Predictors and Socio-Demographic Disparities in STEM Degree Outcomes: A UK Longitudinal Study using Hierarchical Logistic Regression

Abstract

Socio-demographic disparities in STEM degree outcomes impact the diversity of the UK's future workforce, particularly in fields essential for innovation and growth. Despite the importance of institution-level, longitudinal analyses in understanding degree awarding gaps, detailed multivariate and hierarchical analyses remain limited within the UK context. This study addresses this gap by using a multivariate binary logistic model with random intercepts for STEM subjects to analyse predictors of first-class degree outcomes using a nine-year dataset (2014 to 2022) from a research-intensive Russell Group university. We find that prior academic attainment, ethnicity, gender, socioeconomic status, disability, age, and course duration are significant predictors of achieving a first-class degree, with Average Marginal Effects calculated to provide insight into probability differences across these groups. Key findings reveal that Black students face a significantly lower likelihood of achieving first-class degrees compared to White students, with an average 16 percent lower probability, while students graduating from 4-year degree programmes have an average 24 percent higher probability of achieving a first-class degree relative to those on 3-year programmes. Although male students received a higher proportion of first-class degrees overall, our multivariate hierarchical model shows higher odds for female students, underscoring the importance of model choice when quantifying awarding gaps. Baseline odds for first-class outcomes rose considerably from 2016, peaking in 2021, indicating possible grade inflation during the COVID-19 pandemic. Interaction effects between socio-demographic variables and graduation year indicate stability in ethnicity, disability, and socioeconomic awarding gaps but reveal a declining advantage for female students over time.