Ofqual announced that for the duration of the 2019-2020 year, students will receive a calculated grade, which has the same status as examination-based grades awarded in previous years. Ofqual’s guidance to schools and teachers states that calculated grades should be based on performance throughout the year, such as coursework and mock exams in order to approximate the grade students would have received had they sat their examinations. In addition, each school will rank its students within each grade to enable Ofqual to statistically standardize results on a school-wide basis. While Ofqual emphasises the need for teachers to use available data, it is inevitable that a significant element of discretion is being exercised in setting these grades; more than would be the case normally.
The potential for bias
There is concern about the accuracy with which teachers estimate grades, and in particular, the possibility that subconscious bias could discriminate against students based on these protected characteristics.
The evidence for this was highlighted in a 2016 study undertaken by UCL, which found that teachers only accurately predicted 16% of their students’ grades. The study was based on teachers’ predictions of their student performance in A-level exams, which are vital when applying for universities. The raw figures suggested that teachers generally overpredict students’ grades, particularly for students of disadvantaged backgrounds. However, this appeared to be largely due to teachers’ tendency to overpredict the grades of lower performing students, and the fact that students from disadvantaged backgrounds are overrepresented at this end of the scale. With further analysis, the study found:
- For high performing students, teachers tend to underpredict the grades of students from disadvantaged backgrounds.
- High performing female students were also more likely to be underpredicted than males.
The former Department for Business Innovation and Skills, now DEIS, also published studies of predicted grades in 2011 and 2013 which found that black students’ predicted grades had low accuracy rates. In 2011, black students’ grades were the least accurately predicted, with the highest levels of both over and under prediction.
In April 2020, Ofqual released its Equality Impact Assessment, which concluded that the research regarding teacher bias in grading is mixed. The review included research from outside of the UK. Ofqual found that there appears to be some evidence that race and disadvantage affect the accuracy of predicting grades, but the effect itself has not been properly estimated.
Discrimination and the Equality Act 2010
Given the significant impact of education qualifications on access to higher education, employment opportunities and social mobility, it is vital that all practical steps are taken to eliminate the potential for bias in grading. Ensuring fair results is not just the right thing to do, section 149 of the EA specifies that public bodies must have due regard for the need to eliminate unlawful discrimination and advance equality of opportunity. This includes removing or minimising disadvantages, meeting the needs of people with protected characteristics and encouraging their participation in activities – such as education – where participation is low. The Equality Act 2010 is clear on the public sector’s equality duty. The obligation is not intended to be just a box ticking exercise: Ofqual must approach the issue of potential discrimination in grading students “with rigor.”
Students will have the opportunity to appeal their grades or to sit examinations during a later term if they wish to improve the grade they receive. However, students should not bear the burden of correcting inaccurate grades that result from bias against certain protected characteristics.
Ofqual invited submissions to its consultation on its proposed grading scheme, including the issue of making adjustments for potential bias, which closed on 29 April 2020. It will be interesting to see the outcome and what action will be taken as a result.
Research centers, such as FFT Education Datalab, have suggested the use of available data sources, such as the National Reference Test, to assist with quality checking calculated grades and detecting evidence of potential bias.
At this stage Ofqual has only announced plans to adjust the distribution of grades on a schoolwide basis. Students could be moved up or down a grade, depending on whether the school’s overall calculated grade distribution matched the expected grade distribution based on statistical analysis. This adjustment will not revisit how individual students are ranked within schools.
Given that bias could materially affect certain groups of students disproportionately within a school – and could impact the ranking they are given as against their classmates – serious consideration should be given to ensuring grades are fairly adjusted at a more granular level within schools in order to discharge Ofqual’s responsibilities under the EA effectively.