),
∂r q2
∂φ δq + (1 − δ)
= rL L (rL ) + L(rL ) − rL (rL ),
∂rL q
and
∂φ 1−q
= (u(r)D − rR) > 0.
∂δ q
Given q̂, the first-order condition for the banker’s optimal loan
rate is given by
δq + (1 − δ
q̂ rL L (rL ) + L(rL ) − rL (rL )
q
(q̂δ + (1 − δ))(u(r)D − rR)
× 1− = 0.
(1 − δ)(u(r)D − rR) − κq̂ 2
Since the second bracket is strictly positive, the optimal loan rate
satisfies
δq + (1 − δ)
rL L (rL ) + L(rL ) − rL (rL ) = 0.
q
Application of the implicit function theorem yields
∗
drL δ q̂ ∂ q̂ r
∝ −(1 − δ)L (rL ) 1 + − .
dr 1 − δ ∂r q̂
∗
drL δ q̂ ∂ q̂ r
Thus, dr < 0 if and only if 1 + 1−δ < ∂r q̂ .
Proof of Hypothesis 3. From the proof of Proposition 4 it follows

that the reversal rate r̂(δ) is given by the solution to
∂ q̂ r δ q̂
−1− = 0.
∂r q̂ 1−δ
Using the expressions for ∂ q̂/∂r, we can rewrite the latter as
u(r)D − δrR − (1 − δ)u (r)rD − κq̂ 2 = 0. (A.3)
For δ = 0, the above condition is equal to Equation (A.2), imply-

ing that r̂(δ) converges to the value of the reversal rate in Propo-
sition 3. Another application of the implicit function theorem to
Equation (A.3), taking into account that for r = r̂ we have u (r) < ρ
and ∂R/∂r = 0, implies ∂∂δr̂ < 0.
Note further that for δ → 1, Equation (A.3) cannot be satisfied
since q̂ is the larger root, which implies that κq̂ 2 − u(r)D + rR > 0.
Hence, for δ → 1, the reversal rate ceases to exist.
Proof of Lemma 2 and Proposition 5. The adjusted profit function

becomes
z
κq 2
Π = q rL L(rL ) + r (R + x D) dF (x) − rD D − .
−1 2
Because E[x] = 0, we can simplify to the same profit function as in

our baseline model,
κq 2
Π = q (rL L(rL ) + r R − rD D) − .
2
Inserting the participation constraint into the first-order condi-

tion for q implicitly defines the function q̂(rL , D, r)
φ(rL , D, r) = rL L(rL ) + r R
−ρ
u(r)D − (1 − q)r −1
(R + xD) dF (x)
− − κq = 0.
q
Taking the larger of the two roots, we obtain

z
∂φ
= qR − u D + (1 − q) (R + xD) dF (x)
∂r −ρ
−ρ

= R − (1 − q) (R + xD) dF (x) − u D
−1
and
z
∂φ
= q((rL − r)L + L) − (1 − q)rL dF (x)
∂rL −ρ
−ρ

= q(rL L + L) − rL + (1 − q)rL dF (x),
−1
as well as
z
∂φ
= qr − u(r) + (1 − q)r (1 + x) dF (x)
∂D −ρ
−ρ
= r − u(r) − (1 − q)r (1 + x) dF (x) < 0,
−1
which is unambiguously negative for r ≤ u(r).

The first-stage profit function, given the required return for the
expected equilibrium monitoring choice, becomes
Π(rL , D; r) = q̂(rL L(rL ) + r R) − u(r)D

−ρ
q̂ 2
− (1 − q)r (R + xD) dF (x) − κ .
−1 2
Differentiating with respect to D and rL yields the first-order condi-
tions for a profit maximum. As the bank optimally minimizes deposit
costs, we evaluate the first-order condition at D∗ = (1 − σ)L∗ (rL )
and R∗ = −σL∗ (rL ), such that ρ = − 1−σ σ
.
Using the implicit function theorem we obtain
1−σ
σ
−L + (1 − q̂)L −1 dF (x))

σ
1−σ
drL + r L L + L − rL −1 dF (x) ∂ q̂
∂r
=− ∂2Π
.
dr ∂r 2 L
∗
For rL to be the optimal loan rate in equilibrium, we must have
2
∂ Π
∂r 2
< 0. Therefore, dr
dr < 0 if and only if the numerator is negative.
L
L
∂Π
Using the first-order condition ∂rL = 0, we can simplify to
σ
σ
r 1−σ ∂ q̂ 1−σ r ∂ q̂
1− dF (x) > 1 − (1 − q̂) dF (x) ⇔
q̂ −1 ∂r −1 q̂ ∂r
1−σ
σ
1 − (1 − q̂) −1 dF (x)
> 1−σ
σ ,
1 − −1 dF (x)
which corresponds to the condition in Proposition 5. Note that as

σ → 1, the left-hand side approaches zero and the right-hand side q̂
such that the condition can never be fulfilled. If the bank can fund
all loans by borrowing from the central bank, reversal rate cannot
exist.
Proof of Hypothesis 4. The reversal rate r̂ is implicitly defined by
σ

r̂ ∂ q̂ F ( 1−σ )
ψ(r̂, σ) ≡ − 1 − q̂ = 0.
q̂ ∂r 1 − F ( 1−σ
σ
)
∂ r̂ ∂ψ/∂σ
By the implicit function theorem, ∂σ = ∂ψ/∂r < 0, because
σ
F ( 1−σ )
σ
1−F ( 1−σ ) strictly increases in σ as the distribution function F (·),
is an increasing function and at r = r̂, we have ∂ψ/∂r < 0.
Proof of Proposition 6. The banker’s optimal monitoring choice is
the same as in the benchmark model, i.e., q̂ is given by the implic-
itly defined function q̂(rL , r). Substituting q̂ and the deposit rate
into the expected profits yields
u(r) − ρr κq̂ 2
Π = q̂rL L(rL ) − L(rL ) − .
1−ρ 2
The first-order condition determining the bank’s loan issuance is

given by
(u(r) − ρr) (u(r) − ρr)L(rL ) ∂ q̂

q̂rL − L (rL ) + q̂L(rL ) + = 0.
1−ρ q̂ ∂rL
Using the expression for ∂ q̂/∂rL implies that the optimal loan
∗
rate rL must satisfy
∗ (u(r) − ρr)
rL − L (rL
∗ ∗
) + L(rL ) = 0.
q̂(1 − ρ)
The second-order sufficient condition is satisfied when evaluated at
∗
rL . Totally differentiating the first-order condition yields
L (rL ) (u(r)−ρr)L(rL )
∗
1+
q̂
u (r) − ρ
(u(r)−ρr)L(rL )
drL q̂ κ− q̂ 2
= · .
dr ∂2Π
∗
∂rL
1−ρ
Since the term multiplying (u (r) − ρ)/(1 − ρ) is strictly positive, it

follows immediately that
∗
drL
≷ 0 ⇔ u (r) ≷ ρ ⇔ r ≷ r̄,
dr
where r̄ solves u(r) = ρ.
A.2 Proportional Monitoring Cost

This section shows that the key result in Proposition 2, i.e., that
∗
drL ∂ q̂(rL , r) r
< 0 ⇔ ∗ , r) > 1
dr ∂r q̂(rL
remains unchanged if monitoring costs are proportional to loan
issuance, i.e.,
κ
c(q, rL ) = q 2 L(rL ).
2
To show this, we derive the first-order condition determining the
bank’s optimal monitoring effort:
rL L(rL ) − rD D + r(D − L(rL )) − cqL(rL ) = 0.
The optimal monitoring effort q(r, rL ) is implicitly defined by the
first-order condition after substituting for rD :
u(r)D − r(D − L(rL ))
rL L(rL ) − cqL(rL ) − = 0.
q
Application of the implicit function theorem yields
∂ q̂(rL , r) rL L (rL ) + L(rL ) − rq L (rL ) − cqL (rL )

= ≷ 0,
∂rL cL(rL ) − u(r)D−rR
2 q
and
∂ q̂(rL , r) R − u (r)D
= ≷ 0.
∂r cL(rL ) − u(r)D−rR
q 2
Given q̂(rL , r), the bank maximizes its profits by choosing the
loan rate rL . The profit function is given by
c
q̂(rL , r)rL L(rL ) + r(D − L(rL )) − u(r)D − q̂(rL , r)2 L(rL ).
2
Differentiating with respect to rL yields the first-order condition
∗
that pins down rL :
r c
q(rL , r) rL L (rL ) + L(rL ) − L (rL ) − q(rL , r)L (rL )
q(rL , r) 2
∂q(rL , r)
+ (rL L(rL ) − cq(rL , r)L(rL )) = 0.
∂rL
Dividing the latter equation by q̂ and adding and subtracting
cq̂L /2, we obtain
r
rL L (rL ) + L(rL ) − L (rL ) − cq(rL , r)L (rL )
q(rL , r)
∂q(rL , r) 1 c
+(rL L(rL ) − cq(rL , r)L(rL )) + q(rL , r)L (rL ) = 0.
∂rL q̂rL , r) 2
Using the first-order condition for monitoring to replace

rL L(rL ) − cq̂, we obtain
r
rL L (rL ) + L(rL ) − L (rL ) − cq(rL , r)L (rL )
q(rL , r)
u(r)D − r(D − L(rL )) ∂ q̂(rL , r) c
+ + q(rL , r)L (rL ) = 0.
q(rL , r)2 ∂rL 2
Substituting the expression for ∂ q̂/∂rL and collecting terms, we

finally obtain
r
cL(rL ) rL L (rL ) + L(rL ) − L (rL ) − cq(rl , r)L (rL )
q
c u(r)D − r(D − L(rL ))
+ q(rL , r)L (rL ) cL(rL ) − = 0.
2 q(rL , r)2
The second-order condition is strictly negative when evaluated at
∗
the critical point rL that satisfies the latter equation. Thus, from the
∗
implicit function theorem follows that the sign of drL /dr is equal to
the sign of the derivative of the first-order condition with respect to
r, i.e., we have (for simplicity, we have dropped the arguments from
functions q̂ and L):
∗
drL cLL cL u (r)D − (D − L)
∝ − −
dr q̂ 2 q̂
cLL r c2 LL cL u(r)D − r(D − L) ∂ q̂
+ − +
q̂ 2 2 2 q̂ 2 ∂r
cLL cL u(r)D − r(D − L) ∂ q̂
= − − cL −
q̂ 2 q̂ 2 ∂r
cLL r c2 LL cL u(r)D − r(D − L) ∂ q̂
+ − +
q̂ 2 2 2 q̂ 2 ∂r
c ∂ q̂ r
= − LL 1 − ,
q̂ ∂r q̂
∂ q̂
where the second line follows from using the expression for ∂r from
above. Since −cL(rL )L (rL )/q̂(rL , r) > 0, it follows that
∗
drL ∂ q̂ r
< 0 ⇔ > 1,
dr ∂r q̂
which is the same condition as in Proposition 2 where monitoring
costs are independent of L(rL ).
∗
Note, while the condition for the marginal effect of r on rL
remains unchanged, a comparison of the respective first-order con-
ditions shows that if the monitoring costs are proportional to loan
issuance, the bank issues fewer loans (sets a higher loan rate) to
reduce the monitoring costs.
A.3 Different Outside Options for Insured Depositors

In the main text, we assume that uninsured and insured depositors
have the same outside option u(r). Here, we show that Proposition 4
and Hypothesis 3 remain unchanged even if insured depositors have
a different outside option. To this end, let uI I(r) denote the outside
option of insured depositors. The outside option of the uninsured
depositors remains denoted by u(r).
The first-order condition for the banker’s monitoring choice is
given by
rL (L(rL ) + rR − (δuI (r) + (1 − δ)rD )D − κq = 0,
where rD denotes the interest rate on uninsured debt. Substituting
Equation (5) for rD into the first-order condition yields the implicit
function for q̂(rL , r):
φ(q̂, rL , r) ≡ rL (L(rL ))
(δ q̂uI I(r) + (1 − δ)u(r))D − (q̂ + (1 − δ)(1 − q̂))rR
−
q̂
− κq̂ = 0.
Again, choosing the larger root for q̂, we have ∂φ

∂ q̂ ≡ φq̂ < 0 and
∂φ r
≡ φrL = rL L (rL + L(rL ) − (q̂ + (1 − δ)(1 − q̂)) L (rL ).
∂rL q̂
Given the implicitly defined function q̂(rL , r), we next turn to the
banker’s optimal choice of rL . Substituting the uninsured deposit
rate and q̂ into the profit function, we obtain
π(rL ) = q̂rL L(rL ) + (q̂ + (1 − q̂)(1 − δ))rR
κ
− (δuI (r) + (1 − δ)u(r))D − q̂ 2 .
2
The first-order condition for rL is given by
r
π (rL ) = q̂ rL L + L − (q̂ + (1 − δ)(1 − q̂)) L
q̂
∂ q̂
+ (rL L + δ(rR − uI (r)D) − κq) = 0.
∂rL
Substituting from the first-order condition for effort choice,

(δ q̂uI (r) + (1 − δ)u(r))D − (q̂ + (1 − δ)(1 − q̂))rR
rL L − κq̂ = ,
q̂
and the expression for ∂ q̂/∂rL , it follows that the optimal loan rate
∗
rL must solve
r
rL L (rL ) + L(rL ) − (q̂ + (1 − δ)(1 − q̂)) L (rL ) = 0.
q̂
As the second-order condition for a profit maximum must be
negative, it follows from the implicit function theorem that
∗
drL (q̂ + (1 − δ)(1 − q̂))
< 0 ⇔ − L (rL )
dr q̂
r (q̂ + (1 − δ)(1 − q̂))r ∂ q̂
+ − L (rL ) + L (rL ) < 0.
q̂ q̂ 2 ∂r
Rewriting the latter equation yields
∗
drL (1 − δ) r ∂ q̂ q̂δ
< 0 ⇔ − L (rL ) − 1+
dr q̂ q̂ ∂r 1−δ
r ∂ q̂ δ q̂
< 0 ⇔ > 1+ ,
q̂ ∂r 1−δ
which is the same condition as in Proposition 4.
However, note that while the above condition is the same as in
the main text, the magnitude of the thresholds r̄ and r̂ changes
compared to the model with identical outside options. To see this,
consider the sign of

uI (r)
∂ q̂ δ q̂
u (r) + (1 − δ) u (r)
> 0 ⇔ ρ > .
∂r δ q̂ + (1 − δ)
It follows that the relative deposit pass-through, i.e., uI (r)/u (r),
determines whether or not the threshold rates r̄ and r̂ change com-
pared to the baseline model. Whenever the interest rate pass-through
to insured and uninsured depositors is the same, uI (r) = u (r), then
r̄ and r̂ remain unchanged. Otherwise, if, say, uI (r) < u(r), then the
threshold r̄ becomes larger, i.e., the range of risk-free rates where the
bank’s risk-taking incentives increase following a marginal increase
in r.
References
Acharya, V., and H. Naqvi. 2012. “The Seeds of a Crisis: A The-

ory of Bank Liquidity and Risk Taking over the Business Cycle.”
Journal of Financial Economics 106 (2): 349–66.
Allen, F., E. Carletti, and R. Marquez. 2011. “Credit Market Com-
petition and Capital Regulation.” Review of Financial Studies
24 (4): 983–1018.
———. 2015. “Deposits and Bank Capital Structure.” Journal of
Financial Economics 118 (3): 601–19.
Ampudia, M., and S. J. Van den Heuvel. 2022. “Monetary Policy
and Bank Equity Values in a Time of Low and Negative Inter-
est Rates.” Journal of Monetary Economics 130 (September):
49–67.
Arce, O., M. Garcia-Posada, S. Mayordomo, and S. Ongena. 2021.
“Bank Lending Policies and Monetary Policy: Some Lessons
From the Negative Interest Era.” Research Paper No. 21-75,
Swiss Finance Institute.
Basten, C., and M. Mariathasan. 2020. “Interest Rate Pass-through
and Bank Risk-taking under Negative-Rate Policies with Tiered
Remuneration of Central Bank Reserves.” Research Paper No.
20-98, Swiss Finance Institute.
Bechtel, A., J. Eisenschmidt, A. Ranaldo, and A. V. Veghazy.
2021. “Quantitative Easing and the Safe Asset Illusion.” Work-
ing Paper on Finance No. 21-10, Swiss Institute of Banking and
Finance.
Bernanke, B. S., and A. S. Blinder. 1988. “Credit, Money, and Aggre-
gate Demand.” American Economic Review 78 (2): 435–39.
Bernanke, B. S., and M. Gertler. 1995. “Inside the Black Box: The
Credit Channel of Monetary Policy Transmission.” Journal of
Economic Perspectives 9 (4): 27–48.
Bittner, C., D. Bonfim, F. Heider, F. Saidi, G. Schepens, and C.
Soares. 2021. “Why So Negative? The Effect of Monetary Pol-
icy on Bank Credit Supply across the Euro Area.” Unpublished
Working Paper.
Bittner, C., A. Rodnyansky, F. Saidi, and Y. Timmer. 2021. “Mix-
ing QE and Interest Rate Policies at the Effective Lower Bound:
Micro Evidence from the Euro Area.” Working Paper No. 9363,
CESifo.
Borio, C., and H. Zhu. 2012. “Capital Regulation, Risk-taking and

Monetary Policy: A Missing Link in the Transmission Mecha-
nism?” Journal of Financial Stability 8 (4): 236–51.
Brandao-Marques, L., M. Casiraghi, G. Gelos, G. Kamber, and R.
Meeks. 2021. “Negative Interest Rate Policies: Taking Stock of
the Experience So Far.” CESifo Forum 22 (November): 27–32.
Brunnermeier, M., J. Abadi, and Y. Koby. 2023. “The Reversal
Interest Rate.” American Economic Review 113 (8): 2084–2129.
Cordella, T., G. Dell’Ariccia, and R. Marquez. 2018. “Government
Guarantees, Transparency, and Bank Risk Taking.” IMF Eco-
nomic Review 66 (1): 116–43.
Darracq Pariès, M., C. Kok, and M. Rottner. 2020. “Reversal Inter-
est Rate and Macro-prudential Policy.” Working Paper No. 2487,
European Central Bank.
Dell’Ariccia, G., L. Laeven, and R. Marquez. 2014. “Real Interest
Rates, Leverage, and Bank Risk-taking.” Journal of Economic
Theory 149 (January): 65–99.
Döpp, S., A. Horovitz, and A. Szimayer. 2022. “Modeling Non-
maturing Demand Deposits: A Proposed Methodology to Deter-
mining the Idiosyncratic Confidence Level Used for Separating
Stable Deposit Volumes from Volatile Deposit Volumes.” Quar-
terly Journal of Finance 12 (1): 1–20.
Eggertsson, G. B., R. E. Juelsrud, L. H. Summers, and E. G. Wold.
2019. “Negative Nominal Interest Rates and the Bank Lend-
ing Channel.” Working Paper No. 25416, National Bureau of
Economic Research.
Ennis, H. M. 2018. “A Simple General Equilibrium Model of Large
Excess Reserves.” Journal of Monetary Economics 98 (October):
50–65.
Ennis, H. M., and A. L. Wolman. 2015. “Large Excess Reserves in
the United States: A View from the Cross-Section of Banks.”
International Journal of Central Banking 11 (1): 251–89.
European Central Bank. 2017. “Economic and Monetary Develop-
ments.” Economic Bulletin No. 8/2017, European Central Bank.
Flannery, M. J. 1982. “Retail Bank Deposits as Quasi-fixed Factors
of Production.” American Economic Review 72 (3): 527–36.
Freixas, X., and J.-C. Rochet. 1997. The Microeconomics of Banking,
1st Edition. The MIT Press.
Gambacorta, L., and H. S. Shin. 2018. “Why Bank Capital Mat-

ters for Monetary Policy.” Journal of Financial Intermediation
35 (Part B): 17–29.
Heider, F., F. Saidi, and G. Schepens. 2019. “Life Below Zero: Bank
Lending under Negative Policy Rates.” Review of Financial Stud-
ies 32 (10): 3728–61.
Jimenez, G., S. Ongena, J.-L. Peydro, and J. Saurina. 2012. “Credit
Supply and Monetary Policy: Identifying the Bank Balance-Sheet
Channel with Loan Applications.” American Economic Review
102 (5): 2301–26.
Jobst, A., and H. Lin. 2016. Negative Interest Rate Policy (NIRP):
Implications for Monetary Transmission and Bank Profitabil-
ity in the Euro Area. Washington, DC: International Monetary
Fund.
Kashyap, A. K., and J. C. Stein. 1995. “The Impact of Monetary
Policy on Bank Balance Sheets.” Carnegie-Rochester Conference
Series on Public Policy 42 (June): 151–95.
Martin, A., J. McAndrews, and D. Skeie. 2016. “Bank Lending in
Times of Large Bank Reserves.” International Journal of Central
Banking 12 (4): 193–222.
Martinez-Miera, D., and R. Repullo. 2017. “Search for Yield.”
Econometrica 85 (2): 351–78.
Miller, S., and B. Wanengkirtyo. 2020. “Liquidity and Monetary
Transmission: A Quasi-experimental Approach.” Working Paper
No. 891, Bank of England.
Paraschiv, F. 2013. “Adjustment Policy of Deposit Rates in the Case
of Swiss Non-maturing Savings Accounts.” Journal of Applied
Finance and Banking 3 (3): 271–323.
Repullo, R. 2020. “The Reversal Interest Rate: A Critical Review.”
CEPR Discussion Paper No. 15367.
Sharpe, S. A. 1997. “The Effect of Consumer Switching Costs on
Prices: A Theory and Its Application to the Bank Deposit Mar-
ket.” Review of Industrial Organization 12 (1): 79–94.
Wang, Y., T. M. Whited, Y. Wu, and K. Xiao. 2022. “Bank Mar-
ket Power and Monetary Policy Transmission: Evidence from a
Structural Estimation.” Journal of Finance 77 (4): 2093–2141.

Ijcb 24 Q 3 A 6

Uploaded by

Copyright:

Available Formats

Ijcb 24 Q 3 A 6

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Ijcb 24 Q 3 A 6

Uploaded by

Copyright:

Available Formats

Bank Risk-Taking and Impaired Monetary

Philipp J. Koeniga and Eva Schliephakeb

How does risk-taking aﬀect the transmission of interest

Central banks in several advanced economies have until recently kept

Figure 1. Direct Portfolio Adjustment Channel

The portfolio adjustment channel reﬂects the conventional view

incentives. Whenever the safe asset eﬀect dominates the deposit

Second, the banker always holds a non-negligible amount of

Related Literature. Our paper relates to a large body of lit-

policy transmission in a low interest rate environment. Eggertsson

We consider a bank over two periods, indexed by t = 0, 1. The bank

banker and reduces the riskiness of her loans. Depositors cannot

Assumption 1. The depositors’ outside option u(r) ≥ r is bounded

Assumption 2. The amount of deposits, D, is exogenously given.

Bank Assets and Monitoring. The banker is a monopolist in

The amount invested in the risk-free asset is determined endoge-

Assumption 3. The elasticity of the loan demand function, η(rL ) ≡

Figure 2. Sequence of Events

amount of safe assets. We relax Assumption 3 together with Assump-

3. The Portfolio Adjustment and the Loan Risk Channel

Optimal Monitoring Choice. We solve the model backwards by

rL L(rL ) + rR − rD D − κq̂ = 0. (3)

Given that depositors rationally anticipate the bank’s optimal

Lemma 1. The banker’s optimal monitoring choice is given by a

The eﬀects of rL and r on the optimal monitoring level reﬂect

u (r)D < R ⇔ u (r) < ρ. (6)

Optimal Loan Issuance and Reserve Holdings. Substitut-

Π = q̂(rL , r)rL L(rL ) + r(D − L(rL )) − u(r)D

The banker’s remaining choice variable is the loan rate rL . The

At the optimum point, the elasticity of the loan demand exceeds

because the eﬀect of rL on q̂ can also be expressed in terms of the

(+) (−) (+)/(−)

The conventional view of monetary policy transmission holds

Figure 3. Required Marginal Deposit Rate, u (r),

eﬀect dominates, a reduction in the risk-free rate reduces monitor-

Proposition 1. For a suﬃciently small risk-free rate, the safe asset

banker always holds a strictly positive level of reserves, even at low

Proposition 2. The loan risk channel dominates the portfolio

To understand the intuition behind Proposition 2, recall that,

assets, the banker’s monitoring must react strongly enough, i.e., a

Proposition 3. If monitoring costs are suﬃciently high, then the

Proposition 3 translates the condition in Proposition 2 into a

Figure 4. Reversal of Transmission when κ > κ

Figure 5. Monetary Transmission and the Risk-Free Rate

reversal rate in our model is just the most extreme manifestation of

Hypothesis 1. An increase in the banker’s exogenous deposit fund-

Hypothesis 1 follows because an increase in the deposit vol-

Hypothesis 2. The reversal rate is larger if, ceteris paribus,

• the bank has more deposits;

Hypothesis 2 follows from the eﬀects of leverage, and the moni-

Hypothesis 2 is in line with recent evidence by Arce et al. (2021),

4. Extensions and Discussion

4.1 Insured Deposits

(δ q̂ + 1 − δ)u(r) − (1 − δ)(1 − q̂) rR

Substituting r̄D into Equation (3) implicitly deﬁnes the banker’s

Proposition 4. Given a share δ of insured deposits, the loan risk

Comparing Propositions 2 and 4 shows that the condition for the

Hypothesis 3. The reversal rate r̂ is smaller for banks that are

4.2 Endogenous Deposit Choice, Deposit