You could count each n-author paper you are on as 1/n of a paper. Thus each time you submit for publication a 2-author paper, a 3-author paper and a 6-author paper, it's time to referee another paper.

If the journal uses more than one referee per paper, adjust up accordingly. 

In some sense this must be what most people do, on average. 

**Edit:** One could ask the editors how many referees they use per paper. Better yet, the editors should volunteer that information.