November 11, 2012: Primary Drivers of Asset Class Returns

When thinking about the macroeconomy, I think about two continua. We have changes in the amount of real goods services bought as well as changes in the general price level associated with those transactions.

I created an animation that shows how these measurements have changed over time.

The Permanent Portfolio exploits both axes. Gold does particularly well if the black dot is expected to be moving towards the upper left quadrant. Long-duration Treasuries do best when the black dot is expected to be moving towards the lower left quadrant. Equities do well if the dot is expected to be on the right side of the graph. Finally, T-Bills are held because occasionally the Fed tightens to a level where no asset class is attractive. 

From my perspective, the Permanent Portfolio has exposure to each quadrant. That helps ensure that the overall portfolio will be robust in a wide variety of economic environments. If for some reason I began implementing a different strategy, I would be sure to think about the quadrants. For now, I think the Permanent Portfolio remains the ideal choice for retail investors to passively exploit changes in the macroeconomy.


  1. When I read about the All Weather Portfolio on Seeking Alpha I immediately constructed one using VTI,EMB,GLD,JNK,TLT, and TIP. Seems like the author sort of backtracked when questioned about risk weighted versus value weighted. Using the percentages listed in the article for each of the six ETF's listed above, this still looks fairly safe to me. Backtesting this going back 5 years, (even through 2008), still looks impressive. What do you think?


  2. Hi Jeff,

    Adding more asset classes makes coming up with weightings based off of purely theoretical concerns (what the PP does) much more difficult because some of the asset classes can be seen as a hybrid between stocks and bonds (like high yield debt). However, I made a Python script a couple of months ago that solves the problem quantitatively using historical data from Yahoo. Given a set of asset classes, it finds the portfolio where each asset class has the same weighted covariance to the overall portfolio. This is essentially tells you what the perfect risk parity portfolio would have been for the time period.

    When I ran my script for the asset classes you mentioned I generated the following weightings:
    TLT: 21%
    VTI: 12%
    TIP: 26%
    EMB: 16%
    GLD: 12%
    JNK: 13%

    Try plugging it into your backtests and I think you will agree it is balanced. Who knows if it will going forward. What's really cool about the script is that you can lever / delever the portfolio by changing the bonds to shorter duration. So for example a less risky portfolio could have been.

    IEF: 33%
    VTI: 10%
    TIP: 23%
    EMB: 14%
    GLD: 10%
    JNK: 10%

    This has the same fundamental risks as the previous portfolio... but with less leverage. Fun stuff :)

    1. Ryan,

      That's impressive. Thanks for sharing. I feel like I'm on more stable ground already. BTW, nice animation above. Also, not sure if you have received much feedback, but this site has much potential, and I find it quite helpful. (Found it with your link from PragCap.


    2. Hi Jeff,

      I really appreciate the nice feedback. I am glad the site has been useful for you!


    3. Ryan, is your python script shared anywhere?