Understanding the Master’s Engineering Workforce Landscape: Employer Demands and Student Goals
Plain Language Summary
Four times as many engineering master's degrees are awarded each year as doctorates, yet research on the degree has largely treated it as a salary calculation rather than as an educational pathway with its own reasoning behind it. This paper reports the first phase of an NSF-funded project that sets two sources side by side: what employers actually ask for in mechanical engineering job advertisements, and what students nearing the end of their master's programs say they hoped to gain from the degree. The comparison matters because a gap in either direction, employers wanting something students are not developing or students pursuing something employers do not ask about, is a gap a program can act on. As a grantee report on work in progress, what we present here is the state of the methods and a first read of the data rather than settled conclusions.
Contribution
Where prior work on engineering master's degrees tends to examine either labor market returns or student motivation in isolation, and usually at the level of aggregated STEM rather than a named discipline, we bring a 13-year natural language processing analysis of mechanical engineering job postings into direct comparison with interview and expectancy-value survey data from terminal-master's students in that same discipline, using an adapted version of Perna's model of student college choice as the common analytic frame.
Research Questions
- What is the (mis)alignment between the skills associated with jobs in ME that require a master's degree and current master's students' perceptions of the skills needed for employment?
Methods
We take an exploratory concurrent mixed-methods approach with two strands running in parallel. On the industry side, we filtered a Burning Glass Technologies corpus of 372,135,198 postings down to the mechanical engineering subset through a multi-stage programmatic process keyed to job titles, descriptions, desired qualifications, and mentioned software, iteratively developed and validated by the research group against a human-evaluated subset until it reached a high F1 score, yielding over 250,000 ME postings spanning January 2010 through December 2022 out of the roughly 1.3 million the wider project draws on; those postings were then characterized by minimum degree accepted and by BGT-assigned skills, with skills appearing in fewer than 5% of a given year's postings set aside so that attention stayed on broadly desired abilities. On the student side, we conducted in-depth semi-structured interviews with ME master's students (n = 20), recruited specifically for intending the master's as a terminal degree and for being in their last or second-to-last semester, alongside an expectancy-value survey (n = 100) adapted from an instrument originally built for engineering Ph.D. returners by shifting its language toward master's students and removing items unique to doctoral education. Interviews covered pathway and decision-making, desired and actual skill development, perceptions of industry and the job market, and students' underlying assumptions about the role of engineering and how those might meet the values they perceive in the workforce.
Key Findings
What we can report at this stage concerns the instruments nearly as much as the population. The NLP workflow is the firmer result, in that filtering and aggregation produced a reproducible route from raw postings to comparable year-over-year skill frequencies, letting us set skills such as "Scripting Languages" and "Industrial Design" against one another across the 13-year window along with the qualifications that accompany them. BGT's own labels, however, did not hold up under scrutiny; fields such as "SkillClusters" are assembled through a process the vendor does not disclose, and they proved unreliable signals for observing employers' desired job characteristics over time. Preliminary interview findings point to motivations for the degree that are diverse rather than uniformly economic, to a broad correspondence between the skills students hoped to develop and those they report having developed, and to wide-ranging reflections on the role of engineering through which personal goals and values appear to shape the jobs and work experiences students want.
Implications
Because the vendor-supplied labels could not carry the analysis, we are now combing the raw postings by hand to extract tools and technologies, job tasks and responsibilities, employee knowledge and skills, and qualifications against the O*NET Content Model, guided by reflexive content analysis with active member-checking, and building programmatic workflows, generative AI-based among them, to scale those researcher-created labels to the full dataset, an extension we are weighing for how it might obscure our relationship to the data even as it enlarges what we can process. On the student side, thematic analysis of the interviews, survey comparisons across work experience, returner status, and desired sector, and the case study opening created by participants clustering within institutions are together intended to yield descriptions of decision-making detailed enough for programs to act on. The final step joins both strands into a model of the influences bearing on master's enrollment decisions, which is where the project expects to say something usable about where industry demands, student goals, and program offerings actually meet and where they do not.