In Chapter 1 of Numbersense (link), I went through an extensive list of shenanigans that can be used to trick the rankings of colleges and graduate schools. One of them is to allow students to submit the maximum of repeated sittings of GREs, LSATs, GMATs, etc.
This tactic is unabashedly headlined in a recent Wall Street Journal article, "Test Redos Give GMAT Scores a Lift." (link)
In the print edition, the article is printed on six columns. Not until the sixth column does the writer acknowledge that "the test costs $250 for each attempt." Allowing redos is an amazing business. So the stars are aligned: the college admissions people want to report higher GMAT averages; the applicants want to feel good about themselves; the test providers rake in the $$$; the test prep industry sucks even more of the $$$.
The casualty is accurate measurement of whatever the GMAT test is supposed to measure. In an accompanying chart, we learn that the average score of all exams is consistently at least 20 points lower than the average scores sent to MBA programs. A 20-point gap for the average test-taker is a huge difference. Part of this gap is selection: perhaps people who do not score above a certain threshold decide not to apply to business school. Part of this gap is surely due to allowing low scores to be dropped.
Here are the rules of the game:
- College admissions staff only looks at the maximum score (why don't they look at the average or median?)
- The test providers allow test takers to strike low scores from their record within 72 hours, thus preventing these scores from being seen by schools
- You can take the test two, three, four, ... times. There is almost no risk of re-taking the test since you can cancel lower scores, and retain higher scores