New evidence on the robust identification of news shocks: Role of revisions in utilization‐adjusted TFP series and term structure data

• DOI: http://doi.org/10.1002/for.2507
• Date: 01 April 2018
• Published date: 01 April 2018
• Author: Zulfiqar Ali Wagan,Hakimzadi Seelro,Zhang Chen

RESEARCH ARTICLE

New evidence on the robust identification of news shocks:

Role of revisions in utilization‐adjusted TFP series and term

structure data

Zhang Chen

1

| Zulfiqar Ali Wagan

1

| Hakimzadi Seelro

2

1

School of Management, Hefei University

of Technology, Hefei, China

2

Faculty of Social Sciences, Sindh

Agriculture University Tando Jam,

Pakistan

Correspondence

Zulfiqar Ali Wagan, School of

Management, Hefei University of

Technology, 230009, China.

Email: zulfiqarsch@hotmail.com

Funding information

National Natural Science Foundation of

China, Grant/Award Number: 71373065

Abstract

Data revisions and selections of appropriate forwarding‐looking variables have

a major impact on true identification of news shocks and quality of research

findings derived from structural vector autoregression (SVAR) estimation. This

paper revisits news shocks to identify the role of different vintages of total

factor productivity (TFP) series and term structure of interest rates as major

prognosticators of future economic growth. There is a growing strand of

literature regarding the use of utilization‐adjusted TFP series, provided by

Fernald (Federal Reserve Bank of San Francisco, Working Paper Series,

2014) for identification of news shocks. We reestimate Barsky and Sims'

(Journal of Monetary Economics, 2011, 58, 273–289) empirical analysis by

employing 2007 and 2015 vintages of TFP data. We find substantial quantita-

tive as well as qualitative differences among impulse response functions when

using 2007 and 2015 vintages of TFP data. Output and hours initially decline,

followed by quick reversal of both variables. In sharp contrast to results

achieved by the 2007 vintage of TFP data, results achieved by the 2015 vintage

of TFP data depict that output and hours will increase in response to positive

TFP shock. By including term structure data in our VAR specification, total

surprise technology shock and news shock account for 97% and 92% of the

forecast error variance in total TFP and total output respectively. We find that

revisions in TFP series over time ultimately impact the conclusion regarding

news shocks on business cycles. Our results support the notion that term

structure data help in better identification of news shock as compared to other

forward‐looking variables.

KEYWORDS

business cycle, datavintage, news shocks, productivity, termstructure of interest rates, utilization‐

adjusted total factor

1|INTRODUCTION

Forces behind economic fluctuations and transmission of

these fluctuations have fueled a fierce debate over recent

years, as it did 75 years ago in the middle of the Great

Depression. Modern models of the business cycle generate

fluctuations in the business cycle due to changes in funda-

mentals—for instance, changes in technology,

Received: 5 May 2017 Revised: 29 August 2017 Accepted: 2 December 2017

DOI: 10.1002/for.2507

352 Copyright © 2018 John Wiley & Sons, Ltd. Journal of Forecasting. 2018;37:352–370.wileyonlinelibrary.com/journal/for

preferences, and government policy. The hypothesis that

the news shocks can be a basis for economic oscillations

was well recognized in the past.

1

A growing experimental

and hypothetical literature describing the function of

news shocks in generating business cycle fluctuations

has extended over recent years (Jaimovich & Rebelo,

2008; Kamber, Theodoridis, & Thoenissen, 2017;

Schmitt‐Grohé & Uribe, 2012). However, recent interest

on the subject has been revived by the most influential

work of Beaudry and Portier (2006).

2

They find that

novelties in stock prices echo news regarding prospective

total factor productivity and describe a large fraction of

business cycle movements.

In a typical business cycle model, consumption and

leisure move in opposite directions. Real business cycle

models assume that during recession production technol-

ogy is not favorable. Marginal productivity of labor is low,

resulting in low wages. Due to low returns at work,

individuals consume less and increase leisure. Barsky

and Sims (2011) applied a novel structural vector

autoregression (SVAR) approach to identify news shocks

associated with the prospective technology. They used uti-

lization‐adjusted series for total factor productivity (TFP)

provided by John Fernald. Under the proposed orthogo-

nalization scheme, news shock is related to decline in

output, hours worked in the nonfarm business sector,

and increase in consumption.

Their results were in sharp contrast to typical business

cycle models. There are important differences in time

series properties of several vintages of TFP series provided

by John Fernald on his website. In the occasional

revision notes, accompanied by an Excel spreadsheet pro-

duced on November 12, 2015, by John Fernald different

revisions are mentioned during 2012, 2013, 2014, and

2015. In March 2014 he shifted from using the utilization

estimates of Basu, Fernald, and Kimball (2006) to those of

Basu, Fernald, Fisher, and Kimball (2013), which use

recent data. In August 2014 he revised estimates of indus-

try hours, which affect the utilization estimates mainly

prior to 1964 or so. According to John Fernald, these revi-

sions improve estimation. There were also certain revi-

sions during 2015; for example, he corrected

programming errors in mapping hours per worker to

industries.

Data revisions in utilization‐adjusted TFP series over

time might impact our conclusion identifying the impact

of news shocks on business cycle fluctuations (Sims,

2016). There is a significant dearth in the news shock lit-

erature in identifying the role of data revisions and their

impact on business cycle fluctuations. How revisions in

utilization‐adjusted TFP impact business cycle dynamics

is an important question that needs to be addressed.

Earlier research has well documented the importance of

including forward‐looking variables such as consumer

confidence, inflation, and stock prices for proper identifi-

cation of news shocks (Barsky & Sims, 2011; Beaudry &

Portier, 2006). A recent strand of literature has acknowl-

edged the importance of slope of term structure in

predicting future economic activity (Kurmann & Otrok,

2013).

3

Surprisingly, despite the importance of including

various forward‐looking variables for identification of

news shocks, there has been primarily no exploration of

alternative postulation such as whether it is consumer

confidence, stock price data or term structure data which

accounts for the major share of forecast error variance of

TFP and output due to news shock. This paper attempts

to answer the above questions. Mainly we are interested

in answering the following questions: How might our

conclusion regarding the impact of news shocks on busi-

ness cycle fluctuations change by using different vintages

of TFP series? Which forward‐looking variable accounts

for the major share of forecast error variance of TFP and

output due to news shock?

We apply a novel SVAR approach proposed by Barsky

and Sims (2011) for our analysis. They identified

news shock as a shock orthogonal to reduced‐form

innovations in utilization‐adjusted TFP series which best

accounts for the variance share of adjusted TFP over a

40‐quarter horizon.

Their approach has a number of attractive features. It

permits but does not necessitate that either of the shocks

or both have a permanent impact on TFP. Another

advantage is that their approach does not impose any

restriction regarding the common trends in diverse vari-

ables used in VAR. Moreover, due to the partial identifica-

tion scheme, this approach can be extended to other VAR

specifications without imposing illogical assumptions

regarding other shocks.

We revisit the news shocks analysis from Barsky and

Sims (2011) by using the 2007 vintage of adjusted TFP data

provided to the authors by John Fernald and a recent vin-

tage of the TFP series, produced on November 12, 2015, by

John Fernald. Our findings suggest that the vintage of uti-

lization‐adjusted TFP data determines to a large extent the

impulse response functions. There are substantial qualita-

tive differences in some specifications while using a recent

1

As mentioned in the work of Pigou (1927).

2

The role of news shocks in business cycle fluctuations has been well

documented in the works of Christiano, Ilut, Motto, and Rostagno

(2008), Jaimovich and Rebelo (2009), Fujiwara, Hirose, and Shintani

(2011), and Barsky and Sims (2012).

3

Piazzesi and Schneider (2007) and Rudebusch and Swanson (2012) stud-

ied the relationship between the consumption, inflation and term struc-

ture by including macro determinants in multifactor models.

CHEN ET AL.353