2.1. Defining “green jobs”

There is no harmonized definition or categorization of “green jobs” at the international or even European levels. An initial broad definition was given by UNEP (2008, 3), as jobs “that contribute substantially to preserving or restoring environmental quality”. Two main approaches can then be identified: top-down approaches, which classify all jobs within a “green” sector as green, and bottom-up approaches, which differentiate green and non-green jobs at a more granular level – usually at the job level – based on job titles or other occupation-specific characteristics such as task content. While top-down approaches are useful for examining meso-economic trends, they are less appropriate for assessing job quality in terms of wages, type of employment contract, working hours and working conditions. These elements are better measured at the job level.

In light of the above, this article adopts a bottom-up approach. However, various bottom-up approaches exist, and different methods have been developed to identify green jobs at the occupational level. The most widely known and internationally applied is the O*NET green jobs classification developed by the US Department of Labor (Dierdorff et al. 2009 and 2011), which has been adapted to many countries (de la Vega, Porto and Cerimelo 2024; Elliott et al. 2024; Sofroniou and Anderson 2021; Valero et al. 2021). Developed at the end of the 2000s, it identifies three categories of jobs affected by the ecological transition: “green new and emerging” (GNE) jobs, referring to jobs introduced by the ecological transition; “green enhanced skills” (GES) jobs, referring to already existing jobs whose tasks and skills are evolving internally to adapt to the environmental challenge; and “green increased demand” (GID) jobs, which refers to jobs that are expected to grow in volume to support the expansion of the first two categories, without significant changes in the nature of the work.

However, the O*NET approach to green jobs has come under increasing criticism (Villani et al. 2026). The first limitation identified relates to the lack of substantial updates since the initial classification of green jobs in 2009–11. Only minor task additions have been made and the last O*NET release to officially include this green classification was version 24.1 in 2019, after which it was discontinued. More fundamentally, a second limitation relates to how green jobs and green tasks are articulated in O*NET. In spite of its unique approach linking each occupation to a set of (green and non-green) tasks, the green nature of a job was not initially defined by the number or intensity of green tasks actually performed. Instead, green jobs were identified through a review of 60 publications and then classified into one of the three categories based on their alignment with the existing O*NET nomenclature.¹ Green tasks were identified ex post only for jobs from the first two categories (GNE and GES), arbitrarily excluding GID jobs, even though some clearly involved a high intensity of green tasks (Granata and Posadas 2024, 25; OECD 2024, 80). Lastly, the 60 publications reviewed covered only 12 sectors, excluding occupations in other sectors that nonetheless contribute to environmental protection. This limitation essentially reproduces a shortcoming of the top-down approach.

These limitations, which apply to US data, are compounded by another problem when the O*NET categories are adapted to other countries. The crosswalk methods suffer from several shortcomings, mainly relating to aggregation bias due to mismatches between different occupational classifications. The accuracy of these crosswalks has been questioned (Bachelot 2024), and in a report focused on European adaptations, Vona (2021, 24) concluded that their quality “is not sufficient to use [adapted O*NET categories] to measure green employment in Europe”. Therefore, for countries other than the United States, a national ad hoc definition and measure of green jobs would be more appropriate.

France has such an approach, developed by the National Observatory for Employment and Professions of the Green Economy (Onemev), which has worked to identify and monitor green jobs since 2010. The definition proposed by Onemev was revised in 2020 in line with the new French classification of occupations and socio-professional categories (PCS-2020). It now consists of a list of more than 140 green occupations, updated each year and defined at the most detailed level of the French nomenclature – occupational titles. This narrowly defined list covers the “core” green jobs – those most clearly associated with supporting the ecological transition. While not the largest category of green jobs, nor expected to become so, they are strategically critical (Fontaine et al. 2023). Furthermore, this definition is not limited to occupations in the environmental goods and services sector, which is a criticism sometimes levelled at “purist” overly restrictive definitions (e.g. Sofroniou and Anderson 2021).

The approach developed by Onemev classically defines green jobs as “jobs which, through their purpose and/or the skills they employ, contribute to measuring, preventing, controlling and correcting negative impacts and damage to the environment”.² These occupations were identified through an automated textual analysis of job titles collected in a large, nationally representative French survey.³ Despite the large number of park and garden workers, the list includes a wide range of occupations, such as waste sorting agents, asbestos removers, air pollution technicians, energy research and development engineers and renewable energy managers.⁴

Using the Onemev definition for our empirical analysis has several advantages. First, it is directly based on French occupation titles, which avoids translation issues and, most importantly, is more likely to reflect the greenness of jobs in this specific national context. Second, it avoids the crosswalk-related problems that arise when adapting the O*NET approach, which may jeopardize our identification of job quality characteristics in green occupations. In other words, if occupations that contribute substantially to preserving or restoring the environment were grouped – as a result of aggregation bias – with others that have little environmental relevance, our measure of green job quality would be fundamentally flawed.

As there is no perfect definition of green jobs, we also need to acknowledge the limits of the Onemev definition. The main limitation is that it is not transposable to other countries for comparative purposes. However, the conclusions that can be drawn about the relationship between the greenness of jobs and their quality will nevertheless be useful, since the definition ensures that only jobs that contribute substantially to preserving or restoring environmental quality are included in the aggregate.

Another limitation concerns the skill distribution of workers identified empirically with the Onemev definition. Compared with other approaches – particularly O*NET – the Onemev definition leads to a larger proportion of manual workers, especially those working directly with nature, such as park and garden workers,⁵ and a relatively smaller proportion of high-skilled workers, especially professionals and managers (Bachelot 2024). While the over-representation of managers under the O*NET definition has already been noted in other analyses (Maldonado et al. 2024), the preponderance of manual workers under the Omenev definition also seems to be consistent with more prospective French sectoral approaches that highlight the need for blue-collar workers in the ecological transition (Sciberras et al. 2022). Moreover, this skill distribution reflects the structure of the French green labour market and not just a “bias” in the Onemev list of green jobs, since it identifies as many high-skilled job titles as low-skilled ones: of the 143 listed in 2022, 70 refer to managers, directors, engineers, technicians, researchers and consultants, among others. In order to address this difference in composition, we will systematically analyse the job quality of green jobs separately for high-skilled and low-skilled jobs.

2.2. Literature

The literature on the quality of green jobs can be divided into two main strands: (i) quantitative analyses based on various definitions of green jobs that offer a global view of the average quality of green jobs; and (ii) qualitative analyses focusing on a single sector or job particularly exposed to changes stemming from the ecological transition and that shed light on the greening of local working practices.

As regards quantitative analyses, Apostel and Barslund (2024) provide an overview of the literature on developed countries, while Bachelot (2023) focuses on France. In general – and often based on O*NET categories – the literature on countries other than France finds that wages are higher for green jobs (Bluedorn et al. 2023; Bowen, Kuralbayeva and Tipoe 2018; OECD 2024; Valero et al. 2021; Vona, Marin and Consoli 2019), although this premium has diminished over time (OECD 2023; Sato et al. 2023). However, quantitative data on non-pecuniary dimensions of job quality are much rarer. In developed countries, including France (SDES 2023), green jobs appear to be associated with better employment contracts – that is, more secure and less part-time (Peters 2014; OECD 2024; Valero et al. 2021) – whereas the opposite is true in developing countries, where greenness is associated with informality (de la Vega, Porto and Cerimelo 2024; Mathieu 2024).

Quantitative analyses of working conditions are scarce, but an analysis of the French case – based on previous definitions of green and greening jobs – shows that green jobs are more exposed to arduous working conditions, such as mechanical vibrations, harmful noise levels and carcinogenic agents (Havet, Bayart and Penot 2021). This finding holds even when controlling for individual and job characteristics, suggesting a penalty specific to green jobs. In terms of socio-demographic profiles, a particularly striking feature is that, in every country – both in descriptive statistics and when controlling for other factors – green jobs are overwhelmingly held by men (around 80 per cent in France) (Valero et al. 2021).

Regarding the skill levels of these jobs and their holders, the results differ greatly depending on the definition applied. As noted above, the O*NET definition tends to identify higher-skilled occupations. Conversely, most jobs captured by the Onemev definition require no or low qualifications, owing in particular to a high proportion of blue-collar workers (Babet and Margontier 2017).

The second strand in the literature sheds light on sectors at the heart of the ecological transition, addressing other dimensions of job quality – in particular working conditions. These studies document precarious employment and poor working conditions, for instance, in the waste management sector (Corteel and Le Lay 2011; Gregson et al. 2016; Kirov and van den Berge, 2012) or in organic agriculture (Dumont and Baret 2017; Guérillot 2024). Other studies have examined the greening of work practices in jobs or sectors that are not classified as “green”, but which are nonetheless being reconfigured by the integration of environmental considerations (Clos 2022; Zarka 2023). By shedding light on the constraints and intensification that can result from environmental regulations or even voluntary standards (Mahlaoui 2023; Rieucau et al. 2024), this qualitative literature offers valuable insights into the potentially ambiguous effects of greening on working conditions.

Our study is, however, more in line with the quantitative approaches described above. It enhances them in several ways. First, it is the first to operationalize the definition of green jobs recently revised by the French public statistics system, addressing the limitations of both O*NET adaptations and previous definitions applied in France. Together with Havet, Bayart and Penot (2021), it is also the first study to explore the quality of green jobs in France using a multi-dimensional framework that goes beyond wages or skills, conducting controlled regressions for each dimension of job quality. Lastly, it draws on a large representative French survey, whereas previous institutional publications (e.g. Babet and Margontier 2017; SDES 2023) have generally used census data that contain limited information on job quality.

5.1. What are green jobs and who does them?

Descriptive statistics on green job holders (table SA1 in the supplementary online appendix) indicate that men are over-represented in green jobs, while the age distribution is similar to that of other jobs. In terms of occupation, compared with all employees, most workers in green jobs are skilled industrial and blue-collar workers (55.8 vs 14.8 per cent). Conversely, there are almost no clerical, sales and services employees in green jobs (1.9 vs 29.5 per cent). Managers and intellectual professionals are slightly under-represented (16.7 vs 22.7 per cent), and technicians and associate professionals are even more so (18.2 vs 26.1 per cent). Reflecting this occupational structure, workers in green jobs, on average, have lower levels of education: 52.2 per cent of them do not hold the French baccalaureate – the national upper secondary diploma – compared with 33.8 per cent of all employees.

Green job holders are more likely to be employed by territorial authorities (28.7 vs 7.2 per cent) and by private individual employers (5.5 vs 2.4 per cent), but less likely to be employed in the private sector and by non-governmental organizations (54.9 vs 70.5 per cent). They are less likely to work in firms with 250 employees or more (19 vs 25.3 per cent) and more likely to work in firms with 10 to 49 employees (31.5 vs 27.3 per cent).

They are largely over-represented in mining and manufacturing (20.4 vs 13.7 per cent); in professional, scientific and technical activities, and administrative and support service activities (22.6 vs 10.2 per cent); and, to a lesser extent, in public administration, education, human health and social work (37.5 vs 33 per cent) and in other services (8.8 vs 5.6 per cent). Conversely, green jobs are rare in wholesale and retail trade, transportation and storage, and accommodation and food service activities (4.1 vs 22.4 per cent).

These initial descriptive statistics illustrate and confirm the specificity of the Onemev definition of green jobs. As expected, low-skilled workers are over-represented in green jobs. Examining the detailed occupations of workers in these jobs (see table SA2 in the supplementary online appendix), about 45 per cent of them are park and garden workers, often employed by territorial authorities in France. Skilled and unskilled workers from the water, energy and waste treatment industry account for 14.1 per cent of green jobs. Technicians, as well as engineers and technical managers in agriculture, aquaculture, forestry and environment protection represent 11.5 per cent. Lastly, engineers, managers and technicians in quality control and safety risk management make up 9.5 per cent.

5.2. Are green jobs better than non-green jobs?

We run controlled regressions to observe whether having a green job increases or decreases the probability of good job characteristics, using the indicators presented in table 1 as the dependent variables. We also explore differences between high- and low-skilled workers, as defined in section 4, by running all regressions for each subgroup separately.⁹

5.3. Are all green jobs similar in terms of job quality? Building a typology

To move beyond these average (or skill-split) comparisons between green and non-green jobs, we develop a typology of green jobs based on job quality. Given the qualitative nature of most of our job quality variables, we run an MCA based on the sample of workers in green jobs.¹³ We restrict our analysis to the first five axes, which are identified as the only ones containing meaningful information: they explain 53.26 per cent of total inertia in the data, while additional axes provide no further useful information.¹⁴ Using the five dimensions generated by the MCA, we run an HAC to produce the job quality typology.

Four groups emerge, both because this partition presents the greatest relative loss of inertia (see dendrogram in figure SA1 in the supplementary online appendix) and because it is analytically relevant. Figure 1 plots both categories and observations among clusters on the first two axes of the MCA. The first axis is characterized by variables related to wages and socio-economic security, while the second axis is mainly associated with working time organization and working conditions.

Figure 1. Plot of observations and categories (first two axes)

Notes: Dots represent individuals and triangles represent categories of variables. To avoid visual overload, only categories whose squared cosines are greater than 0.3 are displayed in bold.

Source: Our own calculations based on French labour force survey (2021–22).

We label the first cluster, accounting for 22.4 per cent of employees in green jobs, “Good green jobs”. Located on the upper left-hand side of the graph, it comprises workers with high levels of job quality. It is characterized by high earnings (€2,000–€2,999 and €3,000+), stable employment (permanent or civil service contracts) and full-time work. Workers in this group have more frequent access to training than those in other clusters, greater control over working hours and more possibilities for remote work. Women, highly educated workers, managers and professionals are over-represented. We also find more workers employed by the state (58 per cent) or by large firms (with 50–249 and more than 250 employees).

The second cluster, which we label “Green stable average quality jobs”, is the largest group, representing 60.6 per cent of green employees. Located at the bottom of the graph, it comprises workers with stable full-time jobs but with lower earnings and less control over working hours than the first cluster. They earn monthly net wages between €1,000 and €1,999, with most concentrated in the €1,500–€1,999 bracket. They also have stable (either permanent or civil service contracts), full-time contracts, and underemployment is therefore rare. Men are significantly over-represented (88 vs 82 per cent). Workers are often employed within the public administration sector, mostly by territorial authorities, accounting for 80 per cent of green jobs in this cluster. In terms of education and skills, workers mainly have secondary education and are more likely to be blue-collar workers or unskilled clerical workers.

The last two clusters are smaller and composed of workers with lower levels of job quality than in the first two groups. They appear on the right-hand side of the graph, which is marked by low earnings.

The third cluster, representing 10.1 per cent of green jobs, is labelled “Green odd jobs”. As the name suggests, and in contrast to the fourth cluster, these jobs often involve “voluntary” part-time or student contracts (trainees and apprenticeships). Monthly net wages are very low (less than €1,000), and working hours are more frequently irregular – although workers have more control over them. There are no opportunities for remote work and 70 per cent of workers are employed by small firms (1–9 employees) or individual employers. This group mainly includes skilled blue-collar workers with no formal education – possibly because they are still in education. In view of these characteristics, the poorer quality of these jobs may be more transitory than in the last group. However, this does not make them desirable or sustainable, especially if green jobs appear to be over-represented in this type of low-quality employment (compared with the rest of the economy), suggesting that the ecological transition may increasingly rely on objectively degraded jobs.

The fourth cluster, grouping the remaining 7 per cent of green jobs, is labelled “Precarious green jobs”. It is mainly composed of underemployed workers in involuntary part-time work, on fixed-term or temporary contracts, earning less than €1,000 per month. These workers have no control over their working hours and lack access to training or remote work. As in the previous group, workers with no education and skilled blue-collar workers are over-represented. They are also more likely to be employed in the public administration, education, human health and social work sector (52 per cent).

This typology illustrates, beyond the average results of section 5.2, the strong heterogeneity in the quality of green jobs. This challenges the “good green jobs” narrative, since a significant share of green jobs display very poor job quality, especially with respect to socio-economic security.

Lastly, although the clusters may initially appear to reflect differences in skill level and occupations, closer analysis suggests that this is insufficient to explain job quality. Indeed, several occupations are broadly distributed across two or even three clusters (see table SA3 in the supplementary online appendix). For example, park and garden occupations account for a large share of our low-skilled sample and low-quality clusters, but some are nonetheless found in the first “Good green jobs” cluster, albeit in a much smaller proportion.

This clearly shows that there is no “occupational determinism” in terms of job quality but, instead, draws attention to institutional embeddedness. For instance, the employer and the sector appear to play an important role, with some being over-represented in low-quality jobs – such as individual employers in the third group and the broad public administration sector in the fourth. In this respect, the public sector offers a direct policy lever for improving job quality, as it can provide other segments of the green workforce with good jobs, as shown by its strong presence as an employer in the first and second clusters.