The golden rule is that you should return at least as much as you consume from the community.
No one likes peer reviews. Pre-publication reviews, post-publication reviews, no matter what — they are all despised, take too much time, and offer no reward for the reviewer. Despite what we were told in grad school, being a referee does not make you a better scientist beyond the first few reviews; enough of this nonsense.
That being said, until we come up with a better system, peer review is a housekeeping task that is vital to our scientific community. Like washing dishes, folding clothes, and sweeping the floor, peer review is just something we have to do and everyone needs to do his or her part. If you are slacking off - you are a parasite.
How not to be a parasite? quantitative guidelines
So how many papers should you review to be OK? Five papers every year? Ten? Agree to every request? For sure, egalitarianism is a poor choice. If I throw a big party every week at our community club and you visit the club once a month, we cannot simply divide the dishes. One size does not fit all.
The golden rule is that you need to give back at least as much as you consume from the community. If everyone obeys this rule, the system is stable and fair. From here, the math is simple. Let n be the number of papers you submit as a corresponding author in a year. As every paper gets ~3 reviewers on average, you consume -3n review units per submission. Therefore, to break even, you need to review 3n papers in this year. If you review less than 3n papers, you are a parasite. If you reviewed 3n, you are OK and can turn down requests without being considered a parasite.
This simple rule covers many cases. If you are not the corresponding author, don’t consider the paper on your reviewer tally. These are not your dishes. Of course, you can do more, but do not feel bad if you can’t. If you delegate a review to a trainee, it is OK to add +1 to your review tally, because you contributed back to the community (the inside tally in your lab is outside of the scope of this post). If your paper was rejected without review, don’t count it. If your paper was reviewed and rejected, we are all V-E-R-Y SORRY, but please clean your 3 dishes. Resubmitted, reviewed, and rejected? Clean your dishes and do not be a parasite!
One subsidiary outcome of our system is that prolific PIs will have more review tasks. This makes total sense — these people are (usually) the experts in their domain. So the system is not only fair but quite beneficial for the quality of peers in the review system.
A working example
Below you can see my combined activity on Nature Methods/Nature Biotech:
I was the corresponding author on 3 papers, one of which was declined without review, one declined with review, and one published manuscript. So basically, I consumed 6 review units (in fact, one paper saw only two reviews). On the other hand, I reviewed six papers. My dishes are all clean in Nature Methods/Biotech. I am not a parasite. I can sleep well at night.
Enforcing the guidelines:
After we establish quantitative guidelines, the question is how to enforce them. One answer is that we don’t need to enforce them. We are all adults here and people should be ashamed if they leave their dishes for someone else. Another option is for publication houses (e.g. Nature Publishing Group, CSHL press) to present a clear tally for corresponding authors about the number of reviewers versus submissions. They should clearly mark if someone is in review debt or credit. Editors could even mention the tally when requesting a review.
Of course, we can always dream of a community-wide system with a real tally based on a blockchain technology (but see below). Each submission will require sending three satoshis to the journal and each review scores one satoshi back to the reviewer. No satoshis? No submission. Come back after doing your dishes.
Take home message: review three papers for each referred submission. You want to review more? Bless you. You review less — you are a parasite.