Clear Admit Livewire data: high GMAT/low GPA MBA applicants

📅 written

25 Nov 17 - UPDATE: The analysis in this article turned out to be pretty well received, so we built a tool to help you dig into the data yourself. We call it the MBA Livewire Explorer-- check it out & let us know what you think!

Splitters--i.e., applicants with high test scores and low GPAs (or, less commonly, vice-versa)--are near & dear to my heart. There's just something 'American dream-y' about standardized testing as a great normalizer/equalizer. But where do they fit into the conventional wisdom on admissions? Since schools are fairly tight-lipped about the distribution of their intakes' scores, the aspiring splitter's left with little to do but weigh nuggets of wisdom handed out on admissions forums about which schools are 'splitter friendly' (whatever that means). After seeing several questions like this on forums like Wall Street Oasis, we decided to dig into a few years of Clear Admit's Livewire data to see if we could find anything interesting.

The data's always dirty

Lots of caveats with this data set that we've explained before, but it bears mentioning again: Livewire data is (a) self-reported by (b) semi-anonymous accounts and, even if it were perfect, comes from what we imagine is a fairly narrow slice of the broader applicant pool. There's also nothing by way of demographic data, which sucks because, as the HBS Guru pointed out in our last Kreisbot post, that stuff matters--a lot.

But, we gotta play the hand we're dealt, so let's see what we've got.

How do you measure a splitter?

"Splitter" doesn't really have a stable definition. Conservatively, you might say it's a GPA and GMAT in opposite quartiles. For example: a Stanford GSB applicant with a GPA in the 20th % of applicants, but a GMAT in the 80th % of applicants, would probably be a 'splitter.' We thought about creating this kind of definition, but eventually decided against drawing an arbitrary line in the sand. Instead, we calculated the percentile of every applicant's scores with respect to the school to which they were applying. Then, we simply subtracted one score from the other to create a sort of 'score-spread-score.' That's probably not crystal-clear, so let's consider some examples.

Example applicants

Consider the following 5 data points, drawn at random from the past 3 years of Livewire data.

First, we determined which percentile each of these scores fell into--that is, which percentile of that school's applicants over 3 years.
Haas3.7 [59.8 %]770 [92.6 %]
Kellogg3.2 [12.1 %]720 [33.4 %]
Wharton3.2 [8.9 %]710 [10.9 %]
Duke3.1 [11.7 %]680 [11.0 %]
UVA3.5 [44.3 %]720 [44.4 %]

tl;dr: here's what the high/medium/low splitter scores mean:

  • 100% = High GMAT/low GPA, a.k.a. a GMAT-high splitter
  • 0% = GPA & GMAT pretty close in terms of percentiles, a.k.a. a non-splitter
  • -100% = high GPA/low GMAT, a.k.a. a GPA-high splitter

The results

Instead of drawing shaky conclusions from dirty data, we decided to just stick to showing distributions. If you're on mobile, you may have to flip to landscape.

Acceptances & rejections

Box plots of 'splitter scores,' by acceptances & rejections / schoolBox plots of 'splitter scores,' by acceptances & rejections / schoolBox plots of 'splitter scores,' by acceptances & rejections / school

If you're not familiar with the box plot, it's a simple way to compare multiple distributions. Khan explains it better than I can, but in short, the high- and low-bounds of the box represent the 25th - 75th percentile, and the top & bottom tails (also called 'whiskers') represent the 'fences' of the distribution--that is, the min/max once outliers (the dots) are removed.

The boxes on the left are all records of type "Accepted" or "Accepted from Waitlist." The right-hand box includes "Rejected" and "Rejected from Waitlist." "Waitlisted," "Applied," and other record types weren't included.

To help us understand what the data says here, consider Haas, UVA, and Tuck.

A box plot of acceptances and rejections by splitter score at Haas, Darden, and Tuck.
  • Haas - Both Berkeley's acceptance and their rejection boxes are around the same size, both with median splitter-scores of around 0%. In other words, there's evidence to suggest that being a splitter neither harms nor helps you at Haas.
  • UVA - The median & low-bound splitter score among UVA's rejections falls more deeply into the negative range--around -10%--meaning more GPA-high splitters were rejected than GMAT-high splitters. Conversely, their acceptances skew a bit in favor of GMAT-high splitters, which could suggest UVA as a 'splitter-friendly' school (GMAT-high, anyway).
  • Tuck - The story's a little different at Tuck. A big majority (~70%) of their rejections were GMAT-high splitters, and a narrow majority of admits were GPA-high. This might make Tuck something of an anti-Darden, if this data's to be believed.

Rounds 1, 2, and 3

Box plot of acceptances by splitter score / school.Box plot of acceptances by splitter score / school.Box plot of acceptances by splitter score / school.

Some notes:

  1. This chart is only acceptances, not acceptances/rejections.
  2. There were no 'accepts' for a few schools in R3.
  3. A good rule of thumb for this plot is: the wider the box & whiskers, the more 'splitter-friendly.'

While (again) this data isn't good enough to conclude much, there are a few points worth mentioning. It seems that, while several schools tighten-up their splitter-score ranges in later rounds (except for Tuck, where I have no idea what's going on), splitters have a solid chance at a few schools in the later rounds.

This is a little inconsistent with internet wisdom, we think: nearly 25% of Tuck's R3 admits have GMAT-high splitter-scores between ~20 - 50 %. To put that in perspective, a 730 GMAT & 2.9 GPA at Tuck has a splitter-score of 44 %--well within range of R3 admits. Looks like there's always hope for the splitter after all.

The moral of the story is just do what you want

We hope this data helps shed a little bit of light (however dim) on a pretty anxiety-inducing topic. Playing with this data made us sad there isn't better data available (yet!), because the internet admissions echo chamber sorely needs some objective analyses.

At the end of the day, we land back on the only bullet-proof advice we've heard on 'where to apply': if you can see yourself happy there, apply.

How'd we do here? See any glaring flaws or room for improvement in this write-up? Any questions you want us to look into? Let us know in the comments, or drop a note in the little chat bubble in the bottom-right.

About admitbrain

admitbrain is the largest source of objective data on MBA admissions & MBA admissions consultants on the web.

Find your consultant