Temporal Aggregation and the Power of Cointegration Tests: a Monte Carlo Study*
| Published date | 01 September 2002 |
| DOI | http://doi.org/10.1111/1468-0084.00025 |
| Author | Alfred A. Haug |
| Date | 01 September 2002 |
Temporal Aggregation and the Power
of Cointegration Tests: a Monte Carlo Study*
Alfred A. Haug
Department of Economics, York University, 4700 Keele Street, Toronto, Ontario,
Canada M3J 1P3. E-mail: haug@econ.yorku.ca
I. Introduction
Unit root and cointegration tests are commonly used by practitioners in many
empirical applications, and for good reason.
1
There is a controversy about
how long a span of data should be in order to give tests reasonable power and
whether using a higher frequency of observation for a given span would
improve test power at all. This paper uses the Monte Carlo method to
give some quantitative measures in terms of test power for the trade-offs
involved when testing for cointegration, and provides some suggestions for
practitioners.
Since the seminal paper by Shiller and Perron (1985) that concluded from a
Monte Carlo study that the power of unit root tests depends (over a
‘‘substantial range of parameters’’) more on the span of data than on the
frequency of observation, practitioners have often been concerned about
possibly using spans too short and unsuitable for unit root and cointegration
analysis. However, Marcellino (1999) illustrated theoretically that time–
aggregation may increase the local power of cointegration tests but that this
effect may be offset by the effects of the associated decrease in observations
when one deals with finite samples.
2
My paper shows in Monte Carlo
experiments that this is indeed the case and assesses the extent to which this
occurs in samples of typical size used in empirical work. The data generating
*The author thanks, without implicating, the associate editor Anindya Banerjee, anonymous ref-
erees, Neil Ericsson, and participants at seminars at Nuffield College and the University of Can-
terbury for helpful comments and discussions on earlier versions of this paper. Financial support from
the Social Sciences and Humanities Research Council of Canada is gratefully acknowledged.
1
See, for example, Murray and Nelson (2000) for a recent analysis of U.S. GDP behaviour.
2
See also Granger (1990).
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 64, 4 (2002) 0305-9049
399
ÓBlackwell Publishers Ltd, 2002. Published by Blackwell Publishers, 108 Cowley Road, Oxford OX4 1JF, UK and 350
Main Street, Malden, MA 02148, USA.
To continue reading
Request your trial