A Further Look at the Propagation of Monetary Policy Shocks in HANK

• Author: BENJAMIN MOLL,GIOVANNI L. VIOLANTE,GREG KAPLAN,FELIPE ALVES
• DOI: http://doi.org/10.1111/jmcb.12761
• Published date: 01 December 2020
• Date: 01 December 2020

DOI: 10.1111/jmcb.12761

FELIPE ALVES

GREG KAPLAN

BENJAMIN MOLL

GIOVANNI L. VIOLANTE

A Further Look at the Propagation of Monetary

Policy Shocks in HANK

We provide quantitative guidance on whether and to what extent different

elements of Heterogeneous Agent New Keynesian(HANK) models amplify

or dampen the response of aggregate consumption to a monetary policy

shock. We emphasize four ndings. First, the introduction of capital ad-

justment costs does not affect the aggregate response, but does change the

transmission mechanism so that a larger share of indirect effects originates

from equity prices rather than from labor income. Second, incorporating es-

timated unequal incidence functions for aggregate labor income uctuations

leads to either amplication or dampening, depending on the data and esti-

mation methods. Third, distribution rules for monopoly prots that allocate

a larger share to liquid assets lead to greater amplication. Fourth, assump-

tions about the scal reaction to a monetary policy shock have a stronger

effect on the aggregate consumption response than any of the other three

elements.

JEL codes: D14, D31, E21, E52

Keywords: monetary policy, heterogeneous agents, New, Keynesian,

unequal incidence, investment adjustment cost, prot distribution, scal

accommodation, taylor rule

A     microhetero-

geneity into New Keynesian models of the macroeconomy has advanced our

We are grateful to our discussant Gian Luca Benigno for his insightful comments on an early draft,

to Fatih Guvenen and Serdar Ozkan for kindly sharing their data, and to Hugo Lhuillier for his excellent

research assistance. We also thank many seminar participants for their comments.

F A is with Bank of Canada (E-mail: FAlves@bank-banque-canada.ca). G K is

a University of Chicago and NBER (E-mail: gkaplan@uchicago.edu). B M is with London

School of Economics, CEPR, and NBER (E-mail: b.moll@lse.ac.uk). G L. V is a Prince-

ton University, CEPR, IFS, IZA, and NBER (E-mail: violante@princeton.edu).

Received July 16, 2019; and accepted in revised form August 25, 2020.

Journal of Money, Credit and Banking, Supplement to Vol. 52, No. S2 (December 2020)

© 2021 The Ohio State University

522 :MONEY,CREDIT AND BANKING

understanding of the transmission mechanism of monetary policy.1In these Hetero-

geneous Agent New Keynesian (HANK) models, the general equilibrium effects of

an interest rate cut, which operate through an increase in household incomes from

higher labor demand, outweigh the direct effects that primarily operate through in-

tertemporal substitution. This pattern of transmission stands in stark contrast to the

Representative Agent New Keynesian (RANK) models that served as a point of de-

parture for this literature, in which monetary policy affects aggregate consumption

almost exclusively through intertemporal substitution and in which the indirect chan-

nel is negligible.

In this new framework, however, the effect of model assumptions and parameter-

izations on the consumption response to an interest rate cut is less understood. This

is because the HANK framework incorporates several (realistic) elements that are

either inconsequential or not even well dened in representative agent versions. Ex-

amples include the unequal incidence of aggregate uctuations across households,

the distribution of prots and capital gains, the cyclicality of household idiosyncratic

risk and borrowing capacity,and the scal reaction to a monetary expansion. To para-

phrase Sims (1980), once we depart from the representative household, we enter the

“wilderness of heterogeneous agent macro.”

In an attempt to tame this wilderness, a growing literature starting from Werning

(2015) has used stylized versions of HANK models that can be solved analytically

to provide theoretical guidance on the model features that determine the extent of

propagation (see, for example, Acharya and Dogra 2018, Bilbiie 2017, Debortoli and

Gali 2018, Auclert 2019, Bernstein 2019, Bilbiie, Känzig, and Surico 2019). This

literature claries the channels through which HANK model elements contribute to

amplication and dampening. However, little is currently known about which ele-

ments are quantitatively important departures from RANK models, or whether the

insights from these simple analytical models carry through to empirically relevant

versions of HANK models.

In this paper, we address this gap by providing some quantitative guidance on the

relative importance of different candidate propagation mechanisms for the case of a

monetary policy expansion. Our starting point is the two-asset HANK model studied

by Kaplan, Moll, and Violante (2018). We rst incorporate two additional ingredients

that are common in quantitative RANK models, but that were missing in the rst gen-

eration of HANK models: aggregate capital adjustment costs and a Taylor rule with

some degree of smoothing. Next, we use this setup to explore the quantitative impact

of three potential amplication channels that determine how the change in aggregate

labor, capital, and government (taxes and transfers) income induced by the monetary

policy shock is distributed across households. Our main results are as follows.

1. See Guerrieri and Lorenzoni (2011), Oh and Reis (2012), Gornemann, Kuester, and Nakajima

(2014), Den Haan, Rendahl, and Riegler (2015), Luetticke (2015), Werning (2015), McKay and Reis

(2016), Auclert (2019), McKay, Nakamura, and Steinsson (2016), Ravn and Sterk (2017), Bilbiie (2017),

Patterson (2018), Auclert and Rognlie (2018), Kaplan, Moll, and Violante(2018), Bayer, Luetticke, Pham-

Dao, and Tjaden (2019), Kekre (2019), Lenel and Kekre(2019), Cui and Sterk (2019), and Berger, Bocola,

and Dovis (2019), among others.

FELIPE ALVESET AL. :523

Adding capital adjustment costs has a negligible impact on the aggregate consump-

tion response, but changes the dynamics of investment and asset prices dramatically,

with interesting effects on the transmission mechanism of a policy rate cut. While

most of the consumption response still comes from indirect general equilibrium ef-

fects as opposed to direct effects of the real rate change—in line with Kaplan, Moll,

and Violante (2018)—adjustment costs alter the relative contribution of labor versus

nancial income. In particular, they curtail the investmentresponse by increasing the

price of capital. Less investment translates into more moderate movements in output

and hence in households’ labor income. But, at the same time, the rise in the price

of capital boosts nancial wealth and hence more of the gains from the monetary

expansion accrue to wealthy shareholders.

The partial-adjustment Taylor rule has almost no effecton the aggregate consump-

tion and investment responses. Moreover,it does not seem to matter for the decompo-

sition between different channels and for the distribution of gains from the monetary

expansion across households.

Next, we study the different candidate amplication mechanisms based on the in-

sight that in the presence of marginal propensity to consume (MPC) heterogeneity,

redistributing resources across households has real effects. Part of our contribution

here is to provide simple parameterized functional forms for each channel that are

amenable to quantitative analysis, and to discipline these empirically in some cases.

Wepay particular attention to the parameterization and estimation of various “inci-

dence functions”—a concept that has also been used by Werning (2015), Auclert and

Rognlie (2018), Bilbiie (2017), and Patterson (2018). An incidence function describes

a rule for how a time-varying aggregate quantity is allocated across the distribution

of households in the economy. It answers questions such as: when aggregate income

rises by one percent, how is this additional income distributed across the population?

We are interested in short-run incidence functions, for aggregate income uctuations

occurring at the business cycle frequency.

We propose a convenient parameterization for a general class of incidence func-

tions and separately estimate incidence functions for labor income and government

transfer income, using various sources of micro data for the United States: the Annual

Social and Economic (ASEC) supplement of the Current Population Survey (CPS),

the Survey of Consumer Finances (SCF), and tabulated statistics from the Master

Earnings File of the Social Security Administration (SSA).

Depending on the data source used to estimate the incidence functions, the unequal

distribution of income over the cycle can either dampen or amplify the consumption

response to a monetary shock. For example, estimates using ASEC data suggest that

households with low permanent income and higher MPCs are the most heavily ex-

posed to uctuations in aggregate labor income. This leads to an amplication of the

aggregate consumption response of 2% to 20% relative to a model with equal inci-

dence.2In contrast, estimates using SSA data dampen the effect of a rate cut, relative

2. We also explain that these ndings are nonetheless consistent with Patterson (2018), who reports

that her estimated incidence function results in amplication of up to 40%.