Merger Patterns in the European Food Supply Chain

Abstract

This paper investigates merger activity in the food supply chain in Europe as a whole, with an emphasis upon eight individual countries that were most merger active. It finds that M&A activity (vertical, horizontal, inward, and outward) has been substantial in both the production and distribution parts of the supply chain. Using spectral analysis, it also concludes that: (i) there are regular cyclical patterns in merger activity in seven of the eight countries; (ii) most countries exhibit strong coherency with overall EU merger activity in the food industry; (iii) the relative cyclical pattern of mergers in food manufacturing and retailing varies country to country; (iv) there is some evidence that mergers in manufacturing lead or Granger cause mergers in retailing; and (iv) patterns of merger activity in each of the countries studied (except for the UK and the Netherlands) are linked, at least in part, to business and capital market cycles.

Key Words:

• Merger Cycles
• Food Industry
• Spectral Analysis

JEL classifications:

• Q13
• G34
• C22

Notes

1. This is a well-known measure of merger activity used in the merger waves literature (e.g. Clarke and Ioannidis, 1996; Resende, 1996; Resende, 1999).

2. The application of spectral techniques to study the cyclical pattern of mergers also necessitates a large number of observations, that is, a long time series of data. Of the eight countries, six joined the EU prior to the start date of our sample, and Finland (1995) and Spain (1986) joined later.

3. We have explored whether the European single market agenda that has now been pursued for many years has been reflected in more cross border mergers over time. We have not been able to find any trends in the data that would suggest that this, or the opposite, is the case. This holds for both the manufacturing and retailing sectors.

4. As these eight countries also accounted for 80.4% of the GDP of the EU 27 in 2007 (Eurostat), they encompass most of the European economy.

5. Source: Thomson ONE Banker.

6. Since only trend and cycle are assumed to exist, the procedures we use in this paper implicitly assume that seasonal and cyclical components of the series are lumped together, and that irregular (high-frequency) fluctuations are not important.

7. Burns and Mitchell (1946) define business cycles as cyclical components from 6 to 32 quarters in duration. Similar cut-off points are used by Granger and Hatanaka (1964), Levy and Dezhbakhsh (2003), and Lucas (1980).

8. Speculating, it may be that the UK is different because of the larger importance of the market for corporate control in its institutional structures than in any other EU country (Becht, 2005; Jenkinson and Mayer, 1992).

9. Details available from the authors upon request.

10. This test requires first constructing a VAR model of interest and then applying frequency analysis to the estimated coefficients in order to detect causality relations at different frequencies. Using the AIC criterion a VAR(4) model was selected (although other choices of lag length did not qualitatively change the results).

11. Results are available from the authors upon request.

12. Again, using the AIC criterion, a VAR(4) model was selected.

13. The causality test is applied to the cyclical component of stock market prices and real GDP extracted by applying the HP filter, which are by definition stationary.

14. The periodogram is the discrete finite Fourier transform of the complete sample autocovariance function.

15. The Bartlett window is given by for k = 0,1,…M. Parzen’s window was also used in an attempt to check sensitivity of results at the windows used. Different windows did not qualitatively alter the results.