Technology News and the US Economy: Time Variation and Structural Changes
| Published date | 01 July 2015 |
| Date | 01 July 2015 |
| DOI | http://doi.org/10.1111/sjpe.12073 |
TECHNOLOGY NEWS AND THE US
ECONOMY: TIME VARIATION AND
STRUCTURAL CHANGES
Tim Oliver Berg*
ABSTRACT
This article examines the time-varying impact of technology news shocks on the
US economy during the Post-World War II period using a time-varying parame-
ter vector autoregression. The identification restrictions on the sign of the con-
temporaneous responses of observable variables are derived from a dynamic
stochastic general equilibrium model and robust to parameter uncertainty. I find
that the variance of news shocks has decreased over time, contributing to the
Great Moderation in real activity and inflation. The importance of news shocks
is, however, modest compared to technology surprise and non-technology shocks.
Finally, I obtain evidence in favor of a substantial decline in wage rigidity, while
the transmission to other variables has been stable.
II
NTRODUCTION
There is a growing literature on news about future changes in aggregate tech-
nology and their role in explaining business cycle fluctuations. In contrast to
technology surprise shocks, news shocks do not affect aggregate technology
contemporaneously but are incorporated into the decision making of forward-
looking households and firms. For instance, good news about future aggregate
technology increases expected income, so households expand their consump-
tion today. Moreover, firms face lower expected marginal costs and thus cut
their prices now. In recent theoretical studies, Jaimovich and Rebelo (2009),
Christiano et al. (2010a), Beaudry et al. (2011), as well as Schmitt-Groh
e and
Uribe (2012) introduce news shocks into standard business cycle models and
argue that these are an important source for aggregate fluctuations.
Empirical contributions using vector autoregressions (VARs) include Beau-
dry and Portier (2006), who estimate a bivariate system, including a measure
for aggregate technology and an index of stock market value, and isolate a
disturbance that drives long-run movements in technology and instanta-
neously affects stock prices. The authors interpret this disturbance as a shock
to the expected growth rate of technology and find that it may explain an
*Ifo Institute
Scottish Journal of Political Economy, DOI: 10.1111/sjpe.12073, Vol. 62, No. 3, July 2015
©2015 Scottish Economic Society.
227
important fraction of business cycle fluctuations. Their results hence question
the central finding of the real business cycle (RBC) literature, namely that
aggregate fluctuations are driven solely by shocks to the current level of tech-
nology, i.e., technology surprise shocks. In Beaudry and Lucke (2010), a simi-
lar procedure is applied to larger VAR systems, also allowing for the
possibility of non-technology shocks. In the new Keynesian literature, a prom-
inent role for explaining business cycle fluctuations is given to monetary pol-
icy and preference shocks. Beaudry and Lucke (2010) demonstrate, however,
that a substantial fraction of the variation in hours worked may be attributed
to the expected changes in technology, while technology surprise and non-
technology shocks play a much smaller role. Furthermore, Barsky and Sims
(2011) reassess the relevance of expectations driven business cycles and find
that news shocks indeed account for a large bulk of output fluctuations at
medium frequencies but alone fail to explain the decline in output during sev-
eral US recessions since the early 1960s.
In this article, I examine the time-varying impact of technology news shocks
on the US economy during the Post-World War II era using a time-varying
parameter vector autoregressive (TVP-VAR) model. Given that there is con-
siderable evidence that the structure of the US economy has changed over the
last decades, it is surprising that the empirical literature on news shocks has
not yet explored whether the size and transmission of such shocks has been
stable over time or not. The contribution of this article is thus novel in this
respect.
It is a well-documented stylized fact that major US macroeconomic series,
such as output growth and inflation, have been more volatile during the 1970s
and early 1980s than before that period and after. While some authors point
to the heteroscedasticity of economic shocks as an explanation for the chang-
ing pattern in volatility (see, e.g., Stock and Watson, 2003; Sims and Zha,
2006, among others), others argue that the way variables respond to shocks
has also changed. For instance, Champagne and Kurmann (2013) provide evi-
dence for an increase in wage flexibility over the past 25 years due to de-
unionization and a shift toward performance-pay contracts. Since greater flex-
ibility in wage setting should decrease business cycle fluctuations, the observed
decline in output growth volatility may at least in part be the result of struc-
tural changes in the US labor market (see also Blanchard and Gal
ı, 2009, for
a similar view). Moreover, many authors argue that US monetary policy has
become more aggressive against inflationary pressures since the appointment
of Paul Volcker as Chairman of the Federal Reserve in 1979 (see, e.g., Clarida
et al., 2000; Cogley and Sargent, 2002, , 2005; Lubik and Schorfheide, 2004,
among others). In their view, improved monetary policy helped to dampen
the impact of unfavorable economic shocks on the US economy. Taken
together, this leads me to suspect that the size and transmission of news
shocks to the US economy may have also changed over the past decades. If,
for instance, the way monetary policy responds to such shocks varies over
time or wages adjust faster, forward-looking households, and firms will take
228 TIM OLIVER BERG
Scottish Journal of Political Economy
©2015 Scottish Economic Society
this into account when making their decisions, thereby potentially altering the
propagation of news shocks to real activity and inflation.
The TVP-VAR model is developed inter alia in Cogley and Sargent (2005)
as well as Primiceri (2005)
1
and features both time-varying coefficients and sto-
chastic volatility. The model is hence an appropriate framework to address the
question of interest since it allows for smooth and permanent changes in the
structure of the economy via drifting coefficients, while accounting for the pos-
sibility that the size of shocks is not stable over time. For instance, Gal
ı and
Gambetti (2009) employ this model to study whether the remarkable decline in
the volatility of output growth and inflation since the mid-1980s, known as the
Great Moderation, was the result of a drop in the magnitude of technology
and non-technology shocks. At least in part, they reject this good luck hypoth-
esis. Hofmann et al. (2012) use a TVP-VAR model to explore the time varia-
tion in the effects of technology surprise shocks on the US economy. Their
findings suggest that the high-inflation rates of the mid and late 1970s can be
linked to a high degree of wage indexation in combination with a weak reac-
tion of the monetary policy authority to inflation. These authors hence provide
a much richer explanation for the underlying sources of the Great Inflation
than the well-known bad luck story, which suggests that the poor economic
performance of the mid and late 1970s was primarily due to exceptionally large
unfavorable economic shocks. To analyze the time-varying effects of monetary
policy, the model is used, for example, in Baumeister et al. (2008), Canova and
Gambetti (2009), as well as Baumeister and Benati (2013), while Pereira and
Lopes (2010) and Kirchner et al. (2010) provide an application to the United
States and euro area fiscal policy, respectively. Finally, Baumeister and Peers-
man (2013a, b) investigate the time-varying impact of oil supply shocks on the
US economy within the same framework. An application of the TVP-VAR to
news shocks is, however, not yet available.
To identify technology news shocks in the TVP-VAR model, I impose theo-
retical sign restrictions on the contemporaneous impulse responses of observa-
ble variables. The restrictions are derived from a new Keynesian dynamic
stochastic general equilibrium (DSGE) model and robust to parameter uncer-
tainty. Furthermore, the set of restrictions is sufficient to discriminate news
shocks from other supply and demand side disturbances, namely technology
surprise, monetary policy, preference, and labor supply shocks, respectively.
Such a model-based identification strategy is frequently used in the literature
(see, e.g., Canova et al., 2007; Dedola and Neri, 2007; Gambetti et al., 2008;
Peersman and Straub, 2009, among others) and preferable to recursive and
long-run identification schemes for the following reasons. First, the contempo-
raneous zero restrictions implied by recursive systems are often inconsistent
with economic theory and thus hard to defend. For example, the delayed
responses of output and inflation to movements in the interest rate, which
identify a monetary policy shock in recursive VARs, are absent in most
DSGE models and often produce a price puzzle that seems to be an artifact
1
See also the corrigendum to the article in Del Negro and Primiceri (2013).
TECHNOLOGY NEWS AND THE US ECONOMY 229
Scottish Journal of Political Economy
©2015 Scottish Economic Society
To continue reading
Request your trial