Inflation Volatility with Regime Switching
| Author | Phuong V. Ngo,Maksim A. Isakin |
| Date | 01 December 2019 |
| DOI | http://doi.org/10.1111/obes.12313 |
| Published date | 01 December 2019 |
1362
©2019 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 81, 6 (2019) 0305–9049
doi: 10.1111/obes.12313
Inflation Volatility with Regime Switching*
Maksim A. Isakin,† and Phuong V. Ngo†
†Department of Economics, Cleveland State University, 2121 Euclid Avenue, Cleveland, OH
44115, USA (e-mails: m.isakin@csuohio.edu; p.ngo@csuohio.edu)
Abstract
This paper presents a new approach to model U.S. inflation dynamics by allowing regime
switching in an unobserved components stochastic volatility framework.We use a modified
particle filter to construct likelihood and estimate the model using MLE. The number of
regimes is determined based on a bootstrap. We find that a model with three regimes and
regime-dependent constant volatilities has superior performance. In addition, we showthat
since 2000:II, U.S. inflation has entered a regime with moderate volatility where most of
the volatility comes from transitory shocks.
I. Introduction
Since the 1960s, U.S. inflation has exhibited several episodes with distinctive dynamics.
Most prominent episodes include a dramatic rise of the inflation level and its volatility in
1970s, low and stable inflation in the 1990s, and low but volatile inflation in the wake of
the Great Recession. A growing number of studies using DSGE models has documented
the presence of different regime shifts in the U.S. economy.1Thus, it is expected that U.S.
inflation reveals regime-dependent dynamics. For this reason, many reduced-form models
fail to forecast inflation over a long period or reveal changes in the parameter estimates
over time; see Faust and Wright (2013) and Stock and Watson (2007) for more discussion.
Stock and Watson (2007) show that the U.S. inflation rate can be well described by a
parsimonious unobserved components stochastic volatility (UCSV) model. In their model,
inflation is modelled by the sum of a stochastic trend component and a cycle component,
where the log variancesof the innovations for the trend component and the cycle component
follow independent random walk processes. The UCSV model is equivalent to the first-
order integrated moving average IMA(1,1) model used by Nelson and Schwert (1977),
except that the MA coefficient is time-varying to capture important changes in policy and
economic regimes.2Clark and Doh (2014) and Faust and Wright (2013) show that the
UCSV model is one of the best forecasting models.
JEL Classification numbers: C14, C15, C32, E31, E37.
*The authors acknowledge the financial support from the Faculty Scholarship Initiative (FSI) Program of the
Cleveland State University. We are also grateful for the support from the Ohio Supercomputer Center.
1See, for example, Aruoba, Cuba-Borda and Schorfheide (2018), Bianchi (2013), McConnell and Perez-Quiros
(2000), and Liu et al., (2019).
2The UCSV and time-varying IMA(1,1) models are equivalent with an empiricallyplausible constraint on the MA
coefficient (see discussion in section II).
To continue reading
Request your trial