Forecasting Failure
: Assessing Risks to Quality Assurance in Higher Education Using Machine Learning

Student thesis: Doctoral ThesisDoctor of Philosophy


The landscape of UK higher education has changed significantly in the last five years. A tripling of tuition fees, the uncapping of student numbers, and an explosion in the number of ‘alternative providers’ typify a more marketised higher education sector (Brown and Carasso, 2013). With more providers than ever before competing for students, many with little experience and profitdriven motives, there is a clear danger that quality will suffer.

Faced with limited resource and an expanding, fiercely independent sector, the Government sought to protect quality by asking the Quality Assurance Agency for Higher Education (QAA) to adopt a risk-based approach. The 2011 White Paper Student at the Heart of the System directed QAA to prioritise their reviews based on “an objective assessment of a basket of data, monitored continually but at arm’s length” (BIS, 2011, 3.19). There is, however, an evident dearth of empirical evidence to support such an approach . The aim of this thesis is to examine the extent to which available data can predict the outcome of quality assurance reviews, and hence prioritise them.

To fulfill this aim, the outcomes of all QAA reviews comparable with its current inspection methods were gathered along with all available data that could feasibly form part of a data-driven riskbased approach to quality assurance. Using machine learning, this study shows conclusively that a risk-based approach to quality assurance, as envisioned in the 2011 White Paper, cannot work. There is no connection between the available data and the subsequent outcome of QAA reviews.

The final part of this thesis therefore examines the reason why there is no connection between the available data and the outcome of QAA reivews. Three overarching and non-exclusive possibilities are identified. Concerns over the data, the review process, and the definition of ‘quality’ pose significant barriers to the operation of a successful data-driven, risk-based approach. An alternative approach to prioritising quality assurance in higher education is therefore required.
Date of Award2017
Original languageEnglish
Awarding Institution
  • King's College London
SupervisorAlison Wolf (Supervisor) & Henry Rothstein (Supervisor)

Cite this