What Can Stock Correlation Levels Tells Us About Next Day DFM General Index Returns? — GCC Market Analytics
What Can Stock Correlation Levels Tells Us About Next Day DFM General Index Returns? | GCC Market Analytics

Monday 2 August 2010

What Can Stock Correlation Levels Tells Us About Next Day DFM General Index Returns?

What do correlation levels between stocks listed on the Dubai Financial Market tell us about future market activity and price direction of the DFM General Index?  Well, I thought I'd do some testing and find out.




The above chart shows the DFM General Index (red line) along with the 20-day average correlation between a basket of DFM stocks (blue line).  Specifically the 20-day average correlation measures the extent to which the basket of stocks are moving in unison (Note: I used the Spearman rank correlation measure to determine this). 

Interpreting the Correlation Chart

Correlations between stocks are measured over a rolling 20-day period.  When stocks have tended to move in unison over a 20-day period the correlation level will approach 1 (left-hand axis).  When the direction of stock price movements are more dispersed the 20-day average correlation level will be lower in value.  Correlation levels near zero suggest stock price movements over the past 20 days have been varied and moved in different directions.  Well, that's the theory anyway.

Using the 20-day correlation levels between stocks we can investigate if these levels provide any useful information regarding future market action.  For example, the chat below shows the next day performance of the DFM General Index when the 20-day stock correlation is less than 0.5 (blue line) and greater than 0.5 (red line).  (Note: starting capital = $1,000, profits are reinvested, results do not include fees, slippage or dividends):


As can be seen, the DFM General Index has performed significantly worse when 20-day stock correlation levels are greater than 0.5.  When the 20-day correlation has been above 0.5 the average next day return of the DFM General Index has been -0.08% versus an average return of +0.04% for all days.  On the other hand, when 20-day stock correlations have been below 0.5 the next day return of the DFM General Index has been +0.12%, much higher than average.

However, since the market top in 2005 the DFM General Index has fallen in value under both correlation environments.  Since the market top the average daily return for the DFM General Index has been -0.12%.  When 20-day stock correlations have been less then 0.5 the next day returns were -0.07%, negative but much better than the average daily return.  When 20-day stock correlations have been greater then 0.5 the next day returns were -0.18%, much worse than the average daily return.

I think these results tell us more about the expected future volatility of returns rather than the expected direction of prices.  For example, when 20-day stock correlations are less than 0.5 the average absolute  next day return is 1.27% which is much less than the average absolute daily return of 1.48% (with absolute returns we are not concerned with the sign of the returns, positive or negative, just the absolute value of returns).  However, when 20-day stock correlations are greater than 0.5 the average absolute next day return is 1.93%, much higher than normal.

When you think about it these results make sense.  In low correlation environments stock price movements are dispersed with some moving up in value and others moving down. In such environments price movements tend to cancel each other out leading to smaller than normal next day absolute returns.  In a falling market this means next day returns will, on average, fall less than normal whilst in a bull market returns they will rise less than normal.

The opposite is true in higher correlation environments when stock prices tend to moving in unison, whether that be moving up together or down together.  In such environments price returns will tend to be larger than normal.  That's good in a bull market as returns will be higher than normal.  However, it's bad in bear markets where prices in higher volatility environments can be expected to fall more than normal.

Conclusion

This simple look at stock correlation levels and DFM General Index returns seems to suggest that high correlation environments lead to higher than average next day absolute returns.  Instead, low correlation environments lead to lower than average next day absolute returns.  However, on their own these results do not seem to provide any indication of the direction on next day returns, just the extent of next day returns

In future posts on this subject I will go into more detail to see if there is anything else of value in the level of stock correlations and DFM General Index returns.  Specifically I want to cover the following:

  1. This post dealt with next day returns but I want to determine the extent to which stock correlation levels effect returns for longer time periods In this post I divided stock correlations into just two environments; those greater and less than 0.5.  
  2. In future posts I will divide stock correlations into smaller segments and analyse the effects, if any, on future price returns 
  3. Run the same analysis as above but with the addition of a filter or moderator.  For example, use a trend filter to determine the impact of stock correlation levels on future price returns in bull and bear markets 
  4. Perform all the above testing and analysis on other GCC markets 
Anyway, that's all for now.

Enjoy.

Note:
Basket of stocks used for correlation calculation: Air Arabia, Ajman Bank, Aman Insurance, Amlak, Aramex, Arabtec, Takaful House, Deyaar, DFM, Dubai Islamic Bank, DIC, Drake & Scull, DU, Emaar, Emirates NBD, Gulf General Inv., Islamic Arab Ins., Shuaa, Tabrred, Takaful,Tamweel, Union Properties