Yes, but you should be protective of your time. Here is my algorithm.
We have a duty to review papers, but how many?
How many reviews have you consumed?
My rule of thumb is that I will serve as a reviewer for each time when someone else was asked to serve as a reviewer for my paper. Therefore, if I have 4 first author papers, and each was reviewed by 3 people, then I have a duty to review 12 papers. When doing these calculations, I will discount my obligation if I had co-first-authors on the paper, but not for minor authors or senior authors. Senior authors are typically occupied with other obligations, such as serving as editors for journals.
Will you benefit from this review?
Are you interested in the topic? If so, you are getting the first peak at this research, along with an opportunity to get the authors to respond to your questions. Do you respect the editor, and are you happy that she considers you an expert?
Pay attention to how much time you spend on a review
I have a tendency to spend 5-10 hours on a review (biology), but have been told that 2-6 hours is appropriate. I am trying hard to compress that. The appropriate amount of time may depend on your field. A mathematician colleague of mine said that he spends several days reviewing each paper. This is feasible for him because publications in his field are very rare. Sometimes I spend a lot of time because I am interested in the topic but have not previously bothered to read the background literature, so I read several papers while reviewing the one. This may be a helpful or harmful habit for my career. Refuse to review a paper if it will require background reading that you are not interested in.
Don't review unless you can get it done immediately
If you can't make time for it in the next few days, then you probably don't have time for it. It's best for everyone if the review is returned to the authors ASAP.
Reject bad papers ASAP
Refuse to review papers with bad abstracts
I have reviewed a couple of papers even though they seemed pointless based on the abstract. It turned out that they were indeed pointless. My new policy is to refuse to review any paper with a bad abstract and write to the editor that I do not consider the paper publishable. I have not had the opportunity to do this yet, so I don't know how editors actually respond. However, I think this is a legitimate basis for rejecting a paper. If the abstract accurately reflects the content of the paper, then the paper is indeed pointless. If the abstract fails to describe what is notable about the study, then the paper is poorly written and is not ready for publication. Some journals say that their papers are not to be evaluated based on "impact", but if a paper is as pointless/trivial as the stuff I've seen, it is not worth the effort of reviewing it, and therefore it is not publishable.
Try to identify a fatal error quickly; if you find it, stop reviewing
Nobody benefits from detailed nit-picking on a paper that is not going to be published anyway.