A Machine Learning Analysis of the Non-academic Employment Opportunities for Ph.D. Graduates in Australia

Abstract

Can Australia’s Ph.D. graduates be better utilised in the non-academic workforce? There has been a historic structural decline in the ability of Ph.D. graduates to find work within academia for the last couple of decades (Forsyth in A history of the modern Australian University, New South Press, Sydney, 2014). Around 60% of Ph.D. graduates in Australia, now find jobs outside the academy, and the number is growing year on year (McGagh et al. in Securing Australia’s future: review of Australia’s research training system, https://acola.org.au/wp/PDF/SAF13/SAF13%20RTS%20report.pdf , 2016). The Ph.D. is a degree designed in the early twentieth century to credential the academic workforce. How to make it fit contemporary needs, when many, if not most, graduates do not work in academia, is a question that must be addressed by higher education managers and policymakers. Progress has been slow, partly because of the lack of reliable data-driven evidence to inform this work. This paper puts forward a novel hybrid quantitative/qualitative approach to the problem of analysing Ph.D. employability. We report on a project using machine learning (ML) and natural language processing to perform a ‘big data’ analysis on the text content of non-academic job advertisements. This paper discusses the use of ML in this context and its future utility for researchers. Using these methods, we performed an analysis of the extent of demand for Ph.D. student skills and capabilities in the Australian employment market. We show how these new methods allow us to handle large, complex datasets, which are often left unexplored because of human labour costs. This analysis could be reproduced outside of the Australian context, given an equivalent dataset. We give an outline of our approach and discuss some of the advantages and limitations. This paper will be of interest for those involved in reshaping Ph.D. programs and anyone interested in exploring new ML methods to inform education policy work.