Saturday, March 31, 2012

Surprising Correlation Between STI and Dow Jones Avg

Today, I finally got time to sit down and have a good look at STI and Dow Jones data. As I was fiddling with the data, I decided to take a look at its correlation. After all, we often chat up with each other excitedly in the morning if Dow went crazily up or slid frantically down the previous day, making exclamations and predictions about the current day's STI. So it would be indeed enlightening to find out how exactly is such a prediction based on Dow Jones performance.

Analysis 1 set-up 
Using the data from Yahoo Finance, I first weed out the different holidays to match the dates. This was quite tedious given the messy nature of the data available. What to do, it is still free and much better than nothing. And it seems pretty consistent and error free, at least. The date range for the data is from 24/1/2008 to 26/3/2012. I decided to chop off some data at the start of 2008 because there was quite a few missing dates in the Dow Jones data set. Well, 4 years of data should be good and clear enough for a correlation relationship, if any.

Then, I took only the closing day prices for comparison. Note that this is not the change of the day, and just purely the closing day prices.

Here are the results

1. Correlation between STI closing day price with previous day Dow closing day price (ie. the relationship between Dow's price last night and STI's price today)

r = 0.864

2. Correlation between STI closing day price with same day Dow closing day price (ie. the relationship between Dow's price tonight and STI's price today)

r = 0.866

From a statistical point of view, this correlation of around 0.86-0.87 is close to 1 and that suggests a positive relationship. To me however, even if r = 0.86, it does not make a very smart way to trade as it is not entirely close to 1 either. There must be something more to this correlation than this weakly positive value. I was not satisfied so I decided to take another approach to look at this 'correlation'.

Analysis 2 set-up 
Using another approach, I considered only the direction of move and not the magnitude. So it would be easier to sieve out the nuances of noise and volatility in the respective markets and identify if there was any directional correlation. 

Here is what I found

1. Average number of days where a loss or gain in Dow the night before will be matched with a similar direction of move in STI.

Average number of days = 0.617

2. Average number of days where a loss or gain in STI in the current day will be matched with a similar direction of move in Dow at night.
Average number of days = 0.304

Surprisingly, the correlation between Dow and STI is at best weak. Even though the simple linear regression analysis puts the figure at around 0.86 based on closing day prices of both indices, the impact is clearly even lesser when put to perspective in the direction of move of both indices compared. In fact, for every 10 days Dow increased/decreased, there would be 4 diverging days on the STI. That is really some news for short-term traders looking at strategies to employ. Definitely after looking at this, previous day's Dow prices is an inaccurate prediction of the direction of move STI will make in the current day! The opposite is worse. To use STI to predict Dow's price. Of course, this is not too unexpected given the economic impact Singapore exerts on the US - hardly. 

No comments:

Related Posts Plugin for WordPress, Blogger...