Stock market integration: Evidence from returns in developed markets

The purpose of the paper is to assess whether there is increased integration among stock market returns of some major developed markets between 1981 and 2008 while showing that the degree of integration among stock market returns varies depending on the different frequency data.

Data:  We use weekly, monthly and quarterly MSCI Country Equity Price Index in the US dollar abstracted from Datastream for the United States, Canada, France, Germany, United Kingdom, Hong Kong, and Japan. Methods Integration among returns of developed markets are further evaluated using Monte Carlo Integration, which uses the same information as Variance Decompositions (VDCs) and is also based upon the moving average (MAR) representation of the VAR system. Thus, responses with 95 percent confidence bands obtained through Monte Carlo Integration can measure the strength of the correlation among returns of the markets through the ability of returns of Canada, for example, to account for responses of returns of the U.S. market at horizons of interest.  Monte Carlo Integration with 1000 draws was used to generate responses with 95 percent confidence bands of the variables that enter in the VAR system. We also apply cointegration and error-correction techniques to determine the integration behavior of stock market returns.

Conclusions:  The evidence presented here, based on Monte Carlos Integration, supports the following conclusion. The reason why quarterly data show stronger integration among market returns in overall is due to its own innovations capturing a much larger percentage share of responses in weekly data compared to the one in weekly data. For example, in the US case, the lags of market returns of the US capture about 41% share of responses in 6 quarter forward compared to about 91% share of responses in 6 weeks forward in weekly data. About 9% share of responses then must be accounted for by returns of the remaining developed markets. Evidence presented in the vector error-correction analysis shows that quarterly data in general give increased significance in findings of the short-run convergence among returns of developed markets. More important, quarterly data uncover more pairwise convergences among developed markets compared to weekly data. This also hold valid for quarterly data compared to weekly data uncovering stronger convergence in the long-term, as indicated by the t-statistic of the error-correction term.