Policy Research Working Paper 10753 Why Look at Tasks When Designing Skills Policy for the Green Transition? A Methodological Note on How to Identify Green Occupations and the Skills They Require Julia Granata Josefina Posadas Social Protection and Jobs Global Practice April 2024 Policy Research Working Paper 10753 Abstract The coexistence of several definitions of green jobs and methodology with a dictionary of green terms for iden- measurement instruments gives room for mismatches tifying green tasks and occupations. This methodology, between those concepts and their application to research utilizing text analysis, demonstrates superior performance questions. This paper first presents an organizing frame- compared to the well-known O*NET Green Economic work for the existing definitions, measurement instruments, Project classification, particularly for developing countries. and policy frameworks. It then delves into discussing two Lastly, the paper applies this methodology to Indonesia, a appropriate approaches for identifying green occupations middle-income country, and utilizes various data sources to guide skills development policy: the task-content and to showcase the utility of the dictionary and text analysis the skills approaches. In the process, it introduces a novel exercise. This paper is a product of the Social Protection and Jobs Global Practice. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at jposadas@worldbank.org and m.julia.granata@gmail.com. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Why Look at Tasks When Designing Skills Policy for the Green Transition? A Methodological Note on How to Identify Green Occupations and the Skills They Require Julia Granata1 and Josefina Posadas2, 3 JEL classification: J21, J24, J62, O33, O53, O57, P18, Q58 Keywords: green occupations, green skills, green employment, task-content approach, skills anticipation, O*NET, ISCO-08, text analysis, language 1 Labor Economist and World Bank Consultant (m.julia.granata@gmail.com) 2 Lead Economist (jposadas@worldbank.org) 3 This note was financed with the support of the Jobs After COVID-19 MDTF. We thank Wisnu Harto Adi Wijoyo for his excellent support with part of the data analysis, and peer reviewers Wendy Cunningham, Elizabeth Ruppert Bulmer, Ulrike Lher, and Achim Schmillen for the comments provided. 1. Motivation The greening of the economy, like other mega-trends, is changing the world of work. Identifying the occupations that will be in demand in the green economy—often simply referred to as green jobs or green occupations—and the skills they require is paramount to supporting and enabling an environmentally sustainable planet while contributing to economic development. New industries that produce environmentally friendly goods and services for intermediate or final consumption are emerging. Moreover, the greening of the economy requires highly polluting industries not necessarily to disappear, but to transform by implementing different technologies to become more sustainable. This economic transformation is impacting occupations to varying degrees, and it becomes critical to identify which ones are changing and what skills they require to make the green transformation possible. Despite the importance of having a common definition of green jobs to be used by researchers and policy makers, there is currently no consensus on a single concept yet, let alone how to identify them in practice. On the one hand, there is agreement that green jobs aim to reduce environmental impact by both producing environmentally friendly outputs and promoting environmentally friendly production processes. On the other hand, the scope of what green jobs should encompass varies. For some, a definition of green jobs should include all jobs that are somehow associated with the greening of the economy, regardless of whether the job itself contributes to reducing the impact on the environment. Supporters of this view include all environmentally neutral jobs, that is, jobs that do not harm the environment even if they do not help to preserve it. Others include what is usually referred to as "greening jobs" or jobs for a greener economy or indirect green jobs. These are jobs that do not directly contribute to preserving the environment but might be in-demand as part of the green transformation. Lastly, some institutions like the ILO add the "decent job" layer to the definition: a green job not only needs not to damage the environment, but it also needs to pay adequate wages and provide worker protection. The lack of convergence stems from the unfeasibility of translating the concepts into a single measurement methodology that suits existing data (or cost-effective, collectible new data) and different policy questions. Many policy makers prioritize supporting workers moving out from brown jobs to other green jobs. Researchers instead have a broader set of questions, for which they use different definitions and deploy very different specialized surveys. These surveys are usually a good tool for specific questions at hand, but generalizing their use to other policy questions creates important biases. As a single definition may not be possible—or even desirable—in view of the multiple policy questions, this paper seeks to identify which occupations and skills should be the focus of skills development policies needed to support the greening of the economy. Put simply, it aims to identify which occupations may require new skills training to support the green economy. With this objective in mind, this paper shows that green jobs should be understood as those that have specific tasks assigned to lessen consumers' and/or firms' environmental impact. These green tasks can either be assigned with the goal of producing greener outputs or of reducing the firm's environmental footprint. The proposed methodological approach helps to identify the skills that will be needed to carry out such tasks. Some of these jobs will require core skills, some of which will be transversal, while others would be specific 2 technical green skills. This paper leaves out of consideration those indirect or neutral jobs, as it is assumed that advanced labor market information systems would provide the signals to the workforce of which jobs are in demand and hence worth pursuing. This paper also excludes the worker's protection layer from the definition since it is not the main objective of the skills policy, and it is addressed by another set of policy instruments. This note contributes to the understanding of skills policy in the context of the green transition as follows. Section 2 presents a conceptual framework to identify green jobs, departing from the state-of- the-art definition of the production function and resulting in four measurement approaches. Then, it discusses their suitability for skills policy along with their data requirements and limitations when applied in advanced and developing countries, becoming a reference point in the literature. Section 3 discusses the methodologies that can be used to identify green occupations under the task approach and introduces one based on text analysis. Specifically, the proposed methodology constructs a dictionary of green terms and applies text analysis to an occupational database containing task statements with the objective of measuring the green task intensity of occupations. This section also discusses the strengths and weaknesses of the methodology vis-à-vis relying on the O*NET Green Economy Program, a well-recognized list of green occupations in the U.S. Finally, section 4 presents analytic applications of the green occupation taxonomy obtained to the case of Indonesia to illustrate the potential of the proposed methodology. More specifically, it describes selected green occupations' characteristics: gender, education, in-demand skills, and wage premium. Figure 1 visually summarizes the organization of this note. The intended audience of this paper is analysts and policy makers working on skills policy around the world. The proposed definition and measurement methodology can be deployed in any country that relies on ISCO-08 or any other country that uses an occupational classification system containing task statements. We are making available to the community of analysts a toolkit that includes: (i) a readme file; (ii) the dictionary of green terms, which can be easily modified to reflect the country's green economic development; (iii) the STATA do files that classify green tasks and create the green task intensity index for each occupation – with minor modifications, this do-file can be applied to any occupational database with task statements; and (iv) an excel file with the classification of all task statements in ISCO-08 into green, green potential, and non-green, as well as all 4-digit occupations with their respective green task intensity index. The recommendation to analysts is to directly use the excel with the classification if working with a database of occupations that uses ISCO-08 at the 4-digit level. However, if using an occupational database with task statements at a more disaggregated level, running the toolkit to get a more accurate description of the occupation is preferred. 3 Figure 1. Visual organization of the note 2. Green jobs and methodologies to measure them A. A conceptual framework to understand how to align measurement with policy questions Green jobs are being widely discussed in academic literature and among policy makers. Although there is no consensus yet on a definition or on how to measure them, it is widely agreed that green jobs aim to reduce negative environmental impact by both producing environmentally friendly outputs (goods and services) and promoting environmentally friendly production processes. The term 'environmentally friendly' is usually applied and denotes reducing and limiting energy and raw materials consumption, greenhouse gas (GHG) emissions, waste and pollution, protecting and restoring ecosystems, and enabling adaptation to climate change. 4 The scope of the definitions also varies. For example, some envision that all jobs will eventually be green and therefore use definitions that encompass jobs with zero environmental footprint—that is, jobs that neither pollute nor have a negative impact on the environment—and/or jobs that are in demand due to the greening of the economy, even if the job itself does not directly contribute to reducing environmental impact. 5 Others, however, aim for more focused definitions that specifically include jobs that directly contribute to reducing the environmental footprint. This paper follows the latter approach with the objective of advising governments on priority areas. 6 4 Vidican Auktor (2020), ILO (2016), ILO (2019), Eurostat (2009), and BLS (n.d.). Gregg, C.; Strietska-Ilina, O.; Büdke, C. (2015). Appendix A contains each institution's definition. 5 See Ruppert Bulmer et al. (2022) and Stoveska (2017). 6 It should be noted that countries are endorsing to ILO’s definition of green jobs. For example, ASEAN countries and ILO have agreed on a series of recommendations to promote green jobs (ASEAN and ILO 2021), starting by having ASEAN Member States to work together to agree on common and workable definition of green jobs, using a 4 From here on, the term 'green' is used to indicate the proposed definition of jobs, and more broadly, to refer to output, technologies, tasks, or skills that contribute to reducing environmental impact. In the subsections that describe a specific measurement approach, 'green' will refer to the specific definition of that subsection. Accurately identifying green jobs is challenging since a green job involves two different dimensions: the firm and the worker. The firm and worker dimensions complement each other, and both contribute to making a job green. These dimensions sometimes, but not always, overlap. The firm-level dimension relates to the output (goods or services) that the job contributes to producing, and the technologies applied to do so. The firm’s environmental footprint naturally relates to the outputs produced. At one end, firms with a negative environmental footprint produce highly polluting or high carbon-intensive outputs (e.g., coal). At the other end, firms with a positive environmental footprint produce environmentally friendly outputs (here on green outputs or environmental goods and services, EGS) (e.g., solar panels). Green outputs include those that may have carbon-intensive production processes but still contribute to the greening of the economy (e.g., a firm manufacturing electric cars). In the middle, there might be firms that have a neutral environmental footprint, with zero or low carbon-intensive outputs (e.g., a leasing firm). In addition to the firm’s output, the job may be performed at a firm that intentionally utilizes technologies 7 to reduce its own environmental footprint. These are usually referred to as green or environmental technologies. The adoption of green technologies could be triggered by different objectives: increasing profits, compliance with regulations, social responsibility, accessing new market opportunities, or raising consumer awareness on climate change issues (e.g., a paperless office or a firm reducing hazardous waste in the environment). The worker-level dimension, instead, relates to the task-content of the job and the skill requirements to perform it well. Following Acemoglu and Autor (2011), “[…] a task is a unit of work activity that produces an output. In contrast, a skill is a worker’s endowment of capabilities for performing various tasks. Workers apply their skill endowments to tasks in exchange for wages, and skills applied to tasks produce output.” The technological feasibility and the economic cost ultimate determine the combination of capital and skills into tasks that will produce output. This dimension considers whether the job entails carrying out specific tasks that aim to lessen the environmental impact either by contributing to the production of green products and services, or by reducing the firm’s environmental footprint. It also examines the skills required and whether they involve specific knowledge or abilities related to the environment. Figure 2. Approaches to measure green jobs in terms of their relationship with the production function. spectrum approach to identify core green, indirectly green, and non-green occupations across different sectors and geographies. This paper proposes a narrower definition that comes close to what ILO and ASEAN refer as core green occupations. It is consistent with the range of activities that ILO considers as green jobs (see Figure 1 of ASEAN and ILO 2021). 7 Eurostat (2009) defines technologies as technical processes, installations and equipment, and methods or knowledge, the nature or purpose of which is environmentally friendly. 5 Notes: the visualization represents the production function as described in Autor (2013), where the output has a production −1 −1 1 function such as = �∫0 ( ) � , where () is the production level of task and is the elasticity of substitution across tasks. For simplicity, the Autor assumes there are three types of skill levels (but the production function could be generalized to n varieties of skills), each of which when combined with capital produce the task that will generate output. The task production function can be represented as ( ) = ( ) () + ()() + ()ℎ() + ()() where A is the factor- augmenting technology, ,, ( ) are the factor productivity schedules for skills low, medium and high, ( ) is the number of low skill worker units dedicated to produce task i, and () is the number of physical capital units dedicated to produce task i. Distinct from the canonical model, however, a factor-augmenting technological change need not increase the wages of all factors in this setup. Hence, the greening of the economy can influence the technology used, the physical capital, the skills, altering tasks and/or outputs. There are efforts to measure each of the mentioned dimensions. These efforts can be grouped into four approaches, according to the role they play in the production function, as visualized in Figure 2. 8 One of the reasons likely influencing the lack of agreement on a definition of green jobs is that each approach matters to a different extent depending on the research or policy question. The objective of this paper is to inform skills policy and hence, for this purpose, a green job is one which has specific tasks assigned to lessen consumers’ and/or firms’ environmental impact. These green tasks can either be assigned with the goal of producing greener outputs or of reducing the firm’s environmental footprint. They help to reduce and limit energy and raw materials consumption, greenhouse gas emissions, and waste and pollution, or they protect and restore ecosystems, or enable adaptation to climate change. A green job involves specific green tasks and can be at any firm regardless of its economic activity, including at firms with a negative environmental footprint (e.g., a water treatment specialist at a coal mine). And the level 8 See Autor (2013) for a full discussion of the task approach and the state-of-the-art representation of the production function. 6 of greenness of a job can vary depending on the importance of green tasks. Not all jobs in a firm producing a green output are green jobs involving green tasks. Some examples help to better understand the multiple dimensions and how they relate to classifying a job as green or non-green, depending on the question at hand. First, there are jobs that, given green technologies established at the firm, may require workers to perform certain green practices that are not essential to successfully producing the outputs for which the job was created. This is the case of an office worker contributing to a paperless office or a supermarket cashier offering paper bags instead of plastic bags. These types of green practices may require workers to have some environmental awareness or green-citizen skills. 9 However, such practices are not tasks since they are not essential to the production of the output for which the job was created. Thus, this paper 10 does not consider as green those jobs that involve only these types of practices; they do not involve green tasks. Second, there are jobs that are created due to the greening 11 of the economy, but whose task-content is not green. This is the case of a team assistant in a solar power factory. Third, and slightly similarly to the previous case, there are jobs that are performed at firms that have adopted green technologies, but the technology does not alter tasks. This could be the case of a driver of an electric car. These are usually referenced in the literature as indirect green jobs or green increased-demand occupations, but for this note’s definition, these are not green jobs. These cases are still relevant for other policy purposes though, as further explained in the next subsection. B. Measurement approaches and their suitability for skills policy This section describes the four approaches introduced above, looking into what they measure, the challenges of each of them during data collection or the application of the methodology, especially in developing countries, and how each of them contributes to informing skills policy, which is the reason for developing this paper. It also presents international examples of how they have been applied, with more details for the task approach since it is found to be the most suitable for informing skills policy. The output approach The output approach estimates the number of workers employed in firms producing environmentally friendly goods and services. These firms are likely to operate in the environmental goods and services sectors (EGSS). The OECD and Eurostat have laid out the foundation for data collection and analysis of the EGSS.12 Identifying green jobs or occupations through this approach could be quite straightforward: when the firm produces only green outputs in the EGSS, all employees are considered to be performing 9 What the ILO (2021) understands as ‘Basic skills necessary for adapting oneself to related environmental regulation and requirements to curb climate change.' 10 For this paper, from section 3 onwards, green jobs will be those with green tasks. 11 Greening of the economy means the production of greener outputs (products and services) and the use of greener technologies. 12 See OECD and Eurostat (1999), and Eurostat (2009). The OECD and Eurostat have been pioneers and have developed a handbook to determine environmentally friendly goods and services. The handbook has guided the efforts of other institutions and the development of survey instruments like the Green Goods and Services Survey of the BLS. 7 green jobs. The most reliable and common data source to estimate the number of jobs is specific firm surveys. 13 Despite its simplicity, the output approach comes with several important shortcomings. First, specialized data is needed to classify goods and services as green or environmentally friendly. There are several working definitions. Some researchers use estimates of CO2 emissions; 14 others, like the Bureau of Labor Statistics (BLS), collect specialized data for that purpose. 15 Once this is known, a second challenge arises as firms might produce both green and non-green goods and services. In these cases, if there is a rich firm survey with detailed input and output variables, assumptions can be made about the proportion of labor utilized for green outputs. For example, that proportion could be estimated by the percentage of working time or wages spent on the production of green outputs, or the share of revenues coming from green outputs. 16, 17 When output data is not available, then granular industry information about the main activity of the firm can be used to approximate it. This implies further assumptions to assign a category or a share of green outputs to each industry code. Applying the output approach in developing countries can pose additional challenges. Usually, developing countries do not have detailed data to classify locally produced goods and services as environmentally friendly or not. Measurement of CO2 emissions is not widespread 18 and conducting specific surveys mapping industry to occupation, such as the BLS GGS, might be unaffordable and not exempt from its own challenges. Consequently, most analysts appeal to relying on classifications of EGS and EGSS from other countries and estimate emissions based on the industry and input-output data. In the first case, the analyst must assume that the technology of production of goods and services is similar in both countries and will need a detailed mapping from EGS to the industrial classifications, both in the country used as reference and in the country of its application. Such mappings are not usually available in all countries, which means that further assumptions are needed regarding the shares of green and non-green outputs for each industry code, and across country-specific industrial classifications. 19 The output approach is useful for understanding the implications of structural transformation on sectoral employment, as well as the linkages between green industrial policy, 20 green stimulus packages, and employment growth. For example, Bontadini and Vona (2021) are interested in understanding the comparative advantage of green production as a way to reconcile economic growth with environmental preservation and the mitigation of climate change-related risks. The authors find that green production is highly concentrated in a set of high-tech industries producing capital goods, while polluting and green 13 Eurostat (2009). 14 For example, the UK uses multiple data sources including satellite images. For more information consult the UK Office for National Statistics at https://www.ons.gov.uk/economy/environmentalaccounts/bulletins/ukenvironmentalaccounts/2023#measuring- the-data. 15 BLS (2013a). 16 Eurostat data collection handbook (2009) proposes various ways to adjust the estimate. 17 Winkler et al. (2022) forthcoming. 18 Countries report their emissions through what is known as a ‘bottom up’ approach, where national emissions are estimated by combining data on types of activity with the emissions typically produced by those activities. 19 The implications of these assumptions are further explored for the task approach. 20 Including policies related to trade on green outputs and services. 8 production occur in two separate sets of industries that are only related through intra-industry linkages such as the purchase of capital assets. These findings have implications for taxing polluting agencies and subsidizing green outputs. However, the output approach is less useful for informing skills policy. On the one hand, it leaves out jobs in firms that produce non-green outputs but implement green technologies. For example, a leather manufacturing firm may implement technologies to clean wastewater and may employ workers in charge of this process. On the other hand, it includes all jobs regardless of whether they involve green tasks. For example, it would include the receptionist at a firm producing a filter to purify air. This issue cannot be solved by just restricting the count to production workers since not every worker involved in the production of a green output will perform green tasks or require green skills. For example, while an engineer designing an electric car may require energy efficiency knowledge, most assemblers of such cars may perform the same tasks as assemblers of gas-powered cars. Moreover, to inform skills policy it is crucial to profile occupations, which reveals an additional limitation of the output approach for this policy objective. Most notably, surveys that are designed to measure green jobs based on their output (or the data that aligns well with this approach) do not collect or link occupational data. Occupational data is necessary to understand the task content and skills requirements of such jobs. 21 In the best-case scenario, analysts would need to work with multiple surveys or databases to adequately inform skills policy. The U.S., Canada, and the E.U. have good applications of the output approach. In 2011, the U.S. Bureau of Labor Statistics conducted the Green Goods and Services (GGS) survey.22 The GGS survey was a firm- level survey on a sample of 120,000 establishments in industries identified as potentially producing green outputs—333 industries out of 1,200 detailed at the six-digit American Industrial Classification System (NAICS). The selected industries represent 23 percent of all establishments and 20 percent of employment in the U.S. economy. Specifically, the GGS survey asks establishments for the percentage of revenues (or, for firms that do not generate revenue, the percentage of employment) deriving from green outputs according to the BLS definition. This self-reported percentage was then decoded into an establishment’s green percentage. Most sampled establishments ended up not having revenues or employment associated with a green output (confirming limitations of relying solely on standard classifications), only around 2 percent of the U.S. workforce was estimated to do so. 23 Lastly, replicating a survey like the GGS in developing countries would represent a costly endeavor, not only in terms of sampling a substantial number of firms to have statistical power but also in explaining sophisticated questions to respondents. The BLS could overcome the critical limitation mentioned above regarding occupational information by linking the GGS survey to the Occupational Employment Survey (OES), a well-established survey designed to produce employment and wage estimates for nearly 800 narrowly defined occupations 21 To the best of our knowledge, the U.S. is only country that collects data on green outputs and links it to occupational data. 22 BLS (2013a), BLS (2013b), and BLS (2012a). 23 BLS (2013a). 9 using Standard Occupational Classifications (SOCs). For the creation of GGS-OCC estimates, the same sampling frame was used, and OES staffing patterns were matched to the firm’s green percentage calculated with the GGS. The results of the GGS-OCC survey showed that 30.6 percent of 2011 employment was in establishments producing a green output and that occupations whose duties are expected to be directly linked to green activities are also employed in non-green establishments (e.g., environmental engineers, environmental scientists and specialists). The process approach The process approach estimates the number of workers involved in technologies used to reduce the environmental footprint of the firms in which they work, regardless of the outputs the firm produces. This approach is useful when analysts’ main objective is to understand firms’ labor needs to implement new technologies, but without a tight link to occupational and skills data this approach falls short of generating enough information to adequately design skills policy solutions. The Bureau of Labor Statistics (BLS) of the U.S. developed this approach and designed a survey to apply it. The Green Technologies and Practices (GTP) survey is a firm-level survey that captures whether firms use certain environmentally friendly technologies and practices (or GTPs) that can reduce environmental impact.24 The survey also collects information on whether workers are involved in these GTPs by researching, developing, maintaining, using, or installing GTPs, or training other employees on them. In cases where workers spend more than half of their time involved in the GTPs, the survey also collects the occupational job description, which is then used to code jobs into the 4-digit Standard Occupational Classification (SOC), along with wages for such jobs. The following are the GTPs surveyed: 1. Generate electricity, heat, or fuel from renewable sources primarily for use within the establishment 2. Improve energy efficiency within the establishment 3. Reduce greenhouse gas emissions (other than by 1 and 2) 4. Either reduce the creation or release of pollutants or toxic compounds produced, or to remove pollutants or hazardous waste from the environment 5. Reduce or eliminate the creation of waste materials 6. Conserve natural resources In 2011, the BLS carried out the GTP survey. It sampled 35,000 firms stratified by region and 2-digit North American Industry Classification System (NAICS) sector. The survey results showed that 75 percent of firms used at least one GTP. Not surprisingly, the top GTP was improving energy efficiency, for which any firm using at least one Energy Star certified appliance would qualify, followed by reducing the creation of waste materials, for which any firm recycling waste—a widespread practice and mandatory in some cities in the U.S. The GTP is not exempt from the measurement challenges. First, while the survey provides examples of each GTP and includes the most salient types, it does not include an exhaustive list of GTPs, leaving room for respondents’ subjective interpretation on whether they use or not a GTP. Second, the respondents of the survey are not the workers themselves, but employers or CEOs, which might lead to biased answers about the importance and applicability of GTPs by workers. 24 BLS (2011) and BLS (2012b). 10 Although the process approach combines aspects of both the firm and worker levels, it has limitations. Firstly, this approach only includes green jobs that contribute to reducing the firm's environmental footprint, while overlooking other green jobs that generate green outputs to reduce consumers' footprint (see Appendix A for a description of pilot projects that combine both the output and process approaches). Even if the objective is to count green jobs, this limitation cannot be easily resolved by simply summing up jobs obtained from both the output and process approaches, as there will likely be significant overlaps that result in double counting issues. 25 Secondly, while the firm may be reducing its environmental impact by using the listed GTPs, it is not necessarily the case that the worker is the one making the difference by using these GTPs. In other words, even if workers use certain GTPs, their tasks may not necessarily reduce the firm’s environmental impact. This approach does not perfectly fit the objective of identifying jobs that contribute to lessening the impact on the environment and may need skills training. For example, a food manufacturing company using Energy Star certified refrigerators (an example under the GTP improving energy efficiency) will surely reduce energy consumption, but workers using the refrigerator or even installing it may do exactly the same tasks as a similar company using regular refrigerators. According to this measure, the BLS found that Heating, air conditioning, and refrigeration mechanics and installers were the second occupation with the highest GTP employment. 26 For a more recent example affecting a significant part of the global employment, due to the COVID-19 pandemic, many firms have implemented telework programs (an example under GTP reducing GHG emissions). Telework programs for sure reduce firms’ environmental footprint but part of it is transferred to workers, and workers may be just doing the same sort of tasks as they were doing at the firm location. Besides this issue, it is not clear in the survey how workers should be counted when involved in GTPs (should all teleworkers or e-car drivers be counted as green workers?) or how to capture the intensity with which a worker may be performing green and non- green tasks. Applying the BLS process approach to developing countries also brings its own challenges. Besides the challenges already noted, this approach falls short of capturing the informal sector and the agriculture sector, both usually large in the developing world. Moreover, it should be added that the cost associated with these surveys is high, especially as they may require large sample sizes. While a large percentage of firms have adopted GTPs, very few have workers who dedicate time to them. For example, in the U.S., the BLS GTP survey found that 75 percent of firms adopted a GTP and that 0.7 percent of workers use these more than half of their time. 27 Hence, for the occupational module to be relevant, the survey needs to have a large enough sample size to generate statistical power. As one could imagine, these proportions are even lower in developing countries, exacerbating the sample size requirements. 28 For example, we added a few questions similar to the BLS GTP survey in the Environmentally Friendly Industry Study (SIRL) survey conducted in 2021 in Indonesia. It was found that about half of firms have 25 Stoevska & Hunter (2012) also raises this issue. 26 BLS (2012b). 27 BLS (2012b). 28 Of course, the statistical power also depends on the standard deviation of the variables of interest, which even if small they contribute little to reduce the sample size. 11 at least one energy efficiency measure or technology 29 and that 60 percent of firms have dedicated energy teams or personnel. 30 However, only 0.7 percent of workers were involved in GTPs. Given the low percentage of workers directly involved in carrying out GPTs, it would require a very large survey to retrieve occupational information with statistical power, that can later on be used to link it to skills policy. The task-content approach The task-content approach is the most suitable one to inform skills policy. The approach looks into the nature of occupations, their task content, and can be later easily linked to the skills required to perform the job well if that information is available in the country. It allows understanding whether a job has specific tasks assigned to lessen consumers’ and/or firms’ environmental impact. To our knowledge, there are two methods currently applied to classify tasks as green and non-green. The most prominent application of the task-content approach is the O*NET Green Economy Program (O*NET GEP). 31 The Occupational Information Network (O*NET) is the primary source of occupational information in the United States, which is worldwide recognized and used by researchers. The objective of the O*NET GEP was to identify the occupations that are to be impacted by the greening of the economy (i.e., by green economic activities and technologies). 32 The program comprised a desk review of about 60 publications related to the greening of the economy, which was then used by O*NET experts to produce the occupational and tasks classification. 33 The desk review covered 12 major green economic sectors tightly linked to the greening of the economy and selected the occupations commonly employed in these sectors, and hence likely to be subject to change (see appendix B for a list of the 10 steps followed by O*NET to classify occupations and tasks).34 Using the same desk review, it then categorized the resulting occupations into the following groups: 29 The surveyed measures included Heating and cooling the environment, Increased use of climate-friendly energy plants, Upgrade machinery and equipment, Energy management; Waste reduction, recycling, and waste management; Control of air pollution; Water management; Vehicle upgrades; Improved lighting system; Conservation of natural resources (excluding the use of recycled inputs); and other pollution control efforts. The technologies surveyed included: Equipment that has an Energy Star rating; LEED certified buildings; Energy- efficient lighting; Programmable thermostat; Cogeneration (combined heat and power); and Energy-efficient manufacturing equipment. 30 Results from the SILR survey prepared for the Indonesia CCDR. 31 This paper was prepared in 2020 and 2021, with decision review meeting in June 2022. In that month, O*Net released the paper proposing a linguistic approach for green topics (Lewis et al, 2022), which we understand is still in piloting stages, as it is the methodology proposed here. 32 Dierdorff et al. (2009) and Dierdorff et al. (2011). 33 The publications included those “in educational institutions, in academic journals, commissioned reports, industry white papers, and governmental technical reports. Additionally, numerous associated/relevant internet sources on the world of work were reviewed.” Dierdorff et al. (2009). A list of the publications consulted can be found in O*NET (2013). 34 The 12 green economic sectors include: Renewable energy generation, Transportation, Energy efficiency, Green construction, Energy trading, Energy and carbon capture, Research, design, and consulting services (indirect jobs), Environment protection, Agriculture and forestry, Manufacturing (of green technologies and energy efficient manufacturing processes), Recycling and waste reduction, Governmental and regulatory administration (Dierdorff et al., 2009). 12 - Green increased demand occupations are those that will be in demand due to the greening of the economy. Although the work context may be green or become greener, these occupations do not perform green tasks, and their work and work requirements are not expected to change. This group accounts for 64 occupations classified in the O*NET-SOC taxonomy. 35 For example, electrical power line installers in energy efficient and infrastructure upgrades. However, these jobs are not considered green jobs according to the proposed definition in this note. Some researchers refer to this category as indirect green jobs. - Green enhanced skills occupations are those that the greening of the economy will change their work and work requirements –tasks, skills, knowledge, credentials, etc. This group accounts for 62 occupations classified in the O*NET-SOC taxonomy. For example, architects who may be required to have LEED (Leadership in Energy and Environmental Design) certification. - Green new and emerging occupations are those that were not in the O*NET-SOC taxonomy but that O*NET created an O*NET-SOC for them given that they are appearing due to the greening of the economy. Occupations created have distinct work and work requirements and needed to show evidence of being relevant—having at least 5,000 workers, showing evidence of projected growth, having existing accredited education or training programs offering tailored credentials for the occupation, showing evidence of at least one national association serving workers in the occupation, and showing evidence of trade or professional journals for workers in the occupation. In total, the program created 78 new O*NET-SOC occupations. For example, solar system technicians. In addition, the O*NET GEP also classified tasks. O*NET developed new green task statements for both the green enhanced skills occupations and the green new and emerging occupations. 36 The green enhanced skills occupations have 870 tasks in the database, of which 113 were identified as green. In addition, 196 tasks were generated for these occupations. The green new and emerging occupations, on the other hand, had no task in the database since all of these were new occupations to the O*NET-SOC taxonomy. In total, 626 task statements were generated for this group. Two major advantages stand out from the O*NET GEP green classification. First and above all, O*NET is a highly reputable agency whose data is trusted and used by academics all over the world, as well as government agencies, training institutions, and the private sector in the U.S. Second, the O*NET green data constitutes an addition to the existing occupational database that contains detailed descriptions of almost 1,000 occupations in the O*NET-SOC taxonomy, including the importance and level of their tasks, skills, abilities, knowledge, and other work requirements. The O*NET task approach, as well as the O*NET green occupation classification, have been prominently used by academics to estimate the consequences of the greening of the U.S. economy. 37 Despite its advantages, the green occupational database of O*NET has methodological limitations. First, O*NET GEP limited the analysis to occupations in 12 green sectors, excluding green jobs in the rest of the economy that contribute to the sustainability of the environment. Second, the methodology relies 35 The O*NET-SOC is based on the U.S. SOC, by assigning the 6-digit SOC code plus an additional level. 36 O*NET (2010). 37 Consoli et al. (2016); Bowen et al. (2018); Vona et al. (2018); Vona et al. (2019); Rutzer, Niggli, and Weder (2020). 13 on a review of 60 publications of very diverse nature, dated back to 2009 and revised later in 2011. Although such qualitative work is important and should be an integral part of any data driven strategy, on its own it cannot be considered comprehensive or objective, and it is not easy to update. Progress in the adoption of green technologies during the last decade has been phenomenal, and this progress is being left out of the GEP. The use of O*NET GEP outside the U.S. and in developing countries is also subject to important assumptions that cast doubts about its validity in these contexts. When O*NET data is used to study countries outside the U.S. the following assumptions are inevitably made: first, the production function (technology, capital, and labor—tasks and skills) is the same as that of the U.S. Internationally comparable firm data, like that from the World Bank Enterprise Surveys, show that the input mix and the returns of firms vary across countries. Country efforts to replicate O*NET in Indonesia and Uruguay show that there is variation in the task content and skills requirement of occupations across each of these countries and the U.S. 38 A second major problem is that most countries when collecting occupational data use the International Standard Classification of Occupations (ISCO) at the 4-digit level of aggregation or higher. This entails relying on a crosswalk to transform the 8-digit O*NET SOC to the 4- digit ISCO and making assumptions about the structure of employment within each 4-digit occupation. Although crosswalks are available 39 and commonly used by the research community, they have been found particularly problematic for capturing green jobs given the low levels of employment and the high measurement errors produced by the crosswalks. 40 For example, Elliot et al. (2021) applies the O*NET classification and Vona et al. (2019) methodology to study how eco-innovations impact employment in the Netherlands, finding that there is no impact on overall employment but that they increase green jobs. However, the share of green jobs is significantly higher than that in the U.S., bringing into question the quality of the crosswalk to measure green employment. 41 Section four of this paper reports the same issue when applying O*NET classification to estimate green employment in Indonesia. A second modality of the task content approach consists of applying text analysis to occupational manuals with task descriptions. Janser (2019) applied text mining to the German occupational database to calculate the level of "greenness" of occupations. The dictionary included 153 green terms and was applied to a database with occupational requirements at the 8-digit level, containing 14,546 words, to identify green tasks for each occupation. Following Consoli et al. (2016), Janser created a greenness index, which is the share of green requirements (tasks) in the occupation. Janser shows that there was a greening of jobs between 2011 and 2016 in Germany, with slight wage increases, and that the greenness of occupations and employment growth are positively correlated. This second modality also has some methodological limitations. First, the country of interest needs to have an occupational manual that includes task statements as part of the occupational description. The more narrowly defined the occupations are and the more precisely defined the task statements are, the 38 Alatas, Granata and Posadas (2020), Apella & Rovner (2021) for Indonesia and Uruguay respectively. Granata and Moroz (2022) describe Viet Nam’s efforts to replicate and adjust select modules of O*NET to Viet Nam, but do not attempt to compare it with the U.S. 39 https://www.bls.gov/soc/isco_soc_crosswalk. 40 Vona, F. (2021). 41 Vona, F. (2021). 14 richer and more reliable the analysis will be. Second, the occupational manuals need to be regularly revised if the green index is to be updated. While Germany has a rich occupational manual developed at the 6-digit occupational classification, most developing countries simply adhere to ISCO (and at best, at the 4-digit occupational classification), missing the opportunity to profit from the richness of text analysis. The skills requirements approach One of the main limitations of the task approach, for both O*NET and text analysis methodologies, is that updating the green classification over time is not easy and time-consuming. The O*NET methodology would require regularly reviewing the literature to evaluate if there are new green tasks and/or new technologies. The text analysis methodology, however, is also subject to this problem since occupational manuals with task statements are usually updated every 10 years, a period that seems too long for the current pace of technological change. Countries with modern labor market information systems, usually with a more disaggregated occupational classification (6-digit ISCO or equivalent), may overcome this limitation by directly working with online vacancy data to identify green occupations. The skills approach uses green skills required in job postings as a signal of the greening of the economy. LinkedIn, for example, is relying on this approach. Consistent with this note, LinkedIn defines green skills as those that enable the environmental sustainability of economic activities. 42 However, their definition is broader, as these skills are grouped into core green skills (such as recycling) which are most directly related to these sustainability-promoting activities; ambivalent green skills (such as fleet management) which may or may not be used for sustainability, and adjacent green skills (such as biology) which can support the acquisition of core and ambivalent green skills. To our knowledge, their methodology for coding the 38,000 unique skills terms into green and non-green is not publicly available. 43 For LinkedIn, green jobs are occupations that cannot be performed without extensive knowledge of green skills. Instead, Lightcast (formerly Burning Glass, BG) uses a combination of data sources, including O*NET classification, job titles, and job text searches. 44 Like the others, this approach presents advantages and disadvantages for its implementation in advanced and developing countries. The most salient advantage is the granularity and frequency of the data. However, it is difficult to infer from vacancy data if the job to be filled will require the skills to carry out tasks, or if green skills are simply desired or used as a signal for some other job need. Developing countries face the additional challenge that the use of online matching platforms is not yet widespread, and hence the data is usually biased towards large metropolitan areas and the high-skilled segment of the market. 45 The next section illustrates these challenges applied to the Indonesian and U.S. cases in more detail. 42 LinkedIn 2022 Green economy report 2022. https://economicgraph.linkedin.com/content/dam/me/economicgraph/en-us/global-green-skills-report/global- green-skills-report-pdf/li-green-economy-report-2022-annex.pdf. 43 This team has unsuccessfully attempted to connect with LinkedIn to learn about the methodology. 44 Based on our communication with Lightcast. 45 See Granata, Posadas, and Testaverde (2021) for a discussion of biases applied to the Indonesian labor market. 15 It should be noted that even if skills data from job vacancy postings is not best suited to identify green jobs, it brings valuable information to understand and design skills policy. Skills analysis from big data allows profiling of green jobs, including their skills requirements, and understanding other key features such as skills demand variation across industries and regions. 3. Using text analysis to develop an alternative methodology for the task-content approach This section introduces the text analysis methodology that identifies green occupations based on a green dictionary developed by the authors. The methodology falls under the task-content approach, which is our preferred approach for skills policy. It applies the green dictionary to a task database that is used for data collection in many countries, called the international standard classification of occupations (ISCO-08). One advantage of the methodology is that the green dictionary can also be easily applied to any other task database or adjusted to other country settings. This section also discusses the extent to which the methodology can be trusted by conducting two reliability tests. The first test compares the outcome of applying the methodology to the O*NET task database against green occupations listed in the O*NET Green Economy Program. The second test assumes that green skills are necessary to perform green tasks and checks the correlation between a green task index and a green skills intensity measure, both constructed by applying the same methodology to different databases. A. Applying text analysis to the International Standard Classification of Occupations (ISCO) to produce the green task intensity index Classifying occupations according to their level of “greenness” consists of three steps: (1) creating a green dictionary, (2) applying it through text analysis techniques to an occupational database containing task statements, and (3) calculating the GTI index. Step 1: Development of the green dictionary The first step was the development of a green dictionary, a database of terms that were used as the basis for the text analysis. The terms in the dictionary are not tasks, but rather words, word roots, and expressions directly linked to environmentally friendly concepts. Our assumption was that a task statement containing any of these key terms would be related to environmentally friendly tasks. The selection of terms was the result of a careful desk review of diverse sources. Naturally, since terms are short and general, most of them are present in multiple sources. Thus, the creation of the dictionary should be understood as a systematic iterative process of adding and confirming the appropriateness of terms as they consistently came up in the consulted sources. As we continued reviewing sources, fewer new terms were included for consideration in the dictionary. For the creation of the dictionary, the starting point was to examine the terms commonly mentioned in the diverse material from the methodological approaches described in section 2, including the O*NET green project description, and from the labor environmental economics literature. First, the dictionary included terms that appeared in the BLS GTP survey questionnaire, most common terms from machine- 16 based text analysis of all O*NET green task statements, terms within the skills/competences from the ESCO taxonomy, 46 and terms in Eurostat manuals for data collection and analysis (1999 and 2009). Second, it also included commonly mentioned terms that frequently appeared in more than 72 papers and reports in the labor environmental economics literature. The green dictionary was then compared with dictionaries shared by Lightcast (formerly Burning Glass Technologies, BG), a list of 81 job titles within O*NET, and a list of 154 words or phrases from Janser (2019). These dictionaries were informative in the sense that they confirmed the appropriateness of many terms already selected. It should be noted that these dictionaries also included terms that may or may not be environmentally friendly depending on the context applied (e.g., energy engineers, thermograph, battery technology) which received different treatment as explained below or were not included. The desk review resulted in the evaluation of 451 terms, organized into 14 main green categories according to topic areas— categories that frequently appeared in the consulted sources (e.g., GTP survey questionnaire). 47 Appendix C contains a full list of all the consulted sources and table 3.1 provides some examples of terms present in the cited sources that were considered for inclusion. Table 3.1. Examples of terms under consideration Consulted sources Examples of terms present in cited sources BLS GTP survey Renewable source, renewable energy, solar energy, wind energy, landfill gas, energy efficient, carbon offset, composting, remanufacturing, greenhouse gas, LEED Most common environmental words in Environmental, energy, solar, green, waste, wind, carbon, O*NET green tasks biofuels, recycling, hazardous, repair, landfill ESCO taxonomy S3.3.2 complying with environmental impact, erosion control, sea pollution, protect environmental protection laws and standards trees, protect wilderness areas, conserve natural resources Eurostat manuals for data collection and Air pollution control, solid waste management, noise control, analysis wastewater management, soil remediation, hazardous waste collection, waste recovery and recycling, resource-efficient, energy saving, eco-tourism, renewable energy Commonly mentioned terms the in Climate change, global warming, environmental, sustainability, environmental economics literature recycling, renewable energy Emsi-Burning Glass Technologies dictionary Clean energy, renewable energy, wind energy engineers, recycling coordinator, climate change policy analyst, conservation scientist Janser (2019) dictionary Lithium ions, smart grid technology, passive house, tree care, bicycles, lightweight construction Once the first version of the dictionary was completed, it underwent several robustness checks. First, it was manually examined through more than 50 rounds of random tests to ensure that the dictionary correctly identified the task statements in the occupational database. As a result of this process, 104 46 Those within S3.3.2 complying with environmental protection laws and standards. 47 Topic areas were include Agriculture, forestry, and fish production, Clean energy, Climate change common terms, Energy efficiency, Environmental certifications, Environmental knowledge, Environmental regulations and compliance, Environmental software, Greenhouse gas reduction and pollution reduction & removal, Low-carbon mobility, Low-polluting construction, Natural resource conservation, Recycling and reuse of waste and materials, and Repair. 17 terms were removed from the dictionary as they were found to be capturing unrelated concepts. Additionally, 172 exemptions were established for the remaining terms to exclude words that contain the term but are not related to the topic. For instance, the terms 'hazardous material' and 'waste' were removed, while the term 'ecolog' has an exemption 'gynecolog', and the term 'environmental' has an exemption 'environmental sanitation'. Second, the dictionary underwent a sensibility analysis, which consisted of classifying terms into two categories: green and green potential. While some terms strictly captured environmentally friendly tasks (green terms), others captured tasks that may or may not be environmentally friendly depending on the level of greening of the economy in a specific country's context. These latter terms are referred to as green potential terms since, even though the tasks may not be green at present, they have the potential to become green as countries become greener. For example, many terms related to agricultural activities, 'energy engineer,' or 'motor buses.' Green potential terms also include those that capture tasks whose outputs may be less polluting or more resource-efficient than the equivalent average output. 48 For example, 'repair' captures the task of repair activities, which may consume fewer resources or prevent the generation of waste compared to producing a new one. However, it should be noted that this term should be used with caution when repairing old goods leads to consuming more energy than their newer alternatives. The distinction between green and green potential terms provides a narrow and broad definition of greenness, or a lower bound of greenness—what we know for sure is green—and an upper bound—what may be considered green or could be green if greener technologies are adopted. The dictionary contains a total of 308 terms categorized as green and 39 terms were categorized as green potential. Table 3.2 provides a summary of term selection and selected examples. For transparency and replicability purposes, the toolkit with the final green dictionary and do file is available to readers. 49 It contains the exact list of terms with their exemption rules and their categorization into green and green potential. It also includes the 104 removed terms marked as ‘remove.’ Users of the toolkit can adjust the dictionary to the country context and as technology and green jobs evolve. Table 3.2. Green dictionary terms (a) Selection of green terms (b) Select examples of terms in the green dictionary 451 Considered terms Green terms Green Potential terms -104 Not related carbon reduction energy engineer terms climate change passenger train 347 Green and green  Broad dictionary deforestation refuse collection potential terms emission reduction repair - 39 Green potential clean technology Fertiliz terms energy efficien timber construction 308 Green terms  Narrow dictionary 48 This concept is similar to the definition of ‘adapted goods’ in Eurostat (2009). 49 For World Bank staff, it is saved in the Operations Workplace Portal of this task, and colleagues from outside the institution can request it from the authors via email. 18 Source: author's green dictionary Step 2: Application to the ISCO-08 task database The green dictionary is then applied to a task database, created from the occupational manual of the International Standard Classification of Occupations 2008 (ISCO-08). ISCO is a hierarchically structured system that classifies and aggregates occupational information for all jobs in the world into 433 4-digit level occupations (referred to as occupations here on) to be used for the collection of occupational data in statistical census, surveys, and administrative databases (See appendix D for more details on ISCO). 50 For each occupation in ISCO-08 there is an occupational title, a job description that delimits the scope of the occupation, a list of up to 14 tasks performed, 51 and examples of job titles included within the occupation. 52 Hence, the created occupational database contains 433 occupations; 3,245 task statements; and 46,184 words (equivalent to 8,010 unique words). Since ISCO-08 has been widely adopted in several countries, the results of the analysis described here can be applied to any country that uses ISCO-08. The green dictionary captured 78 unique terms and 329 tasks from the ISCO-08 occupational database. Out of the 347 unique terms in the green dictionary, 78 terms were also present in the occupational database (figure 3.1 panel a). 53 Many of the captured terms were related to greenhouse gas reduction, pollution reduction and removal, and natural resource conservation. Out of the 3,245 task statements, the green dictionary captured 329 green tasks (or 10 percent of the total tasks) (figure 3.1 panel b – examples in table 3.3), of which 83 are strictly green. 54 These tasks are related to natural resource conservation (16%), climate change (14%), recycling and reuse of waste and materials (13%), and 36 percent fall into multiple green categories. The sensitivity analysis can also be thought of as providing an upper bound. While the broad dictionary only contains 39 additional terms (or 11 percent additional terms), these terms serve to capture 246 additional tasks (or 7.6 percent of the total number of tasks). These additional tasks will be referred to as having green potential, hereafter. The vast majority of green potential tasks relate to either agriculture, forestry, and fish production (50%) or repair activities (44%). Random quality checks confirmed that the matches were appropriate for the identification of green tasks. 50 ISCO-08 has a total of 436 4-digit occupations, but 3 are Armed forces occupations, which do not include a list of tasks performed. 51 Min=1, p25=6, p50=7, p75=9, max=14. 52 The classification can be downloaded online from the ESCO European Commission website: https://esco.ec.europa.eu/en/use-esco/download. 53 Of the 347 unique terms, 308 are strictly green and 39 are green potential. Of the 58 captured terms, 56 are strictly green and 22 are green potential. 54 Strictly green refers to terms that match the narrow dictionary. 19 Figure 3.1 Green dictionary capturing green tasks a. Green dictionary, total terms b. Total green and green potential task statements captured Total terms Green task statements 0 25 50 75 100 125 0 10 20 30 40 50 60 70 80 Agriculture, forestry, and fish production Agriculture, forestry, and fish… Clean Energy Clean energy Climate change Climate change terms Energy efficiency Energy efficiency Environmental Certifications Green Environmental Knowledge GHG reduction and pollution… Green Environmental Regulations and… Low-carbon mobility GHG reduction, Pollution reduction &… Natural Resource Conservation Low-carbon mobility Recycling and reuse of waste and… Low-polluting construction Natural Resource Conservation Multiple Recycling and reuse of waste and… Agriculture, forestry, and fish… Agriculture, forestry, and fish production Energy efficiency Green potential Energy efficiency Low-carbon mobility Green potential Environmental Knowledge Low-carbon mobility Natural Resource Conservation Low-polluting construction Recycling and reuse of waste and… Natural Resource Conservation Repair Recycling and reuse of waste and… Multiple Repair Terms present in the occupational database Terms not present in the occupational database Source: authors Table 3.3 Select examples of green and green potential task captured by the green dictionary Task type Task statement Green providing environmental engineering assistance in network analysis, regulatory analysis, tasks and planning or reviewing database development; assisting in the development of environmental pollution remediation devices under the direction of an engineer; analyzing workforce utilization, facility layout, operational data and production schedules and costs to determine optimum worker and equipment efficiencies; Green advising on techniques for improving the production of crops, livestock and fish, and potential alternative production options; task fitting, adjusting and repairing electrical parts in domestic appliances, industrial machines and other appliances; driving and tending street tramcars transporting passengers; Source: authors 20 Step 3: Estimation of the green task intensity index The Green Task Intensity (GTI) index measures the green task-content of occupations. It is the proportion of green tasks in an occupation o as described in the formula below. Similar to Vona et al. (2018) measure of the greenness of the occupation, it captures the extensive and intensive margin. # = # While the GTI is a useful measure of greenness, it is not exempt from drawbacks. First, it assumes that each task is performed with the same frequency and has the same importance. In the U.S., Vona et al. (2019) compute weights according to the importance score attributed to each occupation-specific task based on the richness of the task module of O*NET data. Second, it should be noted that the GTI may underestimate the level of greenness of an occupation since the task statements listed under each occupation are not necessarily a fully exhaustive list, but rather a good description of what the occupation involves for the majority of jobs. Third, emerging occupations—many of which are associated with the greening of the economy—may now be under the residual occupation code ('not elsewhere classified occupations') and may not be captured since the task description of these residual occupations tends to be broad. Next, we compute two indexes, one for each dictionary, and label them accordingly: GTI narrow and GTI broad (see appendix E for GTI at 4-digit occupational level). The extensive margin simply measures whether occupations have at least one green task or not. The GTI narrow classifies 36 out of 433 occupations as green (or 8% of all occupations) (figure 3.2 panel a). The intensive margin measures the level of greenness, with some occupations with close to 90 percent of their tasks being green as plotted in figure 3.2 panel b). Green occupations are found across all skill levels and independently of the main activity of the firm; however, larger percentages are found among high skilled occupations. Forty-one percent of the occupations are either professionals or technicians and associate professionals, but almost a quarter are craft and related trade workers, and 14 percent are skilled agricultural, forestry, and fishery workers (figure 3.3). As mentioned above, the sensitivity analysis carried out with the broad dictionary led to classifying more green tasks and in turned 91 additional occupations as having green potential (or 21 percent of all 4- digit occupations). This leaves 306 occupations without any green task regardless of the tolerance of greenness applied. The broad dictionary captures relatively more low-skilled occupations, in the craft and related trade workers, and plant and machine operators and assemblers. Very few occupations within the managerial, service and sales representative, and clerical support workers major occupational groups have green task content. Not surprisingly, the intensive margin changes less when using the broad dictionary. As seen in the green shaded area of figure 3.2 panel (b), the vast majority of the occupations (31 out of the 36 green occupations) that were considered green, only have minor increases in the greenness level. 55 In sum, the expansion of the dictionary acts mostly on the extensive margin rather than on the intensive margin. 55 The five occupations with notable changes are Environmental engineers, Environmental protection professionals, Refuse sorters, Meteorologists, and Garbage and recycling collectors. 21 Figure 3.2 Occupational GTI a. Occupational GTI (narrow) and ranking b. Comparison of GTI between Narrow and Broad Ranking GTI (narrow) 100 90 80 GTI (Narrow) 70 60 50 40 30 20 10 Different 0 Similar 76 51 26 1 426 401 376 351 326 301 276 251 226 201 176 151 126 101 Rank Zero GTI High GTI No Narrow GT, has Broad GT Source: authors Figure 3.3 Share of occupations, all by GTI and green occupations by major occupational group 8 All GTI 14 Zero GTI 23 29 16 GTI Managers 18 Professionals 13 Technicians and associate professionals 92 14 Clerical support workers 71 32 Service and sales workers 24 Skilled agricultural, forestry, and fishery 6 12 workers Craft and related trades workers 8 8 Plant and machine operators, and All occupations GTI occupations All occupations GTI occupations assemblers Elementary occupations Narrow Broad Source: authors B. How much can the proposed methodology be trusted? A natural question is whether the proposed methodology is reliable for identifying green jobs and whether it is preferred over other methodologies. To answer this, two reliability tests are performed. First, the cleanest exercise is to compare the results against another tasks approach methodology in one country. Since the only alternative methodology is the well-respected O*NET GEP run for the U.S., we 22 first examine how well the text analysis works against this gold standard for U.S. occupations. We do this using the narrow dictionary, which only contains strictly green terms. Second, as an alternative exercise, we test the reliability of the identification strategy by applying the same dictionary to a skill database and observing how the GTI correlates with green skills intensity. Reliability test #1: The proposed text analysis methodology applied to O*NET task taxonomy versus O*NET GEP To carry out this test, we apply our dictionary and text analysis techniques to the O*NET task statements database (version 23.3, which uses O*NET SOC 2010 and contains green occupations and task statements). In total there are 19,636 unique task statements for 974 occupations. Our green narrow dictionary successfully captures 77 percent of O*NET GEP green tasks. At the same time, our green dictionary identifies 293 green tasks that the O*NET GEP missed. The differences may be due to the GEP focus being circumscribed to certain sectors and qualitative work. Table 3.4 shows the detailed comparison of the results of applying both methodologies. A thorough analysis of the task statements confirms the validity of our methodology. Tasks that are green to O*NET but are not green according to the text analysis are: ’coordinate with other marketing team members and workers such as graphic artists to develop and implement marketing programs’, or ’prepare, maintain, or revise quality assurance documentation or procedures,’ or to ‘monitor the flow of energy in response to changes in consumer demand‘, or ‘fill out defective equipment reports.‘ Our dictionary also detects 300 green tasks within occupations that are non-green to O*NET. Examples are the tasks ’determine or recommend radioactive decontamination procedures, according to the size and nature of equipment and the degree of contamination,’ or ’conduct audits at hazardous waste sites or industrial sites or participate in hazardous waste site investigations.’ However, a few tasks that are green to O*NET GEP are misclassified and should even be considered brown. For example, Power plant operators, a green enhanced occupation, has as tasks to ‘operate, control, or monitor gasifiers or related equipment, such as coolers, water quenches, water gas shifts reactors, or sulfur recovery units, to produce syngas or electricity from coal.’ The quality check performed found very few tasks that may have been classified as green, such as ‘conduct well field site assessment’ or ‘deposit recoverable materials into chutes or place materials on conveyor belts’. Potentially, these could be fixed by adding terms to the dictionary. Table 3.4 O*NET GEP tasks vs Text analysis methodology Text analysis based on narrow green dictionary O*NET GEP task classification Not green tasks Green tasks Total tasks Non-green task 17,957 293 18,250 Existing green tasks 113 169 282 New green tasks 204 900 1,104 Total tasks 18,274 1,362 19,636 Source: authors As expected, misclassifying green task statements affects the extensive and intensive margins of the GTI. At the occupational level (extensive margin), Table 3.5 shows that our methodology categorizes 862 occupations (or 89 percent) in the same way as O*NET GEP. Specifically, 675 are not on O*NET GEP's list 23 of green occupations; 47 are what O*NET GEP refers to as green increased demand occupations (occupations without green tasks); and most importantly, all green occupations according to O*NET GEP (62 green enhanced skills occupations and 78 green new and emerging occupations) are also green according to the text analysis. However, our methodology also classifies 112 additional occupations as green (meaning they have at least one green task) which are not green according to O*NET GEP (shaded cells in Table 3.5); 17 of these are green increased demand occupations (see Appendix F for the list of all occupations that are green to us but not to O*NET GEP). Table 3.5 O*NET GEP occupations vs Text analysis methodology Text analysis based on narrow green dictionary O*NET occupational classification Zero GTI GTI task Total occupations Non-green occupations 675 95 770 Green increased demand (GID) 47 17 64 Green enhanced skills 62 62 Green new & emerging 78 78 Total occupations 722 252 974 Source: authors When comparing the results of the GTI under both methodologies, although highly correlated (0.9), the text analysis methodology proves to be better for understanding the greenness (or intensive margin) of occupations. Figure 3.4 plots the occupational GTI when calculated through the text analysis methodology versus when calculated based on O*NET GEP. The figure shows that there are many occupations that are either all green or all non-green according to O*NET GEP, while when applying text analysis, occupations have different levels of greenness. Figure 3.4 Occupational GTI text analysis vs O*NET GEP 100 80 GTI O*NET GEP 60 40 20 0 0 20 40 60 80 100 GTI text analysis O*NET Non-green O*NET Green increased demand O*NET Green enhanced skills O*NET New & Emerging Source: authors 24 A detailed look into the task statements of green occupations detected through our dictionary and text analysis but not in the O*NET GEP confirms that these occupations have green task content, even those with the lowest GTI. For example, Home appliance repairers with a GTI of 3.2 have the task ‘conserve, recover, and recycle refrigerants used in cooling systems’, or Real estate brokers with a GTI of 5.2 have tasks such as ‘review property details to ensure that environmental regulations are met’’, or Molecular and cellular biologists with a GTI of 4.7 have a task ‘conduct applied research aimed at improvements in areas such as disease testing, crop quality, pharmaceuticals, and the harnessing of microbes to recycle waste’. Reliability test #2: Correlation between the GTI and green skills intensity Since relying on occupational classifications can be rather static because they are not constantly updated, the second reliability test compares the GTI against a similarly constructed green skills intensity index (GSI) after applying the text analysis methodology to an occupational skills database in the same country. The assumption is that green skills are needed to perform green tasks, and therefore the two indexes should be positively correlated. Recall that the green dictionary is not a list of tasks but of green terms, so the dictionary should be able to pick up skills containing green words. For example, the term ‘environmental’ can pick up the task ‘reporting on the environmental impact of existing and proposed construction’ in the task database and also the skill ‘environmental testing’ in the skills database. It is important to note that merely applying the dictionary to the skills database will not capture all the skills needed to perform a green job since these will be more comprehensive than just skills with green words. This reliability test is carried for two countries: the U.S.—a high-income country with rich data—and Indonesia—a middle-income country with some data. For the U.S., we use the BG skills database for 2017; for Indonesia we also use a database retrieved by BG for 2020. While both databases follow the same protocols, there are important differences related to the size of the online job intermediation market and the localization of the algorithm, mainly the exclusion of postings in Bahasa Indonesian. 56 One problem common to both databases is that they do not cover all occupations. For the U.S., 773 out of 840 SOC titles (version 2010) are observed in the BG database, and for Indonesia, 272 out of 465 occupations appear in the BG database. Given the number of vacancies per occupation in each database, the reliability test includes all occupations for the U.S. and only high-skilled occupations (159) for Indonesia. Applying the (narrow) green dictionary to skills databases resulted in 343 green skills out of 10,798 unique skills for the U.S. and 193 green skills out of 6,263 unique skills for Indonesia. As we did for tasks, a green skills intensity (GSI) index can be defined as the percentage of green skills in an occupation o, as described in the following formula # = # Two things can be concluded from the comparison between the GTI and GSI indexes. First, the applied text analysis methodology is appropriate to capture green tasks, and therefore green jobs, since both 56 For more details on the data collection as well as potential biases see Granata, Posadas, and Testaverde (2021). 25 indexes are positively correlated (Figure 3.5). Second, the skills requirement approach alone is not enough to classify an occupation as green or non-green. While the correlation is high, there is no simple way to determine the level of skills intensity needed to categorize an occupation as green. Leaving aside concerns about biases from online job vacancy data, the extensive margin of the GSI index classifies many more occupations as green than the GTI index. One could argue that the zero cutoff for the extensive margin could be too low for two reasons. First, employers could be tempted to include green skills regardless of their importance for carrying out tasks, simply because they are important for the firm culture or as a signal of other desired skills. Second, the presence of just one vacancy requiring a green skill would result in a positive GSI. And there might be cases in which a green skill is important for a specific job in a particular firm, but may not necessarily be representative of the common skills demand for the occupation. This issue arises because most occupations have at least one vacancy requiring a green skill. Table 3.6 shows that 723 out of the 773 occupations studied in the U.S. and 106 out of the 159 occupations studied in Indonesia have at least one vacancy requiring one or more green skills. To understand where to place the cutoff, we examine how many occupations would be considered green if we raise the cutoff level and compare this with the results of the GTI index. We do this for both countries as described in Table 3.6 and Figure 3.5. It can quickly be inferred that the skills requirement approach leads to high levels of errors (assuming the task approach is correct). Increasing the threshold to be classified as a green job decreases the false positive (type 1 error) but increases the false negative (type 2 error). In other words, when we increase the threshold, fewer non-green jobs are identified as green (false positive), but more green jobs are incorrectly classified as non-green (false negative). This result holds true for both countries. Figure 3.5. GTI vs GSI a) U.S. b) Indonesia 40 30 Incinerator and water treatment plant operators Green Skills Intenisty Narrow 30 Green Skills Intenisty Meteorologists 20 20 Environmental protection professionals 10 Environmental engineers 10 0 0 20 40 60 80 100 0 Green Tasks Intenisty Narrow 0 20 40 60 80 100 Green Tasks Intenisty Yes GT, Yes GS No GT, Yes GS Yes GT, Yes GS No GT, Yes GS Yes GT, No GS No GT, No GS Yes GT, No GS No GT, No GS GS error Fitted values GS error Fitted values Source: authors Table 3.6. Number of green occupations for different cutoff levels of the GSI index and relative to the GTI index U.S. Indonesia If GSI cutoff # of Of which # of non- Of which # of Of which # of non- Of which 26 is… green are green are non- green are green are non- occupati green by occupati green by occupati green by occupati green by ons GTI ons GTI ons GTI ons GTI GSI = 0 723 178 50 46 106 19 53 51 GSI < 0.1 512 167 261 246 89 17 70 66 GSI < 0.25 396 148 377 343 61 16 98 93 GSI < 0.5 284 124 489 431 42 13 117 109 GSI index GSI < 0.75 207 101 566 485 31 11 128 118 GSI < 1 155 84 618 520 24 11 135 125 GSI > 1 Although the skills database may not be the most appropriate tool for identifying green occupations, it can still be very useful in understanding the skills (both green and non-green) that are required in green jobs identified through the task database. This skills profile can then be used to inform training programs. This is discussed further in application #2 in the next section. 4. Application of Green Task Intensity to Indonesia The GTI index at the 4-digit occupational level can then be applied to datasets from countries collecting occupational data with ISCO-8. As an example, this section uses the case of a developing country, Indonesia, to calculate levels of green employment. It shows how different the results are when using the proposed methodology or O*NET data, either the O*NET GEP or the GTI calculated with O*NET task statements. Differences arise from the level of data collection in the country (4-digit)—which is the usual level collected in developing countries—and from the assumptions made when using crosswalks between taxonomies. Next, it describes the skills in demand by green jobs. Lastly, it applies the GTI (binary and continuous) to explore other green employment characteristics: gender and wage premium. Green employment can be measured using the GTI index and Indonesia's labor force survey (Sakernas). In 2016, Sakernas started collecting occupational data using the national standard occupational classification system, Klasifikasi Baku Jenis Pekerjaan Indonesia (KBJI) version 2014.57 The KBJI 2014 corresponds to the adoption of ISCO-08 by Indonesia. However, two important warnings should be kept in mind. First, the illustration is done for Sakernas data for 2017, which is the latest year for which we have access to 4-digit occupational data. 58 Unfortunately, the shared micro data contains KBJI 2002 (based on ISCO-88) instead of KBJI 2014, for which the GTI was computed. Thus, the analysis needs to rely on a crosswalk for employment from KBJI 2002 to KBJI 2014. Such a crosswalk contains several many-to-many matches, which had to be split equally across matches. As a result, employment estimates could be inaccurate for these occupations. Nine out of the 35 green occupations are subject 57 BPS (2014). 58 The team is currently working with BPS to have more recent data including KBJI 2014 classification, and hopefully available by the time of the publication of this note. 27 to these biases. Second, we excluded workers in the agriculture sector and workers in the armed forces (major group 0).59 Application #1: Estimation of employment levels If green employment is defined as working in an occupation that performs at least one green task (meaning a positive GTI), then about 2.3 percent of workers have a green job in Indonesia. When using the broad dictionary, the share of green employment rises to 15 percent, meaning that the greening of the economy could push these workers to perform green tasks if greener technologies were applied in Indonesia (Figure 4.1 panel a). 60 Restricting the analysis now to the narrow definition of green jobs, the majority of jobs are relatively skilled ones: almost two-thirds of green jobs are in occupational group 7:Craft and related trade workers; 29 percent of jobs are high-skilled, either professionals or technicians but not managerial occupations (occupational groups 2 and 3, respectively). Managers, clerical support workers, service and sales representatives (group 1), and elementary occupations (group 9) only represent 6 percent of green jobs. Figure 4.1 Green employment, based on text analysis applied to ISCO-08 taxonomy a) Employment, by GTI b) Share of green employment (narrow definition), by occupation major group 2.3 17.4 Managers Professionals 11 Technicians and associate professionals 18 97.7 Clerical support workers 82.6 Service and sales workers 4 Skilled agricultural, forestry, and fishery workers Craft and related trades workers 61 Plant and machine operators, and Narrow Broad assemblers 4 Elementary occupations Zero GTI GTI Source: GTI and Sakernas 2017 Green employment is significantly larger when using the O*NET taxonomy due to the need to rely on crosswalks to transform occupations from O*NET SOC 8-digit to ISCO 4-digit. Given the lack of 59 Agriculture represents about a third of the employment in Indonesia, and hence it is an important sector to describe. However, the description of the green terms and the task statements were assessed not to be granular enough to produce a meaningful classification. 60 Notice that an alternative way of computing green employment would be to use the GTI to weight employment. This way of computing employment would assume that workers split their time equally across tasks. Since this is a strong assumption, we prefer not to use this measure. 28 employment information at higher disaggregation levels, the common practice is to equally weight all occupations at 8-digits merged to the same 4-digit occupation. Once the crosswalk is done, there are two options to categorize employment into green and non-green: a) assume that the 4-digit occupation is green if it contains at least one green task, or b) assume that only the share of green tasks within the 4-digit occupation represents green employment. If we compute estimates using the O*NET GEP methodology (green enhanced occupations and green new and emerging occupations), then 104 4-digit occupations are green or partially green, 61 which represent between 28 and 15 percent of employment in Indonesia, depending on whether a) or b) is assumed (Figure 4.2). Similarly, when carrying out the text analysis method and applying the narrow green dictionary to the O*NET occupational database, 164 4-digit occupations are green, which represent between 36 and 20 percent of employment, respectively. The employment level is higher using this latter method since the text analysis captured more green tasks and occupations in the O*NET taxonomy than the O*NET GEP. These results imply that, given the level of greening of the economy in Indonesia, the text analysis applied to the ISCO-08 taxonomy may better reflect Indonesia’s level of greening of the economy. Figure 4.2 Green employment using O*NET taxonomy, O*NET GEP and text analysis applied to O*NET taxonomy 15 20 28 36 85 80 72 64 GEP GTI with O*NET taxonomy GEP GTI with O*NET taxonomy Occupation at 4-digit level is green if contains at least Only the share of green tasks is considered green one green task employment Non-green Green Source: authors calculations based on O*NET database Application #2: Skills demanded by online jobs add to high-skilled green occupations Understanding the skills that are in demand for green jobs is essential for supporting the transition to a greener economy. This knowledge can then be used to shape skills policies, particularly in terms of adjusting education and training programs. It can also inform civil society about the specific skills that employers are seeking when hiring for green occupations. The BG online job vacancy data can be a valuable resource for exploring the skills that are demanded for green occupations. Green occupations Elliot et. Al (2021) find that in the Netherlands 106 out of 436 ISCO occupations have some level of greenness 61 when using a crosswalk from O*NET-SOC to ISCO. 29 are identified using the GTI index, which is based on ISCO-08 task statements (described in section 3.A). For the purpose of this discussion, occupations are classified into three categories: Zero GTI (occupations without any green tasks), Low GTI (occupations with less than 30% of their tasks being green), and High GTI (occupations with 30% or more of their tasks being green). It is important to note that, due to the inherent biases of the OV data, only high-skilled occupations with at least 30 job posts in the database are considered for application #2. 62 Project management skills are particularly crucial for high-skilled occupations that involve green tasks. On average, 60% of job advertisements for high GTI occupations require at least one project management skill, which is 10% higher compared to occupations without green tasks (Figure 4.3). Specifically, jobs in high GTI occupations typically require an average of 2.19 management skills. Examples of in-demand project management skills for these occupations include quality assurance and control, planning, people development, and compliance management (Table 4.1). There is a high demand for these skills in green occupations compared to non-green occupations. For instance, the skill of quality assurance appears 4.2 times more frequently in green jobs than in non-green jobs. Similarly, jobs in high GTI occupations are more likely to require cognitive skills and specific field expertise. Table 4.1 illustrates the relative importance of these skills in high GTI occupations compared to occupations without green tasks (column 5). Skills such as fiber optics, research, occupational health and safety, and environmental management are particularly relevant to jobs that involve green tasks. Figure 4.3. Average share of posts with at least one skill demanded in a skill group, by GTI index 0 10 20 30 40 50 60 Cognitive (analytical) Social Character Writing Customer Project management and business People management Financial Computer (basic) Software and information technology Foreign language Field of study and industry knowledge Zero GTI Low GTI <30% High GTI +30% Notes: analysis restricted to high-skilled occupations with at least 30 posts. For each occupation, we calculate the percentage of posts that required at least one skill in each of the skills groups. Then, we calculate the average of this percentage by the three categories: Zero GTI, Low GTI, and High GTI. Table 4.1. Top 20 skills demanded in high GTI occupations Skill ranking Skill 62 See Granata, Posadas, Testaverde (2021) for a discussion on the biases of Indonesia’s online vacancy data. 30 Skill Zero GTI Low GTI High GTI relative Skills group importance Communication Skills 2 4 1 1,03 Social English 1 1 2 0,90 Foreign language Quality Assurance and 40 7 3 4,26 Project management and Control business Fiber Optics 1965 702 4 47,46 Field of study and industry knowledge Research 14 5 5 2,01 Cognitive (analytical) Planning 6 6 6 1,35 Project management and business People Development 621 1156 7 13,85 Project management and business Teamwork / Collaboration 5 11 8 0,93 Social Microsoft Excel 7 15 9 1,12 Computer (basic) Problem Solving 13 12 10 1,24 Cognitive (analytical) Microsoft Office 10 8 11 0,95 Computer (basic) Project Management 39 33 12 2,10 Project management and business Occupational Health and 498 227 13 8,03 People management Safety Environmental 963 369 14 12,06 Project management and Management business Writing 15 18 15 1,07 Writing Engineering 33 22 16 1,52 Cognitive (analytical) Budgeting 18 37 17 1,06 Financial Fundraising 935 478 18 9,77 Customer Policy Research 3684 19 51,22 Cognitive (analytical) Compliance Management 1244 267 20 11,26 Project management and business Source: authors using OV data. Application #3: Gender and education differences in employment Access to green jobs is not even across groups of workers, with women being less likely to hold a green job independently of the definition applied. Females hold 9 percent of green jobs (Figure 4.4 panel a). This is significantly lower when compared to zero GTI (non-green) jobs, for which female represent 41 percent of workers. Furthermore, green jobs employ more educated workers (although not necessarily at the university level) (Figure 4.4 panel b). Two-thirds of green workers have at least a high-school or vocational degree as compared to half zero GTI workers. When using the narrow definition, our analysis also suggests that workers at green jobs are younger (43 percent of non-green workers are under the age of 35 while 54 percent of green workers are), which may be related to new jobs emerging thanks to the greening of the economy. Peters (2013) found similar results in the U.S. 31 Figure 4.4 Share of employment, by workers’ characteristics based on text analysis applying green narrow dictionary to ISCO-08 taxonomy a. Green employment, by gender b. Green employment, by education achievement 13 11 4 3 59 33 51 91 19 41 21 19 9 10 13 4 Zero GTI GTI Zero GTI GTI Less than elementary Elementary school Female Male Middle school High school+vocational Diploma University or higher Source: GTI and Sakernas 2017 Application #4: Green jobs wage premium Green jobs are better paid jobs. This result that is found for the U.S. and other countries is corroborated for Indonesia. Comparing jobs within broad occupational groups which can be thought to require similar levels of effort and training and considering individual workers’ and jobs’ characteristics, 63 green jobs pay on average 6 percentage points more than a job that does not involve any green task (Table 4.2 column A). The higher the GTI, the better on average the job pays: when the GTI increases by 1 percentage point, the wage increases by 0.2 percentage points (column C). However, wage gains depend on the major occupational group: professionals; skilled agricultural, forestry, and fishery workers; and craft and related trades workers, who all represent 72 percent of green jobs, are paid better to those in the same occupational group but at non-green jobs. While managers, Technicians and associate professionals, Service and sales workers, and Plant and machine operators, and assemblers are paid less that those at the same non-green major occupational group. Table 4.2 Green jobs wage premium based on text analysis applying green narrow dictionary to ISCO- 08 taxonomy GTI (binary, green vs non- green) GTI Index (continuous) A B C D ln (wage) ln (wage) ln (wage) ln (wage) GTI 0.062*** 0.100*** 0.002*** 0.007*** (0.000) (0.003) (0.000) (0.000) Major occupational group (1-digit ISCO) *GTI 63 Based on a wage regression controlling for gender, education level, broad occupational groups (1-digit KBJI), province of residence, working hours, and sector of employment. 32 Managers * GTI -0.449*** -0.049*** (0.010) (0.001) Professionals * GTI 0.145*** -0.002*** (0.003) (0.000) Technicians and associate professionals * GTI -0.118*** -0.009*** (0.003) (0.000) Clerical support workers * GTI, omitted - - Service and sales workers * GTI -0.137*** -0.008*** (0.004) (0.000) Skilled agricultural, forestry, and fishery workers * GTI 0.585*** 0.027*** (0.034) (0.002) Craft and related trades workers * GTI -0.028*** -0.004*** (0.003) (0.000) Plant and machine operators, and assemblers * GTI -0.147*** -0.010*** (0.004) (0.000) Elementary workers * GTI, omitted CONTROLS Occupation major group (1-digit) YES YES YES YES Education attainment YES YES YES YES Sector of employment YES YES YES YES Gender YES YES YES YES Age YES YES YES YES Province YES YES YES YES Working hours YES YES YES YES Constant 12.874*** 12.876*** 12.874*** 12.875*** (0.001) (0.001) (0.001) (0.001) Observations weighted 53,022,977 53,022,977 53,022,977 53,022,977 Observations un-weighted 300,518 300,518 300,518 300,518 R-squared 0.515 0.515 0.515 0.515 Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1 5. Conclusions This paper provides a valuable methodology and toolkit for both analysts and policy makers engaged in the examination of skills policies, particularly those related to green jobs. For analysts, the adoption of the task approach for analysis is recommended, as it offers a more nuanced understanding on how to identify occupations that are core to the greening of the economy and the skills required by workers to perform green tasks. The methodology proposed in this paper, based on text analysis, is recommended due to its ability to mitigate biases that may arise from importing classifications from other countries; but it requires having access to a list of task statements, for example those found in an occupational 33 manual. The methodological annex further equips practitioners with a toolkit for effective implementation. Policy makers focusing on skills policy are urged to adopt a narrow definition of green jobs, emphasizing specific tasks aimed at reducing environmental impact. The paper recommends relying on the task approach and the proposed methodology to avoid assumptions inherent in classifications from other countries. Moreover, the utilization of online vacancy data with rich skills information is encouraged for profiling green occupations as well as understanding which skills may become in-demand in non-green occupations as the greening of the economy advances, offering a more dynamic and real-time understanding. The overarching recommendation for policy makers involves the development of a robust labor market information system. This includes investments in detailed occupational classifications or job title taxonomies, updated more frequently than traditional classifications, and informed by ISCO-08, online job vacancy data, ONET-type data, and qualitative insights from labor market stakeholders. Additionally, the collection of skills and tasks data for narrowly defined occupations is emphasized, drawing lessons from established frameworks like ONET, ESCO, PIIAC, and STEP. Regular and comprehensive surveys are proposed to monitor changes in the labor market, inform skills policies, and enhance the understanding of green jobs and the skills they require. It is crucial to recognize that green skills policy is just one facet of broader skills policies promoting the transition to a green economy. The paper underscores the importance of comprehensive policy instruments, such as advanced labor market information systems, to keep the population informed about in-demand jobs across various sectors. The integration of identified green jobs into competency standards, qualification frameworks, curricula, and certification programs is highlighted as an essential step for countries with well-established systems and a critical focus for those in the process of modernizing their educational, TVET, and employment systems. Accelerating this modernization is deemed vital to capitalize on the opportunities presented by the green transition. 6. References Alatas, H.; Granata, J.; and Posadas, J. (2020) Indonesia’s Occupational Tasks and Skills: from occupational employment demand to tasks and skills requirements. World Bank. Apella, I.; Rovner (2021) Cambios ocupacionales y formación a lo largo de toda la vida laboral. Aportes del sistema de información ocupacional de Uruguay (O*NET UY). Working Draft. Acemoglu, D.; Autor, D. H. (2011): Skills, tasks and technologies: Implications for employment and earnings (Chapter 12). In: Ashenfelter, A.; Card D. (Eds.) Handbook of Labor Economics, 4B, 1043- 1171. ASEAN and ILO (2021). Regional Study on Green Jobs Policy Readiness in ASEAN. Final Report. ILO Thailand. ISBN 978-623-6945-15-5 34 BLS. (2011) Counting Green Jobs: Developing the Green Technologies and Practices (GTP) Survey. Conference paper: Stang, S. and Jones, C. Section on Survey Research Methods : JSM 2011 ____. (2012a) Occupational employment and wages in green goods and services: November 2011. U.S. Bureau of Labor Statistics. News release: September 28, 2012 ____. (2012b) Green technologies and practices: August 2011. U.S. Bureau of Labor Statistics. News release: June 29, 2012. ____. (2013a) Green Goods and Services survey: results and collection. U.S. Bureau of Labor Statistics. Monthly Labor Review. September 2013. ____. (2013b) The Green Goods and Services Occupational survey: initial results. Warren, Zack. U.S. Bureau of Labor Statistics. Monthly Labor Review. January 2013. ____. (n.d.). Measuring green jobs. U.S. Bureau of Labor Statistics (BLS). https://www.bls.gov/green/#definition (accessed on 06/03/2022) Bontadini, F.; Vona, F. (2021) Anatomy of Green Specialization: Evidence from EU Production Data, 1995-2015. Sciences Po OFCE Working Paper, n° 21/2020 Bowen, A.; Kuralbayeva, K.; and Tipoe, E.L. (2018) Characterizing green employment: The impacts of ‘greening’ on workforce composition. Energy Economics, 72: 263-275. BPS (2014). Klasifikasi Baku Jabatan Indonesia (KBJI) 2014. Kementerian Ketenagakerjaan dan Badan Pusat Statistik. Consoli, D.; Marin, G.; Marzucchi, A.; Vona F. (2016): Do green jobs differ from nongreen jobs in terms of skills and human capital? In: Research Policy, 45(5), 1046- 1060. Dierdorff, E.; Norton, J.; Drewes, D. W.; Kroustalis, C.; Rivkin, D.; and Lewis, P. (2009) Greening of the World of Work: Implications for O*NET® -SOC and New and Emerging Occupations. The National Center for O*NET Development. Dierdorff, E.; Norton, J.; Gregory, C.; Rivkin, D.; and Lewis, P. (2011) Greening of the World of Work: Revisiting Occupational Consequences. The National Center for O*NET Development. Elliott, R.; Kuai, W.; Maddison, D.; Ozgen, C. (2021) Eco-Innovation and Employment: A Task-Based Analysis. IZA Discussion Papers No. 14028 Eurostat (2009). The environmental goods and services sector. doi 10.2785/31117 Granata, J.; Posadas, J.; Testaverde, M. (2021) Indonesia’s Online Vacancy Outlook: from online job postings to labor market intelligence 2020. World Bank. Granata, J.; Moroz, H. (2022) Identifying Skills Needs in Vietnam: The Survey of Detailed Skills. World Bank. Gregg, C.; Strietska-Ilina, O.; Büdke, C. (2015) Anticipating skill needs for green jobs: a practical guide. Skills and Employability Branch Employment Policy Department. ILO. Geneva. 35 ILO (2016) Technical paper. A just transition to climate-resilient economies and societies: Issues and perspectives for the world of work. Geneva: ILO. ___. (2019) Skills for a greener future: A global view based on 32 country studies. Geneva: ILO. ___. (2021) Global framework on core skills for life and work in the 21st century. Switzerland: ILO Janser, Markus (2018a) The greening of jobs in Germany: First evidence from a text mining based index and employment register data, IAB-Discussion Paper, No. 14/2018, Institut für Arbeitsmarkt- und Berufsforschung (IAB), Nürnberg ___. (2018b) The greening of jobs in Germany: Online Appendix ‘text Mining & Descriptives’ Institut für Arbeitsmarkt- und Berufsforschung (IAB), Nürnberg Lewis, P., Morris, J., and Gregory, C. (2022) Green Topics: Identifying Linkages to Occupations and Education Programs using a Linguistic Approach. O*NET working paper. June 2022 OECD, Eurostat (1999). The environmental goods and services industry: manual for data collection and analysis. O*NET (2010) O*NET® Green Task Development Project. The National Center for O*NET Development. ___. (2013) Greening of the World of Work: O*NET® Project’s Book of References. The National Center for O*NET Development. Oyunbileg, D., Stoevska, V. (2017) Employment in the environmental sector and green jobs in Mongolia. Pilot Study. International Labour Organization (ILO) Stoevska, V. Hunter, D. (2012) Proposals for the statistical definition and measurement of green jobs. Discussion paper prepared for informal consultation, November 2012. International Labour Office. Geneva. Stoevska, V., Elezi, P., and Muraku, E. (2014) Report on the pilot project towards developing statistical tools for measuring employment in the environmental sector and generating statistics on green jobs. Albania. Intenrational Labour Organization (ILO) Stoveska (2017). ILO measuring conference - Measuring green jobs current proposals.pdf Ruppert Bulmer, Elizabeth, Kevwe Pela, Andreas Eberhard-Ruiz and Jimena Montoya. 2021. “Global Perspective on Coal Jobs and Managing Labor Transition out of Coal.” World Bank, Washington, DC. Rutzer, C.; Niggli, M.; and, Weder, R. (2020) Estimating the Green Potential of Occupations: A New Approach Applied to the U.S. Labor Market. WWZ Working Paper 2020/03 UNEP, ILO, IOE, ITUC (2008). Green jobs: Towards decent work in a sustainable, low-carbon world. United Nations Environment Program (UNEP), ILO, International Organisation of Employers (IOE), International Trade Union Confederation (ITUC). Vidican Auktor, Georgeta V. (2020). Green Industrial Skills for a Sustainable Future. Vienna: UNIDO 36 Vona, F. (2021) Labour Markets and the Green Transition: a practitioner’s guide to the task-based approach. Publications Office of the European Union, Luxembourg, ISBN 978-92-76-42260-0, doi:10.2760/65924, JRC126681. Vona, F.; Marin, G.; and Consoli, D. (2019). Measures, drivers and effects of green employment: Evidence from US local labor markets, 2006-2014. Journal of Economic Geography, 19(5): 1021- 1048. Vona, F.; Marin, G.; Consoli, D.; and Popp, D. (2018) Environmental Regulation and Green Skills: An Empirical Exploration. JAERE, vol.5, number 4. The Association of Environmental and Resource Economists. doi.org/10.1086/698859 37 Appendices 38 Appendix A. Pilots combining the output approach and the process approach in developing countries The ILO carried out two interesting pilot surveys to test the output and process approaches in developing countries. These pilots were deployed in Albania and Mongolia. 64 The instrument differed to those of the BLS in three ways. First, instead of a standalone survey the pilots were thought of as modules that could be attached to already established surveys. Second, while the U.S. examples are both enterprise surveys, the ILO piloted an enterprise module and a household module to be attached to already established surveys—the enterprise survey (ES) in Albania and the national labor force survey (LFS) in Mongolia. The reason for testing a household module was that in developing countries the informal sector, the self-employed, and the agriculture sector are a large part of the economy and are usually not captured in enterprise surveys. Third, the pilots included in both modules a combination of questions from both the GGS and the GTP surveys, with some adaptations to developing countries. For example, the instrument includes sustainable and organic agriculture in a separate category to the GTP conserving natural resources. Despite modules parallelism, results are not comparable due to differences in sampling frames and the underlying differences in each of the surveys. For example, the Albanian Enterprise Survey (ES) did not include the agriculture sector while the labor force survey (LFS) did, and the Mongolian ES did not include the banking and financial sector, private health sector, and government agencies while the LFS did. Although the pilots have limitations natural to the approaches, they also present some advantages. First, by including in a single firm survey instrument questions on outputs and processes, jobs can be added up to get a single number of jobs (those involved in the production of an EGS and/or those involved in environmentally friendly processes), without the issue of double counting. Second, since the household module is attached to the LFS, it also provides useful information for occupational profiling of such jobs, including occupation titles, gender, age, educational achievements, and wages. The disadvantage of this combined approach is that it still might be including jobs which do not have any green tasks by counting all jobs at a firm producing a green output and all jobs using green technologies or carrying out green practices (but not necessarily green tasks). 64 Stoevska, Elezi, & Muraku (2014) and Oyunbileg & Stoevska (2017). 39 Appendix B. O*NET GEP Steps for classifying green occupations 1 Locate and review existing literature. About 60 reports Identify and compile job titles. Accumulating a list of all job titles that were 2 mentioned in the reports Review and sort collected job titles. Eliminate too molecular or too molar job titles 3 (2%). Kept 467 job titles 4 Cluster job titles to identify occupations. Similar job titles were clustered 5 Identify occupational sectors. A 12 schema was used to group job titles 6 Determine overlap with ONET. Determine if the occupation was existent in ONET SOC or was a new occupation, and generate the 3 grouping: Green increased demand, green 7 Identify new and emerging. Group job titles into occupations 8 Research potential N&E occupations 9 Build consolidated evidence for final N&E determination 10 Compile and report N&E skills Source : Authors based on Dierdorff et al. (2009), Dierdorff et al. (2011) and O*NET (2010). Appendix C. List of consulted material for the creation of the green dictionary Output approach BLS (2010). Green Goods and Services Survey Questionnaire. O.M.B. No. 1220−0181 BLS (2010). Industries where green goods and services are classified. Excel file with list of green industries at NAICS level. https://www.bls.gov/ggs/ BLS (n.d.) Survey Methods and Reliability Statement for Occupational Employment and Wages in Green Goods and Services . https://www.bls.gov/ggsocc/survey_methods.pdf BLS. (2012) Occupational employment and wages in green goods and services: November 2011. U.S. Bureau of Labor Statistics. News release: September 28, 2012 40 BLS. (2013) Green Goods and Services survey: results and collection. U.S. Bureau of Labor Statistics. Monthly Labor Review. September 2013. BLS. (2013) The Green Goods and Services Occupational survey: initial results. Warren, Zack. U.S. Bureau of Labor Statistics. Monthly Labor Review. January 2013. Canada Survey of Environmental Goods and Services, 2012. Environment Accounts and Statistics Division. Statistics Canada Environmental Goods and Services Sector (EGSS): list of products and activities. Excel. Version as of 6/1/2016 Eurostat (2009). The environmental goods and services sector. doi 10.2785/31117 Federal Planning Bureau (2014). The Environmental Goods and Services Sector in Belgium 2010-2011. Report for Eurostat OECD, Eurostat (1999). The environmental goods and services industry: manual for data collection and analysis. Process approach BLS (2012). Green Technologies and Practices Questionnaire. O.M.B. No. 1220−0184 BLS. (2011) Counting Green Jobs: Developing the Green Technologies and Practices (GTP) Survey. Conference paper: Stang, S. and Jones, C. Section on Survey Research Methods : JSM 2011 BLS. (2012) Green technologies and practices: August 2011. U.S. Bureau of Labor Statistics. News release: June 29, 2012. Oyunbileg, D., Stoevska, V. (2017) Employment in the environmental sector and green jobs in Mongolia. Pilot Study. International Labour Organization (ILO) Stoevska, V., Elezi, P., and Muraku, E. (2014) Report on the pilot project towards developing statistical tools for measuring employment in the environmental sector and generating statistics on green jobs. Albania. Intenrational Labour Organization (ILO) Task approach Dierdorff, E.; Norton, J.; Drewes, D. W.; Kroustalis, C.; Rivkin, D.; and Lewis, P. (2009) Greening of the World of Work: Implications for O*NET® -SOC and New and Emerging Occupations. The National Center for O*NET Development. Dierdorff, E.; Norton, J.; Gregory, C.; Rivkin, D.; and Lewis, P. (2011) Greening of the World of Work: Revisiting Occupational Consequences. The National Center for O*NET Development. Excel shared by Janser with green jobs terms Janser, Markus (2018) The greening of jobs in Germany: First evidence from a text mining based index and employment register data, IAB-Discussion Paper, No. 14/2018, Institut für Arbeitsmarkt- und Berufsforschung (IAB), Nürnberg Janser, Markus. (2018b) The greening of jobs in Germany: Online Appendix ‘text Mining & Descriptives’ Institut für Arbeitsmarkt- und Berufsforschung (IAB), Nürnberg 41 O*NET (2010) O*NET® Green Task Development Project. The National Center for O*NET Development. O*NET. (2013) Greening of the World of Work: O*NET® Project’s Book of References. The National Center for O*NET Development. ONET Green Tasks Statmenets excel https://www.onetcenter.org/dictionary/20.1/excel/green_task_statements.html Skills approach Burning Glass Technologies (2021). After the Storm. The Jobs and Skills that will Drive the Post-Pandemic Recovery ESCO. Skill/Competence S.3.3.2 follow environmentally-sustainablework practices http://data.europa.eu/esco/skill/a992f345-7c06-4982-8fc9-5fab55e316af ESCO. Skill/Competence S.3.3.2 complying with environmentalprotection laws and standards http://data.europa.eu/esco/skill/S3.3.2 Excel shared by Burning Glass with a list of green job titles terms Papers, reports, and websites Bontadini, F.; Vona, F. (2021) Anatomy of Green Specialization: Evidence from EU Production Data, 1995-2015. Sciences Po OFCE Working Paper, n° 21/2020 Bowen, A.; Kuralbayeva, K.; and Tipoe, E.L. (2018) Characterizing green employment: The impacts of ‘greening’ on workforce composition. Energy Economics, 72: 263-275. Chen, Marin, Popp, Vona (2020) Green stimulus in a post-pandemic recovery: the role of skills for a resilient recovery Consoli, D.; Marin, G.; Marzucchi, A.; Vona F. (2016): Do green jobs differ from nongreen jobs in terms of skills and human capital? In: Research Policy, 45(5), 1046- 1060. Elliott, R.; Kuai, W.; Maddison, D.; Ozgen, C. (2021) Eco-Innovation and Employment: A Task- Based Analysis. IZA Discussion Papers No. 14028 Elliott, R.; Lindley, J. (2016). Environmental Jobs and Growth in the United States. http://dx.doi.org/10.1016/j.ecolecon.2016.09.0300921-8009/ ILO (2016) Technical paper. A just transition to climate-resilient economies and societies: Issues and perspectives for the world of work. Geneva: ILO. Marin, G.; Vona, F., (2019) Climate policies and skill-biased employment dynamics: Evidence from EU countries. https://doi.org/10.1016/j.jeem.2019.102253 Martinez-Fernandez,C; Hinojosa, C.; Miranda, G.(2010) Greening Jobs and Skills: Labour Market Implications of Addressing Climate Change. OECD Local Economic and Employment Development (LEED) Papers 2010/02 Peters, D. (2013) Understanding Green Occupations from a Task-Based Approach. Applied Economic Perspectives and Policy (2014) volume 36, number 2, pp. 238-264. doi:10.1093/ aepp / ppt026 Popp, D.; Vona, F.; Marin, G.; Chen, Z. (2020) The Employment Impact of Green Fiscal Push: Evidence from the American Recovery Act. NBER Working Paper No. 27321 Rutzer, C.; Niggli, M.; and, Weder, R. (2020) Estimating the Green Potential of Occupations: A New Approach Applied to the U.S. Labor Market. WWZ Working Paper 2020/03 42 Vona, F. (2021) Labour Markets and the Green Transition: a practitioner’s guide to the task- based approach. Publications Office of the European Union, Luxembourg, ISBN 978-92-76- 42260-0, doi:10.2760/65924, JRC126681. Vona, F.; Marin, G.; and Consoli, D. (2019). Measures, drivers and effects of green employment: Evidence from US local labor markets, 2006-2014. Journal of Economic Geography, 19(5): 1021-1048. Vona, F.; Marin, G.; Consoli, D.; and Popp, D. (2018) Environmental Regulation and Green Skills: An Empirical Exploration. JAERE, vol.5, number 4. The Association of Environmental and Resource Economists. doi.org/10.1086/698859 ASEAN and ILO (2021). Regional Study on Green Jobs Policy Readiness in ASEAN. Final Report. ILO Thailand. ISBN 978-623-6945-15-5 BLS (2013). Green jobs overview. Monthly Labor Review. Dixie Sommers BLS. (n.d.). Measuring green jobs. U.S. Bureau of Labor Statistics (BLS). https://www.bls.gov/green/#definition (accessed on 06/03/2022) Cedefop (2012). Green skills and environmental awareness in vocational education and training. Synthesis report. Luxembourg: Publications Office of the European Union Cedefop (2018). Skills for green jobs: an update. Denmark. European Centre for the Development of Vocational Training Cedefop (2018). Skills for green jobs: an update. Estonia. European Centre for the Development of Vocational Training Cedefop (2018). Skills for green jobs: an update. Germany. European Centre for the Development of Vocational Training Duell, N. (2021) Greening of the labour market – impacts for the Public Employment Services. Small scale study. European Network of Public Employment Services EcoCanada (2016). Competencies for Environmental Professionals in Canada. National Occupational Standards. Environmental Careers Organization of Canada EcoCanada (2021). A National Sector Workforce Strategy to Address Environmental Talent Needs and Gaps. PPT Environmental Entrepreneurs (2020) CLEAN JOBS, BETTER JOBS An examination of clean energy job wages and benefits Gregg, C.; Strietska-Ilina, O.; Büdke, C. (2015) Anticipating skill needs for green jobs: a practical guide. Skills and Employability Branch Employment Policy Department. ILO. Geneva. ILO (2011) Study of occupational and skill needs in green building: final report / International Labour Office, ILO Skills and Employability Department (EMP /SKILLS ). – Geneva: ILO , 2011 ILO (2011) Study of occupational and skill needs in renewable energy: final report / International Labour Office, ILO Skills and Employability Department (EMP /SKILLS ). – Geneva: ILO , 2011 ILO (2011). Comparative analysis of methods of identification of skill needs on the labour market in transition to the low carbon economy: final report / International Labour Office, ILO Skills and Employability Department (EMP /SKILLS ). - Geneva: ILO ILO (2013). Meeting skill needs for green jobs: Policy recommendations ILO. (2019) Skills for a greener future: A global view based on 32 country studies. Geneva: ILO. ILO. (2021) Global framework on core skills for life and work in the 21st century. Switzerland: 43 ILO International Labour Organization (International Institute for Labour Studies) (2011) TOWARDS A GREENER ECONOMY: THE SOCIAL DIMENSIONS Lobao, L.; Partridgea, M.; Heana, O.; Kellya, P.; Chunga, S.; Ruppert Bulmerb, E. (2021). Socioeconomic Transition in the Appalachia Coal Region Some Factors of Success New Hampshire Employment Security, Economic and Labor Market Information Bureau. (2011)New Hampshire Green Jobs Survey. OECD (2012) The jobs potential of a shift towards a low-carbon economy FINAL REPORT FOR THE EUROPEAN COMMISSION, DG EMPLOYMENT Stoevska, V. Hunter, D. (2012) Proposals for the statistical definition and measurement of green jobs. Discussion paper prepared for informal consultation, November 2012. International Labour Office. Geneva. Stoveska (2017). ILO measuring conference - Measuring green jobs current proposals.pdf The Institution of Engineering and Technology (2021). IET skills for net zero and a green recovery 2020 survey. Examining the engineering skills needed to meet net zero. UNEP, ILO, IOE, ITUC (2008). Green jobs: Towards decent work in a sustainable, low-carbon world. United Nations Environment Program (UNEP), ILO, International Organisation of Employers (IOE), International Trade Union Confederation (ITUC). Vidican Auktor, Georgeta V. (2020). Green Industrial Skills for a Sustainable Future. Vienna: UNIDO World Bank (2018). Managing Coal Mine Closure. Achieving a Just Transition for All World Bank (2021) Supporting Transition in Coal Regions A Compendium of the World Bank’s Experience and Guidance for Preparing and Managing Future Transitions Appendix D. ISCO classification system 65 The level of analysis is the occupation at the 4-digit KBJI level. Indonesia adopted the KBJI classification following the International Standard Classification of Occupations 2008 (ISCO-08), which is a hierarchically structured system classifying and aggregating occupational information of all jobs in the world into 436 unit groups to be used for statistical census and surveys as well as administrative databases. ISCO was designed as a tool for international comparisons and not as a replacement of national classification systems since adaptation to the local context is extremely important. However, countries with lack of capacity to develop their own classification systems find it convenient to adopt ISCO-08. Each unit group (referred as occupation now on) is made up of several occupations with high degree of similarity in terms of skill level and skill specialization. Each occupation is defined with a 4-digit code, a title, a job description delimiting the scope of the group, the tasks performed, and examples of occupations included. The unit groups are arranged into minor groups (3-digit codes, 130 in total), which are arranged into sub-major group (2-digit codes, 43 in total), which are arranged into major groups (1-digit code, 10 in total). The major group depends on the occupation skill level: the complexity and variety of tasks to be performed in the occupation, measured mainly by the nature of work performed and also by formal and informal education required. And the occupation’s minor 65 Source: ILO (2012). 44 and sub-major group depends on the skill specialization: the field of knowledge required, tools and machinery used, materials worked on or with, and kinds of goods and services produced. See Tables D1 and D2 for details. Table D1. ISCO mapping of major groups according to skill level. Table D2. ISCO definition of each skill level Skill Job usually involves the performance Type of skills that the job Knowledge and skills usually level of may require obtained through 1 simple and routine physical or Physical strength and/or primary education or the first stage manual tasks. endurance of basic education. A short period of on-the-job training may be required. 2 tasks operating machinery and Relatively advanced First stage of secondary education electronic equipment, driving literacy and numeracy or higher, some may require vehicles, maintenance and repair of skills and good vocation-specific training. In some electrical and mechanical interpersonal skills. Many cases, o-the-job training may equipment, and manipulation, occupations require high substitute for formal education, ordering and storage of information. level of manual dexterity. while in other may require significant specialized vocational education and on-the-job training. 3 complex technical and practical tasks High level of literacy and A higher educational institution for and require extensive body of numeracy skills and well- a period of 1-3 years (first stage of factual, technical, and procedural developed interpersonal tertiary education of short or knowledge in a specialized field communication skills. medium duration, or higher) 4 task that requires complex problem High to very-high levels of A higher educational institution for solving, decision making, and literacy and numeracy a period of 3-6 years (first stage of creativity based on an extensive skills and excellent tertiary education of medium body of theoretical and factual interpersonal duration, or higher) knowledge in a specialized field. communication skills. Source: ILO (2012) 45 Appendix E. GTI for ISCO-08 occupational classification system ISCO-08 Occupation title GTI Narrow GTI Broad 2143 Environmental engineers 88.89 88.89 2133 Environmental protection professionals 85.71 85.71 9612 Refuse sorters 83.33 100.00 2112 Meteorologists 77.78 88.89 9611 Garbage and recycling collectors 75.00 75.00 3132 Incinerator and water treatment plant operators 50.00 50.00 3257 Environmental and occupational health inspectors and associates 40.00 50.00 3119 Physical and engineering science technicians not elsewhere classified 40.00 40.00 2131 Biologists, botanists, zoologists and related professionals 37.50 50.00 7234 Bicycle and related repairers 33.33 50.00 7124 Insulation workers 33.33 33.33 5411 Fire-fighters 33.33 33.33 2263 Environmental and occupational health and hygiene professionals 30.00 40.00 2142 Civil engineers 28.57 57.14 2132 Farming, forestry and fisheries advisers 25.00 100.00 7122 Floor layers and tile setters 25.00 25.00 2114 Geologists and geophysicists 25.00 25.00 2113 Chemists 25.00 25.00 3112 Civil engineering technicians 22.22 33.33 3143 Forestry technicians 20.00 100.00 6210 Forestry and related workers 20.00 80.00 5419 Protective services workers not elsewhere classified 20.00 20.00 8182 Steam engine and boiler operators 16.67 16.67 3131 Power production plant operators 16.67 16.67 3111 Chemical and physical science technicians 16.67 16.67 8142 Plastic products machine operators 14.29 28.57 9623 Meter readers and vending-machine collectors 14.29 14.29 7231 Motor vehicle mechanics and repairers 12.50 25.00 2164 Town and traffic planners 12.50 25.00 2162 Landscape architects 12.50 12.50 6221 Aquaculture workers 11.11 44.44 2163 Product and garment designers 11.11 11.11 2149 Engineering professionals not elsewhere classified 11.11 11.11 2141 Industrial and production engineers 10.00 10.00 1311 Agricultural and forestry production managers 8.33 100.00 7541 Underwater divers 8.33 25.00 9215 Forestry laborers - 87.50 6112 Tree and shrub crop growers - 81.82 6111 Field crop and vegetable growers - 81.82 6123 Apiarists and sericulturists - 66.67 6114 Mixed crop growers - 63.64 8341 Mobile farm and forestry plant operators - 62.50 46 7115 Carpenters and joiners - 60.00 7422 Information and communications technology installers and servicers - 57.14 7412 Electrical mechanics and fitters - 57.14 7233 Agricultural and industrial machinery mechanics and repairers - 57.14 6113 Gardeners, horticultural and nursery growers - 54.55 9624 Water and firewood collectors - 50.00 7215 Riggers and cable splicers - 50.00 6130 Mixed crop and animal producers - 50.00 7213 Sheet-metal workers - 42.86 7232 Aircraft engine mechanics and repairers - 40.00 3151 Ship’s engineers - 40.00 7536 Shoemakers and related workers - 38.46 1312 Aquaculture and fisheries production managers - 38.46 9211 Crop farm laborers - 37.50 3142 Agricultural technicians - 37.50 3118 Draughts persons - 37.50 7222 Toolmakers and related workers - 36.36 8332 Heavy truck and lorry drivers - 33.33 7421 Electronics mechanics and servicers - 33.33 7413 Electrical line installers and repairers - 33.33 7112 Bricklayers and related workers - 33.33 3113 Electrical engineering technicians - 33.33 9622 Odd job persons - 28.57 8331 Bus and tram drivers - 28.57 6310 Subsistence crop farmers - 28.57 9213 Mixed crop and livestock farm laborers - 27.27 7411 Building and related electricians - 25.00 7311 Precision-instrument makers and repairers - 25.00 7127 Air conditioning and refrigeration mechanics - 25.00 5153 Building caretakers - 25.00 3155 Air traffic safety electronics technicians - 25.00 9214 Garden and horticultural laborers - 22.22 6330 Subsistence mixed crop and livestock farmers - 22.22 8311 Locomotive engine drivers - 20.00 7515 Food and beverage tasters and graders - 20.00 7514 Fruit, vegetable and related preservers - 20.00 7126 Plumbers and pipe fitters - 20.00 7114 Concrete placers, concrete finishers and related workers - 20.00 6224 Hunters and trappers - 20.00 6222 Inland and coastal waters fishery workers - 20.00 4223 Telephone switchboard operators - 20.00 3522 Telecommunications engineering technicians - 20.00 3116 Chemical engineering technicians - 20.00 7531 Tailors, dressmakers, furriers and hatters - 18.18 47 8141 Rubber products machine operators - 16.67 7534 Upholsterers and related workers - 16.67 7523 Woodworking-machine tool setters and operators - 16.67 7522 Cabinet-makers and related workers - 16.67 7214 Structural-metal preparers and erectors - 16.67 6122 Poultry producers - 16.67 3230 Traditional and complementary medicine associate professionals - 16.67 6121 Livestock and dairy producers - 15.38 7323 Print finishing and binding workers - 14.29 7224 Metal polishers, wheel grinders and tool sharpeners - 14.29 7221 Blacksmiths, hammersmiths and forging press workers - 14.29 7211 Metal molders and coremakers - 14.29 7113 Stonemasons, stone cutters, splitters and carvers - 14.29 7111 House builders - 14.29 3114 Electronics engineering technicians - 14.29 2153 Telecommunications engineers - 14.29 2144 Mechanical engineers - 14.29 8322 Car, taxi and van drivers - 12.50 7212 Welders and flame cutters - 12.50 6340 Subsistence fishers, hunters, trappers and gatherers - 12.50 6223 Deep-sea fishery workers - 12.50 3152 Ships' deck officers and pilots - 12.50 3117 Mining and metallurgical technicians - 12.50 3115 Mechanical engineering technicians - 12.50 2152 Electronics engineers - 12.50 9332 Drivers of animal-drawn vehicles and machinery - 11.11 8157 Laundry machine operators - 11.11 8111 Miners and quarriers - 11.11 7322 Printers - 11.11 8114 Cement, stone and other mineral products machine operators - 10.00 8113 Well drillers and borers and related workers - 10.00 8112 Mineral and stone processing plant operators - 10.00 5112 Transport conductors - 10.00 3214 Medical and dental prosthetic technicians - 10.00 8155 Fur and leather preparing machine operators - 9.09 7542 Shotfirers and blasters - 9.09 7313 Jeweler and precious-metal workers - 9.09 7312 Musical instrument makers and tuners - 9.09 7533 Sewing, embroidery and related workers - 8.33 2261 Dentists - 8.33 8152 Weaving and knitting machine operators - 7.69 Notes: the occupations not listed in this table have zero GTI. the full list of 4-digit occupations with zero GTI is included in the toolkit 48 Appendix F. O*NET Occupations that are green according to our methodology but not to O*NET GEP O*NET-SOC Occupational title Rank GTI Total Narrow O*NET GEP Code GTI Narrow tasks green classification Narrow tasks 19-2041.00 Environmental Scientists and Specialists, Including 27 68.18 22 15 GID Health 19-2043.00 Hydrologists 48 48.00 25 12 GID 19-1023.00 Zoologists and Wildlife Biologists 52 42.86 14 6 GID 33-3031.00 Fish and Game Wardens 55 41.67 24 10 GID 29-9011.00 Occupational Health and Safety Specialists 107 20.00 20 4 GID 45-4011.00 Forest and Conservation Workers 136 14.29 21 3 GID 19-4093.00 Forest and Conservation Technicians 135 14.29 21 3 GID 47-2131.00 Insulation Workers, Floor, Ceiling, and Wall 154 10.00 10 1 GID 17-2111.01 Industrial Safety and Health Engineers 158 9.52 21 2 GID 51-8021.00 Stationary Engineers and Boiler Operators 174 8.00 25 2 GID 17-2041.00 Chemical Engineers 175 7.69 13 1 GID 11-9121.00 Natural Sciences Managers 194 6.25 16 1 GID 49-9021.02 Refrigeration Mechanics and Installers 221 4.76 21 1 GID 49-9051.00 Electrical Power-Line Installers and Repairers 228 4.35 23 1 GID 45-1011.07 First-Line Supervisors of Agricultural Crop and 231 4.17 24 1 GID Horticultural Workers 47-2073.00 Operating Engineers and Other Construction 244 3.45 29 1 GID Equipment Operators 51-4121.06 Welders, Cutters, and Welder Fitters 251 2.50 40 1 GID 17-2111.02 Fire-Prevention and Protection Engineers 35 61.54 13 8 Non-green 19-1031.02 Range Managers 45 50.00 16 8 Non-green 13-1041.01 Environmental Compliance Inspectors 57 38.46 26 10 Non-green 47-2132.00 Insulation Workers, Mechanical 58 38.46 13 5 Non-green 51-8031.00 Water and Wastewater Treatment Plant and 59 37.50 8 3 Non-green System Operators 19-4051.02 Nuclear Monitoring Technicians 61 36.84 19 7 Non-green 17-2021.00 Agricultural Engineers 73 28.57 14 4 Non-green 33-2022.00 Forest Fire Inspectors and Prevention Specialists 75 28.57 14 4 Non-green 19-1020.01 Biologists 81 27.27 22 6 Non-green 33-1021.02 Forest Fire Fighting and Prevention Supervisors 82 26.92 26 7 Non-green 33-2021.01 Fire Inspectors 87 25.00 24 6 Non-green 19-1032.00 Foresters 89 24.00 25 6 Non-green 19-1022.00 Microbiologists 98 21.43 14 3 Non-green 11-9013.03 Aquacultural Managers 99 21.05 19 4 Non-green 29-1069.09 Preventive Medicine Physicians 106 20.00 15 3 Non-green 19-2012.00 Physicists 105 20.00 15 3 Non-green 51-2021.00 Coil Winders, Tapers, and Finishers 116 18.18 11 2 Non-green 33-1021.01 Municipal Fire Fighting and Prevention Supervisors 118 17.86 28 5 Non-green 17-2121.01 Marine Engineers 120 17.39 23 4 Non-green 49 33-2011.02 Forest Firefighters 122 17.39 23 4 Non-green 19-3092.00 Geographers 127 16.67 12 2 Non-green 17-2151.00 Mining and Geological Engineers, Including Mining 124 16.67 18 3 Non-green Safety Engineers 19-1012.00 Food Scientists and Technologists 129 15.38 13 2 Non-green 49-3091.00 Bicycle Repairers 142 12.50 16 2 Non-green 33-2011.01 Municipal Firefighters 150 11.11 27 3 Non-green 19-1011.00 Animal Scientists 149 11.11 9 1 Non-green 39-7011.00 Tour Guides and Escorts 152 10.53 19 2 Non-green 31-9099.01 Speech-Language Pathology Assistants 160 9.09 11 1 Non-green 43-5111.00 Weighers, Measurers, Checkers, and Samplers, 163 9.09 22 2 Non-green Recordkeeping 11-9161.00 Emergency Management Directors 164 8.70 23 2 Non-green 53-5021.01 Ship and Boat Captains 172 8.33 24 2 Non-green 19-3011.00 Economists 170 8.33 12 1 Non-green 31-1015.00 Orderlies 171 8.33 24 2 Non-green 49-9092.00 Commercial Divers 173 8.00 25 2 Non-green 49-9097.00 Signal and Track Switch Repairers 179 7.69 13 1 Non-green 25-4013.00 Museum Technicians and Conservators 178 7.69 26 2 Non-green 23-1021.00 Administrative Law Judges, Adjudicators, and 177 7.69 13 1 Non-green Hearing Officers 19-1042.00 Medical Scientists, Except Epidemiologists 176 7.69 13 1 Non-green 11-3011.00 Administrative Services Managers 180 7.14 14 1 Non-green 53-7072.00 Pump Operators, Except Wellhead Pumpers 186 7.14 14 1 Non-green 47-2151.00 Pipelayers 184 7.14 14 1 Non-green 33-2021.02 Fire Investigators 182 7.14 14 1 Non-green 33-3052.00 Transit and Railroad Police 183 7.14 14 1 Non-green 19-3091.01 Anthropologists 181 7.14 28 2 Non-green 47-5061.00 Roof Bolters, Mining 185 7.14 14 1 Non-green 47-5021.00 Earth Drillers, Except Oil and Gas 192 6.67 30 2 Non-green 27-1025.00 Interior Designers 190 6.67 15 1 Non-green 25-4012.00 Curators 189 6.67 15 1 Non-green 47-2161.00 Plasterers and Stucco Masons 191 6.67 15 1 Non-green 17-3011.02 Civil Drafters 188 6.67 15 1 Non-green 49-2095.00 Electrical and Electronics Repairers, Powerhouse, 196 6.25 16 1 Non-green Substation, and Relay 49-2098.00 Security and Fire Alarm Systems Installers 197 6.25 16 1 Non-green 51-4052.00 Pourers and Casters, Metal 198 6.25 16 1 Non-green 19-3091.02 Archeologists 195 6.25 16 1 Non-green 53-5031.00 Ship Engineers 202 5.88 17 1 Non-green 39-7012.00 Travel Guides 201 5.88 17 1 Non-green 19-1031.03 Park Naturalists 203 5.56 18 1 Non-green 29-1199.01 Acupuncturists 205 5.56 18 1 Non-green 19-4021.00 Biological Technicians 204 5.56 18 1 Non-green 17-2031.00 Biomedical Engineers 206 5.26 19 1 Non-green 50 41-9021.00 Real Estate Brokers 209 5.26 19 1 Non-green 29-2054.00 Respiratory Therapy Technicians 208 5.26 19 1 Non-green 47-2141.00 Painters, Construction and Maintenance 210 5.26 19 1 Non-green 49-2092.00 Electric Motor, Power Tool, and Related Repairers 212 5.13 39 2 Non-green 11-9013.01 Nursery and Greenhouse Managers 213 5.00 20 1 Non-green 49-3043.00 Rail Car Repairers 215 5.00 20 1 Non-green 49-9052.00 Telecommunications Line Installers and Repairers 216 5.00 20 1 Non-green 39-2021.00 Nonfarm Animal Caretakers 220 4.76 21 1 Non-green 19-1029.02 Molecular and Cellular Biologists 217 4.76 21 1 Non-green 39-1011.00 Gaming Supervisors 219 4.76 21 1 Non-green 19-3093.00 Historians 218 4.76 21 1 Non-green 29-1126.00 Respiratory Therapists 224 4.55 22 1 Non-green 23-1011.00 Lawyers 223 4.55 22 1 Non-green 45-2092.01 Nursery Workers 226 4.35 23 1 Non-green 45-3021.00 Hunters and Trappers 227 4.35 23 1 Non-green 17-2171.00 Petroleum Engineers 225 4.35 23 1 Non-green 47-3013.00 Helpers--Electricians 232 4.17 24 1 Non-green 25-1053.00 Environmental Science Teachers, Postsecondary 230 4.17 24 1 Non-green 53-2022.00 Airfield Operations Specialists 233 4.17 24 1 Non-green 51-9122.00 Painters, Transportation Equipment 235 4.00 25 1 Non-green 25-1064.00 Geography Teachers, Postsecondary 234 4.00 25 1 Non-green 37-3013.00 Tree Trimmers and Pruners 240 3.85 26 1 Non-green 29-1069.05 Nuclear Medicine Physicians 239 3.85 26 1 Non-green 25-1051.00 Atmospheric, Earth, Marine, and Space Sciences 237 3.85 26 1 Non-green Teachers, Postsecondary 19-4099.01 Quality Control Analysts 236 3.85 26 1 Non-green 47-2081.00 Drywall and Ceiling Tile Installers 241 3.85 26 1 Non-green 25-1052.00 Chemistry Teachers, Postsecondary 238 3.85 26 1 Non-green 11-9141.00 Property, Real Estate, and Community Association 242 3.70 27 1 Non-green Managers 29-2011.01 Cytogenetic Technologists 245 3.33 30 1 Non-green 45-3011.00 Fishers and Related Fishing Workers 246 3.33 30 1 Non-green 49-9031.00 Home Appliance Repairers 247 3.23 31 1 Non-green 51-4072.00 Molding, Coremaking, and Casting Machine 248 3.03 33 1 Non-green Setters, Operators, and Tenders, Metal and Plastic 25-9041.00 Teacher Assistants 249 2.94 34 1 Non-green 49-9012.00 Control and Valve Installers and Repairers, Except 250 2.56 39 1 Non-green Mechanical Door 51-9071.01 Jewelers 252 2.50 40 1 Non-green 51