Ex-US Developed Country Solutions

Overview

These models are based on the MSCI EAFE index of non-US developed country indices. There are 21 countries in the EAFE index of which we use 20. We have developed a solution which has been able to significantly differentiate performance between countries over a one to two year horizon.  The model ranks countries based on a combination of macroeconomic, fundamental and momentum factors.  There is a consistent and sequential relationship between our model rankings and country performance.

The Big Picture

The graphic below plots each country’s performance ranking for each of the past 26 years, shaded by top and bottom third. Casual observation of the graphic reveals a mean reverting process where countries take turns to outperform, with the cycles generally lasting for a few years or more. Measuring trend persistence, countries in the top or bottom 20% have a 30% probability of remaining in that bracket the following year. The columns on right hand side shows each country’s 26 year average ranking and average annual return. The standard deviation of the 26 year averages is only 1.22, implying that performance rankings are generally similar over the long term, confirming the mean reverting process. The average annual 26 year returns for all countries was 8.57% with a standard deviation of 3.56%.

The Opportunity Set

The standard deviation of annual returns between countries averages about 12% per year over the 21 years, but did spike up during the late 1990’s tech bubble. Similarly, the difference between the top 20% (4 countries) and bottom 20% averages 13% per year. Being able to identify those countries likely to outperform or underperform provides significant alpha generating opportunities.

Factors affecting country allocation

Understanding what drives the cycles of performance is the key to our model. Our primary theme is to identify countries which are growing and where the runway for continued growth is positive.

By looking at only developed countries as a separate entity from emerging market countries, we take the view that the risk profile within the developed country group is significantly similar, and thus do not concern ourselves with risk in this model. Secondly, while most country allocation models focus on valuations as a key component, we find no significant predictive ability of valuation to future returns.

Our model uses nine factors which fall into three general categories; fundamental, macroeconomic, and momentum. Countries are ranked on each factor, then each factor is assigned a weight, and then a consolidated ranking score is calculated for each country. Lastly, countries are conviction weighted based upon their percentile ranking.

The country rankings are measured against the subsequent returns for each country. There is on average a 27% positive correlation between our country rankings and one year ahead country returns over the 20 year history, and there is a positive correlation between percentile rankings and year ahead performance 90% of the time.

Separating winners from losers

There is a consistent and sequential relationship between our model rankings and country performance. Performance improves sequentially with ranking score improvement, validating the model methodology. The following charts show, by quintile, the one and two year average forward returns and standard deviations over a 20 year period (1994-2013).

Note the return differential widening from top to bottom quintiles and how the differential is still robust in the second year. Based on the models ability to separate winners from losers we provide two implementation options.

Implementation Solutions

The model is implemented using 20 country ETF’s and is available in two solutions:

• EAFE ex-US Best Ideas Solution #3.  A long only model with our top nine picks, conviction weighted and benchmarked against the EAFE equal weighted average.
• EAFE ex-US Long/Short Solution #9.  A market/dollar neutral long/short model utilizing the top and bottom seven countries.

Note that neither of these two solutions includes a market timing component.