INTRODUCTION
The green transition should result in far-reaching changes in the labor market. There are new opportunities for “green” jobs related to environmental sustainability, but also threats to occupations that exist primarily in polluting industries, often labeled “brown jobs” (Vona et al. Reference Vona, Marin, Consoli and Popp2018).Footnote 1 Entire communities face uncertain futures due to their reliance on the extraction or burning of fossil fuels (Gazmararian and Tingley Reference Gazmararian and Tingley2023; Raimi, Carley, and Konisky Reference Raimi, Carley and Konisky2022) or the emission of other pollutants, such as nitrogen. European far-right parties increasingly campaign on the environment as a wedge issue; hoping to peel off dissatisfied voters by breaking from the mainstream party consensus on the need for a green transition (De Vries and Hobolt Reference De Vries and Hobolt2020; Dickson and Hobolt Reference Dickson and Hobolt2025). Does the green transition result in polarization across occupational lines, where workers in brown jobs and communities that rely on these jobs increasingly shift toward the far right?
We build on the existing literature on support for challenger parties and the electoral consequences of adverse economic shocks to develop a theoretical argument about the electoral consequences of the green transition. We argue that the green transition impacts elections if it affects voters’ political preferences (demand side) and if political parties offer divergent political platforms capturing these preferences (supply side). Regarding voters’ preferences, two distinct features are that the green transition is planned and that it has a strong normative dimension. The long-term disappearance of brown jobs is not an undesired side effect of trade policies or automation, but a planned and desired objective of the government-led green transition. The economic and social effects of the green transition can combine to reduce the relative perceived status of individuals in brown occupations. Building on prior research on the repercussions of adverse economic shocks (Baccini and Weymouth Reference Baccini and Weymouth2021; Broz, Frieden, and Weymouth Reference Broz, Frieden and Weymouth2021), we argue that these threats posed to individuals can spill over to members of their communities.
These political and social grievances can translate into electoral consequences if a political party offers opposition. Prior to the green transition, brown industries were generally embraced and protected by mainstream parties. There was little reason for those in brown jobs to move toward anti-establishment parties. This makes the green transition a potential wedge issue for far-right parties, who can claim that mainstream parties no longer represent brown jobs and communities. Based on these considerations, we hypothesize that the green transition and the resulting existential threat to brown occupations may increase far-right vote share in both (i) communities that depend on these jobs and (ii) individuals who work in brown occupations after the far right clearly articulates opposition.
We examine these hypotheses in the context of Germany, which has long pursued relatively aggressive green transition policies and has a significant workforce in polluting occupations. The need for a green transition was a consensus issue among mainstream parties, although the Green party advocated for more aggressive policies. In 2016, the far-right Alternative for Germany (Alternative für Deutschland; AfD) shifted to become the only party that vocally opposed green transition policies, emphasizing that the green transition poses existential threats to the German lifestyle, economic freedom, and key industries.
Our empirical strategy combines a county-level panel of federal election results with administrative labor market data and an individual-level panel survey. We build on prior work by defining a brown occupation as one that is much more likely to be in a highly polluting industry than in any other industry (Cavallotti et al. Reference Cavallotti, Colantone, Stanig and Vona2025; Vona et al. Reference Vona, Marin, Consoli and Popp2018). Individuals in these occupations are likely to be more invested in the success of polluting industries than those whose roles are more general, such as someone working in marketing or financial administration within a polluting industry.
As a first step, we use a difference-in-difference (DiD) design to show that following the AfD’s 2016 platform shift, counties with high pre-existing shares of brown employment experienced a significantly greater increase in AfD support in the 2017 and 2021 elections (relative to the 2013 baseline and to counties with lower shares). In contrast, the Greens lost in counties with higher brown employment shares, particularly in 2021. Importantly, we find no evidence that trends in far-right support prior to the 2013 election were correlated with the prevalence of employment in brown occupations. Assessing alternative explanations, we further show that electoral shifts after 2013 cannot be attributed to a correlation between brown employment and manufacturing employment. In addition, trade exposure proxied by changes in manufacturing employment can also not explain our main result. Finally, we demonstrate that our results do not stem from correlations between brown employment and: (i) preexisting levels of xenophobia, (ii) support for anti-immigrant positions, (iii) regional differences in refugee inflows, (iv) urban–rural differences, (v) center–periphery conflict, and (vi) pre-existing social status and ethno-linguistic differences.
Moving beyond aggregate-level analyses, we use a large-scale annual panel survey to examine individual-level changes in political preferences. We find that shifts toward far-right parties occur both among individuals in brown occupations and those residing in counties with high brown employment, even if they are not themselves employed in brown occupations. Based on the estimated magnitudes and the size of the workforce employed in brown occupations, we derive three implications: first, the shift among people in brown occupations alone cannot fully account for the aggregate-level changes. Rather, the aggregate-level results partially stem from preference shifts among individuals who are not themselves employed in brown occupations. Second, our panel results address the ecological inference problem, demonstrating that aggregate electoral shifts are due to individual vote switching rather than changes in voter turnout among supporters or compositional differences among the electorate. Third, the panel results show that the shift occurred precisely in 2016, the year in which the AfD shifted its platform.
Finally, we present more suggestive evidence related to mechanisms underlying these results. We do not find clear evidence that economic insecurity increases among people in brown occupations after 2015. Similarly, counties with higher shares of brown employment did not experience pronounced declines in wages or unemployment. We do, however, observe lower self-reported social status among individuals in high-emission occupations and among people living in counties with high shares of these jobs. This suggests that far-right parties can tap into social stigmatization and perceptions of lower status among individuals and communities most impacted by the green transition. These findings call for more research on potential backlash effects of growing anti-fossil fuel norms and stigmatization of fossil fuel communities.
Our article contributes to an emerging literature that examines new domestic political cleavages over the green transition. As the distributive consequences of the green transition become more apparent, political entrepreneurs have started to court those who have lost or who expect to lose when their industries are phased out. In many countries, far-right parties have been the most vocal and often the only opponents of green transition policies. This suggests that the environment may become a third pillar of far-right support beyond opposition to immigration and globalization (Colantone et al. Reference Colantone, Di Lonardo, Margalit and Percoco2024; Voeten Reference Voeten2025). Most of the existing literature has focused on political opposition to energy taxes or other measures that affect the cost of carbon-intensive consumption. We analyze the labor market implications, which thus far have only received attention in regard to specific sectors, especially coal (Bolet, Green, and González-Eguino Reference Bolet, Green and González-Eguino2024; Egli, Schmid, and Schmidt Reference Egli, Schmid and Schmidt2022; Gazmararian Reference Gazmararian2025; Goetz et al. Reference Goetz, Davlasheridze, Han and Fleming-Muñoz2019; Stutzmann Reference Stutzmann2025). Expanding on this research, we provide evidence from a broad range of occupations, both at the aggregate and individual levels.
In addition, our study provides new evidence on the role of the supply side in explaining political responses to anticipated economic shocks. As Rodrik (Reference Rodrik2021) notes, most studies on the political effects of economic shocks focus on demand-side factors. This is not surprising, since supply-side changes are often endogenous to voter demands, making it difficult to separate supply-side effects from prior changes in voter preferences. Yet, scholars have also observed that the globalization backlash is not due to large swings in voter opinions about trade or globalization but to the politicization of those issues by political entrepreneurs (Walter Reference Walter2021). Those entrepreneurs would not be effective in the absence of voter demand. Yet, without clearly distinguished political platforms, voters cannot translate their concerns over globalization or the green transition at the ballot box.
Our empirical setting enables us to identify the electoral impact of a supply-side shift when it interacts with pre-existing policy demands. As detailed below, the AfD platform change was largely due to internal party dynamics, such as a leadership change resulting in increased opposition to the Green Party. Based on secondary evidence, the timing and nature of the platform change did not stem from immediate prior voter demands for greater opposition to the green transition. This aligns with the empirical evidence we present, which shows that regions with larger brown employment shares did not diverge electorally prior to the AfD platform shift. Therefore, we can credibly demonstrate the relevance of this supply-side shift in shaping electoral outcomes. This evidence, highlighting how political entrepreneurship can activate latent grievances tied to structural economic changes like the green transition, is particularly important as supply-side mechanisms are understudied, as previous surveys of the literature have noted (see, e.g., Guriev and Papaioannou Reference Guriev and Papaioannou2022).
We highlight that unlike globalization and automation shocks, the economic effects of the green transition are planned and have a strong normative dimension: “dirty” occupations and industries must be phased out over the next few decades. This expands the existing literature on jobs and the far right, which has focused on the displacement of manufacturing jobs due to automation and globalization (e.g., Autor et al. Reference Autor, Dorn, Hanson and Majlesi2020; Colantone and Stanig Reference Colantone and Stanig2018; Dippel et al. Reference Dippel, Gold, Heblich and Pinto2022). The potential losers are clear: individuals with occupations in polluting industries and communities that rely heavily on these industries and occupations. Moreover, civil society actors have actively created social stigmas around brown occupations and industries (Blondeel, Colgan, and Van de Graaf Reference Blondeel, Colgan and Van de Graaf2019). This means that the political effects of the green transition function differently than for trade policy or social welfare cuts: even if the immediate job losses are minimal, transition policies can have political effects by making clear who the long-term losers of policies will be. Moreover, the normative dimension of the green transition lowers the social status of individuals and communities that rely on polluting activities.
One policy implication is that compensatory policies should look beyond economic compensation for individuals. Our findings are consistent with the increasingly recognized need for community-based compensation policies for the economic implications of green transition policies (e.g., Bolet, Green, and González-Eguino Reference Bolet, Green and González-Eguino2024; Gazmararian and Tingley Reference Gazmararian and Tingley2023; Heddesheimer, Hilbig, and Wiedemann Reference Heddesheimer, Hilbig and Wiedemann2024). Yet, our findings also suggest that policymakers should consider the consequences of social stigmatization if they wish to avoid backlash to green transition policies. That is, material compensation may not be enough. The conclusion returns to this issue and the need for more research on understanding potential backlash against the normative dimensions of the green transition.
LITERATURE AND THEORY
Literature
Scholars have attributed the rise of the far right in Europe and the United States to both cultural and economic factors. The political economy literature has identified three plausible economic drivers of support for radical right parties and candidates: deindustrialization due to globalization and automation, austerity following the 2008 financial crisis, and the economic consequences of migration (for recent overviews of the literature, see Colantone and Stanig Reference Colantone and Stanig2019; Margalit Reference Margalit2019; Rodrik Reference Rodrik2021).
An emerging literature focuses on the green transition as a fourth potential economic driver of support for far-right parties and candidates. Most of this literature highlights how transition policies increase the costs of consumption, which may drive some voters toward parties and candidates that oppose environmental policies and away from Green parties or others who take pro-environmental positions. For example, Milanese voters who were negatively affected by a ban on polluting cars became more likely to vote for the Lega Nord (Colantone et al. Reference Colantone, Di Lonardo, Margalit and Percoco2024). Similarly, a Dutch policy that increased household energy taxes and redistributed the revenues as subsidies for renewable energy led affected voters to increase support for the radical right and decrease support for the Greens (Voeten Reference Voeten2025). These findings build on prior literature showing that voters are sensitive to gasoline price increases (e.g., Kim and Yang Reference Kim and Yang2022) and a public opinion literature showing that public support for climate policies depends on their cost to individuals (e.g., Bechtel, Genovese, and Scheve Reference Bechtel, Genovese and Scheve2019; Beiser-McGrath and Bernauer Reference Beiser-McGrath and Bernauer2023; Bernauer and Gampfer Reference Bernauer and Gampfer2015; Egan and Mullin Reference Egan and Mullin2017; Schaffer Reference Schaffer2024; Stokes and Warshaw Reference Stokes and Warshaw2017).
There is some literature on the electoral implications of the decline in coal employment in the United States. The decline of coal in the United States is mostly a consequence of increased competition from natural gas, rather than policies. However, parties and politicians have taken visibly different positions about what the future of coal should look like. Especially Donald Trump campaigned heavily to defend coal; accusing the Obama administration of waging a “war on coal.”Footnote 2 Districts with a higher share of coal-related job losses have relatively higher shares of the Trump vote (Egli, Schmid, and Schmidt Reference Egli, Schmid and Schmidt2022; Goetz et al. Reference Goetz, Davlasheridze, Han and Fleming-Muñoz2019) and there is evidence that the drop in coal employment has led to a partisan shift toward Republicans in U.S. coal communities (Gazmararian Reference Gazmararian2025). There is evidence that the coal phase-out in Germany contributed to electoral losses for the center-left Social Democratic Party (SPD) (Stutzmann Reference Stutzmann2025). Moreover, beyond coal, individuals employed in high-emission industries and occupations are less likely to support climate cooperation (Bechtel, Genovese, and Scheve Reference Bechtel, Genovese and Scheve2019). In a paper close to our study, Cavallotti et al. (Reference Cavallotti, Colantone, Stanig and Vona2025) show that workers who benefit economically from the energy transition exhibit support for the energy transition and parties advocating for it. Using cross-sectional survey data in 15 countries and constructing similar “greenness scores,” Cavallotti et al. (Reference Cavallotti, Colantone, Stanig and Vona2025) find that people in greener jobs are more likely to support environmental policy, Green parties, and parties with pro-environmental policy platforms, while the opposite holds for people with browner jobs.
However, prior work either (i) only focuses on specific sectors (Bechtel, Genovese, and Scheve Reference Bechtel, Genovese and Scheve2019; Gazmararian Reference Gazmararian2025) or (ii) only focuses on demand-side factors. In contrast, we focus on occupations (rather than sectors) and directly show an effect on party preference and voting behavior rather than policy views. Moreover, we propose a theoretical framework that highlights the role of the supply side—that is, the positions that parties and politicians take, as well as the role of the normative dimension of the green transition that might lead to distinct political responses of brown communities and workers. Finally, our study is the first to study both within-person changes using panel data and aggregate vote returns.
Theory
Like any theory of electoral change, our theory must explain why the green transition affects voters’ political preferences (demand side) and why some parties are able to capitalize on this demand (supply side) (De Vries and Hobolt Reference De Vries and Hobolt2020). Economic uncertainty and perceptions of status anxiety can create political grievances. However, these grievances translate into electoral consequences only if a political party offers an opposing stance.
Our theory builds on the globalization backlash literature. Scholars have shown how trade patterns resulted in economic insecurity and status anxiety among workers in import-competing industries and communities that relied on those industries (e.g., Baccini and Weymouth Reference Baccini and Weymouth2021; Ballard-Rosa, Jensen, and Scheve Reference Ballard-Rosa, Jensen and Scheve2022; Gidron and Hall Reference Gidron and Hall2017). Yet, this only affected elections after political entrepreneurs broke the elite consensus on trade and politicized the issue (Walter Reference Walter2021).
Similarly, we argue that the green transition challenges workers in brown occupations and the communities that rely on them. Given that there was elite consensus on the transition among mainstream parties, this made it a potential wedge issue for far-right parties to attract dissatisfied voters from mainstream parties (Haas et al. Reference Haas, Stoetzer, Schleiter and Klüver2023). While the globalization backlash was a response to the unintended negative effects of trade, we argue that the green transition backlash is in large part a response to the planned and intended phasing out of brown jobs.
Demand Side
We argue that the green transition can induce political opposition among workers in polluting occupations and in communities that rely on these occupations. We build on the substantial literature that attributes the rise of the radical right in the United States and Europe at least in part to job losses in import competing industries (e.g., Autor et al. Reference Autor, Dorn, Hanson and Majlesi2020; Colantone and Stanig Reference Colantone and Stanig2018; Dippel et al. Reference Dippel, Gold, Heblich and Pinto2022). Unlike the China shock literature, our emphasis is not on already realized job losses. There is evidence that higher energy prices reduced labor demand in energy-intensive industries across Europe in the 2010s (Bijnens, Konings, and Vanormelingen Reference Bijnens, Konings and Vanormelingen2022; Cox et al. Reference Cox, Peichl, Pestel and Siegloch2014; Marin and Vona Reference Marin and Vona2021). Energy price increases due to the war in Ukraine further decreased the productivity and stock market values of European energy-intensive firms (Ferriani and Gazzani Reference Ferriani and Gazzani2023). However, the strongest economic impacts of the green transition have yet to materialize.
Most transition policies have a long-term focus and are implemented in phases. The threat to brown industries and occupations is planned and existential: some industries and jobs will and should disappear altogether if governments follow through on their climate commitments (Colgan, Green, and Hale Reference Colgan, Green and Hale2021). Green and other parties that advocate for more aggressive transition policies explicitly call for a phase out of coal and other fossil fuels. Such policy positions have clear and drastic consequences for individuals in brown occupations and for communities that either produce fossil fuels or have industries that heavily rely on fossil fuels, especially hard-to-abate sectors like the chemical and steel industries. Similarly, farmers protest that laws that restrict nitrogen emissions, such as the EU’s Natura 2000 legislation, pose existential threats to their occupations and communities (Van der Ploeg Reference Van der Ploeg2020). Although the transition also creates new greener jobs, people rarely move from more carbon-intensive jobs to greener jobs (Bluedorn et al. Reference Bluedorn, Hansen, Noureldin, Shibata and Tavares2023). Moreover, promises of financial compensation or green jobs that could replace brown jobs may not be credible (Gazmararian and Tingley Reference Gazmararian and Tingley2023). However, in our case, the economic effects are mostly anticipated and planned, which clarifies the distributional consequences for voters and potentially makes the direct economic pathway less important.
The green transition also has strong moral and social components to it. Norm entrepreneurs have, with mixed success, advocated for strong moral norms against fossil fuels and fossil fuel producers, including divestment campaigns (Blondeel, Colgan, and Van de Graaf Reference Blondeel, Colgan and Van de Graaf2019; van Asselt and Green Reference van Asselt and Green2023). Large-scale protests, such as the Fridays for Future marches, publicly showed support for ending fossil fuels. Public demands that some industries should be phased out can send a social signal to people occupied in polluting occupations and the communities that depend on them (Spaiser, Nisbett, and Stefan Reference Spaiser, Nisbett and Stefan2022). Moreover, exposure to Fridays for Future protests is associated with increased support for Green parties, who advocate for stronger transition policies (Valentim Reference ValentimForthcoming). Valentim (Reference ValentimForthcoming) also finds that exposure to the protests makes voters who identify with the Green Party more supportive of pro-climate policies, whereas AfD voters become less supportive of climate policies. This suggests a potential backlash among subgroups of voters against social stigmatization.
Consistent with the globalization literature (e.g., Broz, Frieden, and Weymouth Reference Broz, Frieden and Weymouth2021), we argue that these effects can spill over into communities that rely heavily on brown occupations. Phasing out of jobs and industries can affect housing prices, services, and spill over into adjacent industries. Moreover, scholars have argued that “fossil fuel” communities often develop a shared identity and shared existential uncertainty in the face of the energy transitions (Gazmararian and Tingley Reference Gazmararian and Tingley2023; Gazmararian Reference Gazmararian2024). This also occurs in agricultural communities. For instance, even though farmers are a very small part of the labor force even in Dutch rural areas, the farmers’ party was able to attract a plurality of the vote in Dutch provincial elections due to solidarity with farmers over the implications of nitrogen regulations (Otjes and de Jonge Reference Otjes and de Jonge2024). Sociologists have also identified the broader social implications of threats to relatively well-paid jobs for a relatively less educated and male-dominated labor force (for an overview, see Beckfield and Evrard Reference Beckfield and Evrard2023). Finally, there is evidence that people tend to prefer broad-based community approaches to direct compensation for job losses because they understand the shared fate (Gaikwad, Genovese, and Tingley Reference Gaikwad, Genovese and Tingley2022).
In sum, the anticipated impacts of the green transition can generate economic and status insecurity, fostering political opposition among workers in polluting occupations and their communities. Unlike the globalization backlash where voters reacted to already-realized job losses, the green transition’s backlash may stem from prospective and anticipated threats.
Supply Side
A key precondition for social and economic changes to have political effects is that candidates or parties take clearly recognizable positions (e.g., De Vries and Hobolt Reference De Vries and Hobolt2020). For example, in the United States, Donald Trump departed from longstanding Republican support for trade. Scholars therefore identify the effect of manufacturing employment losses on changes in Republican vote shares between 2012, when Romney ran on a traditional Republican platform, and 2016, although the labor market implications of the China shock were already visible in 2012 (Broz, Frieden, and Weymouth Reference Broz, Frieden and Weymouth2021). Identifying whether and when voters have the option to vote for a party opposing the green transition is thus a prerequisite for any study of the electoral effects of the green transition.
De Vries and Hobolt (Reference De Vries and Hobolt2020) argue that successful challenger parties engage in both policy innovation, the mobilization of new and divisive issues, and rhetorical innovation, which refers to the use of language that is highly critical of mainstream elites. The green transition is a promising candidate for challenger parties because even though the transition policies are threatening to some individuals and communities, it has long been a consensus issue in many European countries. Green parties have advocated for more aggressive green transition policies, but there was little active opposition to existing policies (Huber et al. Reference Huber, Maltby, Szulecki and Ćetković2021). This makes it a potential wedge issue where radical right parties can peel off dissatisfied voters from mainstream parties (Dickson and Hobolt Reference Dickson and Hobolt2025).
Based on this discussion, we propose two hypotheses:
-
• H1: Communities with a larger share of voters in brown occupations become more likely to vote for the far right once the far right opposes green transition policies.
-
• H2: Individuals in brown occupations become more likely to identify with the far right once the far right opposes green transition policies.
In sum, we argue that two conditions must simultaneously hold for the green transition to have polarizing effects along occupational lines. First, workers in polluting occupations and their communities must perceive that they stand to lose from the green transition. Second, there needs to be a political party or candidate that clearly articulates opposition to the transition, offering a channel for these grievances. One issue we cannot fully resolve is whether workers in brown jobs realize their unresolved grievances before a political party makes the issue salient in elections. An influential literature in public opinion argues that voters often do not realize their interests until party elites cue an issue (e.g., Bullock Reference Bullock2011; Zaller Reference Zaller1992). We will show in the case discussion that there is little evidence that the AfD’s change of position came about solely due to voter demands; suggesting that the supply shift was relevant in and of itself. However, it would have been difficult for the AfD to fully construct an interest in the absence of people and communities who would credibly lose as a consequence of the transition. Indeed, we show that the AfD framed its opposition to transition policies in terms of its economic burdens and threats to jobs. We thus suggest that both demand and supply conditions must be present simultaneously.
The Case
We examine these theoretical expectations in the context of Germany. Germany presents a compelling case for studying the electoral effects of the green transition due to its ambitious green transition policies and a significant share of polluting industries. On the demand side, these policies have the potential to generate economic and social insecurity among workers in high-emission occupations and their communities. On the supply side, there has traditionally been a consensus among German mainstream parties in support of the green transition. However, the far-right AfD’s platform switch to oppose these policies provides a clear channel for political grievances, making Germany an appropriate setting to explore our theoretical framework.
The German Green Transition
Germany has relatively advanced green transition policies and a relatively large share of the population working in polluting industries. German transition policies have mostly focused on energy and climate, although they have also targeted nitrogen and other pollutants. The 2000 Renewable Energy Act adopted a feed-in tariff and grid priority for renewables. In 2010, the government led by the center-right Christian Democratic Union (CDU) and its Bavarian sister party, the Christian Social Union (CSU), formally endorsed the “Energiewende”—Germany’s planned transition toward a low-carbon (and nuclear-free) economy. These policies were supported by various coalition governments led by the CDU/CSU and joined by either the liberal Free Democratic Party (FDP) or the center-left SPD.
The pre-2016 political status quo consisted of a consensus favoring the expansion of renewable energy sources, while avoiding sharp conflicts over how the economic burdens and benefits of this transition would be distributed. Rather than aggressively phasing out brown industries, governments across the political spectrum shielded coal and other polluting sectors from immediate disruption. Coalitions linking coal unions, energy-intensive industries, and regional politicians from both SPD and CDU defended existing industrial structures (Gazmararian and Tingley Reference Gazmararian and Tingley2023; Mildenberger Reference Mildenberger2020). Germany heavily subsidized hard coal mining until 2018, and lignite remained politically sensitive throughout this period (Furnaro et al. Reference Furnaro, Herpich, Brauers, Oei, Kemfert and Look2021; Hermwille and Kiyar Reference Hermwille, Kiyar, Michael and Jan C.2022). In the absence of a credible anti-green party, this strategy preserved political stability, despite the underlying tensions in Germany’s policy of both safeguarding established industries and pursuing a major energy transition.
Despite Germany becoming a leader in renewable energy industries (Nahm Reference Nahm2017), the country also remains a traditional industrial powerhouse. This means that a sizable share of the workforce is exposed to the green transition, which includes potential adverse effects on their job security and overall livelihoods. For example, about 20% of Germany’s workforce is in manufacturing, compared to just 10% in the United States and 12% in France. The auto manufacturing industry is only just beginning to transition, resulting in significant anticipated job losses as electric vehicles, having fewer parts and simpler powertrains, generally require less manufacturing labor. The chemical industry cannot easily transition away from fossil fuel combustion. For example, BASF, which has more than 50,000 employees in Germany, has announced that it will permanently cut operations in Germany in response to “high energy prices and overregulation.”Footnote 3 There is evidence that higher electricity prices have resulted in lower demand for labor in German industries with high electricity demand, and that the effect is greater among low-skilled workers (Cox et al. Reference Cox, Peichl, Pestel and Siegloch2014).
The 2016 Afd Party Platform Shift
As discussed in the “Theory” section, electoral responses to the green transition require that there is a party that actively opposes the transition. Until the mid-2010s, there was no viable party in opposition to the green transition. This changed in 2016, when the AfD started to vocally oppose green transition policies. Since the AfD’s party platform shift is central to our empirical analysis, we now provide evidence on the nature and origins of this shift. We do so using two data sources: party manifestos and party press releases.
Party Manifestos
The AfD party program for the 2013 Federal elections remained silent on climate change.Footnote 4 The 2014 program for the European elections claims that scientific evidence for anthropogenic climate change is uncertain. However, the program still supports international climate mitigation efforts, claiming that: “In order to take the precautionary principle into account, a gradual reduction in CO
$ {}_2 $
emissions can be agreed within the framework of international agreements.”Footnote 5
This changes completely with the AfD’s 2016 party Grundsatz (basic principles) program, which attacks both climate science and climate policies. The document claims that “Carbon dioxide (CO
$ {}_2 $
) is not a pollutant, but an indispensable component of all life.” The program also tackles the green transition and its existential threat to industry and German ways of life (see also Figure 1). The program states that “[…] the German government is abusing the increasing concentration of CO
$ {}_2 $
for the “great transformation” of society, with the result that personal and economic freedom is massively restricted.”Footnote 6 Since then, the AfD has only increased its attention to this problem. An additional analysis of media articles shows that the AfD’s climate change stance gained public attention only after 2015, with major coverage starting in 2016, highlighting their new election program and job loss criticisms (see Section A.2 of the Supplementary Material).
Figure 1. AfD Election Poster
Note: English translation: “Stop the EEG [German Renewable Energy Sources Act] and the green transition. Electricity should not be a luxury!” Picture taken in September 2017 in Leverkusen. Retrieved from https://www.leverkusen.com/guide/Bild.php?view=48509#google_vignette.
In 2019, AfD leading politician Alexander Gauland declared the opposition to the so-called climate change policies to be among the three most important topics for the AfD.Footnote 7 Its 2021 program again challenges climate science and the “radical restructuring of industry and society” that “threatens our freedom on an increasingly frightening scale.”Footnote 8 Importantly, the AfD emphasized social injustices related to the green transition, especially on the question of “who pays?” for the transition (the AfD’s answer being “the core people”) (Küppers Reference Küppers2024).
Party Press Releases
We substantiate the previous discussion through a supplementary analysis of all press releases by German parties between 2010 and 2019.Footnote 9 Press release data come from the PARTYPRESS database (Erfort, Stoetzer, and Klüver Reference Erfort, Stoetzer and Klüver2023). This analysis yields two main results. First, the share of press releases that discuss the green transition roughly doubles after 2015. This contrasts with all other parties, for which we observe either unchanged or decreasing attention to the green transition in press releases (see Figure A.1 in the Supplementary Material). Second, the AfD is the only party for which the sentiment of press releases on the green transition is consistently negative (see Figure A.2 in the Supplementary Material).
Framing of the Green Transition
Moreover, we use the same press release data to analyze how the AfD framed the green transition in its public communication. We find that the AfD consistently emphasized the economic costs of the green transition (Figure B.1 in the Supplementary Material). These frames are also presented in a distinctly negative tone (Figure B.2 in the Supplementary Material), especially in relation to labor market impacts and national sovereignty. In contrast, the AfD de-emphasizes climate protection and the urgency of climate action compared to other parties. These results support the interpretation that the AfD’s political supply-side shift involved not only adopting a new position on green policies but also actively constructing a narrative of threat to jobs, increasing economic costs, and national autonomy.
Reasons for the AfD Platform Shift
The AfD changed their position on the green transition in 2016 primarily based on strategic considerations. In part, this stemmed from a change in the party leadership in 2015. This leadership change increased the party’s tendency to oppose policies endorsed by the government and all mainstream parties (De Vries and Hobolt Reference De Vries and Hobolt2020; Küppers Reference Küppers2024; Spoon, Hobolt, and de Vries Reference Spoon, Hobolt and de Vries2014). A 2016 internal strategy paper outlined that the AfD aimed to differentiate itself from parties like the Greens by opposing the green transition and by promoting a broader political narrative seeking to return to a “good old” industrial era. Moreover, party professionalization and spillovers from radical right parties outside of Germany who succeeded by running on climate-skeptic positions also contributed to this policy shift. As we discuss in Section A of the Supplementary Material, secondary evidence on the AfD platform shift suggests that this shift was not driven by ideological changes among AfD supporters or potential supporters.
DATA
Our empirical analysis draws on: (i) aggregate-level voting data at the county (Landkreis) level, (ii) administrative labor market data, and (iii) an individual-level panel that includes both occupation and party preferences. Our aggregate-level data link administrative employment data and federal election results at the county level for elections between 2005 and 2021. We present summary statistics for the aggregate-level variables in Table 1. At the individual level, we use the German Socio-Economic Panel (SOEP), a longitudinal household survey first conducted in 1984 and carried out annually (Goebel et al. Reference Goebel, Grabka, Liebig, Kroh, Richter, Schröder and Schupp2019; Richter and Schupp Reference Richter and Schupp2015).Footnote 10
Table 1. Summary Statistics for County-Level Variables in 2013
Note: The table shows summary statistics for aggregate-level data in 2013.
Definition and Measurement of Brown Occupations
Our analyses require us to distinguish employees in brown- and non-brown occupations. To do so, we rely on brownness scores for occupations provided by Vona et al. (Reference Vona, Marin, Consoli and Popp2018). Vona et al. (Reference Vona, Marin, Consoli and Popp2018) first define the most polluting industries as those in the 95th percentile of pollution intensity for at least three out of eight relevant pollutants (CO2, CO, VOC, NOx, SO2, PM10, PM2.5, and lead). Not all of these pollutants are greenhouse gases, but other pollutants have also been targeted with environmental regulations and criticized by the AfD. For example, Germany has adopted restrictions on diesel engines in order to comply with EU NOx emission rules. A July 29th AfD press release warned that such diesel restrictions are: “all about one thing: putting an end to the most successful German industrial product of all time.”Footnote 11
Vona et al. (Reference Vona, Marin, Consoli and Popp2018) define brown occupations as occupations that are relatively more prevalent in these polluting industries. Specifically, this means that an occupation is classified as “brown” if its prevalence in polluting industries is at least seven times higher than its prevalence in any other industry. The original classification by Vona et al. (Reference Vona, Marin, Consoli and Popp2018) derives from the American context. Therefore, we use crosswalks provided by Cavallotti et al. (Reference Cavallotti, Colantone, Stanig and Vona2025) to adapt them to the European context. This mapping results in a continuous measure of the “brownness” of all occupations in the ISCO-08 (International Standard Classification of Occupations) classification. This measure ranges from 0 to 1 and represents the occupation-specific likelihood of being employed in a polluting sector. Because this classification is based on the U.S. labor market, we assume that industry-specific pollution and occupational distribution across industries in the US resemble those in Germany. We believe that this assumption is justified, since production technologies and industrial trends are similar in the US and Germany (Baily, Bosworth, and Doshi Reference Baily, Bosworth and Doshi2020).
Our main brownness scores allow us to distinguish employees in brown- and nonbrown jobs for all ISCO occupations. To then create measures of brown employment at the county level, we obtained administrative labor market data from the German Federal Employment Agency (Bundesagentur für Arbeit). These data measure the share of workers who are employed in these brown occupations in each county for the period 1996–2022. In all county-level analyses, our main treatment is measured as the share of people working in brown occupations in 2013—that is, prior to the AfD platform change. This measure is constructed as follows:
$$ \begin{array}{rl}{D}_c=\frac{{\displaystyle \sum_{a=1}^A}{N}_{ac}\times 1({\mathrm{brownness}}_a>0.3)}{{\displaystyle \sum_{a=1}^A}{N}_{ac}},& \end{array} $$
where
$ {N}_{ac} $
measures the number of people working in a particular occupation a (ISCO-08) in a county c in 2013 and
$ {\mathrm{brownness}}_a $
measures the particular brownness score of that occupation. Following Cavallotti et al. (Reference Cavallotti, Colantone, Stanig and Vona2025), we code occupations as brown if their brownness score is greater than 0.3. Jobs above the threshold can be considered as brown (e.g., sheet-metal workers and metal working machine tool setters), whereas jobs with scores below 0.3 are more ambiguous (e.g., butchers, fishmongers, and food preparers). As we discuss in the “Robustness” section, our main results are not dependent on this threshold—we find similar results when choosing alternative thresholds and an alternative treatment that does not dichotomize the brownness score.
We further probe the robustness of our results to an alternative measure based on German emissions and employment data developed by Südekum and Rademacher (Reference Südekum and Rademacher2024). This variable measures the share of full-time employees who work in industries with CO
$ {}_2 $
emission increases between 2000 and 2019 for each county in the year 2019 (see Section C.1 of the Supplementary Material for more details). Our primary measure, derived from Vona et al. (Reference Vona, Marin, Consoli and Popp2018), identifies “brown jobs” based on specific occupations prevalent in highly polluting industries. This contrasts with industry-based approaches, such as that of Bechtel, Genovese, and Scheve (Reference Bechtel, Genovese and Scheve2019), which classify exposure based on industry-level CO2 equivalent emissions regardless of the specific occupation within that industry.Footnote 12 In our setting, the Südekum and Rademacher (Reference Südekum and Rademacher2024) variable serves as an industry-based alternative measure to our main occupation-based approach. It further has the advantage of using German-specific data. As we show in Figure F.6 in the Supplementary Material and discuss in the “Robustness” section, our main results are very similar when using this alternative treatment.
We present the distribution of brown jobs across counties in 2013 in Figure 2. High-emission jobs cluster in the highly industrialized regions in Baden-Württemberg, North Rhine-Westphalia, and Thuringia, and in the German chemical hubs, such as in Ludwigshafen which hosts the headquarters of chemical company BASF, and Altötting, which is at the center of the “Southeast-Bavarian chemical triangle.” We further list the most common high-emission jobs in the SOEP panel data in Table E.1 in the Supplementary Material. This set of jobs includes machine makers and operators in the agricultural, chemical, and manufacturing industries.
Figure 2. Geographic Distribution of Brown Jobs in Germany in 2013, County-Level
Note: The figure shows the share of brown jobs in each county in 2013. To better convey geographic variation, we use a log scale. Shapefiles from Bundesamt für Kartographie und Geodäsie (2022).
Individual-Level Data
When analyzing individual-level panel data, we rely on individual-level brownness scores using the ISCO-88 and ISCO-08 indicators in the SOEP panel data.Footnote 13 As with the county-level analysis, we create a binary indicator for brown occupation (brownness
$ > $
0.3). For each employed survey respondent, this indicator measures whether the respondent works in a relatively more “brown” occupation. The binary treatment indicator distinguishes the individual-level analysis from the county-level analysis. In the county-level analysis, we use the share of employees for which brownness is above 0.3.
Political Outcomes at the Aggregate and Individual Levels
Aggregate-Level Outcomes
For the county-level analyses, we compile data on German federal election results at the county level between 2005 and 2021. The election data used in this project have been published as part of the German Election Database (GERDA) (Heddesheimer et al. Reference Heddesheimer, Hilbig, Sichart and Wiedemann2025). Some counties experience border changes and mergers during this period. To obtain a balanced panel of comparable counties, all counties that experience boundary changes are mapped to their 2021 boundaries using population-weighted crosswalk files provided by the Federal Institute for Research on Building, Urban Affairs, and Spatial Development (BBSR; Bundesinstitut für Bau-, Stadt- und Raumforschung 2024; see Section C.2 of the Supplementary Material and the accompanying data paper [Heddesheimer et al. Reference Heddesheimer, Hilbig, Sichart and Wiedemann2025] for more details).
Individual-Level Outcomes
When analyzing individual-level panel data, we use a survey item that measures individuals’ party preferences. In particular, participants are asked “Which party do you lean toward?” We construct binary variables for each party that take the value 1 for a respective party if the respondents indicate that they lean toward that party, and 0 otherwise. Note that the party preference variable has substantial missingness, with 77% of observations missing across all years, ranging from 67% in 2003 to 85% in 2021.
County-Level Covariates
At the county level, we collect data for several covariates. In particular, we collect data on the total county population and population density from the BBSR (Reference Heddesheimer, Hilbig and Wiedemann2024). We further obtain data on the proportion of manufacturing workers (as a percentage of the working population) from the Federal Statistical Office of Germany (Statistische Ämter des Bundes und der Länder 2024). Information on average household income is derived from the national economic accounts of the federal states (Volkswirtschaftliche Gesamtrechnung der Länder; Statistische Ämter der Länder 2023). We also collect county-level GDP per capita data, which are available from the Annual Regional Database of the European Commission’s Directorate General for Regional and Urban Policy (ARDECO; European Commission 2024). Finally, we obtain data on unemployment rates and the share of people who graduate with a higher education entrance qualification (Hochschulreife) in a given year from Inkar (Bundesinstitut für Bau-, Stadt- und Raumforschung 2023).
RESEARCH DESIGN
Our empirical strategy follows directly from the theory. Electoral realignment should occur only when two conditions are met: (i) voters are exposed to the green transition, and (ii) at least one party offers an anti-transition platform. Exposure is concentrated in places—and among individuals—whose employment is dominated by “brown” occupations. The supply condition is met only in 2016, when the AfD begins opposing the transition (see the section on the AfD shift above). Therefore, our empirical “treatment” is the interaction of (i) sufficient demand for an anti-transition platform and (ii) the supply of such a platform by the AfD party. Importantly, our design merely requires that the AfD is the single party supplying the anti-transition platform. We do not assume convergence between the AfD and mainstream parties on immigration, economic, or foreign-policy issues (see Section A.1 of the Supplementary Material).Footnote 14
To convert our theoretical argument into an empirical test, we rely on a DiD design. In the aggregate analysis, a county’s 2013 brown employment share (
$ {D}_c $
) represents its exposure to the green transition’s labor market impacts (latent demand). The supply-side shift occurs in 2016. Our DiD design evaluates the effect of this supply-side shift conditional on the pre-existing demand characteristic, by examining the interaction
$ {D}_c\times {1}_{t=k} $
for post-2016 elections. Formally, the treatment variable is the interaction between the pre-2013 brown share and a post-2016 indicator: it equals zero everywhere before 2016 and, thereafter, increases in proportion to a county’s brown exposure. Regions with low brown employment serve as the never-treated control group; we compare their pre- and post-2016 electoral trends with the corresponding trends in high-brown regions. Because supply varies solely over time and demand solely across space, the design identifies the joint effect of the two conditions. The logic is analogous at the individual level: respondents in polluting occupations are coded as treated only in survey waves from 2016 onward, with 2015 providing the last pretreatment observation.
County-Level Analysis
To examine aggregate-level changes in voting behavior, we estimate a specification of the following form:
$$ {Y}_{cjt}={\displaystyle \begin{array}{l}{\gamma}_c+{\delta}_{jt}+\sum_{k=-3}^1{\theta}_k\times {D}_c\times {1}_{t=k}+\sum_{k=-3}^1{\mu}_k\times {M}_c\\ {}\times {1}_{t=k}+\sum_{k=-3}^1{\pi}_k\times {P}_c\times {1}_{t=k}+{\beta}^{\prime }{X}_{ct}+{\varepsilon}_{cjt}.\end{array}} $$
Here, c indexes counties, t indexes elections, and j indexes administrative districts.Footnote 15 The outcome
$ {Y}_{cjt} $
is the vote share of a given party in county c and election t.
As discussed above, the baseline exposure
$ {D}_c $
is the share of people working in brown occupations in county c in 2013 (i.e., prior to the AfD platform shift). We interact
$ {D}_c $
with post-2016 election dummies, so the treatment is “switched on” only once an anti-transition supply option appears. We are interested in the parameters
$ {\theta}_k $
, which are the DiD estimates for the
$ {k}^{th} $
election before or after treatment. Here, k indexes elections relative to 2017, with
$ k=0 $
indicating the 2017 election,
$ k=-1 $
indicating the 2013 election, etc. The term
$ {\gamma}_c $
is a county fixed effect, and the term
$ {\delta}_{jt} $
is an administrative district-specific election fixed effect. The latter accounts for time-varying confounding at the level of the administrative district (Regierungsbezirk) level. As is standard practice in event study specifications, we omit the interaction between
$ {D}_c $
and the last pretreatment period election (2013, which corresponds to
$ k=-1 $
). The 2013 election serves as the baseline for all estimated treatment effects. Standard errors are clustered by county.
The vector
$ {X}_{ct} $
contains the following time-varying control variables, all measured at the county level: average household income, GDP/capita, the share of workers employed in the manufacturing sector, population density, total population, the share of people with a higher education entrance qualification, the share of foreigners, and the unemployment rate. We also include additional terms that interact the share of employees in manufacturing firms (
$ {M}_c $
) as well as population density (
$ {P}_c $
) with election dummies. These terms serve to account for the fact that brown jobs and manufacturing jobs may overlap, and that brown jobs are more prevalent in rural areas. The
$ {M}_c\times {1}_{t=k} $
term allows for separate trends in electoral outcomes, comparing counties with larger or smaller shares of employees in manufacturing. Similarly, the
$ {P}_c\times {1}_{t=k} $
term allows for differential trends when comparing urban and rural areas.
Individual-Level Analysis
In a second step, we assess whether (i) party preferences differ between individuals in brown and nonbrown occupations and (ii) whether this relationship varies over time. To do so, we estimate a specification of the following form:
$$ \begin{array}{rl}{Y}_{it}={\gamma}_i+{\delta}_t+{\displaystyle \sum_{k=2010}^{2021}}{\theta}_k\times {D}_{it}\times {I}_{t=k}+{\beta}^{\prime }{X}_{it}+{\varepsilon}_{it}.& \end{array} $$
Here,
$ {Y}_{it} $
is the reported party preference for individual i in year t. The term
$ {D}_{it} $
indicates if individual i holds a brown occupation in year t, our demand-side characteristic. The model includes individual fixed effects (
$ {\gamma}_i $
) and year fixed effects (
$ {\delta}_t $
), alongside controls
$ {X}_{it} $
(age, education, income, immigration concerns). The coefficients
$ {\theta}_k $
capture the interaction
$ {D}_{it}\times {I}_{t=k} $
, which is the differential party preference for those in brown jobs in year k relative to a baseline year (e.g., 2015, pre-AfD platform shift) and compared to those not in brown jobs. This allows us to examine preference shifts around the AfD’s 2016 platform change.
Changing Sample Composition
Equation 3 is similar to a standard event study specification, but differs with respect to the term
$ {D}_{it} $
. In a typical event study specification, the treatment indicator
$ {D}_{it} $
is constant across t. In our setting, the specification allows for over-time variation in whether individuals are employed in a brown job, and therefore, means that the different coefficients
$ {\theta}_k $
are not estimated using the same groups of people for each year t. Variance in the composition of the groups defined by
$ {D}_{it} $
stems from individuals who switch in or out of brown occupations. In addition, there is some attrition—that is, not all employed individuals observed in 2015 are still observed in the following years.
To address the issue of compositional differences between the groups defined by
$ {D}_{it} $
, we use three strategies. First, we include several time-varying covariates
$ {X}_{it} $
, chiefly income and worries about immigration. We note that individual-level time-invariant characteristics, such as gender, education, and migration background, are already taken into account by including individual-level fixed effects
$ {\gamma}_i $
. Second, we present additional specifications that include weights, which serve to keep the composition of the groups defined by
$ {D}_{it} $
similar across years t. We provide more details on the weighting procedure in Section D of the Supplementary Material, where we also assess the resulting improvement in over-time covariate balance through a comparison of covariate averages over time in Figure D.1 in the Supplementary Material. Although some imbalance remains, this approach significantly improves balance between the brown and nonbrown groups over time. As we discuss in the “Robustness” section, our main results are robust to the inclusion of these weights. Third, we present results from a specification that fixes the value of
$ {D}_{it} $
to its value in 2015. This means that, for this additional analysis,
$ {D}_{it} $
for respondent i is constant across all t. We again find comparable results to our main specification (see Tables G.4 and G.5 in the Supplementary Material).
Aggregate-Level Treatment and Individual Outcomes
In addition to the individual-level treatment discussed above, we further present results in which we assess how individual-level party preferences vary when comparing individuals in counties with higher or lower shares of workers in brown occupations. We provide more details on this analysis in Section G.3 of the Supplementary Material.
RESULTS
County-Level Analysis: Brown Employment and Electoral Results
As described in the “Research Design” section, we expect electoral effects for regions with higher brown employment shares after 2015, when the AfD changes its platform. The corresponding results are shown in Figure 3. The figure shows the relationship between the share of brown employment in 2013 and the change in vote shares between 2013–2017, as well as 2013–2021, at the county level. Since we use a continuous treatment, the control group can be thought of as counties with lower brown employment shares. This means that we present the estimated coefficients
$ {\theta}_0 $
and
$ {\theta}_1 $
from Equation 2.
Figure 3. County-Level Brown Employment in 2013 and Electoral Results
Note: The figure presents the DiD results from the main specification. In particular, we present the coefficients
$ {\theta}_0 $
(for 2017) and
$ {\theta}_1 $
(for 2021) from the specification in Equation 2. Covariates are listed in the “Research Design” section. Standard errors are clustered at the county level. The “far-right parties” term is the sum of all far-right parties, which includes the AfD. For details on the results, see Table F.1 in the Supplementary Material.
In 2017, we find that counties with a higher share of brown jobs have experienced an increase in support for the AfD and the SPD, and a decrease in support for the Left Party, the FDP, and the CDU/CSU, compared to counties with fewer brown jobs. The coefficient for the Greens is negative but insignificant and small in 2017. In 2021, these patterns of differential change are similar, with the main differences being that differential losses in areas with brown occupations for the Green party are considerably larger, while differential gains for the AfD remain comparable to 2017. The 2021 elections similarly saw differential losses for the Left Party in these counties, whereas such losses for the FDP are not observed in 2021.
Apart from the strong positive differential effects for the AfD, we also observe positive but small differential gains for the SPD. This may be explained by the SPD’s traditional alignment with workers in high-emission sectors and their more cautious stance on the green transition, making them a less direct target of backlash than the Greens but perhaps a partial beneficiary in areas more exposed to the transition.Footnote 16 The estimates we present in Figure 3 are consistent with a visual inspection of the relationship between brown employment shares and far-right support, which we present in the form of scatter plots in Figure 4 and Figure E.1 in the Supplementary Material. The bivariate evidence in these figures demonstrates a clear positive relationship between the change in AfD voting after 2013 and levels of brown employment in 2013.
Figure 4. Change in AfD Support 2013–2021 and County-Level Brown Employment in 2013
Note: The figure shows the county-level shares of brown occupations in 2013 and the change in AfD support between 2013 and 2021, measured in percentage points. We highlight select counties.
In the Supplementary Material, we present event study estimates in Figure F.1. Here, we show estimated coefficients
$ {\theta}_k $
from Equation 2. As is standard in DiD designs, we use the estimated coefficients for the periods prior to the 2016 supply-side shift (i.e., interactions of
$ {D}_c $
with election year indicators before 2017, relative to the 2013 baseline) to assess the parallel trends assumption. Regarding our main outcome, voting for the AfD party, we cannot directly assess trends prior to 2013, since the AfD did not compete in earlier elections. We instead analyze the combined vote share of far-right parties, since other far-right parties competed before 2013 (see Section C.4 of the Supplementary Material for more information on these parties). We observe coefficients that are statistically indistinguishable from zero for these pre-2016 periods, suggesting that, with respect to radical-right voting, counties with high shares of brown jobs were not on a diverging electoral trajectory before the AfD’s anti-green platform became salient. Since not all pretreatment estimates across all parties are consistently zero and insignificant, we conduct an additional sensitivity analysis for parallel trend violations (see below and Section F.1 of the Supplementary Material).
We further present descriptive vote share trends over time in Figure E.3 in the Supplementary Material. Here, we split the sample into above- and below-median brown employment shares. The evidence in this figure largely confirms the main results presented in Figure 3 and Figure F.1 in the Supplementary Material. We observe that the AfD gains differentially more votes in counties with above-median brown employment, while the Green party underperforms in areas with higher brown employment shares. For more information, see Section E.2 of the Supplementary Material.
Finally, we present additional results based on an operationalization of the treatment as the change in county-level brown employment shares between 1998 and 2013. This serves to capture long-term structural shifts predating the AfD’s 2016 platform change. Figure F.7 in the Supplementary Material suggests that counties with larger preshift declines in brown employment saw greater increases in AfD support (and decreases for the Greens) post-2013. This suggests that past exposure to employment declines increased susceptibility to the AfD’s new platform. However, we find some evidence for parallel trend violations in this specification, and therefore, consider this finding mainly suggestive.
Contextualizing Effect Size Magnitude
To contextualize the magnitude of the observed effect, we compare the standardized coefficient for baseline brown employment share to those of other common socioeconomic predictors of AfD vote change since 2013 (see Section F.10 of the Supplementary Material for details). We standardize all baseline county-level covariates from 2013 and interact them with election year indicators in our DiD framework. It is important to note that these coefficient estimates can only be interpreted as correlations—we are not identifying distinct causal effects for each of these covariates. These covariates represent established cleavages in voting behavior, including measures of economic disadvantage (unemployment rate, household income, and GDP per capita) and educational attainment (share of graduates with Abitur), as well as industrial structure (manufacturing share) and urban-rural differences (population density). While unemployment and educational attainment show particularly large conditional magnitudes in predicting changes in AfD support, the effect size associated with baseline brown employment is sizable. Its magnitude is larger than that of manufacturing share, household income, and GDP per capita, and comparable to population density. This additional evidence confirms that employment in brown occupations is a relevant predictor of the differential increase in AfD support.
East-West Differences
In the German context, a common source of heterogeneity is the East-West comparison. Figure 5 shows results from our main specification, focusing on either East or West Germany. The findings are generally similar: the AfD gains and the Green Party loses, especially in 2021. Figure 5 further indicates that our main results in Figure 3 are not driven by larger increases in far-right support in East Germany. This is visually confirmed in Figure 4, which shows a positive relationship between brown job prevalence and increases in AfD support for the 2021 election in both East and West Germany.
Figure 5. County-Level Brown Employment and Electoral Results After 2013—Comparing East and West Germany
Note: The figure presents the results from the main specification, separately for East and West Germany. In particular, we present the coefficients
$ {\theta}_0 $
(for 2017) and
$ {\theta}_1 $
(for 2021) from the specification in Equation 2. Covariates are listed in in the “Research Design” section. Standard errors are clustered at the county level. The “far-right parties” term is the sum of all far-right parties, which includes that AfD. For details on these results, see Tables F.2 and F.3 in the Supplementary Material.
Sensitivity to Possible Parallel Trends Violations
As mentioned above, we observe some evidence for violations of the parallel trends assumption prior to 2013 in Figure F.1 in the Supplementary Material. However, the magnitude of the DiD estimates prior to 2013 for these parties is small, even though some estimates are significantly different from zero.
To assess the severity of potential parallel trends violations, we implement the sensitivity analysis proposed by Rambachan and Roth (Reference Rambachan and Roth2023). This test allows us to determine whether our main results hold under the assumption that post-2013 parallel trends violations are as severe or more severe than those observed prior to 2013. We present this analysis in Figure F.5 in the Supplementary Material and Section F.1 of the Supplementary Material. Regarding the far-right party outcome, we find that our results are robust to potential parallel trend violations, even when these violations are larger than those observed before 2017. This finding aligns with the fact that the pre-2013 estimates in Figure F.1 in the Supplementary Material are generally much smaller in magnitude compared to those observed in 2017 and 2021. Moreover, a comparison with a recent meta-analysis shows that the far-right results are more robust to parallel trends violations than many previously published studies employing DiD approaches (Chiu et al. Reference Chiu, Lan, Liu and Xu2025). Therefore, we conclude that potential parallel trends violations before 2017 do not invalidate our main results.
Alternative Explanations
Before moving on, we assess whether our aggregate-level measure picks up other explanations for geographic variation in changes in AfD support after 2013.
First, an alternative economic hypothesis for the rise of the far right is the trade-induced loss of manufacturing jobs (Dippel et al. Reference Dippel, Gold, Heblich and Pinto2022). Our main specification already considers this by allowing for trends in voting behavior that vary with manufacturing employment. Figure F.9 in the Supplementary Material presents results based on adding a term that additionally interacts changes in manufacturing employment with election indicators. This allows for differential trends in AfD support correlated with local shifts in manufacturing. We find that our main results remain unchanged after accounting for differential trends due to changes in manufacturing employment between 2013 and 2017.
Second, aggregate brown employment shares may be correlated with pre-existing anti-immigrant sentiments or xenophobia. Since one dominant political topic after 2013 was the inflow of refugees starting in 2015, our main results may reflect the activation of xenophobia in regions with higher brown employment. To assess whether this is the case, we obtain data on pre-existing levels of xenophobia. First, we measure the number of anti-refugee violent incidents and demonstrations in each county in 2015, obtained from Benček and Strasheim (Reference Benček and Strasheim2016) (see Section C.3 of the Supplementary Material for a description of the data). In our analysis, we use two versions of this measure—one scaled by population and one scaled by the total number of crimes in county c in 2015. We then re-estimate our main specification, allowing trends in party vote shares to vary by pre-existing levels of anti-refugee violence and demonstrations. We further interact election indicators with 2013 levels of far-right support, which should similarly proxy for anti-immigrant sentiments. In Figure F.10 in the Supplementary Material, we again find that the main effect of brown employment is robust to these additional tests. This implies that our main effects are unlikely to be driven by more negative responses to the refugee inflow in areas where brown occupations are more prevalent.
Third, the increase in refugee presence per capita after 2013 may have been greater in areas with more brown employment. Natives might have responded to larger refugee inflows by voting for the AfD. Although this may appear possible, the allocation mechanism for refugees in Germany precludes this explanation. As discussed in the literature on refugee integration in Germany (see, e.g., Gundacker, Kosyakova, and Schneider Reference Gundacker, Kosyakova and Schneider2024), refugee distribution to federal states and counties is proportional to the local population. Therefore, the increase in refugee exposure after 2013 should not vary strongly across counties and should therefore not correlate with brown employment shares.
Fourth, we can exclude the possibility that urban-rural differences drive our main results. As discussed in the “Research Design” section, our main specification already allows for differential trends in vote shares that are correlated with population density. This accounts for the fact that brown jobs are more prevalent in rural areas. In a related robustness check, we obtain data from Ziblatt, Hilbig, and Bischof (Reference Ziblatt, Hilbig and Bischof2024) on dialects as a proxy for historical center-periphery divides, and then interact this variable with election fixed effects. We find that allowing for differential trends in areas with stronger or weaker dialects does not change our main results (see Figure F.11 in the Supplementary Material).
Fifth, we contend that Euroscepticism is likely not a driver of our main results. The AfD was founded on a strongly Eurosceptic platform but later broadened its agenda toward national-conservative and anti-immigrant positions (see Section A of the Supplementary Material). Accordingly, the relative salience of Euroscepticism decreased between 2013 and 2017 (see evidence in Figure A.3 in the Supplementary Material). We can therefore rule out a mechanism where (i) regions with higher brown employment are more inclined toward Eurosceptic positions and (ii) consequently pivot toward the AfD as it becomes more Eurosceptic. While condition (i) may hold, condition (ii) is ruled out by the evidence discussed in the above section on the AfD party platform shift.
Finally, it is worth noting that the inclusion of various controls as well as region-specific time trends does decrease the magnitude of the estimated main coefficients (see Table F.4 in the Supplementary Material). For 2017, the magnitude of the coefficient for the far-right voting outcome is about 18.5% lower in specifications with covariates and region-specific time trends than in the base specification that only includes the interaction between elections and 2015 brown employment levels. For 2021, the magnitude decreases by about 31.6%. This suggests that there is some confounding due to, for example, variation in manufacturing employment, xenophobia, or urban-rural differences that are correlated with brown employment shares. However, this confounding is not sufficient to fully explain our main effects in Figure 3 and Table F.1 in the Supplementary Material.
Targeting of Regions with High Brown Employment Shares
We further examine whether the AfD strategically targeted areas with high brown employment shares following its 2016 platform shift, potentially by investing more resources or fielding different types of candidates. To measure this, we use data from the candidate survey component of the German Longitudinal Election Study (GLES) on whether the AfD fields candidates in national and local elections, as well as on candidate characteristics: whether AfD candidates themselves work in brown occupations, and their socio-economic status (GLES 2014; 2018; 2023). We then employ a DiD approach similar to our main specification, comparing areas with higher or lower brown employment shares across the 2013 and 2017 elections.
We find no clear evidence for differential changes in targeting when comparing 2013 and 2017. It does not seem to be the case that there is a differential increase in the number of candidates fielded, and there is similarly no strong evidence that the AfD nominates different kinds of candidates in areas with higher brown employment shares. For more information on variables, data sources, estimation, and results, see Section F.11 of the Supplementary Material. We caution, however, that our set of variables likely does not cover all possible measures of geographic targeting. Notably, we do not have sufficient coverage for variables like district- or county-level campaign spending or more nuanced candidate characteristics.Footnote 17 We can therefore not conclusively determine whether or not the AfD specifically targeted regions with higher brown employment shares.
Individual-Level Panel Analysis: Brown Jobs and Political Preferences
In the previous section, we have established that counties with a higher share of workers in brown jobs experienced differentially larger increases in support for the AfD following the party’s 2016 platform shift (as observed in federal elections after 2013, relative to the 2013 baseline). We now use individual-level panel data to examine whether we see similar interactive effects when comparing individuals in brown versus nonbrown occupations around this same supply-side change.
In Figure 6, we present results from Equation 3. The figure displays the estimated yearly interaction coefficients (
$ {\theta}_k $
), which capture the differential party preferences for individuals in brown occupations (
$ {D}_{it}=1 $
) each year, relative to those not in brown occupations and benchmarked against a pre-2016 baseline period. We find that this differential support for far-right parties (mainly the AfD) among individuals in brown occupations increases markedly and significantly starting in 2016—the year of the AfD’s platform shift. Concurrently, for this same group of individuals in brown occupations, we observe losses for the CDU/CSU, as well as suggestive evidence for losses among supporters of the Green Party and Left Parties.
Figure 6. Brown Occupations and Individual-Level Partisan Support Over Time
Note: The figure shows estimated yearly interaction coefficients (
$ {\theta}_k $
) from Equation 3, representing the differential partisan support for individuals in brown occupations relative to a baseline year and to those not in brown occupations. All models include base covariates. Standard errors are clustered at the individual level. For details on these results, see Table G.1 in the Supplementary Material.
As discussed previously, the fact that individuals can and do switch out of brown occupations means that the composition of the groups defined by brown and nonbrown employment is not constant over time. The results presented in Figure 6 and Table G.3 in the Supplementary Material address this through the inclusion of time-varying covariates and individual fixed effects. In Tables G.4 and G.5 in the Supplementary Material, we further show that the main result is robust to the inclusion of weights that (partially) account for compositional differences (see also Section D of the Supplementary Material for more details on the weights).
Heterogeneous Treatment Effects
We further assess whether the individual-level effects vary across dimensions that are often shown to be salient for vote choice: gender, age, income, and education. In Table G.7 in the Supplementary Material, we find that the shift toward the AfD is more pronounced among respondents older than 50 years as well as among men. We further estimate heterogeneity by prior concerns about job security, respondents’ economic situations, and migration, but find no evidence for effect moderation (see Table G.8 in the Supplementary Material).
Our results on effect heterogeneity by gender align with recent work by Bush and Clayton (Reference Bush and Clayton2023), who posit that men in wealthy countries are more sensitive to the costs of climate change mitigation than women. Bush and Clayton (Reference Bush and Clayton2023) propose that this can partially be explained by the fact that men are more likely to work in pollution-intensive occupations. Notably, the results we present in Table G.7 in the Supplementary Material suggest that political responses among men are stronger even when holding constant the type of occupation. In our setting, gender differences in responses to climate change mitigation may therefore be primarily related to what Bush and Clayton (Reference Bush and Clayton2023) term “perceived psychological costs” of climate change mitigation.
Aggregate-Level Brown Employment and Individual-Level Outcomes
Since we have information on the place of residence of SOEP survey respondents, we can further examine whether individuals in counties with higher shares of brown employment also report changes in party support even if they themselves are not employed in brown occupations. We present the results from this analysis in Section G.3 of the Supplementary Material. We find that, similar to the aggregate-level results, support for the AfD increases among individuals living in regions with high shares of brown jobs after 2015. Notably, the magnitude of these results is similar to what we find at the aggregate level. Based on the SOEP panel survey results, a one percentage point increase in brown employment shares is associated with an increase in AfD party support of between 0.15 and 0.25 percentage points. At the aggregate level, the effect size is between 0.22 and 0.27 percentage points.
We further note that individual-level panel data in conjunction with the aggregate treatment addresses typical issues with ecological analyses (this is also noted by Gazmararian and Krashinsky Reference Gazmararian and Krashinsky2023). For example, the individual-level results show that our aggregate-level results do not stem purely from turnout changes or compositional differences among the electorate. Rather, we observe that within-individual vote switching can at least partially explain why regions with higher brown employment shares shift toward the radical right.
Comparing Aggregate and Individual Effects
Because we present panel evidence at both the aggregate and individual levels, we can directly examine whether individual-level vote shifts are sufficient to explain aggregate patterns. At the individual level, we find that workers in high-emission jobs are 2–6 percentage points more likely to support the AfD after 2015 (see Figure 6). That is, among hundred employees in these jobs, the AfD attracts 2–6 additional voters compared to their counterparts in nonbrown occupations. When examining county-level vote shares, we find that a one percentage point increase in the proportion of brown jobs within a region in 2013 has an effect of approximately 0.22 to 0.27 additional percentage points in AfD support (see Figure 3 and Table F.1 in the Supplementary Material).
As detailed in the back-of-the-envelope calculation in Section G.4 of the Supplementary Material, comparing these individual and aggregate effect magnitudes reveals a notable discrepancy. If the aggregate effects are solely due to vote switching among employees in brown occupations, the aggregate effects suggest that 22–27 out of every hundred employees in brown jobs would have had to shift their vote to the AfD.
The discrepancy between these figures suggests that the aggregate impact cannot be entirely attributed to shifts in voting behavior among those working in brown occupations. Therefore, one implication of our results is that not only individuals in brown jobs, but also those in nonbrown occupations within areas more reliant on high-emission industries switch toward the far right.Footnote 18
ROBUSTNESS
We now conduct a number of robustness checks for both aggregate- and individual-level analyses.
Aggregate-Level Results
First, we show that our results are robust to alternative definitions of the aggregate-level treatment in Figure F.3 in the Supplementary Material. Second, we show that the results are not driven by one specific state—excluding states one-by-one does not change our main conclusions (see Figure F.4 in the Supplementary Material). Third, we re-estimate the main specification using population weights and show that our results remain largely similar to the results in the main specification (Figure F.8 in the Supplementary Material). Fourth, we re-estimate our main specification using the Sun and Abraham (Reference Sun and Abraham2021) estimator. This estimator accounts for issues that arise from treatment effect heterogeneity and is unbiased under a specific parallel trends assumption as well as assuming no anticipation for the comparison group (see also Chiu et al. Reference Chiu, Lan, Liu and Xu2025). This check further serves to show robustness when using a binary treatment, which is required for the Sun and Abraham (Reference Sun and Abraham2021) estimator. In Figure F.2 in the Supplementary Material, we find that our conclusions are not affected by this additional robustness check.
Fifth, we show that our results are not dependent on the inclusion of specific fixed effects or covariates. In Table F.4 in the Supplementary Material, we find that results are similar using either state-election fixed effects or simply election fixed effects that do not vary by state or administrative district. We further show that the results remain similar when excluding the covariates listed in the “Research Design” section. Sixth, we use an alternative treatment constructed by Südekum and Rademacher (Reference Südekum and Rademacher2024). This variable measures the county-level share of full-time employees in industries that experienced an increase in CO
$ {}_2 $
emissions between 2000 and 2019 (see Section C.1 of the Supplementary Material for more information). The correlation between this variable and our main measure of brown job employment is relatively high at about 0.59. In Figure F.6 in the Supplementary Material, we find very similar results when replacing our main treatment with the Südekum and Rademacher (Reference Südekum and Rademacher2024) measure. Seventh, we account for local socioeconomic status using the International Socio-Economic Index (ISEI), derived from SOEP data (Gidron and Hall Reference Gidron and Hall2017). Our main results remain robust when including an interaction between ISEI and election year fixed effects (Figure F.12 in the Supplementary Material).
Individual-Level Results
In models 4 and 7 in Table G.4 in the Supplementary Material, we assess whether regional differences can account for the observed individual-level relationship between brown occupations and radical-right support. Akin to county-level specification, we do so by including state-specific time trends and find that our results remain largely unchanged. We further show that the results are unchanged when we do not use a time-varying treatment for working in a brown job, but instead use a treatment that is based on whether an individual had a brown job in 2015. In models 5–7 of Table G.4 in the Supplementary Material, we find that our results are, if anything, slightly stronger when we fix the brown job indicator to its 2015 value. We also assess whether differences in social status explain the observed relationship between brown occupations and far-right support. We predict time-varying social status using random forest models trained on socio-economic characteristics and the 2016 and 2018 social status questions. Including these predicted measures as time-varying controls does not attenuate the effect of brown job employment (Figure G.5 in the Supplementary Material).
MECHANISMS
We now examine potential mechanisms to explain our main results. Drawing on the theoretical framework established in the “Theory” section, we explore two primary mechanisms that may drive the political shift to the far right: perceptions of economic insecurity and a loss of social status. We also assess objective measures of economic decline. We analyze these mechanisms for both the community and the individual-level treatment. We note that the mechanism evidence we present is necessarily only suggestive, since the “intermediate outcome tests” we conduct require strong assumptions (Blackwell, Ma, and Opacic Reference Blackwell, Ma and Opacic2024). We draw on SOEP data to assess changes in attitudes related to economic insecurity and social status (see Section G.1 of the Supplementary Material for a more detailed description and Table G.2 for the corresponding survey questions). Table 2 summarizes our mechanism evidence.
Table 2. Evidence for the Mechanisms—Changes After 2015
Note: *For this specification, we fix treatment status to brownness status in 2015. †Evidence from pooled 2016/2018 cross-sectional analysis showing an association, not a DiD estimate of change from the pre-2015 baseline.
Figure 7. Association Between Individual- and Community-Level Brown Employment, and Perceived Social Status
Note: The figure shows the association between (i) having a brown job and (ii) county-level brown employment shares and two survey items that measure social status. The data are from the the 2016 and 2018 SOEP-IS waves, which we pool. The outcomes are standardized. We present cross-sectional evidence based on three specifications. The first only includes the individual or aggregate-level brown employment measure. The second (“Base controls and state FE”) includes age, education levels, sex, and state fixed effects. We further add the ISEI and income as controls in the “Additional controls and state FE” specification. For more information on the outcomes, see Table G.2 in the Supplementary Material. For details on the estimation, see Section G.6 of the Supplementary Material. For details on the results, see Tables G.9 and G.10 in the Supplementary Material.
First, we observe little to no evidence that survey respondents in counties with higher shares of brown employment or respondents in high-emission occupations report higher levels of economic insecurity after 2015. There are no changes in concerns about respondents’ own economic and financial situation for respondents working in brown occupations. At the community level, we only observe suggestive negative effects on concerns about individual job security. However, we caution that there is some evidence for diverging job security perceptions prior to 2015, which suggests that differences in post-2015 trends in job security perceptions may be unrelated to the AfD platform shift.
Second, we do not observe strong evidence for economic decline. While GDP per capita decreases in counties with higher shares of employees in brown occupations, there is no evidence for declines in unemployment and wages after 2015. Among workers in brown occupations, there is some evidence for declines in wages, but this is only apparent in 2020 and 2021—for earlier years, we do not observe clear income declines. This is consistent with the idea that economic losses connected to the green transition have largely not materialized yet.
Third, using cross-sectional data, we observe that self-perceived social status is lower for individuals in brown occupations and for individuals who reside in counties with high shares of polluting jobs (see Figure 7). These results are based on a smaller sample of SOEP respondents for which the status survey items are available (see Section G.6 of the Supplementary Material for a more detailed description). This means that standard errors are larger than for the main SOEP results, and we do not consistently find significant results. Yet, the results for the status items are stronger than for the survey items related to economic insecurity or observed changes in unemployment and income.
Although we do not find evidence that economic insecurity increases after 2015, the AfD platform change in 2016 may tap into pre-existing insecurity related to the green transition. Germany implemented ambitious green transition policies as early as 2000 when the Schröder government enacted the German Renewable Energy Act (Erneuerbare Energien Gesetz [EEG]). In a supplementary analysis in Section G.5 of the Supplementary Material, we examine whether counties with higher shares of brown employment saw increases in economic insecurity after the EEG was passed in 2000. We do not find much evidence that this is the case.
In sum, we do not find evidence for increases in economic insecurity after 2016, suggesting that a demand-side explanation alone (changing preferences among voters) is not sufficient for explaining increasing electoral support for the AfD and declining support for the Greens. However, it may be the case that insecurity in brown areas was already higher prior to the AfD platform shift. The AfD platform change may activate these pre-existing differences in economic insecurity, even if insecurity does not increase disproportionately during the 2010s in areas with higher shares of brown employment. However, we do not find strong evidence that increasing AfD support is accompanied by pronounced shifts in insecurity or objective economic factors. Instead, our mechanism evidence suggests an explanation related to perceptions of lower social status and stigmatization of brown occupations on the demand side, and the activation of pre-existing attitudinal differences through the AfD platform shift on the supply side.
CONCLUSION
The transition to sustainable energy sources is crucial to combating climate change. However, the shift from nonrenewable to renewable energy sources is not only an environmental and economic shift but also a profound social and political transformation. Workers in traditional, nonrenewable industries face economic and social difficulties, as transitioning to greener jobs is often challenging. We expect that workers in “brown” occupations and communities that depend on these jobs will favor political actors that oppose green transition policies, if such actors emerge and articulate this opposition.
We provide evidence for our argument from Germany, a country that has long pursued extensive green transition policies, and where the far-right AfD, particularly after its 2016 platform shift, became the sole vocal opponent of these policies. Using a DiD design, we show that the AfD’s 2016 anti-green platform shift had a differential electoral impact: counties with higher pre-existing shares of brown employment experienced larger increases in electoral support for the AfD. We find similar results when examining individual-level panel data: people in brown occupations disproportionately shifted toward the AfD following this same supply-side change. Importantly, we show that aggregate-level shifts toward the far-right AfD cannot be explained solely by political shifts among workers in brown occupations. Rather, they largely stem from preference changes among other individuals in communities, where brown employment is prevalent, suggesting community-level spillovers of this effect. What is more, the divergence in AfD support in areas with higher shares of brown jobs does not stem from pre-existing differential trends prior to the AfD’s 2016 platform change, underscoring the importance of the party’s strategic shift.
Additional evidence suggests that shifts in perceived economic insecurity starting in 2016 do not play a major role. Instead, we find suggestive evidence that concerns about social status are more pronounced among people in brown occupations and individuals residing in counties dependent on these jobs. The AfD platform shift may have therefore tapped into the stigmatization and resulting perceptions of lower status among individuals and communities most impacted by green transition policies.
Based on our findings, we propose several implications and potential avenues for future research. First, our study highlights the need for policymakers to consider the broader socio-political impacts of green transition policies, particularly on workers in affected industries. This calls for comprehensive strategies that not only promote environmental sustainability, but also address the economic and social challenges faced by these workers. Recent scholarship shows that compensatory policies can alleviate economic and political grievances (Bolet, Green, and González-Eguino Reference Bolet, Green and González-Eguino2024; Heddesheimer, Hilbig, and Wiedemann Reference Heddesheimer, Hilbig and Wiedemann2024) but it is less clear how well they work to ameliorate changes in perceived social status. Future research could test similar programs more broadly, focusing on other occupations and industries. This could include examining the role of retraining programs, economic incentives, and community engagement initiatives in mitigating the adverse effects of this transition. But our research also implies that compensation needs to look beyond purely economic incentives.
Second, our results raise the question of whether mainstream parties might have incentives to accommodate positions opposing the green transition. We posit that dissatisfaction among constituents translates into electoral change when there is a party that offers opposition. In our case, the opposing party is an anti-establishment party. However, prior evidence in the literature on immigration suggests that mainstream left parties may not face penalties when accommodating right-wing populist positions (Hjorth and Larsen Reference Hjorth and Larsen2022). Whether mainstream parties can gain from accommodation depends on the extent to which opposition to the green transition alienates existing supporters (Chou et al. Reference Chou, Dancygier, Egami and Jamal2021; Haas et al. Reference Haas, Stoetzer, Schleiter and Klüver2023). In Germany, a majority currently supports green transition policies,Footnote 19 which suggests that accommodation by mainstream parties may not be a viable strategy. Whether this remains the case likely depends on the economic repercussions of the green transition. If governments fail to implement economically inclusive transitions, a sufficiently large group of economically dissatisfied voters may lead to anti-transition accommodation among mainstream parties. Moreover, the Russian full-scale invasion of Ukraine has led to large increases in energy prices, which have further affected the competitiveness of energy-intensive industries. German businesses have largely refrained from endorsing the AfD (Bergmann et al. Reference Bergmann, Diermeier, Kinderman and Schroeder2024), but they may pressure mainstream parties into adopting stances against the green transition, especially when such policy positions appear more electorally attractive.
SUPPLEMENTARY MATERIAL
To view supplementary material for this article, please visit https://doi.org/10.1017/S0003055425101226.
DATA AVAILABILITY STATEMENT
Research documentation and data that support the findings of this study are openly available at the American Political Science Review Dataverse: https://doi.org/10.7910/DVN/2JZGDR. All raw and processed files used for the county-level election data are freely available in the replication folder. The individual survey data (SOEP), candidate analysis data (GLES), and party press analysis data are not openly available due to licensing restrictions. The replication folder contains detailed guidance on how to access and process these data, including specific contact information and access procedures for each restricted dataset.
ACKNOWLEDGEMENTS
We thank António Valentim, Fiona Bare, Italo Colantone, Robert Huber, as well as audiences at Bocconi, Georgetown, George Washington, Princeton, the Annual Meetings of the American Political Science Association, the European Political Science Association, and the Southern Political Science Association for helpful comments.
CONFLICT OF INTEREST
The authors declare no ethical issues or conflicts of interest in this research.
ETHICAL STANDARDS
The authors affirm this research did not involve human participants.
Open access
Comments
No Comments have been published for this article.