The guidelines for Associate Editors largely follow the Guidelines for Reviewers. However, above and beyond developing a standard review of an article, you should craft a review which effectively integrates your own perspectives with the comments provided by the reviewers. This integration includes not only an identification of the major themes and the most important points provided across reviewers, but also a recognition of when a particular reviewer is recommending an inappropriate course of action on the part of an author (e.g., the recommendation of an unsuitable or superfluous statistical analysis or the addition of theory that does not help to clarify the phenomenon under investigation).
Further, when reviewers offer conflicting advice, you should try to clarify these conflicts for the authors ("e.g., respond more to the comments of Reviewer 2 rather than Reviewer 3 regarding issue X"). You should also recognize and communicate to authors when a reviewer is requesting that the authors undertake an action that is beyond a reasonable scope (for example the collection of additional data when the current data are sufficient to test the study’s hypotheses). Finally, "editors emeriti" of both JSCM and other journals have consistently stated that the key differentiator between a good and bad Associate Editor review is timeliness — good reviews are returned by the deadline, and bad reviews are late!