An Exploratory Study on Nonlinear Causality Among the MILA Markets

ABSTRACT

According to conventional portfolio theory, an increase in the interconnectedness of international financial markets may reduce the potential for constructing diversified portfolios. This article explores the implications of the creation of the Latin American Integrated Market (MILA)¹ over the dependence structure of its members using correlation and cointegration analysis as well as linear and nonlinear Granger causality tests. The creation of MILA aimed to enhance the integration process that Latin American financial markets “naturally” present while still providing diversification opportunities to investors. The results of our empirical analysis suggest that such objective is being achieved. Evidence of a rise in cross-country linear correlations and their linear causal relationship supports the idea of an increasing financial integration process in the region, while the absence of cointegration and the weakening of the nonlinear causal relationship favors the creation of diversified regional portfolios. These findings provide valuable insights for investment portfolio designers, regulators, and supervisors.

KEYWORDS:

• financial integration
• MILA
• nonlinear causality
• stock markets

JEL CLASSIFICATION:

• F36
• G15
• C14
• C58
• C52

Notes

1. “MILA” stands for “Mercado Integrado Latino Americano.” It consists of an electronic trading platform for stocks listed in the Chile, Colombia, Mexico, and Peru national stock markets.

2. Since August 2011, the original MILA members’ performance was measured with the S&P MILA 40 index, designed to provide information about the most liquid and highest capitalization stocks traded in the MILA platform.

3. Every attempt was made to include the studies most referred to in the integration literature. If a work has been overlooked or misinterpreted, we extend our honest apology to the author(s).

4. For a more detailed explanation of “decoupling theory,” see Akin and Kose (2008).

5. As noted by Campbell and Perron (1991).

6. The author Valentyn Panchenko published the program GCTest.exe written in C in order to perform this step.

7. Because of space restraints, the results of these tests are not included but are available upon request.