Why do teachers in higher education grade tests on a curve?
When I worked in industry manufacturing widgets, we did not adjust product quality based on the average quality of the widgets produced (e.g. 70% of inspectable characteristics in conformance to requirements). If we did, all of our customers would have received defective products and demanded expensive repairs, exchanges, or refunds. Likewise, if we evaluate the quality of people’s work performance on a curve using stack rankings, we invite all sorts of difficult and expensive problems. If a student’s course registration process takes 10 steps, and 7 of them worked well while 3 steps failed, would we grade that process on a curve and say that there were actually no bad steps in the process? No, because that would be seen as cheating. Yet, when it comes to the teaching process, faculty freely improve bad test results by adjusting scores higher.
I do not see grading on a curve as a solution to poor test results. Instead, I see grading on a curve as an effective way to hide serious teaching problems from one’s self, administrators, accreditors, and payers. Problems that include poor teaching, poor assessment method, poor test design, inconsistency between what was covered in class and what appeared on the test, etc. Teachers should not use tricks to make things better for themselves and their students. They should not be satisfied with poor test outcomes, simply because they have at their disposal a widely accepted means to adjust test results.
We all know students want to get A and B grades. As a teacher, I want that as well, but that outcome must be coupled with actual learning of the most critical elements of the subject matter. And I want to achieve that outcome without forcing results to look better than they actually are by using artificial means.
If I gave a test that resulted in most students receiving an average score of 70, with some students earning grades above and below that, then I would conclude that I made many serious teaching errors. I would be curious as to what caused the problem and work to eliminate errors that I was responsible for. I would not ignore the causes of poor test scores and raise students’ performance by grading on a curve.
In Lean, we always want to do two things: See reality as it actually is and make errors both large and visible so that they stand out and compel people to understand their cause(s) and correct them quickly. Poor test results makes teaching errors both large and visible to – if one is brave enough to acknowledge their existence.
What about administrators who require faculty to award final grades based on a pre-determined distribution? That demand should be summarily ignored because it hides problems and discourages structured problem-solving. And it also discourages inquiry into individualizing assessment methods that improve student learning outcomes.
The idea of grading on a curve is conceptually absurd. When you have a quantity you can measure, you can use statistical tests of the consistency of actual data with a probability distribution. In a bicycle race, for example, you measure each rider’s gap: how much later than the winner does he or she cross the finish line? Then you can fit a model to these gaps. I tried with Tour de France data for 2011 and 2012, and found that the gaps did not fit a bell curve. “Grading on a curve” means doctoring data to fit a curve.