Ask Stacy Blackman

As a distraction from my studies, I have been trying to do some analysis on what really matters between SCORE and PERCENTILE. I'd like to share my findings and ask for your comments and views.

For those who have studied the GMAT for some time, you will know that:

• 1. loads of people want a GMAT score of 700 or above
  2. slightly fewer people know that a 700+ is around the 92nd percntile mark

FIRSTLY - Psychological barrier between 6__ and 7__
Lets say that a 700+ is at the 92nd percentile mark. Hypothetically, if you achieved this score and were part of a class size of 100, you would be amongst the top 8 "cleverest" people. Now imagine your class size when you were around 16-18 years old; my class size was roughly 30 and applying the same logic, you could say that scoring a 92nd percentile makes you one of the top 2 in the class.

When business schools look at your application, it is widely accepted that GMAT not the sole contributor to their overall decision. However, do business schools look at your actual GMAT score or do they look at your percentile ranking?

Although they should consider the percentile ranking, I believe that due to psychological barriers, admission committees look at your actual score. I.e. a score beginning with 7__ resounds in their mind more than one that begins in 6__ (even though the applicant may have a 690 which for argument sake is a 91st percentile score).

SECONDLY - It's easier to get a high score if you are already getting a high score
I believe the GMAT rankings occur on a normalized bell curve. For a test taker, this basically means 2 things (and I may be wrong on this):

• 1. if you are at the average mark (say 540), it is more difficult to move up the percentile scale
  2. if you are at the upper end of the curve (say 700), it is easier for you to move up the percentile scale

On a bell curve, candidates are distributed with the largest number of candidates obtaining the median score. The higher up you go, the fewer the number of candidates you are grouped with. For example, 10,000 people are assigned a 50th percentile score; 9,000 people are assigned a 60th percentile score; 7,500 people are assigned a 70th percentile score; and so forth. Here is the gem... when you get to the 90th percentile, you may be in a group of say 100 people so you need to beat at most 99 people in order to get into the next percentile bracket!

Therefore, am I accurate in asserting: the higher your ranking, the fewer number of people you have to pass in order to get the next percentile.

----------
Have I been clear in my analysis? Can anyone help me investigate, comment or poke holes in my logic? PLEASE!!!

My questions...

• - Do business schools look at your actual GMAT score or do they look at your percentile ranking?
  - should one study until one is sure of obtaining a 700+ or should one be content with say a 670 (89th percentile)?
  - Am I accurate in asserting: the higher your ranking, the fewer number of people you have to pass in order to get the next percentile?

It is interesting reading why you said, I think people target 700 because it is totally pyschological, it's actually a simple number and it looks good. Also you can always say I scored in the "seven hundreds."

Edit: Forgot to add, because of how 700 is the benchmark now, if you score a 700, no one will ask if you are gonna take it again. 690 is a questional retake.


However, I don't think you really understand the percentile. Actually there should be equal amount of people who score in the same percentile. If there is 100 people in the 99th percentile, there is also 100 people in the 50th percentile. Same with 100 people in the 2nd percentile. (Yes I understand that because of how the test is set up, this isn't always true) But for the case of you saying it is easier or not, it will be true.

Let's put this in real terms.

Let's say you scored a 730 (97th percentile), and on the next exam you scored 10 points higher. At 740, you would be in the 98th percentile, or 1 percentile gain.

However let's say you scored a 540 (50th) and next exam scored also 10 points higher, at 550 you would be 54th percentile, or a 4% gain.

So I would argue the reverse, it is much harder for someone at the top to move up then it is at the middle.


Also, another aspect is if you have 5 questions that you didn't know, do get a 750, you would practically need to guess all of them right, while at 550, you can really get by with 2 or 3 correct.



I really hope this helps. IMO, you really just try the best you can, and forget the scores. I don't think a bad score should really make you change your mind about which school to apply or not.

Can someone verify whether GMAT scores are normally distributed or simply ranked. I am fairly certain that GMAT scores are normally distributed with a mean of about 500. In other words, according to rules of statistics on a normally distributed curve, the following apply:

• * 50% of test takers obtain a score above 500
  * 15.9% of test takers obtain a score above 600
  * 2.3% of test takers obtain a score above 700

In fact, I think the SAT, LSAT and GRE also use a normalised curve.

The GMAC website states "Total GMAT scores range from 200 to 800. Two-thirds of test takers score between 400 and 600." https://www.gmac.com/gmac/thegmat/gmatscores/

I found this on the internet: https://www.uic.edu/classes/mba/mba503/971/503nts5.htm

How can GMAT scores have a normal distribution? The answer is that the raw score (number of items correct) is converted to a score in a normal distribution. This is done as follows. In developing the test a large number of people took it. The median of their raw scores was assigned the number 500. The 95-th percentile of their raw scores was assigned the number 664, which is the 95-th percentile of a normal distribution with a mean of 500 and a std dev of 100. Etc.

Unlike GoBruins (see quote below), I'm not convinced that the same number of people score in each score. E.g. I don't think say 100 people score a 700 and a similar number of people score 600, 500. Due to the nature of a normalised curve, I think that a disproportionate number of people score near 500 and very, very few score 700.

GoBruins wrote:...understand the percentile. Actually there should be equal amount of people who score in the same percentile. If there is 100 people in the 99th percentile, there is also 100 people in the 50th percentile. Same with 100 people in the 2nd percentile...

I'm keen to hear an expert view on the scaling and normalisation of the scores. Any other opinions?

Okay let me clarify what I have said. If there are any experts, please chime in. However I think you misinterpreted some of my statements.

Couple things I will agree with you,
GMAT for all intersive purposes is normally distributed. It isn't gonna be a true bell curve, it's a somewhat skewed bell, and there are limiting factors such as a maximum of 800. But I will say you can assume it's a bell. (agree?)

I never said the same amount of people score 700 as score 500, what I said was 99th percentile and 50th percentil is the same. While scores of 760, 770, 780, 790, 800 all are 99th percentile. 500 is everything from 50th percentile to 54th percentile, which means that there is 5 times as many people who score 500 then there are 760-800 combined. (agree?)

Third point, for your 50%-500, 15.9%-600, 2.3% above 700, how did you come up with those numbers, are you just assuming standard deivation as beeing 100? I read your quote from UIC, and I believe it to be complete false unles it was just an example. BECAUSE, a score of 700 is 91-92% there for 8% of the population is ABOVE 700 (NOT 2.3%) Also 664 is not the 95th percentile, which is closer to 710-720. Therefore I disagree with that example. In fact, you can just see what the actual distribution is on the GMAC website.


But I do agree that the exam is normalized, and that they control the popularion, and that the curve is fairly close to the 'normal' bell curve for all statistical purposes.

With the gmat its all about score. My Math was 500's and my critical thinking 700's not to mention the essay. It gets chopped up. As far as a bell curve we all remember from undergrad statistics that it can be sig. skewed to the left OR right. And the test questions increase in difficulty as you answer correctly and vice versa when you get them wrong they decrease in difficulty. P.S. Most students with an MBA will not be able to find worked unless they have very good connections. Be prepared to drop the MBA of the resume, everyone has an MBA now. Supply and demand.