Correlation Thinking

The intermarket correlations here among Nikkei, Bonds and S&P are interesting. The most useful aspect of the research? Looking at the correlations for risk management reasons as you assemble a portfolio to track. Remember, diversification is not just simplistically examining the uniqueness of assorted markets, but rather examining how markets zig and zag together or not. That’s a critical issue.

A chart to consider…

15 thoughts on “Correlation Thinking

  1. The problem with correlation is that relationships between instruments change with time and correlation potentially fails you when you most need it: in times of crisis. You might set risk management thresholds based on correlation levels “pre-crisis”, which blow out when the crisis arrives with most products “correlating” more than they have in the past and generating larger exposure than anticipated with correlation analysis…

    To quote Taleb: “Anything that relies on correlation is charlatanism” – probably a bit exaggerated (taleb-style) but the paragraphs on correlation in Black Swan are interesting.

    That chart and article “seem” to suggest that S&P, DJIA will get back in tandem with the Nikkei – a basic “pair trading” strategy, which relies on mean reversion rather than Trend Following…
    And we all know what can happen to some mean reversion strategies (Amaranth, LTCM, Niederhoffer, etc.)

  2. I recently read an interesting post on the trading blox forum about a unique way to view correlation and trading. Rather than just looking at the correlation of the instruments themselves, look at the correlation of an instrument’s performance after passing it through a strategy. You may be surprised to see that even though you can have instruments correlate in price movements, because the correlation is not 100%, you may find that the trading output performance for the correlated instruments result in different equity curves from each other.

  3. Just to second Mike on this… it’s only natural for humans to seek certainty and precision over approximation and estimation. If one visits the forum David mentions one may easily see that for many systematic traders the drive for measurement often replaces the need for discipline and continuous assessment of behavior. In my own experience a system is only useful if it is not held with too much rigidity.

  4. That’s why I find trendfollowing so valuable…these philosophical questions become moot as simply allowing a market to takes its natural course from A to B in its usual inefficient manner and jumping on board with proper money management will eventually lead one to riches.

    I quit reading the business pages three years ago and have done extremely well ever since, especially during 2008. I concluded that riding even a single market up or down and staying out when it’s sideways made a lot of sense.

    If I were managing other people’s money, I would care about correlation to some degree in anticipation of that “black swan event” that everyone worries about, but for myself, I rely on stops.

  5. The correlation between all the stocks market is so huge and powerful, that I’ve created a simple World Wide Market Barometer, mixing the trend signals that I obtain on each one of the 15 idices that I consider (E.g. Sensex, FTSE100, DowJones, Nikkey, Dax, Bovespa, etc).

    The Barometer simply tells: Bullish or Bearish
    and these signals applied to any stock market around the world, produce astonishing results.

    This is the final demonstration that on medium term the markets move exactly in the same way.

    If you want to test what I’m speaking of, you can check this signal history:

    02/09/2004 -> Bullish
    25/10/2004 -> Bearish
    28/10/2004 -> Bullish
    24/01/2005 -> Bearish
    28/01/2005 -> Bullish
    23/03/2005 -> Bearish
    08/04/2005 -> Bullish
    11/04/2005 -> Bearish
    05/05/2005 -> Bullish
    30/08/2005 -> Bearish
    02/09/2005 -> Bullish
    13/10/2005 -> Bearish
    02/11/2005 -> Bullish
    15/05/2006 -> Bearish
    26/06/2006 -> Bullish
    21/07/2006 -> Bearish
    25/07/2006 -> Bullish
    28/02/2007 -> Bearish
    20/03/2007 -> Bullish
    26/07/2007 -> Bearish
    23/08/2007 -> Bullish
    29/08/2007 -> Bearish
    03/09/2007 -> Bullish
    02/11/2007 -> Bearish
    29/11/2007 -> Bullish
    20/12/2007 -> Bearish
    21/12/2007 -> Bullish
    28/12/2007 -> Bearish
    13/02/2008 -> Bullish
    22/02/2008 -> Bearish
    26/02/2008 -> Bullish
    04/03/2008 -> Bearish
    26/03/2008 -> Bullish
    27/05/2008 -> Bearish
    30/05/2008 -> Bullish
    04/06/2008 -> Bearish
    21/07/2008 -> Bullish
    22/07/2008 -> Bearish
    23/07/2008 -> Bullish
    21/08/2008 -> Bearish
    22/08/2008 -> Bullish
    26/08/2008 -> Bearish
    29/08/2008 -> Bullish
    05/09/2008 -> Bearish
    30/10/2008 -> Bullish
    07/11/2008 -> Bearish
    08/12/2008 -> Bullish
    14/01/2009 -> Bearish
    06/02/2009 -> Bullish
    17/02/2009 -> Bearish
    13/03/2009 -> Bullish
    18/06/2009 -> Bearish
    14/07/2009 -> Bullish
    29/10/2009 -> Bearish
    10/11/2009 -> Bullish
    22/01/2010 -> Bearish
    16/02/2010 -> Bullish
    28/04/2010 -> Bearish
    31/05/2010 -> Bullish
    02/06/2010 -> Bearish
    11/06/2010 -> Bullish
    01/07/2010 -> Bearish
    13/07/2010 -> Bullish

  6. @Mike – I agree, you probably need to look at correlation from a worst-case historical scenario to set your risk management accordinlgy (ie make sure that the risk management is not “optimised” but rather allows you to stay alive when it hits the fan and correlation spikes up during crisis times..

    @David, would you have the link to the TB forum thread you are referring to please?

  7. “The problem with correlation is that relationships between instruments change with time and correlation potentially fails you when you most need it: in times of crisis”

    Hold on a second here. It is important to differentiate between price and return cross correlation. You are talking about cross correlation in price. Portfolio theory deals with cross correlation in return i.e noise cancellation. This is very different.

  8. See also Pages 94 – 98 of The Complete Turtle Trader…the turtles were taught to be aware of correlated trades in their portfolio.

  9. DG Dye said “I concluded that riding even a single market up or down and staying out when it’s sideways made a lot of sense.”

    On what basis do you decide whether a market is in a trend or going sideways? easy to spot when looking on the left of the chart, however when trying to figure out the ‘what’s next” right end side of the chart…..

  10. You don”t have to know what’s next. You play the probabilities. I personally don’t like staring at a quote screen so I play the long trends, trading maybe 10 times per year. As Seykota says, you can spot a trend by standing back from the chart. And if you’re in doubt about which way it’s going, then it’s going sideways.

    I see a lot of traders staring at their screens all day and seeing lots of trees but no forest. They typically overtrade and panic at the lows and sell their profits short…the opposite of what they should be doing if they could just stand back and keep their perspective.

    I”m a “math guy” but I find the math I use basically confirms this simple visual principle.


  11. …just to add, to guess at the future is gambling; to speculate using trendfollowing is to see the trend in place and go with it with stops in place in case the trend suddenly ends.

    TF is NOT about prediction.


  12. 1. Michael Covel Says:
    August 5th, 2010 at 10:51 am
    Correlations change, and systems have to be built for change. This is not about tight gloves, but rather loose mittens.
    I couldn’t agree more. I would like to add that, for system trading portfolio of instruments, positions have to become correlated to some degree at certain periods in order to make money. These periods I see as run ups and run downs in equity. Not wanting correlation in portfolio makes as little sense as saying that we want to make money only on a part of any given position. I open every position with an idea to make money on it thus if for example portfolio holds 10 long and 4 short positions I want all of them to be correlated and make money for me at the same time at all times. However I have to also admit to my self that the opposite may happen and build into the systems appropriate risk controls to deal with the exact opposite scenario. In 2008 we saw trend followers who understood this make a lot of money as virtually all futures markets moved up and then down in very correlated fashion.

Comments are closed.