The Cyclical Behavior of the Price‐Cost Markup

• Author: CHRISTOPHER J. NEKARDA,VALERIE A. RAMEY
• Date: 01 December 2020
• DOI: http://doi.org/10.1111/jmcb.12755
• Published date: 01 December 2020

DOI: 10.1111/jmcb.12755

CHRISTOPHER J. NEKARDA

VALERIE A. RAMEY

The Cyclical Behavior of the Price-Cost Markup

A countercyclical markup of price over marginal cost is a key transmission

mechanism for demand shocks in New Keynesian (NK) models. This paper

reexamines the foundation of those models by studying the cyclicality of the

price-cost markup in the private economy. We nd that how the markup is

measured matters for its unconditional cyclicality. Measures of the markup

based on the inverse of the labor share are moderately procyclical, but are

moderately countercyclical for some generalizations of the production func-

tion. NK models predict that the cyclicality of the markup should vary de-

pending on the nature of the shock. Consistent with the NK model, we nd

that the markup is procyclical conditional on total factor productivityshocks

and countercyclical conditional on investment-specic technology shocks.

In contrast, we nd that the markup increases in response to a positive de-

mand shock. Thus, the transmission mechanism for the effects of demand

shocks in sticky-price NK models is not consistent with the data.

JEL E3, J3

Keywords: markup, price-cost markup, business cycle, NewKeynesian

T     marginal cost plays a key role

in sticky-price New Keynesian (NK) macroeconomic models. In these models,

a demand shock raises output and marginal cost, but since prices are sticky, the

markup of price over marginal cost falls. As pointed out by Broer et al. (2019),

a lower markup leads to higher output during booms largely through its effect on

prots. In particular, lower markups reduce prots, generating a negative wealth

This paper updates Nekarda and Ramey (2009, 2013). The analysis and conclusions set forth are those

of the authors and do not indicate concurrence by other members of the research staff or the Board of Gov-

ernors of the Federal Reserve System. We are grateful to two anonymous referees, Susanto Basu, Mark

Bils, Larry Christiano, Olivier Coibion, Steven Davis, DavideDebortoli, Martin Eichenbaum, Jordi Galí,

Robert Hall, Christopher House, Greg Kaplan, Loukas Karabarbounis, Antoine Martin, Garey Ramey,

Sergio Rebelo, Matthew Shapiro, Linda Tesar,Harald Uhlig, Kenneth West, Raf Wouters,and participants

at numerous seminars for helpful comments and suggestions. Ben Backes and Myungkyu Shim provided

excellent research assistance. ValerieRamey gratefully acknowledges nancial support from National Sci-

ence Foundation grants SES-0617219 (through the National Bureau of Economic Research) and 1658796.

C J. N is with the Board of Governors of the Federal Reserve System (E-mail:

christopher.j.nekarda@frb.gov). V A. R is with University of California, San Diego and

NBER.(E-mail: vramey@ucsd.edu).

Received June 25, 2020; and accepted in revised form July 30, 2020.

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

© 2021 The Ohio State University

320 :MONEY,CREDIT AND BANKING

effect that induces households to raise their labor supply. Even in the medium-scale

NK models that also incorporate sticky wages, countercyclical movements in the

price markup play a key role in the transmission of monetary and scal policy

shocks. For example, in the estimated dynamic stochastic general equilibrium model

from Smets and Wouters (2007), the price markup is countercyclical in response

to a monetary shock. The two-agent New Keynesian and heterogeneous-agent New

Keynesian (HANK) models also rely heavily on countercyclical price markups to

amplify shocks. As Debortoli and Galí (2018, p. 31) point out, “in both models

the amplication/dampening of aggregate shocks depends critically on the cyclical

properties of markups (or equivalently the labor share). . .” Indeed, price markups

are required to be so countercyclical in the leading HANK models that expansionary

monetary shocks cause prots to fall (Kaplan, Moll, and Violante, 2018, p. 716).

The dependence of Keynesian models on a countercyclical price markup is a

feature only of the models formulated since the early 1980s. From the 1930s through

the 1970s, the Keynesian model was founded on the assumption of sticky wages

(e.g., Keynes 1936, Phelps 1968, Taylor 1980). Some researchers believed that the

implications of this model were at odds with the cyclical properties of real wages,

leading to a debate known as the “Dunlop–Tarshis” controversy.1In response to the

perceived disparity between the data and predictions of the traditional Keynesian

model, the literature shifted in the early 1980s to relying on sticky prices rather than

sticky wages for the transmission of shocks (e.g., Gordon 1981, Rotemberg 1982).

Although the medium-scale NK models add wage stickiness, virtually all current

NK models rely on countercyclical price markups in response to demand shifts.

Is the price markup countercyclical in the data? There is no consensus, because

estimating the cyclicality of the markup is one of the more challenging tasks in

macroeconomics. Theory prescribes a comparison of price and marginal cost;

however, available data typically include only average cost. As we will discuss,

researchers have used a variety of techniques to measure the markup directly, or have

inferred its movements using indirect evidence. Some researchers have estimated

the markup to be procyclical while others have estimated it to be countercyclical.

In this paper, we assess how various measures of the aggregatemarkup move over

the business cycle and how they respond to leading macroeconomic shocks. We nd

that how the markup is measured matters for its unconditional cyclicality—that is,

its relationship with an indicator of the business cycle. Markups measured as the

inverse of the labor share are moderately procyclical, but markups based on more

general production functions are procyclical or countercyclical depending on the

details of the empirical implementation.

Our main emphasis is on the conditional cyclicality of the markup—that is, how

the markup responds to a particular type of shock. If business cycles are driven

1. In fact, Dunlop (1938) and Tarshis (1939) were repeatedly misquoted by the literature as showing

that real wages were procyclical. Neither of them showedthis. Both authors showed that money wages and

real wages were positively correlated, and Tarshiswent on to show that real wages were in fact negatively

correlated with aggregate employment. Dunlop (1998) discusses the debate in his retrospective article.

CHRISTOPHER J. NEKARDA AND VALERIE A. RAMEY :321

by a multitude of shocks, not just demand shocks, the unconditional cyclicality

is not dispositive.2Because sticky-price NK models predict the markup should

behave differently in response to different shocks, the conditional cyclicality is the

appropriate way to evaluate these models.

Unlike our estimates of the unconditional cyclicality, the sign of the conditional

cyclicality does not depend on the empirical measure of the markup. Consistent with

the NK model, we nd that the markup is procyclical conditional on total factor

productivity (TFP) shocks and countercyclical conditional on investment-specic

technology (IST) shocks. In contrast to the sticky-price NK model predictions, we

nd that the markup increases in response to expansionary monetary policy shocks

and government spending shocks. Thus, we conclude that the transmission mecha-

nism for these policy shocks in sticky-price NK models is not consistent with the data.

One possible route to resolving this inconsistency would be for a return to the

traditional Keynesian emphasis on wage rigidities instead of price rigidities. Indeed,

several recent papers have advocated such a shift. For example, Broer et al. (2019)

advocate shifting from price stickiness to wage stickiness based on insights from

heterogeneous agent models, while Auclert and Rognlie (2017) do so based on

undesirable interactions between Greenwood, Hercowitz, and Huffman (1988)

preferences and exible wages.

1. RELATIONSHIP TO THE LITERATURE

Industrial organization economists have a long history of studying the cyclicality

of price-cost margins. Macroeconomists only began studying this issue in the mid-

1980s when macromodels started to emphasize price setting behavior of rms. Four

principal methods have been used to measure the markup directly and two additional

methods have been used to assess the cyclicality of the markup indirectly.

The rst of the direct methods uses the standard industrial organization concept of

a price-cost margin constructed from revenues and variable costs. Domowitz, Hub-

bard, and Petersen (1986) use this method in a panel of four-digit SIC manufacturing

industries and nd that margins are signicantly procyclical. Anderson, Rebelo, and

Wong (2018) use condential detailed data from the retail industry and measure

markups by comparing well-measured individual product prices to the replacement

cost of the good. This latter cost measure should be a very good proxy for marginal

cost. They nd that markups are acyclical or mildly procyclical.

The second method builds on Hall’s (1986) generalization of the Solow residual

to estimate the cyclicality of markups. For example, Haskel, Martin, and Small

(1995) extend Hall’s framework to allow for time-varying markups and apply it

2. Fleischman (1999) made this point forcefully for real wages, demonstrating empirically that real

wages are procyclical conditional on technology shocks, but countercyclical conditional on labor supply

shocks and aggregate demand shocks.