This article is the second part of my first one about the correlation relationships among different commodities and indexes. During the previous research I investigated that such indexes as CSI 200 (China) and Nikkei 225 (Japan) are overvalued now according to the linear regression models with Brent and gold factors. During last 3 days CSI 200 fell more than 11%, Nikkei 225 has not significantly changed.
Source: Google Finance
The main conclusion of my research was based on the 47 months data which did not include the periods of last terrible world crises (DotCom crisis and the Great Recession of 2007-2008).
In this article I tried to:
- Firstly - broaden the research to investigate the correlation change during these crises;
- Secondly - find out if there is an opportunity to forecast the index value for advance if we know the current factor commodity values (lagging correlation model).
In this research I excluded the ASX 200 index and all commodities except gold and Brent.
As in the previous research, main tools were the correlation analysis and the regression analysis. However, the correlation analysis was improved by adding the subtools "moving correlation" and "lagging correlation".
Moving correlation is the correlation for some numbers of data (the so called "window"). For example, if you try to find the moving correlation for the 12 months window on the December 1st, 2014, it means that there are taken 12 previous points, including the point for the December 1st, 2014. In fact, the date range would be from the December 1st, 2013 to December 1st, 2015. Then, the moving correlation for January 1st, 2015 will be from January 1st, 2014 to January 1st, 2015. And so forth.
The lagging correlation is the correlation of 2 samples: the first sample date is today and the second sample date is some date points (e.g. months) from today. Using this method, you may determine if the value of some factor today affects the value of the resulting factor in the future (in some date points).
During using these methods I: 1) found the way that would help forecast the crises; 2) approved that CSI 200 is overvalued.
Body of research:
The first part of my analysis was determining the moving correlation relationships of indexes with Brent and gold. The following diagrams show the moving correlation and the 24 points simple moving average (NYSE:SMA) for the correlation points. Grey fields show the DotCom crisis and the Great Recession.
According to the diagrams I mentioned the following patterns of relationship:
- The correlation for all the indexes is cyclical and rather unstable;
- However, SMAs of all the indexes except CSI 200 (because of lack of data) show that the average period of such a cycle is 4-5 years;
- Crises happened when both the moving correlation and the SMA reached their historical high or low levels (however, not in all cases - did not work with DJIA and Nikkei 225)
- Correlation with CSI 200 shows the change of the tendency to negative which can mean that the economics of China became more dependable on oil price.
The correlation of indexes with gold approves the conclusions for Brent: in the whole, the moving correlation is unstable; crises happen when the moving average is at its high or low levels.
The second part of my analysis was to find the influence of Brent and gold factors on the value of indexes in the future. In the following diagrams you can find the lagging correlation for "0-12 months" (0 means the simple today-today correlation).
And at that moment I discovered rather interesting facts:
- The historical correlation for S&P 500 and DJIA for both Brent and gold is stable for all the lags and is positive; the historical correlation for Nikkei 225 is also stable but negative;
- The correlation for CSI 200 and Brent is unstable and during the time tends to become zero. However, the 3 months correlation is about -0,92 which means a high level of dependence.
I tried to discover the exact formula for this relationship of CSI 200 and Brent. The regression analysis showed that the polynomial regression relationship is more reliable than the linear regression (89% reliability factor against 85% respectively).
The model shows that if the price of Brent tends to be from $50 to $60, the CSI 200 will be from 5500 to 6400 points in 3 months. According to the current price level of Brent ($61,09 on the 07/01/2015) the forecast level of CSI 200 is 5418 points (±400 points) which is 7% higher than the current level (5065) and 4,42% lower than the July 1st level (when my previous research was made).
- Correlation changes during the time and has a 4-5 years cyclical nature
- However, the historical correlation is rather stable for all American indexes and Nikkei 225;
- The correlation for CSI 200 tends to be the strongest at a 3 months lag with Brent;
- Polynomial regression model for Brent and CSI 200 with a 3 months lag shows the 88% of reliability. The forecast for CSI 200 is positive to the current level of index value.
- However, if the oil price will tend to rise, the index value will continue to fall.
My opinion: investors in CSI 200 (companies with high beta or index investors) should have a HOLD position when the oil is at the current levels and SELL position when the oil is high. Investing in American indexes is rather risky because of the changing correlation but nowadays I advise the SELL position as the last trend of correlation is negative.
The data was taken from different sources of information. Here is the table where you can find the links to internet sites.
The date range for Brent, Gold, S&P 500, DJIA and Nikkei 225 is from 06/01/1990 to 07/01/2015. The data range for CSI 200 is from 08/01/2011 to 07/01/2015 (no fair represented monthly or more frequent data was found).
Acknowledgements: I am grateful to wilibear for providing good sources of data.
Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.
I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.
Additional disclosure: All the provided data can not be properly used for making investment decisions until consultation of the professional.