Loading...
Symbols:
Authors:

Asset Class Correlations

by: Bespoke Investment Group July 22, 2008 | about stocks: EEM / EWU / FXY / IEF / IVV / IYZ / OIL / SPY / UDN / USO / UUP / XLB / XLE / XLF / XLI / XLK / XLP / XLU / XLV / XLY    

Bespoke Investment Group

About the author: From Bespoke:
Become a Contributor Submit an Article
  • Font Size:
  • Print

Monday's Wall Street Journal had an interesting article about asset class correlations.  With that in mind, below we highlight (click here for PDF) a correlation matrix of various asset classes including the S&P 500 sectors, oil, gold, the dollar, the yen, emerging markets, the 10-year note and the FTSE 100.  The first matrix highlights the correlation between the daily percent changes of asset classes since the S&P 500 peaked on October 9th, 2007.  Each column (vertical) is color coded from green to red based on highest to lowest correlations.

The second matrix highlights the correlations between the same asset classes, only from a much longer time horizon (1990-present).  Then, in the bottom chart, we highlight the difference between the short-term and long-term correlations to see where differences arise.  Correlations that have increased since the bear market began in 10/07 are shaded in light green, while correlations that have decreased are shaded in light red.  In each column, the biggest increase and decrease in correlation is highlighted in dark green or red. 

As shown, correlations have generally increased among sectors, while stocks have become less correlated with oil, gold and Treasuries.  Correlations between stocks and the yen have increased the most in the short-term compared to their long-term correlations.  To view the matrices in PDF form, please click here.  It's definitely an interesting data set to analyze and it's better to let the info speak for itself.

click to enlarge

Correlation721

Related Articles

This article has 9 comments:

  •  
    Jul 22 08:25 AM
    excellent!
    Reply
  •  
    Jul 22 08:48 AM
    the only problem is that correlations are not static and so if you try and build a portfolio based on "asset class correlation" you might find yourself of mark real quick
    Reply
  •  
    Jul 22 11:11 AM
    In the third chart, the vertical column of correlations for 10-yr is marked as slightly less correlation in the column, but largely less correlation in the row with the same numbers. The inverse is true of the yen, the column disagrees with the row, though for increased correlation.

    The numbers are the same, so I guess this is just coloring error.
    Reply
  •  
    Jul 22 02:20 PM
    I dont understand why treasuries are inversely correlated with the dollar. Can someone explain?
    Reply
  •  
    Jul 22 03:02 PM
    DrBagel,
    I think 10-Yr treasuries in the tables are measured in price, not yield. The dollar drops with short terms rates, while gold rises (-0.64), so do bond prices.
    Reply
  •  
    Jul 22 06:34 PM
    Thanks yuman. that helps.
    Reply
  •  
    Jul 22 11:06 PM
    The majority of correlation coefficients shown are actually not that good. When you get down to 0.2 - 0.6, the association between the pairs of variables can still be jumpy. In fact, if you showed the X-Y scatter plots for a number of these matrix elements, you would probably be surprised at how noisy, jumpy and unrelated a lot of the series are. X-Y scatter plots would likely reveal problems and cause readers to ask why you tried to correlate a lot of these pair-vectors in the first place. At values of r=0.8 (-0.8) you will truly begin to see very tight patterns and tight trending between the data being correlated. The goal is to focus on high negative correlation, and the large negative correlation between the dollar and gold is actually good, since you would want to load a portfolio with something that is going to go up, on average, when the dollar goes down. (remember, though, gold is a commodity so the price is inflated in the direction in which the speculative buyers/sellers think it will go. ). Again, the majority of the coefficients are near-zero and low (less than r=0.2 or greater than r=-0.2) and uninteresting. Last, you are probably showing Pearson correlation, which can be biased by outlier pairs. Try using Spearman correlation, which is not biased by outliers.
    Reply
  •  
    Jul 23 07:26 AM
    More Lep in more places.

    What passes for valid numerical analysis is appalling.
    Reply
  •  
    Jul 23 09:21 AM
    Regarding the increased correlation with the Japanese yen the rate you are referencing is the USD/JPY so increases in stocks is related to increased strength of dollar versus yen and not strength in the yen. This could be closely related to leverage being applied/wound down by hedge funds etc (when they are more bullish on stocks) via the carry trade
    Reply
More Comments on SPY ยป
More by Bespoke Investment Group
Investment Resources
Money Managment
Research
Seeking Alpha provides these ads as a service to its leading contributors
Articles on related themes
Market Strategy
Bonds
Dollar/Currencies
Earnings