identify peers within a firm - sparse matrix only

Dear all:

This is a follow-up post from my previous one, which has been solved by @Image Analyst. Now I have encountered a new problem as I have increased my sample size by a lot.

The original problem is described as follows. The dataset contains worker ID and firm ID in each year. I hope to create a matrix that shows whether worker j is worker i's coworker. For example, if there are 5 workers and 2 firms in the dataset, the data structure look like below.

IDw IDf

1 1

2 2

3 1

4 2

5 2

Ideally, I hope to create a matrix C as follows:

As suggested by @Image Analyst, I could use the following code to create such a matrix:

firm2 = [firms(:), firms(:)];
distances = pdist2(firm2, firm2) == 0;

However, the code pdist2 requires a non-sparse matrix first. In my current case, it is a 0.7 million x 0.7 million matrix. My MATLAB could not handle such a matrix if it is non-sparse.

Therefor, I am wondering if anyone knows an alternative (quick) algorithm that could help me get a sparse matrix of C?

Thank you very much and I look forward to hearing from you!

Best,

Long