The research, led by Anna K. Berryman of the University of Oxford together with Fernanda Senra de Moura of the University of Oxford and the World Bank, and R. Maria del Rio-Chanona of University College London, builds on earlier work showing that productivity-led structural transformation towards less land-intensive sectors, such as manufacturing, can substantially reduce deforestation. That earlier analysis, by J. B. de Souza Ferreira Filho and Marek Hanusch, used a dynamic computable general equilibrium (CGE) model with spatially explicit land-use transitions to show that productivity gains in urban, traded sectors weaken agriculture’s relative competitiveness and lower the economic incentive to clear forest land. What the CGE framework could not reveal, however, is what such a transformation would mean for the millions of Brazilian workers whose livelihoods depend on the sectors being reshaped.

To close that gap, the team developed a spatially and occupationally explicit agent-based labour market model, informed by empirically observed job-transition networks, to simulate how workers respond to labour demand shocks. The empirical backbone of the model is Brazil’s Relação Anual de Informações Sociais (RAIS) dataset, which records all active employer–employee contracts in the country’s formal labour market. From 2011 to 2019 data, the researchers constructed a regional occupational mobility network in which each node represents an occupation–region pair and each weighted edge reflects the historical probability of workers moving between pairs. The network comprises 410 occupations across 16 regions of Brazil.

The network analysis revealed two powerful frictions that constrain how workers can adapt to structural change. Workers were 62% more likely to transition into an occupation within the same broad category than into a different one, and 7.2 times more likely to remain within their home state than to relocate. Quantifying these patterns with assortativity measures, the team found that regional assortativity in the regional occupational network reached 0.77, while occupational assortativity at the one-digit level stood at 0.56. Strikingly, when the regional occupational mobility network was visualized, its structure closely resembled the geographical map of Brazil itself, underscoring how strongly geography, moving costs and familial ties shape labour flows.

The agent-based model itself runs in monthly time steps and tracks employed workers, unemployed workers and unfilled vacancies for every occupation–region pair. Each month, workers may lose jobs through routine turnover or through state-dependent separations driven by mismatches between realized and target labour demand, while vacancies open through analogous spontaneous and state-dependent processes. Unemployed workers search for new positions by applying to vacancies in occupations and regions connected to their previous job through the empirical mobility network, and the matching process uses an urn-ball matching function in which one applicant is drawn at random whenever at least one worker applies. The researchers calibrated the model’s three key parameters using quarterly unemployment and nominal GDP data from 2012 to 2019, achieving unemployment rate correlations exceeding 0.93, and validated it on out-of-sample data from 2020 to 2025, where it successfully reproduced the sharp unemployment spike caused by the COVID-19 pandemic.

The team then coupled this labour market model to the CGE scenarios developed by Ferreira Filho and Hanusch, which simulate how Brazil’s economy and environment might evolve from 2018 to 2030 if productivity grew by 0.5 percentage points annually in manufacturing, agriculture, services or mining. Because the CGE model outputs labour demand at the industry level, the researchers converted it to occupation-level demand using the observed occupational composition within each industry. The coupling is one-directional: labour demand projections inform the labour market model, but labour market outcomes do not feed back into the macro-economic projections. Even though aggregate labour demand in each scenario equals the baseline by design, layoffs and vacancies emerge at the disaggregated occupation–region level, and it is precisely there that the frictions bite.

The results are sobering. Across nearly all productivity-driven scenarios, high-wage and high-skill workers experienced unemployment rates below the baseline, while lower-wage workers, particularly in agriculture, faced increased unemployment. Agricultural workers, who are among the least spatially and occupationally mobile, were negatively affected in almost every scenario, with unemployment rate increases ranging from 0.13 to 2.11 percentage points. The scenario delivering the best environmental outcomes, a nationwide shift towards more productive manufacturing powered by Brazil’s low-carbon energy supply, improved employment prospects for mid- and high-skill workers but raised unemployment for agricultural and other low-skill workers. The only scenario in which agricultural workers fared better than baseline was one where agricultural productivity in the Amazon increased disproportionately, yet in that scenario deforestation and emissions rose significantly, making it the worst environmental performer.

At the fine-grained four-digit occupation level, the model revealed how a worker’s position in the mobility network determines their fate. In São Paulo, garment sewing machine operators and hand weavers, handmade shoe and leather goods craftsmen faced similar demand decreases but divergent unemployment effects: the former suffered increased unemployment because their neighbouring occupations were also declining, while the latter remained nearly unaffected because their neighbours were thriving. Similarly, among agricultural workers in the southern model region comprising Santa Catarina and Rio Grande do Sul, general agricultural and agricultural support workers faced unemployment increases about 0.95 percentage points higher than farm supervisors despite nearly identical demand changes of roughly minus 9%. On the vacancy side, firms struggled to fill positions for middle-level technicians and professionals of sciences and arts. Civil and allied engineers in Roraima, facing 7% higher demand, saw vacancy rates rise by about 0.50 percentage points, whereas CNC machinists in the same state, facing even larger demand growth of 8.8%, saw vacancies remain nearly unchanged because workers from declining neighbouring occupations could flow into those jobs. Spatial frictions compounded the problem in remote northern states such as Acre, Amapá, Rondônia and Roraima, where firms could not fill positions even as unemployment rose elsewhere in the same occupations.

The most hopeful finding concerns policy. When the researchers reran the scenarios with an intervention that increases agricultural workers’ mobility into occupations with similar skill levels in other sectors, such as mid-level, industrial and repair-and-maintenance occupations, unemployment outcomes for Amazon agricultural workers improved by approximately 50% on average under the manufacturing growth scenario, with scenario-led unemployment increases between 0.02 and 0.11 percentage points lower in regions including Acre, Amapá, Amazonas, Matopiba, Pará, Rondônia and Roraima. This demonstrates that well-designed labour market policies, including retraining, reskilling and relocation support, can align productivity gains with equitable employment outcomes.

The study’s implications reach far beyond Brazil. The authors note that their findings echo evidence from the United States showing that a green transition may not dramatically shift employment geographically but requires different skills. Because the analysis relies on formal labour market data covering roughly 67% of the total labour force, informal workers in services, agriculture and industrial-craft occupations remain under-represented, a limitation the researchers say future work should address using sources such as the Brazil National Household Sample Survey. Still, the central message is clear: deforestation is not merely a problem of forest governance and enforcement, but a reflection of deeper structural patterns in the macro-economy. Skills development policies must be implemented early, the authors argue, so that the green transition can be both fast and fair, ensuring that the workers least equipped to move are not left behind as Brazil steers its economy away from the forest frontier.

Subject of Research: Labour market skill and spatial mismatches shaping deforestation-reducing development pathways in Brazil

Article Title: Skill and spatial mismatches for sustainable development in Brazil

Article References: Berryman, A. K., de Moura, F. S., Bücker, J., Barbrook-Johnson, P., Hanusch, M., Mealy, P., Farmer, J. D., & del Rio-Chanona, R. M. (2026). Skill and spatial mismatches for sustainable development in Brazil. Nature Sustainability. https://doi.org/10.1038/s41893-026-01939-z

Image Credits: AI Generated

DOI: 10.1038/s41893-026-01939-z

Keywords: Brazil, Amazon, deforestation, labour market, agent-based modelling, skill mismatch, spatial mobility, structural transformation, green transition, unemployment, Nature Sustainability, economic complexity