Last week, GMAC updated its percentiles for GMAT scores. The organization does this once a year to smooth out any differences in the testing pool.

What do I mean by “differences?” The demographics of the people taking the exam change over time. In particular, over the last ten years or so, GMAC has seen a huge increase in the number of non-United-States-based students taking the test. A majority of these students speak English as a second (or third!) language; a majority also have a better grounding in quantitative skills than the average U.S.-educated student. These differences lead to changes in the data over time.

## Scaled Scores vs. Percentiles

GMAT results are reported using various “scaled scores.” We receive a 2-digit score for quant, a separate 2-digit score for verbal, a Q+V-combined 3-digit score, and two more separate scores for the essay and IR sections.

Think of these scaled scores as “skill levels.” They reflect a specific, measurable level of ability. Here’s the interesting thing: the skills needed to reach a certain level do not change over time. A quant score of 45 today reflects the *same* skill level as a quant score of 45 earned ten or even twenty years ago.

What does change over time is the percentile ranking associated with that score. A percentile ranking reflects how much better you did than a certain percentage of the test-taking population. For example, if you score in the 75^{th} percentile, then you scored better than 75% of the people taking the test—not just that day, or that week, but for the past couple of years (or whatever timeframe is designated for that test).

Imagine that you give a math test to a bunch of 10-year-olds. The scoring algorithm is very simple: if you get a question right, you get one point. You then gather all of the scores and figure out percentile rankings for that group. Let’s say that a certain score (let’s call it 5) represents the 50^{th} percentile. A student who scores 5 earned a better score than 50% of her peers.

Then you take that exact test and give it to a bunch of 14-year-olds. They’re a lot better at math. The same score of 5 might represent only the 25^{th} percentile for this new group, because more of these students have better math skills and can answer more questions correctly. A score of 5 still means the same thing (in this case, 5 questions right), but the pool of testers has changed and so the percentile rankings change too.

This is essentially what happens with the GMAT over time as well. If more people who are good at math start taking the test, then that score of 45 (which represents a certain, fixed level of skill) will drop in the percentile rankings because more people will be capable of performing at that level or higher.

We’ve seen especially big demographic changes on the GMAT over the last 5 to 10 years. In 2006, a quant score of 45 was rated the 78^{th} percentile. Someone scoring at that level had better quant skills than 78% of the people taking the exam around that time.

Today, that *same* skill level of 45 rates the 66^{th} percentile. This does *not* mean that someone scoring a 45 today is worse at math than someone with the same score in 2006; rather, the two students are equally good. Instead, a greater percentage of the *population taking the test today* has stronger math skills.

You might be thinking: oh, great. So that means I have to do even better at math. Actually, the opposite is (sort of) true. Keep reading.

## This Year’s Trends

(more…)