GMAT Scoring Algorithm - My observations

ive been following along. its amazing to me how much scores and percentile especially in quant have increased.

I extend a hearty thank you to jackson24nj for letting me know that people are still reading this interminable thread.

Hjort

hi hjort,
i read your thread and it good, you will be happy to know it sticky now. keep posting. i like the inforatmion abt the test. i would like to ask you which is the most appropriate day to take test to maximise score
-regards,

Another interesting trend over the last few decades has been score convergence across clusters at the 90th percentile. Thus, scores have not only converged within schools as the range from the 10th to the 90th has fallen, but scores have also converged across schools. This is probably a result of the ceiling effect in that there is little room for the 90th percentile of the ultra elite schools to increase.

It should come as little surprise that many of the elite schools caught up with the 90th percentile of the ultraelites. For example, Duke's 90th was more than 100 points behind Columbia's 90th in the early 1980s but in the early 2000s the 90th percentile values of the two schools were the same. Likewise, UCLA and UC Berkeley had a 90th some 50 points below that of Chicago GSB in the early 1980s. Some two decades later, the 90th percentile at the three schools was essentially the same.

What is more interesting is that schools from lower clusters have come very close to the 90th percentile of the ultraelites. For instance, in the early 1980s the 90th percentile of Columbia was near 700. This was some 80 points above the 90th at Emory. Two decades later, Columbia's "lead" had diminished to only about 30 points.

A similar pattern can be observed with the elite schools as well. In the early 80s Cornell had a large lead of about 60 points on its fellow NY school Rochester. Two decades later, this lead had fallen to just 10 points.

Hmmm these are very interesting observations and analyses. I'm impressed with the information you have gathered regarding GMAT and the business schools. It is very helpful.

For myself, I don't even know what the maximum section score is. I thought it was 51 once, but found that in the scale score table the scores are actually up to 60.

I've also read about conjectures about points being deducted for massive guessing or leaving questions unanswered in the end, etc....

The potential maximum scaled score for verbal and quant is 60. However, on many forms of the paper test even if one answered every single item correctly the maximum achievable scaled score was below 60.

Hjort,
I am very confused about the potential maximum score - If I get every question right on say a Poweprep test in quant and verbal, will my scaled quant and scaled verbal score be 51 or 60?

Have you seen anybody getting more than a 51? If not, why shoudn't we assume that the maximum score is in fact 51?
Is the maximum of 60 a statistical extreme and really unatainable (much like infinity)?

You make a good point regarding the theoretical extreme score. I recall seeing scores of 52 and 53 before. Either way, I would not worry too much about some one else "beating" you with a 52 when you have "only" 51.

An excellent point that should be salient in everyone's mind is that schools take many students with GMAT scores below that school's reported mean. It is easy to become fixated on the mean or even the 90th percentile as the score that must be received. Of course, the same goes for any of the stat.s commonly used in admissions such as GPA and nominal work experience. Just about everything in admissions is negotiable if one presents a compelling case- I've seen students waive the BA/BS requirement, I've seen students admitted to major schools with abysmal GPA, low work experience, as well as students admitted with mediocre GMAT scores. Accordingly, it is a good idea to keep in mind that at least a few students are admitted from much of the entire applicant pool distribution at even the top cluster schools. Thus, it is wise to look at the 10th percentile, the mean, and the 90th percentile for all of the major admissions stats.

Note, however, that some schools have official minimum values for GMAT and GPA. For example, AGSM requires V25, Q35, AWA 4.0, overall 550; and a minimum GPA of 2.0. It is wise to take these requirements seriously.

https://www2.agsm.edu.au/agsm/web.nsf/Co ... quirements

Other schools "prefer" particular values (e.g. 550/3.5 40th percentile for Q and for V for University of Victoria). Again, it is wise to take these benchmarks seriously.

https://www.business.uvic.ca/009_FAQ/

Further, make sure to avoid the "explicit expression" trap >> simply because a school does not announce the presence of its cut off scores does not mean that they do not exist. Schools are reluctant to be explicit about having a minimum "cut off" score since this practice is officially frowned upon by GMAC and schools are under pressure to avoid seeming to place too much empasis on the GMAT. Of course, this does not completely dissuade schools from maintaining cutoff scores - they are just less likely to publish them.

I've found a verbal scaled score of 56 from one of the paper tests administered in the 1980s. This suggests that it is possible to see scaled scores this high in the modern era.

Another important issue regarding the use of the GMAT is the composition of the MBA program. It is easiest to use predictors in lockstep programs where every student must take the same or very similar courses in the first year.

Further, the types of classes is extremely important. It should come as little surprise (although it is rarely emphasized) that the predictive strength of the GMAT can vary greatly with the type of course. For instance, for the "softer" human sciences such as Org Behavior, Business Environment, Leadership etc. the GMAT V has greater predictive power than the GMAT Q (which is often an insignificant predictor). On the other hand, for courses such as Statistics, Data Analysis, or Decison Science the GMAT Q score tends to have a higher correlation and GMAT V is an insignificant variable.

Even at the same business school, first year grades in one class (e.g. Organizational Behavior) might give minimal insight into performance in another class (e.g. Statistics). For instance, one school found the correlation between Organizational Behavior and most quantitative courses to be 0.15 or less. What is even more interesting is that there is no selection variable problem here since the grade in the first class is generally not used to screen entrants to the second class (note that using the GMAT to select students and then correlating GMAT scores with student performance tends to lead to range restriction and understate the utility of the selection variable).

Why Does the Ceiling Matter?

Most of the members of this forum would probably prefer a test with a high ceiling. The reason is simple: if the test has a high ceiling then extraordinary performance on the test would provide an advantage in admissions. If most applicants are clustered near the ceiling, then even the maximum score will have little impact.

The GMAT scale was designed to show 6 standard deviations of performance (3 above, 3 below). However, in recent years both the mean and the SD have been increasing. Accordingly, it can now show scores only about 2.4 SD above the mean.

Yield Characteristics

To the surprise of few readers, the admitted class and the enrolled class differ in significant ways. The students with the lowest GMAT scores have an extremely high yield value, while the students with higher GMAT scores are more likely to go elsewhere. Let's look at the University of Iowa.

At Iowa, students with scores below 600 made up 7% of the accepted students while the 10th percentile of enrolled students was 570. Accordingly, students with scores below 600 were over-represented with at least 10% of enrolled students having scores below 600.

Students with scores of 700 or above made up 35% of the accepted students yet the 90th percentile of enrolled students was 710. Thus, these students were greatly under-represented since they made up 35% of accepted but no more than 10% of enrolled.

The AWA Score

The AWA score ranges from 0 to 6 in half-point intervals (mean:4.0, sd: 1.1). Since there are relatively few steps compared to the V and Q scores, the AWA score disribution is rather lumpy. Further, the distribution is far from symmetrical- only 4 percent score below 2.0 while a 4.0 is the 36th percentile! As everyone reading this already knows, the AWA score is NOT included in the GMAT "Total Score."

For a good summary of the scoring process (intended for the people who will are interested in actually doing the scoring) see the following link:
https://www.ets.org/reader/osn/gmatosn.html

Some conventional wisdom on AWA scores:

Purpose of the AWA: To gain some insight into the abilities of students analyze and argue in standard written English.

Practical Purpose of the AWA: To see if students who submit outstanding essays are actually capable of writing them either
1) Under time pressure
and/or
2) On their own

Acceptable scores: The conventional wisdom appears to be that meeting the 4.0 threshold is sufficient for most admissions purposes.

Toronto Rotman's MMPA program requires a minimum score AWA score of 4.0.

Nottingham appears to require a minimum of 5.0.
https://www.nottingham.ac.uk/quality-man ... g-lang.htm

The following link asserts that the minimum AWA for Emory is 4.0.
https://www.bus.emory.edu/clubs/korea/Ad ... s_Info.htm



General AWA Information

Note how Maryland looks at the AWA score before looking at the college transcripts:
https://www.businessweek.com/bschools/co ... _bs002.htm

An old article, but at that time ASU looked for at least a 3.5 score
https://www.businessweek.com/bschools/bs ... saqa13.htm


The following link describes how the Univ of Tennessee uses the AWA score:
https://www.businessweekasia.com/bschool ... saqa40.htm

The early years of the GMAT

The first GMAT was adminstered in 1954 under the name Admission Test for Graduate Study in Business (ATGSB) and was required by a mere nine schools (five of these schools are now in the Ultra Elite group).

The ATGSB/GMAT was a relative late-comer in the world of admissions testing since the SAT was introduced in 1926, the root of the MCAT c. 1930, the root of the GRE c. 1937, and the LSAT c. 1948.

The name was changed to GMAT c. 1976 reflecting part of the change away from an emphasis on "Business" to an emphasis upon "Management."

The test featured 25 minutes of Verbal, 65 minutes of Quantitative, 30 minutes of Best Arguments, and 55 minutes of Quantitative Reading.

The demise of the GRE CAT

ETS is revising the GRE in 2006. One of the most important changes is that the test will no longer be in a Computer Adaptive format like the GMAT. Instead, it will be a "linear" test that more nearly resembles a traditional paper and pencil examination.

The death of the GRE CAT is interesting since it suggests that ETS and major portions of the psychometric community have lost confidence in the CAT methodology. If you recall from the 1990s, the CAT format was supposed to be some sort of testing panacea that overcame the problems with traditional tests. ETS was in many ways the driving force behind the movement toward CATs. It turns out that CATs apparently had far more security problems than the developers anticipated. The new format of the GRE will consist of discrete administrations in which students take the same version of the test. This should allow both more secure testing and more consistent release of test questions.

Of course, the contract between ETS and GMAC will soon end. Nonetheless, the movement toward internet based, linear, discrete administration tests and away from CATs by one their strongest proponents cannot be ignored by GMAC and its new partners. It will be interesting to see if GMAC continues with a CAT format on the GMAT.

A short article on Harvard's decision to resume using the GMAT in the 1990s

https://www.news.harvard.edu/gazette/199 ... hoolA.html

The fact that Harvard stopped using the GMAT and virtually no one joined them is interesting since test critics have long assumed that if one of the big name school stopped using the test there would be some sort of domino effect to the low prestige schools. According to the theory, the local schools could now follow the example of the ultraelite school and be "liberated" from the tyranny of the GMAT. The Harvard experience was not consistent with this theory.