Does the Macroeconomy Predict UK Asset Returns in a Nonlinear Fashion? Comprehensive Out‐of‐Sample Evidence
| Author | Sadayuki Ono,Stuart Hyde,Massimo Guidolin,David McMillan |
| Published date | 01 August 2014 |
| Date | 01 August 2014 |
| DOI | http://doi.org/10.1111/obes.12035 |
510
©2013 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd.
OXFORD BULLETIN OF ECONOMICSAND STATISTICS, 76, 4 (2014) 0305–9049
doi: 10.1111/obes.12035
Does the Macroeconomy Predict UK Asset Returns in
a Nonlinear Fashion? Comprehensive Out-of-Sample
Evidence*
Massimo Guidolin†, Stuart Hyde‡, David McMillan§ and
Sadayuki Ono¶
†Manchester Business School, University of Manchester, Booth Street West, Manchester,
UK; IGIER, Bocconi University, Milan, Italy (e-mail: massimo.guidolin@mbs.ac.uk,
massimo.guidolin@unibocconi.it)
‡Manchester Business School, University of Manchester, Booth Street West, Manchester, UK
(e-mail: stuart.hyde@mbs.ac.uk)
§Stirling Management School, University of Stirling, Stirling, UK
(e-mail: david.mcmillan@stir.ac.uk)
¶Department of Economics, Hiroshima University, Japan (e-mail: sono@hiroshima-u.ac.jp)
Abstract
We perform a comprehensive examination of the recursive, comparative predictive per-
formance of linear and nonlinear models for UK stock and bond returns. We estimate
Markov switching, threshold autoregressive (TAR) and smooth transition autoregressive
(STR) regime switching models and a range of linear specifications including models
with GARCH type specifications. Results demonstrate UK asset returns require nonlinear
dynamics to be modelled with strong evidence in favour of Markov switching frameworks.
Our results appear robust to the choice of sample period, changes in loss functions and to
the methodology employed to test for equal predictive accuracy. The key findings extend
to a similar sample of US data.
I. Introduction
Empirical research overthe past two decades has seen a huge increase in interest in nonlinear
dynamics in macroeconomic and financial time-series. Although the belief in the nonlinear
behaviour of the business cycle has been long-held, it is only over this recent time frame
that a consistent body of work has been established examining and testing such dynamics.
Arguably, this was initiated by business cycle researchers (see, e.g. DeLong and Sum-
mers, 1986; Sichel, 1989) and was then extended to the search for nonlinear dynamics in
*The authors are grateful to the Editor, Christopher Bowdler and two anonymous referees for comments and
suggestions which substantially improvedthe ar ticle.
JEL Classification numbers: C32, C53, E44, G12.
Predicting UK asset returns in a nonlinear model 511
financial variables, including stock returns (see, e.g. Leung, Daouk and Chen, 2000;
Maasoumi and Racine, 2002) and interest rate dynamics (e.g. Balke and Fomby, 1997;
Enders and Granger, 1998).
Concurrently, empirical research has also been devoted to the examination of the links
between macroeconomic and financial variables and in particular to tests of whether the
former can forecast the latter. Although, this line of research has an established history
(see e.g. Keim and Stambaugh, 1986; Fama and French, 1989), interest has been renewed
following the workof Pesaran and Timmermann (1995), who forcefully argued in favourof
predictability. Interestingly, a portion of this research has examined in depth the forecasta-
bility of UK stock and bond returns (interest rates), see for example, Artis, Banerjee and
Marcellino (2005), Pesaran, Schuermann and Smith (2009) and Pesaran andTimmermann
(2000). The main implications of this literature when applied to UK data have been similar
to the typical findings of the ever expanding research concerning linear predictability from
macroeconomic variables to asset returns (see e.g. Timmermann,2008): linear predictabil-
ity tends to be ‘elusive’, that is, subject to frequent structural changes and very hard to
exploit in practice because – due to the presence of breaks and instability – it is hard to
forecast the appearance and structure of the very predictability patterns one would like to
exploit.
Inevitably, these two lines of research merged, with the search for asset return pre-
dictability from macroeconomic variables within a nonlinear framework.This work largely
began in earnest with the articles by Perez-Quiros and Timmermann (2000), McMillan
(2001) and Maasoumi and Racine (2002) who used prominent nonlinear models, such as
the Markov-switching, smooth-transition and threshold regression approaches. However,
it remains an open question whether the nonlinear approach to predictability modelling
provides any substantial benefit over linear alternatives as, despite many forecasting exer-
cises, the evidence is not definitive. The aim of this article is to reconsider the evidence
regarding the usefulness of nonlinear forecasts for UK stock and bond returns. In particular,
we seek answer to the question of whether nonlinear modelling of prediction regressions
linking financial asset returns to macroeconomic variables provides significantlyimproved
forecasting performance over linear alternatives.
With regard to previous research on UK data, there is a wealth of articles that have
investigated whether nonlinear econometric models may provide any payoffs in the fore-
casting space when applied to financial returns. Unsurprisingly, using different data, sam-
ple periods, research designs and especially heterogeneous modelling approaches,
these articles have reached wildly different conclusions. For instance, Sarantis (2001)
uses smooth transition autoregressive (STR) models to predict UK stock returns while
McMillan (2003) demonstrates the ability of such models to predict returns with a
variety of macroeconomic variables. However, though the context is similar, the fore-
casting exercise in McMillan (2003) is limited in terms of breadth of models enter-
tained and of metrics employed (only the root mean squared error) in comparison to the
extensive exercise here. Further evidence on the need to model nonlinear dynamics is
provided by Guidolin and Timmermann (2003, 2005) who employ a Markov-switching
approach. Guidolin and Timmermann (2003) demonstrate that accounting for regimes
in UK stock returns leads to improved forecasting performance, while Guidolin and
Timmermann (2005) establish the need for nonlinear dynamics in both stock and bond
©2013 The Department of Economics, University of Oxford and JohnWiley & Sons Ltd
To continue reading
Request your trial