Predicted grades for university admissions and student application outcomes
4 December 2023
In this blog, ADR UK Research Fellow Dr Konstantina Maragkou discusses her work using the Grading and Admissions Data for England (GRADE) dataset. Her project is exploring the relationship between student and school characteristics, and the accuracy of teacher-predicted grades for university admissions.
The GRADE dataset is a large, linked administrative dataset containing de-identified individual-level records from the National Pupil Database, the Universities and Colleges Admissions Service (UCAS) and the Office of Qualifications and Examinations Regulation (Ofqual). Using this dataset, we can study predicted grades assigned by teachers, and how they affect students’ university application outcomes.
The importance of teacher-predicted grades
In England, students apply to university using teacher-predicted grades rather than their actual exam results. UCAS encourages teachers to make optimistic predictions to motivate students to apply for better university courses and to aim for higher grades. However, a significant concern surrounding predicted grades is the potential for certain groups of students to receive more or less optimistic predictions than others.
If such disparities distort the rates and outcomes of applications, offers, and acceptance, particularly among groups underrepresented in higher education, predicted grades could have substantial implications for social mobility. This means that understanding the causes and consequences of differences between predicted and awarded grades can have important policy implications for the university admissions system.
How can GRADE contribute to our understanding of teacher-predicted grades and their impact on student outcomes?
The GRADE dataset offers us a valuable resource to address these questions, marking the first time we can use individual-level data to explore them. Student demographic information, including gender and ethnicity, appears across all datasets in GRADE including the National Pupil Database, Ofqual, and UCAS. Additional information can be derived from the specific datasets which can help us explore the following topics:
More insight into disadvantage: When it comes to assessing socio-economic background, prior studies using administrative data have primarily relied on eligibility for free school meals, combined with area-level indicators like local deprivation levels. The free school meals eligibility indicator is restricted to identifying students at the bottom of the income distribution, essentially those in receipt of state benefits. On the other hand, the UCAS component in GRADE provides a broader perspective by including self-reported parental occupation information for all university applicants. This measure is a valuable addition as it has the potential to greatly improve how we identify and measure socio-economic disadvantage in administrative datasets.
Comparing different types of grades: GRADE provides teacher-predicted grades for university admissions for all four cohorts of students included in the dataset, awarded grades for the three pre-pandemic cohorts, and Centre Assessed Grades (CAGs) for the 2020 cohort. CAGs are essentially another form of teacher-predicted grades utilised during the pandemic to assign students their final grades in place of their cancelled external exams. This rich data allows us to investigate educational assessment from various perspectives.
Understanding grade thresholds: Finally, the Ofqual component in GRADE offers a first-of-its-kind insight into the total marks achieved by each student in their exams, as recorded by the exam boards. The inclusion of this granular information on student attainment provides a unique opportunity for evaluating predicted grades around certain thresholds, for example, to understand the impact of "just missing out" on a predicted grade.
In summary the research, utilising the GRADE dataset, addresses critical questions about teacher-predicted grades and their influence on student outcomes. It aims to evaluate the accuracy of predicted grades across student socio-demographic and school characteristics, and to investigate the pandemic-induced exam cancellations to shed light on the impact of using teacher-predicted grades for university admissions. By tackling these questions, the research aims to contribute to more equitable education policies and practices.