On a global scale the school education workforce is under stress. However, staffing issues are heterogenous, multi-dimensional, and localised. Surpluses and shortages can co-exist within systems (Edwards et al., 2025). The consequences of any service disruption are significant for equity, social mobility, economic prosperity and nation level human capital (Bekele et al., 2024; Bi & Li, 2025; Graham & Flamini, 2023; Leoni, 2025). Not surprisingly, there is consensus among governments, policy makers, unions, researchers, and industry experts that the window of opportunity to do something about the teacher workforce is fast closing. What to do about it has proven challenging. Government officials have demonstrated being receptive to empirical evidence on ‘what works’ (Crawfurd et al., 2025), but there remains a dearth of data and evidence that is representative, detailed, and timely (Bleiberg & Kraft, 2023).

The lack of a comprehensive understanding of the determinants of teacher labour supply is a major impediment to developing effective workforce policies. The need to go beyond aggregated supply and demand ratios as a basis for policy intervention has been evident for some time. In addition, the teacher labour is not a single labour market. Accreditation requirements in many jurisdictions mean it is segregated across levels (e.g., primary, secondary) and disciplinary areas (e.g., mathematics, science) with limited substitutability, not to mention, geographic inequities (e.g., urban, rural). The policy challenge for government is therefore how to secure the right qualified teachers where they are most needed by the system. This introduces three key problems, ensuring supply is equal to or greater than demand, the proximal housing of teachers to where they are most needed, and having sufficiently attractive working conditions once you have them to keep them. From an explanatory standpoint, and through a systemic lens, meeting school staffing requirements at a systemic scale can be expressed as:

SS ⟺〖 N〗_((supply ≥ demand) )  ⋀ H_((proximal to need) )  ⋀ 〖WC〗_((attractive))

Where SS is systemic staffing, N represents the total number of available workers with the goal of supply being equal to or greater than demand, H is housing within proximity of where teachers are needed, and there are corresponding working conditions WC that need to be attractive to current and prospective workers. Systemic staffing exists, and only exists, if the three criteria (supply greater than demand, proximal housing, and attractive working conditions) are all met at the same time.

This paper asks the deceptively simple question, how sustainable is the school education workforce? More specifically, it asks ‘What areas are under the greatest workforce stress’

Working with the previously outline explanatory logic, this work makes three contributions. Initially, to the best of our knowledge, it is the first state-wide generalisable approach to measuring systemic staffing sustainability including endogenous (supply and demand, working conditions) and exogenous (proximal housing) measures. Second, empirically the paper provides new evidence on workforce sustainability within a specific site (the Australian state of Victoria). Third, the ‘Teacher Workforce Sustainability’ (TWS) matrix and its underlying explanatory logic, represent a viable data product capable of being deployed in other jurisdictions (assuming access to similar data) to provide a comprehensive multi-dimensional measure capturing the complexity of workforce sustainability. Mindful of enduring issues with scaling up interventions (Banerjee et al., 2017), the TWS matrix meets the need for improved data systems capable of balancing the systemic aggregation and the localised nature of teacher shortages (Nguyen et al., 2024).

We know little about the geography of teacher labour markets. This multi-disciplinary research draws on large-scale longitudinal social (e.g., Australian Bureau of Statistics census data, 2006, 2011, 2016, 2021), education (e.g., Australian Curriculum Assessment and Reporting Authority, 2008-2024), and housing (e.g., Australian Property Monitors, 2006-2024) to develop three domains (supply and demand, proximal housing, and working conditions) data with two temporal regimes (structural conditions 2006-2021; and acute labour-market pressure 2018-2024). These two regimes are analysed separately before being recombined to offer policy relevant findings. First, we built a ‘Structural Workforce Stress Index’ (2006-2021). This index uses indicators available at all four census points (2006, 2011, 2016, 2021). It features supply and demand, proximal housing, and working conditions. These variables capture long-run labour balance, housing access and employment conditions (relative to region). This index answers ‘Which areas have been structurally difficult teacher labour markets over the last two decades?’ Second, we use recent data points to establish an ‘Acute Pressure Index’ (2018-2024). This measures attrition, vacancy rates, applications per vacancy, and housing costs). These are aggregated and standardised annually for means (2018-2024), trends (slopes), and volatility (standard deviation). This index answers ‘Where are staffing pressures most intense right now, regardless of long-run structures?’ Finally, we cross-classify areas into a Teacher Workforce Sustainability matrix of high-high (chronic crisis), high-low (enduring disadvantage, currently stable), low-high (emerging pressure), and low-low (resilient). More informative than a single composite score, this approach is temporally coherent, internally consistent, does not conflate structure with acute pressures, and recognises that workforce stress is not a single phenomenon. In doing so, it offers government and system data insights for long-term planning and short-term intervention. Extensions of the work will be to develop a predictive spatial machine learning pipeline (e.g., Graph Neural Networks [GNN]) to assess and forecast areas under greatest stress. GNN has the capacity to represent and learn from spatially structure data that includes both nodal features (e.g., Local Government Areas) and inter-node relationships (e.g., commuting, working conditions) rather than traditional tabular machine learning models such as OLS and tree-based algorithms that treat each area independently, neglecting spatial autocorrelation and network effects that define real world systems. The primary objective is to quantify the spatial mismatch between supply and demand, proximal housing, and working conditions in the present (longitudinal) and into the future. Such data and evidence will have significant implications for teacher workforce strategies.

Preliminary analysis indicates several key findings with implications for school systems. First, supply and demand ratios have remained relatively stable over time. This is contrary to most reporting in the field. There has been changes in how work is undertaken (part time to full time ratio), but these remain relative to other professions at a regional level. Similarly, despite often reported findings of extended working hours, relative to other professions teachers are not spending more time in paid employment. Housing is becoming increasingly unaffordable on a teacher salary, and this is leading to greater workforce distribution and longer commuting times. This matters for two reasons. First, longer commuting times are linked to greater stress, more days absent from work, poorer attitudes to work, burn out and greater likelihood of leaving the areas (with cascading effects on remaining staff). Second, existing evidence indicates teachers like to work near where they live or undertook their training (Edwards et al., 2024; Wang & Chen, 2022). This is stimulus for many grow your own approaches to teacher education. The affordability or accessibility of situationally appropriate housing is a significant, if often over-looked, exogenous factor in teacher workforce planning. In sum, based on preliminary analysis, the Teacher Workforce Sustainability matrix has enabled the identification of areas through long-run explanation of why they struggle with staffing, and others where short-run signals indicate the system is currently breaking.

Banerjee, A., et al. (2017). From Proof of Concept to Scalable Policies: Challenges and Solutions, with an Application. The Journal of Economic Perspectives, 31(4), 73-102. Bekele, M., et al. (2024). Human capital development and economic sustainability linkage in Sub-Saharan African countries: Novel evidence from augmented mean group approach. Heliyon, 10(2). Bi, M., & Li, X. (2025). How does education affect human capital and potential productivity in Africa?. - Empirical evidence based on Thornthwaite Memorial and CS-ARDL models. International Journal of Educational Development, 118, 103404. Bleiberg, J. F., & Kraft, M. A. (2023). What Happened to the K–12 Education Labor Market During COVID? The Acute Need for Better Data Systems. Education Finance and Policy, 18(1), 156-172. Crawfurd, L., et al. (2025). Understanding education policy preferences: Survey experiments with policymakers in 35 developing countries. World Development, 196, 107140 Edwards, D. S., et al. (2025). Teacher Shortages: A Framework for Understanding and Predicting Vacancies. Educational Evaluation and Policy Analysis, 47(3), 703-729. Edwards, W., et al. (2024). Teaching close to home: Exploring new teachers’ geographic employment patterns and retention outcomes. Teaching and Teacher Education, 145, 104606. Graham, J., & Flamini, M. (2023). Teacher Quality and Students’ Post-Secondary Outcomes. Educational Policy, 37(3), 800-839 Leoni, S. (2025). A Historical Review of the Role of Education: From Human Capital to Human Capabilities. Review of Political Economy, 37(1), 227-244 Nguyen, T. D., Lam, C. B., & Bruno, P. (2024). What Do We Know About the Extent of Teacher Shortages Nationwide? A Systematic Examination of Reports of U.S. Teacher Shortages. AERA Open, 10, 23328584241276512 Wang, Y., & Chen, X. (2022). The draw of hometown: understanding rural teachers’ mobility in Southwest China. Asia Pacific Journal of Education, 42(3), 383-397.