Good question - conventional wisdom says to guess rather than leave an item blank. Even some of the folks at GMAC have mentioned that they're often surprised when they model the CAT scoring about how steep the penalty can be for omitting multiple answers. Plus there's always that chance that you guess correctly, in which case you're that much better off.
For further analysis, check out this blog post from the Official GMAT Blog:
https://www.mba.com/mbacommunity/MBA_com ... ssing.aspx
Ultimately, their advice is that you should probably guess if you're planning on getting a higher score, which hopefully corresponds to most of you here!
____________________________________________________________________________
One other theory, which I've borrowed from my colleague David Newland and which I think has some real merit...
-We all know that the GMAT includes a healthy number of unscored, experimental questions (maybe something like 5-7 per section).
-We also know that the GMAT is a timed test and that the pacing tends to give people some difficulty AND that the timed component is something that is important to those behind the test; they want to see how well you make efficient decisions under pressure.
-THEREFORE, it's highly unlikely that the test would "bail you out" by making your last question or two unscored and experimental; in that event, you'd be given a "mulligan" for poor pacing.
-SO...if you know that you'll need to guess at some point - say you have 7 minutes left for the last 6 questions or something like that - that logic above dictates that you'll probably want to use your guess earlier than later, particularly if you see a problem that would take you plenty of time to solve. There's not much likelihood that question #37 on the quant will be unscored, but there's a reasonable likelihood that #32 will be unscored, so at least in terms of probability you're probably better off guessing on 32 or 33 than on 37, especially because you have the choice on 32 or 33 to say "this is too time-consuming...I wouldn't be able to solve this in 5 minutes" and therefore more-strategically use your "guess". By #37, you're just guessing because you're out of options and out of time, and you may end up missing a question that you could/should have gotten right.
Now, overall your goal should be to avoid this situation! But if you do realize earlier-than-later that you'll need to guess, the logic above seems reasonable enough that I'd recommend using your guess on a difficult-looking question that you don't think you'll be able to solve efficiently, before you get all the way to the end.
