Thanks guys.

The first solution was efficient and it is working fine.

The resulting dataset is around 4gb for UK (still have to do US, which will be +-6gb). I know it is not a big dataset for a High Performance Computer, but for my PC it is touching the limits :).

I used the treatment example to help me explain more easily, but the real data are UK and US companies announcing information. Not sure about the next step, I did this as a one-time job and data was requested to me in this format. I am guessing they want to compare the returns of a large set of announcing companies with their non-announcing peers, based on number of yeas that a company is being traded. (age in my example).

Thanks for the help, as usual you have been great.

PA