Yes, it needs to be the complementary sequence to hybridize to the microRNA of interest. We published a study of miR-125b that successfully used this luciferase reporter approach (see PMID: 22331826 for details).
Whether the mature miRNA is the 5p or the 3p one doesn't matter with respect to how it targets its target. If you want to test whether your miR-3p targets a particular 3'UTR, you will need to study that 3'UTR, not its reverse complement. After all, it is the 3'UTR, not its reverse complement, that is expressed. If anything, testing the miR-3p against the reverse-complementary 3'UTR would, in a roundabout way, tell you something about whether the miR-5p can target the 3'UTR (except that the 5p and 3p forms are not exactly reverse-complementary, because they might be offset by a nucleotide or two and because of possible G:U base pairing in the precursor and against the target).
The wording of the original post was a bit confusing to me, Goldi. I did not mean to suggest that you use the complement of the 3'UTR. You should be checking whether the target sequence in the 3'UTR is complementary to the 3p form. As Marc said, simply clone your 3' UTR (unchanged) into the reporter.
In addition, you should be checking whether both forms of your microRNA are really present in your cells (the miR Atlas might be of some use here, http://www.microrna.org/microrna/getExprForm.do).
Goldi- you'll need the sense orientation of the 3'UTR, but this reverse orientation construct could be worth testing anyway. At a minimum, it could serve as a negative control. Alternatively, as Marc points out, maybe it could tell you whether the 5p microRNA form has any activity.