- Countries in MSCI ACWI subregions form regional building blocks for asset allocators, given strong average intraregional equity return correlations.
- Both inter- and intraregional correlations followed the cyclical behavior of the global equity market — they were typically higher in times of market stress.
- Despite increased global integration trends, investing across geographical borders continued to offer significant diversification opportunities.
Global equity investors use regions as building blocks in asset allocation, typically segregating markets by how developed they are and by geography. Frequently, they’ve turned to indexes as market proxies.1 While this approach seems intuitive, has this classification been supported by equity-market performance and hence reflected in stock-return correlations? With increases in international trade and global interdependence, how have these correlations changed over time? Has globalization reduced the potential for geographical portfolio diversification?2
Have stock-return correlations supported the country classification standard?
The correlation matrix below is ordered by six subregions: developed markets (DM) in the Americas, APAC and EMEA and their emerging-market (EM) counterparts. Within each subregion, countries (as measured by their country indexes) are ranked from highest to lowest according to their average correlation within the subregion. These subregions formed correlation clusters during the 20-year study period.
Geographical subregions formed correlation clusters
Data from December 1998 to October 2019. Based on monthly gross index returns.
The only countries that were relatively uncorrelated to their respective clusters were Israel within developed-market EMEA and Pakistan within emerging-market APAC.
For Israel, the lower level of correlation was not surprising, as it is the only country within its cluster that is not located in Europe and is neither a member of the European Union (EU) nor has formed close trade links to the EU. In fact, Israel’s top trading partners are the U.S. and China.3
For Pakistan, the low correlation with other countries in the same subregion was reflected in other indicators of economic integration: For instance, its trade-to-GDP ratio at the end of 2017 was only 25.8%, below the 39% average for South Asian countries and the 58.3% average for global low-income countries.4
The next exhibit shows average intra-cluster correlations for each country over the past 20 years. Selecting a cluster on the bar graph or the map will highlight the corresponding countries in that cluster.
DM regions had much stronger intra-cluster correlations than EM
Data from December 1998 to October 2019. Average correlations with countries of the same cluster based on monthly gross index returns.
The developed-market Americas cluster was the most intra-correlated, though it contains only the U.S. and Canada, which are economically very closely integrated. Developed-market EMEA stood in second place (despite the exception of Israel) due to the close economic integration of developed European countries. Emerging-market EMEA was the least intra-correlated regional cluster.
How have correlations within subregions evolved?
It’s interesting to look at average correlations over a long time period, but it’s also useful to see how those correlations have changed over time. Within subregions, average pairwise correlations declined from the peak reached during the global financial crisis and were at levels close to their values 20 years ago in most regions.5 The decline in correlation was particularly stark in emerging-market EMEA, where levels were even lower today than in the early 2000s. This may reflect the lessening economic dependency within that region and increasing dependency on developed-market Europe. For instance, as of 2018, only three of Poland’s top 10 trading partners were emerging European countries, while Germany alone accounted for 28.2% of Poland’s external trade.6
Correlations inside clusters decreased over the past decade
Data from December 1998 to October 2019. Average intra-cluster pairwise 36-month correlations based on monthly gross index returns.
What about correlations between subregions?
When we looked at the relationship between average pairwise regional correlations between clusters, we also do not see any upward trend over the past two decades. We also observed a cyclical pattern as we did for intra-cluster correlations. While the evolutions of inter- and intra-cluster correlations were similar overall, we note that there was no clear general demarcation between DM and EM clusters at the regional level indicating high levels of integration over the past 20 years.
Interregional correlations have been cyclical over the past 20 years
Data from December 1998 to October 2019. Average inter-cluster pairwise 36-month correlations based on monthly gross index returns.
Developed and emerging regional clusters have formed correlation clusters. Despite globalization trends that have generally increased economic and trade integration between countries and regions, neither intra- nor interregional correlations have consistently increased over the past 20 years. Given that fact, countries and subregions may still be considered a viable starting point in global asset allocation.
1For example, the MSCI Country Classification Standard classifies markets as developed, emerging and frontier. Regional indexes then group countries into larger geographical blocks.
2This blog post is drawn from a paper commissioned by the Norwegian Ministry of Finance.
3“Israel.” Observatory of Economic Complexity.
4World Bank data and definitions as of end of 2017.
5As of August 2019.
6Poland was Eastern Europe’s second-largest economy; it joined the EU in 2004. For information on its trading partners, see: Workman, D., “Poland’s Top Trading Partners.” World’s Top Exports, April 12, 2019.
Selected geographic issues in the global listed equity market