Merger Cycles: A Frequency Domain Approach*
| Author | Zafeira Kastrinaki,Paul Stoneman |
| DOI | http://doi.org/10.1111/j.1468-0084.2012.00691.x |
| Date | 01 April 2013 |
| Published date | 01 April 2013 |
259
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford 2012. Published by Blackwell Publishing Ltd,
9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.
OXFORD BULLETIN OF ECONOMICS AND STATISTICS, 75, 2 (2013) 0305-9049
doi: 10.1111/j.1468-0084.2012.00691.x
Merger Cycles: A Frequency Domain ApproachÅ
Zafeira Kastrinaki† and Paul Stoneman‡
†Imperial Business School, Imperial College, London, UK
(e-mail: z.kastrinaki@imperial.ac.uk)
‡Warwick Business School, University of Warwick, Coventry, CV4 7AL, UK
(e-mail: Paul.Stoneman@wbs.ac.uk)
Abstract
Using frequency domain techniques, a cycle of 6-year duration at the aggregate level
and coherent sectoral cycles of average 5-year duration are found in UK merger activ-
ity between 1969 and 2005. It is shown that business and capital market cycles jointly
are causal for the merger cycle but the capital market cycle alone is not, suggesting that
merger cycles may reflect disequilibria and/or market mis-valuation. Although the possi-
bility of disequilibrium or strong behavioural influences will complicate social evaluation,
no reason is found to advise against the current UK policy stance upon mergers.
I. Introduction
Merger and Acquisitions (M&A) activity has long been an important feature of the UK
economy, helping to shape and reshape patterns of ownership and market power. Hughes
and Singh (1987) report that, in the 1960s, UK mergers accounted for four-fifths of corpo-
rate deaths. In the 1970s, one in three of the largest 730 quoted companies was acquired,
whereas in the mid 1980s, 137 of the largest 1,000 non-financial companies were taken
over in just 4 years (1982–86). Merger activity in the 1990s was particularly impressive in
monetary terms, with transaction values (measured by the value of all domestic mergers)
totalling around £347 billion (authors’ calculation based on Office of National Statistics
(ONS) data) which is more than four times the value in the previous decade.
It has frequently been observed that the time profile of UK merger activity on several
occasions during the last century involved significant increases and then subsequent reduc-
tions in the number of recorded mergers (see, e.g. Hughes, 1993). It has become increasingly
popular to characterize this behaviour as ‘wave-like’ or ‘cyclical’. It is obvious, however,
that merger series do not punctually follow a sine or cosine wave, and thus some research-
ers prefer the more agnostic descriptors of fluctuations or waves as opposed to cycles.
However, merger activity does display what Hillinger (1992) calls a quasi-cycle, meaning
ÅWe would like to thank Michael Waterson, Dimitris Politis and Bertrand Candelon as well as two referees and
the editors of this journal for helpful comments on earlier drafts. Wewould also like to thank Christos Savva for his
advice on the use of Gauss in section III. Of course, all errors that remain are the responsibilities of the authors alone.
JEL Classification numbers: C14, C22, C32, E32, G34.
260 Bulletin
that ‘the length of the period and also the amplitude [is] to some extent variable, their vari-
ations taking place, however, within such limits that it is reasonable to speak of an average
period and an average amplitude’. We thus continue to use the term cycle. It is in such
merger cycles rather than the underlying merger activity per se that we are interested here.
The first objective pursued in this paper is to explore, more robustly than previously,
the existence of cycles in UK merger activity at aggregate and sectoral levels. This analysis
will not only indicate whether such cycles exist but will enable characterization of such
cycles via a set of summary statistics. Such analysis will also provide a set of ‘regularities’
which researchers may then use as a benchmark to examine the validity of alternative
theoretical models.
The limited empirical work to date on merger patterns has confined itself to character-
izations of the stochastic process behind merger activity by focusing on linear (Shughart
and Tollison, 1984; Golbe and White, 1988, 1993) or nonlinear time series models (Town,
1992; Linn and Zhu, 1997; Resende, 1999). In our view, such studies fail clearly to distin-
guish movements at different frequencies and do not characterize, in an intuitive summary
way, the cyclical dynamics of mergers, that is, periodicity, explanatory power and regular-
ity. This paper uses a different, potentially more appropriate, frequency domain approach
to establish robust stylized facts about UK merger cycles.
Specifically, spectral analysis is used to investigate merger cycle regularities, both in
aggregate and at the sectoral level, within a model-free framework, with cyclical fluctu-
ations isolated by the use of filtering methods. The view is then taken that merger cycles
are fluctuations within a specific range of periodicities enabling such cycles to be defined.
Because it is generally argued that merger cycles are unlikely to be longer than business
cycles, which Burns and Mitchell (1946) define as having cyclical components from 6
to 32 quarters in duration, the merger cycle may also be defined as fluctuations with a
range of periodicities from 6 to 32 quarters. Similar cut-off points are used by Granger and
Hatanaka (1964), Lucas (1980) and Levy and Dezhbakhsh (2003). We thus proceed by
filtering out from the data variations at frequencies below 6 quarters in duration and then
use the filtered data to identify regularities in merger patterns. Clear evidence is found of a
long regular cycle in aggregate merger activity with a period of 6 years (24 quarters) with
cyclical activity also being found in most UK industrial sectors.
Having identified the power and the duration of merger cycles, the second objective
pursued is to determine whether there exists any synchronization of UK cycles in different
sectors. For this purpose, we use multivariate spectral analysis. It is found that for most
sectors, merger activity synchronizes with aggregate merger activity. Such synchronicity
might indicate that there are common drivers of merger activity in most sectors of the UK
economy.
The third objective of this paper is to explore the drivers of merger cycles by investigat-
ing the patterns of causality between merger cycles and business and capital market cycles.
This, in addition to being of interest per se, may guide predictions of future patterns of
merger activity and also inform policy analysis. Empirical studies to date on such macro-
economic determinants of merger activity have primarily used time domain techniques and
provide mixed results (Melicher, Ledolter and D’Antonio, 1983; Geroski, 1984; Clark,
Chakrabarti and Chiang, 1988; Guerard, 1989; Clarke and Ioannidis, 1996; Crook, 1996).
Our results are new in that they are generated using frequency domain techniques.
©Blackwell Publishing Ltd and the Department of Economics, University of Oxford 2012
To continue reading
Request your trial