The triggers and clustering properties of merger waves

ABSTRACT

This article studies the triggers and the agglomeration of mergers and acquisitions (M&A) activity within clusters constituted by time, market and industry. Based on almost 500,000 individual transactions, we find that industry factors play a significant role in triggering activity and that M&A agglomerates strongly across related industries. While clustering in time turns out to be insignificant, stock market effects can be either an attracting or a repelling force, depending on the type of deal examined. This supports the view that merger waves are largely driven by industry shocks.

KEYWORDS:

• Merger wave
• clustering
• acquisitions
• industry shocks
• behavioural

JEL CLASSIFICATION:

• L20
• G30

Acknowledgement

I would like to thank Tomaso Duso and Lawrence White for their helpful comments.

Disclosure statement

No potential conflict of interest was reported by the author.

Notes

¹ We use the term ‘merger wave’ in a liberal sense, encompassing mergers as well as different types of acquisitions.

² Harford (2005) and Gugler, Mueller, and Weichselbaumer (2012) are, to a degree, exceptions.

³ A notable exception is the study by Erel et al., which uses almost 200,000 individual transactions.

⁴ Di Giovanni (2005) is an exception; his sample includes almost all nations.

⁵ Data on deal values are available for around 45% of observations. In an effort to keep the real threshold for inclusion in the sample constant, we obtain country- and year-specific CPI data from the World Bank (http://data.worldbank.org) and drop observations with a deal value lower than 1 million in CPI-adjusted 2005 USD.

⁶ The transactions dropped are mostly buybacks, plus a few exchange offers and recapitalizations.

⁷ When firms with large international presences acquire another firm via a subsidiary, the headquarter location of the subsidiary is listed as the acquirer’s nation. We retain this classification.

⁸ The assignment to industry codes is deal-specific, not firm-specific. This is important in the case of diversified companies, doing a large number of acquisitions. For example, Nestlé does the bulk of its acquisitions in SIC 2066 (chocolate and cocoa Products), some in the related SICs 2064 (candy and other confectionery products) and 2086 (bottled and canned soft drinks and carbonated water), but also has deals in more distant industries like 2899 (chemicals and chemical preparations) and 8741 (management services).

⁹ Results are not strongly affected by the choice of industry weights.

¹⁰ To be clear, M&A between different U.S. states are still coded as ‘domestic’ transactions, i.e. the United States still counts as one country.

¹¹ For the remaining 4%, the ownership status of the acquirer could not be determined. These observations are not used in the construction of the public/private subsamples.

¹² Deals are classified as either mergers or acquisitions in the dataset. We retain that classification.

¹³ The 1985–1989 period is included in the figure, but not in the estimation sample because data on industry and financial market performance are unavailable.

¹⁴ Since the industry averages are calculated from accounting data, they vary on the yearly level, whereas the sizes of the clusters are calculated monthly. While this reduces the efficiency with which the available information is employed, it does not pose econometric problems, particularly with a view to the large sample size. Furthermore, calculating sample sizes on the yearly level does not qualitatively change any findings reported in Table 2. We retain the monthly specification for consistency with the results reported in Table 3. The regressions in Tables 2 and 3 contain fewer observations than discussed in Section III, because industry averages are unavailable for ~4% of observations.