How do central banks control inflation?

A guide for the perplexed*

Laura Castillo-Martinez Ricardo Reis
Duke University LSE

July 2024

Abstract

Central banks have a primary goal of price stability. They pursue it using tools that
include the interest they pay on reserves, the size and the composition of their bal-
ance sheet, and the dividends they distribute. We describe the economic theories
that justify the central bank’s ability to control inflation and discuss their relative ef-
fectiveness, in light of both theory and the historical record. We present alternative
approaches as consistent with each other, as opposed to conflicting ideological camps.
While interest-rate setting is often superior, having both a monetarist pillar and fiscal
support is essential, and at times pegging the exchange rate or monetizing the debt is
inevitable.

E31, E52, E61.

Keywords: monetary policy, policy rules, determinacy, effectiveness.

* Contacts: l.castillo-martinez@duke.edu and r.a.reis@lse.ac.uk. We are grateful to students at the LSE, as

well as at the AEA continuing education program, Columbia University, the Kiel Institute, and the Tinber-
gen Institute for many comments on this material, to Maxi Guenneweig and Saleem Bahaj for co-teaching
it with us, to Martina Fazio, Jose Alberto Ferreira, and Edoardo Leonardi for editorial support, and to the
editor, referees, and many colleagues over the years for their comments. This project received funding from
the European Union’s Horizon 2020 research and innovation programme, INFL, under grant number No.
GA: 682288.
 How do central banks keep inflation on target? How do they prevent episodes of
hyperinflation and their tragic consequences for welfare? Can a central bank control in-
flation if the economy goes through secular stagnation, a liquidity trap, or a fiscal crisis?
Why was inflation so high in 2021-24 and will it persist? These are crucial questions that
have answers in current economic theory.
Yet, students coming out of a macroeconomics class are often flummoxed by this topic.
Undergraduates mostly retain that central banks print money and more money means
higher inflation. They are then thoroughly confused when they learn that most central
banks barely mention money in their speeches, that they do not actually choose how
much currency to print, and that the US monetary base increased five-fold between 2008
and 2014 with no visible dent on inflation. Graduate students learn about the setting of
interest rates and the Phillips curve, and perhaps even about the welfare costs of inflation
and the links between monetary and fiscal policy. However, when asked to reconcile the
Fisher relation (higher interest rates are associated with higher expected inflation one-to-
one), the Taylor principle (increasing interest rates more than one-to-one in response to
inflation keeps inflation constant), and the empirical evidence (exogenous positive shocks
to interest rates lower inflation), they struggle with the apparent paradoxes. Discus-
sions of equilibrium determinacy or active-passive regimes attract theoretically-minded
researchers as much as they put off those focused on empirical applications.
The goal of this article is to provide a unified treatment of the theory of how central
banks control inflation. The hope is that researchers will have an accessible entry point
to this literature, so they can make sense of monetary policies and inflation outcomes.
Our approach has three distinctive features.1 The first is that we highlight the common
features of different viewpoints by using a single model of prices. We present alternative
theories not as opposing views, but rather as actively relying on some economic mecha-
nisms while passively allowing other economic mechanisms to accommodate.
The second feature is that we put the central bank at the center of all of them. It is
the central bank whose liabilities define the price level and that has a mandate to target a
value for inflation. The central bank’s policies are always the key determinant of inflation,
whether they can be described as monetary, fiscal or, more accurately, as a mix of both.
The third feature is that we provide one interpretation of the history of inflation as a
result of different strategies followed at different times. Our goal is not to defend it as
either the only or the right explanation for inflation’s movements, but rather to let the
1 Previous surveys, taking a different approach, are McCallum (1999) and Woodford (2003).

1
reader see the different theoretical concepts at work and their relative strengths. We con-
clude that by setting the interest rate on banks’ deposits at the central bank (reserves) in
an aggressive and transparent way, while having a monetary pillar to anchor expecta-
tions, and fiscal support to prevent runs on its liabilities, the central bank can effectively
control inflation. We discuss the circumstances when this is not possible, and how central
banks can regain control of inflation by picking from a menu of unconventional policies.
This is not an article about the optimal way to conduct monetary policy or about how
to trade off variability in inflation versus real activity. We take as given a target for infla-
tion, and study the central bank’s ability to achieve it along three dimensions. The first is
determinacy, on whether policy can deliver a unique price level. We take multiple or inde-
terminate equilibria in a model as a sign of an incomplete policy framework. The second
is effectiveness, on whether one policy leads to a smaller expected variation in the devia-
tions of actual inflation from target. By characterizing the components of this variation,
these can be quantified to then choose between policies based on predicted outcomes. The
third is experience, by comparing different inflation episodes of both advanced economies
and emerging markets and linking them to the monetary policy regimes of the time.
Section 1 sets up the general model of prices we will use, as well as a neoclassical and
a New Keynesian general equilibrium special cases. The key result is that the classical
analysis of supply and demand does not pin down inflation. This is what makes this
topic special.
To discuss inflation, one must introduce a central bank, whose liabilities define what
prices are in the first place. Section 2 lays out the tools at the central bank’s disposal, and
sets up a passive strategy where they are not used, and the government only receives
dividends in the background. Inflation remains indeterminate. The sections that follow
activate one tool at a time, together with different interactions with the fiscal authority.
Each tool is associated with one economic force to pin down inflation.
In section 3, we consider the economic force of arbitrage between savings in nominal
or real investments. The central bank steers inflation by choosing the interest rate that
remunerates reserves, but it must do so with a feedback from inflation to the interest rate.
If that feedback is aggressive, if the central bank forms accurate estimates of the state of
the economy, and if it communicates them transparently, then it can keep inflation close
to its target. The inflation targeting regimes that dominated monetary policy in the 1990s
and 2000s testify to the success of this approach, even though testing it econometrically is
challenging. This section also shows the importance of having a pillar, in the sense of an

2
escape clause, if self-fulfilling expectations make inflation explode.
Section 4 continues with setting interest rates, but in the version that dominated the
actions of the major central banks in the 2010s (and earlier in the case of Japan). We discuss
why there may be bounds on interest rates that lead to unconventional times where the
approach of the previous section cannot be used. Forward guidance, quantitative easing,
or going long are strategies that focus on long-term interest rates, while subsidizing bank
credit focuses on private bank rates. All of these put a great burden on the rationality of
the economic agents interacting with the central bank, as their effectiveness rely on how
far-sighted they are in setting their expectations. The experience with these strategies has
been mixed.
Section 5 considers the economic force that brings the market for banknotes into equi-
librium. The key equation is the demand for currency, while the policy tool is the central
bank’s exclusive right to supply banknotes. We consider money growth rules, fiscal rules
on the seignorage revenues from this activity, and pegs to either commodities like gold,
or to foreign currencies. Determinacy is easy to ensure, but the theory suggests that this
policy strategy is usually not effective. The experience using money growth rules in the
early 1980s in the UK and US and with the pegs of the Bretton Woods system in the post-
war confirms that the monetary approach often leads to volatile inflation. The experience
in Latin America in the 1980s and 1990s shows that it can even come hand in hand with
hyperinflation.
In section 6, the key equation is the intertemporal budget constraint of the central
bank, and the economic force is the unwillingness of private agents to hold the liabilities
of an insolvent institution. The policy tools are the central bank’s expenses, the com-
position of its assets, and especially the dividends it pays to the fiscal authority, all of
them affecting the net shortfall of the central bank. In practice, this strategy is more often
imposed on, rather than adopted by, central banks when they lose their financial inde-
pendence. This section makes clear that ensuring fiscal support for the central bank is the
way to avoid it, and discusses how this can be done.
Section 7 concludes with a brief discussion of the inflation disaster of 2021-24 and how
central banks might have let it happen.

3
1 The challenge of pinning down inflation
Our starting point is a general model of prices that nests most frameworks used to study
inflation. This generality comes from a strict focus on nominal prices, while taking real
outcomes as given. We then incorporate two alternative mappings from nominal to real
variables, which require many assumptions but are nested within the general framework:
a neoclassical model first, and a New Keynesian model next. The take-away is that, in the
absence of an institution like the central bank, inflation is indeterminate.

1.1 A general model of prices

Time is discrete, indexed by t from 0 to infinity. There is a large number of differentiated
goods, indexed by i from 0 to I, as well as assets, indexed by j. Agents behave opti-
mally and interact in markets. First, we describe each of three types of agents, together
with the measure of the price level, and define an equilibrium. Then, after making some
assumptions to simplify the exposition, we try to solve for a unique level of prices.

1.1.1 Building blocks

The first building block is a definition of the price level: an index function over the prices of
many goods in a basket that captures how much is a dollar’s worth in terms of a basket
of goods over time. Given individual prices Pt (i ) ≥ 0, the price level is:
 
Pt = P { Pt (i )}i=0,...I (1)

The function P (.) is non-negative and linearly homogeneous, so that the price level dou-
bles when all individual prices double. The index may be the ideal measure of the cost of
living in the economy. But, it may as well be any other arbitrary index that we choose to
measure inflation over. This is closer to reality, when we use the consumer price index,
the consumption deflator, or core versions of either of them.
The second building block is the optimal behavior by households in allocating spend-
ing across goods. Namely, a representative consumer equates its marginal rate of substitu-
tion between good i and good 0 to the relative price of these goods:

Pt (i )
= Rt (i ) for i = 1, ..., I , (2)
Pt (0)

4
where Rt (i ) captures how many units of good i the consumer would trade for one unit of
good 0. The Rt (i ) may change over time, reflecting shocks to consumer tastes.
Households live for multiple periods and save across time using assets. The Et (.)
operator captures the expectations of the household as of period t, which may not have
full information, but we will assume is rational in the sense of being consistent with the
other equilibrium equations of the model.2 An asset j, held between periods t and t + 1,
pays a nominal return of It+1 ( j). The asset may also provide a non-pecuniary utility to its
holder, which we capture as a convenience yield defined in utility units: Qt ( j).
In making optimal saving decisions, the representative saver equates the marginal rate
of substitution between consumption at time t versus t + 1 to the relative price. This in-
tertemporal marginal rate of substitution is Mt+1 ≥ 0, sometimes also called the stochas-
tic discount factor. If an asset is purely financial, i.e., valued only for the income it pro-
vides, then Qt ( j) = 0 and the relative price is Pt (1 + It+1 ( j))/Pt+1 , since this is the op-
portunity cost of consuming an additional unit today in terms of foregone consumption
tomorrow. More generally, the corresponding relative price is the (inverse of the) nominal
asset return plus its convenience yield:
   
Pt (1 + It+1 ( j))
Et Mt +1 + Qt ( j) = 1 for all j . (3)
Pt+1

There are two specific (pure) financial assets whose returns provide useful bench-
marks for our analysis. The first is a safe bond that pays the same nominal return at
t + 1, no matter the state of the world. We denote its nominal return by It+1 (nom) = It ,
to emphasize that it is known at time t. The second is a safe bond that pays the same real
return at t + 1, no matter the state of the world. We use 1 + Rt = Pt (1 + It+1 (real )/Pt+1
to denote this return, again to emphasize that is it known at time t. Note that, from the
equation above: 1 + Rt = Et (Mt+1 )−1 .
The fourth block refers to firms setting prices to maximize the surplus of production
under imperfect competition. Denote by Ct (i ) the real marginal cost of producing good i,
and Zt (i ) a desired markup, so the desired price charged by the firm, P̃t (i ), is:

P̃t (i ) = Pt Zt (i )Ct (i ) . (4)

Again, both Zt (i ) and Ct (i ) are stochastic and time-varying.

2 Section 4 will relax this assumption where it especially matters, when forming far-away expectations.

5
 These four blocks are common to most models written by macroeconomists. They
define a general model of prices.

Definition 1. Given a path for real outcomes, Nt = {Rt (i ), Zt (i ), Ct (i ), Mt+1 , Qt ( j)}∞ t=0 , an
∞
equilibrium for prices is a path for { Pt , Pt (i ), P̃t (i ), It+1 ( j)}t=0 that satisfies equations (1)-(4).

Our goal in this paper is to study the sequence { Pt }∞

t=0 and the associated inflation
∞
sequence {Πt }t=1 , where Πt ≡ Pt /Pt−1 . 3

1.1.2 Uncertainty

Before proceeding, we make two simplifications regarding the nature of uncertainty. We

invite the interested reader to check an online appendix for derivations without these
simplifications.
First, we assume that all uncertainty regarding the path of real outcomes is resolved
in period 1 and is captured by the state of the world s. At date 0, agents do not know
which s will be realized the following period. But once period 1 arrives, the world is
deterministic, and all exogenous variables are constant. The full path of real outcomes for
t ≥ 1 is then a function of this state: Nt = N(s). In turn, the endogenous future prices
that we want to solve for are then also a function of the state: ( P0 , Πt (s)).
Second, we work with a log-linearized version of this economy around a steady state
where all the real outcomes are constant (for instance Mt+1 = β), and all prices grow over
time at the constant rate Π̄. The four log-linearized equations are:

I
pt = ∑ ωi p t ( i ) (5)
i =0
p t ( i ) = p t (0) + ρ t ( i ) (6)
p̃t (i ) = pt + zt (i ) + ct (i ) (7)
r t = Et ( i t +1 ( j ) − π t +1 ) + q t ( j ) . (8)

Lower-case prices denote the log-linearized counterpart of prices in capital letters, and
the same applies to the real outcomes: for instance ρt (i ) is the log-linear counterpart of
Rt (i ). For returns, rt is the log of the gross yield on the real bond and similarly for the
Q̄( j)
nominal rates, while qt ( j) = 1−Q̄( j) log( Qt ( j)/ Q̄( j)). Finally, ωi represents the elasticity
of function P (.) with respect to its i-th argument evaluated at the steady state. At the start
3 For convenience, we sometimes write Π0 = P0 .

6
of each section, we will refer to the general non-linear model to highlight the economic
forces, but from then on proceed to work with this special log-linear version.

1.1.3 Price level (in)determinacy

Does this model of the economy deliver a unique prediction for inflation? It will if it
obeys the following conditions:

Definition 2. The level of inflation is unique or determinate in equilibrium if:

1. There is a unique scalar P0 in equilibrium.

2. If Π0t (s) and Π00t (s) both satisfy equilibrium conditions, then Π0t (s) = Π00t (s).

In our general model of prices, for a given path of real outcomes, N(s), and a path
for the price level, ( P0 , Πt (s)), the I + 1 equations in (1)-(2) pin down the prices Pt (i ) for
each of the I + 1 goods. In turn, each indexed version of equations (4) and (3) solves for
desired goods’ price P̃t (i ) and the prices of financial assets It+1 ( j), respectively.
But what pins down the price level in the first place? Nothing, there are no equations
left.
Under the current framework, any level of inflation is consistent with the equilibrium
conditions above. If the price level doubles at all dates, then actual and desired prices
will double as well and nominal returns will be the same. If the price level grows twice
as fast, returns will be twice higher, and actual and desired good prices will grow twice
as fast.
This result is what makes the study of inflation so special, and different from other
variables in economics. Its importance and generality dates back to Hume (1752): dol-
lars are just a unit of account with which the prices of goods are determined. If people
started denominating prices in cents instead of dollars nothing would change. There is no
demand or supply that ensures that 100 cents equals one dollar. Nothing in classical eco-
nomics pins down the price level or inflation, in the same way that nothing determines
whether measurements should be in inches or centimeters.

1.2 From prices to quantities

Some readers may only care about inflation insofar as it is associated with changes in real
outcomes that affect well being. To make explicit the link between prices and quantities,

7
we now introduce two, more stylized but general equilibrium, models. While they re-
quire more structure, they endogenize both nominal and real variables as a function of
shocks.
The list of new assumptions, which some readers may choose to skip and jump to the
next sub-section, are:

1. No investment: consumption of each good equals output, denoted by Yt (i ).

2. No good, household, or asset-specific shocks: only aggregate shocks.

3. A CES aggregator of individual goods: R(i ) = (Yt (i )/Yt (0))−σ , where σ > 1 is the
elasticity of substitution across goods;

4. An ideal cost-of-living price index: Pt1−σ = ∑i∞=0 Pt (i )1−σ since with CES preferences
σ −1
= ∑i∞=0 Yt (i )
σ −1
the utility-based measure of aggregate output is Yt σ σ .

5. A CRRA utility function in consumption: Mt+1 = β(Yt /Yt+1 )1/θ , where θ > 0 is the
intertemporal elasticity of substitution and β < 1 is a subjective impatience factor.

6. Constant elasticity disutility of working separable from consumption, and linear

1
+ ϕ1 1+ ϕ1
production function in work: marginal cost is Ct (i ) = Ytθ /At where ϕ is the
Frisch elasticity of labor supply.

7. Markups are exogenous: Zt (i ) is driven by a single aggregate shock.

8. The non-pecuniary benefits from holding an asset depends solely on the asset’s sup-
j
ply relative to output: Qt ( j) = ( Bt ( j)/Pt Yt )−1/η Ut ( j), where Bt ( j) is the nominal
supply of the asset, Ut ( j) is an exogenous shock, and η j > 0 is the elasticity.

The missing ingredient is the link between desired P̃t (i ) and actual prices Pt (i ). This is
where our two stylized models differ. If desired and actual prices are the same, prices are
flexible, and we have a neoclassical economy. If, instead, there is a wedge between P̃t (i )
and Pt (i ), the economy features frictions, which can arise in different markets. Among
these, we study the most popular type, nominal rigidities, in the context of the New
Keynesian model.

8
1.2.1 A neoclassical view

Under flexible prices, the (exact) log-linear solutions for real outcomes is: ynt = ( at −
zt )/( ϕ−1 + θ −1 ) and rtn = − log( β) + (1/θ )(Et ynt+1 − ynt ), where we use the superscript
n to denote the neoclassical, or natural, unique equilibrium.
Real outcomes in equilibrium are independent of nominal variables. This is commonly
referred to as the classical dichotomy: real trade-offs are unchanged regardless of the
price level. That is, in a neoclassical world, Nt is exogenous with respect to inflation.
This implies that the indeterminacy of the price level does not affect real variables, which
are uniquely determined.

1.2.2 The New Keynesian model

This is no longer the case once we move away from a frictionless world. There are many
ways to do so, and we will focus on a single one: nominal rigidities driving a wedge
between desired and actual prices. There are also many models of nominal rigidities
but will now describe, arguably, the most popular one: the Calvo price-setting model.
Note, however, that in each section we will also briefly introduce alternative frictions and
describe their impact on inflation.
We refer the reader to any standard textbook for the derivation of the equations (e.g.,
Gali (2008)). The optimality condition for a safe nominal bond in equation (8) becomes:

ỹt = Et (ỹt+1 ) − θ (it − Et (πt+1 ) − rtn ) , (9)

where ỹt ≡ yt − ynt is the output gap. This is sometimes referred to as the IS curve.
The key assumption of the Calvo price-setting model is that, every period, only a share
of firms are able to reset prices. Thus, when given the opportunity to do so, firms set their
price equal to the discounted sum of expected future desired prices over the period they
expect their current price to be binding. The other three equations combined—(1), (2) and
(4)—lead to a Phillips curve:

πt = β Et (πt+1 ) + κ ỹt + zt . (10)

The parameter κ captures the inverse of the degree of nominal rigidities, so that when
κ → ∞ we are back at the classical dichotomy. Now, changes in inflation affect real
outcomes, which in turn affect inflation as well.

9
 The New Keynesian equilibrium features two equations, (9)-(10) but three unknowns:
it , yt and πt . For a given πt , the it is determined, just as before. But now, for a given
πt , then yt is also solved for and, from there, all other real variables follow. Nominal
outcomes now feed back into real outcomes, without a dichotomy between the two.
There is still nothing in the economy to pin down inflation in the first place. Not
only the indeterminacy of inflation remains, it is now worse, as there is indeterminacy
of output and all other real variables. Intuitively, for any given path for inflation, firms
and workers produce whatever is demanded at these prices. In turn, demand dictates
an amount of savings that determines the real interest rate. Other models of nominal
rigidities will have different mechanisms, but they share this joint nominal and real inde-
terminacy.4

2 Introducing a central bank

In modern digital economies, people use electronic means of payment, like debit and
credit cards, to settle their transactions. For any given transaction, the seller may have an
account in bank A while the buyer has an account in bank B, so there must be a settlement
whereby bank A collects payments from bank B. The central bank is the clearing house
where payments between banks take place. We formally describe the central bank in this
section.

2.1 Central bank tools

The central bank’s assets and liabilities give it a set of monetary policy tools.

2.1.1 Liabilities: reserves

The payments between banks are made using their deposits at the central bank, a digital
means of payment. Often named reserves, these are nothing but liabilities of the central
bank vis-à-vis banks. Because reserves are the ultimate form of payment, they are the
unit of account of the economy. Since reserves in the United States are denominated in
dollars, firms and people choose to denominate their prices in dollars as well. The price
of a good is simply how many units of reserves must be exchanged to obtain such good.
4 Carlstrom and Fuerst (2002) and Nakajima and Polemarchakis (2005) provide further discussion on
sticky prices and real indeterminacy.

10
 The current stock of reserves is a list of entries in a spreadsheet at the central bank, one
for each bank. Given its control over the spreadsheet, the central bank has two tools: it can
choose the amount of reserves, Vt , or the rate at which it remunerates them, It+1 (V ) = Itv ,
which determines their return at t + 1 since they are safe assets. Both are decisions, one
on the sum of the entries in the spreadsheet at a date, and the other on the factor by which
each entry gets multiplied across dates.
From the perspective of households, reserves are just another asset that can be held
(via banks). The savers’ optimality condition to hold them is also given by equation (3).
Since the great financial crisis, central banks have satiated the demand for reserves driv-
ing their convenience yield to zero.5 We will assume this is the case for our main analysis
and then revisit the pre-2008 experience in section 5. Combining equation (3) for both
bonds and reserves implies that It = Itv . Intuitively, any discrepancy between these re-
turns would lead savers to want to borrow an infinite amount of the nominal safe bond
to invest an infinite amount in reserves, or the other way around, and the markets for
neither bonds nor reserves would clear.
Through the power of arbitrage, by choosing the nominal return on reserves, the cen-
tral bank controls the nominal interest rate on safe bonds.

2.1.2 Liabilities: Banknotes

Central banks also issue physical banknotes in the nominal amount Ht ≥ 0.

Banknotes are distinct from reserves in a few ways. To start, banknotes pay no interest.
Therefore, It+1 ( H ) = 0, and the opportunity cost of using banknotes as opposed to digital
means of payment is equal to the interest rate paid on reserves.
Moreover, banknotes can be freely held by anyone in the economy, not just banks, and
they are anonymous as people do not have to declare to the government how much cur-
rency they have or from whom they got it. This is because banknotes are physical, which
in turn may make them easier to use for some payments, as opposed to the electronic
means backed by reserves (while for others the opposite is true). These properties create
a demand for the services provided by banknotes separate to the demand for reserves.
Economic agents are willing to sacrifice returns when holding banknotes because
some prefer not to use the banking system when making payments, some prefer phys-
ical to digital payments, some want anonymity in their transactions, and some find cash
5 Reis (2016) shows evidence for satiation of reserves in the US and argues this is desirable.

11
easier to use. The convenience yield for banknotes in log-linear terms is:

ht − pt − yt
qt ( h) = − + ut ( h) . (11)
ηh

As banknotes provide a convenience yield, but no financial return, they are often de-
scribed as durable goods that the central bank produces and sells for its value 1/Pt , rather
than liabilities. The payoff from these sales of banknotes is called seignorage, and given
by StH+1 = ( Ht+1 − Ht )/Pt+1 .
At the same time, because the central bank commits to exchange them for reserves
one for one, at all times, they are a durable with unlimited refunds. So, they can also be
described as a liability. Either way, the central bank has a tool, the amount of Ht it prints.

2.1.3 Assets

On its asset side, the central bank’s balance sheet holds a portfolio of some of the assets in
the economy. Letting At ( j) be its holdings of asset j, then the real value of these holdings
is At = ∑ j At ( j)/Pt . Another of its tools is the composition of these assets.

2.1.4 Net shortfall

The central bank earns the return on its assets together with seignorage from printing
banknotes. It also has expenses, net of the surcharges it imposes on banks, for instance by
not paying interest on required reserves. We denote their real value by Et . Finally, it pays
a real dividend to the government, Dt . Combining all of these elements, the real resource
constraint of the central bank in period t + 1 is:
(1 + Itv )Vt
 
1 + It+1 ( j) Vt+1
StH+1 + ∑ Pt At ( j) + = Et+1 + Dt+1 + At+1 + (12)
j
Pt+1 Pt+1 Pt+1

The expenses Et and the dividends Dt are choices of the central bank. Together they
determine the income shortfall of the central bank:

It+1 ( j) − Itv
 
St+1 = Et+1 + Dt+1 − ∑ At ( j) − StH+1 . (13)
j
P t +1

This shortfall is an extra monetary policy tool, through dividends, expenses, and the
choice of assets.

12
2.1.5 Net worth

Given its unconstrained ability to issue reserves, the central bank could be tempted to
increase Vt+1 without limit and pay an unlimited dividend Dt+1 to the Treasury. Pri-
vate agents would not want to participate in this Ponzi scheme. Their refusal to do so is
captured by the condition: h i
lim E0 M(s) T WT = 0 . (14)
T →∞

where Wt = At − Vt /Pt is the central bank’s real net worth, i.e., the difference between its
assets and liabilities.
This is not a policy tool, but a constraint that the central bank faces. Note that, with
this definition of net worth, the resource constraint can be written as Wt+1 = ((1 +
Itv ) Pt /Pt+1 )Wt − St+1 or, in its log-linear version:

βwt+1 = itv − πt+1 + wt − (1 − β)st+1 . (15)

2.2 The policy target

The central bank uses these tools to achieve an objective: to keep inflation ( P0 , Πt (s))
close to a target ( P0∗ , Π∗ (s)). The target depends on the real state of the economy, and it
may be arbitrary or optimal given some objectives of policy.6 The key assumption is that,
again, we solve for prices taking as given the target.
The information of the central bank is limited by imperfect real-time estimates of the
state of the economy. While Et (rt+ j ) denoted the public’s expectation at t of what the real
interest rate will be at t + j, we use r̂t+ j to denote the central bank’s expectation at t, and
these may not be the same.
If the central bank is able to determine inflation, its effectiveness is assessed by the size
of the deviations between the log price level and its target. In order words, ε(0) ≡ p0 − p0∗
and ε t (s) ≡ Πt (s) − Π∗ (s).

2.3 A formal definition of a central bank

A central bank is therefore the manager of a spreadsheet of payments on reserves, a seller
of an infinitely-lived durable good in currency, and a borrower and lender from house-
holds through its balance sheet. We sum up the description of the central bank with a
6 Readers interested in the choice of Pt∗ can see Khan, King and Wolman (2003) or Woodford (2010).

13
final definition.

Definition 3. The central bank uses its tools (Vt , Itv , Ht , At ( j), St ) to select an equilibrium price
level path ( P0 , Πt (s)) to get as close as possible to the target ( P0∗ , Π∗ (s)), given its perceptions of
real variables N̂.

Take then the following benchmark where the central bank passively sets its tools:

(i) the interest rate on reserves to satisfy equation (8);

(ii) banknotes to satisfy whatever demand in equation (11);

(iii) shortfalls to ensure net worth is constant in equation (15).

Under this passive regime, the new equations in this section pin down the central bank
tools as endogenous variables. In turn, the choices on the quantity of reserves and the
composition of the balance sheet are irrelevant. Inflation continues to be indeterminate.
Yet, in exercising each of these three functions, the central bank has tools to affect the
price level through each of the new equations. From this benchmark of indeterminacy,
each section will separately relax one of (i) to (iii), while keeping the others unchanged.
Each of them reveals an approach the central bank can follow to achieve its inflation
target.

3 Steering inflation using the interest rate on reserves

For many decades, most major central banks have conducted monetary policy by actively
steering safe interest rates, more recently by setting the interest rate on reserves. We study
this policy regime in this section, while keeping all other policies passive.

3.1 The economic force that drives inflation

Combining the saving optimality condition in equation (3) for the nominal and real safe
bonds, and recalling that by no-arbitrage nominal rates are equal to the rate on reserves,
gives a no-arbitrage relation between real safe bonds and reserves:

1 + Itv
  
Et Mt +1 1 + Rt − = 0. (16)
Π t +1

14
This states that, once adjusted by the stochastic discount factor, savings in real safe bonds
or in reserves at the central bank must yield the same expected return. It is often called the
Fisher equation and it is the key equation of the approach in this section. Its log-linearized
version is:
rt = itv − Et (πt+1 ). (17)

The equation captures the central economic force that determines inflation when cen-
tral banks set interest rates and which relies on the power of arbitrage. It works as follows:
the private sector can choose to hold reserves or real safe bonds. Suppose the price level
today was too low. All else equal, then expected inflation would be too high. Therefore,
the return on reserves would be lower than the return on real safe bonds. In other words,
by holding reserves at the central bank, the private sector gets fewer goods in return than
if they had invested them privately.
Agents would want to hold zero reserves, and invest all of their resources in real terms.
This would not be an equilibrium given a positive supply of reserves.
Rather, as they demand fewer reserves, their value falls. Because reserves are the
unit of account, their real value is 1/Pt , so the price level rises back into equilibrium. A
higher price level means that expected inflation is lower and the real return on reserves
is higher, rising until the point where agents are, once again, indifferent between real safe
bonds and reserves.7
This movement from nominal reserves to real investment can come not only through
real safe bonds as here discussed, but also through the purchase of durables. In a limit
economy without real assets, it would show up as a desire to save less and consume
more. A common saying is that inflation results from too much money chasing too few
goods. From this perspective, it arises rather from economic agents wanting to substitute
nominal assets for real goods or assets.

3.2 Exogenous interest rates

3.2.1 Nominal pegs

By itself, relying on arbitrage and setting interest rates does not determine inflation: it
depends on how it is done. Say the central bank chooses to set the interest rate on reserves
to follow some exogenous sequence itv = xt . The literature has traditionally referred to
7A small literature has studied inflation using the no-arbitrage approach but when incomplete markets
lead to variations of equation (16), see Benassy (2000) and Den Haan, Rendahl and Riegler (2017).

15
this as an interest rate peg. The Fisher equation at all dates then implies that:

E(π1 (s)) = x0 − r0 (18)

πt (s) = x (s) − r (s) for t ≥ 2. (19)

In the second equation, by choosing the right-hand side, the minimal central bank is
able to uniquely pin down inflation at all periods from 2 forwards. Given the target π ∗ (s)
it can choose x (s) to make sure actual inflation is always equal to it.
However, in the first equation, by setting the right-hand side, the central bank only
pins down expected inflation. Inflation at date 1 itself is not determinate. The second
condition for determinacy is not satisfied, as there are an infinite number of inflation
rates at different states of the world that satisfy this equation.8
Moreover, there is no condition pinning down the initial price level p0 . If people expect
higher prices in the future, the price level at date 0 will simply jump up today. The first
condition for determinacy is also not satisfied.9

3.2.2 Real payment on reserves

Suppose instead that the central bank promises to remunerate reserve holders with a
payment in real goods.10 Governments around the world have issued indexed bonds for
a long time, and so could central banks; this is what promising a real payment of goods
amounts to. The nominal return on reserves in dollars would then be itv = xt + pt+1 for
some exogenous xt as before.
Plugging the above into equation (17) and rearranging delivers:

r t = Et ( x t + p t +1 − π t +1 ) ⇒ pt = rt − xt . (20)

At date 0 then p0 = r0 − x0 , and at all future dates p(s) = r (s) − x (s). Since ( x0 , x (s)) is
uniquely chosen by policy, and (r0 , r (s)) is uniquely shaped by real forces, then the above
equation delivers a determinate price level.
No central bank does this, but this peg highlights the no-arbitrage forces behind in-
flation control. The real return on any investment is exogenously fixed by the stochastic
8 Nakajima and Polemarchakis (2005) provide a thorough discussion across different economic environ-
ments.
9 This classic result is due to Sargent and Wallace (1975).
10 This was studied by Hall and Reis (2016), building on earlier work by Hall (1997), which in turn for-

malized a proposal by Irving Fisher.

16
discount factor. If the central bank promises a real payment on reserves, then arbitrage
determines how many goods reserves are worth today. The economic force behind the
Fisher equation is that, since real bonds and reserves both deliver the same payment to-
morrow, they must be worth the same today. But, since reserves are denominated in
dollars, not goods, then this pins down the price level today.
Using its estimates of the future log real rate r̂ (s), the central bank will achieve an
effectiveness of ε(s) = r (s) − r̂ (s). The better the estimates of the real interest rate, the
more effective this policy will be.

3.3 Interest rate feedback rules

While picking interest rates on reserves, the central bank can choose a feedback rule to
adjust the interest rate to inflation (or the price level):

itv = xt + φπt , (21)

where φ > 1, so the response to inflation is more than one-to-one.

Combining with the log-linearized Fisher equation (17) gives a difference equation:

φπt = Et (πt+1 ) + rt − xt (22)

Iterating forwards, even though r (s), x (s) are constant, we cannot just rule out that en-
dogenous inflation rises or falls over time. For now, we impose a terminal condition that
inflation will not explode at a rate higher than φ: limT →∞ φ−T E π T = 0. Then, the differ-
ence equation has a unique solution:

r0 − x0
 
1
p0 = + E(r (s) − x (s)) , (23)
φ φ ( φ − 1)
r (s) − x (s)
π (s) = (24)
φ−1

The price level is determinate at all dates, including date 0.

By having the interest rate respond to inflation as well as to the central bank’s forecast
of real interest rates and the inflation target: x (s) = r̂ (s) − (φ − 1)π ∗ (s), the central bank
can achieve its target in expectation. This makes transparently simple what are the main
tasks of economists at central banks. Some will calculate the optimal target for inflation

17
π ∗ (s), using models that interpret the mandate of the central bank. Others will evaluate
the “state of the economy” in terms of the sufficient statistic r̂ (s). Finally, as reasonable
people disagree on both the state of the economy and the short-run goal for inflation,
committees of policymakers will discuss different views when choosing the right interest
rate x (s).

3.3.1 The virtues of transparency

In period 0, the effectiveness of the interest rate rule is:

r0 − x0 E (π ∗ (s))
 
φ
ε0 = + E(r (s) − r̂ (s)) + − p0∗ (25)
φ φ−1 φ

To choose policy today x0 , the central bank must take into account not just the state of the
economy r0 , but also the public’s expectations of the estimation mistakes that the central
bank will make on the future state of the economy. Even if neither the central bank nor the
public know what r (s) is, and even if their estimates are poor, as long as these estimates
coincide, policy will be effective.
The central bank’s communication is then an important input into its effectiveness.
Speeches and statements of what it thinks the future states of the economy will be are
crucial to keep inflation on target. They will work as long as the public agrees with these
views, a form of transparency. This is sometimes referred to as Delphic forward guidance,
as the central bank works like an oracle that convinces believers.
The Federal Reserve started to release lightly edited transcripts of previous FOMC
meetings in 1993. In 1999, it began issuing statements at the conclusion of every policy
meeting, and including a balance of risks in that statement in 2000. Orphanides (2019)
summarizes some of this evolution in the context of anchoring inflation expectations
when setting interest rates.
Across the world and since the early 1990s, central banks have adopted inflation tar-
geting frameworks. More than announcements of official inflation targets, these have
consisted primarily of transparency and communication efforts with the public about the
central banks’ objectives, plans, and actions (Bernanke and Mishkin, 1997). Empirical
work in this area has shown repeatedly that communication under inflation targeting
works by moving financial markets, the economic force behind feedback rules (Blinder
et al., 2008). Data for 112 countries from 1998 until 2019 shows an almost uniform in-
crease in transparency in the sense of central banks releasing data, sharing their internal

18
forecasts, explaining their framework and deliberations, and disclosing policy decisions
and their rationale (Dincer et al., 2022). Designing a central bank today is as much defin-
ing goals and strategies as it is setting a framework for transparency and accountability
(Reis, 2013). In this line, the IMF has even published a Central Bank Transparency Code
setting international standards (IMF, 2020).

3.4 The Taylor principle

The assumption φ > 1 played a crucial role in delivering the determinacy of inflation. An
aggressive enough response of interest rates to inflation ensured that the sum of future
real interest rates and policies stayed finite when we iterated the difference equation for-
ward. Also, φ > 1 motivated the terminal condition to rule out fast exploding paths for
inflation. This section discusses its role.

3.4.1 Different rules and the aggressiveness of policy

There are broader classes of interest rate feedback rules, partly motivated by the actions
and experience of central banks across countries and times.11
First, most estimates of policy rules also show that interest rates are inertial. Central
banks typically break a desired change in interest rates into 0.25% or 0.5% steps over suc-
cessive policy meetings. We can represent this by having current interest rates responding
to their own past value.
Second, convinced by estimates that monetary policy only affects inflation with a lag,
many central banks adjust interest rates in response to public forecasts of future inflation.
These are obtained from surveys, financial prices or internal models of the central bank.
We can capture this by adding the public’s expectation of future inflation to the interest
rate rule.
Third, many central banks respond to core measures, in order to smooth out the noisy
real-time measures of inflation and capture its permanent trends. Following Muth (1960),
we can model core inflation as a weighted average of past inflation, which is the opti-
mal estimate if actual inflation follows a random walk contaminated with white noise
measurement error.
Fourth, studies of optimal monetary policy often suggest that the central bank should
11 McCallum (1981) introduced these rules and first showed that they lead to determinacy. Taylor (1999),
Clarida, Gali and Gertler (2000) and Woodford (2003) are classic analyses.

19
 Table 1: Determinacy conditions

Rule Condition
Benchmark:
φ>1
xt + φπt
Inertial:
φ+χ > 1
xt + φπt + χitv−1
Forecast targeting:
φ+χ > 1
xt + φπt + χ Et (πt+1 )
Core inflation:
φ>1
x t + φ (1 − χ ) ∑ ∞ j
j =0 χ π t − j
Wicksellian:
φ>0
xt + φpt

target the price level rather than inflation. These Wicksellian rules replace πt with pt in
the policy rule.
The mathematics and economic logic of all these cases are similar to the ones in the
analysis of the Taylor rule. Table 1 formalizes them and shows the determinacy conditions
derived from the same steps as in the previous section. In all of them, the response of
interest rates to inflation must be large enough, although the thresholds differ. In fact, for
all but the last rule, the condition states that the cumulative response of the interest rate
to a persistent unit increase in inflation exceeds one.

3.4.2 Dealing with unobservables

In our general model of prices, the real interest rate r (s) and the optimal short-run target
π ∗ (s) are the sufficient statistics to implement effective policy. In practice, neither can be
observed in real time.12 The two stylized models of section 1.2 provide a mapping from
observables into the real outcomes that our general model of prices takes as given.
First, recall the neoclassical economy where rtn = − log( β) + (1/θ )(Et ynt+1 − ynt ). This
has motivated multiple implementations of the Taylor rule to have x (s) respond to esti-
mates of real activity. It also matches well the experience of central banks responding to
recessions by cutting interest rates.
Since our analysis already allowed for r (s) to be a general stochastic process—the only
restriction was that it was exogenous with respect to inflation—then this does not change
12 Adão,
Correia and Teles (2011) and Holden (2024) suggest using private-sector forecasts of inflation
and returns of real indexed bonds, respectively, as measures of the real interest rate.

20
anything when it comes to the determinacy of inflation. The effectiveness of the policy
rule will differ according to different measures of activity, and their relative variance and
correlation with the short-term inflation target.
Second, in the New Keynesian model, consider a Taylor rule that makes explicit that
the interest rate feedback rule in (21) includes the output gap:

itv = xt + φπt + φy ỹt . (26)

Combining this with equations (9)-(10) to eliminate the interest rate gives a system of two
equations in two unknowns:
! !
 ỹ  ỹ r n−x
t t +1 t
= Φ Et +Ω t , (27)
πt π t +1 zt
!
1 1 − βφ 1
where Φ ≡ Ω and Ω ≡ .
κ κ + β(1 + φy ) 1 + φy + κφ

A system of linear difference equations has a unique non-explosive solution if the

number of eigenvalues of the matrix outside the unit circle is equal to the number of non-
predetermined variables. In this case, both output and inflation can in principle jump, so
both eigenvalues have to have modulus larger than 1. Standard linear algebra shows that
this is the case if the following condition holds:

φy (1 − β)
φ > 1− . (28)
κ

This is a generalized version of the earlier Taylor principle condition. The coefficient
on the output gap relaxes the responsiveness with respect to inflation because output
covaries with inflation in the long run. By responding to output, the central bank is indi-
rectly further responding to inflation.13 The same intuition carries through, together with
the reliance on a terminal condition. Likewise, the characterization of the effectiveness of
the rule, and how it depends on the public’s perceived deviations between natural rates
of interest, inflation targets, and, now, markups over time, is the same. There is an exten-
sive literature that considers many variants of general equilibrium models with nominal
13 Meyer-Gohde and Tzaawa-Krenzler (2023) show that in models of the Phillips curve, like under sticky
information, where fully anticipated long-run monetary policy has no effect on output, the condition reverts
back to φ > 1.

21
rigidities to derive a variety of lower bounds on φ to ensure determinacy.14
The prevalence of a real indeterminacy, in addition to the nominal one, brings a fur-
ther economic force at play. Together with the no-arbitrage channel that is specific to
this approach, there is an aggregate demand channel as well. Changes in the returns
of financial assets affect households’ desire to save, while nominal rigidities make out-
put demand determined. Therefore, changes in the interest rate now also affect inflation
through changes in consumption.15

3.4.3 Testing the Taylor principle

Since different rules put different lower bounds on φ, can we not estimate φ and test this
mechanism? If one goes by the speeches, reports, and statements of central banks, one
would think that they all follow feedback rules and subscribe to the Taylor principle. But
central banks say many other things as well, and it turns out that it is hard to empirically
verify the condition for determinacy.
Going back to the solution for inflation in equations (23)–(24), imagine that the central
bank manages to be fully effective, so πt = πt∗ at all dates. In that case, the interest rate on
reserves will be itv = xt + φπt∗ = rt + πt∗+1 . Since there is no feedback anymore, this rule
is observationally equivalent to a peg. Even if the econometrician had data allowing her
to separate the state of the economy rt from desired inflation πt∗+1 , she could not estimate
φ.
Imagine instead rt = πt∗+1 = 0, so that there are no shocks to the economy or to
the policy goal, but only to monetary policy (mistakes) that follow the stationary process
xt = ρxt−1 + et , where et is iid mean zero. Then, the solution in equation (23) reduces to
πt = − xt /(φ − ρ) so inflation is also autoregressive of order 1. Solving for the interest
rate on reserves: itv = xt + φπt = −(φ − ρ)πt + φπt = ρπt . Therefore, a regression of
the policy rate on inflation would recover the parameter ρ. Since ρ < 1 this estimate
14 A few examples are the inclusion of capital accumulation (Sveen and Weinke, 2005), trend inflation

(Ascari and Ropele, 2009, Khan, Phaneuf and Victor, 2020), constraints to firm credit (Lewis and Roth, 2018),
multiple policy regimes (Barthélemy and Marx, 2019), and dispersion of information (Lubik, Matthes and
Mertens, 2019)
15 Allowing for incomplete markets in the presence of nominal rigidities, as we eventually do in section 6,

does not by itself change the intuition underlying feedback rules, even if it changes the condition (Acharya
and Dogra, 2020, Bilbiie, 2024). However, Acharya and Benhabib (2024) show that in heterogeneous agent
New Keynesian models, if precautionary savings rise in recessions, this is a source of real indeterminacy
in real outcomes that requires interest rates to respond not just to inflation but also to the natural rate of
interest.

22
would mislead the econometrician to think the Taylor principle is violated (Lubik and
Schorfheide, 2004, Cochrane, 2011).
The general result is that since shocks to the feedback rule affect inflation, regressions
of policy rates on inflation inevitably give biased estimates of the feedback coefficient.
To estimate φ one needs to measure (or instrument) for changes in inflation that are un-
correlated with the monetary policy shocks. But, since an effective monetary policy will
respond to all shocks to inflation, in principle there are no such instruments.16

3.4.4 The 1990s and 2000s experience

Monetary policy in the United States during the tenure of Alan Greenspan (1987–2006)
closely conformed to what was prescribed by the rule of Taylor (1993). A version of equa-
tion (21) that includes the difference between the unemployment rate and a time-varying
natural rate plus two lags of the Federal Funds rate has an R2 of 0.97 on quarterly data
during the Greenspan era, but much less during the time of his predecessor, Paul Vol-
cker (Blinder and Reis, 2005). More broadly, feedback interest rate rules that satisfied
the Taylor principle became the established way to conduct monetary policy across the
world during the 1990s and 2000s (Leeson, Koenig and Kahn, 2013). The actions of the
ECB, which started setting monetary policy during this period, can be well described in
reference to a feedback rule for interest rates right up until 2013 (Hartmann and Smets,
2018).
During these close to twenty years, inflation was low and stable. Comparing every
twenty-year period over eight centuries of UK inflation, Reis (2023) finds that the period
1997-2016 had the best inflation outcomes. It had not been so before in all of the G-7
countries, and the most likely explanation was the new monetary regime (Cecchetti et al.,
2007).

3.5 Escape clauses

Why does moving from a peg to a feedback rule make such a difference? Imagine that
inflation is higher at date t by one log unit relative to our solution. Then, the central bank
will raise the interest rate on reserves by φ leading to an increase in expected inflation
between t and t + 1 of φ (the logic is the same for the other rules). But this in turn leads
16 Carvalho, Nechio and Tristao (2021) argue that as long as the output gap is used to proxy for xt , what
is left that drives policy has a small enough variance that the bias will be small.

23
the central bank to raise itv+1 by φ2 , which raises expected inflation between t + 1 and t + 2
by that amount. The process continues so inflation keeps on rising exponentially and the
feedback rule imposes inflation in T periods to be larger by φ T .
Inflation on target is the unique possible solution because the terminal condition ruled
out these deviations by imposing that the random variable Et πt+T − πt∗+T belongs to

O(ln(φ)). That is, if expected inflation deviates from target, those deviations cannot grow
faster than at the rate ln(φ). The larger is φ, the weaker is this condition.
But where did that condition come from in the first place? We discuss it now.

3.5.1 The elusive terminal condition

The terminal condition is not an optimality condition, the way that transversality condi-
tions are. Those apply to the real value of savings, whereas the condition needed here is
on a purely nominal variable, the price level. Additionally, optimal behavior imposes no
money illusion in the optimality conditions for savers or in the transversality condition.
Furthermore, there is no sense in which the economy blows up if this condition does
not hold. In the neoclassical economy, the unit of account may be exploding, but agents
with no money illusion would be indifferent as real outcomes continue to be finite. In the
New Keynesian model, real outcomes would explode, but assuming that prices would
remain sticky as inflation shoots to infinity is absurd.
Some authors argue that explosive paths for inflation are implausible. Perhaps people
would never believe them. More formally, if people’s expectations of inflation deviations
from target in the future are constrained to stay locally bounded, then Et πt+T − πt∗+T

is O(0). This would satisfy the terminal condition for φ > 1.17

3.5.2 Escape clauses as anchors

Escape clauses are valid and realistic justifications for the terminal condition. The idea
is that the central bank commits to the feedback rule only if inflation does not go on an
explosive path. If inflation exceeds a pre-announced threshold, the central bank switches
to a different policy approach. Realistically, if inflation start rising without bound, no
central bank would stick to following blindly a Taylor rule that tells it to raise policy rates
more and more, even as it sees inflation rising faster and faster.
17 Cochrane (2011) makes a scathing critique of these arguments.

24
 If the approach dictated by the escape clause pins down the price level at the date
of the switch, then it provides the terminal condition for the feedback rule. Formally,
the central bank follows the feedback rule only while inflation relative to target is within
some interval [ε L , ε H ]. If, at some date T, inflation π T relative to target π ∗ falls outside
this interval, then it switches to a different policy at T + 1. Take as given that this other
policy is able to determine uniquely π T +1 with some given effectiveness ε(s). It could,
for instance, set a real payment on reserves as we already saw, or involve any of the other
approaches this paper discussed in subsequent sections.18
Going back to the solution for inflation with a Taylor rule, by iterating the Fisher equa-
tion from t ≥ 1 up until a finite date T, we reach:

1 − φt− T
 
πt = (r (s) − x (s)) + φt−T −1 (π ∗ (s) + ε(s)) . (29)
φ−1

The right-hand side is uniquely pinned down, thanks to the switch in regime, so inflation
on the left-hand side is uniquely pinned down as well.
If the regime switch happens very rarely, then T is very large. The effectiveness of
this rule is still approximately given by the formula in equation (25), even if the regime
to which it switches has poor effectiveness. As long as the feedback rule itself is effective,
in keeping inflation close to target and away from the switching bounds, then the switch
may almost never happen.

3.5.3 Escape clauses as off-equilibrium threats

A different way of writing the regime switch ensures that it never happens. In this case,
it works as an off-equilibrium threat.19
Say that monetary policy is still committed to a feedback rule while the difference
of inflation from its target stays in a bounded interval. If inflation at date T is outside
of it, there is still a switch in policy that would uniquely pin down inflation, but now
this switch only happens next period, at T + 1. Moreover, now the new regime pins
down inflation to some level well inside the interval, and in particular to a level such that
18 The classic analysis is Obstfeld and Rogoff (1983), and see also Taylor (1996) and Christiano and Ros-

tagno (2001).
19 Much of this work builds on Bassetto (2005), and includes Atkeson, Chari and Kehoe (2010)’s sophisti-

cated equilibria, Christiano and Takahashi (2018)’s strategy equilibria and Loisel (2021)’s implementability
criteria.

25
π T +1 < π ∗ + ε H − r T .20
The Fisher equation (17) at date T together with the regime switch pins down ivT =
π T +1 − π ∗ + r T < ε H . At the same time, the Taylor rule at T implies that since π T was
larger than π H , and given that the Taylor rule coefficient is larger than one, ivT > π H . This
is a contradiction.
The only way to avoid the contradiction is for inflation to never leave the bounded
interval, and the switch to never happen. If the width of the interval is large enough
such that the size of the exogenous shocks would never send the economy outside the
interval, then the explosions that lead to indeterminacy with a Taylor rule are ruled out.
As the feedback rule implies that inflation explodes at rate φ, then one of the bounds will
be reached for sure in finite time for any inflation path that does not satisfy the elusive
terminal condition. Thus, the condition holds.
Just like in the previous case, the central bank is making the promise that it will not
stick to the Taylor rule if inflation enters one of the explosive paths that violate the ter-
minal condition. But now, the escape clause is inconsistent with equilibrium, and so it is
assumed that rational agents would never expect it to be used. This requires commitment
by the central bank.

3.5.4 Monetary pillars for the regime switch

With a feedback rule, the size of reserves is irrelevant for inflation, as long as it is large
enough to keep their demand satiated. Yet, if the escape clause is triggered, the central
bank can shrink Vt to the point where a convenience yield for reserves emerges. Just like
we discussed for banknotes in section 2.1.2, the Fisher equation now includes the quantity
of reserves. Fixing that quantity to some exogenous path uniquely pins down inflation.
We will carefully explain the intuition for how this leads to determinacy in section 5, but
for now note that this could be equally done by setting a quantity for banknotes.
In this case, the central bank is said to rely on a monetary pillar. This is understood
as a commitment to switch to a monetary approach to pin down inflation if the interest
rate approach leads to an exploding path for inflation. The ECB made this commitment
explicit until 2021.
The pillar also serves as an alert to central banks that have successfully used feedback
rules to control inflation for decades and think this is enough. Even though T may be
large, it is finite. Even in the case of off-equilibrium threats, they must be credible. Having
20 This may require nominal rigidities to make the policy consistent by having deviations from it be costly.

26
a monetary anchor as an escape clause behind the interest rate rule is crucial. Many
central banks have such monetary anchors, often in the form of gold reserves, or holdings
of foreign currency, even if they are rarely used.

3.5.5 The 1970s experience and the monetary pillar

Inflation was high during the 1970s across most advanced economies. One explanation
for why this happened is that central banks did not satisfy the Taylor principle in set-
ting interest rates (Clarida, Gali and Gertler, 2000, Coibion and Gorodnichenko, 2011). A
complementary explanation is that policymakers at the time settled for a higher inflation
target π ∗ (Meltzer, 2010), mis-estimated the state of the economy or mis-understood what
drove it rt (Orphanides, 2003, Romer and Romer, 2002), and neglected the measurement
and management of private-sector expectations (Reis, 2021). Each of these, and all com-
bined, could have contributed to inflation exploding as a result of either indeterminacy
or the lack of a terminal condition anchoring expectations. The conquest of US inflation
by Paul Volcker after 1979 came with a brief switch to monetarism (explained in a later
section), precisely what an escape clause would dictate.
An exception to the dismal inflation performance during the 1970s was West Germany.
The Bundesbank, both set interest rates following a feedback rule that satisfied the Taylor
principle but also had targets for monetary growth that made explicit the potential switch
to a monetary strategy if inflation ever got too far from target (Clarida and Gertler, 1997).
This experience had an important impact on the design of the ECB twenty years later as
following a two-pillar strategy. As described in Rostagno et al. (2021), from the 1998 initial
strategy to the 2003 review, the ECB emphasized the flexibility of potentially switching to
monetarism as a pillar that would stabilize expectations.

4 Unconventional interest rate policies

Starting in 1999, the Bank of Japan found itself unable to use feedback interest rate rules
to control inflation. The Federal Reserve, the ECB, and other major central banks faced
similar struggles following the great financial crisis. The main reason was that feedback
rules prescribed setting a very low policy rate, well below zero.
Central banks throughout the 2010s continued to rely on the forces of no arbitrage, and
to use interest rates as the policy tool. However, they now did it using different strategies,
which were labeled unconventional in spite of their persistent use (Bernanke, 2020). As

27
much, or even more than before, they relied on expectations of private agents, often well
into the future.
This section discusses these alternative interest rate policies. We show that they can
determine inflation, while at the same time arguing that in general they are inferior to
feedback rules for the interest paid on reserves, justifying their unconventional label.

4.1 Bounds on interest rates

Banknotes have a net nominal return ξ that is below but close to 0, since they pay no
interest but have storage costs and risk of theft. As banks would want to substitute all of
their reserves for banknotes if interest rates went below ξ, banknotes impose an effective
lower bound (ELB) on the payment of interest on reserves. The policy rule is now itv =
max{φπt + xt , ξ } (Benhabib, Schmitt-Grohe and Uribe, 2001).
To study the impact of this bound, we focus on what happens from period 1 onwards
so there is no uncertainty. The left panel of figure 1 plots the dynamics of the system
that results from combining the new policy rule with the log-linear Fisher equation (17).
As was the case earlier, if inflation is to the right of the target, π ∗ , then it grows without
bound, violating the terminal condition, and triggering the escape clause. This is not an
equilibrium.
What is new, however, is that once inflation is equal to ξ − r (s), it stays there forever.
This is a global steady state equilibrium of the difference equation: a deflation trap. If
π1 is below the target, inflation will fall. Instead of exploding, it will now converge to
the deflation trap. This is true for any initial π1 , so inflation is again indeterminate: any
initial value between ξ − r (s) and π ∗ (s) is consistent with an equilibrium.21
The same problem arises in the other direction. Sometimes, under pressure from the
Treasury to keep the interest rate on government bonds low, central banks do not raise
interest rates above a certain level (Reis, 2009). This upper bound can be captured as
a policy rule itv = min{φπt + xt , ξ }, where with abuse of notation we use ξ to denote
the upper bound. The right panel of figure 1 shows this case. There is also a second
global equilibrium steady state, but it is now above the inflation target. Thus, there is
indeterminacy of inflation, which may be well above the desired target.
Setting indeterminacy aside, the existence of two steady states implies that, in the
21 Christiano, Eichenbaum and Johannsen (2018) show that an e-stability restriction on the set of equilibria

delivers uniqueness. In the other direction, when expectations are backward-looking, there may be no
steady state and the economy spirals into ever higher deflation.

28
 Figure 1: Phase diagrams for inflation

(a) Lower bound (b) Upper bound

!!"# !!"#
Slope !
#−%
45o
45o
"∗
"∗

#−%
Slope !

&−%
"∗ !!
" ∗
!! !

presence of shocks to the state of the economy, to the inflation target, or to policy mistakes,
there will be two stochastic solutions fluctuating around these steady states. Then, if
there is a sunspot that triggers a change between them, equilibrium inflation will alternate
between being close to target or being close to the deflation trap (Mertens and Ravn, 2014,
Aruoba, Cuba-Borda and Schorfheide, 2017). Depending on the exogenous distribution
of this sunspot, the effectiveness of policy can be arbitrarily poor.
In general equilibrium, if prices are flexible, nothing changes. The interaction of nom-
inal rigidities with the effective lower bound does make a difference, that has been ex-
plored in large strand of the literature (Eggertsson and Woodford, 2003). While this is not
the place to survey it, we focus on how it affects the dynamics of inflation.
Once again, there is a permanent-deflation equilibrium since nominal rigidities do not
bind at a steady state. It is also the case that different initial prices p0 come with different
paths for inflation converging in finite time to that deflationary steady state (Werning,
2011, Cochrane, 2017). However, now small changes in how the nominal rigidities are
modeled, including whether prices are sticky as in Calvo (1983) or as in Rotemberg (1982),
or in how the sunspots that coordinate the equilibria are introduced, or even in what
numerical methods are used to solve the model in a log-linear approximation or globally,

29
seem to matter significantly for the properties of the equilibrium.22

4.2 Unconventional times

In principle, one can eliminate the deflation or high-inflation equilibrium with an escape
clause. However, with an inflation target of 2%, deflation is never too far, and escapes
would be too frequent, making the feedback rule not useful.
To pursue its feedback rule, the central bank would like to remove the bounds. Elimi-
nating the upper bound is usually done by having operational independence to set interest
rates as it sees fit, without any interference from the Treasury. The Treasury-Fed accord of
1952 is the classic example.
Eliminating the lower bound is harder, but there are measures to lower the log return
on banknotes ξ, perhaps all the way to minus infinity. Some suggestions in the literature
on how to lower ξ are to eliminate banknotes, charge a tax on them, or default on the
commitment to exchange currency and reserves one-for-one (Goodfriend, 2016, Rogoff,
2017, Agarwal and Kimball, 2019).
If, after trying these, the central bank finds itself unable to escape the bound, it can
still use interest rates and the power of arbitrage to steer inflation. For the remainder of
this section, we model these unconventional times in the following simple way: at date
0, we assume the interest rate is at its bound ξ, which we set to zero. The Fisher equation
(17) becomes:
r0 = − E0 (π1 ) . (30)

In the long run, from date 1 onwards, the limit does not bind and the central bank can go
back to using feedback rules to hit its inflation target π ∗ (s).

4.3 Forward guidance

From equation (30), the price level at date 0 is: p0 = r0 + E( p1 (s)). If the central bank
chooses to hit its inflation target, then p1 (s) = p1∗ (s). Since r0 + E( p1∗ (s)) 6= p0∗ , then p0
may be very far from p0∗ . Having lost its tool, the central bank cannot deliver inflation on
target.
Forward guidance is an unconventional interest-rate tool to try to remedy this situa-
tion. Say the central bank announces that it will keep itv = 0 for several periods, until
22 See
Fernández-Villaverde et al. (2015), Boneva, Braun and Waki (2016), and Christiano, Eichenbaum
and Johannsen (2018) among others.

30
date T. As a result, in period 1, the price level now is p1 (s) = Tr (s) + p∗T +1 (s). The cen-
tral bank has delayed achieving its target until after period T, when finally it will get to
p T +1 = p∗T +1 (s). In exchange, by manipulating E( p1 (s)), it can bring p0 closer to its target
p0∗ .
This strategy involves a trade-off: to get inflation closer to target in the short run, the
central bank has to sacrifice inflation being away from target in a potentially long medium
run. If it finds itself unconstrained in period 0, so in conventional times, the central bank
would never want to do this. But, in unconventional times where it cannot use i0v , then
since it started by hitting the target exactly in that medium run, by the envelope theorem
it will always want to do some forward guidance whenever it hits the lower bound.
In sum, the central bank chooses to keep the interest rate on reserves pegged in the
future, even though it is not constrained by the state of the economy to do so. This un-
conventional announcement of a path for the policy rate is sometimes called Odyssean
forward guidance, to distinguish it from the conventional communication of the state of
the economy or Delphic forward guidance.23
With the New Keynesian model, forward guidance comes with side effects. To save
on needless terms, assume that all shocks are zero: rtn = 0 and ỹt = yt . This implies that
inflation at T + 1 is on target and output is zero. Then, in period T, equations (9) and
(10) imply that y T = θπ ∗ and that π T = βπ ∗ + κy T , respectively. By keeping the interest
rate at zero, the central bank causes a boom that in turn pushes inflation higher. The
previous period, output is even higher since y T −1 = y T + θπ T , and likewise inflation will
be π T −1 = βπ T + κy T −1 . Continuing until 0, then p0∗ can be brought closer to target by
choosing T adequately, just as before. The combination of the peg with the Calvo Phillips
curve makes forward guidance in the distant future a powerful tool to control inflation in
the present.
What is new, along this path, is that output is above its natural level. Forward guid-
ance causes a boom from period 1 onwards, and if T is sufficiently large it may cause a
boom in period 0 as well.
This result has been called the forward guidance puzzle since it is easily contradicted
by empirical estimates of the effects of forward guidance.24 At the same time, the litera-
23 Disentangling Delphic from Odyssean forward guidance is empirically challenging even with high-
frequency data (Gürkaynak, Sack and Swanson, 2005, Campbell et al., 2017, Andrade and Ferroni, 2021).
24 The puzzle was identified in Del Negro, Giannoni and Patterson (2023) and Carlstrom, Fuerst and

Paustian (2015). For empirical estimates, see Nakamura and Steinsson (2018), Cieslak and Schrimpf (2019)
and Lunsford (2020)

31
ture has found that limits to rationality, incomplete insurance markets that change the IS
relation in equation (9), or different models of price rigidity like sticky information that
change the Phillips curve in equation (10) can weaken the puzzle or make it go away.25

4.4 Quantitative easing

Following the great financial crisis, interest rates in the Euro area stayed near zero for
almost one decade. Following the Fed’s experience in previous years, the ECB purchased
long-term bonds with reserves in 2014. These policies became known as quantitative
easing.
To understand how they work in our general model, consider another financial asset
j = o that is scarce, so it carries a convenience yield qot . The optimality condition for
savers in equation (8) applies to it, as it does to any asset. Importantly, the convenience
yield depends on the quantity held by the private sector, just as with banknotes.
Combining the optimality conditions for this new asset and for pure real bonds at date
0, gives the log-linearized arbitrage condition:

p0 = r0 + E( p∗ (s)) − η o b0 (o ) + u0 (o ) (31)

where η o is the elasticity of the convenience yield with respect to quantities, and u0 (o ) are
shocks to it. When the central bank buys long-term bonds with reserves, it increases vt
and reduces bt (o ). Whereas before the central bank had an interest rate to steer inflation
given a real state of the world, now it has the convenience yield taking its role. Still, by
the same no-arbitrage logic, it is able to ensure determinacy and effectiveness.
This equation also shows the limitations of quantitative easing. The key coefficient η o
is hard to estimate but seems to be small (Krishnamurthy and Vissing-Jorgensen, 2013,
Fabo et al., 2021). In the spirit of the Taylor principle, this implies very large bond pur-
chase programs when inflation is below target, in order to bring it back on target.
Moreover, the price of long-term bonds can be quite volatile, for reasons that seem
divorced from expectations or risk. These are often called liquidity shocks, and are cap-
tured by the u0 (o ) term. With large shocks to u0 (o ), the central bank would have to adjust
its asset purchases to counteract its effect, creating a new source of deviations of inflation
25 Angeletos and Lian (2018), Gabaix (2020), and Garcı́a-Schmidt and Woodford (2019) study deviations
from perfect-foresight rationality in this context; Del Negro, Giannoni and Patterson (2023) and McKay,
Nakamura and Steinsson (2016) explore incomplete insurance against income risks by households, and
Carlstrom, Fuerst and Paustian (2015), Kiley (2016), Eggertson and Garga (2019) explore sticky information.

32
from target. The literature has struggled to find a sizable and persistent impact of liq-
uidity shocks on inflation (Krishnamurthy and Vissing-Jorgensen, 2013, Fabo et al., 2021),
perhaps because, as financial markets adjust to central bank actions, the u0 (o ) shocks are
correlated with monetary policy, x0 .
In effect, through quantitative easing, the central bank is targeting long-term interest
rates. Historically, the source of doing so has been the Treasury rather than the central
bank. Especially in the aftermath of wars, when long-term government debt is high,
fiscal policy imposes low long-term interest rates. This has side effects on the balance
sheet, which we will discuss in section 6.

4.5 Going long

Both forward guidance and quantitative easing focus monetary policy on long-term in-
terest rates. The Bank of Japan went the furthest in this regard by announcing a desired
target for the 10-year interest rate, standing ready to buy and sell government bonds of
this maturity to hit the target. This policy was called yield curve control.
In theory, if the central bank issued bonds of a fixed maturity that were later paid off
with reserves, it could choose how to remunerate these bonds just as it does with reserves.
Concretely, if the central bank issues a two-period bill at date 0, and pays an interest rate
of I0 (2) on it, then the optimality conditions for savers in equation (3) that applies to this
new pure financial asset is:

M(s)2 (1 + I0 (2))
 
E = 1. (32)
Π1 ( s ) Π2 ( s )

By choosing a feedback rule for I0 (2) in much the same way as it did for one-period
reserves, the central bank can control the price level. The condition for determinacy still
requires φ to be larger than some threshold, but the threshold is now equal to the sensi-
tivity of long rates to short rates. The effectiveness of this policy involves similar terms
but with different weights (McGough, Rudebusch and Williams, 2005, Reis, 2019b).
There is an ingenious alternative where the central bank chooses a short-term inter-
est rate, and a sequence of long-term interest rates at different horizons (say by issuing
reserves of different maturities). For each of these, there is an optimality condition like
equation (32). If there are as many of these as there are states of the world at date 1 then,
potentially, inflation can be pinned down.

33
 To see this in action, assume that there are only two states of the world revealed at
date 1, s H and s L , with probabilities f and 1 − f , respectively. From date 1 onwards, the
central bank chooses two separate nominal interest rates I (s H ) and I (s L ) and these pin
down inflation from date t = 2 onwards at two separate levels. If the central bank only
chose the short-term interest rate I0 , then this would only pin down expected inflation at
date 1, but not its separate values per state: Π1 (s H ) and Π1 (s L ).
If the central bank also chooses at date 0 a 2-period rate I0 (2) then we now have two
versions of the optimality condition for savers in equation (3):

f M ( s H ) (1 − f ) M ( s L )
 
(1 + I0 )) + = 1, (33)
Π1 ( s H ) Π1 ( s L )
(1 − f ) M ( s L )
 
f M(s H )
(1 + I0 (2)) + = 1.
Π1 (s H )(1 + I (s H )) Π1 (s L )(1 + I (s L ))

As long as I (s H ) 6= I (s L ), then these two equations have a unique solution in the two
unknowns Π1 (s H ) and Π1 (s L ) (Adão, Correia and Teles, 2014, Magill and Quinzii, 2014).
Intuitively, if the level of the short-term rate pins down expected inflation, now the
long-term rate pins down how it covaries with the intertemporal marginal rate of sub-
stitution. This is sometimes called the inflation risk premium. By pinning it down, the
central bank achieves the determinacy of inflation across states of the world.
This approach has clear limitations. First, it does not pin down the initial price level P0 .
Only the stochastic degree of indeterminacy disappears. Second, counting the number of
possible states of the world, and responding with just as many maturities for reserves, is a
hopeless task. Third, while the effectiveness of a conventional feedback interest rate rule
depends on the real interest rate—the expected value of the stochastic discount factor—
the effectiveness of going long requires evaluating the stochastic discount factor in every
state. Fourth, slight mis-calibrations of the short and long interest rates would induce
savers to take very large positions in one versus the other causing wide fluctuations in
the composition of the liabilities of the central bank.

4.6 The cost of credit

Going back to equation (30), the central bank could try to affect r0 directly in order to
steer p0 . An unconventional tool used in the 2010s were credit policies, whereby the
Bank of England (through the Funding to Lending scheme) and the European Central
Bank (through the Targeted Long-term Refinancing Operations) lent funds to banks at

34
favorable rates under the condition that these funds would then be used to provide loans
to firms. The goal was to lower the cost of bank credit.
Since bank credit is an input in production, if its cost falls, so will the marginal costs
of production Ct (i ). With flexible prices, this real cost of credit would still be determined
with other real variables independently of inflation. It takes a nominal rigidity, like loans
being set in sticky nominal amounts, or their rates in sticky nominal units, for this to lead
to another transmission channel of monetary policy over inflation. Through credit policy,
the central bank can then affect the real costs of credit, marginal costs of production, and
through the Phillips curve, the optimal price set by firms (Christiano and Eichenbaum,
1992, 1995, Barth III and Ramey, 2001, Christiano, Trabandt and Walentin, 2010, Fiore and
Tristani, 2013).
Beyond firm credit, similar mechanisms could operate through household credit, es-
pecially on mortgages that have features set in sticky nominal terms, affecting demand
for goods as opposed to supply (Greenwald, 2018, Berger et al., 2021). A third channel
through which lending rates can affect credit is if they affect the net worth of borrowers
and tighten borrowing constraints (Bernanke, Gertler and Gilchrist, 1999).
While there is strong evidence for a credit channel of monetary policy (Ciccarelli, Mad-
daloni and Peydró, 2015, Gertler and Karadi, 2015), using credit supply or credit rates as
the main strategy to control inflation is rarely used today. When tried in the United King-
dom in the 1950s in the context of the Radcliffe report, it failed (Capie, 2010). Central
banks have an influence on lending conditions, but are very far from controlling them.
There are large financial shocks in lending markets that would translate into large fluctu-
ations in inflation.

4.7 Non-rational expectations

The final term in equation (30) that the central bank could affect is the expectations op-
erator directly. With rational expectations, this is not possible, but moving away from
it creates room for policy to steer “animal spirits”. More generally, all of the unconven-
tional policies require affecting expectations in financial markets. Even in conventional
times, the feedback rule requires that people do not start expecting that inflation in an
arbitrary far-away future will grow (or fall) at an explosive rate, so it relies heavily on ra-
tional expectations into the infinite future. This section discusses the relaxation of rational
expectations.
The literature on non-rational expectations is too rich to cover here and has already

35
been reviewed by Woodford (2013). Instead, we just describe three approaches that have
been used to study the control of inflation.
The first are learning models that assume that expectations are formed by agents that
behave like statisticians using past data to form their beliefs. A learning rule gives a
mapping from past outcomes to current expectations. In turn, recall equation (22), from
combining the Fisher equation with the interest rate rule, which mapped expectations
into outcomes (this is sometimes called a temporary equilibrium). Combining the two
gives the learning equilibrium.
The most popular such model is least-squares learning, where agents use least-squares
regressions on past outcomes to form their beliefs (Eusepi and Preston, 2018). Taking the
limit, as the sample for these regressions goes to infinity, delivers what is known as the
learnable equilibrium. The literature focuses on the e-stability principle, that establishes
that learning converges to the non-explosive rational expectations equilibrium if certain
stability conditions hold. In our simple model with constant r and x after period 1, one
can show that φ > 1 makes this learnable and e-stable (Evans and Honkapohja, 2001,
Bullard and Mitra, 2002, McCallum, 2003).
Another popular class of non-rational expectations models are models of e-duction.
Their central idea is that agents go through a mental process whereby they iterate on
what expectations to have, and what their implications are for equilibrium inflation, until
the two converge. This convergence need not happen at the fixed point of rational ex-
pectations, nor does it have to happen over time, like with learning, but rather occurs in
agents’ minds. For instance, with reflective expectations, at each stage of inference, agents
update their expectations to close the gap to the expectations that are model consistent.
In that case, it turns out that in the limit, as the rounds of reflection go to infinity, only the
non-explosive rational expectations equilibrium is selected.26
Third, there are models of discounting the future through limited foresight (Gabaix,
2020) or the past through imperfect memory (Angeletos and Lian, 2023). Both imply
that, either looking forward or backwards, current inflation depends less on far-away
expectations. Because of that, both can deliver determinacy of inflation without escape
clauses and with conditions on φ that are less strict than the Taylor principle. Sometimes,
limits to rationality or information are enough to select one of the multiple equilibria that
arise even with an interest rate peg.
26 SeeGarcı́a-Schmidt and Woodford (2019), building on the calculation equilibrium of Evans and Ramey
(1992) for the reflective case, and Farhi and Werning (2019) for k-level thinking.

36
 More generally, once one entertains non-rational expectations, then measuring expec-
tations becomes important as an independent source of data and shocks. As much, or
more, than measures of the output gap or natural rates of interest, these data on infla-
tion expectations become part of the state of the economy, xt , that an effective policy rule
should include to keep inflation near its target (Reis, 2022).

5 Steering inflation using the money supply

Monetarism has had many proponents in the past. Yet, it has rarely been voluntarily
adopted by advanced-economy central banks. Sometimes they have been forced to follow
it by Treasuries, usually with dismal consequences. This section discusses the active use
of Ht to control inflation (while interest on reserves and income shortfalls are passive)
and the economic force through which it controls inflation.

5.1 The economic force that drives inflation

The optimality condition for saving in banknotes comes from combining equation (3)
with equation (11) to get:
"    −1/η h #
Pt Ht
Et Mt +1 + Ut = 1. (34)
Pt+1 Pt Yt

This is sometimes called a money demand equation: it links the banknotes that people
want to hold, captured by the convenience yield in the second term on the left-hand side,
to the opportunity cost of holding those banknotes, captured by the interest rate foregone
on pure nominal bonds in the first term on the left-hand side.
Recall that reserves paid an interest rate that the central bank could choose, and they
gave no additional convenience because central banks abundantly supply them. Because
banknotes pay no interest but give a convenience yield they bring a different economic
force to drive the price level.
It works as follows: all else equal, a higher Pt today lowers the real balance of ban-
knotes in the hands of the private sector. This higher scarcity raises their convenience
yield via the second term in the equation. At the same time, a higher Pt raises the first
term in the equation: it lowers expected inflation between the present and the next pe-
riod, so it lowers the shadow nominal interest rate. With both terms higher, only if Pt falls

37
will their expected sum be one again.
This mechanism can be explained using a standard demand-supply story. When the
price level is higher, the supply of banknotes in real terms is lower (the second term). At
the same time, because expected inflation is lower, the opportunity cost of holding ban-
knotes is smaller, so the demand for them is higher (the first term). With lower supply
and higher demand for banknotes, the price level must fall. This re-equilibrates the mar-
ket by both increasing the supply, and by lowering demand through a higher nominal
interest rate.
This Marshallian-sounding supply and demand story is appealing but it can also be
misleading. After all, we saw that nothing about demand and supply pins down the price
level in section 1. The previous paragraph uses a sleight of hand: the price level is not the
price of the banknotes. Changes in the price level bring the market to equilibrium by
both affecting the opportunity cost of currency and by directly changing the quantity of
real banknotes that is held.

5.2 Money growth rules

The log-linearized version of equation (34), from combining equations (8) and (11), is:

ht − pt = ct − η h (rt + Et πt+1 ) + η h uth . (35)

The classic monetarist policy rule is: ht = π ∗ (s)t + xt , so the central bank grows the sup-
ply of currency at the desired inflation target. This rule depends on exogenous variables,
so it is not a feedback rule.

5.2.1 Determinacy

Replacing the policy rule into equation (35) and iterating forward gives a unique solution
for inflation. In the long run, for t ≥ 2, the central bank exactly achieves the target for
inflation, while beforehand, inflation is on target if x0 and x (s) are picked adequately.
As before, this result depends on a coefficient condition and on a terminal condition.
The condition is now that η h > 0. In other words, it is simply that the demand curve of
banknotes slopes down.27 This is not a condition on the policy followed by the central
27 Different
micro-foundations for money imply different forms of the money demand function, and as
such potentially slightly different conditions for determinacy of the price level. Still, the basic result and
economic intuition remains, see Carlstrom and Fuerst (2003).

38
bank, nor is it a requirement for feedback as before. Rather, it is a statement of the sign of
the economic force that we just described.
As for the terminal condition, it is that limT →∞ M (s) T (h T − p T ) = 0. This is a version
of the no Ponzi scheme condition for banknotes, as in equation (14). Again, it is a feature
of the economic environment, not of the policy regime. There is no need for an escape
clause. In fact, monetarism is a possible escape clause, in the sense discussed in section
3.5. Instead of shrinking the supply of reserves, the central bank can switch to targeting
the supply of banknotes, and it will achieve determinacy.

5.2.2 Effectiveness

In the short run, the monetarist rule leads to:

1   ηh
p0 = h
x 0 + η h
r 0 − η h h
u 0 − c 0 + h
E( x (s) + η h r (s) − η h uh (s) − c(s)) + η h E(π ∗ (s)).
1+η 1+η
(36)
Whether the central bank achieves the desired price level depends on how it chooses x0
and x (s). Higher supply of banknotes, in the present or the future, unambiguously raises
inflation.
Just like with interest rate rules, the central bank has to keep track of the state of the
economy and adjust policy in response. That involves estimating both the real interest
rate and the level of consumption.
Even harder, the central bank also needs estimates of uth . This shock to the demand
for central bank currency arises from the difference between the banknotes the central
bank prints and the money that people find useful. Because there are close substitutes
to currency produced by the private market, including foreign currency, the uth are large
and volatile. For instance, uth arises from changes in the availability of ATMs, in the social
norms of what shopkeepers will accept as payment, or in the prevalence of crime that
drives the demand for the anonymity of banknotes. The volatility of money demand,
confirmed by study after study, severely undermines the effectiveness of monetarism at
controlling inflation (Poole, 1970).
The effectiveness of monetarism in the short run is also not necessarily compensated
by effectiveness in the long run. The result that the central bank exactly hits the target in
the long run relies on our assumption that, from period 1 onwards, consumption growth,
real interest rates, and the demand shocks would all be constant. Otherwise, fluctuations
in any of these would lead to deviations of inflation from target, just like they did in the

39
short run. When it comes to the shocks to the demand for banknotes, there is constant
financial innovation on means of payment with long-run consequences that will prevent
the central bank from hitting the inflation target.

5.2.3 General equilibrium breakdowns of the classical dichotomy

With nominal rigidities, the changes in money and inflation come with changes in real
interest rates and output. Again, nominal rigidities bring into play an aggregate demand
channel. As households hold more money, this raises aggregate demand, which leads
to an increase in production and prices by firms. Sticky prices spread the short-term
volatility of inflation that is due to financial shocks into volatility of output.
Monetarism points to alternatives to sticky prices in breaking down the classical di-
chotomy. If banknotes are used in transactions, their outstanding stock will facilitate
trade. For instance, real money balances may enter the utility function (separable from
consumption) and so affect the discount factor M(s). There are several well-justified
models of these interactions (Lucas and Stokey, 1987, Lagos and Wright, 2005). The sup-
ply of banknotes may also affect marginal costs and so the production of goods, for in-
stance by affecting the creation of bank deposits and bank lending (Brunnermeier and
Sannikov, 2016). These additional monetarist channels come with further shocks that
could raise inflation volatility further.

5.2.4 The evidence

In the early 1980s, the US and the UK both briefly adopted money growth rules. Nominal
interest rates were very volatile, as were expected annual inflation rates. An empirical
regularity emerged—Goodhart’s law—stating that once the central bank started using a
policy rule for one measure of ht , the corresponding uth shocks would turn to be even
larger than anticipated before. Monetarism is still a useful strategy in scenarios where the
central banks lacks credibility in the escape clause of its interest rate rule, so that volatile
short-term inflation is tolerable in return for stable long-run inflation.

5.3 Seignorage
Recall that the seignorage earned from printing banknotes is StH = ( Ht − Ht−1 )/Pt .
Seignorage and inflation are tightly linked. A sudden increase in the supply of banknotes,

40
all else equal, raises seignorage. However, as we just saw in the previous section (equa-
tion (36)), it also raises the price level, which lowers seignorage. Moreover, an increase
in expected inflation comes with higher nominal interest rates, which lowers the demand
for currency (equation (35)) and lowers seignorage.
If the net worth of the central bank is constant, and it only holds short safe bonds
as assets, then its dividends to the fiscal authorities are equal to this seignorage minus
expenses Dt = StH − Et . The central bank may be committed to delivering an exogenous
dividend, just like a government fiscal agency that has a target for tax revenues, or a state-
owned company providing a public service with a target for profits. Historically, this was
common, as central banks have been asked for centuries to provide fiscal resources for
the sovereign. Only in the past few decades did inflation targeting replace seignorage as
the primary task for the central bank.

5.3.1 Limits to seignorage

In a steady state, seignorage is given by:

  
H H̄ 1
S̄ = 1− (37)
P̄ Π̄

The first term on the right-hand side falls with inflation, since higher inflation comes
with lower holdings of banknotes. The second term rises with inflation. Overall, higher
steady-state inflation may raise or lower seignorage: it depends on how elastic the de-
mand function for banknotes is.
History suggests that seignorage at first increases with inflation, and then falls. Esti-
mating the peak of this Laffer curve is an understudied topic. But, in practice, once central
banks are turned into fiscal agents, this tends to lead to run-away inflation as the central
bank is asked to deliver more resources than the upper bound on feasible seignorage al-
lows. Moreover, the peak of the Laffer curve likely moves around and is flat so that small
changes in seignorage close to its peak come with large changes in inflation.

5.3.2 Seignorage policy rules

A seignorage policy rule will print banknotes to ensure that seignorage equals an exoge-
nous amount: sth = xt . Log-linearizing the definition of seignorage and replacing for
banknotes using the demand equation (35) gives a difference equation between inflation

41
and seignorage. Its version in the long-run is:

πt + η h Et−1 (πt ) = η h Π̄ Et (πt+1 ) + (Π̄ − 1) x (s) − (Π̄ − 1)(c(s) − η h r (s) + η h uh (s)) (38)

The determinacy of the price level is again ensured by the monetarist forces. Iterating this
equation forward, it is enough that η h Π̄ > 1, which again depends on the features of the
economy.
With a seignorage policy rule, both the interest rate and the stock of banknotes follow
endogenously, from the Fisher equation and the demand for banknotes, respectively. If
the central bank is mandated to hand in a dividend, then the exogenous x (s) that follows
will likely deliver an inflation that is far from the target.
If, instead, the central bank was free from government pressure, then in principle
it could adjust its target for seignorage to control inflation. While we are unaware of
any independent central bank pursuing such an explicit seignorage policy rule, fiscal
authorities do trade off their desires for dividends on the one hand, and for inflation
on target on the other hand. The actual x (s) may then trade off the seignorage policy
rule that delivers the desired dividends with the one that is most effective at delivering
inflation on target.
As the equation above shows, using seignorage as a strategy to control inflation is
likely to perform poorly. As with all monetarist theories, policy will struggle to keep up
with the volatility of the shocks to the demand for banknotes uh (s). Moreover, calibrat-
ing the response to these via x (s), depends on pinning down the steady-state level of
real banknotes held, which is likewise hard given the extent of innovation in payment
systems.

5.3.3 Experience with seignorage

Seignorage policy rules are not a theoretical curiosity, but a common occurrence, espe-
cially one that is imposed on the central bank by the Treasury (Sargent and Wallace, 1984).
In Latin America in the 1980s and 1990s, the extent to which these fiscal pressures fluctu-
ated from year to year can explain some of the movements in actual inflation rates (Kehoe
and Nicolini, 2022). Further back in history, many hyperinflations were associated with
seignorage policy rules (Cagan, 1956), and their ends with fiscal reforms that lowered the
government’s demand for fiscal revenue from the central bank (Sargent, 1982).

42
5.4 Different monies: banknotes, reserves, and more
Both reserves and banknotes are sometimes referred to as money. Yet, they provided
different approaches to control inflation. Substantially, the two differences between these
two monies were whether they paid interest (currency no, reserves yes) and whether they
gave a convenience yield (currency yes, reserves no). This section relaxes these differences
and, in doing so, further expands on the link between the two approaches to inflation.
Finally, most people make most transactions using neither currency nor reserves, but
with a third type of money, deposits at private banks. We discuss its link to inflation.

5.4.1 Reproducing monetarism with an interest rate rule

We can rewrite the equilibrium in the market for banknotes in equation (35) as:

pt ct + uth − ht
it = + . (39)
ηh ηh

This is mathematically equivalent to a Wicksellian interest rate feedback rule. Since

1/η h > 0, it satisfies the determinacy condition. But while in section 3 this was a pol-
icy rule, here it emerges as an equilibrium condition.
The link to interest rates under a monetarist approach arises because the nominal in-
terest rate it adjusts endogenously so that the market for banknotes clears. Canzoneri,
Henderson and Rogoff (1983) blur this distinction by specifying a feedback rule for cur-
rency that depends on the nominal interest rate: ht = xt + φit . In this case, the central
bank can limit the volatility of the nominal interest rate. In fact, it can even peg it to
follow a pre-determined path, while inflation remains determinate.

5.4.2 Non-interest paying reserves

Before 2008, reserves paid no interest in the United States. Yet, even though nominal
interest rates were well above zero, banks still voluntarily chose to hold (small) amounts
of reserve balances at the central bank. These allowed the banks to settle unexpected
transactions with other banks at the end of each day, especially when people ran to the
banks to withdraw their deposits. In other words, reserves paid a convenience yield.
The central bank then engaged in open market operations whereby it would change
the supply of reserves to hit a target for the nominal interest rate. Just as with banknotes,

43
in equation (39), the central bank would target it by varying reserves and their conve-
nience yield. The level of it would then follow a feedback rule, just as in section 3.
Given that level of the interest rate and the level of inflation, the central bank would
passively accommodate the demand for banknotes following equation (35). The final
piece of the puzzle was to stand ready to exchange reserves for banknotes one-to-one at
all times as people preferred one or the other. Effectively, the central bank would only
control the sum of banknotes and reserves (the monetary base).
The control of inflation was a hybrid of the monetarist and the no-arbitrage approaches.
On the one hand, it was a monetarist principle applied to the narrow market for reserves
that allowed it to control the interest rate. But, on the other hand this interest rate then
followed a feedback rule that through the force of no arbitrage allowed it to control infla-
tion.
Since the great financial crisis, the market for reserves has been satiated, eliminating
the monetarist force. Interest has been paid on these reserves, with this policy rate used
to directly shift the safe interest rate (Ihrig, Meade and Weinbach, 2015).

5.4.3 Scarce interest-paying reserves

In principle, central banks could pursue a novel approach, where they keep reserves
scarce, but pay interest on them. This may be particularly relevant in a future where
households can have digital deposits at the central bank just like banks do today. The
benefits from using currency for payments will extend to reserves, and the demand for
them may be at a much higher level than in the past.
The optimality condition for savers in equation (3) applied to reserves, when they
pay interest, combined with the same optimality condition for pure nominal bonds and
ignoring uvt for simplicity, gives in log-linear terms:

vt − pt − ct = η v (it − itv ). (40)

Whereas before, with satiated reserves, the interest rate on reserves was equal to the inter-
est rate on bonds, now the gap between them it − itv gives the opportunity cost of holding
reserves that households trade off against their convenience benefit.
The central bank can now choose both vt and itv (Diba and Loisel, 2021). In particular,
consider the case where it follows a Wicksellian rule, whereby the interest on reserves
responds to pt with a coefficient φ. In that case, the price level is determinate as long as

44
φ > −1/η v . This includes the case where φ = 0, that is where there is a pure interest rate
peg. The logic is that of the monetarist approach. With two policy tools, the central bank
can potentially get closer to tracking the variables it must offset to keep inflation close to
its target.

5.4.4 Bank deposits as money

Most households use their bank deposits to engage in transactions. The version of the
optimality condition for savers in equation (3) that would hold for the demand for bank
deposits hdt relates them to their opportunity cost, the gap it − itd , where the interest rate
paid on deposits is itd .
The central bank does not control either hdt or itd . Both are determined by the equilib-
rium in the banking sector. However, banks also deposits reserves at the central bank and
can invest in financial assets (Piazzesi, Rogers and Schneider, 2022). Optimality in their
portfolio choice leads to a log-linearized relation of the form it − itd = `(it − itv ). In section
3, implicitly there were competitive frictionless banks absorbed into the private sector, so
that in equilibrium ` = 1, leading to it = itv . With market power of banks, or financial
frictions, ` < 1. Combining these two equations:

hdt − pt = ct − η d `(it − itv ). (41)

If the production of deposits by banks was exogenous with respect to the price level,
then by choosing the interest on reserves, the central bank could again control inflation.
Even though the policy tool is the interest rate, the economic logic is the monetarist one,
as the key equation is the demand curve above.
If, instead, the quantity of reserves affects the amount of deposits—a money-multiplier
process—then we are back at the previous case, of scarce-interest paying reserves, where
the central bank has two tools, vt and itv , with which to improve the effectiveness of infla-
tion control.
Either way, while central banks’ digital currencies, more realistic banking sectors, or
scarce reserves all affect the dynamics of inflation, the economic logic and the policy ap-
proach by which the central bank can control it are unchanged.

45
5.5 Pegs
For many emerging and developing economies, the most common approach today to pin
down inflation is to peg their currency to another country’s currency. This was also the
case for most of the advanced world during the Bretton Woods regime, where the peg
was to the US dollar between 1944 and 1976 (Bordo, 2017). Until 1971, the US dollar in
turn was convertible to gold, following an even older tradition that started in the 1870s
of pegging the currency to gold. This is a type of monetarist approach that does not
involve banknotes or rules for their supply. Instead, the central bank is following a rule
of exchanging domestic reserves for either a commodity or a foreign money.

5.5.1 Commodity pegs

Combining the equality of the marginal rate of substitution of good i with good 0, ρt (i ),
to their relative prices, pt (i ) − pt (0), and the definition of the price index, pt , in equations
(5)-(6) gives:
I I
pt = ∑ ωi p t ( i ) = p t (0) + ∑ ωi ρ t ( i ). (42)
i =0 i =1

Recall that the parameters ωi are non-negative and sum to one, reflecting the weights of
each good in the price index.
The central bank can announce it will denominate reserves in the units of good 0. Since
it can issue reserves in unlimited amounts, the central bank can enforce this denomination
by buying good 0 with the reserves and holding it. Seignorage is no longer distributed as
dividends to the fiscal authority. This way the central bank can always buy and sell good
0 with reserves to keep their relative price at one forever.
This uniquely determines inflation. From equation (42), having defined that pt (0) = 1,
the price level pt is unique. No expectations of the future or terminal conditions are
involved, because the central bank is relying on pegging the value of reserves relative to
a commodity.
With this strict peg, changes in relative prices would lead the price level to devi-
ate from target. The central bank could adjust the peg to estimates of relative-price
movements using a rule pt (0) = p∗t − ∑iI=1 ωi ρ̂t (i ). The effectiveness would then be:
ε t = ∑iI=1 ωi (ρt (i ) − ρ̂t (i )).
Changes in the supply of good 0, or in the public’s taste for it, become sources of
deviations of inflation from target. Moreover, if good 0 is a complement with others in

46
consumption, then the impact on relative prices across all goods ρt (i ) can be large. The
ideal commodity to peg the price level to has to: be storable, have a stable supply, and
not be complementary or substitutable with many other goods.
Gold or other precious metals meet these criteria and this is why they have often been
used with this approach. Still, relative-price movements are large enough that commodity
pegs have tended to generate large ε t (Bordo, 2005).

5.5.2 Exchange rate pegs

Today, it is more common to peg to a foreign currency. This is especially the case in
small open economies, which import goods from other countries, often denominated in
a dominant foreign currency. A currency board consists of using the same strategy as
in a commodity peg, but where reserves are now exchanged for a foreign currency (or a
basket of currencies).
The economic logic of how they work is the same. Take the first J + 1 goods to be
domestic, but the remaining I − J goods are foreign. Their price in domestic units is then
equal to a foreign price pt ( j) plus the log exchange rate between the domestic and the
foreign units of account: et . Letting α denote the measure of home bias in consumption,
equation (5) for the domestic price level is then equal to:

J I J
p t = α ∑ ωi p t ( i ) + (1 − α ) ∑ ωi ( pt ( j) + et ) = α ∑ ωi pt (i ) + (1 − α)( pt + et )
f
(43)
i =0 j = J +1 i =0

f
where pt is the price index of the imported goods in foreign currency.
The optimality condition between any two domestic and foreign goods in equation (6)
is: ρt (i, j) = pt (i ) − pt ( j) − et , where ρt (i, j) is the marginal rate of substitution between
consumption of domestic good i and foreign good j. It then follows that: ∑ jI= J +1 ω j ρt (i, j) =
f
pt (i ) − pt − et . Replacing for pt (i ) in equation (43) delivers:

J I
+α∑ ∑
f
pt = et + pt ωi ω j ρt (i, j). (44)
i =0 j = J +1

The second and third term on the right-hand side are exogenous with respect to the price
level. An exchange-rate target peg is a choice of et . Thus, it uniquely pins down the price
level.

47
5.5.3 Experience with pegs

The peg of the Hong Kong dollar to the US dollar is perhaps the most famous case of
a successful currency peg. In place since October of 1983, the standard deviations of
the monthly log nominal exchange rate between the two currencies has been 0.004 over
these almost 40 years. And yet, inflation in Hong Kong dollars has been significantly
more volatile than inflation in US dollars, peaking at 11% in 1991 and bottoming at -4%
in 1999. In part, this happens because choosing et to be constant over time implies that
changes in ρt (i, j), and consequently in the real exchange rate, lead to wide fluctuations
in pt (Obstfeld and Rogoff, 1995, Ilzetzki, Reinhart and Rogoff, 2019).
Currency boards are rarely adopted for two practical reasons. First, central banks
often have conflicting goals for the desired price level πt∗ and for the exchange rate et .
When pursuing one of them has unpleasant consequences for the other one, the currency
board is abandoned.28 Second, currency boards require that the central bank keeps all the
foreign currency it buys with its reserves, so it is ready to buy and sell it as needed. In
reality, there is pressure on central banks to distribute some of these assets as dividends,
or to exchange the foreign currency for domestic government bonds. They then run out
of the foreign currency when they need to sell it, and the peg fails.
Without foreign assets to back the reserves, countries that try to maintain exchange
rate pegs rely either on choosing the interest rate on reserves to mimic movements on
foreign interest rates, or in adjusting the supply of money to control its relative scarcity
relative to the foreign currency, or a mix of both. This translates into particular interest
rate rules or money growth rules that we have already covered. One way to interpret
these pegs is that the value of the exchange rate is seen as a useful indicator of the state
of the economy or of the shocks to demand for currency that the central bank aims to
track to have a more efficient policy rule. The adoption and abandonment of these pegs
follows the usefulness of this indicator as wedges arise between the domestic economy
and its foreign counterpart.

6 Steering inflation using income shortfalls

The final set of tools of the central bank are the composition of its assets At ( j), its spending
Et , and its dividend policy Dt . These policies drive the net income shortfall of the central
28 See
Frankel (2010) for the difficulty of controlling inflation when the central banks cannot commit to a
monetary policy strategy.

48
bank, St . This section studies them, while now keeping the interest on reserves and the
supply of banknotes passive. This is the most controversial approach to controlling infla-
tion, because its economic force is disputed in theory and has been hard to test. It often
goes under the name of the fiscal theory of the price level. It is important to understand
it even if only for central banks to know how to avoid it.

6.1 The economic force that drives inflation

The key equation is the no Ponzi scheme condition for the central bank in equation (14).
By forward iterating the resource constraint in equation (12), imposing the no-Ponzi con-
dition, and using the definition of net worth we get:

V0 M(s)
= A0 − S ( s ). (45)
P0 1 − M(s)

This provides one equation in one unknown, the price level.29 The economic force at
play is the following: a higher shortfall of the central bank leaves fewer real resources
available to back its debt, reducing its real value. As reserves are default-free, they have
a fixed value in nominal terms. Given their role as unit of account, the only way for their
real value to fall is for the price level to rise.
The price level adjusts because banks choose to hold more or fewer reserves in re-
sponse to them becoming a Ponzi scheme. When the central bank follows a path for
shortfalls that makes it insolvent, then banks run on the central bank’s liabilities. This
causes these liabilities—reserves—to lose their real value until they are back in line with
the central bank’s assets and net shortfall so the central bank is solvent. It is the control of
the real resources earned by the central bank that gives it control over inflation.30

6.2 Shortfall policy rules

Imagine that the central bank adjusts its balance sheet tools to follow a feedback rule for
its shortfall:
(1 − β)st+1 = φwt + xt+1 . (46)
29 The original analysis is Woodford (1994) and Sims (1994) and our approach is closer to that in Benigno
(2020) or Cochrane (2005).
30 For criticism of this mechanism, see Niepelt (2004) and Buiter (2017).

49
If the net worth falls, the central bank may pay less dividends or cut spending to lower
its shortfall, in which case φ > 0. The exogenous policy shock xt+1 is now an action by
the central bank that increases its shortfall.

6.2.1 Determinacy

Combining equation (46) with the law of motion for net worth in equation (15) to replace
out shortfalls gives a difference equation for net worth. In turn, the no-Ponzi scheme
condition for the central bank’s liabilities in equation (14), which is limT →∞ β T wT = 0 in
its linear form, provides a terminal condition. Together they give a solution for net worth
as long as φ < 1 − β.
Recalling that net worth is just the difference between assets and reserves delivers the
solution for the price level in the long run:

x (s) − r (s)
   
W̄ Ā
p(s) = + v(s) − a(s) (47)
1−β−φ V̄/ P̄ V̄/ P̄

The determinacy condition φ < 1 − β calls for the central bank to not lower its shortfall
by too much when its net worth falls. In fact, a reckless central bank that spends more as
its net worth falls φ < 0 would be ideal. Its irresponsible behavior forces banks to run on
the central bank, exchange their reserves for real goods and assets, and so raise the price
level to its new level. Following Leeper (1991), the literature has called feedback rules
that satisfy this condition non-Ricardian policies.
It seems paradoxical that inflation control results from the central bank actively pur-
suing insolvency to trigger runs by banks. Any private agent that tries to do this would
find that its liabilities become worthless, so it can get no real resources in exchange for the
new debt it issues. What makes the central bank special is that its liabilities are the unit
of account. It can honor these liabilities in nominal terms, by just issuing more reserves,
but also in real terms as long as the price level adjusts. Therefore, when it follows a rule
that would lead to a Ponzi scheme for an unchanged price level, the required fall in the
real value of reserves requires the price level to rise.
Inflation results from the condition that the central bank must stay solvent. It is a
capital gain to the central bank that comes at the expense of a capital loss of the private
sector holding the nominal reserves.31
31 If
reserves earn a convenience yield that is tied to their safety, then the shortfall policy rules of the
central bank can affect inflation now through a monetarist channel, like the one in the previous section, see

50
6.2.2 The role of dividends

As far as we know, no central bank has ever voluntarily pursued this strategy. Its effec-
tiveness is dubious. A non-Ricardian reckless behavior would get a central banker fired
before she got to implement it. Even if not, committing to a precise pattern for future
shortfall would be as hard for the central bank as it is for any government agency. In-
stead, the solvency approach is useful because it sheds light on how the central bank
should manage its surpluses st to stay away from a reckless shortfall policy.
The shortfalls in equation (13) are the sum of four parts. The first is seignorage from
banknotes. A policy rule for seignorage would affect inflation via the supply of currency,
so it could not be used to control it via solvency at the same time. The second compo-
nent are the expenses of the central banks. These are typically too small to be relevant.
The third component are the returns from holding assets net of the interest paid on re-
serves. By choosing the composition of these assets, and so the risk in their returns, the
central bank could, hypothetically, guide its net shortfall and so control inflation. How-
ever, since financial returns are volatile, most fluctuations in net shortfall are outside the
central bank’s control.
This leaves the fourth component of shortfalls as their main driver: the dividends to
the treasury dt . Ensuring that the central bank is solvent at all times, or that φ > 1 − β, is
referred to as the central bank having full fiscal support (Hall and Reis, 2015, Del Negro
and Sims, 2015, Benigno, 2020). In practice, this is achieved by requiring the central bank
to pay out all of its net income as dividends. In principle, this would imply that wt is
constant, the assumption we made in previous sections. In practice, because net income
is calculated differently across central banks, the net worth fluctuates, but the condition
we derived shows that what is important is that φ is sufficiently large so that the central
bank is not running a Ponzi scheme.

6.2.3 Experience

If the central bank has a “narrow” balance sheet, holding almost only short-term bonds,
then the term on asset gains is close to zero. Because seignorage from banknotes greatly
exceeds expenses, the dividend rule that guarantees solvency has traditionally always
implied positive dividends. These would sometimes be larger or smaller, but they have
raised little attention.
Bassetto and Cui (2018), Brunnermeier, Merkel and Sannikov (2020).

51
 After the great financial crisis, as central banks in advanced economies held large
amounts of long-term safe bonds together with risky bonds, this was no longer the case.
Most central banks in the world experienced negative dividends as a result of holding for-
eign assets whose value fluctuates with the foreign exchange rate (Reis, 2019a). Between
2021 and 2024 this led to large losses, and a series of approaches to ensure fiscal support
ranging from fiscal transfers from the Treasury (Bank of England), lower future dividends
in exchange for setting present negative dividends to zero (the Federal Reserve) or ex-ante
provisioning, by paying lower dividends in the past, that allows net worth to now fall to
its steady state (European central banks).

6.2.4 Monetizing the deficit

Historically, with some regularity, the Treasury imposes a dividend process. The central
bank is no longer independent to set a rule for its shortfall, whether it is one that ensures
fiscal backing or not.
Section 5.3 derived the consequences for inflation if the central bank uses seignorage
to satisfy the Treasury’s request. The printing of banknotes pins down the price level
through monetarist forces, often at high and volatile levels.
The solvency approach shows an alternative. The central bank could refuse to print
the banknotes. Insofar as the dividends are exogenous, the resulting shortfalls would be
exogenous as well. In that case, φ = 0, and the price level would be pinned down by
solvency forces. In effect, the central bank would be issuing reserves to borrow from the
private sector and send the resources to the Treasury. Every time it does so, the reserves
are worth less because of the inflation it generates.
Whether through seignorage or insolvency, this state of affairs could be referred to as
monetizing the fiscal deficit. It is the monetary base, currency and reserves, that is adjusting
to provide the necessary funding for the Treasury.

6.3 The solvency of other agents and inflation

Even if the solvency of the central bank is ensured, the solvency of other agents can affect
the level of inflation.

52
6.3.1 Fiscal dominance

The bulk of the literature on the fiscal theory of the price level focuses on the intertem-
poral budget constraint of the government linking the value of public debt to the present
value of its primary shortfalls, the deficits (Leeper and Leith, 2016). It makes three strong
assumptions.
The first is that the government does not default on its liabilities, so that government
bonds are perfect substitutes for reserves. Yet, unlike reserves, government bonds are not
the unit of account, and sovereign defaults are frequent (Reinhart and Rogoff, 2009).
Second, it assumes that the central bank dividends are passive and can take any value.
In that case, the intertemporal budget constraints of the central bank and the Treasury are
no longer two separate constraints, but rather a single consolidated one that states that
the sum of reserves and government liabilities (netting out the government bonds held
by the central bank) equals the present value of shortfalls of both the government and
the central bank (netting out the central bank’s dividends). This assumption effectively
denies central bank independence over its balance sheet.
The third assumption is that the Treasury solely chooses this shortfall, and any actions
of the central bank on its expenses, seignorage or composition of its assets are exactly
offset by the Treasury.
Combining these three assumptions, the same logic that allowed the central bank to
control inflation is now applied to the Treasury. That is, the solvency of the Treasury
becomes tied to the price level. This predicts that exogenous shocks to primary shortfalls
are linked to movements in inflation. There is some support for this in the data, especially
during the 1970s in the US, when high inflation coincided with large and persistent fiscal
deficits (Sims, 2011, Bianchi and Melosi, 2018, Cochrane, 2022a).

6.3.2 General equilibrium and nominal rigidities

Nominal rigidities do not alter the underlying logic of the solvency approach (Woodford,
1996, Sims, 2013, Cochrane, 2023). In terms of the basic New Keynesian model, the equi-
librium condition that is combined with the IS and Phillips curve in equations (9)–(10)
together with the net worth of the central bank in (15) give three equations with three
unknowns—the output gap, inflation, and net worth—as a function of nominal interest
rates and shortfalls. Instead of a feedback interest rate rule, as in section 3, or a money
growth rule, like in section 5, now there are two parts of the policy rule: an exogenous

53
process for the nominal interest rate set by the central bank, and a rule for the central
bank’s shortfalls as in equation (46). The first part fixes expected inflation, while the sec-
ond part determines its response to shocks.
As before, nominal rigidities add complementary channels working through output
and, especially, real interest rates. Now, after a loss, the jump in inflation to reestablish
the solvency of the central bank will lower the real interest rate given a nominal interest
rate peg. This stimulates more consumption, which further raises inflation via aggre-
gate demand. In turn, the transfer of wealth from the private sector towards the cen-
tral bank happens not just via inflation but also because of lower real interest rates over
time. Therefore, the dynamics of inflation change relative to when the classical dichotomy
holds, becoming more drawn out.
These dynamics feed back into the size of the inflation response because the real value
of the central bank’s net worth depends on the maturity of its assets. The value of long-
dated assets depends on both unexpected inflation and on the path of real interest rates.
Quantitative easing strategies that set the maturity of the central bank’s assets become a
determinant of the persistence of inflation deviations from target (Cochrane, 2001).

6.3.3 Alternatives to break the classical dichotomy

The focus on solvency brings to light different mechanisms that can break the classical di-
chotomy even with flexible prices. The transfer of wealth to the central bank comes at the
expense of banks holding these reserves (Reis, 2016). If banks’ net worth constrains their
willingness to extend private credit, then this provides a credit channel that complements
the ones we already discussed in section 4.
More generally, in an economy where different agents hold different mixes of nominal
and real assets, the inflation that is driven by the central bank’s solvency requirement will
induce redistribution of wealth. With incomplete markets, these will matter for aggregate
demand and so for the dynamics of inflation (Auclert, 2019, Kaplan, Nikolakoudis and
Violante, 2023).
Incomplete markets raise another channel for breaking the classical dichotomy through
the solvency channel. The total stock of reserves outstanding is a relevant variable for in-
flation because it determines by how much inflation must change to keep the real value
of these reserves in line with solvency. Under the fiscal dominance version, the size of
the government debt matters as well. In incomplete markets models, the net supply of
assets available to agents affects their equilibrium choices. The reason is that, unable to

54
diversify individual risks, agents engage in precautionary savings through these assets,
so their relative availability determines the cost and limits of doing so. When inflation re-
sults from the central bank’s choices regarding its shortfall, this both ex post redistributes
wealth across different agents and ex ante affects the expected returns on different sav-
ings vehicles. Both affect the desire to consume and produce, and so the real outcomes in
the economy (Hagedorn, 2018a,b).
In the other direction, breaking the classical dichotomy affects the source of shocks to
inflation. Changes in output and inflation directly affect the demand for currency and
the seignorage revenue of the government. Changes in the path of real interest rates over
time affect the capital gains and losses in the central bank’s portfolio. And, the effect of
monetary policy on real outcomes has an effect on tax collections and government spend-
ing and so on the primary shortfalls of the fiscal authority. Through fiscal dominance,
these would trigger changes in the demand for dividends from the central bank (Reis,
2019a).

7 Conclusion: a unified approach and the 2021-24 inflation

disaster
Each of the previous sections emphasized one policy tool that leans on one particular
economic force to bring inflation close to target. They all co-existed in the same dynamic
general equilibrium model, not as contradictory theories, but as different policy options.
The central bank can choose a strategy that relies on arbitrage forces, on money market
forces, or on insolvency forces, and within each it can choose one rule to implement it,
whether it is an interest rate rule, or a money growth rule, or a net shortfall rule, or one
of their many variants.
Importantly, if more than one strategy is active, most likely they will be in conflict with
each other; mathematically the economic system for prices becomes over-determined.32 .
We showed how to measure the expected effectiveness for each policy strategy. Our hope
is that the academic debate shifts from arguments on which assumptions are perceived
as being more convincing to attempts to measure the effectiveness of each strategy. We
32 A
different use of the word active and passive is to describe which of the two institutions, the central
bank or the Treasury, is imposing its decisions on the other. If they are playing a game with each other, this
will affect how the policy approach is chosen and set. Unfortunately, both definitions of active/passive are
used in the literature, generating confusion.

55
also provided our own take of the history of monetary policy and inflation in post-war
advanced economies.
During Bretton Woods, most countries followed a monetarist strategy by pegging to
the US dollar. During the 1970s, the US or the UK followed an interest rate strategy
without satisfying the conditions for determinacy, both in the responsiveness of the feed-
back rules and in having a clear pillar as an escape clause. This led in the early 1980s to
both countries having brief experiences with money growth rules. At the same time, in
countries through Latin America, fiscal authorities imposed a combination of monetarist
seignorage rules or net shortfall rules that flirted with insolvency, leading to high and
volatile inflation.
The conquest of stable inflation between 1990 and 2010 came from a coherent strat-
egy across the different elements. Central banks used feedback interest rate rules that
satisfied the determinacy principle. They adopted inflation targeting regimes to manage
expectations, while having escape clauses reliant on monetary anchors. Central bank in-
dependence imposed rules on the dividends paid to Treasuries that kept the central bank
solvent at all times. These ruled out seignorage or insolvency from driving inflation.
Between 2010 and 2020 this framework was refined through forward guidance, quan-
titative easing, going long, and communication strategies to overcome the effective lower
bound, even if these tools relied heavily on rationality of expectations and were generally
less effective.
Between 2021 and 2024, unusual shocks that hit the economy justified an optimal in-
flation rate well above 2%, but the actual inflation rate was much higher than that. Why
this happened is understandably still the subject of debate. Perhaps central banks kept
to this same framework but they made mistakes in judging the state of the economy, in
anchoring expectations, and in using unconventional interest rate tools that are less effec-
tive (Reis, 2023, Eggertsson and Kohn, 2023). Or, perhaps the expansion of the balance
sheet of central banks through quantitative easing and the large increase in public debt
during the pandemic have made central bank independence untenable and it is concerns
about solvency that are driving inflation (Bianchi and Melosi, 2022, Cochrane, 2022b). The
future will show.

References
Acharya, Sushant, and Jess Benhabib. 2024. “Global Indeterminacy in HANK

56
 Economies.” Bank of Canada manuscript.

Acharya, Sushant, and Keshav Dogra. 2020. “Understanding HANK: Insights from a
PRANK.” Econometrica, 88: 1113–1158.

Adão, Bernardino, Isabel Correia, and Pedro Teles. 2011. “Unique Monetary Equilibria
with Interest Rate Rules.” Review of Economic Dynamics, 14(3): 432–442.

Adão, Bernardino, Isabel Correia, and Pedro Teles. 2014. “Short and Long Interest Rate
Targets.” Journal of Monetary Economics, 66: 95–107.

Agarwal, Ruchir, and Miles Kimball. 2019. “Enabling Deep Negative Rates to Fight Re-
cessions: A Guide.” IMF Working Paper 19/84.

Andrade, Philippe, and Filippo Ferroni. 2021. “Delphic and Odyssean Monetary Policy
Shocks: Evidence from the Euro Area.” Journal of Monetary Economics, 117: 816–832.

Angeletos, George-Marios, and Chen Lian. 2018. “Forward Guidance without Common
Knowledge.” American Economic Review, 108(9): 2477–2512.

Angeletos, George-Marios, and Chen Lian. 2023. “Determinacy without the Taylor Prin-
ciple.” Journal of Political Economy, 131(8): 2125 – 2164.

Aruoba, Boragan, Pablo Cuba-Borda, and Frank Schorfheide. 2017. “Macroeconomic

Dynamics Near the ZLB: A Tale of Two Countries.” Review of Economic Studies, 85(1): 87–
118.

Ascari, Guido, and Tiziano Ropele. 2009. “Trend Inflation, Taylor Principle, and Indeter-
minacy.” Journal of Money, Credit and Banking, 41(8): 1557–1584.

Atkeson, Andrew, Varadarajan V Chari, and Patrick J Kehoe. 2010. “Sophisticated Mon-
etary Policies.” Quarterly Journal of Economics, 125(1): 47–89.

Auclert, Adrien. 2019. “Monetary Policy and the Redistribution Channel.” American Eco-
nomic Review, 109(6): 2333–2367.

Barthélemy, Jean, and Magali Marx. 2019. “Monetary Policy Switching and Indetermi-
nacy.” Quantitative Economics, 10(1): 353–385.

Barth III, Marvin J, and Valerie A Ramey. 2001. “The Cost Channel of Monetary Trans-
mission.” NBER Macroeconomics Annual, 16: 199–240.

57
Bassetto, Marco. 2005. “Equilibrium and Government Commitment.” Journal of Economic
Theory, 124(1): 79–105.

Bassetto, Marco, and Wei Cui. 2018. “The Fiscal Theory of the Price Level in a World of
Low Interest Rates.” Journal of Economic Dynamics and Control, 89: 5–22.

Benassy, Jean-Pascal. 2000. “Price Level Determinacy under a Pure Interest Rate Peg.”
Review of Economic Dynamics, 3(1): 194–211.

Benhabib, Jess, Stephanie Schmitt-Grohe, and Martin Uribe. 2001. “The Perils of Taylor
Rules.” Journal of Economic Theory, 96(1-2): 40–69.

Benigno, Pierpaolo. 2020. “A Central Bank Theory of Price Level Determination.” Amer-
ican Economic Journal: Macroeconomics, 12(3): 258–283.

Berger, David, Konstantin Milbradt, Fabrice Tourre, and Joseph Vavra. 2021. “Mort-
gage Prepayment and Path-Dependent Effects of Monetary Policy.” American Economic
Review, 111(9): 2829–2878.

Bernanke, Ben S. 2020. “The New Tools of Monetary Policy.” American Economic Review,
110(4): 943–983.

Bernanke, Ben S, and Frederic S Mishkin. 1997. “Inflation Targeting: A New Framework
for Monetary Policy?” Journal of Economic Perspectives, 11(2): 97–116.

Bernanke, Ben S, Mark Gertler, and Simon Gilchrist. 1999. “Issues in the Design of
Monetary Policy Rules.” In Handbook of Macroeconomics. Vol. 1, , ed. J. B. Taylor and M.
Woodford, 1341–1393. Elsevier.

Bianchi, Francesco, and Leonardo Melosi. 2018. “Constrained Discretion and Central
Bank Transparency.” Review of Economics and Statistics, 100(1): 187–202.

Bianchi, Francesco, and Leonardo Melosi. 2022. “Inflation as a Fiscal Limit.” In Reassess-
ing Constraints on the Economy and Policy. Federal Reserve Bank of Kansas City Jackson
Hole Symposium.

Bilbiie, Florin. 2024. “Monetary Policy and Heterogeneity: An Analytical Framework.”

Review of Economic Studies, forthcoming.

58
Blinder, Alan S., and Ricardo Reis. 2005. “Understanding the Greenspan standard.” In
The Greenspan Era: Lessons for the Future. 11–96. Jackson Hole Symposium:Federal Re-
serve Bank of Kansas City.

Blinder, Alan S, Michael Ehrmann, Marcel Fratzscher, Jakob De Haan, and David-Jan
Jansen. 2008. “Central Bank Communication and Monetary Policy: A Survey of Theory
and Evidence.” Journal of Economic Literature, 46(4): 910–945.

Boneva, Lena Mareen, R. Anton Braun, and Yuichiro Waki. 2016. “Some Unpleasant
Properties of Loglinearized Solutions when the Nominal Rate Is Zero.” Journal of Mon-
etary Economics, 84: 216–232.

Bordo, Michael. 2005. The Gold Standard and Related Regimes: Collected Essays. Cambridge
University Press.

Bordo, Michael D. 2017. “The Operation and Demise of the Bretton Woods System: 1958
to 1971.” In The Bretton Woods Agreements. , ed. Naomi Lamoreaux and Ian Shapiro,
217–235. Yale University Press.

Brunnermeier, Markus K, and Yuliy Sannikov. 2016. “The I Theory of Money.” NBER
Working Paper 22533.

Brunnermeier, Markus, Sebastien Merkel, and Yuliy Sannikov. 2020. “The Fiscal Theory
of the Price Level with a Bubble.” NBER working paper 27116.

Buiter, Willem H. 2017. “The Good and the Bad Fiscal Theory of the Price Level.” CEPR
Discussion Paper 11975.

Bullard, James, and Kaushik Mitra. 2002. “Learning about Monetary Policy Rules.” Jour-
nal of Monetary Economics, 49(6): 1105–1129.

Cagan, Phillip. 1956. “The Monetary Dynamics of Hyperinflation.” In Studies in the Quan-
tity Theory if Money. , ed. Milton Friedman, 25–117. University of Chicago Press.

Calvo, Guillermo A. 1983. “Staggered Prices in a Utility-Maximizing Framework.” Jour-

nal of Monetary Economics, 12(3): 383–398.

Campbell, Jeffrey R, Jonas DM Fisher, Alejandro Justiniano, and Leonardo Melosi.

2017. “Forward Guidance and Macroeconomic Outcomes since the Financial Crisis.”
NBER Macroeconomics Annual, 31(1): 283–357.

59
Canzoneri, Matthew B., Dale W. Henderson, and Kenneth S. Rogoff. 1983. “The Infor-
mation Content of the Interest Rate and Optimal Monetary Policy.” Quarterly Journal of
Economics, 98(4): 545–566.

Capie, Forrest. 2010. The Bank of England: 1950s to 1979. Cambridge University Press.

Carlstrom, Charles T, and Timothy S Fuerst. 2002. “Taylor Rules in a Model that Satisfies
the Natural-Rate Hypothesis.” American Economic Review, 92(2): 79–84.

Carlstrom, Charles T, and Timothy S Fuerst. 2003. “Money Growth Rules and Price Level
Determinacy.” Review of Economic Dynamics, 6(2): 263 – 275.

Carlstrom, Charles T., Timothy Fuerst, and Matthias Paustian. 2015. “Inflation and Out-
put in New Keynesian Models with a Transient Interest Rate Peg.” Journal of Monetary
Economics, 76: 230–243.

Carvalho, Carlos, Fernanda Nechio, and Tiago Tristao. 2021. “Taylor Rule Estimation by
OLS.” Journal of Monetary Economics, 124: 140–154.

Cecchetti, Stephen G, Peter Hooper, Bruce C Kasman, Kermit L Schoenholtz, and

Mark W Watson. 2007. “Understanding the Evolving Inflation Process.” US Monetary
Policy Forum.

Christiano, Lawrence J, and Martin Eichenbaum. 1992. “Liquidity Effects and the Mon-
etary Transmission Mechanism.” The American Economic Review, 82(2): 346–353.

Christiano, Lawrence J, and Martin Eichenbaum. 1995. “Liquidity Effects, Monetary Pol-
icy, and the Business Cycle.” Journal of Money, Credit and Banking, 27(4): 1113–1136.

Christiano, Lawrence J., and Massimo Rostagno. 2001. “Money Growth Monitoring and
the Taylor Rule.” NBER Working Paper 8539.

Christiano, Lawrence J., and Yuta Takahashi. 2018. “Discouraging Deviant Behavior in
Monetary Economics.” NBER Working Paper 24949.

Christiano, Lawrence J., Martin S Eichenbaum, and Benjamin K Johannsen. 2018.

“Does the New Keynesian Model Have a Uniqueness Problem?” NBER Working Paper
24612.

60
Christiano, Lawrence J, Mathias Trabandt, and Karl Walentin. 2010. “DSGE Models for
Monetary Policy Analysis.” In Handbook of Monetary Economics. Vol. 3, 285–367. Elsevier.

Ciccarelli, Matteo, Angela Maddaloni, and José-Luis Peydró. 2015. “Trusting the
Bankers: A new Look at the Credit Channel of Monetary Policy.” Review of Economic
Dynamics, 18(4): 979–1002.

Cieslak, Anna, and Andreas Schrimpf. 2019. “Non-Monetary News in Central Bank
Communication.” Journal of International Economics, 118: 293–315.

Clarida, Richard H, and Mark Gertler. 1997. “How the Bundesbank Conducts Mone-
tary Policy.” In Reducing Inflation: Motivation and Strategy. , ed. Christina D Romer and
David H Romer, 363–412. University of Chicago Press.

Clarida, Richard, Jordi Gali, and Mark Gertler. 2000. “Monetary Policy Rules and
Macroeconomic Stability: Evidence and Some Theory.” Quarterly Journal of Economics,
115(1): 147–180.

Cochrane, John H. 2001. “Long-term Debt and Optimal Policy in the Fiscal Theory of the
Price Level.” Econometrica, 69(1): 69–116.

Cochrane, John H. 2005. “Money as Stock.” Journal of Monetary Economics, 52(3): 501–528.

Cochrane, John H. 2011. “Determinacy and Identification with Taylor Rules.” Journal of
Political Economy, 119(3): 565–615.

Cochrane, John H. 2017. “The New-Keynesian Liquidity Trap.” Journal of Monetary Eco-
nomics, 92: 47–63.

Cochrane, John H. 2022a. “Fiscal Histories.” Journal of Economic Perspectives, 36(4): 125–
146.

Cochrane, John H. 2022b. “Inflation Past, Present and Future: Fiscal Shocks, Fed Re-
sponse, and Fiscal Limits.” In How Monetary Policy Got Behind the Curve–and How To
Get Back.. , ed. Michael D. Bordo, John H. Cochrane and Joyn Taylor, 63–114. Hoover
Institution Press.

Cochrane, John H. 2023. The Fiscal Theory of the Price Level. Princeton University Press.

61
Coibion, Olivier, and Yuriy Gorodnichenko. 2011. “Monetary Policy, Trend Inflation,
and the Great Moderation: An Alternative Interpretation.” American Economic Review,
101(1): 341–370.

Del Negro, Marco, and Christopher A. Sims. 2015. “When Does a Central Bank’s Balance
Sheet Require Fiscal Support?” Journal of Monetary Economics, 73: 1–19.

Del Negro, Marco, Marc P Giannoni, and Christina Patterson. 2023. “The Forward
Guidance Puzzle.” Journal of Political Economy Macroeconomics, 1(1): 43–79.

Den Haan, Wouter J, Pontus Rendahl, and Markus Riegler. 2017. “Unemploy-
ment (Fears) and Deflationary Spirals.” Journal of the European Economic Association,
16(5): 1281–1349.

Diba, Behzad, and Olivier Loisel. 2021. “Pegging the Interest Rate on Bank Reserves: A
Resolution of New Keynesian Puzzles and Paradoxes.” Journal of Monetary Economics,
118: 230–244.

Dincer, Nergiz, Barry Eichengreen, Petra Geraats, et al. 2022. “Trends in Monetary Policy
Transparency: Further Updates.” International Journal of Central Banking, 18(1): 331–348.

Eggertson, Gauti B., and Vaishali Garga. 2019. “Sticky Prices versus Sticky Information:
Does it Matter for Policy Paradoxes?” Review of Economic Dynamics, 31: 363–392.

Eggertsson, Gauti B, and Donald Kohn. 2023. “The Inflation Surge of the 2020s: The Role
of Monetary Policy.” Hutchins Center Working Paper 87.

Eggertsson, Gauti B., and Michael Woodford. 2003. “The Zero Bound on Interest Rates
and Optimal Monetary Policy.” Brookings Papers on Economic Activity, 34(1): 139–235.

Eusepi, Stefano, and Bruce Preston. 2018. “The Science of Monetary Policy: An Imperfect
Knowledge Perspective.” Journal of Economic Literature, 56(1): 3–59.

Evans, George W., and Garey Ramey. 1992. “Expectation Calculation and Macroeco-
nomic Dynamics.” American Economic Review, 82(1): 207–224.

Evans, George W., and Seppo Honkapohja. 2001. Learning and Expectations in Macroeco-
nomics. Princeton University Press.

62
Fabo, Brian, Martina Jancokova, Elisabeth Kempf, and Lubos Pastor. 2021. “Fifty
Shades of QE: Comparing Findings of Central Bankers and Academics.” Journal of Mon-
etary Economics, 120: 1–20.

Farhi, Emmanuel, and Iván Werning. 2019. “Monetary Policy, Bounded Rationality, and
Incomplete Markets.” American Economic Review, 109(11): 3887–3928.

Fernández-Villaverde, Jesús, Grey Gordon, Pablo Guerrón-Quintana, and Juan F.

Rubio-Ramı́rez. 2015. “Nonlinear Adventures at the Zero Lower Bound.” Journal of
Economic Dynamics and Control, 57: 182 – 204.

Fiore, Fiorella De, and Oreste Tristani. 2013. “Optimal Monetary Policy in a Model of
the Credit Channel.” The Economic Journal, 123(571): 906–931.

Frankel, Jeffrey. 2010. “Monetary Policy in Emerging Markets: A Survey.” In Handbook of

Monetary Economics. Vol. 3, , ed. Benjamin M. Friedman and Michael Woodford, 1439–
1520. Elsevier.

Gabaix, Xavier. 2020. “A Behavioral New Keynesian Model.” American Economic Review,
110(8): 2271–2327.

Gali, Jordi. 2008. Monetary Policy, Inflation, and the Business Cycle. MIT Press.

Garcı́a-Schmidt, Mariana, and Michael Woodford. 2019. “Are Low Interest Rates De-
flationary? A Paradox of Perfect-Foresight Analysis.” American Economic Review,
109(1): 86–120.

Gertler, Mark, and Peter Karadi. 2015. “Monetary Policy Surprises, Credit Costs, and
Economic Activity.” American Economic Journal: Macroeconomics, 7(1): 44–76.

Goodfriend, Marvin. 2016. “The Case for Unencumbering Interest Rate Policy at the Zero
Bound.” In Designing Resilient Monetary Policy Frameworks for the Future. 127–160. Jack-
son Hole Symposium:Federal Reserve Bank of Kansas City.

Greenwald, Daniel. 2018. “The Mortgage Credit Channel of Macroeconomic Transmis-

sion.” MIT manuscript.

Gürkaynak, Refet, Brian Sack, and Eric Swanson. 2005. “Do Actions Speak Louder Than
Words? The Response of Asset Prices to Monetary Policy Actions and Statements.”
International Journal of Central Banking, 1(1).

63
Hagedorn, Marcus. 2018a. “A Demand Theory of the Price Level.” CEPR Discussion Pa-
per 11364.

Hagedorn, Marcus. 2018b. “Prices and Inflation when Government Bonds are Net
Wealth.” CEPR Discussion Paper 12769.

Hall, Robert E. 1997. “Irving Fisher’s Self-Stabilizing Money.” American Economic Review
Papers and Proceedings, 87(2): 436–438.

Hall, Robert E., and Ricardo Reis. 2015. “Maintaining Central-Bank Solvency under
New-Style Central Banking.” NBER Working Paper 21173.

Hall, Robert E., and Ricardo Reis. 2016. “Achieving Price Stability by Manipulating the
Central Bank’s Payment on Reserves.” NBER Working Paper 22761.

Hartmann, Philipp, and Frank Smets. 2018. “The European Central Bank’s Monetary
Policy during its First 20 Years.” Brookings Papers on Economic Activity, 2018(2): 1–146.

Holden, Tom D. 2024. “Robust Real Rate Rules.” Econometrica, forthcoming.

Hume, David. 1752. Political Discourses. A. Kincaid & A. Donaldson.

Ihrig, Jane E, Ellen E Meade, and Gretchen C Weinbach. 2015. “Rewriting Monetary
Policy 101: What’s the Fed’s Preferred Post-crisis Approach to Raising Interest Rates?”
Journal of Economic Perspectives, 29(4): 177–198.

Ilzetzki, Ethan, Carmen M. Reinhart, and Kenneth S. Rogoff. 2019. “Exchange Arrange-
ments Entering the Twenty-First Century: Which Anchor Will Hold?” Quarterly Journal
of Economics, 134(2): 599–646.

IMF. 2020. “The Central Bank Transparency Code.” IMF Policy Paper 20/38.

Kaplan, Greg, Georgios Nikolakoudis, and Giovanni L Violante. 2023. “Price Level and
Inflation Dynamics in Heterogeneous Agent Economies.” NBER Working Paper 31433.

Kehoe, Timothy J, and Juan Pablo Nicolini. 2022. A Monetary and Fiscal History of Latin
America, 1960–2017. University of Minnesota Press.

Khan, Aubhik, Robert G. King, and Alexander L. Wolman. 2003. “Optimal Monetary
Policy.” Review of Economic Studies, 70(4): 825–860.

64
Khan, Hashmat, Louis Phaneuf, and Jean Gardy Victor. 2020. “Rules-Based Monetary
Policy and the Threat of Indeterminacy When Trend Inflation Is Low.” Journal of Mone-
tary Economics, 114: 317–333.

Kiley, Michael. 2016. “Policy Paradoxes in the New-Keynesian Model.” Review of Eco-
nomic Dynamics, 21: 1–15.

Krishnamurthy, Arvind, and Annette Vissing-Jorgensen. 2013. “The Ins and Outs of
LSAPs.” 57–111. Jackson Hole Symposium:Federal Reserve Bank of Kansas City.

Lagos, Ricardo, and Randall Wright. 2005. “A Unified Framework for Monetary Theory
and Policy Analysis.” Journal of Political Economy, 113(3): 463–484.

Leeper, Eric M. 1991. “Equilibria under Active and Passive Monetary and Fiscal Policies.”
Journal of Monetary Economics, 27: 129–147.

Leeper, Eric M, and Campbell Leith. 2016. “Understanding Inflation as a Joint Monetary–
Fiscal Phenomenon.” In Handbook of Macroeconomics. Vol. 2, , ed. John B. Taylor and
Harald Uhlig, 2305–2415. Elsevier.

Leeson, Robert, Evan F Koenig, and George A Kahn. 2013. The Taylor Rule and the Trans-
formation of Monetary policy. Hoover Press.

Lewis, Vivien, and Markus Roth. 2018. “Interest Rate Rules under Financial Domi-
nance.” Journal of Economic Dynamics and Control, 95(C): 70–88.

Loisel, Olivier. 2021. “The Implementability and Implementation of Feasible Paths by

Stabilization Policy.” Theoretical Economics, 16(2): 677–716.

Lubik, Thomas A., and Frank Schorfheide. 2004. “Testing for Indeterminacy: An Appli-
cation to U.S. Monetary Policy.” American Economic Review, 94(1): 190–217.

Lubik, Thomas A., Christian Matthes, and Elmar Mertens. 2019. “Indeterminacy and
Imperfect Information.” FRB Richmond working paper 19-17.

Lucas, Robert, and Nancy Stokey. 1987. “Money and Interest in a Cash-in-Advance Econ-
omy.” Econometrica, 55(3): 491–513.

Lunsford, Kurt G. 2020. “Policy Language and Information Effects in the Early Days of
Federal Reserve Forward Guidance.” American Economic Review, 110(9): 2899–2934.

65
Magill, Michael, and Martine Quinzii. 2014. “Term Structure and Forward Guidance as
Instruments of Monetary Policy.” Economic Theory, 56(1): 1–32.

McCallum, Bennett T. 1981. “Price Level Determinacy with an Interest Rate Policy Rule
and Rational Expectations.” Journal of Monetary Economics, 8(3): 319–329.

McCallum, Bennett T. 1999. “Issues in the Design of Monetary Policy Rules.” In Handbook
of Macroeconomics. Vol. 1, , ed. J. B. Taylor and M. Woodford, Chapter 23, 1483–1530.
Elsevier.

McCallum, Bennett T. 2003. “Multiple-Solution Indeterminacies in Monetary Policy

Analysis.” Journal of Monetary Economics, 50(5): 1153–1175.

McGough, Bruce, Glenn D. Rudebusch, and John C. Williams. 2005. “Using a Long-
Term Interest Rate as the Monetary Policy Instrument.” Journal of Monetary Economics,
52(5): 855 – 879.

McKay, Alisdair, Emi Nakamura, and Jón Steinsson. 2016. “The Power of Forward Guid-
ance Revisited.” American Economic Review, 106(10): 3133–58.

Meltzer, Allan H. 2010. A History of the Federal Reserve, Volume 2: 1970–1986. University of
Chicago Press.

Mertens, Karel R. S. M., and Morten O. Ravn. 2014. “Fiscal Policy in an Expectations-
Driven Liquidity Trap.” Review of Economic Studies, 81(4): 1637–1667.

Meyer-Gohde, Alexander, and Mary Tzaawa-Krenzler. 2023. “Sticky Information and

the Taylor Principle.” IMF working paper 189.

Muth, John. 1960. “Optimal Properties of Exponentially Weighted Forecasts.” Journal of

the American Statistical Association, 55: 290–306.

Nakajima, Tomoyuki, and Herakles Polemarchakis. 2005. “Money and Prices under Un-
certainty.” Review of Economic Studies, 72(1): 223–246.

Nakamura, Emi, and Jón Steinsson. 2018. “High-Frequency Identification of Monetary

Non-Neutrality: The Information Effect.” Quarterly Journal of Economics, 133(3): 1283–
1330.

66
Niepelt, Dirk. 2004. “The Fiscal Myth of the Price Level.” Quarterly Journal of Economics,
119(1): 277–300.

Obstfeld, Maurice, and Kenneth Rogoff. 1983. “Speculative Hyperinflations in Maxi-

mizing Models: Can we Rule them Out?” Journal of Political Economy, 91(4): 675–687.

Obstfeld, Maurice, and Kenneth Rogoff. 1995. “The Mirage of Fixed Exchange Rates.”
Journal of Economic Perspectives, 9(4): 73–96.

Orphanides, Athanasios. 2003. “The Quest for Prosperity Without Inflation.” Journal of
Monetary Economics, 50(3): 633–663.

Orphanides, Athanasios. 2019. “Monetary Policy Strategy and its Communication.” 22–
24. Jackson Hole Symposium:Federal Reserve Bank of Kansas City.

Piazzesi, Monika, Ciaran Rogers, and Martin Schneider. 2022. “Money and Banking in
a New Keynesian Model.” Available at: https://web.stanford.edu/~piazzesi/MB\
inNK.pdf.

Poole, William. 1970. “Optimal Choice of Monetary Policy Instruments in a Simple

Stochastic Macro Model.” Quarterly Journal of Economics, 84(2): 197–216.

Reinhart, Carmen M., and Kenneth S. Rogoff. 2009. This Time is Different: Eight Centuries
of Financial Folly. Princeton University Press.

Reis, Ricardo. 2009. “A Sticky-Information General-Equilibrium Model for Policy Analy-

sis.” In Monetary Policy under Uncertainty and Learning. , ed. Klaus Schmidt-Heubel, Carl
Walsh and Norman Loyaza, 227–283. Banco Central de Chile.

Reis, Ricardo. 2013. “Central Bank Design.” Journal of Economic Perspectives, 27(4): 17–44.

Reis, Ricardo. 2016. “Funding Quantitative Easing to Target Inflation.” In Designing Re-
silient Monetary Policy Frameworks for the Future. Jackson Hole Symposium:Federal Re-
serve Bank of Kansas City.

Reis, Ricardo. 2019a. “Can the Central Bank Alleviate Fiscal Burdens?” In The Economics
of Central Banking. Oxford University Press Handbooks in Economics, , ed. David Mayes,
Peter Siklos and Jan-Egbert Strum. Oxford University Press.

67
Reis, Ricardo. 2019b. “Central Banks Going Long.” In Monetary Policy and Financial Sta-
bility: Transmission Mechanisms and Policy Implications. , ed. Markus Brunnermeier Al-
varo Aguirre and Diego Saravia, 43–81. Central Bank of Chile.

Reis, Ricardo. 2021. “Losing the Inflation Anchor.” Brookings Papers on Economic Activity,
307–361.

Reis, Ricardo. 2022. “Expected Inflation in the Euro Area: Measurement and Policy Re-
sponses.” In ECB Forum on Central Banking. ECB.

Reis, Ricardo. 2023. “The Burst of High Inflation in 2021-22: How and Why Did We Get
Here?” In How Monetary Policy Got Behind the Curve–And How to Get it Back. , ed. Michael
Bordo, John Cochrane and John Taylor, 203–252. Hoover Institution Press.

Rogoff, Kenneth. 2017. “Dealing with Monetary Paralysis at the Zero Bound.” Journal of
Economic Perspectives, 31(3): 47–66.

Romer, Christina D., and David Romer. 2002. “The Evolution of Economic Understand-
ing and Postwar Stabilization Policy.” In Economic Policy Symposium. 11–78. Federal Re-
serve Bank of Kansas City.

Rostagno, Massimo, Carlo Altavilla, Giacomo Carboni, Wolfgang Lemke, Roberto

Motto, Arthur Saint Guilhem, and Jonathan Yiangou. 2021. Monetary Policy in Times
of Crisis: A Tale of Two Decades of the European Central Bank. Oxford University Press.

Rotemberg, Julio. 1982. “Monopolistic Price Adjustment and Aggregate Output.” Review
of Economic Studies, 44(4): 517–531.

Sargent, Thomas J. 1982. “The Ends of Four Big Inflations.” In Inflation: Causes and Effects.
41–98. National Bureau of Economic Research.

Sargent, Thomas J, and Neil Wallace. 1975. “”Rational” Expectations, the Optimal Mon-
etary Instrument, and the Optimal Money Supply Rule.” Journal of Political Economy,
83(2): 241–254.

Sargent, Thomas J, and Neil Wallace. 1984. “Some Unpleasant Monetarist Arithmetic.”
In Monetarism in the United Kingdom. 15–41. Springer.

Sims, Christopher A. 1994. “A Simple Model for Study of the Determination of the Price
Level and the Interaction of Monetary and Fiscal Policy.” Economic Theory, 4(3): 381–399.

68
Sims, Christopher A. 2011. “Stepping on a rake: the role of fiscal policy in the inflation of
the 1970s.” European Economic Review, 55: 48–56.

Sims, Christopher A. 2013. “Paper Money.” American Economic Review Papers and Proceed-
ings, 103(3): 563–584.

Sveen, Tommy, and Lutz Weinke. 2005. “New Perspectives on Capital, Sticky Prices, and
the Taylor Principle.” Journal of Economic Theory, 123(1): 21–39.

Taylor, John B. 1993. “Discretion versus Policy Rules in Practice.” Carnegie-Rochester Con-
ference Series on Public Policy, 39: 195–214.

Taylor, John B. 1996. “Policy Rules as a Means to a More Effective Monetary Policy.”
Monetary and Economic Studies, 14(1): 28–39.

Taylor, John B., ed. 1999. Monetary Policy Rules. University of Chicago Press.

Werning, Ivan. 2011. “Managing a Liquidity Trap: Monetary and Fiscal Policy.” NBER
Working Paper 17344.

Woodford, Michael. 1994. “Monetary Policy and Price Level Determinacy in a Cash-in-
Advance Economy.” Economic Theory, 4(3): 345–380.

Woodford, Michael. 1996. “Control of the Public Debt: A Requirement for Price Stabil-
ity?” NBER Working Paper 5684.

Woodford, Michael. 2003. Interest and Prices: Foundations of a Theory of Monetary Policy.
Princeton University Press.

Woodford, Michael. 2010. “Optimal Monetary Stabilization Policy.” In Handbook of Mon-

etary Economics. Vol. 3, , ed. Benjamin M. Friedman and Michael Woodford, Chapter 14,
723 – 828. Elsevier.

Woodford, Michael. 2013. “Macroeconomic Analysis Without the Rational Expectations

Hypothesis.” Annual Review of Economics, 5: 303–346.

69