1. Introduction

The climate emergency demands substantial reductions in greenhouse gas emissions to prevent irreversible and catastrophic outcomes. In parallel, the severe degradation of the environment and biodiversity loss due to rapid industrialization necessitate a new economic model that ideally decouples economic growth from further resource use. To address these environmental challenges, conventional policy approaches have primarily focused on transitioning from polluting energy sources to cleaner and renewable ones. However, achieving a successful green transition and ensuring environmental sustainability may also require action in other policy domains, including labour markets and labour supply. As Ostrom (2009) suggests, economic activities are inherently embedded within a social-ecological system, involving intrinsic linkages between labour markets and the environment. These interconnections, coupled with the ongoing climate crisis, call for a rethinking of how labour is organized (ILO 2019).

Given these intrinsic linkages, this article – along with others such as Schor (2005), Knight, Rosa and Schor (2013), Kreinin and Aigner (2022) and Fitzgerald, Givens and Briscoe (2024) – argues that one way to address climate change and limit its negative impacts on the environment could be through a labour market instrument: the reduction of working time. To assess whether this approach could serve as a viable option – among other mitigation strategies – in addressing climate change, this article aims to establish and quantify the relationship between hours worked and several environmental indicators across a set of European countries over the past few decades.

In this context, the European Green Deal and its Fit for 55 package set out a road map towards a net-zero economy in the European Union (EU). Similar decarbonization plans have been adopted by many countries worldwide, aiming to reduce emissions and, to a varying extent, limit resource extraction. However, these plans often assume the maintenance of similar levels of consumption and growth patterns, which have already contributed to the climate emergency and are typically associated with high energy demand and resource extraction. This dynamic risks placing further strains on the planet, even as efforts are made to shift from polluting energy sources to cleaner alternatives (Bohnenberger 2022; Urban, Chapman and Rizos 2024). While emerging research is increasingly recognizant of the impact of the decarbonization process on work patterns and labour markets more generally, policy discussions tend to focus on the consequences of sectoral restructuring of fossil-fuel-intensive industries, which are often regionally concentrated and, therefore, may necessitate localized policy responses (Black, McKinnish and Sanders 2005; Beckfield et al. 2020; Hanson 2023; OECD 2024). However, if the objective is to transition to a net-zero economy within planetary boundaries, more ambitious policies beyond energy transition (Schor 2005) and “far-reaching changes in countries’ production and consumption” systems (Hanson 2023) will be necessary.

One promising area for policy action is working time reduction, given its potential aggregate impact on the environment through several channels. One commonly advocated mechanism relates to the redistribution of productivity gains. The proponents of working time reduction argue that the extensive productivity gains over recent decades could be translated into reduced working time while keeping the aggregate output at the same level – or even lower levels for some high-income and industrialized countries – by offering workers more non-working time, where the latter is an invaluable yet scarce asset and is associated with improved well-being (Hayden and Schandra 2009; Schor 2010). In other words, alongside conventional efforts in energy transition, working time reduction could kill two birds with one stone by, on the one hand, alleviating negative repercussions on the environment and addressing a key driver of climate change, and, on the other hand, improving the well-being of workers (Schor 2005; Knight, Rosa and Schor 2013). Another way in which reduced working time may ease environmental pressures is by encouraging behavioural changes at the individual, household or organizational level, thereby fostering more sustainable lifestyle choices and consumption patterns. These channels may operate independently or complementarily in shaping the relationship between hours worked and environmental pressures.

However, two key issues must be considered when advocating this policy option to address climate change. First, although productivity has increased steadily over recent decades, the past few years – particularly within the EU – have been characterized by stagnation or even slowdown, as documented by several studies (Draghi 2025; Eurostat 2025). This suggests that the available margins may be limited, together with lower growth rates, given the current geopolitical or economic context. In such cases, the behavioural effects of reduced working hours may be more relevant for addressing climate change. Nevertheless, cross-country variations in productivity and growth potential suggest that some countries may still have room to manoeuvre in this area. The second issue concerns the nature of the leisure activities that fill the reduced working hours. Increased leisure might involve environmentally intensive activities (e.g. air travel) and thus not necessarily be less polluting. However, growing evidence suggests that reduced working hours are generally associated with lower greenhouse gas-related behaviours (e.g. less commuting by car and fewer clothing expenditures) (Hanbury, Bader and Moser 2019; Neubert et al. 2022).

Building on a growing literature on the juncture of these two areas, this article first replicates and then extends the empirical analyses linking working time to environmental effects in order to quantify the potentially diverse interrelations between the two with a relatively larger country set in Europe over the past several decades. Compared to similar studies, the article’s novel contribution lies in its exploration of possibly differing effects of working time across the distribution – namely, the quantiles – of environmental indicators, while controlling for a range of independent variables (e.g. sectoral composition, energy intensity and the share of energy from renewables) at the aggregate level. The heterogeneity of linkages between hours worked and environmental variables at different points of the distribution of the latter highlights where there might still be significant margins to benefit from a potential reduction in working time to alleviate environmental pressures.

The main findings confirm that there is a significant and positive relationship between hours worked and environmental pressures across countries in Europe, where the most stable associations with hours worked are estimated for carbon dioxide (CO₂) emissions and carbon footprint variables. In other words, more hours worked are associated with higher CO₂ emissions and a higher carbon footprint, all else being equal. Results are robust when controlling for cross-sectional dependence and slope heterogeneity and when additional control variables are included (thereby addressing omitted variable bias). While reverse causality remains a concern due to other potential omitted variables or in the absence of strictly exogenous instrumental variables within a macro-comparative setting, the relatively stable empirical association between hours worked and several environmental pressure indicators suggests that the reduction of working time may ease pressures on the environment at the aggregate level and thus potentially mitigate an important driver of climate change.

The rest of the article is organized as follows: section 2 reviews the literature and situates the contribution of this article within existing studies; section 3 outlines the methodology, starting with the underlying theoretical framework and empirical strategy for econometric estimations, followed by a description of key variables and data sources; section 4 presents the descriptive analysis and the main estimation results, followed by a discussion of the augmented model results as well as robustness checks; and section 5 concludes the article.

2. Literature review

Academic research studying sustainability issues such as greenhouse gas emissions, biodiversity degradation and air pollution have mostly remained within the boundaries of the ecological field until recently (Schor 2005; Schor 2010; Urban, Chapman and Rizos 2024). However, Grossman and Krueger (1995) introduced the idea of the environmental Kuznets curve, proposing a dynamic relationship (i.e. inverse-U shaped) between various indicators of environmental degradation and per capita income. According to the environmental Kuznets curve, economic development initially leads to a deterioration of the environment, but after a certain level of economic growth, the relationship between environmental degradation and prosperity flips and becomes negative.

Given the increasing evidence – including various reports by the Intergovernmental Panel on Climate Change (IPCC)¹ – supporting the anthropogenic nature of climate change and environmental degradation, a larger span of social science dimensions has been considered to address climate change, including labour market policies, which were rarely considered in the context of climate change in the past (Bohnenberger 2022). One such policy instrument is the reduction of hours worked (Knight, Rosa and Schor 2013; Fitzgerald, Schor and Jorgenson 2018; Fitzgerald, Givens and Briscoe 2024). There has been an emerging interest in exploring the relationship between hours worked and the environment, connecting the two strands of literature that often evolved in siloes until recently. This interest initially emerged among social scientists wishing to broaden the understanding of how human behaviour affects the planet. While the main reason for the growing interest in these issues relates to the role of human behaviour in the climate crisis, as the IPCC report from 2023 suggests, the increasing availability of comparable and harmonized data providing information on a number environmental and labour market indicators has facilitated the pursuit of this new area of research.

Working time reduction is the practice of decreasing the number of working hours over a given period (day, week, month or year) and might involve reduced pay or the same level of pay, depending on the company, industry, region or country. There are several ways to reduce working time: by reducing the number of hours worked per day, while maintaining the same number of days worked per week; by decreasing the number of days worked per week, without changing the number of hours worked per day; by decreasing both the number of hours worked per day and the number of days worked per week, and so on. Working time reduction has been an emerging topic of interest in academic research, particularly in the fields of labour economics and labour sociology, in organizational theory and in psychological studies looking at work–life balance. Various studies have sought to explore the potential effects of working time reduction on various issues, ranging from employment (Victor 2008) and productivity (Garnero, Tondini and Batut 2022) to job satisfaction and work–life balance (Lepinteur 2019). At the individual level, the main findings point to increased job satisfaction, improved work–life balance, individual productivity gains, lower stress levels and thus improved worker well-being (De Spiegelaere and Piasna 2017; Lepinteur 2019; Hanbury et al. 2023). At a more aggregate level, the economic impact of working time reduction is still poorly understood, despite some indications of a slightly positive effect on employment (Raposo and van Ours 2010) and value added (Batut, Garnero and Tondini 2023), which might be suggestive of limitations of redistribution from productivity gains, particularly in countries where margins are small and conventional economic growth paradigm is prevalent.

Increased working time could have several effects on the environment.² First, longer working hours can lead to increased energy and resource consumption in the workplace (e.g. through office heating/cooling and consumption of electricity, water, office materials and so on) and thus result in higher greenhouse gas emissions (Schor 2010; De Spiegelaere and Piasna 2017; Eurofound 2022). Second, longer working hours can impact mobility patterns, which have a bearing on transport emissions (Cerqueira et al. 2020). Third, working time impacts work–life balance (Persson, Larsson and Nässén 2022) and reduced working hours could improve worker well-being, which might lead individuals to make environmentally more sustainable choices (Neubert et al. 2022). At a more aggregate level, proponents of economic degrowth advocate that reduced working time leads to slower economic growth, resulting in both social and environmental benefits (Fitzgerald, Jorgenson and Clark 2015). However, the link between working hours and environmental impact is complex and influenced by various contextual factors, such as industry, location, institutions and individual circumstances. These elements shape how reduced working time affects the environment. Antal et al. (2021) offer a thorough and critical review of the literature on this topic.

Among existing studies, several methodological approaches have emerged to understand the relationship between the environment and working time. The first group of studies looks at the individual or household level, often working with time-use data to relate consumption, production and leisure patterns to various environment-related indicators such as energy use and mobility patterns (Fremstad, Paul and Underwood 2019; Nässén and Larsson 2015; Neubert et al. 2022; Persson, Larsson and Nässén 2022). The second group of studies analyses the linkages between environmental variables (e.g. carbon emissions and fine particulate matter) and average working hours at the regional level within a country, mainly comprised of studies from the United States at the state level (Fitzgerald, Schor and Jorgenson 2018; Fitzgerald 2022; Jorgenson et al. 2020; Mallinson and Cheng 2022). The third group of studies adopts a macro-comparative approach across countries, relying on aggregate-level data on environmental variables such as greenhouse gas emissions, carbon footprint and energy consumption and relating them to conventional macroeconomic variables such as GDP output, population size and employment rate (Hayden and Shandra 2009; Knight, Rosa and Schor 2013; Fitzgerald, Jorgenson and Clark 2015; Fitzgerald, Givens and Briscoe 2024).

Regardless of the level of analysis, the focus of most existing studies has been on high-income countries (Schor 2005; Knight, Rosa and Schor 2013; Shao 2015; Shao and Shen 2017; Fitzgerald, Givens and Briscoe 2024), even though several articles have suggested including lower- and middle-income countries in addition to high-income countries in their sample (Fan et al. 2006; Fitzgerald, Jorgenson and Clark 2015). While the objective of some articles is to provide a transatlantic comparison of aggregate hours worked between Europe and the United States (Alesina, Glaeser and Sacerdote 2005) with implications for the environment (Rosnick and Weisbrot 2007), others focus on the methodological aspects of linking hours worked to environmental outcomes (e.g. Shao 2015; Shao and Shen 2017).

Among the household-level studies, Fremstad, Paul and Underwood (2019) find strong evidence that households with longer working hours emit more CO₂ in the United States, based on calculations of the carbon intensity of goods using input–output tables and combining them with spending data. Using Swedish household time-use and consumption data, Nässén and Larsson (2015) estimate that a decrease in working time by 1 per cent may reduce energy use and greenhouse gas emissions by about 0.7 per cent and 0.8 per cent, respectively. Based on a longitudinal study with Swiss employees, findings by Neubert et al. (2022) suggest that decreased working time is associated with decreased greenhouse gas-related behaviours and increased individual well-being. Using a survey conducted among municipal employees in Gothenburg, Sweden, Persson, Larsson and Nässén (2022) find that working fewer hours improved socio-ecological outcomes among individuals engaged in relatively low-carbon activities. Although it is difficult to draw generalizations from household-level studies, they provide a complementary perspective to the more aggregate studies summarized below.

Focusing on the state level in the United States, Fitzgerald, Schor and Jorgenson (2018) analyse the relationship between CO₂ emissions and working hours over the period 2007–13 and find that there is a strong and positive relationship between working hours and state-level emissions, net of controls for political, economic and demographic drivers of emissions. Mallinson and Cheng (2022) provide a replication study of Fitzgerald, Schor and Jorgenson (2018) and confirm their findings, with even stronger relationships found using data for the period 2014–17. Fitzgerald (2022) elaborates on the previous studies by showing how inequality moderates the relationship between average working hours and CO₂ emissions across US states.

At a more aggregate level, Hayden and Shandra (2009) use a cross-national framework on ecological footprint to explore whether working hours increase environmental impacts, using data from 45 countries. Arguing that ecological footprint is one of the best available aggregate indicators of consumption-based environmental impact, the authors find support for the hypothesis that working hours are positively related to ecological footprint. However, one of the major shortcomings of their analysis is that it is based on a single year of observation (2000). Using a broader framework, Knight, Rosa and Schor (2013) focus on 29 high-income member countries of the Organisation for Economic Co-operation and Development (OECD) over the period 1970–2007 and find evidence of a strong association between working hours and environmental pressures, measured by three environmental variables: ecological footprint, carbon footprint and CO₂ emissions. Focusing on a different environmental variable and working with data from both developing and developed countries over the period 1990–2008, Fitzgerald, Jorgenson and Clark (2015) examine the link between energy consumption and hours worked and study the extent to which the effect of hours worked on energy consumption changes over time. Their results indicate that the effect of working hours on energy consumption has increased over time for all the countries included in their sample. Fitzgerald, Givens and Briscoe (2024) use two relatively new composite indicators of sustainability, namely the carbon intensity of well-being and the ecological intensity of well-being, by taking the ratio of these environmental variables to average life expectancy. Their results, based on longitudinal data from 1970–2017 across a set of high-income OECD countries, suggest that longer working hours are positively associated with the proposed sustainability indicators, after controlling for other key social and economic factors.

This article takes a similar approach to the one used by Knight, Rosa and Schor (2013) and Fitzgerald, Givens and Briscoe (2024), with several differences. First, it uses a slightly different set of countries, focusing on the EU countries as well as Norway, Switzerland and the United Kingdom. The time period covered is longer than that in most existing studies (except Fitzgerald, Givens and Briscoe 2024).³ Second, this article proposes an extension of the commonly used theoretical framework by adding sectoral as well as energy-related additional control variables, which are expected to be strongly associated with environmental pressures and thereby affect the relationship between hours worked and environmental variables. Third, this article uses an extended set of environmental indicators, including CO₂ emissions, carbon and ecological footprints (as in Knight, Rosa and Schor 2013), as well as total energy use (as in Fitzgerald, Jorgenson and Clark 2015). Furthermore, the dynamic empirical specification is similar to a lagged dependent variable model in levels. It therefore somewhat resembles the specification in Fitzgerald, Givens and Briscoe (2024), who estimate a first-differenced model with a lagged dependent variable, but differs from the specification in Knight, Rosa and Schor (2013), who estimate a first-differenced model without a lagged dependent variable. Finally, this article investigates heterogeneities across countries by looking at the distribution of environmental variables through the estimation of quantile regressions. Conducting quantile analysis helps to uncover potential heterogeneities in how hours worked relate to environmental outcomes across different levels of environmental indicators within a macro-comparative framework.

3. Empirical methodology

4. Results

4.1 Descriptive analysis

Table 2 provides key summary statistics of the variables used in the empirical analysis. It is important to note that these numbers correspond to averages over the pooled sample and might vary by country and year. Given that the raw environmental-impact-related variables were significantly large numbers (with many more digits than conventional figures), I report them in billions in their respective units. The empirical estimations use logarithmic conversions.

Table 2. Summary statistics

	N	Mean	Standard deviation
CO2 emissions (Mt)	2 100	127	200
Total carbon footprint (bn gha)	1 129	3 051 934	5 941 546
Total ecological footprint (bn gha)	1 455	3 496 855	7 327 204
Total energy use (bn kg)	1 452	56.93	76.88
Average annual hours worked	1 580	1825.4	229.2
GDP output per hour worked (US$/PPP)	1 516	32.56	19.86
Population (m)	2 160	15 787 995	20 481 224
Employment share (% of population)	2 160	45.3	7.65
Urbanization (% of population)	1 800	67.8	14
Services, value added (% of GDP)	985	59.4	8.8
Energy intensity (kg per €1,000, PPP)	781	185	86.2
Renewable energy sources (%)	505	19.9	14.6
Source: Own calculations.

Figure 1 shows the overall distribution of annual hours worked across the sample countries over the period 1950–2020, suggesting that annual hours worked stood at just over 1,800 hours on average.

Figure 1. Overall distribution of annual hours worked (1950–2020)

Source: Own calculations, based on data from the Conference Board Total Economy Database (1950–2020).

To understand how working hours have evolved over time, figure 2 focuses on the distribution of annual hours worked by decade. The data show a consistent decline in average hours worked since 1950 – from over 2,200 hours in 1950 to approximately 1,600 hours by 2020.

Figure 2. Distribution of hours worked by decade

Source: Own calculations, based on data from the Conference Board Total Economy Database (1950–2020).

While average hours worked have declined steadily over recent decades, labour productivity has increased sharply, as shown in figure 3, likely reflecting rapid technological advancements.

Figure 3. Average labour productivity

Notes: Average annual labour productivity is defined as GDP output per hour across countries in a given year.

Source: Own calculations, based on data from the Penn World Tables (1950–2020).

Figure 4 shows the evolution of the four main environmental indicators used in the empirical analysis. Most of the environmental indicators show a sharp increase until the 1990s, followed by a plateau and slight decline for CO₂ emissions and a relatively strong decline for the footprint variables.⁹ The trends in energy use differ from the others insofar as they mostly increase until the 2010s and then begin to decline.

Figure 4. Evolution of environmental variables

Notes: Linear fits are added to the logarithmic trends of the respective environmental variables. Time breaks in the data are observed around the early 1970s (the oil crisis) for total energy use, and around the early 1990s (the dissolution of the former Soviet Union) for the footprint and energy use variables (but not for CO₂ emissions).

Source: Own calculations, based on CO₂ emissions data from the Global Carbon Budget, total carbon and ecological footprint data from the Global Footprint Network and energy use data from the World Bank.

4.2 Baseline results

The baseline estimation results are provided in table 3, followed by the corresponding coefficient plots in figure 5, where the coefficient estimates of working hours as well as the baseline controls in equation (6) are plotted per environmental variable, providing a quick overview of the main coefficient of interest (i.e. hours worked). This model demonstrates the effect of working time on the environment through its contribution to the output (GDP), thereby measuring its scale effect in line with the underlying STIRPAT framework.¹⁰ The baseline results suggest that increases in annual hours worked are significantly and positively associated with all environmental indicators when the 2DFE method is used, while the coefficients for total footprint and energy use are statistically insignificant when the DCCE method is used. In other words, a reduction in working hours is associated with less pressure on the environment, measured by CO₂ emissions and carbon footprint.¹¹ For example, a 1 per cent increase in average annual hours worked is estimated to increase CO₂ emissions by 0.236 per cent based on the 2DFE method (column 1), and by 0.443 per cent based on the DCCE method (column 2). Similarly, significant and positive coefficient estimates of hours worked are obtained for the carbon footprint indicator. The other coefficient estimates are mostly in line with expectations, such as the lagged dependent variable, which is significant and positive in all 2DFE models.

Table 3. Estimation of the effects of hours worked on environmental indicators (baseline)

	CO2 emissions			Total carbon footprint			Total ecological footprint			Total energy use
	(1)	(2)		(3)	(4)		(5)	(6)		(7)	(8)
Hours worked	0.236*** (0.058)	0.443** (0.198)		0.190*** (0.048)	0.682*** (0.199)		0.273*** (0.045)	0.730 (0.532)		0.149*** (0.046)	0.100 (0.209)
Output per hour	0.070*** (0.014)	0.263*** (0.081)		0.079*** (0.020)	0.135 (0.121)		0.070*** (0.010)	0.091 (0.246)		0.058*** (0.014)	0.164*** (0.063)
Employment share	0.103*** (0.028)	0.368 (0.261)		0.161*** (0.036)	0.541 (0.336)		0.064** (0.029)	-0.230 (0.963)		0.090*** (0.028)	0.561*** (0.173)
Population	0.018 (0.032)	0.215 (0.414)		0.102* (0.050)	1.062** (0.471)		0.158*** (0.032)	1.181 (1.629)		0.034 (0.043)	-0.216 (0.566)
Lagged dependent variables
CO2 emissions (t-1)	0.939*** (0.009)	0.303*** (0.055)
Total carbon footprint (t-1)				0.930*** (0.008)	0.301*** (0.060)
Total ecological footprint (t-1)							0.882*** (0.012)	-0.006 (0.079)
Total energy use (t-1)										0.923*** (0.011)	0.113 (0.113)
Obs.	1 501	1 117		1 016	768		1 224	1 162		1 214	968
Year effects	Yes	Yes		Yes	Yes		Yes	Yes		Yes	Yes
Country fixed effects	Yes	-		Yes	-		Yes	-		Yes	-
R-squared	0.99	0.99		0.99	0.99		0.99	0.98		0.99	0.99

• Notes: All models are logged and include year effects. Columns 1, 3, 5 and 7 are based on the 2DFE model with robust, clustered standard errors, while columns 2, 4, 6 and 8 are based on the DCCE mode, using the mean group estimation method. Coefficient estimates should be interpreted as follows: a 1 per cent increase in hours worked is associated with a 0.236 per cent increase in CO₂ emissions (column 1) and so on. Cluster-robust standard errors are reported in parentheses. *, ** and *** indicate statistical significance at the 10, 5 and 1 per cent levels, respectively.

  Source: Own calculations.

Figure 5. Coefficient plots of baseline model estimations

Note: Coefficient plots are based on the results reported in table 3 (columns 1, 3, 5, and 7), which are all logged and estimated using the 2DFE model. Coefficient estimates in the plots should be interpreted as follows: e.g. a 1 per cent increase in hours worked is associated with a 0.236 per cent increase in CO₂ emissions (upper left panel) and so on. More detailed estimation results are reported in table 3.

Source: Own calculations.

4.3 Augmented model estimations with additional controls

To better capture linkages not addressed in the baseline model, I extend the model with additional controls (e.g. urbanization, sectoral structure, energy intensity and energy mix). This helps to assess whether the relationship between working hours and environmental indicators remains robust. Table 4 presents two sets of estimates per outcome: one excluding renewable energy share, and one including it, though the latter has a smaller sample due to limited time-series data. In light of the reduced sample size (T<30) and given that the tests generally fail to reject the null hypothesis of weak cross-sectional dependence – especially in the most comprehensive models – the 2DFE estimator is preferred over the DCCE, as the former performs better asymptotically under such conditions (Thombs 2022).

Table 4. Estimation of the effects of hours worked on environmental indicators (augmented)

	CO2 emissions			Total carbon footprint			Total ecological footprint			Total energy use
	(1)	(2)		(3)	(4)		(5)	(6)		(7)	(8)
Hours worked	0.434*** (0.112)	0.446** (0.199)		0.260** (0.124)	0.329* (0.183)		0.188 (0.125)	0.130 (0.219)		0.249** (0.108)	0.197 (0.184)
Output per hour	0.096 (0.066)	0.199* (0.113)		0.227*** (0.034)	0.256*** (0.056)		0.051* (0.030)	0.163** (0.067)		0.230*** (0.071)	0.486*** (0.101)
Employment share	0.225 (0.134)	0.279* (0.140)		0.593*** (0.062)	0.771*** (0.106)		0.178*** (0.058)	0.242** (0.101)		0.471*** (0.076)	0.761*** (0.109)
Population	0.002 (0.076)	0.034 (0.131)		0.421*** (0.069)	0.489*** (0.106)		0.131** (0.059)	0.102 (0.110)		0.280*** (0.081)	0.550*** (0.141)
Lagged dependent variables
CO2 emissions (t-1)	0.780*** (0.051)	0.687*** (0.062)
Total carbon footprint (t-1)				0.634*** (0.028)	0.519*** (0.046)
Total ecological footprint (t-1)							0.725*** (0.032)	0.596*** (0.050)
Total energy use (t-1)										0.523*** (0.045)	0.321*** (0.052)
Additional controls
Services sector as % of GDP	-0.171* (0.099)	-0.187 (0.117)		-0.037 (0.071)	-0.058 (0.110)		-0.181*** (0.064)	-0.167 (0.116)		-0.048 (0.076)	-0.042 (0.098)
Share of urbanization	-0.104 (0.106)	-0.245 (0.222)		-0.031 (0.092)	0.024 (0.213)		-0.006 (0.099)	-0.234 (0.251)		-0.170* (0.098)	-0.039 (0.165)
Energy intensity	0.071 (0.056)	0.285** (0.109)		0.179*** (0.029)	0.328*** (0.050)		0.031 (0.023)	0.167*** (0.056)		0.194** (0.070)	0.494*** (0.073)
Share of renewable energy		-0.043*** (0.011)			-0.038** (0.039)			-0.052*** (0.012)			-0.042*** (0.009)
Obs.	725	448		506	286		638	364		602	329
Year effects	Yes	Yes		Yes	Yes		Yes	Yes		Yes	Yes
Country fixed effects	Yes	No		Yes	No		Yes	No		Yes	No
CD statistic (p-value)	0.085	0.608		0.004	0.023		0.015	0.131		0.017	0.219
R-squared	0.9976	0.9942		0.9985	0.9972		0.9995	0.9984		0.9980	0.9927

• Notes: All models are logged and estimated using the 2DFE model. Given that the null hypothesis of weak cross-sectional dependence cannot be rejected at 1 per cent for all specifications (except in column 3) and that the time dimension of the panel data is relatively short, it was not appropriate to use the DCCE estimator. Coefficient estimates should be interpreted as follows: a 1 per cent increase in hours worked is associated with a 0.434 per cent increase in CO₂ emissions (column 1) and so on. Cluster-robust standard errors are reported in parentheses. *, ** and *** indicate statistical significance at the 10, 5 and 1 per cent levels, respectively.

  Source: Own calculations.

Table 4 shows that longer working hours are significantly associated with higher CO₂ emissions and carbon footprints, even after accounting for lagged dependent variables and controls such as the share of the service sector in GDP, urbanization, energy intensity and renewable energy share. Employment share and population size are also positively and significantly linked to environmental pressures in most models. Urbanization is not statistically significant, probably because its effects are captured by population and employment variables. The service sector share is significantly estimated with a negative sign only in the specification that includes ecological footprint, aligning with expectations that a larger service sector is associated with lower environmental impact due to the level of technological development in industrialized countries (though not always). On the contrary, energy intensity, reflecting manufacturing activity, shows a strong positive association with environmental pressures. Finally, renewable energy share is significantly and negatively associated with environmental outcomes, indicating its mitigating effect on the latter. However, once this variable is included, the significance of working hours also disappears – probably because the type of energy source explains much of the variation in environmental outcomes. Figure 6 summarizes the main coefficient estimates for hours worked across environmental indicators, showing two values per indicator – one excluding and one including the renewable energy variable, as in table 4.

Figure 6. Coefficient plots of hours against environmental indicators (augmented models)

Notes: Coefficient plots are based on the results from table 4, which are all logged and estimated using the 2DFE model. Only the coefficient estimates of hours worked are plotted. Specifically, estimates indicated with square symbols (the upper ones in blue) are based on results from columns 1, 3, 5 and 7 of table 4, while those indicated with circle symbols (the lower ones in red) are based on results from columns 2, 4, 6 and 8 of table 4, which include additional controls. Coefficient estimates in the plots should be interpreted as follows, taking the upper left panel as an example: a 1 per cent increase in hours worked is associated with a 0.434 per cent increase in CO₂ emissions (blue square) in the more parsimonious specification (column 1, table 4) and with a 0.446 per cent increase in CO₂ emissions (red circle) in the most extended specification (column 2, table 4).

Source: Own calculations.

4.4 Quantile regressions of environmental variables on hours worked

Thus far, results have focused on estimating the conditional mean of environmental outcomes in relation to working hours. In this final section, I extend the model using quantile regression to explore how the relationship varies across the wider distribution of environmental indicators. Figure 7 summarizes these results, showing the estimated effects of working hours at different quantiles, with shaded areas representing confidence intervals.

Figure 7. Quantile regressions of impacts of hours worked on environmental indicators

Notes: All models are logged and estimated using quantile regression at the 1st, 20th, 40th, 60th, 80th and 99th quantiles, using the baseline specification. Only the coefficient estimates of hours worked (along with confidence intervals shown in shaded areas) are plotted against the distribution of the respective environmental indicator.

Source: Own calculations.

As regards CO₂ emissions, there seems to be a stark declining pattern of the hours worked coefficient, suggesting that the estimated effect of hours worked is larger at the lower quantiles of CO₂ emissions than at the higher ones towards the right tail of the distribution in figure 7. In other words, the estimated effect of hours on emissions is larger in countries with relatively lower emissions. On the one hand, this could indicate that countries with longer working hours have greater scope for policy manoeuvre in working hours regulation, which may have a substantial effect on emissions (the potential causal link). On the other hand, the quantile results could also be evidence that wealthier countries tend to combine lower emissions per capita and shorter working hours, and that the environmental Kuznets curve decreases steeply when working hours start to decrease. A similar downward trend – albeit slightly flatter than that for CO₂ emissions – is observed for total carbon footprint and energy use as outcome variables. For example, more hours worked seem to be associated with a significant carbon footprint approximately between the 20th and 60th quantiles, but the estimated coefficient is not statistically significant at the tails (both left and right) of the carbon footprint distribution. As regards energy use, a pattern (albeit a declining one) seems to emerge in which an increase in hours worked is associated with higher energy demand across most of the distribution; however, the estimated link does not seem to hold for the highest quantiles of the distribution (about one fifth of the right tail). It is possible that, in countries with the highest energy demand, other structural factors – such as the economy’s production and/or consumption patterns – drive energy demand, so that even a reduction in hours worked is unable to offset this effect.

5. Conclusions

Growing evidence of human-driven climate change has expanded policy efforts beyond the ecological domain. This article contributes to the emerging body of work at the intersection of labour economics and environmental research by examining the role of working hours on environmental pressures. Leveraging harmonized cross-country panel data from 30 countries in Europe over several decades, it explores how working hours relate to four environmental indicators: CO₂ emissions, carbon footprint, ecological footprint and energy use. Some estimations rely on shorter panels due to data availability. Using a theoretical framework informed by previous research and extended with new dimensions, this article provides quantitative estimates of the impact of working hours on environmental pressures. It also investigates how these relationships vary across the distribution of environmental outcomes, offering insights on heterogeneities as well as potential areas for policy intervention.

Descriptive analysis shows that while average working hours have declined over recent decades, productivity per hour has increased. Environmental indicators worsened until the 1980s–1990s but have generally stabilized since then. The baseline econometric model builds on existing literature (e.g. Knight, Rosa and Schor 2013; Fitzgerald, Givens and Briscoe 2024) and extends it by applying a dynamic STIRPAT framework and testing for cross-sectional dependence and directional asymmetry. Results reveal a significant positive link between working hours and most environmental indicators, especially CO₂ emissions. Adding control variables does not, in general, alter this relationship, with the exception of energy use. Further analysis of heterogeneity across the distributions of environmental variables shows that the impact of working hours is strongest at lower quantiles of CO₂ emissions. To the best of my knowledge, this is the first study to explore such distributional dimensions, offering a promising direction for future research.

Although the established empirical link between working hours and environmental pressures is robust and significant to various extensions of the model employed in this article, there are some limitations. First, while the models have moved beyond the theoretical framework of the STIRPAT model through inclusion of lagged dependent variables, additional control variables – somewhat addressing omitted variable bias – and the stability of the results to various robustness checks are reassuring, the causality of the results cannot be claimed. Extending the current theoretical model would require further research, and quasi-experimental analyses based on working time reduction reforms in one or more countries would have to be carried out to explore potential causal relationships. Second, it is important to understand the extent of working time reduction and investigate whether reduced hours of work might lead to rebound effects (e.g. increased air travel or consumption of luxury goods with high ecological footprints), even though emerging research suggests that this is not necessarily the case (Neubert et al. 2022). Moreover, the institutional and economic context of a country – for example, whether a country is able to maintain economic output at constant levels to allow the translation of productivity growth into reduced working time – determines the feasibility and desirability of such an instrument, if there are still margins left to distribute gains from productivity as reduced working hours. In a similar vein, the way working time reduction is implemented (e.g. voluntary, at sectoral or national level, level of pay, etc.) also has a bearing on how it affects environmental pressures.

Given these considerations, I remain cautious about the causal interpretation based on the empirical links between hours worked and environmental pressures, as the findings could also be related to a steeply decreasing environmental Kuznets curve with decreasing working hours. In other words, the demonstrated associations between working hours and environmental indicators might reflect the features of an environmental Kuznets curve, whereby wealthier countries tend to combine lower working hours with lower emissions per capita (except for countries like the United States). In this case, the observed association might not necessarily be evidence that working time reduction reform could be an effective environmental policy tool, but rather reflect broader structural development patterns. For these reasons, further research is needed to examine the relationship between working time reduction and environmental pressures.

Acknowledgements

I would like to thank Jakob Wall for his excellent research assistance and support with data collection and preparation during the early phases of writing this article. Feedback and comments received during the “A just transition beyond growth?” conference at the European Trade Union Institute (December 2022), the 8th Regulating for Decent Work Conference (July 2023), the European Work-Time Network webinar (September 2023), the Trade Union Related Economists network meeting (June 2024), the International Labour and Employment Relations Association (ILERA) conference (June 2024) and the Interuniversity Research Centre on Globalization and Work (CRIMT) conference (November 2024) were greatly appreciated. Finally, I would like to thank two anonymous referees for their very helpful and constructive feedback, which helped me tremendously to improve this article.

Competing interests

The author declares that they have no competing interests.