GMAT Getting Harder

I definitely agree. Take a look at a couple of posts back. I think the JJs have a discernible impact and with the prevalent cheating of some internationals, the percentiles are probably even more impacted.

I guess my point was that startup knows his gmatprep; if I remember correctly, he prepared almost exclusively from it.

Indeed. I'm just guessing, based on past data. I suspect that scores have increased because people have taken GMAT preparation more seriously in recent years. There's a limit to how much time test-takers can invest in preparation, and my guess is we're getting close to that limit. Still, forums like this one are of great benefit to GMAT takers, and there may still be further improvement to come in the test population- we'll see. But I think at some point the score creep will slow, and my guess is we're getting close to that point.



No, that's a myth. The scoring algorithm is not a secret; it's based on a body of psychometric testing theory that's been widely published. I've implemented a simulation of the scoring algorithm, and anyone else willing to do the research could as well. It is true that the first questions 'count more' if you manage to finish the test 'normally', that is, if you have time to attempt to answer the questions at your true ability level even at the end of the test. In that case, yes, the scoring system will have made bigger adjustments to your score at the beginning, and smaller adjustments later. But if you need to guess at several questions at the end of your test, your score will crater, without question, unless by blind luck you guess extremely well. And it's certainly true that questions at the end of the test can 'count' as much (i.e. adjust your score estimate as much) as questions at the beginning of the test.


I'm not sure whether StartupAddict is exaggerating or not, however, he would know since he exhausted the GMATprep more than anyone. He's a bit of a clown who poops himself, but I think he would know about the number of questions on GMATPrep.[/quote]

I'll only say that 5000 questions is an extraodrdinarily large number- far more than the diagnostic tests need to make an ability assessment. Each test could make do with about 350 verbal and math questions apiece- probably less- so I'd guess there are about 1200 questions total in the GMATPrep database. Anyway, I'll wait for his response.

I'd add as well- it is well-known that questions from the pool are not all equally likely to appear on a the test. There are several variables involved when the GMAT determines what question you will see next- the estimate of your ability level (difficulty of the question); how often the question has appeared on other tests (exposure, limited for security reasons); and question content (the test tries to ensure a spread of content, so you don't have 36 geometry questions on your test). Because there are many dimensions to question selection, it can be unpredictable what questions will appear on a test. Indeed, it's normal for about 20% of the questions in the pool to not appear at all during the month of testing. And while the test-designers do as much as they can to limit 'over-exposure' of questions, GRE test-makers found that 20% of the question pool made up 50% of question that test-takers actuallly saw- that is, they showed up 2.5 times more often than you'd expect by random selection. The point is, some questions are much more probable than others, and if you know the answers, in advance, to those questions, you have a very substantial advantage. Knowing the answers to JJ questions would certainly have helped test-takers, and quite significantly.

This makes sense to me. And I have to think that if I had already seen and knew the answers to even 2 or 3 questions on my GMAT, it would have had a significant positive impact on my score. First, "cheaters" would get a problem correct that they may have otherwise gotten incorrect. Second, and possibly more important, each question that "cheaters" would have seen before would bank them an additional minute and half (estimate, of course) to be spent on other difficult questions.

When I consider that on my practice tests I usually missed ~10 quant questions when scoring a 49, I have to guess that reducing this number to ~7 or ~6 (taking into consideration answers that would have been known due to JJ's as well as the extra time available for more difficult problems) would increase my quant score (assuming I'm still at the same difficulty level with these questions) and this could have been the difference between a 740 and a 760 (which I think is significant - not from an adcom's perspective, but from the viewpoint of simply increasing the overall score). And this is all if I had only seen a couple of JJ's before...can you imagine if someone had seen 10 JJ's?

Of course this is all speculation, but these are my thoughts.

I agree with all you're saying, but wanted to mention a study by ETS about the value of having additional time on the GRE test, which I consider reasonably comparable to the GMAT. ETS conducted a study five years ago, in which some test-takers were given 50% more time to complete one quantitative GRE section, and on the 800 scale, they only scored 7 points higher on average than those who had the normal time limit. Weaker scorers benefitted more than high scorers. It may seem counterintuitive, but additional time on the test may be of less benefit than you'd expect.

edit- That's not to say I don't think people benefitted from seeing JJs- in fact, as I've said above, I think they will have benefitted substantially, but I'm not convinced it's because they had additional time to complete their test.

Very interesting, I wouldn't have guessed that. By the way, I find your posts about statistics and studies concerning standardized tests very insightful. Out of curiosity, how did you come across this information? Was it just personal interest and research, or did you have a job function related to it?

Thanks- happy to hear it's of interest. I'm a mathematician, and have been teaching GMAT for several years. I do want to respect this forum's rules about mentioning external sites, but since you asked, I hope it's acceptable to mention that I also co-founded a GMAT/MBA-focused website, which launched recently (gmatix.com). It's really a beta site now; at the moment, its main content is an international directory of freelance GMAT tutors and admissions consultants, which is free to use for both test-takers and tutors/admissions consultants. There'll be some new material on the site quite soon- likely within a week- including some articles based on the research we've been doing into the scoring algorithm and test structure.