How, exactly, do you determine the word count for a manuscript? This seems to remain one of the huge mysteries of the publishing industry. And I'm going to tell you right here and now that you're going to get a different answer from virtually every person you ask. We at TKA tend to submit in Times New Roman, 12 pt., double spaced, and for the word count I'll actually run a count according to Word, round it off, and label it "approximate." I've had several editors tell me this is pretty much what they do on their end as well, at least up until it gets down to type setting. But the wonder of the computer age is that as long as your word count is in the ball park of what they need, they can make it fit through tiny adjustments in layout, type size, spacing, etc.
Have I heard other methods? Of course. There's the Courier, estimate 250 words per page (which gives you an extraordinarly inflated sense of how long your manuscript is); the counting of words on a selection of pages and then averaging them based on the manuscript's length; and so on. I even heard someone once claim to weigh their manuscripts on a postal scale and figure it out from there through some bizarre calculation of their own. (I kid you not.)
So, in keeping with a policy of providing as much information as possible to the world at large, I offer you
Anna Genoese's explanation for counting words at Tor. It's quite detailed, but fascinating. And an eye-opener for those of you who were under the impression that we in the publishing industry have escaped from doing math...