The Unequal Costs of Carbon Pricing: Economic and Political Effects Across European Regions

Layer 1: Overview

This paper asks whether carbon pricing through the EU Emissions Trading System (EU ETS) imposes economic costs that are unequally distributed across European regions, and whether those economic costs translate into political costs in the form of votes for extremist and populist parties. The motivation is both practical — political opposition has blocked or rolled back climate policies in several countries — and analytical: no prior study had systematically estimated the political consequences of carbon pricing at the subnational level.

The authors build a panel dataset covering 224 NUTS2 regions from 20 European countries (covering 97% of EU GDP, plus Norway) over 2000–2019. Economic data come from the European Commission’s ARDECO database; emission data from EDGAR (aggregate GHG) and the EU ETS Transaction Log (verified ETS emissions from regulated installations, mapped to NUTS2 via zip codes); voting data from the EU-NED dataset with party classifications from The PopuList. Household expectations are measured from 34 Eurobarometer survey waves (2004–2019). The dataset spans 114 elections (110 national, four European Parliament).

Identification rests on the carbon policy shocks of Kanzig (2023), constructed from high-frequency movements in EU carbon allowance futures prices around 126 regulatory events between 2005 and 2019, instrumented in a monthly VAR and aggregated to annual frequency. These shocks are orthogonal to contemporaneous economic conditions by construction, and are normalized so that the on-impact effect equals a 1% rise in Euro Area HICP energy prices. The main estimator is Jorda (2005) local projections in a panel with region fixed effects, lagged controls, and Driscoll-Kraay standard errors, estimated over a four-year horizon.

Main economic findings (average region): A 1%-energy-price-equivalent carbon shock reduces real GDP by approximately 0.7% — a contraction that persists for four years. Employment, real net disposable household income, real GVA, real compensation, real investment, and hours worked all decline significantly and persistently. GHG emissions fall by roughly 1% one year after the shock, confirming the policy’s effectiveness.

Main political findings: The combined extremist vote share (far-left plus far-right) rises by 0.3 to 0.4 percentage points two years after the shock and remains elevated. Populist and Eurosceptic vote shares also rise significantly in the medium term. Political fragmentation (1 minus the HHI) increases persistently. The shift is primarily toward far-right parties.

Survey-based expectations: The share of respondents citing environmental issues as a top concern falls by approximately 2 percentage points and remains depressed for four years. Respondents become significantly more pessimistic about national economic and employment prospects and their own financial situation.

Role of the economic channel: Using the Holm-Paul-Tischbirek (2021) decomposition, up to two thirds of the total rise in the extremist vote share over the four-year horizon is attributed to the decline in GDP, employment, and household income. The first year is more dominated by non-economic attribution effects (roughly 25% of the effect is explained by the economic channel at h=1), consistent with voters initially blaming the government’s policy choice rather than responding to realized economic deterioration.

Regional heterogeneity and inequality: Regions one standard deviation above mean ETS emission intensity experience a meaningfully larger output contraction and a 20–50% larger and more persistent rise in the extremist vote share relative to the average region. Regions receiving fewer free ETS allowances face analogously larger economic and political costs. The within-country 90–10 ratio of real disposable household income rises by approximately 0.05 percentage points, with widening concentrated at the lower tail (the median-to-10th-percentile gap), meaning poorer regions bear disproportionate costs. These heterogeneous effects imply that carbon pricing contributes to regional inequality within countries.

Policy implication: The EU ETS lacks direct redistribution mechanisms. The authors argue that progressive revenue recycling — household rebates calibrated to income — is necessary to cushion vulnerable regions, limit inequality, and rebuild public support for climate policy. These concerns are especially pressing given the EU ETS’s scheduled expansion to buildings and transportation in 2027.

Layer 2: Deep Dive

What is the identification strategy and what are the main threats to it?

The key identifying assumption is that the carbon policy shocks of Kanzig (2023) are exogenous with respect to regional economic conditions. The shocks are constructed from high-frequency daily movements in EU carbon allowance futures prices on days of regulatory announcements, relative to wholesale electricity prices on the prior day; the narrow event window ensures that confounding macroeconomic factors are already priced in. The shocks are then instrumented in a monthly VAR to extract structural shocks with a higher signal-to-noise ratio before being aggregated to annual frequency. The main threat would be if major regulatory announcements coincidentally coincided with other economic news. The authors defend against this by showing robustness to controlling for unemployment, stock market indices, monetary policy rates, oil prices, and a global financial crisis dummy. For the heterogeneity analysis, ETS intensity and free allowance share are fixed at their pre-sample values (end of ETS pilot phase, 2008) to rule out reverse causality from carbon pricing to the exposure measures.

How is the economic voting channel distinguished empirically from other channels?

The authors use the decomposition approach of Holm, Paul, and Tischbirek (2021). They re-estimate the extremist vote share local projection while controlling for the contemporaneous path of GDP, employment, and household income over the same h-year horizon. The residual coefficient on the carbon shock captures voting effects not attributable to economic deterioration. Comparing the controlled and uncontrolled responses shows that over the full four-year horizon, roughly two thirds of the voting increase is explained by economic variables. In the first year, the economic channel explains only about 25% of the response, consistent with non-economic attribution effects — voters blaming a government policy choice rather than an exogenous shock — being more prominent early on.

What additional evidence distinguishes ETS-driven political effects from other energy price effects?

Two benchmarks are used. First, national carbon taxes, which prior literature shows have muted economic effects, produce no statistically significant response in either real GDP or the extremist vote share (Appendix A.2), consistent with the economic channel being essential for the political response. Second, oil supply news shocks (Kanzig, 2021), constructed with a comparable high-frequency methodology and producing a similarly sized GDP decline, generate a statistically significantly smaller increase in the extremist vote share over the first two years (Appendix A.3). The excess political response to carbon shocks over oil shocks is interpreted as reflecting voters attributing policy-driven economic pain to the government, analogously to Gabriel, Klein, and Pessoa (2023) finding that austerity-induced recessions elicit stronger political responses than general downturns.

What heterogeneity across regions is documented and how is it measured?

Two exposure dimensions are explored. First, ETS emission intensity (verified ETS emissions scaled by GDP) captures direct agglomeration of installations covered by the carbon market. Second, the share of freely allocated ETS allowances relative to verified emissions captures the effective carbon price faced by firms in the region. Regions one standard deviation above mean ETS intensity experience meaningfully larger output and employment contractions, and 20–50% larger and more persistent increases in the extremist vote share. Regions with fewer free allowances bear analogously larger costs. Results hold when GHG intensity (covering non-ETS sectors) replaces ETS intensity, and when sectoral composition is controlled in the free allowance analysis. A country-level inequality analysis using local projections on the 90–10 ratio of regional household income shows that carbon pricing raises within-country dispersion by approximately 0.05 percentage points, driven primarily by widening of the lower tail (50th to 10th percentile gap), indicating that poorer regions suffer most.

What robustness checks are run?

Vote share results are robust to: (a) excluding parties coded as borderline by The PopuList; (b) excluding European Parliament elections and using only national elections; (c) averaging national and European election outcomes in years when both occur; (d) a minimal control set of only lagged dependent variable and region fixed effects; (e) an expanded control set adding country-level unemployment rate, stock market index, monetary policy rate, Brent oil price, and a GFC dummy variable. The inequality results are robust to using the 75–25 ratio and the Gini coefficient in addition to the 90–10 ratio. The heterogeneity results are robust to including time fixed effects, which absorb the aggregate carbon shock but preserve cross-sectional variation, confirming that heterogeneous responses are not driven by aggregate confounders. Driscoll-Kraay standard errors are used throughout to allow for cross-sectional and serial dependence; clustering at region-year level delivers nearly identical results.

How does this paper relate to and differ from closely related prior work?

Most directly related is Mangiante (2024), which documents that regions in poorer Euro Area countries are more exposed to carbon policy shocks. The present paper complements this by identifying within-country variation driven by ETS intensity and free allowance allocation, and by adding the political dimension. Kanzig and Konradt (2024) establish country-level economic effects of EU ETS shocks; this paper confirms those findings carry to the regional level and confirms comparable magnitudes. Gabriel, Klein, and Pessoa (2023) use the same econometric approach to study the political costs of austerity in European regions; the present paper finds analogous results for carbon pricing and attributes the political response similarly to economic deterioration. The finding that national carbon taxes lack economic or political bite echoes Metcalf and Stock (2023) and Konradt and Weder di Mauro (2023). The paper adds to the globalization-and-populism literature (Funke et al., 2016; Pastor and Veronesi, 2021; Colantone and Stanig, 2018) by identifying carbon pricing as another channel through which economic shocks drive extremist voting.

What is the direction of the political shift — toward far right or far left?

The decomposition in Appendix A.2 shows the increase in the combined extremist vote share is driven primarily by far-right parties. The far-right vote share rises significantly, while the far-left vote share shows a smaller and less precisely estimated increase. This is consistent with prior literature (Funke, Schularick, and Trebesch, 2016) documenting that far-right parties disproportionately benefit from recessions. A small decline in voter turnout is also documented, which may amplify measured increases in extremist vote shares by reducing the denominator (valid votes).

What do the results imply for environmental concern and the political sustainability of climate policy?

Eurobarometer data show that the share of respondents ranking environmental issues among the two most important problems facing their country falls by approximately 2 percentage points following a carbon policy shock, a persistent decline lasting four years. The authors interpret this as a self-interest crowding-out effect: when carbon pricing imposes economic costs, concern for the environment is displaced by concern for living standards, consistent with Douenne and Fabre (2022). This creates a potential self-undermining dynamic: carbon pricing erodes the popular support needed to sustain and strengthen climate policy over time, particularly given that carbon-intensive regions — which suffer most economically — also see the largest decline in public support for environmental issues.

What are the scope conditions on the policy implications?

The findings pertain to ETS-style cap-and-trade pricing based on regulatory-driven supply restriction, not to national carbon taxes, which the paper shows have much smaller economic and political footprints. The sample covers 20 European countries with NUTS2 regional data over 2000–2019. The carbon policy shocks are derived from EU ETS regulatory events and are specific to that institutional context; generalization outside the EU ETS requires caution. Political effects operate primarily over a two-to-four-year horizon coinciding with electoral cycles. The paper’s redistribution prescription (progressive revenue recycling) presupposes a policy instrument capable of targeting household income; the EU ETS currently lacks such a mechanism, which is precisely the gap the authors flag as most urgent given the ETS expansion to buildings and transportation scheduled for 2027.

Key Concepts

Carbon policy shock: A series of exogenous regulatory surprises in EU ETS carbon allowance markets, constructed by Kanzig (2023) from high-frequency futures price movements around 126 regulatory events (2005–2019), instrumented in a monthly VAR, and normalized to produce a 1% on-impact increase in Euro Area HICP energy prices. Distinct from carbon price levels or oil shocks; isolates policy-driven changes in the supply of emission allowances, orthogonal to contemporaneous economic conditions.

ETS emission intensity: Verified ETS emissions from regulated industrial installations in a NUTS2 region, scaled by regional GDP. The primary measure of a region’s direct exposure to EU carbon pricing; regions with higher ETS intensity experience larger economic contractions and larger shifts toward extremist parties when carbon prices rise.

Share of free allowances: The ratio of freely allocated ETS emission permits to a region’s verified ETS emissions, used as a second regional exposure measure. A higher share implies a lower effective carbon price faced by firms; regions with fewer free allowances bear larger economic and political costs from carbon policy shocks. Free allowances were originally granted to protect energy- and trade-intensive sectors from rapid cost increases.

Extremist vote share: The combined vote share of far-left and far-right parties in a region-election observation, using party classifications from The PopuList expert-coding database. The primary political outcome variable in the paper; empirically driven mainly by the far-right component in response to carbon policy shocks.

Political fragmentation: Defined in the paper as one minus the Herfindahl-Hirschman Index computed over all parties’ vote shares in an election (1 − sum of squared vote shares). Captures the dispersion of votes across parties beyond the extremist vote share; used as a summary indicator of political polarization.

Economic voting channel: The mechanism by which voters respond to carbon-pricing-induced economic deterioration — falling GDP, employment, and household income — by shifting support away from mainstream parties toward extremist alternatives. Isolated empirically via the Holm-Paul-Tischbirek (2021) decomposition; accounts for approximately two thirds of the total extremist voting response over the four-year impulse response horizon.

Regional inequality (90–10 ratio): Within-country dispersion of regional real disposable household income (or employee compensation) measured as the difference between the 90th and 10th percentile NUTS2 regions. Carbon pricing raises this measure persistently, with widening concentrated at the lower tail (the median-to-10th-percentile gap), indicating that poorer regions bear disproportionate economic costs.