News Shocks, Business Cycles, and the Disinflation Puzzle

• Published date: 01 December 2023
• Author: HAFEDH BOUAKEZ,LAURENT KEMOE
• Date: 01 December 2023
• DOI: http://doi.org/10.1111/jmcb.12939

DOI: 10.1111/jmcb.12939

HAFEDH BOUAKEZ

LAURENT KEMOE

News Shocks, Business Cycles, and the Disination

Puzzle

We arguethat key ndings of the empirical literature on the effects of news

about future technology—including their tendency to generate negative co-

movementof macro-economic aggregates, and their puzzling disinationary

nature—are due to measurement errors in total factor productivity (TFP). In

this paper, we estimate the macro-economic effects of news shocks in the

United States using an agnostic identication approach that is robust to mea-

surement errors. We nd no evidence of negativecomovement conditional

on a news shock, and the disination puzzle essentially vanishes under our

identication strategy. Our results also indicate that news shocks have be-

come an important driver of business-cycle uctuations in recent years.

JEL codes:C32, E20, E32, O47

Keywords:identication, measurement errors, news shocks, sign

restrictions, technology, total factor productivity

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

macroeconomics is: what causes business-cycle uctuations? Following the seminal

work of Beaudry and Portier (2006), interest has been rekindled in Pigou (1927)’s

theory of business cycles, according to which changes and revisions in expectations

about future fundamentals can give rise to boom–bust cycles. A number of empirical

studies—based on vector autoregressions (VARs)—have therefore attempted to

gauge the importance of news shocks about future productivity in generating the

We thank the Editor, Sanjay K. Chugh, and two anonymous refereesfor very helpful comments and

suggestions. We also thank our discussants Nadav Ben Zeevand Daniele Siena, as well as Paul Beaudry,

Giacomo Candian, Ryan Chahrour, Patrick Fève, André Kurmann, Michel Normandin, and seminar par-

ticipants at the Canadian Macro Study Group Meeting and the T2M Conference for useful feedback, and

David Gutkovsky for excellent research assistance. Financial support from the HEC Montréal Founda-

tion is gratefully acknowledged. The views expressed in this paper are those of the authors and do not

necessarily represent the views of the IMF,its Executive Board, or IMF management.

H B is at the Department of Applied Economics, HEC Montréal, and CIREQ

(E-mail: hafedh.bouakez@hec.ca).L K is at International Monetary Fund (E-mail:lke-

moe2@imf.org).

Received August 28, 2020; and accepted in revised form March 7, 2022.

Journal of Money, Credit and Banking, Vol. 55, No. 8 (December 2023)

© 2022 The Ohio State University.

2116 :MONEY,CREDIT AND BANKING

type of positive comovement of macro-economic aggregates observed in the data,

and, more generally, in explaining business cycles.1

Beaudry and Portier (2006) were the rst to document using U.S. data that news

shocks lead to positive comovement of consumption, hours worked, and investment,

and account for the bulk of their variability at business-cycle frequencies. Beaudry

and Lucke (2010) and Beaudry and Portier (2014) reach essentially the same conclu-

sions.2These ndings have been challenged, however, by some scholars who ques-

tioned the underlying identication strategies.3Using an alternative, more exible,

identication approach, Barsky and Sims (2011) nd that good news about future

technology tend to raise consumption but to decrease output, hours worked, and in-

vestment in the short run.4They also nd that ination declines sharply and persis-

tently in response to a positive realization of the news shock; a result deemed puz-

zling in light of the standard New Keynesian model (see, for instance, Jinnai 2013,

Kurmann and Otrok 2014, and Barsky,Basu, and Lee 2015). Though Barsky and Sims

(2011) nd that news shocks account for a signicant fraction of output variability

at business-cycle frequencies, they invokethe negative comovement to conclude that

these shocks are unlikely to be a major driver of business cycles. These ndings are

conrmed by subsequent studies that propose alternative but related methodologies

to Barsky & Sims’ (e.g., Forni, Gambetti, and Sala 2014, Barsky,Basu, and Lee 2015,

Kurmann and Sims 2021).

Existing empirical approaches to identify news shocks about future productivity

are based on the premise that total factor productivity (TFP) is entirely and exclusively

driven by two orthogonal disturbances: unanticipated and newsshocks, the latter gen-

erally affecting TFP with a lag. This assumption is consistent with the standard treat-

ment of TFP in theoretical macro-economic models. Hence, the above-mentioned

studies invariably include a measure of TFP in the information set when attempting

to identify technological news shocks from the data.

1.Two alternative approaches to evaluate the importance of news shocks exist in the literature. The

rst consists in estimating/calibrating dynamic stochastic general-equilibrium (DSGE) models that feature

anticipated shocks to technology (e.g., Fujiwara, Hirose, and Shintani 2011, Karnizova 2012, Khan and

Tsoukalas 2012, Schmitt-Grohé and Uribe 2012, Görtz and Tsoukalas 2017, Görtz, Tsoukalas, and Zanetti

Forthcoming). The second uses directly observable news such as the standardization of new technologies

(Baron and Schmidt 2014), oil and gas discoveries(Arezki, Ramey, and Sheng 2017), and changes in rms’

stock-market valuation due to announcements of patent grants (Cascaldi-Garcia and Vukoti´

c 2022).

2.Di Casola and Sichlimiris (2018) extend the methodology proposed by Beaudry and Portier (2006)

and nd that news shocks about future technology are inationary.

3.Beaudry and Portier (2006),2014) and Beaudry and Lucke (2010) estimate small-scale systems (two

to ve equations) in which news shocks are identied using a mix of short- and long-run restrictions. Kur-

mann and Mertens (2014) show that Beaudry and Portier (2006)’s identication scheme does not have a

unique solution when applied to a vector error correction model (VECM) with more than two variables.

This identication scheme is therefore uninformativeabout the effects of news shocks and their importance

for business cycles. Kurmann and Mertens (2014) further point out that the validity of the identication

strategy proposed by Beaudry and Lucke (2010) critically depends on the plausibility of zero restrictions

for other nonnews shocks necessary to identify news shocks. Finally, Forni, Gambetti, and Sala (2014)

argue that small-scale VARsand VECMs do not contain enough information to recover anticipated tech-

nology shocks from observable variables, a problem commonly known as nonfundamentalness.

4.Barsky and Sims (2011) identify the news shock as the shock that best explains future movements

in TFP not accounted for by its own innovation.

HAFEDH BOUAKEZ AND LAURENTKEMOE:2117

In this paper, we argue that the TFP measures typically utilized in the empirical

literature contain important measurement errors that call into question the interpreta-

tion of measured TFP as a proxy for technology. This is despite the corrections aim-

ing at purging TFP of its nontechnological component by controlling for unobserved

variations in labor and capital. Most importantly, we demonstrate that the negative

comovement of macro-economic aggregates and the disination puzzle documented

in recent empirical studies are spurious and are just an artifact of using a polluted

measure of technology.

We document the severity of measurement errors in the adjusted TFP measure

constructed by Fernald (2014)—which is the most widely used TFP series—by ex-

amining the dynamic effects of an unanticipated technology shock, identied as the

reduced-form innovation to TFP, as is done in all existing VAR-based studies on

news shocks.5The most revealing symptom of the presence of measurement errors

is that unanticipated technological improvements are found to be inationary,an out-

come that runs against the conventional interpretation of surprise technology shocks

as supply shocks, and violates the prediction of standard theories of aggregate uctu-

ations. A favorable surprise technology shock is also found to have counterintuitive

effects on stock prices and consumer condence, which are initially unresponsive to

the shock but fall persistently in the subsequent periods. We interpret these anoma-

lous responses as an indication that the TFP series used in the empirical literature is

an uncleansed measure of technology. Using articial data from a simple NewKeyne-

sian model, we illustrate how, for instance, failing to account for nonconstant returns

to scale when measuring TFP could lead one to conclude that a surprise technology

shock identied as the reduced-form innovation to TFP is inationary whereas the

true shock is not. Since a correct identication of news shocks hinges on the surprise

technology shocks being properly identied, measurement errors in TFP are likely to

undermine existing identication approaches.

We then propose an agnostic identication strategy that is robust to the presence

of measurement errors in TFP. Our methodology relaxes the assumption that only

technological shocks affect measured TFP, and assumes instead that the latter can

be perturbed by nontechnological disturbances at any given horizon. To identify the

surprise technology shock, we select the linear combination of reduced-form inno-

vations that explains most of the forecast error variance of TFP at the one-quarter

horizon, subject to the constraint that a positive realization of the shock yields a neg-

ative impulse response of ination. Hence, by construction, our strategy avoids the

ination anomaly engendered by identication schemes that associate surprise tech-

nology shocks with reduced-form innovations to TFP. We then extractthe news shock

as the linear combination of reduced-form innovations that is orthogonal to the sur-

prise technology shock, and that maximizes the contribution of the news shock to the

forecast error variance of TFP at a long but nite horizon. The argument underlying

this criterion, originally proposed by Francis et al. (2014) and commonly referred to

5.The only exception is the study by Kurmann and Sims (2021), in which there is no attempt to

identify surprise technology shocks.