Compensating affected parties necessary for rapid coal phase-out but expensive if

Identifying countries with coal phase-out pledges and compensation policies

We build a database of national coal phase-out pledges and compensation policies combining systematic document review, web searches and expert consultations (Supplementary Fig. 1).

First, we identify all countries with a time-bound coal phase-out pledge and installed coal capacity at the time of making the pledge (Supplementary Table 3)²³. This database was built through a systematic review of national and international documents including National Energy and Climate Plans (NECPs); National Recovery and Resilience Plans (NRRPs); Nationally Determined Contributions (NDCs); the Powering Past Coal Alliance (PPCA); the Global Coal to Clean Power Statement (GCCP); and other policy documents. It covers all explicit national coal phase-out pledges, but does not include coal phase-out implied by national net-zero or other climate targets since such plans may or may not feature coal phase-out. It also excludes countries which have joined the PPCA, but where there is no official date associated with the coal phase-out pledge (the United States⁵⁸, Kosovo⁵⁹, and Mexico⁶⁰).

Second, we identify all compensation policies in these countries using a systematic Google search and the terms “coal phase-out”, “coal”, “just transition” and “coal compensation”. We identified 23 countries that have both coal phase-out pledges and related compensation policies. To confirm our case identification, we consulted experts in two surveys and three workshops. The first survey was conducted in September 2021 with a selection of 15 coal phase-out experts. The second was conducted in January/February 2022 and distributed to the same 15 experts as well as via Twitter (now called X). We received 14 expert responses. In both versions of the survey, we presented respondents with our criteria for case selection (explicit coal phase-out pledge and a compensation policy) and our initial set of cases, and asked them two questions: (1) Are you aware of any governments not included in the list above that are planning to phase out coal and compensate affected actors? (2) Are you aware of any other governments that compensate affected actors of coal sector declines without a deliberate coal phase-out policy?

We asked the same two questions to attendees at three online workshops on fossil fuel decline – two associated with the CINTRAN project⁶¹ and one associated with the Contractions project⁶². Through the surveys and expert consultations, we found that while Poland has not finalized its law, the country has plans for compensation⁶³. The study was exempt from ethical review and approval subject to Swedish higher education regulation, because no sensitive personal data were collected or processed as part of the surveys and the workshops.

We also include compensation for coal phase-out through the EU Just Transition Fund (JTF) and the Just Energy Transition Partnerships (JETPs), both of which were announced during our analysis. At the time of writing, three countries have JETPs connected to the phase-out or phase-down of coal power: South Africa⁶⁴, Indonesia²¹ and Vietnam²⁰ (Supplementary Note 1). Indonesia and Vietnam have coal phase-out pledges, but South Africa does not. We thus include the JETP for South Africa in our analysis (Table 1, Table 3), but cannot consider it as a case in our average compensation per avoided tCO₂ emissions or regression analysis.

To calculate the global share of countries with coal power and compensation policies, we used data on installed coal capacity by country from ref. ⁶⁵. We consider all countries with installed coal capacity >100 MW (77 countries in 2022). The global installed capacity in 2022 was 2084 GW.

Quantifying and mapping financial compensation for coal phase-out

We code each compensation policy for: the amount of compensation; the type of support; and the funding or budgetary source from which compensation is paid (Supplementary Table 1 and ref. ⁶⁶).

We only consider public finance to enable consistent comparison across cases and since private investment is likely to follow another logic.

In the majority of cases we rely on official governmental (laws, national budgets, strategies, plans and press releases) and international sources (the JETPs^20,21,67, the EU JTF Allocation^68,69, NRRPs^70,71,72, and EU case law^73,74) – Supplementary Fig. 1. We identify government sources by searching national and ministerial websites for each of our national cases with the search terms “coal phase-out”, “coal”, “just transition” and “compensation”. We use these terms in the English version of government websites, and where English versions are not available, we translate terms into the national language, and the search results into English using Google translate.

In four cases (Poland, Netherlands, Greece, and for Germany’s auction system) we found evidence of compensation in third-party sources but could not identify a government source (Supplementary Table 1), but in the case of Poland, could confirm the existence of the compensation policy through the expert consultation.

For the JETP agreements, we include South Africa, Indonesia, and Vietnam in our analysis but not Senegal since the Senegalese JETP does not refer to coal phase-out⁷⁵. The EU JTF includes support for regions in EU member states expected to be negatively affected by climate mitigation measures including coal phase-out⁶⁹; we use the Territorial Just Transition Plans (TJTPs) and other European Commission documentation to estimate how much of this funding is likely to be related to coal phase-out⁶⁹. We also include NRRPs because the Recovery and Resilience Facility supports not only economic recovery from the Covid 19 pandemic, but also low-carbon transitions⁷⁶. NRRPs explicitly describe what measures funding is intended for so we include amounts related to coal phase-out.

In 13 cases there is uncertainty about the amount of compensation either because we could not verify part of the compensation policy in government sources; in the case of the EU JTF it was not possible to identify how much funding relates to coal phase-out versus other climate change mitigation measures; and in the case of Vietnam and Indonesia the JETPs are only pledged for an initial period.

We capture these uncertainties with a lower and upper estimate for compensation (Table 1, Supplementary Table 2). The lower estimate includes: only funding mechanisms we could confirm in official government sources; for the EU JTF, an estimate of the likely share of overall funding based on nationally-specific documentation (Supplementary Table 3); and for JETP countries, a plausible lower estimate based on the JETP agreements and an earlier coal phase-out (Supplementary Notes 1 and 2). The upper estimate includes: funding mechanisms we could only confirm in third-party sources; for the EU JTF, the entire amount of compensation pledged or in the case of Bulgaria all JTF funding since at the time of writing, the country’s TJTP had not been approved (Supplementary Table 1); and for JETP countries, a plausible upper estimate based on the JETP agreements and a later coal phase-out (Supplementary Notes 1 and 2).We calculate a central estimate as the mean of the upper and lower estimates.

We excluded three specific funding mechanisms for which the situation has substantially changed since the compensation policy was announced. For Germany we excluded potential compensation to electricity consumers dependent on future electricity price changes due to the coal phase-out since the policy was formulated prior to the Russo-Ukrainian war and the resulting energy crisis; given the current situation it is difficult to estimate compensation in this context and highly uncertain whether it should be attributed to coal phase-out. In the Netherlands we excluded requests for compensation from two coal power plant owners since they have been struck down by the courts⁷⁷. We also excluded Ukraine from our analysis. While Ukraine declared coal phase-out in 2020 and specified compensation to coal companies in its 2022 budget⁷⁸, the start of the war in February 2022 has made implementation of these plans highly uncertain.

We could not quantify compensation for two countries: Chile pledges to compensate power companies based on a capacity mechanism but does not specify how this capacity mechanism will be calculated^79,80,81. North Macedonia’s NECP mentions funds for coal phase-out and a just transition but has not yet specified the amount or funding source⁸².

All compensation is reported in US Dollars (USD) 2020 using the exchange rate⁸³ and Gross Domestic Product (GDP) deflator⁸⁴ for the respective year and country. We used the year in which the compensation is numerated in, if specified, or the year in which the respective document was published.

We calculated the average for compensation per ton of avoided emissions using the total compensation divided by the sum of avoided emissions in all countries with compensation policies total. We replicated this calculation using both an optimistic and pessimistic interpretation of the coal phase-out pledges and the upper and lower estimates for compensation (Supplementary Table 10).

We identify the funding source and type of support implied by compensation policies based on the national and international documents described above and government budgets. We retrieve government budgets for the years of and following coal phase-out pledges (for example, France’s coal phase-out pledge was made in 2017 so we use government budgets from 2017-2021). We search government budgets for the term “coal” (or its equivalent in the national language, for example “charbon” in French) and code budget entries specifically related to coal phase-out. To ensure transparency over how we code compensation by the scope or type of support, we provide a “description of support” for each individual funding mechanism in Supplementary Table 1.

Calculating avoided emissions from coal phase-out

We follow the method developed in refs. ^23,26 to calculate avoided emissions resulting from national coal phase-out pledges⁸⁵, and for India and China under 1.5 °C-, 2 °C- and 2.5 °C-compatible pathways^1,86 from the Sixth Assessment Report of the Intergovernmental Panel on Climate Change (IPCC AR6), versus a reference scenario where coal plants are retired following the average historical national lifetimes. We calculate the average historical lifetime of coal power plants since 2001 in each country using the S&P database⁶⁵. If a country has fewer than four retirement events in that period, we use the global average lifetime (42 years), except for Asian countries where we use a regional average which is markedly shorter (30 years)^23,26.

For the reference scenario, coal plants are retired using truncated normal distribution based on the historical national lifetime and its standard deviation (see ref. ^23,26). For this calculation, we assume that no new plants are built beyond those already under ‘construction’ at the time each phase-out pledge was made. We assume that plants currently under construction come online at the planned date specified for each plant in ref. ⁶⁵. In the reference scenario, coal plant retirements begin in 2022 for countries who recently adopted coal phase-out pledges or in 2018 for countries which adopted coal phase-out pledges prior to 2019 (see ref. ²⁶). This is done to best capture the expected effect of the pledges at the moment of taking them.

In the phase-out scenario, we assume that plants follow a natural retirement trajectory – that is they follow the same logic as the reference scenario – until the pledged retirement date, when all remaining plants are abruptly retired. For countries with ranges in their coal phase-out pledge, we calculate several phase-out scenarios. For example, Vietnam and Indonesia pledge to phase out coal “in the 2040s”⁸⁷, thus we assume an optimistic phase-out by 2040, a central phase-out by 2045, and a pessimistic phase-out by 2049. For Germany, the phase-out scenarios correspond to the multi-stage coal phase-out plan proposed by the Coal Commission, and the new phase-out year envisioned by the current government⁸⁸ (Supplementary Text 4 in ref. ²⁶).

We calculate the avoided emissions from coal phase-out as the difference in emissions between the reference and the coal phase-out scenarios for each country. We do this by multiplying the capacity of prematurely retired coal power plants by the number of years between the retirement under the reference and coal phase-out scenarios and accounting for the historical national load factor as well as technology-specific efficiencies and emission rates for the thermal content of different coal types to convert avoided generation into avoided emissions (see ref. ²⁶ for more details).

For avoided emissions of China and India under 1.5 °C-, 2 °C- and 2.5 °C- compatible IPCC AR6 pathways we use a similar methodology and calculate the difference between coal emissions under a reference scenario and estimated coal emissions in IPCC AR6 pathways from 2022, essentially seeing the former as a carbon budget for coal generation (see also ref. ²³). For the reference scenario, we calculate emissions from coal power for all countries in the China+ and India+ regions from the set of ten regions (R10) using the same approach we describe above.

We then calculate unabated coal power generation under 1.5 °C-, 2 °C- and 2.5 °C- compatible pathways as the difference between total electricity generation from coal (variable “Secondary Energy|Electricity|Coal”) and generation from coal with CCS (“Secondary Energy|Electricity|Coal|w/ CCS”).

For 1.5 °C-compatible pathways, we use categories C1 (no/low overshoot) and C2 (high overshoot), for 2 °C-compatible pathways categories C3 (likely below 2 °C) and C4 (below 2 °C) and for 2.5 °C-compatible pathways we use category C5 (below 2.5 °C). We interpret AR6-scenarios in line with ref. ²³ which includes pathways which return to 1.5°C after a high overshoot as 1.5 °C-consistent (corresponding to C2 pathways in AR6) and ‘higher-2C’ pathways as 2 °C-consistent (corresponding to C4 pathways). This approach follows the original formulation of 1.5 °C -consistent pathways as those with and without overshoot⁸⁹ or both C1 and C2 category pathways¹. This approach is on the broader end of a range of different interpretations in the literature. On the narrower end of interpretations, some view only a subset of C1-pathways as Paris-consistent⁹⁰ and others a subset of C2-C3-pathways⁹¹ as 1.5 °C-consistent. Since we take a broader approach, we use the term 1.5°C-compatible rather than 1.5 °C-consistent. We test the effect of using individual pathways for 1.5 °C- and 2 °C-compatible compensation and find that this does not significantly affect our results (Supplementary Table 8).

We convert unabated coal generation to emissions using the same emission intensity as in the reference scenario for the respective region in order to estimate required avoided emissions under each 1.5 °C-, 2 °C- and 2.5 °C- compatible IPCC AR6 pathway.

We also estimate the coal phase-out year in the China+ and India+ regions in climate mitigation pathways with the median (and interquartile range (IQR)) of the first reported year when unabated coal power generation falls below 1% for each region across the sets of 1.5 °C-, 2 °C- and 2.5 °C- compatible pathways respectively (see Table 3).

Multivariable regression analysis

We conduct a multivariable regression analysis to measure the relationship between coal phase-out ambition and compensation while controlling for variables reflecting characteristics of the coal sector and the national context which are likely to affect compensation. We identify these variables based on theoretical and empirical evidence from previous literature (Supplementary Note 3).

Our sample includes all countries with coal phase-out pledges and installed coal capacity at the time of making the pledge (Supplementary Table 3), and for which data on compensation and all control variables were available. We could not quantify compensation for Chile, Ukraine or North Macedonia and there was no state capacity data for Brunei Darussalam which resulted in a sample of 39 countries.

Our outcome variable is the central estimate of coal phase-out compensation. For countries with a coal phase-out pledge but no compensation (20 countries, Supplementary Table 3), we set compensation equal to zero.

We group our independent variables in six categories representing similar mechanisms:

First, variables related to the ambition of national coal phase-out. This includes (1) Avoided emissions (Megatonne (Mt) CO2). We use three sets of avoided emissions estimates – a central estimate, an optimistic estimate where coal is phased-out at the earliest possible date, and a pessimistic estimate where coal is phased-out at the latest possible date (Supplementary Table 3) based on our own calculation as described above. (2) Number of years over which coal phase-out is pledged based on our own calculation as the difference between the year in which each phase-out pledge was made, and the end-year by which coal is pledged to be phased-out.

Second, control variables related to the strength of the coal sector or third, the level of vested interests. This includes (1) Installed capacity of operating coal power (Gigawatt (GW)) in the year of the phase-out pledge based on ref. ⁶⁵. (2) Average coal power generation (2016-2020) based on ref. ⁹². We use an average since coal power generation fluctuates due to e.g. energy demand changes or availability of other electricity generation sources (3) Average coal mined (Mt) based on ref. ⁹³ for most countries, ref. ⁹⁴ for Greece, Bulgaria and Slovakia, and ref. ⁹⁵ for Vietnam. We use the average over the last five available years due to fluctuations in annual coal production. (4) Coal share in power generation based on data from ref. ⁹², for the year in which coal phase-out was pledged. (5) Number of coal workers; own calculation based on national employment factors from ref. ⁹⁶, multiplied by installed coal capacity, coal capacity in construction, and amount of coal mining (see sources above), respectively. (6) A variable measuring the regional concentration of coal power plants within a country. Own calculation based on the Shannon Evenness Index (SEI)⁹⁷:

where

P_i = the proportion of coal capacity in each region within a country, using data on regional distribution of coal capacity from ref. ⁶⁵.

m = the total number of regions within a country. We define regions as administrative subdivisions at the highest level (for example, Zambia is divided into ten provinces, which each include several districts. We use the higher level, province.)

Finally, we include three types of control variables related to the national context: First, variables measuring state capacity: (1) Hanson and Sigman’s (HS) index⁹⁸ which incorporates indicators of extractive, coercive and administrative dimensions of capacity⁹⁸, and has been shown to be a robust predictor of coal phase-out³². (2) The Government Effectiveness Index from the World Bank which captures quality of public and civil service, the quality of policy formulation and the commitment of a government to its policies⁹⁹; and “focuses strongly on the administrative aspects of state capacity”¹⁰⁰. Second, variables measuring economic capacity: (1) The size of the national economy (GDP) for the year in which coal phase-out was pledged converted to USD2020 from the IMF World Economic Outlook⁸⁴. (2) GDP/capita (Purchasing Power Parity (PPP)) for the year in which coal phase-out was pledged converted to USD2020 from the Penn World Table¹⁰¹. And third, a variable on access to international funding capturing whether a country is a donor or recipient of either Official Development Assistance (ODA) or EU funds. We code this variable based on data from ref. ¹⁰² and a report on EU finances¹⁰³.

All models include a measure of the ambition of coal phase-out pledges, since our goal is to test the relationship between ambition and compensation while controlling for other potentially relevant variables. We limit the number of independent variables to a maximum of four in each model due to the relatively small sample (39 countries). To avoid multi-collinearity, we also only use one variable per variable category and exclude any variable combinations with high correlation (Pearson’s R² > 0.7).

We run five sets of multivariable regression analyzes: a central set with our central estimates of compensation and coal phase-out pledge ambition; two sets where we vary ambition using an optimistic and pessimistic interpretation of coal phase-out pledges; and two sets where we vary compensation policies using both an upper and lower estimate for compensation (Supplementary Tables 12–16).

This returns 820 machine-generated models. We find that there are conflicting results on the relationship between economic and state capacity and compensation, where for example poorer countries have higher compensation, when we don’t control for access to international funding. We thus pool measures of economic or state capacity with access to international funding, meaning that we include only economic or state capacity as control variables in models which also control for access to international funding.

This results in 330 remaining models: – 66 models from each of the five regression analyzes. We rank the models from each respective analysis according to Akaike’s Information Criterion (AIC) (Supplementary Note 3). AIC indicates goodness of fit, penalizing for additional independent variables¹⁰⁴. A lower AIC means a better model fit.

We report our top ten models for all five sets of regression analyzes (Supplementary Tables 12–16).

Estimating compensation and its uncertainty range for China and India under climate pathways

To estimate compensation for China and India, we use the central average of compensation/ton avoided emissions ($37.5/tCO₂), calculated based on all countries with coal phase-out and compensation policies, and the median of avoided emissions in 1.5 °C-, 2 °C- and 2.5 °C -compatible pathways from the IPCC AR6 database for the China+ and India+ regions⁸⁶. We use the China+ and India+ regions to estimate compensation in China and India respectively, since each country accounts for at least 97% of coal power generation in their respective region. In identifying the median among IPCC AR6 pathways, we exclude a handful of pathways (ten for 2 °C and 25 for 2.5 °C) which depict a coal power expansion, and thus higher emissions from coal power than in the reference scenario as this has been criticized as unrealistic¹⁰⁵.

We calculate the uncertainty ranges for compensation estimates for China and India accounting for three types of uncertainties (Supplementary Note 2): (1) Parametric uncertainties arising from coal phase-out pledges and compensation policies; (2) Model uncertainties arising from using different control variables and the confidence intervals across different regression models; and (3) Pathway uncertainties arising from coal phase-out trajectories for China and India envisioned in different IPCC AR6 pathways and leading to different levels of avoided emissions

We calculate the uncertainty ranges of compensation for China and India using two methods: first, we use the top performing models across our five sets of regression analyzes and their confidence interval (Supplementary Tables 12–16). Second, we use a range of average compensation per ton of avoided emissions based on varying coal phase-out pledge ambition and compensation estimates (Supplementary Table 10 and Supplementary Note 2). We apply both methods to the IQR of avoided emissions in IPCC pathways.

Comparing compensation to national and international policy support

We also benchmark compensation against several domestic energy and climate policies, international financial support mechanisms, as well as recent coal power plant costs.

For domestic energy and climate policies we consider average carbon price data from 2017 to 2022 under the EU emissions trading scheme from emission spot primary market auction reports¹⁰⁶; and annual coal production subsidies from the Organisation for Economic Co-operation and Development (OECD)¹⁰⁷ and International Institute for Sustainable Development (IISD)¹⁰⁸ (Supplementary Table 7 and Supplementary Fig. 3). We do not include coal consumption subsidies, since compensation for coal phase-out generally focuses on producers (companies, workers, and regions) rather than consumers; we also do not include investments in state owned enterprises in our main estimate because these are made under the assumption that enterprises are operational and will return a profit, while compensation is paid without an expected return. We could not identify coal subsidy information for Vietnam, Bulgaria, Croatia and Romania.

For international financial mechanisms, we include only public finance which is most comparable to the policy effort for accelerated coal phase-out and since private investment is likely driven by a different logic. We include average annual gross Official Development Assistance (ODA) disbursements over the period 2013-2022 (Aid type: “Memo: ODA Total, Gross Disbursements”)⁴³; annual climate finance first pledged by developed countries to developing countries at the 15^th Conference of the Parties (COP15)²⁵; and a climate finance request from India’s prime minister for $1 trillion at COP26¹⁰⁹. To compare this to annual compensation estimates (Fig. 3), we divide the request by the median duration of coal phase-out in line with 1.5 °C- and 2 °C- compatible pathways (Supplementary Table 7).

To compare coal phase-out subsidies and the international financial mechanisms to compensation, we calculate an annual compensation rate by dividing the upper and lower estimates for compensation by the number of years from the announcement of the coal phase-out pledge to the year of planned coal phase-out. For countries with uncertainty in the pledge date (Supplementary Table 3) we use the longer coal phase-out duration for the lower estimate and the shorter duration for the upper estimate, assuming that more ambitious pledges would be accompanied by higher compensation. For annual compensation estimates for China+ and India + , we divide pathway-specific coal compensation by pathway-specific phase-out dates (the first year in which unabated coal power generation declines below 1%).

Finally, for coal power plant costs, we identify all recently-constructed coal plants in the EU between 2010 and 2022 from the S&P database⁶⁵ using a systematic Google search with the terms: “[power plant name]” + “construction cost” + “[year of construction]” (Supplementary Table 6). To compare these costs to compensation, we normalize compensation to the installed coal capacity in the year the coal phase-out pledge was made⁶⁵.

Limitations

Our analysis is based on currently evolving coal phase-out pledges and compensation policies. This dataset needs to be updated as new countries develop compensation policies, and more information about the design and implementation of such policies can be collected. Additionally, given the early phase of such policies it is not possible to evaluate their implementation or impact. It will be particularly important to understand how regions, workers and industries supported by compensation fare in the long term and what role compensation plays in this development. As compensation policies evolve along their implementation and evaluation phases, uncertainties around the amount of compensation will decrease – such as current uncertainties around future compensation particularly for JETP countries. Our database and analysis should be updated as more information becomes available. We also do not examine the difference between grants versus loans for international compensation such as the JETPs – future research is needed to understand the conditions under which grants or loans are pledged, and how this affects the implementation of coal phase-out compensation.

Additionally, while the policy documents we review provide some information on how compensation policies are financed, compensation likely originates from additional sources. Where we could not access national policy documents in a language known by the authors, we reached out to country experts (for example for Poland) or retrieved information from international organizations (for example for Greece) to minimize the effect of language barriers on our data collection. However, it is possible that additional resources may be accessible to native speakers for certain countries.

Finally, the predictive power of statistics tends to increase with the number of cases. Currently, a relatively small number of countries have both coal phase-out pledges and associated compensation policies. If more countries are added to the database in the future, the regression analysis should be updated to understand whether the relationship we currently observe between avoided emissions and amount of compensation remains stable.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.