A note on cointegration methodology in studying stock markets integration

Abstract

Market integration literature employing cointegration approach often carelessly uses different categories of data, such as price indices, total return indices or indices adjusted with risk-free rate. This article proves using only data adjusted with risk-free rate can be compatible with Capital Asset Pricing Model (CAPM). Based on this finding, links between cointegration system and international CAPM are established. Assuming a group of benchmark markets in a cointegration system is fully integrated with global market, this article applies Gonzalo and Granger (1995) decomposition method to show the single common trend in the system is a proxy for accumulated world portfolio excess return and its loading vector is a vector of estimated betas in CAPM.

Keywords:

• stock market integration
• cointegration
• permanent–transitory decomposition
• international CAPM

JEL Classification:

• C32
• G15
• F36

Acknowledgements

I am grateful to Kate Phylaktis and Lorenzo Trapani for their very helpful comments and suggestions. All errors are my own responsibility.

Notes

¹ We searched for keywords ‘cointegration and ‘stock market integration’ in Web of Science powered by Thomson Reuters, then randomly chose 10 related papers.

² Arshanapalli and Doukas (1993), Arshanapalli et al. (1995), Chen et al. (2002), Yang et al. (2003), Tian (2007), Awokuse et al. (2009) and Mylonidis and Kollias (2010).

³ Kasa (1992) used price indices and dividend indices adjusted by US Consumer Price Index (CPI). Richards (1995), with the purpose of checking Kasa's result, used the total return indices adjusted by risk-free rate.

⁴ To avoid confusing with in CAPM, we used to represent the cointegration vector.

⁵ Readers can refer to Gonzalo and Granger (1995) for detailed mathematical derivations.

⁶ One can normalize to let sum of its elements be 1, so that the size of common shock can be similar to .