2. Literature

This paper mainly builds on two strands of the literature. First, it builds on the literature examining the effects of climate policies, such as carbon taxes and ETS, on carbon emissions. Second, it builds on the literature studying macroeconomic determinants of carbon emissions.

The effects of climate policies, such as carbon taxes and ETS, on carbon emissions have been studied in several papers. A comprehensive review of the literature on the effects of carbon pricing on carbon emissions was carried out by Green (Reference Green2021). She highlighted that surprisingly few papers have conducted an ex post empirical analysis of how carbon pricing has actually affected CO₂ emissionsFootnote ² and that the vast majority of these papers are focused on Europe. She noted that in most cases, studies have estimated emissions reductions in the sectors covered by the carbon pricing policy, although some extrapolate to broader jurisdictional effects (e.g. Murray and Maniloff, Reference Murray and Maniloff2015; Bayer and Aklin, Reference Bayer and Aklin2020; Rafaty et al, Reference Rafaty, Dolphin and Pretis2020).Footnote ³ She concluded that the majority of the studies suggest that the aggregate reductions from carbon pricing on emissions are limited, generally between 0% and 2% per year, with considerable variation across sectors. She also concluded that in general, carbon taxes perform better than ETS and that studies of the European Union’s ETS indicate limited average annual reductions in carbon emissions of 0% to 1.5%.

Our paper contributes to filling this gap in the literature by providing ex post empirical analysis of the effects of carbon pricing on carbon emissions at the aggregate national level, for a very broad sample of countries. There is also little empirical literature on the effects of broader climate policies on emissions, and we contribute to filling this gap by studying the effects of an index of broad climate policies on CO₂ emissions.

The recent systematic review and meta-analysis of ex post evaluations of the effects of carbon pricing by Döbbeling-Hildebrandt et al. (Reference Döbbeling-Hildebrandt, Miersch, Khanna, Bachelet, Bruns, Callaghan, Edenhofer, Flachsland, Forster, Kalkuhl, Koch, Lamb, Ohlendorf, Steckel and Minx2024) highlighted a critical evidence gap regarding the carbon price elasticity of emissions reductions (i.e. the effect of a marginal change in the carbon price on emissions). It noted that only nine primary studies estimated such a carbon price elasticity, four of which only studied the carbon tax in the Canadian province of British Columbia to estimate elasticities for the transport and buildings sectors there. They concluded that having only nine price elasticity studies provided them with too few effect sizes for meta-analysing these price elasticities separately. Instead, they only conducted a meta-analysis of the effect of introducing carbon pricing, i.e. of the treatment effect (Döbbeling-Hildebrandt et al., Reference Döbbeling-Hildebrandt, Miersch, Khanna, Bachelet, Bruns, Callaghan, Edenhofer, Flachsland, Forster, Kalkuhl, Koch, Lamb, Ohlendorf, Steckel and Minx2024). Our paper also contributes to filling this gap in the literature of estimating carbon price elasticities of carbon emissions reductions, by providing evidence on the effectiveness of carbon pricing relative to the level of the carbon price, rather than just on the effect of introducing a carbon pricing scheme.

There is limited empirical evidence that higher carbon prices reduce carbon emissions (Arlinghaus, Reference Arlinghaus2015; Martin et al., Reference Martin, Muûls and Wagner2016; OECD, 2021). Sen and Vollebergh (Reference Sen and Vollebergh2018) found that for OECD economies, an increase in a broad-based tax on energy consumption of €10/tCO₂ is expected to lead to a 7.3% reduction in carbon emissions from fossil fuel consumption in the long run. Metcalf and Stock (Reference Metcalf and Stock2023) found that Europe’s carbon taxes led to a cumulative reduction of around 4% to 6% for a $40/tCO₂ tax covering 30% of emissions. They argued that emissions reductions would likely be larger for a broad-based US carbon tax, since European carbon taxes do not include sectors with the lowest marginal costs of carbon pollution abatement in the tax base. Related recent papers are Best et al. (Reference Best, Burke and Jotzo2020), D’Arcangelo et al. (Reference D’Arcangelo, Pisu, Raj and van Dender2022) and Schroeder and Stracca (Reference Schroeder and Stracca2023). Based on sectoral analysis, Rafaty et al. (Reference Rafaty, Dolphin and Pretis2020) found an imprecisely estimated semi-elasticity of a 0.5% reduction in carbon emissions growth per average $10/tCO₂ carbon price.Footnote ⁴ An overview of the results from 24 ex ante models in the IPCC’s AR6 Scenario Database for the effects of carbon taxes on CO₂ emissions, in comparison with the results from ex post empirical models, was provided in Tol (Reference Tol2023).

The review of Döbbeling-Hildebrandt et al. (Reference Döbbeling-Hildebrandt, Miersch, Khanna, Bachelet, Bruns, Callaghan, Edenhofer, Flachsland, Forster, Kalkuhl, Koch, Lamb, Ohlendorf, Steckel and Minx2024) noted that the contributions of carbon pricing schemes to carbon emissions reductions remain a subject of heated debate in science and policy. It mentioned that there is a critical evidence gap with regard to dozens of unevaluated carbon pricing schemes. It concluded from its meta-analysis of ex post studies that introducing a carbon price led to substantial emission reductions for at least 17 of 21 carbon pricing schemes studied, with statistically significant emissions reductions ranging between −5% and −21% across the schemes (or −4% and −15% after correcting for publication bias).

The effects of macroeconomic variables on carbon emissions were studied in Raupach et al. (Reference Raupach, Marland, Ciais, Quere, Canadell, Klepper and Field2007), Sadorsky (Reference Sadorsky2014), Kasman and Duman (Reference Kasman and Duman2015), Menyah and Wolde-Rufael (Reference Menyah and Wolde-Rufael2010), Gonzalez-Sanchez and Martin-Ortega (Reference Gonzalez-Sanchez and Martin-Ortega2020), Feng et al. (Reference Feng, Davis, Sun and Hubacek2015), Peters et al. (Reference Peters, Marland, Quere, Boden, Canadell and Raupach2012), Doda (Reference Doda2014), Wang (Reference Wang2012) and International Energy Agency (2020), among others. Our approach for controlling for macroeconomic determinants of carbon emissions follows the study of Kohlscheen et al. (Reference Kohlscheen, Moessner and Takats2021). For a review of the literature studying the macroeconomic determinants of carbon emissions, readers are referred to that study.

Ellis et al. (Reference Ellis, Nachtigall and Venmans2019) reviewed ex post empirical assessments on the impact of carbon pricing on competitiveness in OECD and G20 countries in the electricity and industrial sectors. They concluded that there are no significant effects on competitiveness. In turn, Lilliestam et al. (Reference Lilliestam, Patt and Bersalli2021) reviewed the empirical evidence available in academic ex post analyses of the effectiveness of carbon pricing schemes in promoting technological change necessary for full decarbonisation. They considered the European Union, New Zealand and Scandinavia. They found that there is little empirical evidence in this direction so far, with most of the papers on the topic being theoretical. Recent papers studying the macroeconomic effects of carbon taxes include Metcalf and Stock (Reference Metcalf and Stock2023, Reference Metcalf and Stock2020).

3. Data

In order to assess the empirical drivers of CO₂ emissions, we collected data from several sources. Data on CO₂ emissions per capita, measured in metric tons per capita, were from the World Development Indicators (WDI) of the World Bank. They measure CO₂ emissions stemming from the burning of fossil fuels and the manufacture of cement and include CO₂ produced during the consumption of solid, liquid and gas fuels and gas flaring. Demographic, economic and energy use data were from the World Bank (GDP per capita, GDP growth, the urbanisation rate, the share of manufacturing in GDP and the share of coal, oil and renewables in electricity generation).Footnote ⁵ Our database spanned the 1971–2016 period and 121 countries. The country selection was based solely on data availability, and the list of countries in the sample is presented in the Appendix.

Our main variables of interest are CO₂ emissions and climate policies.

Carbon emissions emanate mostly from the burning of fossil fuels (to heat, transport goods and people or generate electricity) and the manufacture of steel and cement. The level of carbon emissions, following that of economic development, is highly uneven across countries. Highly developed advanced economies tend to emit large quantities of carbon per capita, while less developed economies, particularly in Africa, tend to emit less (Figure 1). However, there is no definite advanced–emerging economy divide. Due to fast economic development, many emerging market economies (including energy producers and fast-developing East Asian economies) already have high carbon emission levels per capita. Besides, given their larger populations, their contribution to global CO₂ emissions has grown very rapidly.

Figure 1. Carbon dioxide emissions per capita.

Notes: CO₂ emissions per capita in metric tons per capita. CO₂ emissions measure carbon dioxide emissions stemming from the burning of fossil fuels and the manufacture of cement and include carbon dioxide produced during the consumption of solid, liquid and gas fuels and gas flaring. Source: World Development Indicators (WDI) of World Bank, code EN.ATM.CO2E.PC.

Primary climate policies are carbon taxes and ETS. The first carbon tax was introduced in Finland in 1990. In turn, emissions trading has been considered a possible tool for mitigating greenhouse gas emissions since the early 1990s and formed a key part of the Kyoto Protocol agreement (Philibert and Reinaud, Reference Philibert and Reinaud2004). Our database contains information on carbon taxes and ETS implemented at the national and supranational levels since 1990 from the World Bank (World Bank, 2021). Country coverage and prices of these policies are shown in Figures 2 and 3 separately for carbon taxes and ETS, for the start and the end of our sample period in 2016. The share of global greenhouse gas emissions covered by the carbon taxes and ETS at the national and supranational levels is shown in Figure 4. As the figure shows, there has been a huge increase in this proportion.

Figure 2. Prices of carbon taxes as of 2016 and 1971.

Note: Carbon taxes implemented at the national and supranational levels. Nominal prices as of 1 April each year in $/tCO₂ (US dollars per metric ton carbon dioxide emissions) equivalents. In the case of the UK, the number refers to the carbon price floor. Source: Carbon Pricing Dashboard, World Bank (2021).

Figure 3. Prices of carbon ETS as of 2016 and 1971.

Note: Prices of the carbon emission trading system (ETS) implemented at the national and supranational levels. Nominal prices as of 1 April each year in $/tCO₂ (US dollars per metric ton carbon dioxide emissions) equivalents. EU: European Union. Source: Carbon Pricing Dashboard, World Bank (2021).

Figure 4. Share of global greenhouse gas emissions covered by carbon taxes and ETS at the national and supranational levels.

Notes: Share of global greenhouse gas emissions covered by carbon taxes and emissions trading systems (ETS) at the national and supranational levels, in percentage. The coverage of each carbon pricing initiative is presented as a share of annual global GHG emissions for 1990–2015 based on data from the Emission Database for Global Atmospheric Research (EDGAR) version 5.0, including biofuels emissions. From 2015 onwards, the share of global GHG emissions is based on 2015 emissions from EDGAR. The greenhouse gas emissions coverage for each jurisdiction is based on official government sources and/or estimates.

Source: Carbon Pricing Dashboard, World Bank (2021).

In the case of carbon taxes, governments set the price of carbon emissions and let private agents determine emissions reductions. ETS have two main forms, cap-and-trade and baseline-and-credit ETS. For cap-and-trade ETS, governments set a limit on emissions, and allowances up to this limit are auctioned or allocated according to certain criteria. These permits are then traded, and carbon prices are determined by supply and demand in the market. For baseline-and-credit ETS, baselines for emissions are set for regulated emitters. Emitters with emissions above their designated baseline need to give up credits to make up for these emissions, while those with emissions below their baseline receive credits for these reductions, which they can sell to other emitters (World Bank, 2021).

Besides carbon taxes and ETS, we also evaluated the effectiveness of climate policies using a broader index for the overall stringency of climate policies for OECD member countries and major emerging economies. This was based on the Environmental Policy Stringency Index (EPS), which is compiled by the OECD and available from 1990 to 2015. The index covers a much wider range of climate policies, also considering other regulatory policies (Botta and Kozluk, Reference Botta and Kozluk2014). Generally, stringency is defined as the degree to which environmental policies put an explicit or implicit price on polluting or, more generally, on environmentally harmful behaviour.Footnote ⁶

We conducted several checks of the robustness of our baseline results. In particular, to control for the quality of governance in a country, we used a measure of the control of corruption from the World Bank. It reflects the perceptions of the extent to which public power is exercised for private gain, including both petty and grand forms of corruption, as well as “capture” of the state by elites and private interests.Footnote ⁷

4. Methodology

We quantified the effects of climate policies on carbon emissions through a dynamic panel model. The dynamic specification accounts for the high degree of persistence in CO₂ emissions. Throughout the analysis, we controlled for macroeconomic factors, such as economic development (GDP per capita), GDP growth, urbanisation, the share of manufacturing in total output and the energy mix used in electricity production. Note that we included only the respective shares of the energy mix, not the intensity of use. Formally, we explored the effects of climate policies (denoted by CP) on the logarithm of CO₂ emissions per capita (denoted by lnCO₂) according to the following equation:

(1) $$ {\displaystyle \begin{array}{c}\mathrm{lnCO}{2}_{i,t}={\alpha}_i+{\beta}_t+\rho \hskip0.32em \mathrm{lnCO}{2}_{i,t-1}+\lambda \hskip0.32em {\mathrm{CP}}_{i,t-1}+\gamma \hskip0.32em {\mathrm{GDPpc}}_{i,t}+\theta \hskip0.32em {\mathrm{growth}}_{i,t}\\ {}\hskip2.6em +\hskip0.32em \unicode{x03C9} \hskip0.32em {\mathrm{urbanisation}}_{i,t}+\vartheta \hskip0.32em {\mathrm{manufacturing}}_{i,t}+{\mu}_1{\mathrm{share}}_{\mathrm{oil},i,t}\\ {}\hskip-5.2em +\hskip0.32em {\mu}_2\hskip0.32em {\mathrm{share}}_{\mathrm{coal},i,t}+{\mu}_3\hskip0.32em {\mathrm{share}}_{\mathrm{renewables},i,t}+{\varepsilon}_{i,t}\end{array}} $$

Besides the lagged dependent variable, we included climate policies as key explanatory variables of interest. To address potential endogeneity concerns, we used lagged climate policy variables, which minimised the risk of reverse causality. As climate policies, we considered carbon taxes and prices of permits in ETS implemented at the national and supranational levels, as well as the broader index for the stringency of climate policies described in Section 2. We also included a number of macroeconomic variables as controls.

Our models included country fixed effects to capture unobserved heterogeneities across countries that might affect the rate of CO₂ emissions. These included fixed institutional factors such as enforcement of environmental laws. They also included natural factors such as average median temperatures, which tend to correlate with heating or cooling needs. We also included the full set of yearly time dummies to control for the effects of global factors. These subsumed, for instance, technological advances that may reduce environmental effects, as well as other global trends or global shocks.

In this analysis, we used fixed effect panel estimations and based inference on cluster robust standard errors. As a robustness check, we controlled additionally for the quality of governance in a country, using the measure of the control of corruption as described in Section 3.

5. Estimation results

5.1. Effects of climate policies

We first present baseline estimates of our dynamic panel equation (1). The estimations for the effects of carbon taxes, prices of permits in the ETS and the EPS policy index separately are shown in columns I, II and IV of Table 1, respectively. The results when including carbon taxes and prices of permits of ETS together are shown in column III, and those when including all three climate policies together in column V.

Table 1. Effects of climate policies on CO₂ emissions

Note: Sample period: From 1971 to 2016, annual data. Cluster-robust standard errors reported in the second line are clustered at the country level. ***/**/* denote statistical significance at the 1%/5%/10% level. (I): nominal price of first carbon tax in USD/tCO₂ equivalents; (II): nominal price of ETS in USD/tCO₂ equivalents; (IV): EPS index.

Overall, the specifications describe the country-specific evolution of carbon emissions quite well. Our models with climate policies are able to explain 97%–98% of the variation in per capita CO₂ emissions across countries and across time (Table 1). Importantly, this is driven by variation both within and between countries, as R² within ranges from 0.82 to 0.88 and R² between ranges from 0.97 to 0.99. The lagged dependent variable has a coefficient of just above 0.7 and is clearly statistically significant, which confirms that a dynamic panel specification is indeed appropriate. Put differently, more than 70% of CO₂ emissions can be explained by previous year emissions alone.Footnote ⁸ Most of the macroeconomic control variables are statistically significant with the expected signs.

Consistently, we found that higher carbon taxes significantly reduce carbon emissions (p-values always below 0.01). On average, an increase in carbon taxes by $10 per ton of CO₂ (tCO₂) reduces CO₂ emissions per capita by 1.3% in the short run and by 4.6% in the long run (baseline model in column III of Table 1).Footnote ⁹ Importantly, this result is robust to controlling for ETS prices (see column III).

We also found that higher prices of ETS permits reduce carbon emissions (p-values below 0.01 in columns II and III of Table 1). An increase in prices of ETS permits by $10/tCO₂ reduces CO₂ emissions per capita by 1.4% in the short run and by 5.0% in the long run, when carbon taxes are controlled for (column III).

Moreover, we found that the broad EPS index of climate policy stringency has a negative effect on carbon emissions at the 10% significance level (column IV of Table 1). A one standard deviation increase in the EPS index reduces CO₂ emissions per capita by 1.6% in the short run and 6.2% in the long run. It is striking that these results are broadly similar to the results we obtained for carbon taxes, for which a standard deviation increase leads to a reduction in carbon emissions of 1.1% in the short run and 3.9% in the long run.

When we included all three measures together, the coefficients on carbon taxes and the EPS index remained significantly negative, with a somewhat smaller magnitude (column V). While the coefficient on the prices of ETS permits remained negative, it lost significance. This is most likely due to the much smaller sample size, as we lost more than 80% of the sample when including the EPS index. Furthermore, the EPS index also reflects carbon prices, which introduces the problem of multicollinearity of the variables—which is particularly acute for small sample sizes.

5.2. Robustness

We performed several robustness checks. First, we controlled for the quality of governance in a country, proxied by a measure of control of corruption described in Section 3 (Table 2).Footnote ¹⁰ We found that higher carbon taxes and more stringent climate policy based on the broader EPS policy index significantly reduce carbon emissions when controlling for governance in all specifications (Table 2). By contrast, the coefficient on the prices of ETS permits remained significant only in one specification (column II). These results imply that the effects of carbon taxes and the broad climate policy index on carbon emissions are generally more robust than those for the effects of ETS permit prices.

Table 2. Effects of climate policies on CO₂ emissions: with control of corruption

Note: Sample period: From 1971 to 2016, annual data. Cluster-robust standard errors reported in the second line are clustered at the country level. ***/**/* denote statistical significance at the 1%/5%/10% level. (I): nominal price of first carbon tax in USD/tCO₂ equivalents; (II): nominal price of ETS in USD/tCO₂ equivalents; (IV): EPS index.

Furthermore, we also estimated the effects of carbon policies when using real carbon prices instead of nominal ones; that is, we considered prices obtained by deflating the nominal USD carbon price by the US consumer price index (Table 3). The results for the significantly negative effects of higher carbon taxes, of higher prices of ETS permits and of more stringent climate policy based on the EPS index on CO₂ emissions are robust to using these alternative measures.

Table 3. Effects of climate policies on CO₂ emissions: using real carbon prices

Note: Sample period: From 1971 to 2016, annual data. Cluster-robust standard errors reported in the second line are clustered at the country level. ***/**/* denote statistical significance at the 1%/5%/10% level. (I): real price of first carbon tax in USD/tCO₂ equivalents (deflated by US CPI); (II): real price of ETS in USD/tCO₂ equivalents (deflated by US CPI); (IV): EPS index.

Finally, we excluded countries whose economic development, as measured by GDP per capita, is below that of the country with the lowest GDP per capita among those that have implemented carbon taxes or ETS. We did so by restricting GDP per capita to above $1500 in constant 2010 dollars in columns I to III of Appendix Table A1. Very similar results were obtained when this restriction was imposed.

Our main takeaway from the robustness analysis is that the results for the carbon price elasticities of emission reductions are generally robust to the different empirical models, which enhances confidence in these results. Moreover, the results of this paper for the ex post effects of carbon pricing on carbon emissions are quantitatively similar to those obtained by D’Arcangelo et al. (Reference D’Arcangelo, Pisu, Raj and van Dender2022), Metcalf and Stock (Reference Metcalf and Stock2023) and Rafaty et al. (Reference Rafaty, Dolphin and Pretis2020) (see Table 1 of Tol, Reference Tol2023), even though these papers have used different methods. The results of Rafaty et al. (Reference Rafaty, Dolphin and Pretis2020) are, for example, based on sectoral analysis. This consistency of our results with those of other papers with different methodologies also enhances confidence in our results.