Absolute vs. Relative Grading System: Pros, Cons, and Experiences

[gen x]

I am currently weighing the pros and cons of the absolute grading system versus the relative grading system (grading on a curve). I would love to hear your insights, experiences, and academic perspectives on this. Specifically, how do these systems impact student motivation, fairness, stress levels, and cooperation among peers in rigorous STEM fields? Which system do you personally prefer, either as a student or as an educator, and why?

If you use relative system so average must fit in 60-65%, now you get every year have same average grade, but every generation is different...So what is point of this, how can you judge how this generation is bad or good if every year you have same grades?

here are some links about problems in grading systems:

https://academia.stackexchange.com/...-too-high-for-the-department-what-should-i-do

https://academia.stackexchange.com/...at-all-are-there-any-arguments-in-favor-of-it

https://academia.stackexchange.com/...s-without-creating-competition-among-students


https://www.linkedin.com/pulse/impact-relative-grading-university-education-sohail-khan-ph-d--q8enf

https://www.linkedin.com/pulse/abso...undergraduate-education-rajendra-prasad-hp5cc

 

[.Scott]

You haven't made it clear whether this is college or high school. From your links, I am suspecting college - but I cannot tell for sure.

I think that the first thing that needs to be said is that grades (numbers or A, B, etc) are a terse way of describing some part of the learning situation. When I tutor, I don't give a grade - I just relate the progress to the learning goals in narrative terms. A "grade" focuses on some piece of that narrative.

Just for perspective (practical or not), I would start with a baseline description of what one would presume is happening: There is an institutional objective to the course. By the end of the course, some target goal or goals may be achieved and those goals should be expressible as a statement about the student that goes beyond "this student got a B-". Those target goals are development from stakeholder requirements:
a) The school "certification" standards - and how you course fits in with them.
b) The school itself. Its organizational methods, it public image, its business objectives.
c) The students. Are they there just for grade or are there actual learning requirements they will need.
d) You. What do you want to contribute to the course - for no other reason than to make your job easier or more rewarding?

In the easiest case, all of these requirements naturally align and are consistent with a particular grading method. I am guessing that that is the exception.

If the course you are teaching is a prerequisite to other courses or to a third-party achievement test (scholastic, professional licensing, etc) and the only (or the prevalent) goal is to serve as that prerequisite, then some approximate pass/fail criteria can be described. Presumably, any student that is ready for those follow-on courses/trials should pass and any that are not ready should not. Alternatively, you could pass them with the certain knowledge that they are so soured to the subject that they would never attempt to tackle those follow-ons ... (not my recommended strategy).
Beyond the final grade, interim grades provide feedback - encouragement, progress measurement, etc. So those interim reports are intended as a coded message to the student, parents, school, or other stakeholder and are used (intentionally or not) to reflect the progress, encourage the student, satisfy the schools image, or address issues with the schools credentials. That can be a lot of messaging for a single 2-digit number.

If the content of the course you are teaching is simply intended to provide general support for an unknown future employment or other such mission, then you need to recognize from the outset that you are not standing on rock. And if there will never be any tracking for how your students ultimately succeeded in their mission, you're probably standing on sand.

If you're standing on sand, there really is no solid argument against "grading on the curve". If it pleases any of the stakeholders listed above (including yourself), then why not?

Decades ago, I did teach as part of my job. It was a 2-week course (80 hours) with a very practical objective - the ability to add new features to an existing CAD/CAM system.
I wrote the entire curriculum. The only exam was at the start of the course. The stated prerequisite was experience with Fortran programming - and the students were asked to demonstrate that skill. If the failed, then I asked if they would prefer to spend the next two weeks dining out with sales/support people who could answer any of their questions. Otherwise we got down to business.

 

[gen x]

You haven't made it clear whether this is college or high school.

Yes I am talking about university, not high school. Especially for math-physics-engineering.

If you're standing on sand, there really is no solid argument against "grading on the curve". If it pleases any of the stakeholders listed above (including yourself), then why not?

How do you then explain that USA, India, South Korea, Oxford, Cambridge STEM etc use relative grading?

 

[kuruman]

So what is point of this, how can you judge how this generation is bad or good if every year you have same grades?

I view the administration of a test as a physical measurement of how well a student can do physics. In other words, I set the goals and then I test how well a particular student has met these goals. The grade I assign is a numerical representation summarizing that assessment. I use the standard scale
90 $\leq$ G $\leq$100 Range of A grades
80 $\leq$ G $<$ 90 Range of B grades
70 $\leq$ G $<$ 80 Range of C grades
60 $\leq$ G $<$ 70 Range of D grades
G $<$ 60 Failure

To effect the measurement,

• 
  I construct the test and assign numerical values to each question or part thereof.
• I add all the values to get a test total, say 84 points, (I stay away from totals of 100 to avoid preconceived ideas about scores.)
• For each question, I set grading criteria, i.e. what correct statement (if present) is worth how many points.
• I then review the test, question by question, asking myself "What is the very least that students should be able to do here and how many points does that correspond to? This sets a minimum score for each question.
• I add all minimum scores and get, say, 18 points out of 84.
• I then construct a graph showing a piece-wise continuous straight line with Raw score (0 - 84) on the abscissa and Scaled Score (0 - 100) on the ordinate as shown above on the right.
• Finally, I use the graph to map the raw scores onto the standard numerical and letter grade scale shown above. All of this is done, of course, on a spreadsheet.

There is no curving based on preconceived distributions. From exam to exam, the maximum raw score need not be fixed and neither is the passing line (scaled 60). Pluses and minuses are assigned to scaled letter grades strictly depending on where, within the letter grade interval, the scaled score happens to fall. For example, $70 \leq \text{G} < 73\frac{1}{3}$ translates to a C -.

I have found this system of assigning grades satisfactory and I see no reason for adopting a curve. My expectations for a test are set before administering it. If all students perform at A level, they will all get A's; if they all perform dismally, they will all get F's. Neither happened in the 39 years that I taught university-level courses. In my mind, a curve is not an honest assessment of a student's performance.

I viewed grading as an integral part of my teaching. I was hired to teach and to assess how well the students mastered what I taught. If anyone told me, "department policy" requires me to assign grades in a certain way, I would dump the entire stack of tests on their desk and say, "if you don't trust me to do my job right, grade them yourself."

 

[gen x]

I viewed grading as an integral part of my teaching. I was hired to teach and to assess how well the students mastered what I taught. If anyone told me, "department policy" requires me to assign grades in a certain way, I would dump the entire stack of tests on their desk and say, "if you don't trust me to do my job right, grade them yourself."

I didn't even know that relative grading exist, so I was shocked when read that some uni use this system, because to me at first looks like cheating.

Does it make sense use relative grading if exams are set brutally hard(genius level) so no one can get above 80%, in that way you are secure that all class will not fail? Because this hard exam using absolute grading everyone will fail.. So system is set to produce/find new Newtons, Einsteins..

What do you think why they use relative grading? Must be some reason, it is not possible that all people who decide this are all stupid and dont know what they are doing..

 

[symbolipoint]

I didn't even know that relative grading exist, so I was shocked when read that some uni use this system, because to me at first looks like cheating.

Does it make sense use relative grading if exams are set brutally hard(genius level) so no one can get above 80%, in that way you are secure that all class will not fail? Because this hard exam using absolute grading everyone will fail.. So system is set to produce/find new Newtons, Einsteins..

What do you think why they use relative grading? Must be some reason, it is not possible that all people who decide this are all stupid and dont know what they are doing..

To setup identifiable or strict basic standards the way kuruman ( post #4) describes may be best.

 

[Andy Resnick]

I am currently weighing the pros and cons of the absolute grading system versus the relative grading system (grading on a curve). I would love to hear your insights, experiences, and academic perspectives on this. Specifically, how do these systems impact student motivation, fairness, stress levels, and cooperation among peers in rigorous STEM fields? Which system do you personally prefer, either as a student or as an educator, and why?

I think it's worth pointing out that (at least in my experience as a university STEM educator) that we instructors receive no guidance or training or education on this point; having a PhD in a subject matter is taken as sufficient evidence that I am imbued with the ability to teach the subject.

Both absolute and relative grading systems have serious deficiencies; if I use an absolute grading scheme then grades will sensitively depend on how accurately I can match the difficulty level of problems to student capability, whereas if I use a relative scheme then some students *must* fail, regardless of how hard they work.

My 'official' grading scheme is the following:

A	Work of excellent quality
A-
B+
B	Work of good quality
B-
C+
C	Work of satisfactory quality
D	Work of poor, but passing quality
F	Failure to meet the minimum standards for passing the course

Before someone points out that this grading scheme is entirely subjective (which it is), please keep in mind that *any* grading scheme is purely subjective: who picked the questions, and why were those specific questions chosen?

As for "how do these systems impact student motivation, fairness, stress levels, and cooperation among peers in rigorous STEM fields", I would refer you to the enormous body of existing literature (71k results!)

https://scholar.google.com/scholar?hl=en&as_sdt=0,36&q=STEM+grading+schemes&btnG=

 

[kuruman]

Before someone points out that this grading scheme is entirely subjective (which it is), please keep in mind that *any* grading scheme is purely subjective: who picked the questions, and why were those specific questions chosen?

I am 100% with you here. The course instructor knows exactly what was taught, how it was taught, the class's response when it was taught and is, the "sticky" points as divulged by students' questions and homework performance. It follows that the course instructor is best qualified to decide what to put on a test and how to attach a numerical score to the students' answers.

 

[berkeman]

When I was in undergrad at UC Davis in the late 1970s, I took the main Physics sequence for Physicists and Engineers. There were 5 classes in all, with 2 of them optional for Engineers (but I took them anyway because I enjoyed Physics so much). I don't remember how the first 4 were graded, but the 5th one was graded on a curve.

There were about 200 students in those main sequence classes, and for the midterms and final the scores were posted the day after the exam (impressive work by the professor and many TAs). The scores were also tabulated as a bar graph showing the number of students with the scores 0-100, with dividing lines for the various letter grades. I was impressed by the fact that the curves were always very Gaussian in their shape. (I was also happy that my scores were out near the upper tail of the distribution.)

 

[kuruman]

... and for the midterms and final the scores were posted the day after the exam (impressive work by the professor and many TAs).

One of the semesters that I graded exams as a TA, the professor in charge offered incentives to get the job done quickly. He assembled together all graders in a conference room and provided pizza and soft drinks to get going. The beer came after everybody was done and the scores were tallied.

 

[Herman Trivilino]

In my opinion a novice instructor needs to grade on a curve and an experienced instructor can and should grade on an absolute scale. How the student feels about it is not relevant.

 

[gen x]

@kuruman
@symbolipoint
@.Scott
@berkeman
@Andy Resnick
@Herman Trivilino
etc

Is it a higher probability that we have differences in knowledge between generations or that the exam make a difference because it is too hard or too easy?

 

[Andy Resnick]

I am 100% with you here. The course instructor knows exactly what was taught, how it was taught, the class's response when it was taught and is, the "sticky" points as divulged by students' questions and homework performance. It follows that the course instructor is best qualified to decide what to put on a test and how to attach a numerical score to the students' answers.

One issue I am increasingly concerned about is when there are multiple sections (instructors) for a single course. There is external pressure for 'uniformity', which is in (unacknowledged) tension with the basic student-teacher relationship. I guess when we all are replaced by AI, the goal of uniformity will finally be realized.

 

[Andy Resnick]

Is it a higher probability that we have differences in knowledge between generations or that the exam make a difference because it is too hard or too easy?

Never mind generations- there are significant differences in preparation/knowledge/curiosity within a single classroom.

It's trivially easy for me to make exams where everyone will either score 100% or 0%.

 

[.Scott]

Is it a higher probability that we have differences in knowledge between generations or that the exam make a difference because it is too hard or too easy?

There are a couple of issues here.

The first is that when you ask a question about "probabilities" and leave the data collection method unspecified or up to the responder, you are leaving the question so open that even if I said "knowledge between generations is 45% and difficulty is 35%", that response is still wide open to different interpretations.

In the US, STEM SAT scores have gone down even as test questions have been made easier. But there are lots of considerations. More kids are deciding to take the tests - shooting for college when earlier generations would not have.
Even worse (from a data collection view point), historically truancy was not well managed. Back in the late 60's a day at Lowell High School with 85% attendance was a day with crowded corridors. And some unknown percent of the potential student population simply never registered for school at all. (As a student there, I did the attendance there for '67 to '72).

The next issue relates to the "easiness" of the test. I would not presume that the course target were the same decades ago as they are now. Would we really want them to be? And the students that walk into the course at the start of the semester have very different skills (such as Google vs. Dewey Decimal System).

 

[.Scott]

How do you then explain that USA, India, South Korea, Oxford, Cambridge STEM etc use relative grading?

This question is asked in the context of lack of measurable requirements makes absolute grading harder to defend.

I think that @kuruman response gives you the clue to this.
He creates hard measurable course objective himself and then grades to that teacher-imposed standard.
He said: "In other words, I set the goals and then I test how well a particular student has met these goals."

The teacher is certainly a stakeholder - and lacking any push back from other stakeholders, teachers are free to impose a standard.

 

[gen x]

Are you talking about physics or some other discipline?
What does "differences in knowledge" mean? Background, preparation?

Yes physics/engineering, how good are students in class, one class can be bad, other very god etc

 

[kuruman]

One issue I am increasingly concerned about is when there are multiple sections (instructors) for a single course. There is external pressure for 'uniformity', which is in (unacknowledged) tension with the basic student-teacher relationship.

When you say "multiple sections" do you mean recitation sections that support a single lecture course, or independent lecture sections with different instructors for a course with the same catalogue number and description?

If the former, it is the task/duty/obligation/job of the course lecturer to enforce uniformity internally.
If the latter, I think that would be an oddity that most deans would not support these days.

 

[kuruman]

Yes physics/engineering, how good are students in class, one class can be bad, other very god etc

From my experience, there is always a distribution of talents and abilities in every class. I never had a class with only bad students or only good students. I retired from teaching 10 years ago and I can only base my opinions on dated data. Compared with the students that I have known, the post-Covid19 generation might be different from them, not in talent or ability, but in habits, motivation and expectations.

 

[gen x]

In my mind, a curve is not an honest assessment of a student's performance.

What do you think why some uni (all in USA) use curve grading?

 

[berkeman]

[Mentor Note: a couple replies invoking AI have been placed into Moderation for review by the Mentors. Please remember that AI cannot be used as a reference in the technical forums, which includes this forum. Thank you.]

 

[kuruman]

What do you think why some uni (all in USA) use curve grading?

This is what I think and it is only a guess.

Consider a top institution in the USA that accepts the best of the best from all over the country. What kind of exam do you give to these students at introductory level courses? You don't ask questions to determine what they can do as is usually the case. You ask questions to determine what they cannot do by making the exam tough and use the results to teach them how to do what they cannot do. However, to make the process fair, you score the exams on a relative scale. These students already know that they are good. They want to find out how good they are relative to the rest of the class and try harder if they are near the bottom of the best. It makes sense, no?

 

[symbolipoint]

What do you think why some uni (all in USA) use curve grading?

The reflex response is that students are not expected to be identical and therefore a Central Tendency in the grades is chosen to find which score or range of scores is Average (for a letter grade of C).

 

[symbolipoint]

gen x
maybe one combination of ways to say is like this:
Grading by absolutes is for competing with the course. Grading by Relative is for competing against other classmates.

That is only a characterization; not an explanation.

 

[PAllen]

I am a strong believer in absolute grading. In effect, there is a contract between teacher and student: this is the material to be mastered; each grade is a degree of mastery. In principle, that may mean almost no As one year, or 35% the next, depending on actual achievement.

 

[gen x]

This is what I think and it is only a guess.

Consider a top institution in the USA that accepts the best of the best from all over the country. What kind of exam do you give to these students at introductory level courses? You don't ask questions to determine what they can do as is usually the case. You ask questions to determine what they cannot do by making the exam tough and use the results to teach them how to do what they cannot do. However, to make the process fair, you score the exams on a relative scale. These students already know that they are good. They want to find out how good they are relative to the rest of the class and try harder if they are near the bottom of the best. It makes sense, no?

Yes that make sense, but you can still get them very hard exam with absolute grading.

In my country at engineering uni best grade get less than 1% of students(absolute grading), often is case that everyone fail the exam. Graduate or first year pass rate is very low..
If all class solve only 40% of exam, in relative grading 40% will become new best grade, in absolute everyone will fail. Do we want that engineer(40%) that don't know how to "project bridge"? How safe it that?
That I don't like in relative grading..

In Cambridge best grade get 40% of students and pass rate is 99%!!

 

[symbolipoint]

Post #26:
Keep trying!

 

[PAllen]

In my opinion a novice instructor needs to grade on a curve and an experienced instructor can and should grade on an absolute scale. How the student feels about it is not relevant.

I agree there is something to this point of view.

 

[symbolipoint]

In my opinion a novice instructor needs to grade on a curve and an experienced instructor can and should grade on an absolute scale. How the student feels about it is not relevant.

Some institutions do not allow the teacher or instructor to make the choice. In some of these cases, no matter if new, inexperienced or long-time instructor, if the system adopted some form of an Absolute Grading system, then that is what must be applied.

 

[Andy Resnick]

When you say "multiple sections" do you mean recitation sections that support a single lecture course, or independent lecture sections with different instructors for a course with the same catalogue number and description?

I mean the second- for example, each semester we have approximately 200 students enrolled in Intro Physics I (PHY 221 is the course number). PHY 221 usually has 4 independent sections taught at different days and times to accommodate the various scheduling requirements of students. The lecture component is independently taught by different instructors.

This also is the situation for Physics 2 (PHY 222), and calculus-based physics (PHY 241 and PHY 242).

For all of these "Gen-Ed" STEM courses, instructors are given a list of topics to be covered by our State's Education Department, but the list is fuzzy- there are (say) 15 different topics and each instructor is required to cover at least 80% of those topics. Consequently, each section will cover (somewhat) different topics in (somewhat) different order. As a specific example, in my Physics 2 courses I include a lecture on eyes and vision (and optical illusions!) within the "optics" sequence.