Climate policy accelerates structural changes in energy employment

The employment implications of decarbonizing the energy sector have received far less attention than the technology dimension of the transition, although being of critical importance to policymakers. In this work, we adapt a methodology based on employment factors to project future changes in quantity and composition of direct energy supply jobs for two scenarios (1) relatively weak emissions reductions as pledged in the nationally determined contributions (NDC) and (2) stringent reductions compatible with the 1.5 C target. We find that in the near-term the 1.5C-compatible scenario results in a net increase in jobs through gains in solar and wind jobs in construction, installation, and manufacturing, despite significant losses in coal fuel supply; eventually leading to a peak in total direct energy jobs in 2025. In the long run, improvements in labour productivity lead to a decrease of total direct energy employment compared to today, however, total jobs are still higher in a 1.5 C than in an NDC scenario. Operation and maintenance jobs dominate future jobs, replacing fuel supply jobs. The results point to the need for active policies aimed at retraining, both inside and outside the renewable energy sector, to complement climate policies within the concept of a “just transition”.


Introduction
Reduction of emissions to reach the goals of the Paris Agreement will require a drastic energy transition-not only replacing fossil fuels by renewables for power generation, but higher end-use electrification, adoption of other low-carbon fuels, greater energy efficiency, and behavioural change (Dubois et al., 2019;Luderer et al., 2018;Weber, 2015).
The employment implications of decarbonizing the global energy sector system are of critical importance for the political salience of global mitigation pathways towards the goals of the Paris Agreement. Employment in the energy sector represents a tiny fraction (~1.2%) of total global employment. In 2019, against a total world employment of around 3.3 billion (15+ age) (ILO, 2020) there were only 40 million total world energy jobs (including production, distribution, and transportation) (IEA, 2020). Despite the insignificant overall share, energy sector jobs, especially on the supply side are directly and visibly linked to energy policy, are a source of indirect job creation, and important revenue for states and sub-regions. Thus, their consideration is critical to the speed and direction of energy transition.
An energy transition will lead to a change in the number, structure, and required skill of jobs in the energy sector. To be sure, such a 'conscious', policy-induced transition will be superimposed to an autonomous long-term trend towards more service-based economies and increasing endogenous labour productivity (often accompanied by shift in factors of production from labour to capital). Whatever the type of transition, two things are cleari) job creation and employment will continue to be a major political force affecting political decisions at all administrative levels, and ii) unlike a "natural transition", there will be higher resistance to a "conscious transition", especially because the long-history of incumbent technologies has resulted in strong political affiliations and lobbying power of relevant stakeholders, both regionally and nationally (Caldecott et al., 2016;Spencer et al., 2018). In some cases, such as coal, a which have a long history in certain regions, the loss would not only be that of employment but whole culturesfestivals, language etc. The emerging field of "just transition", defined as "a process by which economies that progress towards a green economy also strengthen each of the four pillars of decent work for all (i.e. social dialogue, social protection, rights at work and employment)" (ILO, 2018) has partly arisen to ease the opposition from people/groups, whose jobs which will be lost or at risk of being lost due to an energy transition and to dilute the climate-specific focus of energy policies to include broader societal goals as put down in the Sustainable Development Goals SDGs, thereby increasing political support (McCauley and Heffron, 2018).
To contribute to both the ongoing effort on societal implications of energy transitions and political feasibility of scenarios (in the context of employment) requires first and foremost, a technology-specific quantification of energy sector jobs under alternative policy pathways. Thus, the first objective of the paper is to bring forth a transparent and indepth methodology of estimating employment, including identifying its most important drivers. The second objective is to find differences in near-term and long-term structure of (global) jobs in the energy supply sector. Lastly, using this information, we seek to identify if employment transitions in the near-term could hinder decarbonisation.
The current study provides a method to account for employment effects of global climate policy scenarios, which although widely used in international scientific and policy communities (e.g., IPCC), hardly provide employment impacts. Previous studies using this method, include energy models (Dominish et al., 2019;Ram et al., 2019Ram et al., , 2022 and global integrated assessment models (IAM) (Pai et al., 2021). We go beyond these studies by discussing in detail the uncertain but crucial determinants for energy employment.

Overview of methodologies for estimating energy employment
Approaches to measure employment effects from renewable deployment or energy policies can broadly be broken down into two typesi) those using input-output (IO) or computable general equilibrium models (CGE) of the economy, and ii) those relying on analytical approaches (Kammen et al., 2004). The former includes flow of goods and services between different sectors of the economy, i.e., everything produced either serves as an input to the next level of production or an end-use purpose (IRENA, 2014). This allows finding the macro-economic impacts, including employment, of various energy and climate policies (Lambert and Silva, 2012). However, their coarse sectoral coverage prevents detailed breakdown of jobs by technology and/or fuel (for studies using the GEM-E3 CGE see, for e.g., Vandyck et al., 2016;Vrontisi et al., 2020).
The second approach and more relevant for this paper involves calculation of job intensities or employment factors (EF), defined as the number of jobs resulting from a unit investment or unit production of a physical commodity. When combined with energy transformation pathways they yield gross employment (only direct jobs) in that sector, although multipliers have been used to extend the scope of the approach to include indirect jobs (IRENA, 2014). A schematic of the employment factor approach is shown in Fig. 1. Direct jobs in the energy supply sector are broken down into stages or activities commonly associated with the supply chain or life cycle of a fuel/technologymanufacturing, construction and installation (C & I), operation and maintenance (O & M), fuel supply or production. Each activity and technology require a separate employment factor and when estimating jobs globally, also country-specific factors. Since manufacturing and C & I jobs are only created during the capacity addition, they are multiplied to the added capacity, O & M to the existing capacity, and fuel supply factors to the fuel production. The distinction of jobs in the value chain also helps to distinguish between the temporal (short-term manufacturing and C & I jobs), spatial (export-oriented manufacturing jobs vs. regional C & I and O & M jobs), and to some extent worker-skills characteristic of each technology.

Estimation of employment factors
An important pre-requisite for the calculation of jobs in this study is the employment factor (EF). EFs have been reported in literature either through I/O models, industry surveys, or back-calculation based on employment and capacity figures in a particular year (Cameron and van der Zwaan, 2015). Several studies have aggregated and analysed these EFs, providing important insightsi) Renewables EFs are reported more often than conventional technologies (Cameron and van der Zwaan, 2015;IRENA, 2014;Lambert and Silva, 2012), ii) Most studies are for/from OECD countries (Cameron and van der Zwaan, 2015;Rutovitz et al., 2015), iii) EFs for RE technologies are much higher than conventional technologies (measured in MW or MWh) (del Río and Burguillo, 2008), and solar PV C & I + manufacturing EFs are higher than corresponding wind EFs (Cameron and van der Zwaan, 2015), iv) Large variation exists in EFs for similar technologies (Breitschopf et al., 2012;Cameron and van der Zwaan, 2015).The large variation in turn exists because of unclear boundaries between-direct and indirect jobs and the various activities in the supply chain; local and export/import component of jobs; specific country-contexts; and methodology (for e.g., not considering or reporting effects of economies of scale) (Cameron and van der Zwaan, 2015;IRENA, 2014). (Llera et al., 2013) further note that even with a consistent methodology for data collection, differences could still arise from maturity of the industry or availability of skilled workers.

Estimation of jobs using the EF approach
The EF approach has been used by many studies to calculate existing or future jobs in the energy sector, although most of them have been limited to regional or sub-regional scales and included only renewable technologies (Kammen et al., 2004;Stavropoulos and Burger, 2020). On the other hand, Rutovitz et al. have used the EF approach to calculate regionally differentiated global energy supply job estimates, divided by technology and type (Rutovitz et al., 2015;Rutovitz and Atherton, 2009;Rutovitz and Harris, 2012). Their methodological updates, published every three years, have updated the employment factors (as more data is discovered and/or made published) and expanded sectoral scope of jobs, e.g., from heat supply. Additions have also been made by (Ram et al., 2019(Ram et al., , 2022 notably the inclusion of decommissioning, storage-battery, and power-to-X jobs, although these have not been used in this study due to very few (empirical) data points.

Approach and employment factors used in the study
The following sub-sections provide an in-depth review of methodology of Rutovitz et al. (2015) (the first paragraph) and is accompanied by changes and additions undertaken in the current work (second paragraph). These sub-sections coverthe source of employment factors (Calculation of employment factors), assumptions on how EFs evolve into the future (Evolution of employment factors), inclusion of trade of fuels to differentiate import and export components of jobs (Trade), differentiating manufacturing to account for uneven distribution of manufactured components (Manufacturing jobs), the accounting for technologies not included but important in energy employment (Share of sub-technologies), and the comparison of the resulting employment numbers with other literature (Comparison with other sources). Additional information, wherever required, is provided in the Supplementary Information (SI) and indicated in the paragraphs. The limitations of the study for e.g., the limited sectors where employment is estimated is provided in section 3.4.

Employment factors
As mentioned in the section before, the majority of EFs are only available for OECD countries and where available, often show a spread of values. To tackle these two issues Rutovitz and Atherton (2009) i) calculated EF per technology for OECD countries by a weighted average or average spanning the country-specific studies ii) calculated the EF for non-OECD countries by assuming a regional adjustment factor. For the base year (2015) this factor is the labour productivity (measured as GDP (or value added) per worker, of the whole economy excluding agriculture) for different nations relative to OECD. The adjustment factors are assumed to be the same across the different activities and technologies. iii) For a few studies that report EFs for non-OECD countries (for specific technologies), values from step ii) are replaced.
All EFs from Rutovitz et al. (2015) for OECD countries, are taken as values for the year 2020 in the current study but updated according to recent literature (Fragkos and Paroussos, 2018;IRENA, 2017aIRENA, , 2017b. Following the methodology of Rutovitz et al. (2015), for countries without empirical data (mostly non-OECD), the EF is calculated by multiplying the OECD EF from Rutovitz et al. (2015) to a regional adjustment factor. Next, wherever possible, the resulting EFs have been replaced by country-specific values using studies mentioned in Rutovitz et al. (2015), recent studies (CEEW, 2019; Rutovitz et al., 2020;The Solar Foundation, 2020), or own calculations, e.g., coal EF. Lastly, some EFs are modified for specific countries/technologies by comparing the resulting jobs from the EF approach with bottom-up regional and global studies providing job estimates. See SI section 1.1 for all these details.

Evolution of employment factors
To estimate jobs into the future, the employment factors need to be projected into the future. They calculate this by considering two developmentsimprovements in LP and decline factors.
a. Improvement in labour productivity -EF for all countries into the future evolve with the improvements in labour productivity and are assumed to be equal to (inverse of) future GDP per capita (relative to OECD). For OECD countries, the factor is 1. The data on GDP per capita comes from the energy model used in the study. b. Decline factorsaccount for the reduction in EF as technologies mature. They are assumed to evolve with the changing capital costs of technologies, which is an input to the energy model. Decline factors are undifferentiated across activities, except for coal fuel supply. No decline factors are assumed for oil, gas, and nuclear fuel supply (Rutovitz et al., 2015, sec. 6).
To better understand the dynamics of the system in the future, the method is slightly simplified. The employment factors calculated IEA, 2020 are subjected to -i) (inverse of) future GDP per capita for all countries (relative to 2020), which is used as a proxy for improvements in labour productivity (SI section 2.3) and ii) capital costs of technologies relative to 2020 (SI section 2.2). Data for both comes from REMIND, except the improvements in labour productivity for coal fuel supply which are exogenous (see SI Section 1.1.4).

Trade
Since the import of fuels (coal, gas, and biomass) does not lead to creation of fuel supply jobs in the consuming country, it is important to differentiate between the amounts of fuel produced in the country vs. amount of fuel exported/imported. Rutovitz et al. (2015) therefore make these assumptions for each region and time step.
Trade of coal, gas, oil, biomass, and nuclear 1 is endogenous to REMIND. This means that production/import/export of a fuel are readily available as outputs of the model. The fuel supply jobs (per region, fuel, and time step) are calculated by multiplying the employment factor with the amount of fuel produced in the country.

Manufacturing jobs
As for fuels, the manufacturing of components required for each technology are unevenly distributed in the world and need to be differentiated. For each region and time step, the proportion of local manufacturing and share of import from all other regions is assumed. It is also assumed that countries become self-sufficient over time. The same shares are applied for wind, solar PV, solar thermal power, geothermal power, and ocean (wave and tidal) technologies. All manufacturing for fossil fuel, biomass, hydro and nuclear technologies occurs within the region.
Instead of appropriating local vs. import shares for each region and time step, the current study assigns the share of total world production/ manufacture to each region, although only for solar PV and wind. All other technologies are assumed to be manufactured domestically/ regionally. These shares evolve such that those regions manufacture their own share of technology deployment locally by 2050. This assumption reflects that countries will promote domestic manufacturing to create jobs locally and for reasons of energy security; at the same time 1 Employment factor for nuclear fuel supply is based on the secondary energy of nuclear-based electricity. Furthermore, no trade in uranium is assumed, i.e., all extraction and processing jobs are created place within the consuming regions.
income convergence assumed in the SSP2 socio-economic scenario underlying our results, and spill-over effects and diffusion of technological know-how will make manufacturing available widely. The methodology is also flexible to consider fixed manufacturing shares at current levels or other assumptions for the exploration of alternative socio-economic futures.

Share of sub-technologies
The energy model used by Rutovitz et al. (2015) includes sub-technologies or variants of traditional RE technologies -wind onshore and offshore; small and large hydro. Solar PV is however not differentiated into rooftop and utility.
REMIND currently includes a generic technology representation each for solar PV, solar CSP, wind, and hydro as power-generating technologies, i.e., it does not differentiate between solar PV utility and Solar PV rooftop, wind offshore and onshore, and small and big hydro. For the parametrization of costs and potentials, mostly the cheaper variants (utility scale PV, onshore wind and large hydro) have been considered, but some adjustments are done to account for additional potentials (e.g., of rooftop solar in densely populated countries like Japan and India). When the only consideration is cost, only larger and cheaper variants of the technology would be installed. In reality, different constraints, e.g., land, political feasibility, energy security etc. make it impractical to exclusively deploy the dominant variants (for e.g., Germany has 60% of its installed solar capacity as rooftop). Nevertheless, the share (in terms of installed capacity) of the alternative more expensive variants for most countries remains small. When estimating energy-related jobs, these sub-technology differentiations can play an important role because the more expensive variants tend to have higher employment factors (depending on the technology and activity) (CEEW, 2019; Rutovitz et al., 2015;The Solar Foundation, 2020).
To capture this effect to some extent, an external share controls how much of the additional and existing installation from REMIND (for solar PV, wind, and hydro), is supposed to be of the different sub-technology variants. A detailed explanation of this assumption is provided in SI Section 4.1.
Comparison with other sources. Due to the different methods and boundaries (for e.g., between direct and indirect jobs) of measuring jobs, there is no 1:1 comparison between jobs estimates from this study with the previous literature. However, comparisons can still be useful to get an indication if the numbers from this study make sense and assess the relative confidence of estimates for different technologies/regions. Such a preliminary comparison reveals that REMIND job estimates are well consistent to other national and global estimates (see SI section 1.2 for more a detailed comparison).

Scenario setup
The global integrated assessment model REMIND in its version 2.1 (Baumstark et al., 2021)was used to run two policy scenarios "NDC" and "1.5C" (described in Table 1). The evolution of employment factors into the future was explored by building EF-scenarios (ex-post). The eventual EF scenario selected included both capital costs and improvements in labour productivity driving the results. See SI section 2.1 on the process and explanation.

Future energy sector jobs
IEA, 2020, the total direct energy supply jobs in the world are around 20.4 million, with a roughly equal proportion between fossil and non-fossil jobs (Fig. 2a). These jobs are dominated by coal (4.3 million) followed by Hydro, solar PV, and Gas (~3 million each) (Fig. 2b and c). Fig. 2c shows that most of the current coal jobs are in fuel supply (purple) whereas solar PV jobs are almost equally split between construction and installation (C & I) (green), and manufacturing (rust yellow). Furthermore, fuel supply constitutes the largest share (~50%) of the current energy supply jobs (see also Fig. S19, which is directly expressed as shares).
In 2050, the total energy supply jobs decrease to 16 million under the NDC scenario and to 18.2 million with 1.5 • C policies (Fig. 2d). Independent of the policy scenario, the share of fuel supply jobs (purple) strongly decreasesfrom 50% IEA, 2020 (mainly in coal, biomass, oil, and gas) to 27% in NDC and 24% (mainly in biomass and oil) in the 1.5 • C scenario. Also, in absolute terms, the fossil fuel supply jobs decrease strongly in both scenarios, although the absolute fossil fuel supply differs by a factor of 2 in 2050 (see Fig. S17b). On the other hand. the share of fossil jobs decreases to 12-25% of total jobs, depending upon the scenario. In both NDC and 1.5 • C policy case, wind jobs dominate in 2050, and operation and maintenance (O & M) becomes the activity employing the most people ( Fig. 2b and Fig. S19). In both scenarios, there is a steep increase in power generation and shift from fossils to non-fossils (mainly wind and solar), which accompanied by wide-scale end-use electrification reduces the need for conventional fuels (Fig. S17). Although renewable technologies require more jobs per MW, 2 their exponential uptake is also accompanied by steeply decreasing employment factors (due to decreasing capital costs and improvements in labour productivity) (Fig. S16). This eventually leads to lesser people employed directly in the energy supply sector than now. 3 A good contrast on how employment factors eventually influence the shape of the curve is between wind and solar PV. Although installations for both these technologies increase steeply in the future (with solar growths higher than wind) (see Fig. S17), EFs for solar also have a sharper decline (Fig. S16). The net effect is that while jobs in wind increase almost linearly over time and become the largest employer in 2050 (Fig. 2b), solar PV jobs might be prone to periods of boom, bust, and eventual stagnation (Fig. 2b).

Near-term jobs
In the near term, there is a net increase in jobs for a 1.5 scenario (+838,500), compared to a net decrease in the NDC scenario Immediate introduction of regionally differentiated carbon prices which converge in 2050, iteratively adjusted to fulfil a constraint in carbon budget (900 GtCO2) from 2011 to time of net-zero global CO2 emissions. Carbon prices after reaching net-zero increase moderately, leading to moderate net-negative emissions thereafter, and a 66% chance of limiting temperature increase below 1.5 • C at the end of century (2100).
2 Comparison across per technology per MW fails to capture fuel supply jobs for fossil technologies. The correct unit of measurement to compare employment across technologies should be per GWh. However, the point here is the rapid decrease in employment factors of VRE technologies in comparison to traditional technologies. 3 A big caveat here is that jobs which might become significant in the futurestorage, transmission and distribution, decommissioning, hydrogen production, and all jobs on the demand side, including energy efficiency have not been included due to data limitations. See Section 3.4.
(− 760,000) (in 2030 compared IEA, 2020) (Fig. 3a). Losses in the fossil sector are mainly from coal fuel supply (85-90%) and amount to 2.2-3.7 million (depending on the scenario), while gains are mainly from C & I and Manufacturing activities in solar and wind (~90%) and amount to 2.5-4.3 million (Fig. 3b). The competing effect of capacity/production and employment factor on jobs for different technologies and scenarios can be seen in Fig. 3c. The black diagonal line represents the case where change in jobs is entirely due to change in capacity/production. The further the dot is from the linear line, the stronger the effect of the employment factor on total jobs. For e.g., while capacity increases almost 700% for solar PV in the 1.5C case, jobs increase only 94%. For coal mining in NDC while production decreases 12%, jobs decrease 51%. Comparing this with other fossil fuels (still for the NDC), we see that an increase of 9% of oil production, leads to 15% decrease in jobs, while for gas a 17% increase in production, leads to a 7% increase in jobs implying a lower effect of labour productivity improvements. 4

Implications of the results
In Fig. 2a we showed that even with a significant expansion in renewables, energy supply jobs, after peaking around 2025, decrease because of the strong effect of increasing labour productivity and decrease in capital costs. 5 This result is not surprising. Under the assumption of decreasing capital costs, increasing wages and constant share of labour in total capital costs, employment factors must fall faster than the capital costs.
The employment-related challenges in the oncoming energy transition for both the NDC and policy cases, can be understood from Fig. 3b. In NDC, almost 2.7 million jobs are lost in fuel supply, compared to 2.5 million gained through solar PV and wind. Thus, even under an ideal assumption that all coal workers manage could be retrained to be employed in these new jobs, these will not be enough. How many of the new jobs go eventually to people who lost them depends on many factors skill requirements and location of created and lost jobs, options for retraining and relocating, incentives and compensation. Additionally, in the case of manufacturing, not all jobs might be created locally (at least in the near term), particularly for late entrant countries that fall back in technology development. On the other side, countries currently mostly relying on fuel imports and therefore sustaining fuel supply jobs abroad can potentially increase their domestic energy-related jobs by transitioning to local renewable energy forms.
For a 1.5 scenario, the challenges already present in the NDC scenario are somewhat exacerbated. Although there is a net increase in jobs, an extra 1.3 million jobs in fuel supply will be lost over the same period. Since most of these lost jobs would be regionally concentrated, this can result in additional challenges for policy. To alleviate the related political economy concerns will require faster and more intensive engagement between firms, employees, state government and local administration. This might not be easy considering thati) areas of and around coal mines might not be prime areas for renewable energy development (Pai et al., 2020) and thus might not immediately gain from increasing RE deployment, ii) unlike C & I and O & M jobs, manufacturing jobs are often not local, and thus wouldn't necessarily contribute to job growth, iii) Subsequent increases in the same amount of RE capacity will employ lesser people and could lead to conditions of boom and bust.
In either case, the first step in the direction of reducing challenges and increasing support of the energy transition would be constituting regional and sub-regional studies on what, how, and how much of new RE technologies could (either theoretically or economically) substitute job losses in the fuel supply (mainly coal) at a sub-regional level (see for example Alves Dias et al., 2018;Pai et al., 2020), and then progressively move to options directly outside the energy sector. These should pay attention to the fact that, i) C & I jobs could be considered as long-term jobs (and thus on par with O & M and fuel supply jobs), under increasing RE capacity, ii) sub-technologies like solar-rooftop, small-hydro often employ much more people than large-scale solar-utility and hydropower projects.
Over decades, the results show various promising aspects of largescale RE deployment. Firstly, fuel supply jobs in fossil are progressively replaced by O & M jobs, both of which offer job stability (see Fig. S19 and Fig. S20 for share of activity until 2050). Secondly, unlike the present day where majority of supply jobs are concentrated in coal, oil, and gas-producing countries; in a RE-based energy system jobs would be more evenly spread across the whole world, with the possible Fig. 3. Difference in Global jobs (2030-2020) for (a) different technologies (with cross denoting the net gain/loss) and (b) technologies and activities, and (c) relative percentage change in jobs vs. relative percentage change in capacity (GW)/production (EJ) (in 2030 relative to 2020) for the two policy scenarios. 5 Both these factors are essentially improvements in labour productivity, with the former mostly occurring outside the energy sector (e.g., artificial intelligence, drones etc.), while the latter occur mostly within the sector. exception for manufacturing of components, though increasing transport costs in decarbonized futures and a focus to create jobs locally might incentivize stronger local production. Thirdly, within a country, energy supply will be distributed both across remote parts containing utility-scale solar and wind farms but also in cities as solar rooftop, though the relative importance of these options has implications for costs, grid requirements and broader sustainability considerations, and largely depends on policy settings.

Limitations and future research
Our result must be read with the important caveat that we only include direct energy jobs from the currently existing supply technologies. We thus do not include many other energy-related jobs on both the supply and demand sides, which could become significant into the future. These include Transmission and Distribution (T & D), Battery storage, Decentralised PV, Heating (solar thermal, heat pumps etc.), hydrogen production, and energy efficiency. Previous studies have shown that significant investment would need to go to these sectors/ technologies/fuels McCollum et al., 2018), thus also highlighting their importance. Employment factors for some of these have been provided by Ram et al. (2019Ram et al. ( , 2022, however given that they are based on a few empirical studies and/or are immature technologies, their values are highly uncertain and have not been used here. Furthermore, given the specific scope of our methodology, we are unable to comment and quantify how mitigation policies would influence job numbers and structure outside (direct) energy supply, for e.g., in the automotive or chemical industry sector.
The employment factor approach relies on accurate estimation of employment factors for a technology. Moreover, an estimation of global energy supply jobs requires such values for major countries around the world. As mentioned previously, although most of the energy supply jobs exist in non-OECD countries, employment factor studies are mostly available for OECD countries. Thus, besides the need of studies calculating employment factors for both conventional and new technologies, the spatial scope needs to cover more non-OECD countries.

Conclusions and policy implications
Our estimation of employment in the energy supply sector was based on the employment factor approach, whose different assumptions were explored before. Using this approach, we quantified direct jobs in the energy supply sector for two scenarios -NDC (weak climate mitigation) and 1.5C (strong climate mitigation). We showed that for both policy and NDC scenarios the direct energy supply jobs decrease in the future compared to 2020, however ambitious policy jobs are higher than the latter. Secondly, the increase in cumulative solar and wind capacity, against the decrease in total fuel production means that the O & M jobs overtake fuel supply as the major share of total jobs. Lastly, in the nearterm, net gains are seen only in the 1.5C policy case, however, lead to considerable losses in the coal mining sector. This exposes the trade-off of ambitious climate policyboth of increasing job losses and gains, and eventually the dichotomy of political supportin the form of winners and losers. To align both towards a strong climate ambition will require that (people/regions/firms) currently working on the fossil side are made available the opportunities in the new RE energy world or compensated through other means.

Data availability
The input data, including the code to produce the figures, both in the main text and SI, is available here -https://gitlab.pik-potsdam.de/amali k/energy-employment. The scenarios were prepared using the opensource integrated assessment model REMIND (https://github.com/rem indmodel/remind and https://zenodo.org/record/4091409).