Banning GCSE grades before year 11

6 min readJul 2, 2019

There is nothing so useless as doing efficiently that which should not be done at all — Peter Drucker

Last month I gave a talk at ResearchED Rugby looking at the misuse of GCSE grades. Here is a short summary of why using GCSE grades before year 11 is wrong, a waste of time, or unsubstantiated. I caveated my talk on the day by saying I often talk in absolutes when really there is some nuance; please bear that in mind and please challenge anything you disagree with.

Using GCSE grades a progression model

This is where a student is expected to slowly rise up through the grades from grade 4 in Year 9 up to a grade 9 at the end of Year 11. This is commonly known as a flight path and pervades schools.

GCSE grades are a measure of attainment for students at the end of year 11 who have studied the whole course. The grade describes that students attainment against other students in their year group and against historic and future year 11 students.

However, the logic of flight paths if that if we know that the ‘best’ students at the end of year 11 get a grade 9, then when they were in year 10 they will have been worse, so therefore we could say they would have got a grade 8. If we track this back then in year 9 they would have got a 7 etc etc.

My first issue, that I won’t linger on is that the interval of the differences is made up. What’s to say that students shouldn’t make 2 grades progress every year?

The bigger issue is that GCSE grades are not intended to be used to describe attainment before the end of year 11.

Let’s use a marathon runner as an example. Elite runners can run a marathon in just over 2 hours. Runners who are worse will run a marathon in a longer time. We also find that over time runners will get quicker. Seems ok so far but now imagine I hire a running coach to get be from couch to marathon. We have our first training session, we go for a 5k run and I ask her how I’m doing. “5 hours”, she says. “What do you mean?”, I ask. “You’ve run a marathon in 5 hours”. But I haven’t run a marathon! Hopefully, you can see how ludicrous it would be to describe the ability of a runner using a marathon time when they have never run a marathon.

You can not use a GCSE grade to describe the attainment of someone who hasn’t studied the whole course.

Predicting GCSE outcomes

Continuing my analogy of marathon training. My coach has taken on board my feedback and agrees it wasn’t a good way to feedback. I ask her again how I’m doing and she says “3.5 hours”. “What do you mean?”, I ask again. “Based on your training to date I think by the end you’ll run a marathon in 3.5 hours. My prediction might change a little over time but I don’t expect it to change very much”

The premise here is that there is a correlation with the baseline data and GCSE outcomes. This is usually measured as a correlation between the two sets of data; the predictor and the outcome. You will read different standards in different places but in general 0.7 is considered to be good. Before teachers start doing this they are usually supplied with predicted or target grades which may be based on KS2 results or a commercial test. These usually have reasonably high correlations which may be down to the length of the test and the standardised environment in which they’re sat. The graph above shows a study I did on my own school’s end-of-year exam data that showed similarly high correlations.

Your commercial test may output a predicted grade of 6 but this isn’t the full picture at all. In fact, the supplier will probably give you a predicted grade as a decimal like 5.9. So what does this mean? It’s the average result of similar students in the past. This is usually supplied as a changes graph. What this means is that there is a 55% chance of the student achieving a grade 6. There is a 42% chance that the student will achieve a 5 or a 7.

If you have 100 students in your cohort, and you predict them all a grade 6 (based on 5.9), 23 of them will fall short of their predictions. This is because they aren’t predictions, they are just an average.

Even if we accept all this, we now need to think about what we’re basing our predicted outcome on when we give students a GCSE grade in year 9. This is usually end-of-topic tests, so to check how well these predict outcomes I produced a similar set of correlations for our A-Level end-of-topic tests. We have correlations ranging from -0.2 to 0.7. It’s a mixed bag and it doesn’t look good! End-of-topic tests aren’t a good predictor of final outcomes.

If you are using your end-of-topic tests to predict final grades, I think this needs to be justified with data. I haven’t seen any strong arguments yet, but I’m all ears.

Basing grades on criteria

Lastly, I see people arguing that they’re not using a progression model or predicting the outcome but that they know from the content and quality of students work what grade they will achieve.

The premise here is that GCSE grade can be used to describe the difficulty of work. You might see this as graded questions or learning objectives.

The problem is two-fold.

1 Teachers are making these grades up. You can find two different resources with the same construct assigned two different grades. They can’t both be right.

2 A GCSE grade is used to describe the overall attainment at the end of the course. A grade is awarded based on the accumulation of marks across a paper.

To test this out I have plotted the marks awarded for each question of Physics paper 2 2018 broken down by the final grade awarded to each student. On question 1, for example, students who got 5 marks were awarded grades 6–9. The last question, which would commonly be described as a ‘grade 9’ question, saw students who achieved a grade 6 gain the same number of marks as students who were awarded a grade 9.

One argument I hear is that “I’m just using grade 9 question as shorthand for ‘hard question’”. Then just say it’s a hard question!

So what’s the big deal?

When you give students a grade it means something to them. Saying they are on a grade 6 in year 9 but will improve is a fabrication. Predicting a student a grade 6 when they are just as likely to achieve a 5 or 7 is worrying. Calling a question a ‘grade 9’ question means that students who don’t get all the marks don’t think they can get a grade 9. This is unfair.