At the moment there is a lot of fuss about the A-level grades awarded in the UK. This year, due to Covid-19 the grades awarded were based off of predicted grades from the teacher of the students. Now, what is interesting is that every year in the UK teachers must give predict grades anyway! Why is this?
In the UK students use their predicted grade to apply to Uni. If their grades don’t cut the muster they won’t be offered a place but if their grades are good enough then they’ll get an offer letter. This means if they achieve the grades on the offer letter (which should be the same or lower than their predicted grades) they’ll get the place in Uni. So, this means we have good data on how good teacher predictions are, each year teachers make predictions and then students either get those grade or don’t.
This whole system is managed by UCAS (this is the UK version of the CAO). So, what did UCAS find in 2019 when it comes to the question of students reaching their predicted grades from their teacher? Do they achieve their predicted grades? Do better? Or worse? Well, here is a little snippet from a UCAS report from 2019.
Now, notice what this is saying. It is saying 80% fell short of their predicted grade. Therefore, any model that tries to predict actual exam grades from teacher’s predicted grade must discount the teacher’s prediction 80% of the time. So, this year, in 2020 how many grades were revised down from the teacher’s predicted grade? Well from the Telegraph (and other tabloids) running with this story:
Well, that is interesting, it is only 40%. This means that (the other) 40% of students got grades higher than what they would have got had they have taken the exam. So on the face of it is seems students are doing better than they should be doing. It also means that 40% of students got a grade lower than predicted (half the normal amount). It would seem those 40% are kicking up a fuss and now the UK government is going to U-turn and give everything their predicted grade.
In the end, instead of 80% of students being downgraded (the usual story) only 40% were downgraded and even that is now being reduced to 0% so the crazy situation is that for 2020 we will have 80% of grades inflated. Top grades for all I guess… I do understand the fuss though. The problem seems to be in the design of the model that predicts grades, it seems those in fancy public schools are getting a boost (and those on the other end a drop). But surely the answer is not to abandon the model but to fix it i.e. modify it? Is this the path of least resistance though?