## 2.1    Empirical framework

## Source details

**Canonical URL:** [2.1    Empirical framework](https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022066-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2022/english/wpiea2022066-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2022/english/wpiea2022066-print-pdf.pdf.json)

---

### Empirical expectations and causal framework
- Expectation: positive statistical association between the number of climate policies other than carbon pricing (referred to as the size of the climate policy portfolio) and both the adoption and the intensity of a pricing policy.
- Two causal channels:
  - Ex-post channel: earlier climate policies increase the probability of subsequent adoption of a pricing policy.
  - Ex-ante channel: an anticipated later adoption of a pricing policy motivates prior adoption of other policies.
- Mechanisms and barriers:
  - Concerns about high mitigation costs for firms in energy-intensive industries competing internationally.
  - High costs for consumers and possibly a regressive distribution of these costs.
  - Political opposition to climate change policies and to pricing policies.
  - Concerns about effects on employment in fossil-fuel–reliant industries.
- Policy implication: other climate policies (e.g., technology subsidies) can lower mitigation costs through technological innovation or gradually transform the economy away from carbon-intensive activities, potentially addressing barriers and increasing the likelihood of subsequent carbon pricing.

### Data
- Primary dataset: climatepolicydatabase.org; dataset composition and validation noted (Nascimento et al., 2021; validation with national stakeholders and experts).
- Coverage:
  - Comprehensive for G20 economies (including EU member countries that are individual members of the G20, but not other EU members) and 18 additional countries (mostly advanced and emerging economies in Europe, Asia and Latin-America, some less developed countries, and two countries in Africa).
  - Countries with non-comprehensive data are dropped from the sample; this includes Norway, Sweden, Finland, Poland, Denmark.
- Policy filtering and treatment of EU policies:
  - Include only policies that have climate change mitigation as one of their objectives.
  - Neglect policies at the subnational level.
  - Apply all EU policies to member countries’ portfolios; if a country joined the EU after the policy decision, use the year of joining as the date of policy adoption.
- Instrument and sector taxonomy:
  - Original dataset distinguishes 72 instrument categories, aggregated to seven instrument types based on IEA typology; combined with Carbon Pricing Dashboard to create an eighth instrument type carbon pricing.
  - Final eight instrument types: Regulatory instruments; Grants, subsidies, and other financial incentives; Information and education; Policy support; Research, development, and deployment; Voluntary agreements; Procurement and investment; Carbon pricing.
  - Sectors: Electricity and heat production; Transport; Buildings; Industry; Agriculture, forestry, and other land use; General.
- Portfolio scoring:
  - Count how many of the eight instrument types have been implemented in each of the six sectors, yielding a score between 0 and 48 for every country and every year.
- Complementary country characteristics:
  - GDP per capita (PPP) from the World Bank.
  - Education index from UNDP Human Development Indicators.
  - Control of corruption from World Governance Indicators (World Bank).
  - Fossil fuel reserves from US Energy Information Administration.
- Data scale and frequencies:
  - Every instrument type has been used in every sector in at least one country.
  - Number of observations for specific instrument-sector combinations ranges from 18 to 1902.
  - Total observed instrument-sector combinations: 14,540.

### Statistical methods
- Policy sequencing:
  - Eight instrument types imply 40,320 possible sequences.
  - Approach: consider all possible pairs (28 pairs); for each pair, examine which instrument tends to be adopted first across sectors and countries using conditional frequencies f(Y_{t−1} | X_{t}) = n(Y_{t−1} ∧ X_{t}) / n(X_{t}), where X_t and Y_{t−1} are binary indicators of policy adoption up to year t and t−1 respectively.
  - Exclude observations after the first time a specific instrument is adopted in a specific sector in a specific country.
- Matching and inference for adoption of carbon pricing:
  - Treated group: countries that adopted a national carbon price in a given year.
  - Control group: countries that did not adopt a carbon price neither earlier nor in the same year.
  - Matching: for each treated country assign one randomly chosen control country; iterate random assignment 1000 times and compare average portfolio sizes of treated and control countries.
  - Inference: calculate bootstrapped confidence intervals.
  - Additional regression: logit model with adoption of carbon pricing as binary dependent variable to include country characteristics as controls.
- Analysis among adopters:
  - Linear regression with portfolio size at adoption as dependent variable and control variables including year of adoption.
  - Parsimonious model with only GDP per capita and reserves of fossil fuels (coal and oil included; oil and gas highly correlated so gas excluded in some specifications).
  - Lasso model selection to identify most important explanatory variables.
  - Linear regression with economy-wide average carbon price as dependent variable to examine association with portfolio size.

### Results — temporal sequence of climate policies (summary)
- Overall sequencing:
  - Pooling across countries and sectors: carbon pricing tends to be the last instrument type.
  - Pooling across countries but keeping sectors separate: carbon pricing is the last instrument type in every sector.
  - In 12 out of the 15 countries examined, carbon pricing tended to be used for the first time in a specific sector after the use of any of the other instrument types.
  - Conclusion: carbon pricing tends to be implemented last, after adoption of all or almost all other seven instrument types.
- Two groups of instrument types by timing:
  - Early instrument types: Regulatory instruments; Grants, subsidies, and other financial incentives; Information and education; Policy support.
  - Late instrument types: Research, development, and deployment; Voluntary agreements; Procurement and investment; Carbon pricing.
  - Exception: voluntary agreements and financial incentives in Agriculture.
- Cross-country variation:
  - Most frequent national pattern: early adoption of Regulatory instruments and Policy support; late adoption of Procurement and investment and Carbon pricing.
  - Greater variation across countries in relative positions of Grants, subsidies, Research, development and deployment, Voluntary agreements, Information and education.

### Results — build-up of climate policy portfolios over time
- Three portfolio trajectory types among countries that eventually adopted carbon pricing:
  - Group 1 (e.g., Canada, Japan, South Africa): relatively rapid expansion of portfolio followed by slow further expansion over several years, eventually including carbon pricing.
  - Group 2 (e.g., Argentina, Switzerland): steady gradual expansion up until introduction of carbon pricing.
  - Group 3 (e.g., current EU members in the sample): rapid expansion of policies almost immediately followed by introduction of carbon pricing.
- Average time between adoption of new instrument type–sector combinations and adoption of carbon pricing (examples):
  - Canada: about 18 years.
  - Japan: about 13 years.
  - South Africa: about 14 years.
  - Argentina: about 11 years.
  - Switzerland: about 10 years.
  - Current EU countries in sample: about 5 years.

---

### 3.3    Policy portfolios of adopters versus non-adopters of carbon pricing

### Comparison of policy portfolios: adopters vs non-adopters
- At the time adopters adopted carbon pricing, average portfolio size: 29.5 instrument type - sector combinations.
- By the end of 2020, countries without a national carbon price had used on average about 23.9 instrument type - sector combinations.
- In 2015 (the average year of adoption of carbon pricing) adopters had used on average about 19.7 instrument type - sector combinations.
- Matching procedure result: countries that adopted a carbon price in a given year had policy portfolios on average about 11.12 instrument type-sector combinations larger than portfolios of those that did not; difference significant at α= 0.05.
- Bootstrapped confidence intervals used for inference.

### Statistical inference and robustness
- Logit model with adoption as binary dependent variable: statistically significant positive association between portfolio size and adoption (Table A2).
- Lasso estimation: size of portfolio and prevalence of gas reserves are the strongest determinants of adoption.
- Including additional control variables reduces the coefficient and significance of portfolio size, indicating potential confounding by country characteristics.

### Key regression/logit coefficients (Table A2 excerpts)
- Policy portfolio coefficient, Column 1: 0.1072 (∗∗∗)
- Policy portfolio coefficient, Column 2: 0.0485
- Policy portfolio coefficient, Column 3: 0.0808 (∗∗∗)
- Policy portfolio coefficient, Column 4 (Lasso-selected): 0.0994 (∗∗∗)
- Gas reserves coefficients: -0.0085 (∗), -0.0073, -0.0108
- Education coefficient example: 19.0651 (∗∗∗)
- Intercept examples: -4.4354 (∗∗∗), -20.3759 (∗∗), -8.8369 (∗∗∗), -3.9841 (∗∗∗)
- Ps. R2 examples: 0.33, 0.51, 0.39, 0.35
- AIC examples: 641.66, 479.73, 595.02, 625.47
- Sample: No. countries 36; N 1188
- Figure statistics:
  - Figure 6 (size of portfolio vs year of adoption): R2 = 0.79
  - Figure 6 (price intensity vs size of portfolio): R2 = 0.44

### Interpretation and heterogeneity
- Countries adopting carbon pricing in a given year tended to have significantly larger climate policy portfolios than non-adopters.
- Positive association robust across matching, logit, and Lasso, but sensitive to additional controls.
- Lasso highlights portfolio size and gas reserves prevalence as leading predictors.
- Size of portfolios at adoption has decreased over the last 20 years after controlling for country characteristics (Figure 6 left).
  - Example: Canada’s 2019 portfolio was substantially smaller than France’s 2003 portfolio after controls.
  - Decreasing pattern robust to different model specifications (Table A3) and to dropping outliers including scenarios treating the group of EU ETS countries as one observation or dropped.
- Regression of average carbon price at first year on portfolio size (controlling for same characteristics) shows a positive association:
  - Larger portfolios at adoption tended to implement higher price levels (Figure 6 right).
  - Japan is an outlier: relatively low price given its relatively large portfolio.
  - Association is less significant and less robust to dropping individual countries than the portfolio–adoption association.

### Mechanisms, trajectories, and policy implications
- Evidence consistent with prior climate policies removing barriers to subsequent carbon pricing and enabling higher initial price stringency.
- Carbon pricing generally adopted relatively late in countries’ policy sequences.
- Three observed trajectories prior to carbon pricing:
  - Gradual expansion of policy portfolios over time.
  - Rapid expansion followed immediately by adoption of national carbon pricing.
  - Rapid expansion followed by several years’ delay before adopting national carbon pricing.
- Possible drivers of trajectories: exposure to domestic and international events, domestic barriers to carbon pricing, whether carbon pricing was part of a long-term climate strategy.
- Over last 20 years, declining size of portfolios at adoption consistent with declining abatement costs and international influences such as technological diffusion and growing economic opportunities for green technologies.
- For countries with exhausted extensive margins of instrument types and sectors, future climate policy avenues:
  - Higher stringency of existing policies.
  - Additional policies of existing sector and instrument type combinations.
  - Carbon pricing as a final step in the sequence.

### Implications for future research
- Observed associations are not necessarily causal; causal interpretation should be cautious.
- Future empirical work should:
  - Expand data coverage to include earliest adopters of carbon pricing as data become available.
  - Investigate how early climate policies influence the effectiveness of subsequent carbon pricing (beyond association with initial price levels), including impacts on GHG reductions and emission leakage.
  - Explore how the size of climate policy portfolios influences GHG reductions obtained with a given carbon price.

---

### 1.  Size of portfolio is measured as described

### Regression analysis: determinants of portfolio size at adoption and carbon price at implementation
- Dependent variables:
  - Size of portfolio at time of adoption (Columns 1–4).
  - Carbon price at implementation (Columns 5–8).
- Significance notation: * p < 0.1, ** p < 0.05, *** p < 0.01.
- Coefficients and standard errors (in parentheses):

- log GDP per capita PPP
  - Column 1: -0.3065 (0.6389)
  - Column 2: -0.3598 (0.3817)
  - Column 3: -0.0551 (0.4820)
  - Column 4: 0.3032 (0.2632)

- Coal reserves
  - Column 1: -0.2631 (0.3648)
  - Column 2: -0.4717 *** (0.1061)
  - Column 3: -0.3787 *** (0.0904)
  - Column 4: 0.1396 (0.2898)
  - Column 5: 0.0563 (0.1056)

- Oil reserves
  - Column 1: 0.6041 (0.2159)
  - Column 2: 0.4866 ** (0.1808)
  - Column 3: 0.3336 *** (0.0669)
  - Column 4: -0.1718 (0.1832)
  - Column 5: -0.1624 ** (0.0571)
  - Column 6: -0.1193 (0.0722)

- Gas reserves
  - Column 1: -0.9966 (1.0017)
  - Column 2: 0.3461 (0.9932)

- Control of corruption
  - Column 1: -0.1574 (0.3191)
  - Column 2: 0.5966 (0.3191)
  - Column 3: 0.3323 ** (0.1357)

- Education
  - Column 1: 0.1547 (0.1926)
  - Column 2: -0.2415 (0.1894)

- Policy portfolio
  - Column 1: 0.5960 *** (0.1507)
  - Column 2: 0.4805 * (0.2066)
  - Column 3: 0.4957 ** (0.1577)
  - Column 4: 0.3960 ** (0.1320)

- Year of adoption
  - Column 1: -0.5067 ** (0.2043)
  - Column 2: -0.9301 * (0.4650)
  - Column 3: -0.9112 ** (0.3138)
  - Column 4: -0.6283 *** (0.1679)

- Intercepts (selected)
  - Column 1: -0.0336 (0.1883)
  - Column 2: 0.0176 (0.1798)
  - Column 3: 0.1769 (0.1281)
  - Column 4: 0.0722 (0.1873)
  - Column 5: -0.1687 (0.1959)
  - Column 6: -0.0713 (0.3606)
  - Column 7: -0.1265 (0.2123)
  - Column 8: -0.1332 (0.2012)

- Goodness of fit and sample size:
  - R2 by column: 0.27, 0.76, 0.71, 0.66, 0.42, 0.66, 0.53, 0.58
  - N reported: 15 for columns 1–8; separate line shows 24 (context unclear in excerpt).
- Additional note: Carbon price at implementation is calculated as an economy-wide average price.
- Additional note: The positive coefficient of oil reserves in Columns 2–4 becomes insignificant if Canada is dropped from the sample.

### Policy portfolios and carbon pricing status by end-2020 (summary)
- Policy portfolios by end-2020: categories include Carbon pricing and policy portfolios by end of 2020; instrument-type counts for countries without a national carbon price include examples "6/8 instrument types" and "7/8 instrument types".
- Carbon price (national) and Carbon price (subnational) distinguished in source figures; some cases marked "Not enough data" or "Fewer instrument types".

### Sample inclusion and exclusions (examples from tables)
- Selected included countries and carbon pricing policies by end-2020 (preserving exact labels where given):
  - ARG Yes 2018 XXXXX
  - CAN Yes 2019 2007 XXXXX
  - CHN Yes 2013
  - DEU Yes 2005 XXX
  - FRA Yes 2005 XXX
  - GBR Yes 2005 XXX
  - JPN Yes 2012 XXXXX
  - MEX Yes 2014 XXXXX
  - USA Yes 2009
  - ZAF Yes 2019 XXXX
  - CHE Yes 2008 XXXX
  - CHL Yes 2017 XX
  - COL Yes 2017 XXXXX
  - KAZ Yes 2013 XXX
  - UKR Yes 2011 XXX
- Selected countries excluded from main analysis due to insufficient data but with national carbon price by end-2020 (preserved exactly):
  - SGP 2019 XX
  - HRV Yes 2013 XXX
  - ISL Yes 2013 XXXXX
  - LIE Yes 2008 XXXX
  - NZL 2008 XXXX
  - BGR Yes 2007 XXX
  - AUT Yes 2005 XXX
  - BEL Yes 2005 XXX
  - CZE Yes 2005 XXX
  - EST Yes 2005 XXX
  - ... (table continues in source)

### Instrument categories and observed frequencies (sample of 37 countries, selected entries)
- Grants, subsidies, other fin. incentives:
  - Grants and subsidies 342
  - Tax relief 191
  - Feed-in tariffs or premiums 128
  - Loans 96
  - Fiscal or financial incentives (other) 91
  - Energy and other taxes 74
  - GHG emission reduction crediting and offsetting mechanism 43
  - GHG emissions allowances 42
  - Market-based instruments (other) 34
  - Tendering schemes 28
  - Net metering 19
  - other CO2 taxes 19
  - Economic instruments (other) 16
  - Retirement premium 11
  - User charges 8
  - Removal of fossil fuel subsidies 1
- Information and Education:
  - Information provision 315
  - Advice or aid in implementation 172
  - Endorsement label 78
  - Comparison label 69
  - Professional training and qualification 39
  - Information and education (other) 29
  - Green certificates 27
  - Performance label (other) 14
  - White certificates 13
  - Barrier removal (other) 1
- Policy Support:
  - Strategic planning 692
  - Institutional creation 177
  - Policy support (other) 153
  - Political & non-binding climate strategy 72
  - Political & non-binding GHG reduction target 67
  - Political & non-binding renewable energy target 62
  - Formal & legally binding renewable energy target 43
  - Formal & legally binding climate strategy 40
  - Formal & legally binding GHG reduction target 39
  - Political & non-binding energy efficiency target 29
  - GHG reduction target (other) 21
  - Formal & legally binding energy efficiency target 20
  - Renewable energy target (other) 20
  - Coordinating body for climate strategy 18
  - Target (other) 12
  - Energy efficiency target (other) 7
  - Climate strategy (other) 3
- Procurement and investment:
  - Infrastructure investments 123
  - Procurement rules 62
  - Funds to sub-national governments 44
  - Direct investment (other) 18
- Regulatory Instruments:
  - Other mandatory requirements 192
  - Monitoring 178
  - Product standards 129
  - Sectoral standards 115
  - Regulatory Instruments (other) 100
  - Building codes and standards 99
  - Vehicle fuel-economy and emissions standards 93
  - Obligation schemes 88
  - Auditing 75
  - Codes and standards (other) 58
  - Grid access and priority for renewables 29
  - Industrial air pollution standards 3
  - Vehicle air pollution standards 2
- Research, Development and Deployment:
  - Technology deployment and diffusion 131
  - Technology development 112
  - Demonstration project 108
  - RD&D funding 93
  - Research & Development and Deployment (RD&D) (other) 72
  - Research programme (other) 11
- Voluntary Approaches:
  - Negotiated agreements (public-private sector) 153
  - Public voluntary schemes 25
  - Voluntary approaches (other) 23
  - Unilateral commitments (private sector) 9

### Figures referenced
- Figure A2: Carbon pricing policies and the number of instrument types in countries’ policy portfolios (map as of end-2020). Number of instrument types shown only for countries without a national carbon price.
- Figure A3: Frequency heatmap of instrument types in different sectors (based on all policies adopted by countries in the sample). AFOLU = Agriculture, Forestry, and other Land Use.

*Source: wpiea2022066-print-pdf - 2.1 Empirical framework; 3.3 Policy portfolios of adopters versus non-adopters of carbon pricing; 1. Size of portfolio is measured as described.*

### 2.1    Empirical framework

### 2.1    Empirical framework

### Empirical expectations and causal framework
- Expectation: positive statistical association between the number of climate policies other than carbon pricing (referred to as the size of the climate policy portfolio) and both the adoption and the intensity of a pricing policy.
- Two causal channels:
  - Ex-post channel: earlier climate policies increase the probability of subsequent adoption of a pricing policy.
  - Ex-ante channel: an anticipated later adoption of a pricing policy motivates prior adoption of other policies.
- Mechanisms and barriers: Possible barriers to adoption of carbon pricing include concerns about high mitigation costs for firms in energy-intensive industries competing internationally, high costs for consumers and possibly a regressive distribution of these costs, political opposition to climate change policies, opposition to pricing policies, and concerns about effects on employment in fossil-fuel–reliant industries.
- Policy implication from framework: other climate policies (e.g., technology subsidies) can lower mitigation costs through technological innovation or gradually transform the economy away from carbon-intensive activities, potentially addressing barriers and increasing the likelihood of subsequent carbon pricing.

### Data
- Primary dataset: climatepolicydatabase.org; dataset composition and validation noted (Nascimento et al., 2021; validation with national stakeholders and experts).
- Coverage: dataset can generally be considered comprehensive for G20 economies (including EU member countries that are individual members of the G20, but not other EU members) and 18 additional countries (mostly advanced and emerging economies in Europe, Asia and Latin-America, some less developed countries, and two countries in Africa).
- Exclusions: countries for which the data cannot be considered comprehensive are dropped from the sample; this includes some early pioneers of carbon pricing (Norway, Sweden, Finland, Poland, Denmark).
- Policy filtering and treatment of EU policies:
  - Focus only on policies that have climate change mitigation as one of their objectives.
  - Neglect policies at the subnational level.
  - Apply all EU policies to member countries’ portfolios; if a country joined the EU after the policy decision, use the year of joining as the date of policy adoption.
- Instrument and sector taxonomy:
  - Original dataset distinguishes 72 instrument categories, aggregated to seven instrument types based on IEA typology; combined with Carbon Pricing Dashboard to create an eighth instrument type carbon pricing.
  - Final eight instrument types: Regulatory instruments; Grants, subsidies, and other financial incentives; Information and education; Policy support; Research, development, and deployment; Voluntary agreements; Procurement and investment; Carbon pricing.
  - Sectors: five sectors based on IPCC AR5 WGIII (Electricity and heat production; Transport; Buildings; Industry; Agriculture, forestry, and other land use) plus one additional sector General.
- Portfolio scoring:
  - Count how many of the eight instrument types have been implemented in each of the six sectors, yielding a score between 0 and 48 for every country and every year.
- Complementary country characteristics: GDP per capita (PPP) from the World Bank; education index from UNDP Human Development Indicators; control of corruption from World Governance Indicators (World Bank); fossil fuel reserves from US Energy Information Administration.
- Data scale and frequencies:
  - Every instrument type has been used in every sector in at least one country.
  - Number of times a specific instrument-sector combination is observed ranges from 18 (Research, development and deployment in Agriculture, forestry, and other land use) to 1902 (Policy support introduced without targeting a specific sector).
  - Total observed instrument-sector combinations: 14,540.

### Statistical methods
- Policy sequencing:
  - Eight instrument types lead to 40320 possible sequences.
  - Approach: consider all possible pairs of instrument types (28 pairs); for each pair, examine which instrument tends to be adopted first across sectors and countries using conditional frequencies.
  - Conditional frequency examined: f(Y_{t−1} | X_{t}) = n(Y_{t−1} ∧ X_{t}) / n(X_{t}), where X_t and Y_{t−1} are binary indicators of policy adoption up to year t and t−1 respectively.
  - Exclude observations after an event is observed for the first time (i.e., after the first time a specific instrument is adopted in a specific sector in a specific country).
- Matching and inference for adoption of carbon pricing:
  - Treated group: countries that adopted a national carbon price in a given year.
  - Control group: countries that did not adopt a carbon price neither earlier nor in the same year.
  - Matching procedure: for each treated country assign one randomly chosen control country; iterate this random assignment 1000 times and compare average portfolio sizes of treated and control countries.
  - Inference: calculate bootstrapped confidence intervals.
  - Additional regression: estimate a logit model with adoption of carbon pricing as binary dependent variable to include country characteristics as controls.
- Analysis among adopters:
  - Examine size of policy portfolios at time of adoption using linear regression with portfolio size at adoption as dependent variable and the same control variables.
  - Include year of adoption to examine trends over time.
  - Because of small sample relative to explanatory variables, also estimate a parsimonious model with only selected explanatory variables: GDP per capita and reserves of fossil fuels (include only reserves of coal and reserves of oil because reserves of oil and gas are highly correlated).
  - Use Lasso model selection to identify most important explanatory variables.
  - Also estimate linear regression with economy-wide average carbon price as dependent variable to examine association with portfolio size.

### Results — temporal sequence of climate policies (summary)
- Overall sequencing patterns:
  - Pooling policy adoption across all countries and sectors: carbon pricing tends to be the last instrument type.
  - Pooling across countries but keeping sectors separate: carbon pricing is the last instrument type in every sector.
  - In 12 out of the 15 countries examined, carbon pricing tended to be used for the first time in a specific sector after the use of any of the other instrument types.
  - Conclusion: carbon pricing tends to be implemented last, after adoption of all or almost all other seven instrument types.
- Two groups of instrument types by typical timing:
  - Early instrument types: Regulatory instruments; Grants, subsidies, and other financial incentives; Information and education; Policy support.
  - Late instrument types: Research, development, and deployment; Voluntary agreements; Procurement and investment; Carbon pricing.
  - With the exception of voluntary agreements and financial incentives in Agriculture, policies of the first group tend to be implemented before policies of the second group in all sectors.
- Cross-country variation:
  - Most frequent national pattern: relatively early adoption of Regulatory instruments and Policy support; relatively late adoption of Procurement and investment and Carbon pricing.
  - Greater variation across countries in the relative positions of Grants, subsidies, and other financial incentives; Research, development, and deployment; Voluntary agreements; Information and education.

### Results — build-up of climate policy portfolios over time
- Portfolio trajectories for countries that eventually adopted carbon pricing reveal at least three types:
  - Group 1 (e.g., Canada, Japan, South Africa): relatively rapid expansion of portfolio followed by slow further expansion over several years, eventually including carbon pricing.
  - Group 2 (e.g., Argentina, Switzerland): steady gradual expansion up until introduction of carbon pricing.
  - Group 3 (e.g., current EU members in the sample): rapid expansion of policies almost immediately followed by introduction of carbon pricing.
- Average time between adoption of new instrument type–sector combinations and adoption of carbon pricing differs across groups:
  - Canada: about 18 years.
  - Japan: about 13 years.
  - South Africa: about 14 years.
  - Argentina: about 11 years.
  - Switzerland: about 10 years.
  - Current EU countries in sample: about 5 years.

*Source: wpiea2022066-print-pdf - 2.1    Empirical framework.*

### 3.3    Policy portfolios of adopters versus non-adopters of carbon pricing

### 3.3    Policy portfolios of adopters versus non-adopters of carbon pricing

### Comparison of policy portfolios: adopters vs non-adopters
- At the time countries of the adopter group adopted carbon pricing, they had used on average 29.5 instrument type - sector combinations.
- By the end of 2020, countries without a national carbon price had used policies with on average about 23.9 instrument type - sector combinations.
- In 2015 (the average year of adoption of carbon pricing) adopters had used on average about 19.7 instrument type - sector combinations.
- Using a matching procedure with random assignment (which permits control countries that adopted later), countries that adopted a carbon price in a given year had policy portfolios that were on average about 11.12 instrument type-sector combinations larger than the portfolios of those that did not; this difference is significant at a confidence level of α= 0.05.

### Statistical inference and robustness checks
- Bootstrapped confidence intervals were calculated for inference (see Section 2).
- A logit model with adoption of carbon pricing as the binary dependent variable finds a statistically significant positive association between the size of the policy portfolio and adoption (Table A2).
- Results of a Lasso estimation suggest that the size of portfolio and the prevalence of gas reserves are the strongest determinants of the adoption of carbon pricing.
- As additional control variables are included, the estimated coefficient of the size of the portfolio becomes smaller and less significant, suggesting part of the positive association is due to country characteristics that influence both portfolio size and adoption.

### Key regression/logit coefficients (from Table A2)
- Policy portfolio coefficient, Column 1: 0.1072 (∗∗∗)
- Policy portfolio coefficient, Column 2: 0.0485
- Policy portfolio coefficient, Column 3: 0.0808 (∗∗∗)
- Policy portfolio coefficient, Column 4 (Lasso-selected): 0.0994 (∗∗∗)
- Gas reserves coefficient examples: -0.0085 (∗), -0.0073, -0.0108
- Education coefficient (Column showing significance): 19.0651 (∗∗∗)
- Intercept examples: -4.4354 (∗∗∗), -20.3759 (∗∗), -8.8369 (∗∗∗), -3.9841 (∗∗∗)
- Ps. R2 examples: 0.33, 0.51, 0.39, 0.35
- AIC examples: 641.66, 479.73, 595.02, 625.47
- Sample: No. countries 36; N 1188

### Figure statistics referenced
- Figure 6 (size of portfolio vs year of adoption): R2 = 0.79
- Figure 6 (price intensity vs size of portfolio): R2 = 0.44

### Interpretation of statistical results
- Countries that adopted carbon pricing in a given year tended to have significantly larger climate policy portfolios than those that did not.
- The positive association between portfolio size and adoption is robust across matching, logit, and Lasso methods, but sensitivity to additional controls indicates potential confounding by country characteristics.
- Lasso variable selection highlights portfolio size and gas reserves prevalence as leading predictors in the available data.

### Heterogeneity among adopters and temporal patterns
- Large variation exists in the size of countries’ policy portfolios at the time of adoption.
- After controlling for country characteristics, the size of policy portfolios at the time of adopting a carbon price has decreased over the last 20 years (Figure 6 left).
  - Example: When Canada adopted a national carbon price in 2019, its policy portfolio was substantially smaller than France’s portfolio in 2003, after controls.
  - This decreasing pattern is robust to different model specifications (Table A3) and relatively robust to dropping possible outliers, including scenarios where the group of EU ETS countries is considered as one observation or dropped.
- Regressing the average carbon price of the first year of implementation on portfolio size (controlling for the same characteristics) yields a positive association:
  - Countries with larger policy portfolios at adoption tended to implement carbon prices with higher price levels (Figure 6 right).
  - Japan is a notable outlier, implementing a relatively low price given its relatively large policy portfolio.
  - This association is less significant and less robust to dropping individual countries than the portfolio–adoption association.

### Mechanisms and interpretation in the Discussion and Conclusions
- The evidence is consistent with the idea that prior climate policies can remove barriers to subsequent adoption of carbon pricing and can pave the way for higher initial price stringency.
- Carbon pricing has generally been adopted relatively late in countries’ individual policy sequences.
- Three broad trajectories of portfolio buildup prior to carbon pricing were observed:
  - Gradual expansion of policy portfolios over time.
  - Rapid expansion followed immediately by adoption of national carbon pricing.
  - Rapid expansion followed by several years’ delay before adopting national carbon pricing.
- Possible drivers of different trajectories include exposure to domestic and international events, domestic barriers to carbon pricing, and whether carbon pricing was part of a long-term climate strategy.
- Over the last 20 years, the declining size of portfolios at adoption is consistent with declining abatement costs and international influences such as technological diffusion and growing economic opportunities for green technologies.
- Countries with exhausted extensive margins of instrument types and sectors may have three avenues for future climate policy:
  - Higher stringency of existing policies.
  - Additional policies of existing sector and instrument type combinations.
  - Carbon pricing as a final step in the sequence.

### Implications for future research and policy
- The observed associations are not necessarily causal because some confounders cannot be fully controlled for; causal interpretation should be cautious.
- Future empirical work should:
  - Expand data coverage to include some of the earliest adopters of carbon pricing as data become available.
  - Investigate how early climate policies influence the effectiveness of subsequent carbon pricing (beyond their association with initial price levels), including impacts on GHG reductions and emission leakage.
  - Explore how the size of climate policy portfolios influences the reductions in GHG emissions obtained with a given carbon price.

*Source: 3.3 Policy portfolios of adopters versus non-adopters of carbon pricing (excerpt).*

### 1.  Size of portfolio is measured as described

### 1.  Size of portfolio is measured as described

### Regression analysis: determinants of portfolio size at adoption and carbon price at implementation
- Dependent variables:
  - Size of portfolio at time of adoption (Columns 1–4)
  - Carbon price at implementation (Columns 5–8)
- Notes on significance notation preserved:
  - * p < 0.1, ** p < 0.05, *** p < 0.01
- Coefficients and reported standard errors (in parentheses) by explanatory variable and column:

- log GDP per capita PPP
  - Column 1: -0.3065 (0.6389)
  - Column 2: -0.3598 (0.3817)
  - Column 3: -0.0551 (0.4820)
  - Column 4: 0.3032 (0.2632)

- Coal reserves
  - Column 1: -0.2631 (0.3648)
  - Column 2: -0.4717 *** (0.1061)
  - Column 3: -0.3787 *** (0.0904)
  - Column 4: 0.1396 (0.2898)
  - Column 5: 0.0563 (0.1056)

- Oil reserves
  - Column 1: 0.6041 (0.2159)
  - Column 2: 0.4866 ** (0.1808)
  - Column 3: 0.3336 *** (0.0669)
  - Column 4: -0.1718 (0.1832)
  - Column 5: -0.1624 ** (0.0571)
  - Column 6: -0.1193 (0.0722)

- Gas reserves
  - Column 1: -0.9966 (1.0017)
  - Column 2: 0.3461 (0.9932)

- Control of corruption
  - Column 1: -0.1574 (0.3191)
  - Column 2: 0.5966 (0.3191)
  - Column 3: 0.3323 ** (0.1357)

- Education
  - Column 1: 0.1547 (0.1926)
  - Column 2: -0.2415 (0.1894)

- Policy portfolio
  - Column 1: 0.5960 *** (0.1507)
  - Column 2: 0.4805 * (0.2066)
  - Column 3: 0.4957 ** (0.1577)
  - Column 4: 0.3960 ** (0.1320)

- Year of adoption
  - Column 1: -0.5067 ** (0.2043)
  - Column 2: -0.9301 * (0.4650)
  - Column 3: -0.9112 ** (0.3138)
  - Column 4: -0.6283 *** (0.1679)

- Intercept
  - Column 1: -0.0336 (0.1883)
  - Column 2: 0.0176 (0.1798)
  - Column 3: 0.1769 (0.1281)
  - Column 4: 0.0722 (0.1873)
  - Column 5: -0.1687 (0.1959)
  - Column 6: -0.0713 (0.3606)
  - Column 7: -0.1265 (0.2123)
  - Column 8: -0.1332 (0.2012)

- Goodness of fit and sample size
  - R2 by column: 0.27, 0.76, 0.71, 0.66, 0.42, 0.66, 0.53, 0.58
  - N reported: 15 for columns 1–8 (table shows 15 repeated), and a separate line shows 24 (context unclear in excerpt)

- Additional analytic notes reported in source text:
  - Size of portfolio is measured as described in Section 2.
  - Carbon price at implementation is calculated as an economy-wide average price.
  - The positive coefficient of oil reserves in Columns 2–4 becomes insignificant if Canada is dropped from the sample, consistent with the results by Best and Zhang (2020).

### Policy portfolios and carbon pricing status by end-2020
- Policy portfolios by the end of 2020 (map summary in source):
  - Categories shown include: Carbon pricing and policy portfolios by end of 2020
  - Instrument-type counts reported visually for countries without a national carbon price: examples noted as "6/8 instrument types" and "7/8 instrument types"
  - Carbon price (national) and Carbon price (subnational) distinguished; some cases marked "Not enough data" or "Fewer instrument types"

### Sample of countries included and exclusions
- Table A4: List of countries included in the main analysis and their carbon pricing policies by the end of 2020.
  - Columns indicate: ISO code, G20, Other major emitter, National, Subnational, EIBTA, and sectors (E = Electricity and Heat Production, I = Industry, B = Buildings, T = Transport, A = AFOLU)
  - Examples from the included list (preserving exact labels as in source):
    - ARG Yes 2018 XXXXX
    - CAN Yes 2019 2007 XXXXX
    - CHN Yes 2013
    - DEU Yes 2005 XXX
    - FRA Yes 2005 XXX
    - GBR Yes 2005 XXX
    - JPN Yes 2012 XXXXX
    - MEX Yes 2014 XXXXX
    - USA Yes 2009
    - ZAF Yes 2019 XXXX
    - CHE Yes 2008 XXXX
    - CHL Yes 2017 XX
    - COL Yes 2017 XXXXX
    - KAZ Yes 2013 XXX
    - UKR Yes 2011 XXX
  - (Table lists additional countries; excerpt preserves entries above without inventing missing rows.)

- Table A5: Countries not included in the main analysis because of insufficient data on policy adoption, but which had a national carbon price implemented by end-2020.
  - Examples preserved exactly as listed:
    - SGP 2019 XX
    - HRV Yes 2013 XXX
    - ISL Yes 2013 XXXXX
    - LIE Yes 2008 XXXX
    - NZL 2008 XXXX
    - BGR Yes 2007 XXX
    - AUT Yes 2005 XXX
    - BEL Yes 2005 XXX
    - CZE Yes 2005 XXX
    - EST Yes 2005 XXX
    - ... (table continues in source)

### Instrument types, categories, and frequencies (sample of 37 countries)
- Table A6 and A7 list instrument categories, their instrument type, and their observed frequency in the sample of 37 countries. Selected entries (exact numbers preserved):

- Grants, subsidies, other fin. incentives
  - Grants and subsidies 342
  - Tax relief 191
  - Feed-in tariffs or premiums 128
  - Loans 96
  - Fiscal or financial incentives (other) 91
  - Energy and other taxes 74
  - GHG emission reduction crediting and offsetting mechanism 43
  - GHG emissions allowances 42
  - Market-based instruments (other) 34
  - Tendering schemes 28
  - Net metering 19
  - other CO2 taxes 19
  - Economic instruments (other) 16
  - Retirement premium 11
  - User charges 8
  - Removal of fossil fuel subsidies 1

- Information and Education
  - Information provision 315
  - Advice or aid in implementation 172
  - Endorsement label 78
  - Comparison label 69
  - Professional training and qualification 39
  - Information and education (other) 29
  - Green certificates 27
  - Performance label (other) 14
  - White certificates 13
  - Barrier removal (other) 1

- Policy Support
  - Strategic planning 692
  - Institutional creation 177
  - Policy support (other) 153
  - Political & non-binding climate strategy 72
  - Political & non-binding GHG reduction target 67
  - Political & non-binding renewable energy target 62
  - Formal & legally binding renewable energy target 43
  - Formal & legally binding climate strategy 40
  - Formal & legally binding GHG reduction target 39
  - Political & non-binding energy efficiency target 29
  - GHG reduction target (other) 21
  - Formal & legally binding energy efficiency target 20
  - Renewable energy target (other) 20
  - Coordinating body for climate strategy 18
  - Target (other) 12
  - Energy efficiency target (other) 7
  - Climate strategy (other) 3

- Procurement and investment
  - Infrastructure investments 123
  - Procurement rules 62
  - Funds to sub-national governments 44
  - Direct investment (other) 18

- Regulatory Instruments
  - Other mandatory requirements 192
  - Monitoring 178
  - Product standards 129
  - Sectoral standards 115
  - Regulatory Instruments (other) 100
  - Building codes and standards 99
  - Vehicle fuel-economy and emissions standards 93
  - Obligation schemes 88
  - Auditing 75
  - Codes and standards (other) 58
  - Grid access and priority for renewables 29
  - Industrial air pollution standards 3
  - Vehicle air pollution standards 2

- Research, Development and Deployment
  - Technology deployment and diffusion 131
  - Technology development 112
  - Demonstration project 108
  - RD&D funding 93
  - Research & Development and Deployment (RD&D) (other) 72
  - Research programme (other) 11

- Voluntary Approaches
  - Negotiated agreements (public-private sector) 153
  - Public voluntary schemes 25
  - Voluntary approaches (other) 23
  - Unilateral commitments (private sector) 9

### Figures referenced
- Figure A2: Carbon pricing policies and the number of instrument types in countries’ policy portfolios (map as of end-2020). Number of instrument types shown only for countries without a national carbon price.
- Figure A3: Frequency heatmap of instrument types in different sectors (based on all policies adopted by countries in the sample). AFOLU = Agriculture, Forestry, and other Land Use.

*Policy Sequencing Towards Carbon Pricing — Working Paper No. WP/2022/066*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2022/english/wpiea2022066-print-pdf.pdf_
