Modelling the Link Between US Inflation and Output: The Importance of the Uncertainty Channel
| Published date | 01 November 2015 |
| DOI | http://doi.org/10.1111/sjpe.12083 |
| Date | 01 November 2015 |
MODELLING THE LINK BETWEEN US
INFLATION AND OUTPUT: THE
IMPORTANCE OF THE UNCERTAINTY
CHANNEL
Christian Conrad* and Menelaos Karanasos**
ABSTRACT
This article employs an augmented version of the UECCC GARCH specification
proposed in Conrad and Karanasos (2010) which allows for lagged in-mean
effects, level effects as well as asymmetries in the conditional variances. In this
unified framework, we examine the twelve potential intertemporal relationships
among inflation, growth and their respective uncertainties using US data. We
find that high inflation is detrimental to output growth both directly and indi-
rectly via the nominal uncertainty. Output growth boosts inflation but mainly
indirectly through a reduction in real uncertainty. Our findings highlight how
macroeconomic performance affects nominal and real uncertainty in many ways
and that the bidirectional relation between inflation and growth works to a large
extent indirectly via the uncertainty channel.
II
NTRODUCTION
The nature of the relationship between inflation and output (or unemploy-
ment) has been an issue of considerable debate in the macroeconomic litera-
ture. While much of the debate has focused on the levels of the two series,
there are many theories that highlight the importance of effects due to the
interaction between levels and volatilities. For example, Friedman’s (1977)
famous argument about the negative welfare effects of inflation consists of
two claims: higher inflation increases nominal uncertainty, which then
decreases output growth.
1
Thus, the negative welfare effects of inflation may
(at least partly) work indirectly via nominal uncertainty.
A series of articles published during the last 30 years (see, for example,
Logue and Sweeney, 1981; Evans, 1991; Brunner, 1993; Evans and Wachtel,
1993; Ungar and Zilberfarb, 1993; Holland, 1993, 1995; Fuhrer, 1997; Grier
and Perry, 1998, 2000; Grier et al., 2004; Elder, 2004; Balcilar and Ozdemir,
*Heidelberg University
**Brunel University
1
We will use the terms variance, variability, uncertainty and volatility interchangeably in
the remainder of the text.
Scottish Journal of Political Economy, DOI: 10.1111/sjpe.12083, Vol. 62, No. 5, November 2015
©2015 Scottish Economic Society.
431
2013) highlights the importance of nominal and real uncertainty for macro-
economic modelling and policy making.
Brunner and Hess (1993) was one of the first articles to employ a univariate
GARCH model to test for the first stage of the Friedman hypothesis (see also
Baillie et al., 1996). During the last decade, researchers have employed various
bivariate GARCH-in-mean models to investigate the relation between the two
uncertainties (see, for example, Conrad et al., 2010) and/or to examine their
impact on the levels of inflation and growth (see, for example, Elder, 2004;
Grier et al., 2004). However, the econometric specifications which are
employed in most of these studies are typically characterized by one or more
of the following three limitations.
First, the impact from the variabilities on the levels (the so-called in-mean
effects) is typically restricted to being contemporaneous (as, for example, in
Shields et al., 2005). However, as the theoretical rationale for the in-mean
effects usually suggests that it takes some time for them to materialize (e.g. in
the Cukierman and Meltzer, 1986, theory it requires a change in monetary
policy), it appears more appropriate to investigate such effects within a specifi-
cation that includes several lags of the variances in the mean equations (see
also Elder, 2004 and Conrad et al., 2010).
Second, the existing literature focuses almost exclusively on the impact of
macroeconomic uncertainty on performance, but neglects the effects in the
opposite direction (level effects). Moreover, the few studies that take level
effects into account focus on own but not cross-level effects. In sharp contrast,
the empirical results in Logue and Sweeney (1981) suggest that higher nominal
uncertainty produces greater variability of real growth. That is, inflation, via
the nominal uncertainty channel, affects not only growth (the Friedman
hypothesis) but real variability as well. In addition, Brunner (1993) points out
that while the second stage of Friedman’s hypothesis is plausible, the negative
causation between nominal uncertainty and growth could also work in the
opposite direction. Therefore, higher growth rates via nominal uncertainty
could reduce real variability. In the first stage, high growth rates reduce infla-
tion uncertainty (the Brunner conjecture). In the second stage, this reduced
inflation variability lowers real uncertainty (the Logue-Sweeney theory). Thus,
a meaningful empirical analysis should allow for bidirectional causality
between the four variables.
Third, the two most commonly used specifications are the diagonal constant
conditional correlation (CCC) model (see, for example Grier and Perry, 2000;
Fountas et al., 2006) and the BEKK representation (see, for example, Shields
et al., 2005; Grier and Grier, 2006; Bredin and Fountas, 2009). Both specifica-
tions are characterized by rather restrictive assumptions regarding potential
volatility interaction. While the CCC model assumes that there is no link
between the two uncertainties, the BEKK specification only allows for a posi-
tive variance relationship (see Conrad and Karanasos, 2010). In sharp con-
trast, several economic theories predict either a positive or a negative
association between the two volatilities (for more details and a review of the
literature, see Arestis et al., 2002; Karanasos and Kim, 2005).
432 CHRISTIAN CONRAD AND MENELAOS KARANASOS
Scottish Journal of Political Economy
©2015 Scottish Economic Society
To continue reading
Request your trial