What Determines the Sacrifice Ratio? A Bayesian Model Averaging Approach
| Author | Hajime Katayama,Natalia Ponomareva,Malvin Sharma |
| DOI | http://doi.org/10.1111/obes.12304 |
| Published date | 01 October 2019 |
| Date | 01 October 2019 |
960
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 81, 5 (2019) 0305–9049
doi: 10.1111/obes.12304
What Determines the Sacrifice Ratio? A Bayesian
Model Averaging Approach*
Hajime Katayama,†Natalia Ponomareva,‡ and Malvin Sharma§
†School of Commerce, Waseda University, Tokyo, Japan
(e-mail: hajime.katayama@waseda.jp)
‡Faculty of Business and Economics, Macquarie University, Department of Economics
Building E4A, Room 428 Sydney, NSW, 2109, Australia
(e-mail: natalia.ponomareva@mq.edu.au)
§Macquarie University, Sydney, NSW, 2000, Australia
(e-mail: malvin.sharma@students.mq.edu.au)
Abstract
Existing empirical studies on the sacrifice ratio (measuring the output cost of disinflation)
consider a large number of potential explanatory variables including the length of disin-
flation, various institutional settings, economic conditions, and the political climate. Some
results are robust across different studies, while others are not. We address the presence of
model uncertainty by using the Bayesian model averagingmethod to identify the important
determinants of the sacrifice ratio, without relying on ad hoc model selection. Our results
show that the length of disinflation is the most important variable. This supports the ‘cold
turkey’ argument for faster disinflation.
I. Introduction
The effectiveness of monetary policy depends to a large extenton the slope of the inflation-
output trade-off or the Phillips curve. Thus, it is important to understand the determinants
of this slope. One of the most commonly used measures of the output-inflation trade-off
is the sacrifice ratio – the ratio of the total output loss to the change in trend inflation over
the course of disinflation. It has certain advantages as a proxy for output-inflation trade-
off. As indicated by Temple (2002), the shocks over the course of a disinflation episode
are mainly demand shocks. It is also safe to assume that demand contractions are due to
monetary rather than fiscal policy.In addition, given the nature of calculation of the sacrifice
ratio, understanding its determinants is crucial for policymakers who aim to engineer a
JEL Classification numbers: C50, E31 and E41.
*We are grateful to the editor Jonathan Temple and two anonymousreferees for many valuable suggestions that
helped to improve our paper. We also thank Laura Billington, Lance Fisher, Chris Heaton, Roselyne Joyeux, Jef-
frey Sheen, Vladimir Smirnov, Rami Tabri, Andrew Wait and the participants of 2013 Australasia Meeting of the
Econometric Society for useful comments.
What Determines the Sacrifice Ratio? 961
disinflation. This understanding wouldgive the authorities an idea of what they should take
into account in order to reduce inflation with minimum costs.
This study assesses the relevance of the previously identified determinants of the sacri-
fice ratio. It also examines the degree of model uncertainty in the sacrifice ratio regression
– the relationship between the sacrifice ratio and its determinants. For this purpose, we
apply a Bayesian model averaging (BMA) method. For a given set of potential determi-
nants, we evaluate all linear models that result from any subset of those determinants. For
each model, the posterior model probability (i.e. the posterior probability of being the ‘cor-
rect’ model) is computed. This allows us to examine how model probability is distributed
across different specifications, providing insights into the degree of model uncertainty in
the sacrifice ratio regression and the robustness of the results in previous studies.
Our paper relates to a growing literature examining determinants of the sacrifice ra-
tio both theoretically and empirically. The empirical literature began with an influential
paper by Ball (1994), who found the sacrifice ratio to be lower for quicker disinflation
and more flexible wages. In subsequent studies, many additional potential determinants
were considered, including some measures of central bank independence (Brumm and
Krashevski, 2003; Diana and Sidiropoulos, 2004; Daniels, Nourzad and VanHoose, 2005);
trade openness (Daniels et al., 2005; Loungani and Razin, 2005; Bowdler, 2009; Daniels
and VanHoose, 2009) and financial openness (Loungani and Razin, 2005); a measure of
central bank transparency (Chortareas, Stasavage and Sterne, 2002); inflation targeting
policy (Daniels and VanHoose, 2009; Goncalves and Carvalho, 2009; Pickering andValle,
2012); and a variable reflecting political stance (Caporale and Caporale, 2008).
Although these studies are helpful in understanding the determinants of the sacrifice
ratio, there is an issue with the robustness of results: while some results hold across various
studies, many do not. For instance, the majority of studies found that the sacrifice ratio
is negatively associated with the length of disinflation while positively associated with a
change in inflation. At the same time, evidence is mixed for trade openness and central
bank independence. For example, Ball (1994) andTemple (2002) did not find a statistically
significant relationship between trade openness and the sacrifice ratio, whereas it wasfound
to be positive in Daniels et al. (2005) and Bowdler (2009), but negative in Pickering and
Valle(2012). Similarly, Caporale and Caporale (2008) found a positive association between
a measure of central bank independence and the sacrifice ratio, while Daniels et al. (2005)
and Bowdler (2009) found a negative relationship.
These differences in results may be attributed to several issues. First, estimation meth-
ods differ across these previous studies. While the majority of studies use ordinary least
squares, some use fixed-effects methods to control for unobservabletime-invariant country
characteristics (e.g. Caporale and Caporale, 2008) or dynamic panel methods to address
time-invariant heterogeneity, and endogeneity (e.g. Brumm and Krashevski, 2003). Sec-
ond, variablesare not always defined in the same manner across studies. For example,some
studies define trade openness as the ratio of imports to GDP, while others define it as the
ratio of exports and imports to GDP. Similarly, several different measures of central bank
independence have been used in the literature.
In addition, differences in model specification can be another important source of in-
consistencies across results in the previous studies. To understand these differences, it is
important to know how models are typically specified. The initial model usually consists
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial