Location is a major barrier for transferring US fossil fuel employment to green jobs

Climate change is intensifying storms, flooding, and wildfires, necessitating policy to limit warming. Absent scalable carbon removal, phasing out fossil fuels by 2050 is required to stay within 2°C, but such a phaseout could displace 1.7 million US fossil fuel workers and harm communities dependent on fossil industries. A Just Transition aims to support affected workers. Stakeholders view green jobs as a solution—e.g., the US Inflation Reduction Act (IRA) projects millions of jobs and the EU’s Just Transition Mechanism provides substantial funding. However, the feasibility of placing fossil fuel workers into green jobs hinges on two conditions: transferable skills and geographic co-location of workers and new jobs. While many assert skills are transferable, it is unclear whether both skill and location requirements are met. Regions specialized in fossil occupations (e.g., Western Pennsylvania) are unlikely solar hubs due to resource constraints (cloud cover), illustrating potential geographic mismatch. This study takes a macroeconomic perspective, assessing whether fossil fuel extraction workers’ skills align with green occupations and whether these workers are co-located with current and projected green job opportunities.

Prior work highlights that skill similarity shapes transitions and labor market resilience, and that geographic frictions can hinder mobility. Existing studies have begun examining co-location issues between fossil and green jobs but often focus on narrower occupational sets or lack mobility estimates. Research has also addressed labor reallocation during transitions, the role of matching functions, and the importance of related skills and tasks. Policy-oriented literature emphasizes Just Transition frameworks and regional vulnerabilities, with debates on place-based versus people-based strategies. This paper builds on these strands by jointly evaluating skill similarity, historical mobility patterns, and geographic co-location relative to current renewable energy infrastructure and projected green employment growth.

Data sources: (1) O*NET occupation skill profiles (BLS) for more than 750 SOC occupations, normalized to 0–1 importance scores across 232 skills; (2) BLS employment by occupation and region (MSA/NMSA) from 2005–2019, plus national occupation-by-industry employment breakdowns; (3) US Energy Information Administration data on current solar, wind, hydro, and biomass power plants; (4) US Census Bureau Job-to-Job Origin-Destination (J2JOD) flows (2006–2019) capturing inter-industry and inter-regional transitions at MSA/NMSA levels; (5) BLS 2029 employment projections and additional demographic/economic covariates (population by age, gender, race, education; state-level GDP per capita by NAICS 2-digit). Skill similarity: Real-valued Jaccard similarity computed between occupations using min/max aggregations across the 232 normalized O*NET skills. Industry-level skill profiles were constructed as employment-weighted averages of occupation skill vectors; industry-level similarity computed analogously. Mobility modeling: Poisson regressions of J2JOD worker flows between industry-region pairs include log geodesic distance (Haversine, centroid-to-centroid), skill similarity, and employment in origin/destination regions; variables are standardized (z-scores). Model comparisons include (i) random mixing baseline (employment only), (ii) baseline + skill similarity, (iii) gravity model with distance and employment, and (iv) gravity + skill similarity; robustness controls include indicators for staying in same industry or location. Co-location analyses: Cross-sectional mapping and Poisson regressions relate 2019 fossil extraction employment to counts of nearby green power plants (solar, wind, hydro, biomass), controlling for labor market size. Forecasting green jobs: A random forest regression is trained to predict 2029 green employment by MSA/NMSA using lagged (10-year) historical employment, demographics, and economic features; 10-fold cross-validation, temporal holdout (2019 as test), and comparisons to OLS/Lasso. Out-of-sample performance achieved R²=0.82 and RMSE=0.57. Transition projections: Apply the best-performing mobility model (with distance and skill similarity; controlling for stayers) to combine 2019 extraction worker distributions, predicted 2029 green employment by region, skill similarity, and distances, estimating expected transitions. Scenario simulations: Vary the size and spatial allocation of newly created green jobs—non-targeted (proportional to total employment) versus geo-targeted (proportional to 2019 fossil fuel employment)—at 1M, 5M, and 10M jobs to quantify effects on transition rates. Robustness: Alternative green occupation definitions, validation against external projections (e.g., Princeton Net-Zero America), and analyses of transitions to non-green sectors with high and low skill similarity to extraction (e.g., construction, manufacturing, transportation/warehousing).

- Skills are transferable: Fossil fuel extraction workers exhibit high skill similarity to green occupations (average skillsim(f,g)=0.79; two-sample t-test p<0.0001), higher than to other industries. Some re-skilling may still improve transitions, as transitions are more likely when skillsim≥0.9. - Distance dominates mobility: In Poisson models of historical J2JOD flows, adding distance to employment (gravity model) yields pseudo-R²=0.72; adding skill similarity raises pseudo-R² to 0.81. Distance is the largest effect driver; skill similarity adds predictive power but is secondary. - Weak current co-location: Mapping 2019 fossil extraction employment against locations of solar and wind plants shows most regions have either many plants or many extraction workers, not both. Poisson regressions controlling for labor market size reveal weak associations between green plant counts and extraction worker concentrations. - Projected mismatch persists: Random forest predictions of 2029 green employment (validated with R²=0.82, RMSE=0.57) show several Great Plains regions will approach parity between green and fossil employment, but many large fossil regions (Nevada, New Mexico, Western Pennsylvania, North Dakota) will not see comparable green job growth. - Very low baseline transitions: The model predicts 98.97% of extraction workers will not transition to green jobs. Under idealized co-location, 13.7% would transition; under perfect skill match (skillsim=1) with other factors equal, 5.51% would transition—indicating geography is the primary constraint. - Limited relocation: Most transitions occur without relocation; few exceed 20 miles. In the 15 most extraction-intensive regions, <1.5% are expected to transition; the highest observed local transition rate (Dallas, TX) is ~4%. - Policy targeting matters: Creating 1M geo-targeted green jobs in fossil-intensive regions induces higher transition rates than creating 5M non-targeted jobs distributed by total employment. Geo-targeted 5M jobs can exceed transition rates to construction/manufacturing, which otherwise absorb more workers than green jobs in baseline. - Alternative sectors: Without targeted policy, transition rates to manufacturing and construction exceed transitions to green jobs, reflecting higher co-location and/or strong skill similarity (construction and utilities have highest industry-level similarities to fossil extraction).

The study asks whether a Just Transition can re-employ fossil fuel extraction workers in green jobs by assessing skill transferability and geographic co-location. Findings confirm a strong skill match between extraction and green occupations, suggesting that large-scale re-skilling is not the primary hurdle. Instead, geographic mismatch and low worker spatial mobility are the main barriers. Historical mobility patterns indicate distance strongly constrains transitions, and current/prospective green job locations do not align with where extraction workers live. Policy implications are clear: effectiveness hinges on addressing geography. Place-based strategies that concentrate new green investments in fossil fuel communities or people-based policies that reduce relocation frictions could meaningfully improve transition rates. Simulations show geo-targeted green job creation can outperform much larger non-targeted expansions, emphasizing the importance of co-location. The results also suggest that adjacent sectors (e.g., construction, manufacturing) can temporarily absorb displaced workers, but optimal outcomes for a Just Transition likely require aligning green job creation with existing worker locations.

The paper contributes evidence that US fossil fuel extraction workers possess skills closely aligned with green occupations, but geographic mismatch between worker locations and current/projected green job sites is the dominant barrier to a Just Transition. Quantitatively, baseline projections suggest very low transition rates absent policy changes; even under idealized scenarios, distance remains pivotal. Policy should prioritize geographic targeting of green investments and/or reduce relocation frictions. Future research should investigate social and cultural factors that shape relocation hesitancy and design optimal regional diversification strategies that leverage existing worker distributions. Absent attention to geographic mismatch, large-scale climate policy initiatives risk underperforming in labor market outcomes for fossil fuel workers.

- Assumes all current fossil fuel workers would be willing to transition to green jobs, ignoring potential social, cultural, or identity-based barriers. - Applies an average mobility model from industry-level J2JOD data to occupation-level projections, potentially obscuring heterogeneity across sub-sectors and communities. - Ignores competition from workers in other industries for green jobs. - Focuses primarily on extraction workers, excluding other fossil sector roles (e.g., engineers, legal, managerial) that would also be affected. - Assumes 2019 O*NET skill requirements persist through 2029; new occupations and evolving skills may alter matches. - Uses forecasts of green job growth; while validated and robustness-checked (alternative green definitions, power-plant-based estimates, external comparisons), forecast uncertainty remains. - Does not directly model non-employment social barriers or broader community-level effects beyond the labor market.