GCSE maths and English grades should be validated using the national reference test to control for any potential bias in teacher assessments, according to a respected academic.
The recommendation by Professor John Jerrim of UCL's Institute of Education follows the government's announcement last Friday that teacher assessments and other data will be used to produce a calculated grade for this year's A-level and GCSE cohorts.
In a blog post, Professor Jerrim said that the national reference test could be used as a potential "get out of jail free card" in case teacher assessments lead to disadvantaged pupils being awarded lower grades than their more advantaged peers.
Live: Coronavirus and schools
Coronavirus: Teacher assessments will be used for exams
Coronavirus: Ofsted won’t judge on 2020 exam grades
Professor Jerrim wrote: "As many people have pointed out, one of the potential problems with teacher-determined grades is that they could be biased for or against certain groups (eg, children from lower socio-economic status backgrounds receiving worse grades than their more advantaged peers).
"It is therefore critical that (a) such predictions are underpinned by data wherever possible, and (b) that the guidance issued by Ofqual (and the approach taken by teachers in making their predictions) has been validated."
He said it was therefore critical that any teacher assessments or predicted grades used to calculate results for this year's GCSE cohort must be validated, and suggested the national reference test could be used to do so.
Professor Jerrim said the test in maths and English – which is sat by around 14,000 pupils in approximately 600 schools each year – could be used to control against any possible teacher bias when awarding "calculated grades" based on teachers' assessments.
He pointed out that the national reference test has already been taken in late February and early March, before the advent of widespread cases of coronavirus.
"Schools and pupils are not provided with any information on how they did on this test, which means it cannot be used by teachers in these schools to inform their view on GCSE grades," he said.
"This makes it the best available data source to validate the guidance Ofqual will publish – and the approach teachers will take – on how to award GCSE grades."
He said the NRT sample of 600 schools could be split in half at random by Ofqual, asking 300 schools – 150 of which would have sat the test for maths, and 150 for English – to follow Ofqual's provisional guidance, and provide grades for students in English and maths GCSE.
Ofqual would also ask schools to provide any additional data on pupils, for example, marks from internal assessments.
Professor Jerrim said this could be used to create a database of Year 11 pupils' key stage 2 results, NRT scores and any other data, which could then be used to "investigate the psychometric properties of teacher-determined grades, including if there is any evidence of bias for/against certain groups".
If problems of bias were identified, this could be used to inform grade moderation.
"Ofqual could then repeat this process with the other 300 schools that participated in the NRT, to test out their revised guidance," he said.
He said this would require "a lot of goodwill" from NRT-participating schools, as well as hard work from staff at Ofqual, but added: "To be credible, it is vital that the approach to awarding GCSE grades is tested and to some extent validated first."
Ofqual has been contacted with a request for comment.