Most industrialised countries have peaked carbon dioxide emissions during economic crises through strengthened structural change

The paper examines how and under which conditions national CO2 emissions peak and start to decline, particularly in the context of major economic crises such as the oil shocks of the 1970s, the collapse of the Soviet Union, the Global Financial Crisis (GFC), and the COVID-19 pandemic. While global emissions typically exhibit only short-lived dips during crises, national-level outcomes vary substantially. Theoretically, crises may support decarbonisation via structural change—through Schumpeterian creative destruction, policy windows for transformative energy and climate policies, and destabilisation of socio-technical regimes—yet they may also impede progress by elevating uncertainty, discouraging investment, and shifting priorities away from climate action. Empirically, prior studies provide mixed evidence on whether crises lead to lasting emission reductions and have typically not addressed structural change explicitly. This study links emissions drivers analysis with crisis impacts, investigating whether and how crises affect the GDP effect and structural change (energy and carbon intensity) contributing to emissions peaks across 45 OECD and G20 countries from 1965–2019. The central research questions are whether national CO2 emission peaks are associated with economic crises and, if so, which economic and structural mechanisms explain such crisis-related peaks.

Prior research on emissions drivers shows that rising activity levels (GDP, population) generally increase emissions, while improvements in energy intensity are key reducing drivers. Reviews indicate moderate decarbonisation in Europe and North America since 1990, driven by fuel switching and renewable deployment, with fossil energy use expanding in rapidly industrialising regions. Studies on decoupling have identified many countries—especially in Europe—with peaked emissions and absolute decoupling (GDP growth with declining emissions) for both production- and consumption-based metrics. Early assessments of crisis impacts (e.g., the GFC, COVID-19) suggested short-lived reductions followed by rebounds, while later work and some broader cross-country analyses point to substantial crisis-related impacts on emissions and decoupling. However, most prior studies did not systematically investigate the effects of crises on structural change, leaving a gap this paper addresses by linking crises to changes in energy and carbon intensity that underpin durable decarbonisation.

Scope: The analysis covers 45 countries: all 37 OECD members plus 8 G20 members not in the OECD (all major emitters except Iran), representing 77% of global CO2 emissions in 2019. Period: 1965–2019. Data: CO2 emissions from fossil fuel combustion and primary energy consumption are drawn from BP Statistical Review. Emissions reflect territorial, combustion-related sources using IPCC default factors. Economic and population data are from the World Bank; GDP per capita for former Soviet countries and Poland is from the Maddison Project. The focus is on production-based emissions to assess national economic and energy system responses to crises. Identification of peaks: For each country, the year of maximum CO2 emissions is identified using a 5-year moving mean to smooth short-term anomalies. Sustained peaks are defined only if they occur at least 10 years before 2019 (i.e., peaks after 2009 are not considered sustained). Potential peaks post-GFC for Australia, New Zealand, Israel, and South Africa are noted but excluded from the peak group due to insufficient post-peak duration. A peak is associated with an economic crisis if it occurs within ±2 years around the crisis onset. Decomposition approach: The Kaya identity is used with Log-Mean Divisia Index (LMDI) multiplicative decomposition to attribute changes in CO2 to four factors: population (P), GDP per capita (G), energy intensity of GDP (EI), and carbon intensity of energy (CI). For reporting and statistical testing, factors are grouped into two effects: GDP effect (population + GDP per capita) and structural change effect (energy intensity + carbon intensity). For each crisis, growth rates of contribution factors are computed for pre-crisis and post-crisis periods, typically 10 years before the first crisis year with negative GDP growth and 10 years after the last crisis year, adjusted when overlapping crises shorten available periods. Growth rates r are calculated as r = [(K_t / K_1)^(1/n) − 1] × 100%. Crises examined include the first oil crisis (1973–75), second oil crisis (1979–81), dissolution of the Soviet Union (1989–1991), and GFC (2007–09). Statistical analysis: Because data are non-normal, Wilcoxon signed-rank tests assess whether post-crisis contribution factors (GDP and structural change effects) are significantly lower than pre-crisis values for peak, non-peak, and all countries. Hypotheses test for reductions in GDP and structural change effects post-crisis for both groups and the full sample.

- Global impact: Global CO2 emissions display only temporary dips during major crises (e.g., −2.1% in 2009; −5.9% in 2020 with a 5.6% rebound in 2021), indicating limited global persistence of crisis impacts. - National peaks: Of 45 OECD/G20 countries, 28 have passed a sustained national peak in energy-related CO2 emissions since 1965. Of these, 26 peaked just before or during a crisis (0–2 years), notably around the oil crises, the Soviet Union collapse, and the GFC. Denmark (1998) and Switzerland (2001) are exceptions without proximate deep crises. - Peak-and-decline countries (26 crisis-related peaks): - GDP effect: Median decrease of 1.5 percentage points per year (interquartile range −3.3 to −1.1) in post-crisis GDP-related contribution (p < 0.01). In 23/26 cases, GDP’s contribution to emissions growth was lower post-crisis; GDP continued to grow in most cases but added less to emissions. In a few deep recessions (e.g., Lithuania, Latvia, Greece, Italy), the GDP effect slightly reduced emissions post-crisis. - Structural change effect: Intensified by −0.7 percentage points per year (−1.8 to −0.1), statistically significant (p < 0.01). In 17/26 cases, pre-crisis improving trends in energy and/or carbon intensity accelerated post-crisis; in 3 cases trends bent from increasing to decreasing emissions; in 6 cases improvements slowed but still reduced emissions. This intensified structural change combined with lower GDP growth explains sustained absolute decoupling. - Mechanisms of structural change: 1) Energy efficiency responses to price shocks and tight conditions (notably during the oil crises and GFC), supported by policies and firm-level shifts (e.g., smaller, more efficient cars; building and industrial efficiency). 2) Economic restructuring: decline of energy/carbon-intensive sectors and rise of less intensive sectors (e.g., post-Soviet transitions; Spain’s post-GFC shift away from construction; Poland’s 1990s reforms). 3) Energy mix changes reducing carbon intensity: nuclear build-outs (e.g., France, Sweden) after oil shocks; post-GFC coal-to-gas and renewables deployment driven by policy and market conditions. - Non-peak countries: GDP growth was less impacted by crises and structural change effects were weaker or even increased emissions. Post-crisis GDP effect decreased by −1.0 percentage point (−2.4 to 0.1), significant (p < 0.01), but remained positive; structural change effect improved by −0.2 percentage points (−2.2 to 1.3), not significant (p = 0.102). Emissions typically rebounded and continued rising due to strong GDP-driven increases outweighing weak structural change. - Statistical tests (Wilcoxon signed-rank, change pre- vs post-crisis contributions): - GDP effect: significant decreases for peak (median −0.015; p = 0.000), non-peak (median −0.010; p = 0.000), and all countries (median −0.012; p = 0.000). - Structural change effect: significant decrease only for peak countries (median −0.007; p = 0.003); not significant for non-peak (median −0.002; p = 0.102) or all countries (median −0.002; p = 0.126).

The findings show a robust temporal association between major economic crises and national CO2 emissions peaks in industrialised economies. Peaks are explained by a combination of reduced GDP-driven emissions growth during and after crises and, crucially, by accelerated structural change manifested as faster improvements in energy and carbon intensities. Structural change often magnifies pre-existing decarbonisation trends; countries already improving efficiency and decarbonising energy supply were best positioned to convert crises into durable emissions reductions. In contrast, in non-peaking countries—often emerging economies—crises had limited effects on GDP growth and structural change, leading to quick rebounds and continued emissions growth. The results support theories of creative destruction, policy windows, and regime destabilisation enabling transitions, but also acknowledge contexts where crises impede decarbonisation. Absolute decoupling is possible and observed over long periods, though sustained reductions require sufficient rates of intensity improvements relative to GDP growth. Policy implications include the importance of pre-crisis policy frameworks for energy efficiency and clean energy so that recovery phases can accelerate structural change rather than restore high-carbon trajectories.

The study contributes by (1) extending analyses of emissions drivers to explain how major economic crises relate to emissions peaks in OECD/G20 economies, and (2) providing rare ex post evidence that crises can accelerate structural change relevant for long-term decarbonisation. Most industrialised countries that peaked did so around crises, with structural change strengthened during and after the crisis, while GDP growth resumed at lower emissions intensity. However, crises do not automatically trigger structural change; preparatory policies and ongoing transitions are key. Future research should investigate consumption-based emissions responses to crises, disentangle the relative roles of market forces versus policy in driving structural change, and conduct country-specific case studies to trace mechanisms. The long-term impacts of COVID-19 and the energy crisis linked to the war in Ukraine remain to be seen and warrant continued monitoring.

- Emissions boundary: Focus on territorial, production-based CO2 emissions; potential effects of outsourcing/embodied emissions are not captured. Prior work suggests offshoring effects have slowed in the past 15 years, but consumption-based analyses could yield additional insights. - Causality and mechanisms: The study identifies temporal associations and contribution factor changes but does not fully disentangle specific policy versus market drivers in each country; detailed case studies are needed for causal tracing. - Data and scope: GDP per capita data sources differ for post-Soviet countries; analysis relies on relative changes. Peaks after 2009 are excluded as sustained status is uncertain; this may omit recent peakers. - Crisis heterogeneity: Crises vary in depth, duration, and policy responses; generalisations may not apply uniformly. The short, sharp GDP drop during COVID-19 may limit creative-destruction effects compared to earlier crises. - Path to zero: Observing a peak and decline does not imply rapid progress to net-zero; intensity improvements rarely exceed −4%/year, and continued GDP growth can constrain the speed of decarbonisation without strong structural shifts.