I have two time series of prices. Both price series has increased with more than 100% the last two years.I estimate the crosscorrelation at different lags and I see a significant crosscorrelation between the two time series.I take log returns on both time series, and compute the crosscorrelation between these transformed time series. To my surprise, it is impossible to see any crosscorrelation between the two time series at any lag. I was a bit surprised by this results. Is it not common to look at the log returns, when estimating the crosscorrelation between two financial time series?

Generally, correlation between the price series is not useful in finance. If there is correlation between the return series, whether simple returns or logs or even just differences, you have the potential for profit.Lots of things go up (or down) with time. Population, GDP and the price of beer tend to go up, while the size of MP3 players goes down. But they don't go up or down at the same time or for the same reasons. Knowing that one has just gone up a lot doesn't tell you to place big bets on the others going up or down.Only if the changes are correlated do you suspect an underlying relation, and only then can you make money from the knowledge.

Thanks for the input.The time series I am looking at are oil and gas prices. In the european market there are several gas contracts that sell gas with price linked to the oil price. These contracts make a lot of the total volumes, so there is also strong correlation between oil and gas prices. Often with lag, cause the gas contracts has prices such as the average of the last x month's quotations on an oil product.There is also a strong correlation on the price levels, but it totatlly disappears when I look at the returns. Was a surprise to me!

Sounds like the sample of prices is at different times during the day.Try weekly returns as a correlation input.

The samples were daily closing prices. But it really didn't matter. I looked at average monthly prices, and get exactly the same thing..I agree with the statement from Aaron, but I would assume that there is a correlation between gas and oil prices. However spot prices in energy are a bit complicated... The forward curves show a much stronger correlation!

Another question on correlation. Assume you estimate historical cross correlation to 0,5 for each lag up to 100 days between two time series, looking at returns. All correlation beyond 100 days is zero.How much of the return sin the second time series can be explained by the returns over the last 100 days in the other time series.If you have instant (lag=0) correlation 0,5. You can argue that 0,5^2 = 0,25 =25% of the return is explained by the return in the other series. What is the formula I am looking for here?