A long time ago, in a blog far far away, I ran a small poll about paper refereeing. The poll asked "What is your ratio of reviewed to submitted manuscripts?". The results were
- >=6 reviewed for every 1 submitted: 7 votes (8 percent)
- 5 reviewed for every 1 submitted: 3 votes (4 percent)
- 4 reviewed for every 1 submitted: 9 votes (10 percent)
- 3 reviewed for every 1 submitted: 12 votes (14 percent)
- 2 reviewed for every 1 submitted: 13 votes (15 percent)
- 1 reviewed for every 1 submitted: 20 votes (24 percent)
- 1 reviewed for every 2 submitted: 6 votes (7 percent)
- 1 reviewed for every 3 submitted: 5 votes (6 percent)
- 1 reviewed for every 4 submitted: 2 votes (2 percent)
- 1 reviewed for every 5 submitted: 0 votes (0 percent)
- 1 reviewed for every >=6 submitted: 7 votes (8 percent)
This works out to an average 2.2 papers reviewed for every one submitted.
But the question I didn't ask is what should your ratio of reviewed (refereed) to submitted be?
I think, the way the question is worded, it is too open to interpretation. For instance, an eminent physicist near the end of his or her career might not be publishing much anymore, but might enjoy reviewing and thus would expect his or her ratio to be weighted toward reviewing. On the other hand, someone at the beginning of their career might be spitting out papers left and right but, until they're recognized, might have not been asked yet to review many papers and would fully expect that and thus would be inclined to answer with a ratio weighted toward submissions.
The answer to this question depends heavily on what you mean by "submitted". If you count only first author papers, then the answer is "3", since each paper should have 3 reviewers. However, if you count all papers, then it'll depend on the number of authors on your papers. Assuming everyone follows the same strategy, you should review 3/N papers for each N-author paper you co-author.
I'd consider going higher to make up for all the slackers out there. :)