Waves and Determinants in Mergers and Acquisitions: The Case of Latin America

ABSTRACT

This article contributes to the current literature on mergers and acquisitions (M&As) by identifying the existence of waves and the determinants of M&A activity in the economies of Argentina, Brazil, Chile, Colombia, Mexico, and Peru. From a sample of 2,391 M&A announcements reported by Thomson One on these countries, applying the methodology proposed by Harford (2005), evidence of M&A waves is found for the periods 1995–2002 and 2003–2010, as reported for other regions in various international studies. After controlling for economic and business environment variables, as well as for profitability and book-to-market variables at the industry level, we find evidence that supports neoclassical theory as a main explanation for M&A activity but not for the misvaluation effect.

KEY WORDS:

• business environment
• Latin American economics
• M&A waves
• macroeconomic environment
• mergers and acquisitions

JEL CLASSIFICATION:

• C23
• G34
• K21

Acknowledgments

This article originated from a thesis completed by the first author in partial fulfillment for the degree of MSc. in Finance at Universidad EAFIT under the guidance of the two other authors. We thank the anonymous referees for their useful suggestions that helped us to improve this article. We are also grateful to Centro de Investigaciones Económicas y Financieras (CIEF) for its support and management. We also thank Professors Andrés Ramírez, Diego Restrepo and participants in XI International Finance Conference in 2011 and CLADEA’s Annual Assembly 2012, for useful discussions. All errors remain our own responsibility.

Notes

^1. Thompson One database.

^2. This represents 90% of the deals reported by Thompson One for Latin American target countries in that period. Specifically, we left out deals in Bolivia, Costa Rica, Dominican Republic, Ecuador, Panama, Uruguay, and Venezuela.

^3. Golbe and White (1993), the first study on M&A waves, model the time-series behavior of US M&A activity from 1895 to 1989 using a sinusoidal function to approximate the wavelike pattern.

^4. Privatizations are excluded, despite sharing some drivers with M&As, for being actively used as an instrument of economic policy (Megginson, Nash, and Van Randerborgh 1994; Megginson and Netter 2001).

^5. We define as focalization when the announced deal relates two firms with the same first two digits of the SIC code, following Martynova and Renneboog (2006). Appendix Table A details the sector classification.

^6. For example, the adoption of “inflation targeting” monetary policies has been widely implemented in Latin America in the past three decades.

^7. We explored a possible relationship between the M&A wave in Latin America 1995–2002 and the one reported by Harford (2005) for the United States around the same period. Specifically, we ranked and identified the 20 industries that account for 77% of the M&A deals in Latin America for the period 1995–2002. Out of those, 83% are from an industry with an M&A wave in the United States during that period as identified by Harford (2005), and 79% are from a sector with an M&A wave in Latin America as identified in this article. This result suggests a possible relationship between the first M&A wave in Latin America and the corresponding one in the United States for most industries. However, a detailed examination of the deals would be needed to confirm this, and we leave it for future studies. We thank an anonymous referee for suggesting this.

^8. According to the neoclassical hypothesis, some macroeconomic variables for the acquirer’s country are also determinants of M&A. We thank an anonymous referee for bringing this to our attention.

^9. As an additional robustness test, suggested by an anonymous referee, we run the panel-data model for the entire database of 2,391 deals, before excluding the regulated industries and the industries with too little M&A activity. The results are qualitatively very similar and are available from the authors upon request.

^10. The most commonly used is the Poisson regression. However, the negative binomial regression is a more flexible alternative that does not assume that the mean and the variance are the same (Cameron and Trivedi 2005). In fact, the Poisson model is found inadequate by the overdispersion-in-variance test. Instead, the negative binomial regression model generalizes the Poisson model by introducing an unobserved individual effect to the variance of the model: $E\left(y_{\mathrm{i}\mathrm{c}\mathrm{t}}|x_{\mathrm{i}\mathrm{c}\mathrm{t}}\beta \right)=\mathrm{e}\mathrm{x}\mathrm{p}\left(\mathbf{x}{ }_{\mathrm{i}\mathrm{c}\mathrm{t}}^{\prime }\beta \right)$ and $\mathrm{V}\mathrm{a}\mathrm{r}\left(y_{\mathrm{i}\mathrm{c}\mathrm{t}}|x_{\mathrm{i}\mathrm{c}\mathrm{t}}\beta \right)=\mathrm{e}\mathrm{x}\mathrm{p}\left(\mathbf{x}{ }_{\mathrm{i}\mathrm{c}\mathrm{t}}^{\prime }\beta \right)\left[1+\alpha \mathrm{e}\mathrm{x}\mathrm{p}\left(\mathbf{x}{ }_{\mathrm{i}\mathrm{c}\mathrm{t}}^{\prime }\beta \right)\right]$.

^11. We thank an anonymous referee for bringing this to our attention.