International ETFs: Correlations with Broad U.S. Market

Assembling assets with low correlation is fundamental to reducing portfolio volatility, a key type of risk. We have shown in prior articles that most US stocks by market capitalization or by style (value or growth) have high correlations to the broad US stock market (here represented by the Russell 3000). We have also shown in prior articles that some US industrial sectors have varied correlation to the broad US market, some with only modest correlation. This article reviews the correlation between major world regional stock markets and the broad US market.

Correlation is not a static relationship. Correlations change and shift over time. For that reason, we analyzed three time periods; 1 year daily, 3 years weekly, and 5 years weekly. Because the ETF or CEF proxies for the regional indices have not all existed for 5 years, we analyzed the indices themselves (except for India where we analyzed the CEF which does not follow an index). Since the tracking error by the proxy securities is low, we feel that index correlation can be reasonably used as a substitute for the proxy securities.

The China stock market has much lower correlation with broad the US market than India. EAFE, Europe, Latin America and India have the highest correlation with the US market. Emerging markets, Pacific Ex Japan, Japan and China have the lowest correlation with the US market.

Comments

[simon cassidy:]

Hi Richard,

I've been looking for correlations similar to the ones you provided but have come up with vastly different numbers. I don't think that any of the ETFs you mention should have a correlation (based on 1 year) to the U.S. of close to zero. PortfolioScience.com lists most of these correlations as above .50, while State Street lists some at above .90 (!). While I'm not sure how accurate Portfolio Science is, State Street and IShares seem way off. Any thoughts on this? I'd love to know what the right numbers are.

Simon

2007 Jan 30 01:13 PM

[QVM Group:]

The first issue is that the State Street online correlation tool measures price correlation, whereas the iShares online correlation tool measures return correlation. They are different measures and develop quite different results.

Price correlation tends to be used by short term traders to find highly correlated securities that experience temporary mispricing for arbitrage opportunities. Return correlation is used by long-term investors to find low correlation securities to reduce portfolio risk.

This matter is discussed more fully in a prior article: etf.seekingalpha.com/a... which also provides instructions for hand calculation of correlations. You might try recalcuating one or two of the numbers yourself to be certain of the numbers.

The correlation figures in this article were taken directly from the iShares online calculator to save time. They do, however, specifically state on their site that they measure correlatiion of returns, unlike State Street whose site says they correlated prices (and upon inquiry verify that they do not correlate returns).

2007 Jan 30 03:46 PM

[simon cassidy:]

I hadn't realized how big the difference between correlations for returns and prices are. Thanks for directing me to the other article. However, the figures still seem off. When I ran a one year correlation for IWV and EFA using daily returns, the figure was .80. When using prices, it was .94. Both of these are a long way from the .09 IShares claims on their calculator. I'd guess that EPP, EEM, FXI, etc. are also off. These figures would drastically overstate how easy it is to develop a diversified portfolio.

2007 Jan 30 10:47 PM

[QVM Group:]

Simon, I warned against using black boxes in my other article and possibly I have been sucked in again by using the iShares black box. I believed them when they said they correlate returns and they are doing if for their own products. Maybe I have been a sucker twice, and maybe not. I will run some manual tests to see how they come out myself. It is also important to note that the period used has an impact on the results too (whether by minute, by hour, by daily, by week, my month, by quarter, etc).

Tomorrow, re-examine the iShares site, cross-check a few using the technique described in my other article and post a response of some kind. Thanks for sticking with it.

2007 Jan 30 10:56 PM

[QVM Group:]

Simon,

Here's the story on the correlations.

First, the language on the site tool on which I based my belief that Barclay's is correlating returns as they should be is shown below. I have no reason to believe that they did the math incorrectly -- it's just the data on which they calculate that should reasonably be questioned.

"Correlation: A measurement of how closely a portfolio's performance correlates with the performance of a benchmark index, and thus a measurement of what portion of its performance can be explained by the performance of the overall market or index. Readings for correlation may range between -1 and +1, where +1 means perfect positive correlation and -1 means perfect negative correlation."

Second, I took the position as the #1 ETF sponsor in the world with enormous credibility at risk by offering the tool in their "financial professional" section of their site, they would have validated their methodology. However, you correctly note that State Street numbers vary. I have previously discovered and published a report indicating that the State Street site which does not correlate returns (they correlate prices only – verified by phone contact with State Street) and therefore is not useful for my purposes or relevant to the correlation focus of this article.

Third, I did not recalculate by hand on my own, but I did contact the Barclay's wholesaler division to verify their methodology. I gathered information from two wholesalers after telling them about the challenge to my published numbers, about the difference between their numbers and those from State Street, and how the State Street tool does not correlate returns. They assured me that the Barclay's tool correlates returns not prices, and I am confident that their math formula must be right.

Therefore, I rest comfortable that the numbers are correct. If the 1-year numbers look funny, it may be the shorter time intervals produce result less in step with numbers from longer periods of time.

I encourage you to do some manual calculations with the Yahoo data source and Excel method listed in my other correlation article. etf.seekingalpha.com/a...

If numbers you calculate contradict the figures from Barclay's we can take it up with them at another level. If you would like to call them yourself, they can be reached at 800-474-2737.

State Street numbers are irrelevant for the purposes of this article and should be ignored. As for PortfolioScience.com, I cannot comment, because I do not have access to their tool and did not sign up for their free trail.

As a practical matter, virtually all of the data we use in investment decisions is provided by somebody who we rely upon to use reasonable care – correlations, portfolio statistics, even basic things like prices and dividends, and of course narrative reports and news. We live in a complex world that requires that we rely on each other and we cannot do everything from the ground up individually. This is one of those cases. I’m going to go with the assumption that the #1 ETF company with a portfolio construction tool based in part on correlation has been developed properly and with care. If someone can show me a specific error in their tool, I will change, but until then, life is too short.

I appreciate you comment and the fact that you caused me to voice verify the Barclay’s methodology with their staff.

2007 Feb 01 04:49 PM