Market Correlation

[tommog]

Hi,

Please could you tell me how to workout the correlation between two markets, im guessing its a combination of change in notional values and ATR's over a certain period.

If there is a website that publishes this data even better

Thanks

 

[Rhody Trader]

Many charting packages have correlation studies that will plot it over time, which can be quite informative. Alternatively, you could run it in Excel (there's a function). It doesn't have anything to do with ATRs, though, unless you run it specifically on the ATRs.

 

[tommog]

cheers will look into it

 

[tommog]

Hi Rhody,

One other question if you dont mind (got a bit of a project on the go), if I was running a correlation study on the closing prices of 2 markets, would I have to include normalising factors and spread ratios or would the close vs the close do? Say I was running a correlation study between oil and the dow, would I need to multiply the closes by a spread multiplier?

i have managed to find the excel formula for correlation function (thanks for the pointer) and have downloaded historical closing prices and run a correlation function but it doesnt look right.

Many thanks

 

[cr6196]

i don't know what a "spread multiplier" is, but the way to normalize prices is to look at the percentage return over each period. i would add that linear regression/R-squared is helpful here as well and if your running a long period it may be best practice to look at the log return.

 

[Rhody Trader]

i have managed to find the excel formula for correlation function (thanks for the pointer) and have downloaded historical closing prices and run a correlation function but it doesnt look right.

Not sure what you mean by not looking right. I'd agree with cr6196 that if you really want to do it in a fully normalized fashion you'd want to use periord returns rather than absolute prices. I'd also run the correlation in a rolling fashion, like a moving average, as correlations most definitely change considerably over time.

 

[tommog]

cheers by "not looking right" I was seeing negative correlations in markets that i could see on the chart were both moving in the same direction. Have realised that the data wasnt pulled accurately though, i.e times and dates didnt match which obviously made a difference.

Just wanted to check that close vs close is the correct data and you dont need to adjust the closing prices by a normalising factor. i.e one trades in half ticks the other full ticks etc

appreciate the help

 

[amit1986]

You don't need to adjust the data to compute a simple correlation, but most of academia normally analyses the logarithmic of financial data, instead of just the raw data itself.

 

[tommog]

think i have it now, thanks a lot guys

 

[badi]

remainders:

there are two types of correlation:
positive: A and B go in the same direction
negative: A and B go in opposite directions, I noticed that the late type is often undervalued.

also remember that correlation doesn't mean causality