Monetary Policy and Heterogeneity:

An Analytical FrameworkI
Florin O. BilbiieII

August 2021; First draft: December 2017III

Abstract

THANK is a tractable heterogeneous-agent New-Keynesian model that captures analytically key

micro-heterogeneity channels of quantitative-HANK: cyclical inequality and risk as separate but re-
lated channels; idiosyncratic uncertainty and self-insurance, precautionary saving; and realistic in-
tertemporal marginal propensities to consume. I use it for a full-fledged New Keynesian macro
analysis: determinacy with interest-rate rules, solving the forward-guidance puzzle, amplification
and multipliers, and optimal monetary policy. Amplification requires countercyclical while solving
the puzzle requires procyclical inequality: a Catch-22, resolved in theory if the separate "pure" risk
channel is procyclical enough. Price-level-targeting ensures determinacy and is puzzle-free, even
when both inequality and risk are countercyclical, thus resolving the Catch-22. The same holds for
a rule fixing nominal public debt in the model version with liquidity. Optimal policy with hetero-
geneity features a novel inequality-stabilization motive generating higher inflation volatility—but it
is unaffected by risk, insofar as the target equilibrium entails no inequality.
JEL Codes: E21, E31, E40, E44, E50, E52, E58, E60, E62
Keywords: heterogeneity; inequality; liquidity; tractable HANK; optimal monetary policy; for-
ward guidance puzzle; Taylor and Wicksellian rules; determinacy; multipliers.

II am grateful for comments to the Editor, Kurt Mitman and three anonymous referees, as well as in particular to Adrien
Auclert, Edouard Challe, Keshav Dogra, Gauti Eggertsson, and Marcus Hagedorn, and to Sushant Acharya, Christian Bayer,
Davide Debortoli, Marco Del Negro, Axelle Ferrière, Jordi Galí, Marc Giannoni, Keith Kuester, Eric Leeper, Olivier Loisel, Karel
Mertens, Christina Patterson, Xavier Ragot, Vincent Sterk, Ludwig Straub, Paolo Surico, Andrea Tambalotti, Mirko Wiederholt,
and Michael Woodford, and a great many other colleagues and seminar and conference participants. I gratefully acknowledge
without implicating the support of Banque de France Chair at PSE, and of Institut Universitaire de France, as well as the
hospitality of New York University and CREI.
II University of Lausanne and CEPR; [email protected]. http://florin.bilbiie.googlepages.com.
III This paper supersedes the December 2017 "A Catch-22 for HANK Models: No Puzzles, No Amplification" (then "Hetero-

geneity and Determinacy: Amplification without Puzzles").

1 Introduction

A spectre is haunting Macroeconomics—the spectre of Heterogeneity. Some of the world’s leading

policymakers have been asking for research on it, and its other name, “Inequality”, in connection
with stabilization, monetary and fiscal policies. For until recently, research on these two topics has
been, with few exceptions, largely disconnected. Yet a burgeoning field emerged as a true synthesis
of these two lodes: heterogeneous-agent (HA) and New Keynesian (NK), leading to HANK.1
The vast majority of contributions consists of quantitative models, involving heavy machinery
for their resolution, the price to pay to achieve the realism conferred by matching the micro data.2
Yet given that much of the post-crisis bad press of existing DSGE models refers to their being too
complex and somewhat black-box, it seems important to build simple tractable representations of
these models to gain analytical insights into their underlying mechanisms and make their policy
conclusions sharper and easier to communicate. The two, quantitative and analytical approaches
are thus strongly complementary and reinforce each other.
With this paper, I wish to propose a tractable HANK model, THANK, to achieve two purposes.3
First, argue that it is a good representation, along several key dimensions, for rich-heterogeneity
quantitative HANK models. And second, use it for a full-fledged positive and normative NK analy-
sis in closed form: determinacy of interest rate rules, curing the forward guidance puzzle, ampli-
fication and fiscal multipliers; and optimal monetary policy. The ethos is thus to maximize micro
heterogeneity into a macro model under the constraint of tractability.
THANK is a three-equation model isomorphic to the textbook representative-agent (RANK)
model, which it nests; yet it captures several dimensions that the recent quantitative literature
finds important for the study of macro fluctuations with heterogeneity. First, it features a key
aggregate-demand (AD) amplification: the "New Keynesian cross" present in any HANK model
where some households are constrained hand-to-mouth. Heterogeneity shapes aggregate outcomes
through cyclical inequality: how the distribution of income between constrained and unconstrained
changes over the cycle, e.g. who suffers more in recessions. This originates in the TANK model in
Bilbiie (2008, 2020) and generalizes to the subsequent rich-heterogeneity literature (Auclert’s (2019)).
Recently, Patterson (2019) provides empirical evidence for countercyclical inequality.
Second, my analytical model incorporates uninsurable risk, precautionary, self-insurance saving,
and the distinction between liquid and illiquid assets, staples of quantitative HANK models, e.g.
Kaplan et al (2018). The paper proposes a decomposition of cyclical income risk into one part due
to cyclical inequality and one related to skewness: cyclical variations in the likelihood of ending up
in the bad state. I argue below that this distinction is key for some of the theoretical properties of
the model. Third, the version with liquidity allows an analytical solution for the key "intertemporal
MPC" statistics that Auclert, Rognlie, and Straub (2018) introduced in a quantitative HANK for
their distinct, "intertemporal" Keynesian cross. Finally, it delivers key stylized statistical properties
1 Theabbreviation is due to Kaplan, Moll, and Violante (2018); the opening sentence is a paraphrase of Marx and Engels.
2 Overwhelming evidence was long available for the failure of an aggregate Euler equation, for a high fraction of households
having zero net worth and a high marginal propensity to consume MPC, "hand-to-mouth". Important work clarified the link
with liquidity constraints: some "wealthy" households behave as hand-to-mouth if wealth is illiquid (Kaplan and Violante
(2014)), perhaps because it consists of a mortgaged house (Cloyne, Ferreira, and Surico (2015)).
3 The T in THANK stands for "tractable" and for "two" (states/types, and assets), symbolizing the model’s bridging HANK

and (two-agent) TANK, the analytical version in Bilbiie (2008) centered on the asset market participation distinction. Galí,
Lopez-Salido and Valles (2007) embedded a different distinction in a quantitative model, between holding or not physical capital.

1
of idiosyncratic income emphasized by a large empirical literature: autocorrelation, cyclical variance
and (negative) skewness, and leptokurtosis (e.g. Guvenen et al (2014)).
To the best of my knowledge and to this date, THANK is the only tractable framework, of the
several reviewed below, that simultaneously captures all of these important features of the rich
micro-heterogeneity models. This, in my view, makes it particularly suitable for the full (short-term)
macro analysis pursued in this paper.
This paper’s contribution is manifold, as afforded by the tractable model. The determinacy
analysis provides a first modified Taylor principle in this environment, with the differential effect
of pro- and countercyclical inequality on the threshold. Understanding determinacy properties is
crucial for being able to even solve quantitative models, and for policy discussions: what rules or
institutions can anchor expectations when distributional mechanisms are at play. Another key con-
tribution is the decomposition of "income risk" into income inequality and "pure" (skewness-related)
risk, with different implications for transmission. The paper identifies a Catch-22 for HANK mod-
els, that amplification and multipliers in HANK models require countercyclical inequality, but this
also aggravates the forward guidance puzzle (the resolution of which would require procyclical in-
equality). This can be resolved if the "pure risk" channel is procyclical, but is further aggravated if it
is countercyclical. This paper is also the first to show that a Wicksellian rule delivers determinacy in
a HANK economy regardless of how countercyclical inequality and risk are; as a corollary, it allows
preserving amplification while ruling out the FG puzzle, thus sidestepping the Catch-22. These are
virtues shared with a fiscal rule setting nominal debt originally proposed by Hagedorn, that I prove
analytically in the version with liquidity. A final novel contribution is the optimal monetary policy
analysis, emphasizing the different roles of inequality and risk.
Adding a standard Phillips curve, I study monetary policy both as interest rate rules and as
aggregate welfare-maximizing optimal policy. Under a further inconsequential simplification, the
model reduces to one first-order difference equation whose root governs aggregate dynamics and
depends chiefly on the cyclicality of inequality, that is on the constrained agents’ income elasticity
to aggregate income χ. AD-amplification occurs with countercyclical inequality, χ > 1: a demand
increase leads to a disproportionate increase in constrained agents’ income and a further demand
expansion, the intertemporal version of which delivers compounding in the aggregate Euler equation.
Conversely, procyclical inequality χ < 1 leads to AD-dampening and Euler-equation discounting.
The determinacy properties of Taylor rules reflect this intuition. When inequality is countercyclical,
the central bank needs to be possibly much more aggressive than the "Taylor principle" (increas-
ing nominal interest more than inflation) to rule out indeterminacy. Whereas in the discounting,
procyclical-inequality case, the Taylor principle is sufficient but not necessary: for a large region
there is determinacy even under a peg, undoing the Sargent-Wallace result.
The Catch-22 is that the condition for amplification relative to RANK and multipliers—that much
of the literature uses HANK models for—is countercyclical inequality χ > 1. Yet ruling out the for-
ward guidance puzzle—that the later an interest rate cut takes place the larger its effect today (Del Ne-
gro, Giannoni, and Patterson, 2012)—requires the opposite: procyclical inequality χ < 1 (evidently,
χ > 1 implies an aggravation of the puzzle). A possible way out is for the distinct, non-inequality-
related "pure" risk to be procyclical enough to compensate. My model includes a novel formalization
of such a channel (emphasized previously by others as reviewed below) that can also give rise to
Euler discounting, through a different mechanism: if uninsurable risk increases in expansions, pre-

2
cautionary saving leads agents to cut back demand. A nagging policy implication is that empirically,
both channels seem to be countercyclical, implying that the puzzle is (double-)aggravated and de-
terminacy requirements with a Taylor rule become very stringent.
While indeterminacy and puzzles are pervasive with heterogeneity under countercyclical in-
equality and risk, I show that the Wicksellian price-level targeting rule, introduced in RANK by
Woodford (2003) and Giannoni (2014), ensures determinacy and rules out the FG puzzle in THANK
no matter how strong the Euler-equation compounding, resolving the Catch-22. Similar virtues are
shared by the debt quantity rule proposed by Hagedorn (2020) and shown by Hagedorn et al (2018)
to sidestep the Catch-22. I prove analytically both this and Auclert et al’s numerical determinacy
criterion based on intertemporal MPCs in my model’s version with liquidity.
Optimal monetary policy in quantitative HANK is subject to phenomenal technical challenges,
many of which are resolved in a recent contribution by Bhandari, Evans, Golosov, and Sargent
(2021); I calculate optimal policy analytically in THANK, approximating aggregate welfare to second-
order to derive a quadratic objective function for the central bank. This encompasses a novel in-
equality motive relative to RANK, implying optimally tolerating more inflation volatility when more
households are constrained. While inequality is of the essence for optimal policy, risk is not—insofar
as the policymaker shares society’s first-best (perfect-insurance) objective. Risk does matter for im-
plementation: with countercyclical inequality and idiosyncratic risk, the interest-rate rule that im-
plements optimal discretionary policy may entail cutting real rates, when in RANK it would imply
increasing them. Furthermore, optimal policy under commitment ensures determinacy regardless
of heterogeneity and inequality-cyclicality and, while affected by similar inequality considerations,
amounts to a form of price-level targeting.
Related Literature—Quantitative HANK models with rich heterogeneity and feedback effects
from equilibrium distributions to aggregates are increasingly used to address a wide spectrum of
issues in macroeconomic policy.4 This paper develops an analytical representation of the richer-
heterogeneity models in order to gain insights into their mechanisms and thus belongs to an emerg-
ing literature reviewed below and in more detail in Appendix A, including my own previous work.
Most other analytical studies focus on the role of cyclical income risk without disentangling the
role played by cyclical inequality. The clearest example is Acharya and Dogra (2020), which isolates
the cyclical-risk channel by using CARA preferences to simplify heterogeneity and shows that in-
tertemporal amplification may occur purely as a result of income volatility going up in recessions.
With this different mechanism, it also studies determinacy and the forward guidance puzzle mak-
ing explicit reference to the analysis in this paper’s previous version.5 Ravn and Sterk (2020) study
a complementary analytical HANK with endogenous (through search and matching) unemployment
risk and analyze determinacy and shock transmission, while Challe (2020) analyzes optimal mone-
tary policy in that model. Werning (2015) studies the possibility of AD amplification/dampening of
monetary policy in a different, general model of cyclical risk and market incompleteness, without
discussing the distinction between inequality and risk and without analyzing any of the topics of
4 The effects of transfers (Oh and Reis, 2012); liquidity traps (Guerrieri and Lorenzoni, 2017); job-uncertainty-driven re-
cessions (Ravn and Sterk, 2017; den Haan, Rendahl, and Riegler, 2018); monetary transmission (Gornemann, Kuester, and
Nakajima, 2016; Auclert, 2018; Debortoli and Gali, 2018; Auclert and Rognlie, 2017); portfolio composition (Bayer et al, 2019
and Luetticke, 2021); fiscal policy (Ferrière and Navarro, 2018, Hagedorn, Manovskii, and Mitman, 2018; Auclert, Rognlie, and
Straub, 2018; McKay and Reis, 2016; Cantore and Freund, 2019); the FG puzzle (McKay et al, 2016).
5 Also subsequently to this paper, Auclert et al (2018) provided numerical determinacy results emphasizing the cyclicality of

risk in quantitative HANK; Acharya and Dogra (2020) stemmed from a discussion of it meant to provide analytical insights.

3
this paper (determinacy with various rules, Catch-22, optimal policy).6 Broer, Hansen, Krusell, and
Oberg (2020), in another analytical HANK, show that wage rigidity can cure some of the uncomfort-
able implications brought about by the dynamics and distribution of profits, some of which occur in
TANK in Bilbiie (2008). The companion paper Bilbiie (2020) abstracts from cyclical risk and liquidity
and analyzes different issues, focusing on the important role of TANK’s cyclical-inequality channel
in HANK transmission in and of itself. Both that paper and Debortoli and Galí (2018) use the TANK
version in Bilbiie (2008) to approximate some aggregate implications of some HANK models.
Fiscal multipliers under heterogeneity have been analyzed in several quantitative HANK models
cited above and in TANK for spending (Galí et al (2007)), transfers (e.g. Bilbiie, Monacelli and
Perotti (2013)) or both, in liquidity traps (Eggertsson and Krugman (2012)). Hagedorn (2020) has
shown that the demand for nominal bonds inherent in incomplete-market economies coupled with
a supply rule leads to price-level determinacy and rules out the FG puzzle;7 I provide an analytical
application of this in my model. Hagedorn et al (2018) explored the implications of such a rule for
fiscal multipliers, and thus for sidestepping what this paper calls the Catch-22.
Finally, this paper is related to studies of optimal policy: in RANK (i.a. Woodford (2003)) and
with heterogeneity in TANKs (Bilbiie (2008); Nistico (2016); Curdia and Woodford (2016)). Recent
different analytical HANKs providing complementary insights include Challe (2020) and Bilbiie
and Ragot (2020). Important optimal-policy studies in rich-heterogeneity quantitative HANK imply
deviations from price stability: Bhandari, Evans, Golosov, and Sargent (2021) emphasize inequality
motives, and Nuño and Thomas (2021) redistribution with nominal assets.

2 THANK: An Analytical HANK Model

This section outlines THANK, an analytical HANK model that captures several key channels of
complex HANK models: cyclical inequality, self-insurance in face of idiosyncratic uncertainty, and
a distinction between liquid and illiquid assets. While related to several studies reviewed in the
Introduction, the exact model is to the best of my knowledge novel to this paper and its companion
Bilbiie (2020), which uses a special cases of it (with this paper as its reference for the full model),
focusing on AD amplification of monetary and fiscal policies through a "New Keynesian Cross" and
on using it as a one-channel approximation to richer HANK models.
A unit mass of households j 2 [0, 1] discount the future at rate β, derive utility from consumption
j j
Ct and dis-utility from labor supply Nt , and have access to two assets: a government-issued riskless
bond (with nominal return it > 0), and shares in monopolistically competitive firms. Households
participate infrequently in financial markets and freely adjust their portfolio and receive dividends
from firms when they do. When they do not, they receive only the payoff from previously accumu-
lated bonds. Denote these two states, for ease and anticipating what equilibrium we will focus on,
S for participating (from "savers") and H from "hand-to-mouth" for non-participants.
The exogenous change of state follows a Markov chain: the probability to stay type S and H
is respectively s and h, with transition probabilities 1 s and 1 h; later on, I assume that s is a
6 Holm (2021) shows that the effectiveness of monetary policy is reduced with (yet another model of) procyclical risk; see

also Bernstein (2021) and Caramp and Silva (2021) for other recent analytical frameworks.
7 Other modifications of the NK model that solve its puzzles include changing the information structure (Garcia-Schmidt

and Woodford (2019), Gabaix (2019), Angeletos and Lian (2017), Farhi and Werning (2019), Woodford (2018)), pegging interest
on reserves (Diba and Loisel (2017)), wealth in the utility function (Michaillat and Saez (2017), Hagedorn (2018)).

4
function of aggregate activity. I focus on stationary equilibria whereby by standard results the mass
of H is the unconditional probability λ (with 1 λ the mass of S):8

1 s
λ= , (1)
2 s h

that is the ergodic distribution (in the idiosyncratic dimension) around which we approximate the
model with respect to aggregate shocks. At one extreme stands TANK: permanent idiosyncratic
shocks (s = h = 1 ) and λ fixed at its initial free-parameter value. Other useful special cases include
s = h = 0 with agents oscillating between states every other period and λ = 21 ; and iid idiosyncratic
shocks s = 1 h = 1 λ (being S or H tomorrow is independent on today’s state).
Key assumptions on the asset market structure simplify the equilibrium and afford an analytical
solution; the precise combination of assumptions used is novel to this paper, although subsets have
been used by existing literature. Notably, the setup builds on the seminal contributions of Lucas
(1990) and Shi (1997) in monetary theory.9 First, households belong to a family whose utilitarian
(equally-weighted) intertemporal welfare its head maximizes facing limits to risk sharing. House-
holds can be thought of as being in two states or "islands", with all participants on the same island S
and all non-participants on island H. The family head can transfer all resources across households
within the island, but can only transfer some resources between islands.
In face of idiosyncratic risk there is thus full insurance within type, after idiosyncratic uncertainty
is revealed, but limited insurance across types. At the beginning of the period, the family head pools
resources within the island. The aggregate shocks are revealed and the family head determines the
consumption/saving choice for each household in each island. Then households learn their next-
period participation status and have to move to the corresponding island accordingly, taking only
bonds with them. There are no transfers to households after the idiosyncratic shock is revealed,
and this is taken as a constraint for the consumption/saving choice. Different assets thus have
different liquidity: only one of the two assets (bonds) can be used to self-insure before idiosyncratic
uncertainty is revealed, i.e. is liquid. While stocks are illiquid—they cannot be used to self-insure.
The flows across islands are as follows. The total measure of households leaving island H each pe-
riod (who participate next period) is (1 h) λ; the rest λh stay. Likewise, a measure (1 s) (1 λ)
leaves island S for H at the end of each period. Total welfare maximization implies that the family
head pools resources at the beginning of the period in a given island and implements symmet-
j
ric consumption/saving choices for all households in that island. Denote by Bt+1 per-capita, real
beginning-of-period-t + 1 bonds on island j = S, H: after the consumption-saving choice, and also
after changing state and pooling. The end-of-period-t (after the consumption/saving choice but before
j
agents move across islands) per capita real values are Zt+1 . We have the following relations (stocks
stay on the S island, so we ignore them):

(1 λ) BtS+1 = s (1 λ) ZtS+1 + (1 h) λZtH+1

λBtH+1 = (1 s ) (1 λ) ZtS+1 + hλZtH+1 .

8 The h 1 h
stationary distribution (1) is found by solving: λ 1 λ = λ 1 λ .
1 s s
9 Closer to this literature, this way of reducing heterogeneity and eliminating the wealth distribution as a state variable also

extends Challe and Ragot (2011), Challe et al (2017), Heathcote and Perri (2018) and Bilbiie and Ragot (2020). NK models with
two switching types were studied by Curdia and Woodford (2016) and Nistico (2016) but with different insurance structures.

5
or rescaling by the relative population masses and using (1):

BtS+1 = sZtS+1 + (1 s) ZtH+1 (2)

BtH+1 = (1 h) ZtS+1 + hZtH+1 .

The program of the family head is (with π t denoting the net inflation rate):
h i
W BtS , BtH , ω t = max (1 λ) U CtS + λU CtH + βEt W BtS+1 , BtH+1 , ω t+1
fCtS ,ZtS+1 ZtH+1 ,CtH ,ω t+1 g

subject to the laws of motion for bond flows (2) and to standard budget constraints for the respective
j
households, where post-tax incomes are Yt , ω t is the per-capita fraction of the portfolio of shares
(with price vt ) held only by S who receive dividends Dt ; all households receive the real return on
their respective bond holdings and face positive constraints on new bond holdings (4).

1 + it 1 S
CtS + ZtS+1 + vt ω t+1 = YtS + B + ω t ( v t + Dt ) ,
1 + πt t
1 + it 1 H
CtH + ZtH+1 = YtH + B (3)
1 + πt t
ZtS+1 , ZtH+1 0 (4)

The Kuhn-Tucker conditions with complementary slackness are:

1 + it h 0 i
U 0 CtS βEt sU CtS+1 + (1 s) U 0 CtH+1 (5)
1 + π t +1
1 + it h 0 i
and 0 = ZtS+1 U 0 CtS βEt sU CtS+1 + (1 s) U 0 CtH+1 ;
1 + π t +1
1 + it h i
U 0 CtH βEt (1 h) U 0 CtS+1 + hU 0 CtH+1 (6)
1 + π t +1
1 + it h i
and 0 = ZtH+1 U 0 CtH βEt (1 h) U 0 CtS+1 + hU 0 CtH+1 ;
1 + π t +1
v t + 1 + Dt + 1 0
U 0 CtS βEt U CtS+1 and ω t+1 = ω t = (1 λ) 1 .
vt

The key is the Euler equation (5), governing the bond-holding decision of S self-insuring against
the risk of becoming H and taking into account that bonds can be used when moving to the H island.
Equation (6) determines the bond choice of agents in the H island; both bond Euler conditions are
written as complementary slackness conditions. With this market structure, the Euler equations (5)
and (6) are of the same form as in fully-fledged incomplete-markets model of the Bewely-Huggett-
Aiyagari type. In particular, the probability 1 s measures the uninsurable risk to switch to a bad
state next period, risk for which only bonds can be used to self-insure—thus generating a demand
for bonds for "precautionary" purposes. The last Euler equation corresponds to illiquid shares: there
is no self-insurance motive, for they cannot be carried to the H state, so it is the same as with a

6
representative agent, merely determining the price vt residually.10
To deliver the simple equilibrium representation, I focus on equilibria where the constraint of H
agents always binds and their Euler "equation" (6) is in fact a strict inequality, whatever the reason:
for instance, because the shock is a "liquidity" or impatience shock making them want to consume
more today (decreasing β in (6) to a low enough value β H ), or because their average income in that
state is lower enough than in the S state, e.g. if average profits are high enough; or simply because
a technological constraint prevents them from accessing any asset markets.
I consider two equilibria, according to whether liquidity is supplied or not: as a benchmark,
the zero-liquidity limit reminiscent of Krusell, Mukoyama and Smith (2011)11 ; and then in section
4, an equilibrium with government-provided liquidity. In the former case, we assume that even
though S’s demand for bonds is well-defined (their constraint is not binding), the supply is zero so
there are no bonds held in equilibrium. Under these assumptions the only equilibrium condition
from this part of the model is the Euler equation for bonds of S (5) holding with equality.12 The
H’s constraint binding and zero-liquidity implies that they are hand-to-mouth CtH = YtH . Because
transition probabilities are independent of history and there is full insurance within type, all agents
who are H in a given period have the same income and consumption.
The rest of the model is exactly like the TANK version in Bilbiie (2008, 2020), nested with s = 1.
The λ households who are "hand-to-mouth" H do make an optimal labor supply decision deter-
mining their income. The income of H is YtH = Wt NtH + Tt H , where W the real wage, N H hours
and Tt H fiscal transfers to be spelled out. The remaining 1 λ agents also work, and receive
the profits from the illiquid shares net of taxes and transfer. The choice of hours worked deliv-
j j j j
ers the standard intratemporal optimality condition for each j: UC Ct = Wt UN Nt . Defining
j j j j
σ 1 UCC C j /UC as risk aversion and ϕ UNN N j /UN as the inverse labor supply elasticity, and
small letters log-deviations from steady-state (to be discussed below), we have the labor supply for
j j
each j: ϕnt = wt σ 1 ct . Assuming for tractability that elasticities are identical across agents, the
same holds on aggregate ϕnt = wt σ 1 ct .
The supply, firms’ side is standard and outlined for completion in Appendix A.2. A notable
feature is that I assume as a benchmark the standard NK optimal sales subsidy inducing marginal
cost pricing. This policy is redistributive: since steady-state profits are zero D = 0, it taxes the
firms’ shareholders and results in the "full-insurance" steady-state used here as a benchmark C H =
C S = C. Loglinearizing around it, with dt ln ( Dt /Y ), profits vary inversely with the real wage:
dt = wt , an extreme form of the general property of NK models. This series of assumptions
(optimal subsidy, steady-state consumption insurance, zero steady-state profits) is not necessary for
the results and can be easily relaxed, but makes the algebra more transparent.
Firms’ optimal pricing under Rotemberg costs implies the loglinearized Phillips curve:

π t = βEt π t+1 + κct + ut , (7)

10 As households pool resources when participating (which would be optimal with t = 0 symmetric agents and t = 0

trading), they perceive a return conditional on participating next period. This exactly compensates for the probability of not
participating next period, generating the same Euler equation as with a representative agent.
11 Other zero-liquidity HANK include i.a. Ravn and Sterk (2020), Werning (2015), and Broer et al (2020).
12 The Euler equation prices these possibly non-traded bonds, just like in RANK or TANK—but now the pricing takes into

account the possible transition to the constrained H state.

7
where ut are cost-push shocks that I abstract from until studying optimal policy in Section 5. To
obtain maximum tractability and closed forms, I first focus on the simplest special case:

π t = κct , (8)

nested in (7) above with myopic firms (β = 0), used previously in a different context in Bilbiie (2019).
Appendix A.2 microfounds this assuming that firms pay a Rotemberg cost relative to yesterday’s
market average price index, rather than to their own individual price (the latter leads to (7)). That
is, firms ignore the impact of today’s price choice on tomorrow’s profits. While over-simplified,
this nevertheless captures the key supply-side NK trade-off between inflation and real activity and
allows us to isolate and focus on the essence of this paper: AD. The results reassuringly generalize
to the standard Phillips curve (7), as I show in Appendix C.
The government conducts fiscal and monetary policy. The former consists of a simple endogenous
redistribution scheme: taxing profits at rate τ D and rebating the proceedings lump-sum to H who
D
thus receive τλ Dt per capita; this is key here for the transmission of monetary policy, understood as
changes in the nominal interest rate it . In the version with liquidity, the government also supplies
liquid nominal bonds and levies lump-sum uniform taxes/transfers on households.
Market clearing implies for equilibrium in the goods and labor market respectively Ct λCtH +
λ) CtS = 1 Yt and λNtH + (1 λ) NtS = Nt . With uniform steady-state hours N j = N
ψ 2
(1 2 πt
by normalization and the fiscal policy assumed above inducing C j = C, loglinearization around a
zero-inflation steady state delivers yt = ct = λctH + (1 λ) cSt and nt = λntH + (1 λ) nSt .

2.1 Cyclical Income Risk and Inequality in THANK

A keystone to this paper’s analysis is to define and distinguish income inequality and risk, and their
cyclicality. I define income inequality as the ratio of income in the two states Γt YtS /YtH ; this is
proportional to the unconditional variance of log income:

j
var ln Yt = λ (1 λ) (ln Γt )2 ;

in Appendix B.1, I show that this is also proportional to other standard inequality measures like
the Gini coefficient and generalized entropy. Importantly, as we will see in the model’s equilibrium
inequality is cyclical: it depends on aggregate output Γ (Yt ).
In the data and in quantitative HANK models alike, income risk is generally cyclical. Other
analytical HANK frameworks model cyclical idiosyncratic risk as either unrelated (Acharya and
Dogra (2020)) or differently related (Challe et al (2017); Holm (2021); Ravn and Sterk (2020); Werning
(2015)) to liquidity constraints and hand-to-mouth behavior. To capture a component of cyclical
risk that is distinct from cyclical inequality and thus further differentiate from the cited papers, I
assume that the probability of becoming constrained depends on tomorrow’s aggregate demand
1 s (Yt+1 ).13 If the first derivative of 1 s (.) is positive s0 (Yt+1 ) > 0, the probability is higher
in expansions so, insofar as being constrained leads on average to lower income, this makes income
13 Ina model with endogenous unemployment risk like Ravn and Sterk or Challe et al, this happens in equilibrium through
search and matching. This is also related to Werning’s Section 3.4, where nevertheless it is unconditional probabilities (and
population shares) that are cyclical. Here, to capture purely idiosyncratic (as opposed to "aggregate") variation, λ is invariant.

8
risk procyclical (go up in expansions). Conversely, s0 (Yt+1 ) < 0 makes risk countercyclical.
A precise definition of "income risk" is notoriously controversial. The literature often employs
the conditional variance of idiosyncratic (log) income, found to be countercyclical in the data by
Storesletten, Telmer, and Yaron (2004). This is easily calculated in my two-state model as:

var ln YtS+1 j ln YtS = s (Yt+1 ) (1 s (Yt+1 )) (ln Γt+1 )2 . (9)

Of particular interest for the dynamic properties of the model is its cyclicality. Taking a first-order
Taylor expansion around an arbitrary steady state Γ, this is made of two components:
2 3
6 7
6 ΓY 1 7
var ln YtS+1 j ln YtS s (1 2
s) (ln Γ) ' 2 ln Γ 6
6 Γ s (1 s ) sY s ln Γ7
7 y t +1 (10)
4| 2
{z }| {z }5
inequality pure risk

The first component is due to cyclical inequality: when ΓY < 0, inequality is countercyclical
and so is risk because income at the bottom overreacts, increasing variance in both expansions and
recessions. This "intensive" margin of risk operates even when the second channel is absent, i.e. with
constant s or symmetric distribution s = 21 . The second component is intuitively and formally related
to the cyclicality of conditional skewness that Guvenen, Ozkan, and Song (2014) argued forcefully in
favor of using as a measure of cyclical income risk, see also Mankiw (1986).14 I derive this formally
in Appendix B, but the intuition is simple: skewness is negative whenever s > 21 (there is left-tail
risk). When 1 s (Y ) is decreasing with aggregate activity sY < 0, it becomes more likely to draw
from the left tail in recessions; hence, (skewness) risk is countercyclical: upward income movements
become less likely and downward movements more likely in a recession. This "extensive" margin
of risk operates even with acyclical inequality ΓY = 0.
This simple setup allows for an arbitrary, general relationship between risk and inequality: it
depends on both their levels and cyclicalities, and their interactions, as parameterized by Γ, s, ΓY
and sY . It is useful for the further analysis to consider the following polar cases:
a. no inequality in levels, Γ = 1: risk is acyclical to first order only. As we will see, risk has
no impact on the first-order equilibrium in this benchmark. (naturally, risk is still cyclical to higher
orders and away from this steady state).
b. no risk in levels, s = 1 or s = 0: the former is TANK (no transition between states), the latter
has agents oscillating between states every other period and 12 mass in each state.
c. acyclical inequality, ΓY = 0; risk is cyclical only through the second channel sY 6= 0, and only
if there is level inequality Γ > 1 and skewness s 6= 21 .
d. acyclical (pure) risk, sY = 0; risk is cyclical only through the inequality channel ΓY 6= 0, and
(to first order) only if there is level inequality Γ > 1.
Thus, the model nests several scenarios that are useful to disentangle the importance of the
corresponding economic mechanisms, as we will see below. While the precise decomposition is of
course model-specific, the general idea and mechanisms and their equilibrium implications derived
below transcend the simple model used here.
p
14 Appendix B derives higher moments: formally, conditional skewness (1 2s) / s (1 s) < 0 when s > 21 .

9
2.2 Cyclical Inequality and Aggregate Demand in THANK

We derive an aggregate Euler-IS equation by taking an approximation (in the aggregate-shocks di-
mension) around the ergodic idiosyncratic distribution with relative shares given by (1). To isolate
the role of cyclical inequality, we first approximate around a symmetric steady-state with Γ = 1 and
C H = C S . Start from the individual self-insurance Euler equation (5):

cSt = sEt cSt+1 + (1 s) Et ctH+1 σ (i t Et π t+1 ) . (11)

j
To express this in terms of aggregates, we need individual ct as a function of aggregate ct and yt .
Once idiosyncratic uncertainty is revealed and asset markets clear, this part—one of many possible
examples of how the income distribution depends on aggregate income—is exactly as in the TANK
model in Bilbiie (2008, 2020), for simplicity. I summarize its main implications here and refer to the
reader to the Appendix for a complete derivation and to those papers for a thorough discussion.
In equilibrium, individual consumption/income is related to aggregate income by:

τD
ctH = ytH = χyt , χ 1+ϕ 1 7 1; (12)
λ
1 λχ
cSt = yt .
1 λ

The composite parameter χ is the model’s keystone: a sufficient statistic that in this specific model
depends on fiscal redistribution 1 τ D /λ and labor market characteristics ϕ. It is important to
stress that this is but one possible simple theory of the income distribution.15 Equilibrium cyclical
income inequality γt , the log deviation of Γt , reveals that ΓY /Γ in (10) is proportional to 1 χ:

yt
γt ySt ytH = (1 χ) . (13)
1 λ

Inequality is procyclical (∂γ/∂y > 0) iff χ < 1 and countercyclical (∂γ/∂y < 0) iff χ > 1.
Distributional considerations make χ different from 1. In RANK, such considerations are absent
since one agent works and receives all the profits. When aggregate income goes up, labor demand,
real wages and marginal cost increase. This decreases profits, but because the same agent incurs both
the labor gain and the profit loss, the redistribution of income across factors is neutral.
Take now the case with heterogeneity and χ > 1, i.e. fiscal redistribution of profits that is not
skewed towards H, τ D < λ and upward-sloping labor supply ϕ > 0.16 If demand goes up, the wage
goes up, H’s income increases and so does their demand. Thus aggregate demand increases by more
than initially, shifting labor demand and increasing the wage even further, and so on. In the new
equilibrium, the extra demand is produced by S, whose decision to work more is optimal given the
income loss from falling profits of which they get a disproportionate share even post-redistribution.
With χ < 1 (when H receive a disproportionate share of the profits τ D > λ) the AD expansion
is instead smaller than the initial impulse, as H recognize that this will lead to a fall in their income;
15 Several subsequent models deliver different distributional implications, e.g. using sticky wages. See Colciago (2011), Ascari
et al (2017) and Walsh (2018) in TANK; Broer et al (2020), Hagedorn et al (2018), and Auclert et al (2018) in HANK.
16 The benchmark used by Campbell and Mankiw’s (1989) seminal paper is χ = 1, which occurs when profits are uniformly

redistributed τ D = λ or labor is infinitely elastic ϕ = 0; income inequality is then acyclical. See also Bilbiie (2008, footnote 14).

10
while S, given the positive income effect from profits, optimally work less. As the income of H now
moves less than proportionally with aggregate income, inequality is procyclical.
Replacing the individual (12) in the self-insurance equation (11), we obtain the aggregate Euler-IS:

1 λ 1 s
ct = δEt ct+1 σ (i t Et π t+1 ) , where δ 1 + (χ 1) . (14)
1 λχ 1 λχ

The contemporaneous AD elasticity to interest rates is the TANK one, σ 11 λχ λ

, reflecting the described
New Keynesian Cross logic. Although the direct effect of interest rates is scaled down by 1 λ, the
indirect effect is increasing with λ at rate χ; with χ > 1 the latter dominates, delivering amplification
relative to RANK (dampening, for χ < 1). λχ is thus akin to an aggregate-MPC slope of a planned-
expenditure line: it yields amplification when χ > 1, for then the slope increase dominates the
decrease in the shift of this line due to adding λ agents who are directly insensitive to policy.17
The key novelty of THANK’s aggregate Euler-IS equation relative to TANK is that it is charac-
terized by compounding δ > 1 iff inequality is countercyclical χ > 1 and discounting δ < 1 if
procyclical χ < 1 (see Proposition 3 in the companion paper Bilbiie (2020)). In RANK and TANK,
good future income news imply a one-to-one demand increase today as households (who can) sub-
stitute consumption towards the present and, with no assets, income adjusts. Discounting occurs
when procyclical inequality meets idiosyncratic uncertainty: When good news about future aggre-
gate income arrive, households recognize that in some states they will be constrained and not benefit
fully from it. They self-insure, increasing their consumption less than if there were no uncertainty:
the saving demand increase cannot be accommodated (there is no asset), so income falls accordingly.
Countercyclical inequality leads to compounding instead. The Keynesian-cross amplification
extends intertemporally: good aggregate income news boost today’s demand because they imply less
need for self-insurance. Since future income in states where the constraint binds over-reacts to good
aggregate news, households demand less saving. With zero savings in equilibrium, households
consume more than one-to-one and income increases more than without risk.
The foregoing focuses on cyclical inequality and embeds a notion of idiosyncratic risk that is
intimately related to whether liquidity constraints bind or not but is by construction acyclical as can
be seen by recalling the discussion after (10): locally around Γ = 1 variance is acyclical because it is
proportional to ln Γ. Idiosyncratic risk itself may still be cyclical, but this has locally no first-order
effect on the variance, on precautionary saving, and thus on Euler discounting-compounding.
The other useful special case illustrating the independent role of cyclical inequality is the limit
s = 0, with no risk at all as the conditional variance is nil.18 Even in that extreme case, my model
implies Euler discounting-compounding with, replacing s = 0 and λ = 21 in (14):

χ
δ j s =0 = 7 1 iff χ 7 1. (15)
2 χ

There is again discounting with pro- and compounding with counter-cyclical inequality. These
two observations illustrate clearly that cyclical risk is not necessary for Euler-equation discount-
ing/compounding: cyclical inequality is sufficient, combined with a self-insurance motive.
17 As shown in Bilbiie (2020) the aggregate MPC out of aggregate income is a convex combination of the two out-of-own-
income MPCs weighed by the elasticities to the aggregate, mpc = (1 λ) (1 β) 11 λχ λ +λ 1 χ = 1 β (1 λχ) .
18 This limit case is akin to Woodford (1990), abstracting from the endogenous income distribution that is of the essence here.

11
2.3 Cyclical Inequality and Risk

I now turn cyclical risk back on, both by allowing the probability s to depend on the cycle and
approximating around a steady state with inequality Γ > 1. Risk is cyclical through both channels,
see (10), and matters to first order—the loglinearized aggregate Euler-IS becomes (see Appendix B):

1 λ
ct = [ δ̃
|{z} + η ] Et ct+1 σ (it Et π t+1 ) (16)
|{z} 1 λχ
cycl.-ineq HANK cycl.-risk HANK | {z }
cyclical-inequality TANK
(χ 1) (1 s̃)
with δ̃ δ + Γ1/σ 1 (δ 1) s̃ = 1 +
1 λχ
sY Y 1/σ 1 λ
and η 1 Γ (1 s̃) σ ,
1 s 1 λχ

(1 s)Γ1/σ
where 1 s̃ = s+(1 s)Γ1/σ > 1 s is the inequality-weighted transition probability measure of risk.
There are thus two effects of income risk cyclicality. The first comes from income inequality,
which around a steady-state with Γ > 1 and s > 0 also makes risk cyclical, as discussed above and
captured by Γ1/σ 1 (δ 1) s̃. When δ > 1 there is an additional source of compounding that
increases with the level of inequality. The second effect encapsulates "pure" (independent on cyclical
inequality) cyclical risk through its key determinant: the elasticity sY Y/ (1 s), the second term
in (10) above, which determines the novel parameter η. Dampening/amplification of future shocks
only occurs through η even with acyclical inequality δ = 1. Procyclical risk η < 0 implies Euler
discounting: good news generate an expansion today to start with, which increases the probability
of moving to the bad state and triggers precautionary saving, containing the expansion. Conversely,
countercyclical risk generates compounding: an aggregate expansion reduces the probability of mov-
ing to the bad state and mitigates the need for insurance, amplifying the initial expansion.19
This formalization of cyclical risk has similar reduced-form implications for the link between
current and future consumption to the cyclical-inequality channel, but the underlying economic
mechanism is different. Furthermore, while η is observationally equivalent to Acharya and Dogra’s
(2020) different formalization with CARA preferences leading to a P(seudo-)RANK, the underlying
implications for risk are also different. In my model, η captures the cyclicality of skewness, a key ele-
ment of the reviewed evidence; whereas Acharya and Dogra’s PRANK relies on symmetry (normal
shocks), abstracting from skewness to focus on variance. Ravn and Sterk (2020)’s model delivers
something akin to η based on search and matching, but abstracts from cyclical inequality. While
Werning’s (2015) general non-linear model contains both channels but without the distinction and
decomposition and without identifying their differential effects on transmission.
The pure-risk channel captured by η operates only if there is long-run level inequality Γ > 1, the
literal risk of moving to a lower income level. Whereas the cyclical-inequality channel relies only on the
cyclicality of income when constrained χ. Both channels capture precautionary saving: the former,
through the effect of uncertainty and the third derivative of the utility function (η is proportional
to prudence σ); the latter, through the effect of constraints, a separate source of concavity in the
consumption function, a manifestation of the general results in Carroll and Kimball (1996).
19 The Appendix studies an alternative where s depends on current Yt , delivering contemporaneous amplification.

12
 Comparing (16) with its RANK counterpart reveals directly three separate endogenous wedges
corresponding to each channel; in a recent contribution, Berger et al (2021) provide a valuable em-
pirical "accounting" of related wedges, including a decomposition between the ability to smooth
shocks versus the volatility of labor income. Such exercises may help disentangle the signs of δ 1
and η which, as we will see next, are crucial for the model’s properties.

3 THANK Analytics: Determinacy, Puzzles, and Amplification

This section exploits tractability to conduct a pencil-and-paper, full analysis of some main NK topics:
determinacy, solving the FG puzzle, and conditions for amplification-multipliers.

3.1 HANK, Taylor, and Sargent-Wallace

I now solve the model under further assumptions delivering a one-equation representation that may
be useful in different contexts where reducing dimensionality is necessary for closed-form solutions.
In particular, first, the nominal rate it follows a Taylor rule (we study other policies momentarily):

it = φπ t . (17)

The model is completed by adding the simple aggregate-supply, Phillips-curve specification (8); all
the results carry through with the more familiar forward-looking (7) as I show in Appendix C.
With this simple RANK-isomorphic HANK we first revisit a classic determinacy result and de-
rive a HANK Taylor principle. Replacing (8) and (17) in (14), THANK collapses to one equation:

δ + κσ 11 λ
λχ
ct = νEt ct+1 , where ν (18)
1 + φκσ 11 λχ
λ

captures the effect of good news on AD, and the elasticity to interest rate shocks.
There are three channels shaping this key summary statistic. First, the "pure AD" effect through
δ coming from cyclical inequality, operating even with fixed prices or fixed real rate it = Et π t+1 .
Second, a supply feedback cum intertemporal substitution: the inflationary effect (κ) of good income
news triggers ceteris paribus a fall in the real rate and intertemporal substitution towards today, the
magnitude of which depends on static amplification/dampening 11 λχ λ
. Finally, all this demand
amplification generates inflation and real rate movements. When policy is active φ > 1, a higher real
rate and a contractionary effect today ensue, the strength of which depend on cyclical-inequality.
These considerations drive Proposition 1 (the case with NKPC (7) is in Appendix C.1).

Proposition 1 The HANK Taylor Principle: The HANK model under a Taylor rule (18) has a determinate,
locally unique rational expectations equilibrium if and only if (as long as λ < χ 1 ):

δ 1
ν<1,φ>φ 1+ .
κσ 11 λχ
λ

The Taylor principle φ > 1 is sufficient for determinacy if and only if there is Euler-IS discounting: δ 1.

13
 The proposition follows by recalling that the determinacy requirement is that the root ν be inside
the unit circle; in the discounting case δ < 1, the threshold φ is evidently weaker than the Taylor
principle, while in the compounding case it is stronger. With countercyclical inequality and risk δ >
1, a future aggregate sunspot increase in income generates a disproportionate increase in income in
the bad state, and thus incentives to dis-save and a further demand boost today–making the sunspot
self-fulfilling even with a fixed real rate φ = 1. The central bank needs to do more to counteract this
and prevent the sunspot from becoming self-fulfilling. (The opposite holds in the discounting case
and the Taylor principle is sufficient for determinacy.) The Taylor threshold φ > 1 reappears for
either of χ = 1 (acyclical inequality), s ! 1 (no risk), or κ ! ∞ (flexible prices). The determinacy
region for φ squeezes very rapidly with countercyclical inequality because of a complementarity
(χ 1)(1 s̃)
between idiosyncratic and aggregate risk apparent from φ = 1 + κσ(1 λ) . The threshold depends
on price stickiness because policy responds to inflation, but the relevant amplification goes through
real demand’s equilibrium response, and price stickiness modulates the relationship between the
two. If instead policy responds to real activity it = π t + φc ct , the determinacy threshold is φc >
(χ 1)(1 s)
(1 λ ) σ
and no longer depends on price stickiness because policy then acts directly on demand.

φ 20 φ
1.0

0.8 15

0.6
10
0.4
5
0.2

0.0
λ 0 λ
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0

χ = 0.5 (procyclical inequality) χ = 2 (countercyclical inequality)

Fig. 1: Taylor threshold φ with 1 s = 0 (dash, TANK); 0.04 (solid); λ (dots). Note: determinacy above the curve.

Figure 1 plots the threshold φ as a function of λ (for λ < χ 1 ) for different 1 s, with procyclical
inequality in the left panel and countercyclical in the right. The parametrization assumes κ = 0.02,
σ = 1, and ϕ = 1. In the countercyclical-inequality case, the threshold increases with λ and does so
at a faster rate with higher risk 1 s. The required response can be large: for the calibration used in
Bilbiie (2020) to match Kaplan et al’s quantitative HANK aggregate outcomes (χ = 1.48, λ = 0.37,
1 s = 0.04) it is φ = 2.5 and can be as high as 5 for other calibrations therein. With procyclical
inequality, the Taylor principle is sufficient but not necessary for determinacy. For a large subset of the
region, there is in fact determinacy even under a peg φ = 0, undoing the classic Sargent and Wallace
(1975) result, namely if and only if φ < 0 or:

1 λ
ν0 δ + κσ < 1. (19)
1 λχ

With enough discounting, the sunspot is ruled out by the economy’s endogenous forces, unlike in

14
RANK where ν0 = 1 + κσ 1; as we shall see, (19) also rules out the forward guidance puzzle.
Finally, since adding cyclical risk is isomorphic to a change in δ, the threshold becomes:

δ̃ 1+η
φ 1+ , (20)
κσ 11 λ
λχ

with the different intuition discussed above for AD amplification/dampening through η.20

3.2 A Catch-22 for HANK: No Puzzle, No Amplification?

We are now in a position to state the Catch-22: the closed-form conditions for amplification in THANK
are the opposite of those needed to solve the forward guidance puzzle. To state this formally, we
introduce two policy shocks: discretionary exogenous changes in interest rates it in the Taylor rule
it = φπ t + it ; and public spending: the government buys an amount of goods Gt with zero steady-
state value (G = 0) and taxes all agents uniformly to finance it.21 Straightforward derivation delivers
the aggregate Euler-IS, starting with cyclical inequality only (extending (14)):

1 λ λ ( χ 1)
ct = δEt ct+1 σ (i t Et π t+1 ) + ζ ( gt Et gt+1 ) + (δ 1) Et gt+1 , (21)
1 λχ 1 λχ

where ζ ϕσ/ (1 + ϕσ). Together with the static PC π t = κct + κζgt and AR(1) spending Et gt+1 =
µgt , this delivers Proposition 2 (see Appendix C.2 for the general case with NKPC (7)).

Proposition 2 A Catch-22 for HANK: In THANK with cyclical inequality, there is amplification of mone-
tary policy relative to RANK and the fiscal multiplier on consumption is positive if and only if:

χ > 1,

∂2 c t
whereas the forward-guidance puzzle is ruled out ( < 0) only if
∂( it+T )∂T

χ < 1.

The first part pertains to amplification with respect to RANK, the focus of the majority of quan-
titative HANK studies. Kaplan et al (2018) show that HANK yields monetary policy amplification
through "indirect", general-equilibrium forces; similar insights apply to Auclert (2019), Gornemann
et al (2015), and Debortoli and Galí (2018). Such amplification occurs only with countercyclical in-
equality (Bilbiie (2020) calibrates TANK and the acyclical-risk zero-liquidity THANK to match the
20 Ravn and Sterk (2020) show that SaM delivers η > 0 endogenously, making the Taylor principle insufficient. In works

subsequent to this paper’s determinacy Proposition 1: Acharya and Dogra (2020) derived a Taylor principle and Auclert et al
(2018) provided numerical simulations in a quantitative HANK on the role of cyclical risk for determinacy.
21 The redistribution of the taxation financing spending is essential for the multiplier, see Bilbiie (2020) in TANK: I sidestep it

assuming uniform taxation. See Bilbiie, Monacelli, and Perotti (2013) in TANK, and Oh and Reis (2012), Ferrière and Navarro
(2018), Hagedorn et al (2018) and Auclert et al (2018) in quantitative HANK for multipliers with progressivity.

15
aggregate predictions of these quantitative models). The fiscal multiplier in THANK is:
2 3

1 6 λ (1 µ ) + (1 s) µ 1 λ 7
∂ct ζ 6 7
= 6(χ 1) κσ (φ µ)7 . (22)
∂gt 1 νµ 1 + φκσ 11 λ 4 1 λχ 1 λχ 5
λχ | {z } | {z }
TANK + HANK AD RANK AS

With fixed prices κ = 0 and proportional incomes χ = 1, one recovers the benchmark zero-
multiplier derived in RANK by Bilbiie (2011) and Woodford (2011). Positive multipliers occur with
countercyclical inequality χ > 1: spending has a demand effect that increases labor demand, wages,
the income of H, and so on: the "new Keynesian cross" channel.22 If the stimulus is persistent
(µ > 0), there is an extra kick through self-insurance: as agents expect higher demand and income,
with χ > 1 they expect even higher income in the H state and thus less need to self-insure.23
The second part of Proposition 2 pertains to solving the forward guidance puzzle. The condition
is ν0 < 1, i.e. determinacy under a peg, found by iterating forward (21) with φ = 0 to obtain:

1 λ 1 λ j
ct = ν0 Et ct+1 σ it = ν0T̄ Et ct+T̄ σ Et ∑ T̄j=01 ν0 it+ j .
1 λχ 1 λχ

The time-t response to time-t + T interest rate cut is, for any T 2 (t, T̄ ): ∂ct
= σ 11 λ T
λχ ν0 , which
∂ ( it+ T )
is decreasing in T iff ν0 < 1 (the derivative is σ 11 λχ
λ T
ν0 ln ν0 ). Since with ν0 < 1 the term ν0T̄ Et ct+T̄
vanishes in the limit as T̄ ! ∞, we can solve the equation forward finding a unique solution, i.e. the
earlier determinacy under a peg (φ < 0) result. The condition ν0 < 1 captures a powerful intuition
when rewritten as:
1 λ
1 δ > κσ .
1 λχ
To rule out the puzzle, the HANK-discounting on the left side needs to dominate the right-side
AS-compounding of news that is the source of trouble in RANK. This entails jointly idiosyncratic
uncertainty 1 s > 0 and procyclical enough inequality χ < 1 σκ 11 λs < 1. One implication is
an interpretation of McKay et al (2016), where procyclical inequality holds because profits are redis-
tributed disproportionately to low-productivity households, "as if" τ D > λ in my model (see also
Hagedorn et al (2019) for a quantitative illustration).
Proposition 2 purposefully abstracts from cyclical risk. Turning it back on can in theory resolve
the Catch-22, providing amplification without the puzzle, as emphasized next.
Proposition 3 THANK with cyclical inequality and risk resolves the Catch-22 if and only if, with counter-
cyclical inequality (χ > 1), pure risk is procyclical enough:

η<1 δ̃ 0. (23)

The condition requires that risk procyclicality through the separate s (Y ) channel dominate the
countercyclicality through χ > 1, meaning high enough steady-state inequality in levels Γ and pru-
22 The channel is at work in Gali et al’s (2007) earliest quantitative model on this (but convoluted with several other channels),

as well as in Bilbiie and Straub (2004), Bilbiie, Meier and Mueller (2008), and Eggertsson and Krugman (2012).
23 The last term extends the usual RANK channel: spending is inflationary, which when φ > 1 increases real rates generating

intertemporal substitution towards the future.

16
dence σ, i.e. a strong enough precautionary motive due to uncertainty:

1/σ χ 1
sY Y 1 Γ σ > (1 s) > 0.
1 λ

This reflects the dependence of the risk-cyclicality channel on the level of inequality; recall that with
no inequality in levels Γ = 1, risk cyclicality is irrelevant for Euler discounting/compounding, while
without risk in levels (1 s = 0) the cyclicality of inequality is irrelevant. When the two channels
coexist and go in opposite directions with the right relative strengths, the Catch-22 can be resolved.
However, when the η < 1 δ̃ condition is not met, cyclical risk aggravates the Catch-22. In partic-
ular, when risk and inequality are both countercyclical THANK delivers amplification and aggravates
the puzzles further, while the determinacy conditions with a Taylor rule become even more strin-
gent, see (20). Existing empirical evidence suggests that this is the more plausible scenario. Insofar
as it is a counterpart to η, Guvenen et al’s (2014) measure of skewness risk cyclicality (with acyclical
variance) can be viewed as implying η > 0; while Storesletten et al’s (2004) countercyclical-variance
estimates suggest that the overall sign of the elasticity in the variance approximation (10) is negative,
and thus the aggregate Euler (16) is compounded δ̃ + η > 1. Recent evidence on inequality control-
ling for MPCs suggests that it is also countercyclical (Patterson (2019)). A fruitful complementary
approach is accounting of Euler wedges directly, see Berger et al (2021).
Two caveats to interpreting this evidence through the model’s lens are necessary. First, all the
existing evidence pertains to earnings, whereas the model-relevant object is total income inclusive of
both financial income and transfers. And second, to my knowledge none of the available evidence
isolated and disentangled the two channels in a model-equivalent way, i.e. provide identified para-
meter estimates for Γ, χ, s, and η. In that regard, the paper provides theoretical restrictions to inform
that measurement, as well as reasons for why that is important for macro transmission.
Both the Catch-22 and the FG puzzle, however, are model properties contingent upon the Taylor
rule and intimately related to determinacy: I next study two different policy rules that alleviate these
issues, ensuring determinacy and sidestepping the Catch-22.

3.3 Virtues of a Wicksellian, price-level targeting rule in THANK

Indeterminacy under Taylor rules is pervasive in HANK economies with countercyclical inequality
and risk. What can the central bank do to anchor expectations, when for a standard calibration it
would need to change nominal rates by 5 percent if inflation changed by one percent? One solution
is to adopt the "Wicksellian" policy rule of price level targeting which, as shown by Woodford (2003)
and Giannoni (2014), yields determinacy in RANK:

it = φ p pt with φ p > 0, (24)

This rule is especially powerful in HANK, as emphasized in the following Proposition.

Proposition 4 Wicksellian rule in HANK: In the THANK model, the Wicksellian rule (24) leads to local
determinacy even when δ > 1. Thus, the model delivers "amplification" without also aggravating the FG
puzzle even when both inequality and risk are countercyclical.

17
 The simple proof is outlined in Appendix C.3 (and extended to NKPC in Appendix C.4); the
intuition is that, no matter how strong the AD-amplification, this rule anchors long-run expectations.
Agents recognize that bygones are not bygones: adjustment will eventually take place, as inflation
will a fortiori imply future deflation. The same intuition applies for ruling out the FG puzzle while
delivering amplification, resolving the Catch-22. The essence is that under the Wicksellian rule
THANK reduces, instead of one difference equation (18), to a second-order equation; replacing (8)
rewritten with the price level pt pt 1 = κct and (24) in (16) delivers:

1 λ 1 λ
Et pt+1 1 + ν0 1 1 + σ φ κ p t + ν0 1 p t 1 =σ κν 1 i . (25)
1 λχ p 1 λχ 0 t

It is easy to show that (25) has a unique solution iff φ p > 0, and that the effect of an interest cut
∂ct /∂ it+T decreases with the time T, i.e. the FG puzzle disappears. Intuitively, the puzzle’s
source is indeterminacy under a peg and a Wicksellian rule provides determinacy under a "quasi-
peg": some, no matter how small response to the price level anchors long-run expectations. This
is particularly important in HANK, for even when heterogeneity aggravates the puzzle, the rule
restores standard logic and resolves the "Catch-22".24
This paper assumes throughout a passive-Ricardian fiscal policy. A different route to determi-
nacy and solving the puzzle is to resort to an active, non-Ricardian fiscal rule that does not en-
sure debt repayment for any price level (Leeper (1991); Woodford (1996); Cochrane (2017)). In
incomplete-market economies, yet a different fiscal policy can deliver determinacy, as argued by
Hagedorn (2020); to study it in my model, we need to turn to the equilibrium with liquidity.

4 THANK with Liquidity

In the version with liquidity in the form of government bonds used for precautionary saving, I
derive the savings demand and study the intertemporal propagation and determinacy implications.

4.1 Savings-Liquidity Demand

Denote by BtN+1 the total nominal quantity of government bonds outstanding at the end of each
period. In nominal terms, BtN+1 = (1 + it 1 ) BtN Pt Tt , and in real terms:

Bt+1 = Rt Bt Tt (26)

1+ i
where Rt = 1+tπ t1 is the gross interest rate. The bond market clears Bt+1 = λZtH+1 + (1 λ) ZtS+1 .
Recall now that ZtH+1 = 0, so that Bt+1 = (1 λ) ZtS+1 and using the flow definitions:

1 h 1 s
BtH+1 = (1 h) ZtS+1 = Bt+1 = Bt+1 ,
1 λ λ
24 A hitherto unnoticed to my knowledge corrollary is that in RANK too, the puzzle disappears under a Wicksellian rule.

18
 s
and for S similarly BtS+1 = sZtS+1 = 1 λ Bt+1 . The respective budget constraints imply:

1 s
CtH = ŶtH + Rt Bt (27)
λ
1 s
CtS + Bt+1 = ŶtS + Rt Bt ,
1 λ 1 λ
j
where Ŷt is j’s disposable (net of taxes) income. Savers hold all bonds for next period; because bonds
are liquid a fraction 1 λ s of the payoff, including interest, accrues to next period’s hand-to-mouth.
In Appendix D, I derive the steady-state demand for liquidity-bonds, or savings function:
2 3
1 4 1 1 5,
B= (28)
1 R 1 1 s
1+ 1
1 1 λ 1 + (1 λ) (Γ 1)
λ βR 1 s

under log utility σ = 1, normalizing Y = 1; (28) is an increasing convex function of R under standard
restrictions. The Appendix explores this analytically in detail, but is is worth noticing that as βR ! 1
debt becomes proportional to (1 λ) (Γ 1) and thus tends to zero with no steady-state income
inequality Γ = 1. The condition for a positive liquidity demand B > 0 is:

1 s
>1 β; (29)
λ
this is strictly true when βR ! 1 and translates into more general restrictions on 1 s, λ, β, see ap-
pendix D. Essentially, (29) requires some idiosyncratic risk and liquidity; it is violated e.g. in TANK.
One intuitive interpretation is that positive long-run liquidity, self-insurance savings requires that
the present discounted value of the risk of becoming constrained (the infinite discounted sum of
1 s at rate β) be larger than the unconditional, long-run probability of being constrained.

4.2 Liquidity, Inequality, and Intertemporal MPCs in THANK

This version embeds a distinct amplification channel orthogonal to the NK Cross, the "intertemporal
Keynesian cross" of Auclert et al (2018) (see also Hagedorn et al (2018)), and allows a novel analytical
solution for their key summary statistics, the intertemporal MPCs (iMPCs). Loglinearizing (27)
around a zero-liquidity steady state with R = β 1 delivers (see Appendix D):

11 s
ctH = ŷtH + β bt , (30)
λ
1 s
cSt + bt+1 = ŷSt + β 1
bt ,
1 λ 1 λ

where bt is in shares of steady-state Y. Aggregating (30), we have:

1
ct = ŷt + β bt bt + 1 . (31)

The iMPCs are the partial derivatives of aggregate consumption ct with respect to aggregate dis-
posable income ŷt+k at different horizons k, keeping fixed everything else (i.e. taxes and public

19
debt). To find them, we solve for equilibrium liquidity bt replacing (30) in the self-insurance equa-
tion (11). The general case is analyzed in Appendix D, but the intuition is clearest in the oscillating
case s = 0, with agents saving when they expect lower income and vice versa:

ŷSt Et ŷtH+1 (2 χ) ŷt χEt ŷt+1

bt + 1 = = . (32)
1 1
2 1+β 2 1+β

The consumption function follows by substituting this into (31), delivering Proposition 5.

2 χ + βχ 2 χ βχ
ct = ŷt + ŷ + ŷ ; (33)
2 (1 + β ) 2 (1 + β ) t 1
2 (1 + β ) t +1

Proposition 5 The iMPCs for THANK with s = 0 in response to a time-T disposable income shock are:

dc T 2 χ + βχ dc T +1 2 χ dc βχ dct
= ; = ; T 1 = ; = 0 for any t < T 1 or t > T + 1.
dŷ T 2 (1 + β ) dŷ T 2 (1 + β) dŷ T 2 (1 + β) dŷ T

j
This illustrates the key points transparently. With acyclical inequality χ = 1 (ŷt = ŷt , Auclert
et al’s case) a current income shock induces agents to self-insure, saving in liquidity to maintain
higher future consumption. While a future shock makes them consume in anticipation, depleting liq-
uid savings. The second point concerns adding cyclical inequality: higher income cyclicality when
constrained χ makes agents consume more out of news and less out of past and current aggregate
income. When self-insuring, agents take into account how the aggregate shock affects income in
each state and change their asset demand and equilibrium liquidity consequently. Even this sim-
plest s = 0 case can then match the two key iMPCs matched by Auclert et al, the contemporaneous
dc0 /dŷ0 = 0.55 and one-year-after dc1 /dŷ0 = 0.15 with β = 0.95 annually and χ = 1.47.

0.6 iMPC(t,0)

0.5

0.4

0.3

0.2

0.1

0.0
1 2 3 4 5
year (t)

Figure 2: iMPCs in THANK (blue solid); TANK (red dash); Data (dots)

The expressions for THANK with s > 0 are still analytical and convey the same intuition, but
are more tedious (see Proposition 9, Appendix D). Figure 2 plots the iMPCs for THANK, along with
TANK and the data from Fagereng et al. In THANK, I match the two target MPCs with λ = 0.33,

20
s = 0.82 (0.96 quarterly) and χ = 1.4.25 The intertemporal path is remarkably in line with both
the data and Auclert et al’s quantitative HANK: the effect dies off a few years after, whereas TANK
misses this intertemporal amplification altogether, the iMPCs being:

dc T dct
= λχ + (1 λχ) (1 β) βT ; = (1 λχ) (1 β) β T 8t 6= T.
dŷ T dŷ T

4.3 Determinacy with Liquidity

I now provide two determinacy results that are analytical counterparts of results in the quantitative
literature: a criterion based on iMPCs as in Auclert et al; and a debt quantity rule à la Hagedorn.
My HANK Taylor principle in Proposition 1 is intimately related to subsequent results developed
for quantitative HANK by Auclert et al (2019) showing that the Taylor principle is sufficient when
the sum of iMPCs is larger than 1. The intuition is that if the total MPC out of an income shock far
into the future is larger than 1, the model is "explosive" (stable forward) and thus determinate. The
connection can be clearly seen by summing up the iMPCs in Proposition 5 to obtain:

1 β
µimpc = 1 + (1 χ) > 1 iff χ < 1. (34)
1+β

There is determinacy (the Taylor principle is sufficient) with procyclical inequality and indetermi-
nacy otherwise. This holds in the general model, as I prove analytically in Appendix D.
In an incomplete-markets economy, a further route to determinacy was discovered by Hagedorn
(2020). If the government chooses the quantity of nominal debt, the price level can be determined
without an interest-rate rule. Consider a rule that chooses nominal debt and sets nominal taxes so
as to balance the budget intertemporally for any price level, thus making policy passive-Ricardian
and ruling out fiscal theory. In this arrangement, the central bank sets freely the nominal interest
rate that clears the liquid-bond market, with no need to respond to any endogenous variable.
Given the steady-state demand for bonds (28), determinacy of the steady-state price level is im-
mediate: the proof is exactly as in Hagedorn (2020).26 More interesting is that local determinacy
under a debt quantity rule holds in my model too, as shown in the next Proposition.

Proposition 6 A nominal debt rule: The THANK model with a well-defined demand for liquid bonds ((29)
holds) leads to local determinacy even when δ > 1 under the nominal-debt quantity rule:

BtN+1 = B N fixed ! btN+1 = 0. (35)

Intuitively, condition (29) requires that H agents receive a fraction of savings that is larger than
the interest income, β 1 1 λ s > β 1 1 = r; more generally, it requires that in steady state H agents
receive positive net income from liquidity, see Appendix D.
The proposition generalizes to btN+1 = φb pt with φb < 1 so that real debt bt+1 = (φb 1) pt falls
when the price level increases. Furthermore, it can be easily shown using the same proof structure
25 This is coincidentally close to the calibration in Bilbiie (2020) matching general-equilibrium statistics with the zero-liquidity
model. Figure D1 in the Appendix provides a comparison of different calibrations, and iMPCs in response to future shocks.
See Cantore and Freund (2021) for a subsequent, simpler analytical iMPC calculation with portfolio adjustment costs.
26 In a nutshell, monetary policy chooses steady-state i, which given π determines R and thus the steady-state real B. The

fiscal authority’s choice of nominal B N (and its growth rate) then immediately determines P (and steady-state π).

21
as for the Wicksellian rule that it holds with forward-looking Phillips curve and that it rules out the
FG puzzle even in the "amplification" region, thus sidestepping the Catch-22. See Hagedorn (2020)
for further discussion and a more general version of those arguments, and Hagedorn et al (2018) for
an analysis of the implications for fiscal multipliers and the earliest illustration of how a quantitative
HANK model with this policy rule sidesteps the Catch-22.

5 Optimal Policy in THANK

THANK is also useful for studying optimal monetary policy analytically, in the version without
liquidity. This provides a benchmark that helps elucidate some key mechanisms operating in the
rich-heterogeneity quantitative-HANK studies featuring several additional relevant channels, such
as Bhandari et al (2021). I build on Woodford’s (2003, Ch. 6) analysis in RANK. In Appendix E,
I spell out the full Ramsey problem and derive a linear-quadratic problem equivalent to it under
certain conditions, taking a second-order approximation to aggregate welfare around a flexible-
price equilibrium that is efficient. The target equilibrium of the central bank is the socially-desirable,
perfect-insurance equilibrium induced by a fiscal policy generating zero profits to first order under
flex prices, following the TANK analysis in Bilbiie (2008, Proposition 5). This delivers Proposition 7.
Proposition 7 Solving the welfare maximization problem is equivalent to solving:
8 9
>
> >
>
1 ∞< =
min E0 ∑ βt π 2t + αy y2t + αγ γ2t , (36)
fct ,π t g 2 >
> | {z } | {z } >
>
t =0 : ;
RANK inequality-THANK
s.t. (7),(13), and (14),

where the optimal weights on output and inequality stabilization are, respectively:

σ 1+ϕ 1 1
αy ; αγ λ (1 λ) σ ϕ αy .
ψ

Several results are worth emphasizing. While the weight on output (gap) stabilization αy is the
same as in RANK, there is an additional term pertaining to income inequality.27 This evidently affects
the central bank’s stabilization tradeoff, introducing a redistribution motive. However, idiosyncratic
risk and its cyclicality are irrelevant for optimal policy, insofar as the target flexible-price equilibrium
is the first-best with perfect insurance, without inequality; thus, the aggregate implications of the
distributional channel for optimal policy in THANK happen to be the same as in TANK.
This is different from Challe (2020), which abstracts from inequality altogether but where an iso-
morphism occurs between RANK and a different analytical HANK with cyclical risk through search
and matching. The common point is that my framework too features irrelevance of income risk, but
in THANK relative to TANK. Both the optimal allocation and the interest rate policy instrument are
radically different, and depend crucially on the cyclicality of inequality here. Furthermore, as we
shall see, in my framework cyclical risk does not matter for implementation either.
27 Other studies found additional stabilization motives using different TANK extensions, e.g. Nistico (2016) and Curdia

and Woodford (2016) for a financial-stability motive, and Bilbiie and Ragot (2016) for a liquidity-insurance motive with an
imperfect-insurance target equilibrium giving rise to a linear term in the approximation.

22
 An important observation concerns the interest rate, which is both residual to the policy problem
and unaffected by risk cyclicality. Recalling that we approximate around the efficient equilibrium
Γ = 1, the IS curve (14) is not a constraint: as in RANK, it just determines it once we found the
optimal allocation (yt , π t ). And since the IS curve approximated around Γ = 1 is also independent
of cyclical risk, so will the interest rate that implements optimal policy.28
Consider for simplicity only shocks that drive no wedge between inequality and aggregate output
gap, which stay proportional: (13) holds; the analysis of shocks that do drive a wedge is relevant for
capturing further mechanisms in richer HANK, but beyond the scope of this paper and pursued in
follow-up work. We can simplify the problem by replacing (13), obtaining the per-period loss:

λ
π 2t + αy2t , with α αy 1 + σ 1
ϕ 1
(χ 1)2 (37)
1 λ

The inequality motive thus amounts, in my benchmark THANK relative to RANK, to a higher
weight on output stabilization that increases with λ. Importantly, this holds regardless of whether
inequality is counter- or pro-cyclical, as long as it is cyclical: the extra stabilization motive is pro-
portional to (χ 1)2 . The simple intuition is based, as in TANK, on the key role of profits which are
eroded by inflation volatility. With higher λ, less agents receive profits; the weight on inflation falls,
and vanishes in the λ ! 1 limit, where there is no rationale for stabilizing profit income.
We can now study optimal policy in THANK starting with discretion, or Markov-perfect equi-
librium. The ability to study this more realistic, time-consistent policy is an appealing feature of
the tractable framework, since computing Markov-perfect optimal policies in quantitative models is
cumbersome. This is obtained by solving (36) assuming that the central bank lacks commitment and
treats expectations parametrically, without internalizing its actions’ effect on them; this amounts to
re-optimizing every period subject to (7) with fixed expectations at the decision time t. The problem
being mathematically identical to RANK, we go directly to the solution:

α
πt = yt . (38)
κ
This targeting rule under discretion requires engineering an aggregate demand decrease for a given
inflation increase. Assuming AR(1) cost-push shocks Et ut+1 = µut , the equilibrium is:

α κ
π dt = ut ; ydt = ut . (39)
κ2 + α (1 βµ) κ2 + α (1 βµ)

Optimal policy under discretion implies that both output and inflation deviate from target: a trade-
off between inflation and output stabilization. Since α is increasing in λ, it follows directly that
optimal policy in THANK requires greater inflation volatility and lower output volatility than in RANK.
One instrument rule implementing this equilibrium is found by using the aggregate IS (14):

1 κ (1 λχ)
it = φd Et π t+1 , with φd 1+ µ δ .
ασ (1 λ)
28 Thisis no longer the case—and risk then matters—in the model with liquidity, where the interest rate has direct distribu-
tional consequences and thus novel interactions with fiscal policy that are left to future work.

23
Unlike in RANK, the instrument rule implementing optimal policy may be passive φd < 1 with
enough compounding δ > µ 1 , i.e. with countercyclical enough inequality: optimal policy requires
a real rate cut in THANK when in RANK it would require an increase. Whereas with procyclical
inequality δ < 1, the required instrument rule is even more active than in RANK. However, it is
independent of the cyclicality of risk in this benchmark.
Optimal (timeless-) commitment policy, the time-inconsistent Ramsey equilibrium, requires com-
mitting to the different targeting rule, by similar arguments as in RANK (Woodford, 2003, Ch. 7):

α
πt = (yt yt 1) . (40)
κ
It is straightforward to show that commitment to (40) delivers determinacy regardless of hetero-
geneity. The difference from RANK is still captured by the inequality motive shaping α, but optimal
commitment policy still amounts to price-level targeting, like in RANK.

6 Conclusions

THANK, a tractable HANK model with two types and two assets, captures analytically several key
channels of quantitative HANK models. I use it for a full analysis of the main themes of the NK lit-
erature of the past decades: determinacy properties of interest rate rules, amplification, multipliers,
resolving the forward guidance puzzle, and optimal monetary policy.
The key channel is cyclical inequality: whether the income of constrained hand-to-mouth agents
comoves more or less with aggregate income. This channel is already the main focus of TANK in
Bilbiie (2008), but interacts with idiosyncratic uncertainty and self-insurance in THANK, as in quan-
titative HANK models. Procyclical inequality delivers discounting in the aggregate Euler equation,
which makes the Taylor principle not necessary for determinacy and can cure the forward guidance
puzzle. Conversely, countercyclical inequality generates Euler-equation compounding, making the
Taylor principle insufficient for determinacy and aggravating the puzzle. This is a Catch-22, for
countercyclicality is precisely the condition for amplification or multipliers in HANK models, which
is what many studies focus on, exploiting a New Keynesian cross inherent therein.
The paper proposes a decomposition of cyclical variations in income risk into one component re-
lated to cyclical income inequality, and one due to cyclical skewness—variations in the probability
of the bad, constrained state. The Catch-22 can be resolved in theory if the latter channel is procycli-
cal enough, as it delivers Euler-discounting without mitigating amplification. As discussed in text,
available evidence does not speak directly to this decomposition—the theory developed here can
in fact be viewed as potentially informing such measurement. However, existing empirical stud-
ies reviewed in text seems to support the view that both (inequality and pure-risk) channels are
countercyclical, case in which determinacy conditions become very stringent and the puzzle is ag-
gravated. Nevertheless, I show that a Wicksellian rule of price-level targeting resolves this tension
by making THANK determinate and puzzle-free, even with countercyclical inequality and risk. This
virtue is shared by a rule setting nominal debt proposed by Hagedorn (2020), as I show analytically
in my model’s version with liquidity.
Optimal monetary policy, solved for analytically in THANK, requires a separate inequality ob-
jective, in addition to stabilizing inflation and real activity around an efficient perfect-insurance

24
equilibrium. Regardless of risk, optimal policy implies tolerating more inflation volatility as a result
of distributional concerns when inequality is cyclical. While timeless-optimal commitment policy
ultimately still amounts to price-level targeting, even though along the adjustment path there it still
entails tolerating more inflation.
It is conceivable that for the analysis of many important macroeconomic questions the tractable
HANK framework proposed here, THANK, is sufficient and one does not always need a full-heterogeneity
model; the latter is certainly needed for many important questions, e.g. for identifying the most rel-
evant micro heterogeneity dimensions. To date and to the best of my knowledge, THANK is the
only tractable framework, among the several reviewed, to capture all of these channels found to be
key in rich-heterogeneity models: cyclical income inequality, precautionary self-insurance saving,
intertemporal marginal propensities to consume, and features of idiosyncratic income uncertainty
and risk (cyclical variance and skewness, and kurtosis).
As models of the economy as a whole become larger and more complex, with many sectors,
frictions, and sources of heterogeneity, the quest for tractable representations seems important for
entropic reasons. It is my hope that this framework is thus useful for policymakers and central
banks, for communicating to the larger public, for students and colleague economists from other
fields seeking to enter the fascinating realm of macro stabilization policy in a world where hetero-
geneity and inequality are of the essence.

References

Acharya, S. and K. Dogra, 2020 "Understanding HANK: Insights from a PRANK," Econometrica

Acharya, S., E. Challe, and K. Dogra, 2020 "Optimal Monetary Policy According to HANK," Mimeo

Aiyagari, R., 1994, “Uninsured Idiosyncratic Risk and Aggregate Saving,” Quarterly Journal of Economics, 109,
659–684.

Angeletos, G.M. and C. Lian, 2017, Forward Guidance without Common Knowledge, American Economic Review

Ascari, G., A. Colciago and L. Rossi, 2017, "Limited Asset Market Participation, Sticky Wages, and Monetary
Policy", Economic Inquiry

Auclert, A. 2019, "Monetary Policy and the Redistribution Channel," American Economic Review

Auclert, A. and M. Rognlie 2016, "Inequality and Aggregate Demand," Mimeo

Auclert, A., M. Rognlie, and L. Straub 2018, "The Intertemporal Keynesian Cross," Mimeo

Bayer, C., R. Luetticke, L. Pham-Dao, and V. Tjaden, 2019, "Precautionary Savings, Illiquid Assets, and the Aggre-
gate Consequences of Shocks to Household Income Risk," Econometrica

Benigno, P. and G. Benigno, 2003, "Price stability in open economies," Review of Economic Studies 70 (4), 743-764

Berger, D., L. Bocola, and A. Dovis, 2020, "Imperfect Risk Sharing and the Business Cycle," Mimeo

Bernstein, J., 2021, "A Model of State-Dependent Monetary Policy," Journal of Monetary Economics, 117, 904-917

Bewley, T., 1977, “The Permanent Income Hypothesis: A Theoretical Formulation,” Journal of Economic Theory,
16, 252–292.

25
Bhandari, A., D. Evans, M. Golosov, and T. Sargent, 2021, "Inequality, Business Cycles and Monetary-Fiscal Policy",
Econometrica

Bilbiie F. 2008. "Limited Asset Market Participation, Monetary Policy, and (Inverted) Aggregate Demand Logic".
Journal of Economic Theory 140, 162-196.

Bilbiie, F, 2011, " Non-Separable Preferences, Frisch Labor Supply and the Consumption Multiplier of Government
Spending: One Solution to a Fiscal Policy Puzzle," Journal of Money, Credit and Banking

Bilbiie F. and R. Straub, 2013, "Asset Market Participation, Monetary Policy Rules and the Great Inflation", Review
of Economics and Statistics, 95(2), 377-392

Bilbiie F., Monacelli T. and R. Perotti 2013, “Public Debt and Redistribution with Borrowing Constraints", The
Economic Journal, F64-F98

Bilbiie, F. 2019, "Optimal Forward Guidance", American Economic Journal: Macroeconomics

Bilbiie, F. and X. Ragot, 2021, "Optimal Monetary Policy and Liquidity with Heterogeneous Households", Review
of Economic Dynamics

Bilbiie, F. 2020, "The New Keynesian Cross," Journal of Monetary Economics

Bilbiie, F. 2018, "Neo-Fisherian Policies and Liquidity Traps," American Economic Journal: Macroeconomics

Broer, T., N. J. Hansen, P. Krusell, and E. Oberg, 2020 "The New Keynesian transmission mechanism: a heteroge-
neous agent perspective," Review of Economic Studies

Campbell, J. Y. and Mankiw, N. G. 1989. ”Consumption, Income, and Interest Rates: Reinterpreting the Time Series
Evidence,” NBER Macroeconomics Annual 185-216

Cantore, C. and L. Freund, 2019. “Workers, Capitalists, and the Government: Fiscal Policy and Income
(Re)Distribution.” Journal of Monetary Economics

Caramp, M. and D. Silva, 2021, "Monetary Policy and Wealth Effects: The Role of Risk and Heterogeneity," Mimeo

Carroll, C. and M. Kimball, 1996, “On the Concavity of the Consumption Function,” Econometrica, 64(4), 981–992

Challe, E. 2020. "Uninsured Unemployment Risk and Optimal Monetary Policy in a Zero-Liquidity Economy."
American Economic Journal: Macroeconomics, 12 (2): 241-83

Challe, E. and X. Ragot, 2011 “Fiscal Policy in a Tractable Liquidity-Constrained Economy,” The Economic Journal

Challe, E. and X. Ragot, 2016 “Precautionary Saving Over the Business Cycle,” The Economic Journal

Challe, E, J. Matheron, X. Ragot and J. Rubio-Ramirez, 2017, "“Precautionary Saving and Aggregate Demand",
Quantitative Economics

Cloyne, J., C. Ferreira and Surico, P., 2015 "Monetary Policy when Households have Debt: New Evidence on the
Transmission Mechanism" Review of Economic Studies

Cochrane, J. 2011 "Determinacy and Identification with Taylor Rules,"Journal of Political Economy, 119 (3), pp.
565-615

Colciago, A. 2011, Rule-of-thumb consumers meet sticky wages, Journal of Money Credit and Banking 43(2),
325-353.

Curdia, V. and M. Woodford 2016, “Credit Frictions and Optimal Monetary Policy,” Journal of Monetary Eco-
nomics

26
Debortoli, D., and J. Galí, 2018, "Monetary Policy with Heterogeneous Agents: Insights from TANK models";
Mimeo

Del Negro, M., M. Giannoni, and C. Patterson 2012: “The Forward Guidance Puzzle,” Mimeo

Den Haan, W. J.; P. Rendahl, and M. Riegler 2016, "Unemployment (Fears) and Deflationary Spirals" Journal of the
European Economic Association

Diba, B. and O. Loisel, 2017, "Pegging the Interest Rate on Bank Reserves: A Resolution of New Keynesian Puzzles
and Paradoxes," Mimeo

Eggertsson, G. 2010, "What Fiscal Policy Is Effective at Zero Interest Rates?" NBER Macroeconomics Annual

Eggertsson G. and P. Krugman 2012, "Debt, Deleveraging, and the Liquidity Trap: A Fisher-Minsky-Koo Ap-
proach," Quarterly Journal of Economics, 127(3): 1469-1513.

Fagereng, A., M. Holm and G. Natvik, 2018, “MPC Heterogeneity and Household Balance Sheets" American
Economic Journal: Macroeconomics

Farhi, E. and I. Werning, 2017, "Monetary policy with bounded rationality and incomplete markets" American
Economic Review

Ferrière, A. and G. Navarro, 2017 "The Heterogeneous Effects of Government Spending: It’s All About Taxes,"
Mimeo

Gabaix, X., 2019, "A Behavioral New Keynesian Model," American Economic Review

Galí, J. 2008, "Monetary Policy, Inflation, and the Business Cycle: An Introduction to the New Keynesian Frame-
work and Its Applications", Princeton University Press

Galí, J., D. López-Salido, and J. Vallés, 2007. "Understanding the Effects of Government Spending on Consump-
tion". Journal of the European Economic Association, March, vol. 5 (1), 227-270.

Garcia-Schmidt, M. and M. Woodford, 2014. “Are Low Interest Rates Deflationary? A Paradox of Perfect Foresight
Analysis,” American Economic Review

Gornemann, N., K. Kuester, and M. Nakajima, 2016, "Doves for the Rich, Hawks for the Poor? Distributional
Consequences of Monetary Policy," Mimeo

Guerrieri, V. and G. Lorenzoni, 2017, "Credit Crises, Precautionary Savings, and the Liquidity Trap" Quarterly
Journal of Economics

Guvenen, F., S. Ozkan, and J. Song, “The Nature of Countercyclical Income Risk,” Journal of Political Economy,
122 (3):621-660.

Hagedorn, M., 2020, "A Demand Theory of the Price Level", Mimeo

Hagedorn, M., 2018, “Prices and Inflation when Government Bonds are Net Wealth,” mimeo

Hagedorn, M., I. Manovskii, and K. Mitman, 2018, "The Fiscal Multiplier," Mimeo

Hagedorn, M., J. Luo, I. Manovskii, and K. Mitman, 2019, "Forward Guidance," Journal of Monetary Economics

Heathcote, J., F. Perri, and G. Violante, 2010. "Unequal We Stand: An Empirical Analysis of Economic Inequality
in the United States: 1967-2006," Review of Economic Dynamics

Heathcote, J. and F. Perri, 2018, "Wealth and Volatility," Review of Economic Studies

27
Holm, M., 2021, "Monetary Policy Transmission with Income Risk," Mimeo

Huggett, M., 1993, “The Risk-Free Rate in Heterogeneous-Agent Incomplete-Insurance Economies,” Journal of
Economic Dynamics and Control, 17, 953–969.

Kaplan, G., G. Violante, and J. Weidner, 2014 “The Wealthy Hand-to-Mouth,” Brookings Papers on Economic
Activity.

Kaplan, G., B. Moll and G. Violante, 2018, "Monetary Policy According to HANK", American Economic Review

Krusell, P., T. Mukoyama, and A. Smith. 2011. "Asset Prices in a Huggett Economy," Journal of Economic Theory,
146 (3), 812-844

Leeper, E., 1991, "Equilibria Under "Active" and "Passive" Monetary and Fiscal Policies", Journal of Monetary
Economics

Lucas R. E. Jr., 1990 "Liquidity and Interest Rates." Journal of Economic Theory 50, 237 - 264.

Luetticke, R., 2021 "Transmission of Monetary Policy with Heterogeneity in Household Portfolios," American
Economic Journal: Macroeconomics

Mankiw, N. G. 1986. “The Equity Premium and the Concentration of Aggregate Shocks.”Journal of Financial
Economics, 17: 211–19

Mankiw, N. G. 2000. ”The Savers-Spenders Theory of Fiscal Policy,” American Economic Review, 90 (2): 120-125.

McKay, A. and R. Reis, 2016 “The Role of Automatic Stabilizers in the U.S. Business Cycle,” Econometrica, 84-1

McKay, A., E. Nakamura, and J. Steinsson 2015: “The Power of Forward Guidance Revisited,” American Economic
Review

McKay, A., E. Nakamura and J. Steinsson 2016: "The Discounted Euler Equation", Economica

Michaillat, P. and E. Saez, 2021, "A New Keynesian Model with Wealth in the Utility Function", Review of Eco-
nomics and Statistics

Nistico, S., 2016, "Optimal monetary policy and financial stability in a non-Ricardian economy." Journal of the
European Economic Association

Nuño, G. and C. Thomas, 2017, "Optimal Monetary Policy with Heterogeneous Agents" Mimeo

Oh, H. and R. Reis, 2012, "Targeted transfers and the fiscal response to the great recession." Journal of Monetary
Economics, 59(S), S50-S64

Patterson, C., 2019, "The Matching Multiplier and the Amplification of Recessions," Mimeo

Ravn, M. and V. Sterk, 2017 “Job Uncertainty and Deep Recessions,” Journal of Monetary Economics.

Ravn, M. and V. Sterk, 2020, "Macroeconomic Fluctuations with HANK and SAM: an analytical approach," Journal
of the European Economic Association

Sargent, T. and N. Wallace, 1975, "Rational" Expectations, the Optimal Monetary Instrument, and the Optimal
Money Supply Rule," Journal of Political Economy, 83(2), 241-254

Storesletten, K., C. Telmer, and A. Yaron, 2004. “Cyclical Dynamics in Idiosyncratic Labor Market Risk,” Journal
of Political Economy, 112(3), 695-717.

28
Walsh, C., 2016, "Workers, Capitalists, Wages, Employment and Welfare", Mimeo

Werning, I., 2015, "Incomplete markets and aggregate demand", Mimeo

Woodford, M., 1990, "Public Debt as Private Liquidity," American Economic Review Papers and Proceedings, 80
(2), pp. 382-388

Woodford, M., 1996, "Controlling the public debt: a requirement for price stability?" Mimeo

Woodford, M., 2003, Interest and prices: foundations of a theory of monetary policy, Princeton University Press

Woodford, M. 2011. "Simple Analytics of the Government Expenditure Multiplier," American Economic Journal:
Macroeconomics, 3(1), p 1-35.

29
 Appendix to Monetary Policy and Heterogeneity: An Analytical Framework
Florin O. Bilbiie, University of Lausanne and CEPR

A Model and Literature Details

This Appendix presents in detail the model and reviews the connection to the literature.

A.1 Relation to Literature: Details

Relation to other analytical HANK Others studies also provide different analytical frameworks, both
because they isolate different HANK mechanisms and focus on different questions. The clearest separa-
tion in terms of channels is illustrated by the subsequent paper by Acharya and Dogra (2020) reviewed
in text, that is explicitly set to isolate cyclical risk using CARA preferences. That paper shows that indeed
intertemporal amplification may occur purely as a result of uninsurable income volatility going up in re-
cessions, even when inequality is acyclical. (the paper also studies determinacy and puzzles referring to
this paper’s results from the previous version.) In a previous contribution, Werning (2015) similarly em-
phasizes the possibility of AD amplification/dampening of monetary policy relative to RANK in a more
general model of income risk and market incompleteness where inequality and risk coexist. My paper’s
subject is very different, a full analysis of NK topics. So is the mechanism, although some of its equi-
librium implications pertaining to intertemporal amplification or dampening have a similar flavor. But
the key here is cyclical inequality: the distribution of income (between labor and "capital" understood as
monopoly profits) and how it depends on aggregate income, as summarized through χ, the chief feature
of my earlier TANK model Bilbiie (2008). Whereas the discussion in Werning emphasizes the cyclicality
of income risk: as uninsurable income risk goes up in a recession, agents increase their precautionary
savings and decrease their consumption, amplifying the initial recession which further increases idio-
syncratic risk, and so on—a mechanism previously emphasized through endogenous unemployment
risk by Ravn and Sterk (2020) and Challe et al (2020). My model’s mechanism is instead an intertempo-
ral extension of the cornerstone amplification (dampening) mechanism in TANK, when any agent can
become constrained in any future period and self-insures (imperfectly) using liquid assets against the
(acyclical) risk of doing so. This puts the cyclicality of income of constrained, and thus of inequality, at the
core of transmission; whereas Werning emphasizes the cyclicality of income risk, although the two are
convoluted in the different, more general framework therein.
To incorporate this distinction, I embed a separate cyclical-risk channel in THANK, assuming that the
probability of becoming constrained is a function of aggregate output. With this different formalization,
the two different channels of cyclical inequality and risk jointly determine AD amplification. Not only
are the two channels naturally separate: my analysis implies that they better be distinct, for in order to
resolve the Catch-22 they need to go in opposite directions. Which channel prevails empirically is a very
interesting and hitherto unexplored topic that I pursue currently.
Additionally, my analysis is conducted in a loglinearized NK model that nests not only the three-
equation textbook RANK but also: TANK, a HANK with cyclical inequality and acyclical risk, and a
HANK with cyclical risk and acyclical inequality. Since it is so simple and transparent and close to
standard NK craft, it may be of independent interest to some researchers.
My results imply an analytical reinterpretation of McKay et al’s (2016, 2017) incomplete-markets
based resolution of the FG puzzle. My framework underscores the procyclicality of inequality as sufficient
for delivering Euler-equation discounting in the presence of (albeit acyclical) idiosyncratic risk. Procycli-
cality of inequality occurs in my model through labor market features and fiscal redistribution making
the income of constrained agents vary less than one-to-one with the cycle χ < 1. If inequality is instead
countercyclical, the Euler equation is compounded in my model, implying an aggravation of the FG puzzle.
Furthermore, my paper addresses a wide range of NK topics as mentioned above.

A-1
 Broer, Hansen, Krusell, and Oberg (2020) study a simplified HANK whose equilibrium has a two-
agent representation, underscoring the implausibility of some of the model’s implications for monetary
transmission through income effects of profit variations on labor supply—and showing that a sticky-
wage version features a more realistic transmission mechanism; Walsh (2017) provides another analyt-
ical model with heterogeneity emphasizing the role of sticky wages (see Colciago (2011), Ascari, Colci-
ago, and Rossi (2017), and Furlanetto (2011) for earlier sticky-wage TANK).
Auclert, Rognlie, and Straub (2018) also use a "Keynesian cross" version to capture a distinct, comple-
mentary HANK channel. In particular, they abstract from the cyclical-inequality channel emphasized
here to focus on the role of liquidity in the form of public debt; they unveil key summary statistics per-
taining to the marginal propensities to consume out of past and future income (labelled iMPCs) and how
they shape the responses of the economy to past and future income shocks. Their quantitative HANK
model with liquid and illiquid assets can in fact be viewed as the closest generalization of my THANK
model; or alternatively, it is among the wide spectrum of quantitative HANK models the one to which
my THANK model is the closest reduced representation. I use THANK to calculate analytically Auclert
et al’s iMPCs and provide insights into the important propagation mechanism they emphasize. Indeed,
self-insurance to idiosyncratic risk is necessary and sufficient in the presence of liquidity to generate
the tent-shaped path of iMPCs in THANK; whereas cyclical inequality is not of the essence to generate
persistent iMPCs, but is important to fit the magnitudes under realistic calibrations.
Ravn and Sterk (2020) also study an analytical HANK but with search and matching (SaM), that is
different from and complementary to my model and focusing on a different (sub)set of the issues stud-
ied here; Challe (2020) studies optimal monetary policy therein. Their models include endogenous unem-
ployment risk through labor SaM, risk against which workers self-insure. The simplifying assumptions
used to maintain tractability, in particular pertaining to the asset market, are orthogonal to mine.29 Their
framework delivers an interesting feedback loop from precautionary saving to aggregate demand (see
also Challe et al (2017)). My benchmark model does much the opposite: in the zero-liquidity case, it
gains tractability assuming exogenous transitions and a different asset market structure, but emphasizes
the NK-cross feedback loop through the endogenous constrained income that is absent in Ravn and Sterk
and Challe. While my extension to cyclical risk can be viewed as a reduced-form formalization of their
channel. This paper addresses additional topics: forward guidance and the FG puzzle, (restoring deter-
minacy under a peg), the Catch-22 and a way out of it, the virtues of a Wicksellian rule of price-level
targeting, the version with liquidity, and optimal monetary policy.
Relation to Bilbiie (2020) and (2008) The THANK model proposed here is an extension of the TANK
model in Bilbiie (2008), which analyzed monetary policy introducing the distinction between the two
types based on asset markets participation:30 H have no assets, while S own all the assets, i.e. price bonds
and shares in firms through their Euler equation. That paper analyzed AD amplification of monetary
policy and emphasized the key role of profits and their distribution, as well as of fiscal redistribution, for
this in an analytical 3-equation TANK model isomorphic to RANK. In recent work, Bilbiie (2020) and
Debortoli and Galí (2018) both used this TANK model to argue that it can approximate reasonably well
some aggregate implications of some HANK models: several models from the HANK literature cited
above, for the former; and the authors’ own, for the latter. This suggests that the cyclical-inequality
channel plays an important role in HANK transmission in and of itself.
The first extension here pertains to introducing self-insurance to idiosyncratic uncertainty: the risk
of becoming constrained in the future despite not being constrained today, a key HANK mechanism
29 Inmy model savers hold, price, and receive the payoff (profits) of shares. In Ravn and Sterk, hand-to-mouth workers
get the return on shares but do not price them. Their mechanism creates an "unemployment-trap", a breakup of the Taylor
principle complementary to the one here, and fixes the puzzling NK effects of supply shocks in a LT, which I abstract from.
30 Thus abstracting from physical investment, the element of distinction in previous two-agent studies: Mankiw (2000) had

used a growth model with this distinction, due to pioneerig work by Campbell and Mankiw (1989), to analyze long-run fiscal
policy issues. Galí, Lopez-Salido and Valles (2007) embedded this same distinction in a NK model and studied numerically the
business-cycle effects of government spending, with a focus on obtaining a positive multiplier on private consumption. They
also analyzed numerically determinacy properties of interest rate rules, that Bilbiie (2008) derived analytically.

A-2
that is absent in TANK; this gives the model another margin to fit the aggregate findings of quantitative
HANK, as shown in Bilbiie (2020).31 That paper introduces the New Keynesian Cross as a graphi-
cal and analytical apparatus for the AD side of HANK models, expressing its key objects—MPC and
multipliers—as functions of heterogeneity parameters. It studies the implications for monetary and
fiscal multipliers, the link between MPC and multipliers with the "direct-indirect" decomposition of Ka-
plan et al, and the ability of this simple model to replicate some aggregate equilibrium implications of
several quantitative, micro-calibrated HANK models. Finally, Bilbiie and Ragot (2021) builds a different
analytical HANK with three assets—one ("money") liquid and traded in equilibrium, two (bonds and
stock) illiquid—and studies Ramsey-optimal monetary policy as liquidity provision.
This paper’s novel elements include: adding cyclical risk from several sources, related or unrelated
to inequality, and pertaining to either variance or skewness; liquidity and a calculation of the iMPCs; an
aggregate supply side and closed-form conditions for determinacy with Taylor rules (the HANK Taylor
principle), for determinacy under price-level targeting, and for ruling out the forward-guidance puzzle;
a formal statement of the "Catch-22" and of the conditions on the cyclicalities of risk and inequality to
rule it out; an analysis of optimal monetary policy.

A.2 Aggregate Supply: New Keynesian Phillips Curve

All households consume an aggregate basket of individual goods k 2 [0, 1], with constant elastic-
R1 (ε 1)/ε
ε/(ε 1)
ity of substitution ε > 1: Ct = 0
Ct ( k ) dk yielding the standard demand Ct (k ) =
R 1 1
( Pt (k) /Pt ) Ct and aggregate price index Pt1 ε = 0 Pt (k) ε dk. Good are produced by monopolis-
ε

tic firms using labor: Yt (k ) = Nt (k ), with real marginal cost Wt .

The profit function is: Dt (k ) = 1 + τ S [ Pt (k )/Pt ] Yt (k ) Wt Nt (k) TtF .
The individual goods producers solve:
" #
∞ 2
ψ Pt (k )
maxE0 ∑ Q0,t 1 + τ Pt (k )Yt (k ) Wt Nt (k )
S S
1 Pt Yt ,
Pt (k) t =0 2 Pt 1

where I consider two possibilities for the reference price level Pt 1 , with respect to which it is costly for
firms to deviate. In the first scenario, this is the aggregate price index Pt 1 which small atomistic firms
take as given—this delivers the static Phillips curve. In the second, Pt 1 is firm k’s own individual price
1
βt P0 C0S /Pt CtS
S σ
as in standard formulations. Q0,t is the marginal rate of intertemporal substitution
S
of participants between times 0 and t, and τ the sales subsidy. Firms face demand for their products
from two sources: consumers and firms themselves (in order to pay for the adjustment cost); the demand
function for the output of firms z is Yt (k ) = ( Pt (k )/Pt ) ε Yt . Substituting this into the profit function, the
first-order condition is, after simplifying, for each case:
Static PC case Pt 1 = Pt 1
" #
ε 1
Pt (k ) S Wt Pt (k ) Pt (k ) 1
0 = Q0,t Yt 1+τ (1 ε) + ε Q0,t ψPt Yt 1
Pt Pt Pt Pt 1 Pt 1

In a symmetric equilibrium all producers make identical choices (including Pt (k ) = Pt ); defining net
inflation π t ( Pt /Pt 1 ) 1, this becomes:

ε 1 ε
π t (1 + π t ) = wt 1 + τS ,
ψ ε 1
31 Thatpaper also discusses the differences with earlier work using type-switching to analyze monetary policy, e.g. Nistico
(2016) and Curdia and Woodford (2016). I spell out the differentiating assumptions below.

A-3
loglinearization of which delivers the static PC in text (8).32
Dynamic PC case Pt 1 = Pt 1 ; the first-order condition is

" #
ε 1
Pt (k) S Wt Pt (k )
0 = Q0,t Yt 1+τ (1 ε) + ε
Pt Pt Pt
Pt (k ) 1 Pt+1 (k ) Pt+1 (k)
Q0,t ψPt Yt 1 + Et Q0,t+1 ψPt+1 Yt+1 1
Pt 1 (k ) Pt 1 (k) Pt (k ) Pt (k )2

In a symmetric equilibrium, using again the definition of net inflation π t , and noticing that Q0,t+1 =
1
σ 1
Q0,t β CtS /CtS+1 (1 + π t +1 ) , this becomes:

CtS 1/σ Yt ε 1
π t (1 + π t ) = βEt [( S
) π t+1 (1 + π t+1 )] + wt ,
Ct+1 Yt+1 ψ m

1 ε
with the post-subsidy markup m 1+ τ S ε 1
. Loglinearizing this delivers the NKPC in text (7); notice
that this nests the static PC when the discount factor of firms β = 0.
I assume as a benchmark that the government implements the standard NK optimal subsidy induc-
ing marginal cost pricing m = 1, i.e. τ S = (ε 1) 1 . Financing its total cost by taxing firms TtF = τ S Yt
gives total profits Dt = Yt Wt Nt . This policy is redistributive: since steady-state profits are zero D = 0,
it taxes the firms’ shareholders and results in the "full-insurance" steady-state used here as a benchmark
C H = C S = C. Loglinearizing around this and denoting dt ln ( Dt /Y ) we have dt = wt : profits vary
inversely with the real wage an extreme form of the general property of NK models.

B Income Processes, Inequality, and Cyclical Idiosyncratic Risk

B.1 Inequality, Gini Coefficient, and Generalized Entropy

This section discusses the relationship between our measure of inequality Γt and the more standard
measures: first, Gini coefficient, and then generalized entropy.
The income Gini with two levels is given by

(1 λ) YtS YtS
Φt = (1 λ ) = (1 λ) 1
Yt Yt
and is between 0 and λ (when S get all income). Rewrite it using our measure as

Γt (1 λ ) ( Γ t 1)
Φ t = (1 λ) 1 =λ
λ + (1 λ ) Γ t 1 + (1 λ ) ( Γ t 1)

and conversely Γt = 1 + (λ ΦΦt )(t 1 λ)

. Using the log-deviation of inequality γt Γt Γ
Γ = ySt ytH we have
the log-deviation of the Gini:

YS S λ (1 λ ) Γ
υ t = (1 λ) yt yt = γ,
Y λ + (1 λ ) Γ t

which around a symmetric SS simplifies to υt = λ (1 λ) γt .

A generalized entropy measure (with largest sensitivity to small incomes) is:
32 In
a Calvo setup, this amounts to assuming that each period a fraction of firms f keep their price fixed, while the rest can
re-optimize freely but ignoring that this price affects future demand. This reduces to β f = 0 only in the firms’ problem (not
recognizing that today’s reset price prevails with some probability in future periods).

A-4
 YtH YtS
Ξt = λ ln (1 λ) ln
Yt Yt
Subtracting the steady-state value of this same measure we obtain the deviation (note, in a uniform
steady-state this measure is zero so we express this deviation in levels)

YtH Y YtS Y
ξ t = Ξt Ξ= λ ln (1 λ) ln
Y H Yt Yt Y S
YH YS
= yt λytH (1 λ) ySt = λ 1 ytH + (1 λ) 1 ySt
Y Y
YS YH Γ 1 Γ 1
= λ (1 λ) γ t = λ (1 λ) γt = υt .
Y λ + (1 λ ) Γ Γ

B.2 Income processes

Higher moments of the income process are readily calculated in my model, since individual income fol-
lows a two-state Markov chain with values YtS and YtH in the respective states. The analytical character-
ization of this process’ key moments is useful both to illustrate a key dimension along which this model
is a representation of complex HANK models, and for calibration and quantitative analysis. (Note that
this is essentially just an analytical two-state version of the Rouwenhorst method.)
Conditional variance is found using Et ln YtS+1 j ln YtS = s (Yt+1 ) ln YtS+1 + (1 s (Yt+1 )) ln YtH+1 as:
!2
YtS+1
var ln YtS+1 j ln YtS = s (Yt+1 ) (1 s (Yt+1 )) ln = s (Yt+1 ) (1 s (Yt+1 )) (ln Γt+1 )2 . (B.1)
YtH+1

Conditional skewness and kurtosis are easily calculated as:

1 2s (Yt+1 )
skew ln YtS+1 j ln YtS = p ; (B.2)
s (Yt+1 ) (1 s (Yt+1 ))
1
kurt ln YtS+1 j ln YtS = [s (Yt+1 ) (1 s (Yt+1 ))] 3.

The first-order autocorrelation of the income process for any of the two states j = S, H:

j j 1 s
corr ln Yt+1 , ln Yt = s+h 1=1 . (B.3)
λ
As standard for Bernoulli distributions there is negative skewness for s > .5 and leptokurtosis (positive
excess kurtosis kurt (.) 3) outside of the 21 p1 interval, i.e. for s smaller than 0.21 or larger than 0.79.
12
Notice that s > 0.79 ensures both negative skewness and leptokurtosis, with s 1 h ensuring positive
autocorrelation.
Of special importance to fit key micro facts on income distribution in the cross-section are the relative
skewness and kurtosis of the two types: evidence in e.g. Guvenen et al (2014) suggests that the income
of an empirical proxy of S is relatively more negatively skewed and more leptokurtic. It can be easily
shown, comparing (B.2) with the equivalent formulae for H that both properties are satisfied in the
model if and only if: s > h. This simple two-state model features, albeit in a stylized way, some key
elements of the literature pertaining to income heterogeneity and uncertainty: conditional idiosyncratic
variance that can be cyclical, autocorrelated income processes with left-skewness and leptokurtosis. The
combined conditions for matching the key micro facts are s > 1 h, s > h and both s and h larger than
.79. We use this when calibrating the model in text.

A-5
 The cyclicality of skewness is:
d (skew) sY
= 3 (B.4)
dY 2 [ s (1 s)] 2
and is entirely determined by the cyclicality of the probability to become constrained. When the proba-
bility to become constrained 1 s is increasing in recessions, sY < 0, risk (in the Mankiw and Guvenen
et al sense) is countercyclical: negative skewness becomes more negative in recessions, making upward
income movements less likely and downward income movements more likely therein. Notice that this
does not depend on the size of income inequality.
THANK thus captures analytically persistent and conditionally volatile idiosyncratic income, and
also, albeit in a stylized way given the coarse two-state implicit discretization, the key feature of con-
comitant left-skewness and leptokurtosis that a discretization with more states matches very well. Using
(B.2), the conditional skewness and excess kurtosis are, for the calibration used in text to match iMPCs,
1.66 and 0.77 respectively; while the quarterly autocorrelation is (s + h 1)1/4 = (1 (1 s) /λ)1/4 =
0.819 (corresponding to the quarterly transition probability 1 s = 1 0.821/4 = 1 0.952 ' .04). Given
the coarse two-state discretization, it is no surprise that these moments are not perfectly aligned with
the micro data.

B.3 Cyclical inequality: derivations

Take first the hand-to-mouth, who consume all their income and loglinearize the budget constraint: ctH =
D
ytH = wt + ntH + τλ dt . Substituting the wage schedule derived using the economy resource constraint,
production function, and aggregate labor supply wt = ϕ + σ 1 ct ; the profit function dt = wt ; and
their labor supply, we obtain H 0 s consumption function given in text: ctH = ytH = χyt , with

τD
χ 1+ϕ 1 7 1.
λ

Cyclical distributional effects make χ different from 1. The other agents, S, with income ySt = wt +
D
nSt + 11 τλ dt , face an additional (relative to RANK) income effect of the real wage, which reduces their
profits dt = wt . Using this and their labor supply, we obtain: cSt = 11 λχ λ yt , so whenever χ < 1 S’s
income elasticity to aggregate income is larger than one, and vice versa.
In RANK, such distributional considerations are absent since one agent works and receives all the
profits. When aggregate income goes up, labor demand goes up and the real wage increases. This
drives down profits (wage=marginal cost), but because the same agent incurs both the labor gain and
the "capital" (monopolistic rents) loss, the distribution of income between the two is neutral.
Income distribution matters under heterogeneity; to understand how, start with no fiscal redistrib-
ution, τ D = 0 and χ > 1. If demand goes up and, with upward-sloping labor supply ϕ > 0, the wage
goes up, H’s income increases. Their demand increases proportionally, as they do not get hit by profits
falling. Thus aggregate demand increases by more than initially, shifting labor demand and increasing
the wage even further, and so on. In the new equilibrium, the extra demand is produced by S, whose de-
cision to work more is optimal given the income loss from falling profits. Since the income of H goes up
and down more than proportionally with aggregate income, inequality is countercyclical: it goes down
in expansions and up in recessions.
Redistribution τ D > 0 dampens this channel, lowering χ. Through the transfer, H start internaliz-
ing the negative income effect of profits, and increase demand by less. The benchmark considered by
Campbell and Mankiw’s (1989) seminal paper is χ = 1, which occurs when the distribution of profits is
uniform τ D = λ (the income effect disappears) or when labor is infinitely elastic ϕ = 0 (all households’
consumption comoves perfectly with the wage); income inequality is then acyclical.
Finally, χ < 1 occurs when H receive a disproportionate share of the profits τ D > λ. The AD

A-6
expansion is now smaller than the initial impulse, as H recognize that this will lead to a fall in their
income; while S, given the positive income effect from profits, optimally work less. As the income of H
now moves less than proportionally with aggregate income, inequality is procyclical.

B.4 Cyclical risk: derivations

The self-insurance equation when the probability depends on aggregate demand (tomorrow) is
1 1 1
1 + it
CtS s (Yt+1 ) CtS+1 s (Yt+1 )) CtH+1
σ σ σ
= βEt + (1 . (B.5)
1 + π t +1

We loglinearize this around a steady-state with inequality; in the context of our model, that requires
assuming that steady-state fiscal redistribution is imperfect and that a sales subsidy does not completely
undo market power (generating zero profits). In particular, we focus on a steady state with no subsidy,
so that the profit share is D/C = 1/ε and the labor share W N/C = (ε 1) /ε. Under the same arbitrary
redistribution scheme, the consumption shares of each type are respectively

τD
CH WN + λ D 1 τD
= =1 1
C C ε λ
τD
CS W N + 11 λ D 1 λ τD CH
= = 1+ 1 > iff τ D < λ.
C C ε1 λ λ C

CS
Denoting steady-state inequality CH
Γ we loglinearize around a steady state:
h 1
i
1 = β ( 1 + r ) s (Y ) + ( 1 s (Y )) Γ σ , (B.6)

where I restrict attention to cases with positive real interest-rate r (the topic of "secular stagnation" in this
framework is interesting in its own right—it can occur for high enough risk and high enough inequality).
Loglinearization delivers, denoting by rt the ex-ante real interest rate for brevity, and the steady-state
s 0 (Y )Y
value of the probability by s (C ) = s and its elasticity relative to the cycle (consumption) is 1 s(Y ) :

s (1 s) Γ1/σ s 0 (Y ) Y σ (1 s) 1 Γ1/σ
cSt = σ (it Et π t+1 ) + E cS +
1/σ t t+1
Et ctH+1 + Et ct+1
s + (1 s) Γ s + (1 s) Γ1/σ 1 s (Y ) s + (1 s) Γ1/σ

which replacing individual consumption levels as function of aggregate becomes

1 λ 1 s̃ s 0 (Y ) Y σ (1 λ )
cSt = σ (it Et π t+1 ) + 1 + (χ 1) (1 s̃) 1 Γ 1/σ
Et ct+1
1 λχ 1 λχ 1 s (Y ) 1 λχ

(1 s)Γ1/σ
denote by 1 s̃ = s+(1 s)Γ1/σ
>1 s the inequality-weighted transition probability, the relevant inequality-
adjusted measure of risk given steady-state inequality coming from financial income Γ Y S /Y H 1.
Γ1/σ (χ 1)
There can be discounting as long as risk is procyclical enough η > . But the contemporary
σ(1 λ)(Γ1/σ 1)
AD elasticity to interest rates is unaffected by the cyclicality of risk (this is thus isomorphic to Acharya
and Dogra’s different formalization of cyclical risk based on CARA utility).

B.5 Current aggregate demand

For the case where the probability depends on current (today) aggregate demand s (Yt ), the aggregate
Euler-IS is

A-7
 1 s 0 (Y ) Y 1
cSt = σ (it Et π t+1 ) + β (1 + r ) sEt cSt+1 + β (1 + r ) (1 s) Γ σ Et ctH+1 + σβ (1 + r ) (1 s) 1 Γ σ ct
1 s (Y )

Replacing β (1 + r )
1
s (1 s) Γ σ
1
s 0 (Y ) Y σ (1 s) 1 Γσ
cSt = σ (it Et π t+1 ) + 1 Et cSt+1 + 1 Et ctH+1 + 1 ct
s + (1 s) Γ σ s + (1 s) Γ σ 1 s (Y ) s + (1 s) Γ σ

Replace the consumption functions of H and S we obtain:

1 λ
ct = θδEt ct+1 θσ (it Et π t+1 ) (B.7)
1 λχ
1
s 0 (Y ) Y 1 λ
with θ 1+ 1 Γ 1/σ
(1 s̃) σ ,
1 s (Y ) 1 λχ

where the notation is as previously. Notice that now the two channels (cyclical inequality and cyclical
risk via s0 (.)) are intertwined for both the amplification/dampening of current interest rates and for
future consumption. A previous working paper version contained a full analysis of this version of the
model and its implications for curing puzzles and the Catch-22.

C The 3-equation THANK with NKPC

This section derives the same results as in text but with the forward-looking NKPC (7).

C.1 The HANK Taylor Principle: Equilibrium Determinacy with Interest Rate Rules
Determinacy can be studied by standard techniques, extending the result in text (there will now be
two eigenvalues). Necessary and sufficient conditions are provided i.a. in Woodford (2003) Proposition
0 0
C.1. With the Taylor rule (17), the system becomes Et π t+1 Et ct+1 = A π t ct with transition
matrix: " #
β 1 β 1κ
A=
δ 1 σ 11 λχ
λ
φπ β 1 δ 1 1 + σ 11 λχ
λ
β 1κ

with determinant det A = β 1 δ 1 1 + κσ 11 λχ λ

φπ and trace trA = β 1 + δ 1 1 + σ 11 λχ λ
β 1κ .
Determinacy can obtain in either of two cases. Case 2. (det A trA < 1 and det A+trA < 1) can
be ruled based on sign restrictions. Case 1. requires three conditions to be satisfied jointly:

det A > 1; det A trA > 1; det A + trA > 1

The third condition is always satisfied under the sign restrictions, so the necessary and sufficient condi-
tions are:
( δ 1) (1 β )
φπ > 1 + (C.1)
κσ 11 λχ
λ

βδ 1 (1 β)(δ 1)
together with φπ > max κσ 11 λχ
λ , 1+ κσ 11 λχ
λ . The second term is larger than the first iff (2β 1) δ <
κσ 11 λχ
λ
+ β, which holds generically for most plausible parameterizations. Condition (C.1) thus gener-
alizes the HANK Taylor principle to the case of forward-looking Phillips curve.

A-8
C.2 Ruling out the FG Puzzle
The analogous of Proposition 2 for the case with NKPC (7) is:

Proposition 8 The analytical HANK model (with (7)) under a peg is locally determinate and solves the FG puzzle
2
( ∂ ct < 0) if and only if:
∂( it+T )∂T
1 λ κ
δ+σ < 1,
1 λχ 1 β

Notice that the condition nests the one of Proposition 2 when β ! 0. Indeed, it has exactly the same
interpretation with δ + σ 11 λχ
λ κ κ
1 β being the "long-run" effect of news, and 1 β being the slope of the
long-run NKPC.
Point 1. (determinacy under a peg with NKPC) follows directly from (C.1): a peg is sufficient if
(1 β)(δ 1)
both δ < β 1 and 1 + κσ 1 λ < 0, the latter implying δ < 1 1 κ β σ 11 λχ
λ
< β 1 , which delivers the
1 λχ
threshold in the Proposition.
Point 2 requires solving the model; focusing therefore on the case where the condition holds, and the
model is determinate under a peg, we rewrite the model in forward (matrix) form as:

πt 1 Et π t+1 1 λ κ
=A σ it (C.2)
ct Et ct+1 1 λχ 1

where !
1 β + κσ 11 λχ
λ
κδ
A =
σ 11 λχ
λ
δ
0
is the inverse of matrix A defined above under a peg φ = 0. To find the elasticity of π t ct to
T κ
an interest rate cut at T, it+T we iterate forward (C.2) to obtain σ 11 λχλ
A 1 . But notice that
1
we know by point 1 that the eigenvalues of A are both outside the unit circle; it follows by standard
T
linear algebra results that the eigenvalues of A 1 are both inside the unit circle and therefore A 1
is decreasing with T. (the eigenvalues to the power of T appear in the Jordan decomposition used to
compute the power of A 1 ). This proves that the FG puzzle is eliminated.
Point 3 requires computing the equilibrium given an AR1 interest rate with persistence µ as before
Et it+1 = µit ; since we are in the determinate case, the equilibrium is unique and there is no endogenous
persistence, so the persistence of endogenous variables is equal to the persistence of the exogenous
process. Replacing Et ct+1 = µct and Et π t+1 = µπ t in (C.2) we therefore have:

πt 1 λ 1
1 κ
= σ I µA it .
ct 1 λχ 1

Computing the inverse we obtain

" #
1
1 1 1 δµ κδµ
I µA = ,
det σ 11 λ
λχ µ 1 β + σ 11 λ
λχ κ µ

where det µ2 βδ µ δ + σ 11 λ
λχ κ + β µ + 1. Replacing in the previous equation, differentiating, and
simplifying, the effects are:
∂π t
!
∂it 1 λ 1 κ
= σ
∂ct
∂it
1 λχ det 1 µβ

A-9
Therefore, neo-Fisherian effects are ruled out iff det > 0, i.e.:

1 βµ σ 11 λ
λχ κµ
δ< .
µ (1 βµ)

But this is always satisfied under the condition in the proposition (for determinacy under a peg) δ < 1
σ 11 λχ
λ
κ 1 βµ σ 11 λχ
λ
κµ
1 β µ(1 βµ)
where the second inequality can be easily verified, it implies (1 βµ) (1 β) +
βσκµ 0.
Figure C1 illustrates the threshold level of endogenous redistribution sufficient to deliver determi-
nacy under a peg and thus rule out the FG puzzle, as a function of λ and for different 1 s. Close to the
TANK limit (small 1 s), no level of redistribution delivers this (red dash); as idiosyncratic risk 1 s
increases (blue solid), the region expands and is largest in the iid case (blue dots).

1.0 τmin
D

0.8

0.6

0.4

0.2
λ
0.0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

D in TANK 1
Fig. C1: Redistribution threshold τ min s ! 0 (dash); 0.04 (solid); λ (dots).
Note: The crosses represent the threshold above which the IS slope is positive λχ < 1.

C.3 Ruling out puzzles with Wicksellian rule and Contemporaneous PC

Replacing (8) and the policy rule (24) in the aggregate Euler-IS (16) we have

1 λ
ct = ν0 Et ct+1 σ φ p pt + it ; (C.3)
1 λχ
the static PC rewritten in terms of the price level is:

pt pt 1 = κct . (C.4)

Combining, we obtain:

1 λ 1 λ
Et pt+1 1 + ν0 1 1 + σ φ κ p t + ν0 1 p t 1 =σ κν 1 i . (C.5)
1 λχ p 1 λχ 0 t

Notice that the RANK model is nested here for λ = 0 (or χ = 1, the Campbell-Mankiw benchmark),
which would yield a simplified version of Woodford and Giannoni’s analyses.
Recall that we are interested in the case whereby ν0 1 (as the paper shows, for ν0 < 1 there there is
determinacy under a peg in HANK and thus no puzzles). The model has a locally unique equilibrium
(is determinate) when the above second-order equation h has one root inside and one
i outside the unit
1
circle. The characteristic polynomial is J ( x ) = x2 1 + ( ν0 ) 1 + σ 11 λ
λχ φ p κ x + ν0 1 where by

A-10
standard results, the roots’ sum is 1 + ν0 1 1 + σ 11 λ
λχ φ p κ and the product is ν0 1 < 1. So at least one
root is inside the unit circle, and we need to rule out that both are; Since we have J (1) = ν0 1 σ 11 λ
λχ φ p κ
and J ( 1) = 2 + 2ν0 1 + ν0 1 σ 11 λχ
λ
φ p κ, the necessary and sufficient condition for the second root to be
outside the unit circle is precisely φ p > 0—coming from J (1) < 0 and J ( 1) > 0. This completes the
proof of Proposition 4.
To find the solution, denote the roots of the polynomial by x+ > 1 > x > 0; the difference equation
is solved by standard factorization: The roots of the characteristic polynomial are
rh i 2
1 + ν0 1 1 + σ 11 λ
λχ φ p κ 1 + ν0 1 1 + σ 11 λ
λχ φ p κ 4ν0 1
x =
2
x+ > 1 > x > 0

Factorizing the difference equation (25):

1 1 1 λ
L x L x+ pt 1 =σ κν 1 i
1 λχ 0 t

we obtain:
1 λ 1
pt = x pt 1 σ κν 1 x 1 i
1 λχ 0 + 1 ( x+ L) 1 t

∞
1 λ
κν0 1 x+1 ∑ x+ it+ j
j
= x pt 1 σ
1 λχ j =0

Let ∆t+ j σ 11 λ 1 1
λχ κν0 x+ it+ j denote the rescaled interest rate cut:
" #
∞ ∞ ∞ ∞
pt = x t +1
p 1 + ∑ ( x+ ) j
∆t+ j + x ∑ ( x+ ) j
∆t 1+ j + ... + x t 1
∑ ( x+ ) j
∆ 1+ j + x t
∑ ( x+ ) j
∆j
j =0 j =0 j =0 j =0

Normalizing initial value to zero (since x < 1 it vanishes when t goes to infinity), the solution is made
of a forward and a backward component:
t +1 t k
1 x x+1 ∞ j t 1 1 x x+1
pt =
1 x x+1
∑ x+1 ∆t+ j + ∑ x 1+ k 1 x x+1
∆t 1 k
j =0 k =0

Lagging it once and taking the first difference we obtain the solution for inflation:
t +1 t
1 x x+1 ∞ j 1 x x+1 ∞ j
πt =
1 x x+ 1 ∑ x+1 ∆t+ j
1 x x+ 1 ∑ x+1 ∆t 1+ j
j =0 j =0
t k t 1 k
t 1 1 x x+1 t 2 1 x x+1
+ ∑ x 1+ k 1
∆t 1 k ∑ x 1+ k ∆t 2 k
k =0 1 x x+ k =0 1 x x+1
∞ j
= A (t ) ∑ x+1 ∆t+ j + Ψt 1.
j =0

A-11
 t +1 t +1
1 ( x+1 )+(x )t ( x+1 ) (x x+1 )
where A (t) 1 x x+1
(if we put ourselves at time 0 this simply becomes A (0) =
σ 11 λχ
λ
ν0 1 ), while in Ψt 1 we grouped all terms that consist of lags of ∆t (∆t 1 and earlier) which are pre-
determined at time t and will not be used in any of the derivations of interest here—where we consider
shocks occurring at t or thereafter. This delivers, for consumption:
∞ j +1
ct = A (t) Et ∑ x+1 it+ j + Ψt 1 (C.6)
j =0

where Ψt 1 is a weighted sum of past realizations of the shock and A (t) > 0 is a function only of
calendar date; both Ψt 1 and A (t) are irrelevant for our purpose because they are invariant to current
and future shocks.
The effect of a one-time interest rate cut at t + T is now:
∂ct T +1
= A (t ) x+1
∂ it+ T

which, since A (.) > 0 and x+ > 1, is a decreasing function of T: the FG puzzle disappears.33 Notice that
the Wicksellian rule also cures the FG puzzle in the (nested) RANK model (this follows immediately by
replacing λ = 0 or χ = 1 above).

C.4 Determinacy with Wicksellian rule and NKPC

Rewrite the system made of (14), (7) and the definition of inflation as (ignoring shocks):

1 λ 1 λ
ct = δEt ct+1 φ p pt + σ
σ Et π t+1
1 λχ 1 λχ
πt = βEt π t+1 + κct
pt = π t + pt 1

0 0
Substituting and writing in canonical matrix form Et ct+1 Et π t+1 pt = A ct π t pt 1 with
transition matrix A given by
0 1
δ 1 1 + β 1 σ 11 λχ
λ
κ δ 1 σ 11 λχ
λ
φp β 1 δ 1
σ 11 λ
λχ φ p
B C
A=@ β 1κ β 1 0 A.
0 1 1

We can apply Proposition C.2 in Woodford (2003, Appendix C): determinacy requires two roots outside
the unit circle and one inside. The characteristic equation of matrix A is:

J ( x ) = x 3 + A2 x 2 + A1 x + A0 = 0
33 Likewise for neo-Fisherian effects: take an AR(1) process for i with persistence µ as before; the solution is now both 1.
t
uniquely determined (by virtue of determinacy proved above) and 2. in line with standard logic—an increase in interest rates
∂ct
leads to a fall in consumption and deflation in the short run: ∂i = A (t) x+1 µ ,which is negative as A (.) > 0 and x+ > 1 > µ.
t
Notice that in the long-run, i.e. if there is a permanent change in interest rates, the economy moves to a new steady-state and
the uncontroversial. long-run Fisher effect applies as usual.

A-12
with coefficients:
1 σκ 1 λ1
A2 = +1 1<0
β β 1 λχδ
1 1 σκ 1 λ 1
A1 = + 1 + φp + 1 + >0
β δ β 1 λχ β
1
A0 =
βδ

To check the determinacy conditions, we first calculate:

1 σκ 1 λ
J (1) = 1 + A2 + A1 + A0 = φ >0
δ β 1 λχ p
J ( 1) = 1 + A2 A1 + A0
" #
2 1 σ 11 λχ
λ
κ σ 11 λχ
λ
κ 2
= 2 2 + φp + 2 + <0
β δ β β β

Since J (1) > 0 and J ( 1) < 0 we are either in case Case II or Case III in Woodford Proposition C.2;
Case III in Woodford implies that φ p > 0 is sufficient for determinacy if the additional condition is
satisfied:
σ 11 λχ
λ
κ+β
A2 < 3 ! δ < . (C.7)
2β 1
This is a fortiori satisfied in RANK (and delivers determinacy there), but not here with δ > 1. Therefore,
we also need to check Case II in Woodford and to that end we need to check the additional requirement
(C.15) therein:
A20 A0 A2 + A1 1 > 0,
which replacing the expressions for the Ai s delivers:

(1 β) (δ 1) + σ 11 λ
λχ κ
φp > (1 δβ)
σ 11 λ
λχ κδβ

Since the ratio is positive, this requirement is only stronger than the already assumed φ p > 0 when

1
δ<β ; (C.8)

It can be easily checked that the δ threshold C.8 is always smaller than the threshold C.7; therefore,
whenever δ < β 1 , Case III applies and φ p > 0 is sufficient for determinacy. While when C.7 fails (for
large enough δ), Case II applies and φ p > 0 is still sufficient for determinacy.

A-13
D Liquidity in THANK

D.1 Model derivations

The two budget constraints (after asset market clearing) for the case with liquidity (27), evaluated at the
steady state deliver, respectively, with Y j denoting income of agent j:

1 s
CH = YH + 1 R+1 B (D.1)
λ
λ 1 s
CS = YS 1 R + 1 B, (D.2)
1 λ λ

where we already imposed the steady-state budget T = ( R 1) B and assumed uniform taxation. The
steady-state self-insurance Euler equation for bonds is:
2 1
3
C S σ
1 = βR 4s + (1 s) 5 (D.3)
CH

To derive the steady-state savings function (demand for liquidity), replace individual consumptions in
the Euler equation
2 ! σ1 3
S λ 1 s
Y 1 R+1 B
1 = βR 4s + (1 s) 1 λ
1 s
λ 5,
H
Y + 1 R + 1 B
λ

which rewritten and under the particularly tractable case of log utility σ ! 1 becomes:
" !#
1 s Y
1 = βR 1 + 1
1 λ YH + 1λ s 1 R + 1 B

Using SS inequality Γ = Y S /Y H to write:

YH 1 YS Γ
= ;=
Y 1 + (1 λ) (Γ 1) Y 1 + (1 λ) (Γ 1)

we finally obtain (normalizing Y = 1 wlog) (28) in text:

1 1
1+ 11 λ 1
1 1+(1 λ)(Γ 1)
s βR
B= 1 s
.
λ 1 R+1

Notice that SS inequality is determined in the model: using symmetric SS hours and the expression
1
for the real wage w = m 1 where m is the steady-state markup with arbitrary subsidy m ε
1+ τ S ε 1
,
τD 1 τD m 1 1 τD
we have Y H = W N H + λ D = mN + λ m Y = m 1+ λ (m 1) Y and for savers Y S = W N S +
1 τD 1 1 τD
1 λ D = m 1+ 1 λ (m 1) Y. Replacing both of these:

1 τD
1+ 1 λ (m 1)
Γ= D ,
1 + τλ (m 1)

so there is SS inequality Γ > 1 iff m > 1 and τ D < λ.

Start by noticing that to have a self-insurance motive C S > C H the standard condition from incomplete-

A-14
market models applies:
1
R<
β
When is SS debt positive? The denominator being positive requires:

1
R< 1 s
, (D.4)
1 λ

which essentially implies, see (D.1), that H get positive income from liquidity, net of taxes. A positive
numerator instead requires:
1
R> . (D.5)
β [1 + (1 s) (Γ 1)]
For an equilibrium interest rate to exist we thus need

1 1 1 s
< ! >1 β [1 + (1 s) (Γ 1)]
β [1 + (1 s) (Γ 1)] 1 1 s
λ
λ

1
With large enough Γ, always satisfied as RHS negative, β 1 < (1 s) (Γ 1)
The combined condition is thus:
!
1 1 1
< R < min , ,
β [1 + (1 s ) ( Γ 1)] β 1 1λ s

which in the zero-liquidity (no SS inequality) case used in the text is:
1
1 1 s
R=β < 1 .
λ

1 (1 λ)(Γ 1)
Several properties are worth noticing. As βR ! 1 debt tends to B ! B̄ =
1+ β ( 1 λ s 1) 1+(1 λ)(Γ 1)
1

which is > 0 iff 1 λ s > 1 β. Moreover, B ! 0 when R ! β 1

(1 + (1 s) (Γ 1)) 1 . Thus, in the
"bondless", zero-liquidity limit we have
R β 1,
with equality in the no-SS-inequality case Γ = 1.
As standard in incomplete-market models, the demand for debt features and asymptote. In particu-
lar, B ! ∞ when ( )
1
1 1 s
R ! min β 1 , 1 .
1 11 λs λ
1
First, at given numerator debt becomes infinite whenever 1 + R 1 λ s 1 ! 0 so R ! 1 1 λ s >
1
β under the positive-debt requirement. Second, at given denominator, B can also become infinite
1 1 λ 1 1 1
when 1 + βR 1 1 s ! 0 so R ! β 1 1 s >β (assuming s > λ). The former threshold is smaller
1 λ
than the latter whenever 1 1 λ s < β 1 11 λs . Notice, however, that the threshold is always larger
than β 1 .
Loglinearization. The loglinearized budget constraint of H agents (27) is:

CH H YH H 1 s 1 s
ct = y tt + Rbt + RBY rt , (D.6)
Y Y t λ λ

A-15
where bt ( Bt B) /Y.

D.2 Derivation of analytical iMPCs

For analytical convenience, for this section I derive results loglinearizing around a long-run steady-state
with zero public debt (and thus zero liquidity) B = 0, implying R = β 1 and no consumption inequality
C H = C S , so (D.6) and the corresponding equation for S agents become (30), where we imposed asset
market clearing.
The equilibrium dynamics of private liquid assets bt are found by replacing these individual budget
j j
constraints (30) into the loglinearized self-insurance Euler equation for bonds (11), with ŷt yt tt
denoting disposable income as in text, obtaining:

1 1 λh i
Et bt+2 Θbt+1 + β bt = sEt ŷSt+1 + (1 s) Et ŷtH+1 ŷSt , (D.7)
s
1 1 1 s 1 s
where Θ +β 1+ 1 .
s s λ

As clear from (D.7), finding the derivatives of bt+k with respect to ŷt requires a model of how individ-
ual disposable incomes are related to aggregate, such as this paper’s. Furthermore, since the calculation
of iMPCs keeps fixed by definition all the other variables (in particular taxes, their distribution, and thus
public debt), the partial derivatives of individual disposable incomes with respect to aggregate dispos-
able income are respectively dŷtH = χdŷt and dŷSt = 11 λχ λ d ŷt .
34 Solving the asset dynamics equation

taking this into account delivers:

∞
1 λχ
dbt+1 = xb dbt +
s ∑ ( βxb )k+1 (dŷt+k δdŷt+k+1 ) , (D.8)
k =0
q
1 1
where the roots of the characteristic polynomial of (D.7) are xb = 1
2 Θ Θ2 4β and ( βxb ) ,
with 0 < xb < 1 as required by stability whenever β > 1 1 λ s .
Substituting (D.8) in (31) delivers the aggregate consumption function, the key equation for calculat-
ing the analytical iMPCs in Proposition 9:
∞
1 λχ
dct = dŷt + β 1
(1 βxb ) dbt +
s ∑ ( βxb )k+1 (δdŷt+k+1 dŷt+k ) . (D.9)
k =0

Proposition 9 The intertemporal MPCs (iMPCs) for the THANK model, in response to a one-time shock to
disposable income at any time T and for any t 0: (i) are given by:
8
> 1 λχ δ βxb T t t
>
> s 2 ( βx b ) 1 xb + xb (1 βxb ) βxb2 , if t T 1;
>
< 1 βx b
dct 1 ( βx2 )
T
=
> 1 1 sλχ βxb (δ βxb ) xb 1 sλχ (1 βxb ) 1 βxb2 , if t = T;
dŷ T > b
>
>
: 1 λχ 1 βx2b x t T 1 xb δ + xb (δ βxb ) βxb2 T , if t T + 1.
s 1 βx b b

and (ii) are increasing with the cyclicality of inequality χ when t < T and decreasing with χ when t T >0
(keeping fixed the time-0 contemporaneous MPC dc0 /dŷ0 ).
34 In particular, any model would deliver a reduced-form ŷ H = χ ŷ + χ t , χ
t t tax t tax being an equilibrium elasticity depending
on the tax distribution, labor elasticity, etc. But for calculating iMPCs, we look at a partial equilibrium wherein dtt /ŷt = 0.

A-16
 It is useful, in order to isolate this liquidity-amplification channel, to follow Auclert et al’s paper that
discovered it and start with the benchmark of acyclical inequality χ = 1. This amounts to replacing
j
individual disposable incomes with aggregate disposable income ŷt = ŷt , obtaining the expressions in
Proposition 9 with χ = 1 and δ = 1. The path of the iMPCs is apparent in this special case: faced
with a current income shock, agents optimally self-insure, saving in liquid wealth to maintain a higher
consumption in the future. While when facing a future income shock agents consume in anticipation,
decreasing their stock of liquid savings.
Ceteris paribus, countercyclical inequality χ > 1 leads to a higher contemporaneous MPC but to
lower future MPCs (without affecting persistence as described by xb which is independent of χ). Per-
sistence is instead increasing with the share of hand-to-mouth and decreasing with the level of idiosyn-
cratic risk (it can be directly verified that ∂xb /∂λ > 0 and ∂xb /∂ (1 s) < 0).
Figure D.1 illustrates this by plotting the iMPCs for four models: TANK, and three cases of THANK
(encompassing both liquidity and cyclical inequality) for pro- and counter-cyclical inequality, and the
benchmark acyclical-inequality akin to Auclert et al’s quantitative HANK, respectively. The left panel
looks at a date-0 aggregate income shock and calibrates the THANK with acyclical inequality to closely
follow Auclert et al, i.e. β = 0.8 and λ = 0.5; this requires s = 0.84 to match both the contemporaneous
and next-year MPCs (0.55 and 0.15 respectively). The discount rate is very large, even for the yearly
calibration adopted here; in the models with cyclical inequality (both TANK and THANK) I set β = 0.95
and match the two target MPCs with λ = 0.33, s = 0.82 and χ = 1.4. This is coincidentally close to
the calibration used in Bilbiie (2020) to match other (aggregate, general-equilibrium) objects with the same
model.
The intertemporal path of the iMPCs is remarkably in line with that documented by Fagereng et
al and Auclert et al in the data; in particular, the effect of the income shock dies off a few years after;
whereas the model with acyclical inequality implies unrealistically high persistence while TANK implies
no persistence at all. The reverse side of it is that, as clear from the right panel that compares iMPCs out
of current and future income shocks for THANK with acyclical and countercyclical inequality, the latter
implies larger iMPCs out of future income—an illustration of part (ii) of the Proposition; this is due,
intuitively, to the same self-insurance forces that generate Euler-compounding in general equilibrium
illustrated in the previous section. Direct differentiation of the analytical expressions in Proposition 9
reveals in fact that the iMPCs out of future income (news) are increasing in χ while the iMPCs out of
past income are decreasing in χ.

0.6 iMPC(t,0) 0.6 iMPC(t,T)

0.5 0.5

0.4 0.4

0.3 0.3

0.2 0.2

0.1 0.1

0.0 0.0
2 4 6 8 10 12 14 2 4 6 8 10 12 14 16 18 20
year (t) year (t)

Figure D.1: iMPCs in THANK with χ = 1 (thin black dot-dash); TANK (red dash); THANK with counter- and
pro-cyclical inequality (thick and thin blue solid). Left: T = 0; right: T = 0; 10

An important remark is that countercyclical inequality is, nevertheless, not necessary for the THANK

A-17
model to match the iMPCs. Indeed, the model with procyclical inequality χ < 1 also does it. To illustrate
this, consider the model with χ = 0.8. Clearly, we need to re-calibrate the model for a lower χ implies,
by the logic of the cyclical-inequality channel, a lower contemporaneous MPC and a higher MPC out of
past income; matching the two MPCs thus requires re-calibrating λ = 0.64 and s = 0.74. The resulting
path (the thin solid line in the Figure) illustrates our intuition: the MPC out of past income is virtually
identical, which is not surprising since we matched the one-period-ago MPC. But the whole path of the
"forward" MPCs is below the countercyclical-inequality case (with the acyclical-inequality case between
the two), which is a direct implication of the Euler discounting through δ discussed at length above.
Notice, however, that while discounting/compounding in the Euler equation is not per se of the essence
for matching the iMPCs (although it certainly matters quantitatively), idiosyncratic risk is.

D.3 Proof of Proposition 9

The solution of the asset-accumulation equation implies the following recursions for the responses of
assets to income shocks:
dbt+1 dbt 1 λ
t T = xb
1: + ( βxb )T t
( βxb δ)
dŷ T dŷ T s
dbt+1 dbt 1 λ
t = T: = xb + βxb
dŷ T dŷ T s
dbt+1 dbt
t > T: = xb
dŷ T dŷ T

The solutions of these equations are (setting initial debt equal to steady-state without loss of generality):

dbt+1 1 λχ 1 ( xb βxb )t+1

t T 1: = ( βxb )T t
( βxb δ)
dŷ T s 1 xb βxb
dbT +1 1 λχ 1 ( xb βxb )T 1 λχ
t = T: = xb βxb ( βxb δ) + βxb
dŷ T s 1 xb βxb s
dbt+1 db
t T+1 : = xbt T T +1
dŷ T dŷ T

Taking derivatives of the consumption function D.9, we have:

dct dbt 1 λχ
t T 1: =β 1
(1 βxb ) + ( βxb )T t
(δ βxb )
dŷ T dŷ T s
dct dbt 1 λχ
t = T: = 1 + β 1 (1 βxb ) βxb
dŷ T dŷ T s
dct dbt
t > T: = β 1 (1 βxb )
dŷ T dŷ T

A-18
 Replacing the solution for assets:

dct t +1 1 λχ 1 ( xb βxb )t 1 λχ
t T 1: =β 1
(1 βxb ) ( βxb )T ( βxb δ) + ( βxb )T t
(δ βxb )
dŷ T s 1 xb βxb s
dct 1 1 λχ ( xb βxb )T
1 1 λχ
t = T: = 1+β (1 βxb ) βxb ( βxb δ) βxb
dŷ T s 1 xb βxb s
dct 1 dbt 1 dbT +1
t T+1 : =β (1 βxb ) =β (1 βxb ) xbt T 1
dŷ T dŷ T dŷ T
!
1 1 λχ ( xb βxb )T 1 λχ
1
= β (1 βxb ) xbt T 1
xb βxb ( βxb δ) + βxb
s 1 xb βxb s

Rewriting and simplifying, we obtain the expressions in Proposition 9. Notice that, as argued by
Auclert et al, the present discounted sum of the iMPCs needs to be 1 (the increase in income is consumed
entirely, sooner or later). To prove that the iMPCs in THANK derived here satisfy this property, replace
the respective solution into the sum:
T 1 ∞
dct dc dct
∑ βt T
dŷ T
+ T + ∑ βt
dŷ T t=T +1
T
dŷ T
t =0

obtaining

1 λχ δ βxb T h T
i
β xb 1 βxb2
s 1 βxb2
1 λχ 1 λχ 1 ( xb βxb )T
+1 βxb (δ βxb ) xb (1 βxb )
s s 1 xb βxb
1 λχ βxb T
+ 1 xb δ + xb (δ βxb ) βxb2
s 1 βxb2
= 1
dc0 1 λχ dc1
The calibration in text following Auclert et al concerns two iMPCs, dŷ0 = 1 s βxb and dŷ0 =
1 λχ
s (1 βxb ) xb .
Part (ii) of the Proposition concerns the dependence on χ (and δ Euler-compounding), keeping fixed
the time-0 contemporaneous MPC ddcŷ00 ; denote this by:

dc0 1 λχ
m00 =1 βxb
dŷ0 s

Replacing in the Proposition and rewriting the iMPCs, taking the derivative with respect to the cyclical-
ity of inequality χ we obtain:
8
> 1 m00 δ βxb T t t
>
> βxb 1 βxb2 ( βxb ) 1 xb + xb (1 βxb ) βxb2 , if t T 1;
dct >
<
∂ dŷT ∂ 1 ( βxb2 )
T
jm___00 = 1 m00 1 m00
1 βx b ( δ βx b ) ( 1 βx b ) , if t = T;
∂χ ∂χ >
> βx b β 1 βxb2
>
>
: 1 m00 1 βx2b x t T 1 xb δ + xb (δ βxb ) βxb2 T , if t T + 1.
βxb 1 βx b b

∂δ 1 λ
It follows directly that "anticipation iMPCs" (t < T) are increasing in χ (using = (1 s) > 0);
∂χ (1 λχ)2
T
iMPCs out of past income (t > T) are decreasing in χ (the derivative is proportional to xb 1 βxb2 ∂δ
∂χ <0),

A-19
and decrease the contemporaneous MPC at given T.

D.4 Determinacy and iMPCs

Auclert et al (2019) show that determinacy occurs when the unweighted sum of iMPCs for an income
shock occurring at T ! ∞ is larger than 1. In my model, this object is calculated using the expressions
in Proposition 9.
!
T 1 ∞
dct dc T dct
µimpc = lim ∑
dŷ T t=∑
+ +
t=0 d ŷ T dŷ T
T !∞
T +1

Replacing the expressions in Proposition 9 and taking the limit for T ! ∞ we obtain, for the first term:

1 λχ 1 xb
(δ βxb ) βxb
s (1 xb βxb ) (1 βxb )

for the second term (contemporaneous iMPC):

1 λχ 1 1 λχ
1 + (1 βxb ) xb ( βxb δ) βxb
s 1 xb βxb s

and for the third sum:

1 λχ 1 1 xb δ
(1 βxb ) xb .
s 1 xb 1 xb βxb
Taking the total sum:
" β (1 x b )
#
1 λχ (δ βxb ) (1 xb βxb )(1 βxb )
β
µimpc = 1 + xb 1 βxb
s (δ βxb ) 1 xb βxb + (1 βxb ) 1 1xb 11 xbxβx
bδ
b

1 λχ (1 β ) x b
= 1 + (1 δ) (D.10)
s (1 βxb ) (1 xb )

Thus, the condition for determinacy (and for the Taylor principle to be sufficient) of Auclert et al µimpc >
1 is equivalent to my condition δ < 1.

D.5 Determinacy with a nominal debt quantity rule

In this Appendix, I prove Proposition (6) for the case of static Phillips curve. As we shall see, the algebra
and intuition are both very similar to the proof of determinacy under a Wicksellian rule—and so is the
generalization to forward-looking Phillips curve, which I ignore for brevity.
The loglinearized government budget constraint is (with BY steady-state debt share in output):

bt+1 + tt = R (bt + BY it 1 BY π t ) . (D.11)

Next, notice that one important dimension of Hagedorn’s policy is that taxes adjust automatically to
ensure that the government budget constraint is indeed a constraint for any price level (thus ruling out
fiscal-theory equilibria); that is, tt is endogenous and determined by (D.11) residually
To derive the aggregate Euler equation for the model with liquidity and with steady-state inequality
and non-zero steady-state debt, I adopt the following assumption that simplifies the algebra without
losing generality: to obtain that consumption is equalized in steady-state across agents (while income
is not, and there is positive debt), assume a pure redistributive transfer taken as given by agents (so
it does not preclude bonds demand). Since without the transfer the share of consumption of H in Y

A-20
 1
is C H /Y = [1 + (1 λ) (Γ 1)] , this transfer—an additive term in (D.1)—is trivially equal to 1
h i 1
1
1 + 11 λs βR 1 , while for S agents we have a tax equal to the same amount times λ/ (1 λ). In
YH
addition, let again χ denote the equilibrium elasticity: Y ytH = χyt .
Under these assumptions, the budget constraint of H (D.6) and S become:

1 s 1 s
ctH = χyt tt + Rbt + RBY (it 1 πt )
λ λ
1 λ 1 s 1 s
cSt = ct χyt tt + Rbt + RBY (it 1 πt )
1 λ 1 λ λ λ

Replacing these in the self-insurance Euler equation we obtain (with the usual δ notation):

1 λ
ct = δEt ct+1 (it Et π t+1 )
σ
1 λχ
λ 1 λ s 1 λ s1 s λ 1 s
tt Et tt+1 + Rbt+1 + Rbt
1 λχ 1 λχ 1 λχ λ 1 λχ λ
1 λ s1 s λ 1 s
+ RBY (it Et π t+1 ) + RBY (it 1 π t )
1 λχ λ 1 λχ λ

When is the equilibrium determinate under the debt-quantity rule fixing the nominal amount of debt
proposed by Hagedorn? Replace the rule btN+1 = 0 ! bt+1 = btN+1 pt = pt , taxes from (D.11), the
assumed exogenous interest rates it = it , and the static Phillips curve ct = κ 1 ( pt pt 1 ) to obtain,
rearranging (take log utility wlog):

" 2 #
1 s 1 s
1 λ λ 1 λ 1
δκ 1 + λ
RBY λ
Et pt+1
1 λχ 1 λχ 1 λχ
8 h 2
i 9
< 1 λ λ 1 s
1
2 λ 1 + 1λ s 1 R λ 1 s
1 =
1 λ λ
(1 + δ ) κ + RBY + + RBY pt
: 1 λχ 1 λχ 1 λχ 1 λχ ;
" #
1 λ 1λ s 1 λ 1λ s 1
+ κ R+ RBY pt 1 =
1 λχ 1 λχ
2 !
1 λ λ 1λ s 1 λ 1λ s 1
RBY it RBY it 1
1 λχ 1 λχ 1 λχ

This is a second-order difference equation with characteristic polynomial

J ( x ) = A2 x 2 A1 x + A0 ,

where A2 , A1 , A0 are the coefficients on Et pt+1 , pt , pt 1 . By the same standard results invoked before,
the necessary and sufficient condition for determinacy is J (1) J ( 1) < 0, where

1 s s 1
J (1) = 1+R 1
1 λχ λ
1 λ λ 1 s 1 s
J ( 1) = 2κ 1 (1 + δ) + 2 +2 1 RBY 2
1 λχ 1 λχ λ λ
λ 1 s 1 s
+ 2 1+ 1 R
1 λχ λ λ

A-21
 First, we have J (1) < 0 if:

1 s 1
1+R 1 >0!R< 1 s
,
λ 1 λ

which is always satisfied (see the restrictions needed for the steady-state debt demand). Next, note that
J ( 1) > 0 holds in e.g. the most extreme case of pure oscillation s = 0, as 2κ 1 (1 + δ) + 2 11 λχ
λ
> 0. In
the more general case, with 1 s < λ, the third term in J ( 1) can in principle become negative when
BY > 0 but we can in fact prove that the sufficient condition for its being positive (sum of third and
fourth term is always positive): 2 1 λ s 1 RBY + 1 + 1 1 λ s R > 0 implies (using 1 1 λ s R < 1)
1
1 s +1
(1 λ )R
BY < >1
2
Replacing the equilibrium expression for debt, this implies
0 1
1 λ 1 2
1 s βR 1 1 1+ 1 s
1 R
@ + A< λ
,
1+ 1 λ 1
1 1 + (1 λ) (Γ 1) 2 1 1 s
λ R
1 s βR

which is always satisfied since the LHS is negative and the RHS positive. This proves the Proposition in
the general case with BY 0. In the special case, zero-liquidity limit BY = 0 the determinacy conditions
are much simpler with J ( 1) > 0 following immediately and J (1) < 0 requiring:

1 s
>1 β.
λ
The proof that the debt quantity rule eliminates the FG puzzle and expansionary interest rate in-
creases is entirely analogous to the one for a Wicksellian rule outlined in Appendix C.3 above, with
the appropriate change of notation (implying different roots x+ and x ). Likewise, the extension to
forward-looking Phillips curve is also analogous.
Indeed, notice that in the iid idiosyncratic risk case 1 s = λ, the key difference equation becomes
(using the same notation ν0 = δ + κ 11 λχλ
):

κ 1 λ
Et pt+1 1 + ν0 1 1 + λ p t + ν0 1 p t 1 = κν 1 i .
1 λχ 1 λχ 0 t

This has exactly the same form as the equation under Wicksellian rule, but with λ instead of σ (1 λ) φ p
inside the bracket of the coefficient on pt ; here, the real demand for liquidity is elastic to the price because
of risk, combined with a fixed nominal quantity; under the Wicksellian rule, because of the direct effect
σ (1 λ) of interest rates (and their dependency on p via φ p ) on the saving decision of S.
Considering instead the extension whereby the debt rule responds to the price level btN+1 = φb pt !
bt+1 = (φb 1) pt and replacing in the Euler equation above doing the same (tedious) manipulations,
we obtain the determinacy-relevant characteristic polynomial objects:

λ 1 s 1 s
J (1) = (1 φb ) 1+ 1 R
1 λχ λ λ

A-22
 1 1 λ λ 1 s 1 s
J ( 1) = 2κ (1 + δ ) + 2 +2 2 1 RBY
1 λχ 1 λχ λ λ
λ 1 s 1 s
+ (1 φb ) 2 1+ 1 R
1 λχ λ λ

These satisfy the necessary and sufficient condition for determinacy if φb < 1 so that J (1) < 0 and
J ( 1) > 0.

E Optimal Policy in THANK

First, we write explicitly the Ramsey problem, and then we derive the second-order approximation
around an efficient equilibrium that allows transforming it into a linear-quadratic problem.

E.1 The Ramsey Problem in THANK

The Ramsey problem of maximizing a utilitarian welfare objective is:
∞
max E0 ∑ βt fλU CtH , N tH + (1 λ) U CtS , N St (E.1)
fCtH ,C St ,N tH ,N St ,π t g t =0
!
UN NtS UN NtH
+ς1,t
UC CtS UC CtH
UN ( NtH ) H τD ψ 2 UN ( NtH )
+ς2,t CtH + N (1 π + )(λN tH +(1 λ) NtS )
UC (CtH ) t λ 2 t UC (CtH )
ψ 2 H S
+ς3,t λC tH +(1 λ)CtS (1 π )(λN t +(1 λ) N t )
2 t
UC (CtS+1 ) λN tH+1 +(1 λ) N St+1
f π t (1 + π t ) βEt [ UC (CtS ) λN tH +(1 λ) N St
π t+1 (1 + π t+1 )]
+ς4,t H
U (N )
+ ε ψ 1 [ ε ε 1 UN (C Ht ) +1 + τ S ]g
C t

where ς j,t the co-state Lagrange multipliers associated to them (with arbitrary initial values).
ψ 2 ψ 2
In the above Ramsey constraints, we already substituted Ct = 1 2 πt Yt = 1 2 πt Nt ... ,
Wt = UN NtS /UC CtS = UN NtH /UC CtH , and eliminated Dt = 1 2 Π2t Wt λNtH + (1 λ) NtS
ψ

Importantly, notice that the self-insurance equation is not a constraint—just as in RANK the Euler-IS
curve is not a constraint. In other words, the equation

1 + it
UC (CtS ) = βEt s(Ct+1 )U C (CtS+1 ) + (1 s(Ct+1 ))U C (CtH+1 )
1 + π t +1

determines it residually once we found the allocation.35

Note that it is trivial to show that the first-best equilibrium amounts to perfect insurance. And solving
the above Ramsey problem and finding the optimal steady-state inflation can be easily shown to deliver
long-run price stability (π = 0) as the optimal long-run target.
35 The interest rate is no longer orthogonal in models where it affects the mapping between the income and consumption

distribution and aggregate income, for instance because it determines the MPC as in the Acharya and Dogra’s PRANK model.
See Acharya, Challe, and Dogra (2020) for a recent exploration of the optimal policy implications of this.

A-23
E.2 A Second-Order Approximation to Welfare
We approximate the economy around an efficient equilibrium, defined as an equilibrium with both flexi-
ble prices and perfect insurance; this is the case in our baseline economy under the assumed steady-state
fiscal policy, because the optimal subsidy inducing zero profits in steady state implies that consumption
shares are equalized across agents. In particular, since the fiscal authority subsidize sales at the constant
rate τ S and redistribute the proceedings in a lump-sum fashion T S such that in steady-state there is
marginal cost pricing, and profits are zero. The profit function becomes Dt (k ) = 1 + τ S Pt (k )Yt (k)
2
Pt (k)
Pt Yt + TtS where by balanced budget TtS = τ S Pt (k )Yt (k ). Efficiency requires
ψ
Wt Nt (k ) 2 Pt 1 1
1
τ S = (ε 1) , such that under flexible prices Pt (k ) = Wt and hence profits are Dt = 0 (evidently,
with sticky prices profits are not zero as the mark-up is not constant). Under this assumption we have
that in steady-state:
UN N H UN N S W Y
H
= S
= =1= ,
UC (C ) UC (C ) P N
where N j = N = Y and C j = C = Y.
Suppose further that the social planner maximizes a convex combination of the utilities of the two
types, weighted by the mass of agents of each type: Ut (.) λU H CtH , NtH + [1 λ] U S CtS , NtS . The
second-order approximation to type j0 s utility around the efficient flex-price equilibrium delivers:

Ûj,t Uj Cj,t , Nj,t Uj Cj,t , Nj,t =

1 1 1+ϕ
j σ j 2 j j 2
= UC C j ct + ct UN N j nt + nt + t.i.p + O k ζ k3 , (E.2)
2 2

where we used that flex-price values are equal to steady-state values (because of our assumption of no
j j
j Ct j Nt
shocks to the natural rate) ct log C = ct = 0 and nt log N = nt = 0.
Approximating the goods market clearing condition to second order delivers:

1
λCH,t + (1 λ) CS,t ' λc H,t + (1 λ) cS,t + λc2H,t + (1 λ) c2S,t
2
1
= λNH,t + (1 λ) NS,t ' λn H,t + (1 λ) nS,t + λn2H,t + (1 λ) n2S,t
2
The linearly-aggregated first-order term is thus found from this second-order approximation of the econ-
omy resource constraint as:

1
λc H,t + (1 λ) cS,t λn H,t (1 λ) nS,t + λc2H,t + (1 λ) c2S,t λn2H,t + (1 λ) n2S,t = 0 (E.3)
2
The economy resource constraint is

ψ 2 ψ 2
Ct = 1 π Yt = 1 π Nt
2 t 2 t

which approximated to second order is:

ψπ 1 ψ
ct = nt ψ 2 πt π2
1 2 1 ψ π2 t
2π 2

It is straightforward to show that the optimal long-run inflation target in this economy is, just like in
RANK, π = 0. Replacing, we obtain the second-order approximation of the resource constraint around

A-24
the optimal long-run equilibrium:
ψ 2
π , ct = nt (E.4)
2 t
where the second term captures the welfare cost of inflation.
Note that since UC C j and UN N j are equal across agents we can aggregate the approximations of
individual utilities above (E.2), using (E.3) and (E.4) to eliminate linear terms, into:

σ 1 2 2 ϕ 2 2 ψ 2
Ût = UC C λ ctH + (1 λ) cSt + λ ntH + (1 λ) nSt + π + t.i.p + O k ζ k3 .
2 2 2 t

Quadratic terms can be expressed as a function of aggregate consumption (output). Notice that in eval-
uating these quadratic terms we can use first-order approximations of the optimality conditions (higher
order terms imply terms of order O k ζ k3 ). Recall that up to first order, we have that ctH = χyt and
cSt = 11 λχ
λ yt and (after straightforward manipulation for hours worked):

λ
ntH = 1 + ϕ 1
σ 1
(1 χ) yt ; nSt = 1+ϕ 1
σ 1
(χ 1) y t
1 λ

To second order we thus have

2 2 h i2
ctH = χ2 y2t + O k ζ k3 ; ntH = 1+ϕ 1
σ 1
(1 χ) y2t + O k ζ k3
2 2
2 1 λχ 2 λ
cSt = y2t + O k ζ k3 ; nSt = 1+ϕ 1
σ 1
(χ 1) y2t + O k ζ k3
1 λ 1 λ

Replacing in the per-period welfare and considering only the "real" part (abstracting from inflation)
for notational convenience, and ignoring terms independent of policy and of order larger than 2 we
have:
" #
2
σ 1 2 1 λχ
λ (χct ) + (1 λ) ct
2 1 λ
" #
ϕ h i 2 λ 2
+ λ 1 + ( ϕσ) 1 (1 χ) ct + (1 λ) 1 ( ϕσ) 1 (1 χ) ct
2 1 λ

Collecting terms, multiplying by 2 for simplicity, and rearranging, using ct = yt :

λ λ h i2
σ 1
1+ (χ 1)2 y2t + ϕ 1 + ( ϕσ) 1
(χ 1) y2t
1 λ 1 λ
λ
= σ 1
+ϕ+σ 1
(χ 1)2 1 + ( ϕσ ) 1
y2t
1 λ
λ
= σ 1
+ ϕ y2t + σ 1 1 + ( ϕσ) 1 [(χ 1) y t ]2
h 1 λ i
= σ 1 + ϕ y2t + σ 1 ϕ 1 λ (1 λ) γ2t

where the last line used the expression for γt (13). Adding back the inflation term, we obtain the loss
function in (36) in Proposition 7 in text.

A-25