Hi Dave,

Thanks for reaching out!

I would go for the residuals of the series. The dependence within each original series will probably result in misleading conclusions.

I suggest you look into pre-whitening operations before measuring cross-correlation between series.

Hope this helps.