Introduction

As climate change mitigation efforts advance globally, concerns have emerged regarding the potential for rapid decarbonisation to create labour market challenges. This transition from fossil fuels towards low-carbon energy sources involves structural transformations on an exceptional scale, with few historical precedents. Within the European Union, extensive discussions are focused on how a ‘just transition’ should be designed to mitigate disruptions, supported by policy measures totalling approximately EUR 55 billion for the 2021–2027 period. These initiatives include vocational training, reskilling programmes, and support for the economic transformation of affected regions. While previous research has documented the effects of environmental regulation on labour market outcomes, empirical evidence on the extent to which decarbonisation itself affects wage inequality remains limited, particularly in Japan.

Conceptual framework: channels linking CO₂ emissions and wages

The study organises the mechanisms linking environmental performance and wages into six primary channels. Four of these channels predict a positive relationship: compensating wage differentials (offering higher wages for hazardous brown-sector work), worker preferences and sorting on amenities (where environmentally responsible firms may pay lower wages due to non-pecuniary benefits), rent sharing in carbon-intensive production, and differences in production processes or capacity utilisation. Two channels predict a negative relationship: productivity-linked rent sharing (where more productive, cleaner firms pay higher wages) and the green skill premium (where workers with low-carbon technological skills command higher pay). Controlling for total factor productivity (TFP) is critical, as it separates the productivity-linked rent-sharing channel from the remaining mechanisms.

Data

The analysis utilises an unbalanced panel dataset of Japanese manufacturing installations constructed by linking government statistics from 2000 to 2019. Primary data sources include the Census of Manufacture conducted by the Ministry of Economy, Trade and Industry (METI) and the Economic Census for Business Activity. CO₂ emissions data are sourced from the Greenhouse Gas Emissions Calculation, Reporting and Publication System (SHK system) administered by the Ministry of the Environment (MOE). The system covers approximately 53 per cent of Japan’s total greenhouse gas emissions. The final estimation sample comprises 59,316 installation-year observations covering 6,611 installations. The dataset disproportionately represents relatively large installations, which typically pay higher average wages than the census average.

TFP estimation and empirical specification for wages

Total factor productivity is estimated using the Wooldridge (2009) methodology to account for unobserved productivity shocks affecting the production function. The empirical model for average wages employs a fixed-effects specification, regressing average wages on CO₂ emissions while conditioning on output (shipment values), labour, capital, and TFP. This specification describes how average wages differ across periods in which an installation emits more or less CO₂, holding its output and measured inputs constant. The study examines both CO₂ intensity, defined as emissions divided by shipment values, and absolute emission levels.

Results

The baseline results show that CO₂ intensity is negatively associated with average wages when TFP is not controlled for (-0.0589), but becomes positively associated once TFP is included (0.0325). This reversal suggests that more productive installations tend to emit less per unit of output while paying higher wages. Holding productivity constant, the within-installation relationship is positive: periods of higher CO₂ emissions correspond with periods of higher average wages. Industry-specific analyses show significant positive relationships in many sectors, including food products (0.0583), textile products (0.0713), and production machinery (0.0812). Conversely, general-purpose machinery exhibits a negative and statistically significant relationship (-0.0478), possibly due to value creation shifting towards knowledge-intensive, lower-emission activities such as design, software, and lifecycle support.

Heterogeneity and sources of variation

Two-stage least squares (2SLS) estimates indicate that wages are more closely associated with persistent levels of emissions (0.1446) than with short-term fluctuations (0.0345). Analysis across TFP quintiles reveals that the positive association is weaker amongst the most productive installations (0.0103), where workers are more likely to possess skills not tied to carbon-intensive production. Furthermore, the association between emissions and wages increases monotonically with an installation’s value-added share of shipment values, rising from 0.0214 in the lowest quintile to 0.0462 in the highest. This provides the most direct evidence that rent-sharing mechanisms in carbon-intensive production play a significant role in the observed wage patterns.

Discussion and conclusion

The study documents a positive association between CO₂ emissions and average wages within Japanese manufacturing installations when productivity is held constant. These findings are consistent with mechanisms such as rent sharing and differences in working conditions rather than a dominant green skill premium during the 2006–2019 period. While the findings provide a useful descriptive benchmark, they do not imply that reducing emissions will cause worker wages to fall. However, the observed association suggests that the distributional consequences of the low-carbon transition warrant further investigation, particularly for workers and regions tied to carbon-intensive production. Future research is suggested to examine how these relationships evolve under more ambitious decarbonisation policies, such as the GX Promotion Act enacted in 2023.