## Annex Table 3.1.1. Aggregate Effect of Environmental Policy on Clean Innovation ## Source details **Canonical URL:** [Annex Table 3.1.1. Aggregate Effect of Environmental Policy on Clean Innovation](https://www.imf.org/-/media/files/publications/weo/2020/october/english/annexch3.pdf) ## Other formats - [Markdown version](/-/media/files/publications/weo/2020/october/english/annexch3.pdf.md) - [Structured JSON version](/-/media/files/publications/weo/2020/october/english/annexch3.pdf.json) --- ### Overview and identification strategy - Dependent variable in the main regressions: CCM energy patent families / CCM energy patent applications (climate change mitigating). - All regressions include country and year fixed effects. T-statistics reported in parentheses. - Definitions and abbreviations preserved: EPS = environmental policy stringency; CCM = climate change mitigating; tech stock = patent stock in specific technology; all tech = total patenting in all technologies; bb = billions of barrels; ETCR = energy, transport and communication regulation. - Year fixed effects from the baseline regression were regressed on the country-specific EPS indicator and oil prices to assess the contribution of these global factors to the time trend in innovation. ### Contribution of EPS and oil prices to global innovation trend - Change in the EPS indicator is responsible for 37 percent of the increase between 1990 and 2010 in global innovation captured by year fixed effects. - Oil prices account for about twice as much: contribution from the increase in oil prices is about double the EPS contribution. - Individual contributions to variation in year fixed effects: - Environmental tightening (EPS): 30 percent. - Oil prices: 46 percent. - Joint contribution of EPS and oil prices: 61 percent. - Note: indirect, second-round effects through a larger knowledge stock are acknowledged but considered of second order. ### Effect of individual environmental policies on clean innovation (selected regression coefficients) - Regression framework: country and year fixed effects; dependent variable: CCM Energy patent applications. - Key coefficient estimates (selected, with t-statistics in parentheses): - Log tech stock t-1: - Column (1): 0.551*** (14.20) - Column (2): 0.531*** (13.72) - Column (3): 0.596*** (11.29) - Column (4): 0.543*** (12.32) - Column (5): 0.508*** (13.22) - Column (6): 0.519*** (10.48) - Log all tech t-1: - Column (1): 0.492*** (9.28) - Column (2): 0.506*** (10.40) - Column (3): 0.438*** (7.16) - Column (4): 0.448*** (10.23) - Column (5): 0.593*** (12.66) - Column (6): 0.525*** (9.71) - Policy coefficients (selected): - CO2 taxes: 0.0105 (0.47) and −0.016 (0.55) — insignificant. - Trading schemes: 0.0333** (2.03) and 0.0320*** (2.79). - Feed-in tariffs: 0.0278*** (3.10) and 0.0207* (1.67). - Emission limits: 0.0511** (2.29) and 0.0388* (1.65). - R&D subsidies: 0.0693*** (5.41) and 0.0616*** (3.16). - Number of observations by column: 788, 785, 788, 788, 788, 785. - Statistical significance legend: *** p<0.01, ** p<0.05, * p<0.1. - Interpretation: - Both non-market policies (emission limits, R&D subsidies) and market policies (trading schemes, feed-in tariffs) made a statistically significant contribution to clean innovation. - Carbon taxes show an insignificant effect, plausibly due to limited use: only slightly more than 10 percent of countries used carbon taxes in 2015, while other policy tools were used by 60-100 percent of countries in the sample. ### Electricity sector: clean, gray, and dirty innovation (selected coefficients) - Focus: electricity-related patent families classified as Clean, Gray (environmental improvements to dirty tech), Dirty, and Total. - EPS t-1 coefficients: - Clean (column 1): 0.0688*** (0.02) - Gray (column 2): 0.151*** (0.04) - Dirty (column 3): −0.0405** (0.02) - Total (column 4): 0.175*** (0.04) - Other selected coefficients: - Log all tech t-1: - Clean: 0.0522 (0.10) - Gray: −0.338*** (0.10) - Dirty: 0.0543** (0.02) - Total: 0.450*** (0.07) - Log knowledge stock specific to the type of electricity t-1: - Gray: 0.531*** (0.07) (Clean and Dirty not significant in some specs) - Log oil and gas reserves t-1: - Clean: −0.129** (0.05) - Gray: −0.0843 (0.06) - Dirty: 0.0398 (0.02) - Number of observations by specification: 738, 743, 743, 781. - Interpretation: - EPS tightening increases the relative share of clean innovation and increases gray innovation even more. - Dirty innovation declined (negative coefficient), and overall electricity innovation (Total) increased; the decline in dirty innovation was more than offset by increases in clean and gray innovation. ### Synthesis on innovation and electricity sector - Environmental policy tightening has a statistically and economically significant effect on clean energy innovation. - In the electricity sector, policy tightening: - Shifts innovation away from dirty toward clean and gray technologies. - Yields a net positive effect on electricity innovation overall. - Market-based support policies (feed-in tariffs, green certificates) and electricity market deregulation are particularly effective in increasing the renewable electricity share. - Carbon taxes show an insignificant effect on clean innovation in the sample, likely reflecting limited adoption. --- ### Annex Table 3.2.1. Main Results (1990–2014): Renewable electricity share and drivers ### Data and EPS index - Dependent variable: annual change in the share of non-hydro renewable energy in electricity generation (∆y_i,t), using IEA (2019) data for 1990-2017. - EPS index: OECD Environmental Policy Stringency (scale 0 to 6) covering 32 OECD and emerging market countries between 1990-2015 and including taxes on NOx/SO2/PM, CO2 and SO2 trading schemes, renewable/green certificates, white certificates, feed-in tariffs, emission limits, and government R&D for renewable energy. ### Main findings (Model 1 and disaggregations) - Model 1 (average tightening of market-based environmental policies between 1990 and 2010): - Can explain a 0.38 percentage point increase in the share of renewable electricity generation per year. - Contextualization of 0.38 percentage points: - Equivalent to 29 percent of the average model-implied increase in the share of renewable electricity in 2014. - Equivalent to 55 percent of the actual increase of 0.69 percentage points in the sample of 32 countries in 2014. - Model 2 (disaggregating market-based policies): - A one standard deviation tightening of: - Feed-in tariffs increases renewable share by 0.118 percentage points per year. - Trading schemes increases renewable share by 0.183 percentage points per year. - Illustrative case: Germany scoring 4 or higher on OECD feed-in tariff variable between 1997 and 2007 — Model 2 implies such a policy for a decade would add 2.5 percent to its share of renewable electricity, with a cumulative indirect effect adding another 7.5 percent over the same decade. - Model 3 (trading-scheme subtypes): - Green certificates are the only trading-scheme subtype with statistically significant evidence. - A one standard deviation change in green certificates (controlling for other policies) increases renewable share by 0.116 percentage points per year. - CO2 trading schemes show no significant effect, attributed to limited sample variation and relatively weak average policy strength. - Robustness (Models 4–6): - Coefficients on environmental policy variables are broadly insensitive to additional controls. - Statistically significant negative effects found for income and the share of nuclear power in electricity generation. - Electricity market deregulation: - The average de-regulation of electricity markets in OECD countries between 1990 and 2010 supported an annual increase of 0.38 percentage point in the share of renewable electricity generation. - One standard deviation increase in deregulation corresponds to a 0.223 percentage point increase in renewable share per year. ### Selected regression coefficients (Δ(electricity share of renewables)) - Solar wind and biomass share MA t-1 coefficients (columns 1–6): - 0.0852** (standard error 0.0304) - 0.0791* (0.0361) - 0.0826* (0.0365) - 0.0808* (0.0361) - 0.0741+ (0.0410) - 0.0799+ (0.0419) - Policy variables (selected with standard errors preserved exactly): - Market EPS: 0.231* (0.0974) and 0.249* (0.101) - Non-market EPS: −0.124 (0.106), −0.147 (0.113), −0.146 (0.113), −0.109 (0.122), −0.140 (0.131), −0.138 (0.130) - Market EPS feed-in tariff: 0.0621* (0.0287), 0.0625* (0.0290), 0.0658* (0.0305), 0.0666* (0.0307) - Market EPS trading: 0.166** (0.0543), 0.174** (0.0622) - Market EPS trading green certificates: 0.0918* (0.0373), 0.108** (0.0322) - Market EPS trading CO2: 0.0392 (0.0360), 0.0298 (0.0388) - Log electricity PMR t-1 coefficients (columns 1–6): −0.440** (0.159), −0.441** (0.149), −0.445** (0.141), −0.563** (0.197), −0.557** (0.184), −0.568** (0.174) - Sample and fit: - Number of observations: 652, 652, 652, 558, 558, 558 - Number of countries: 32, 32, 32, 28, 28, 28 - R2: 0.28 across columns - Notes: Robust standard errors clustered at the country level in parentheses; variables in logarithmic scale. ### Electricity mix and electricity generation per capita (Annex Table 3.2.2 summary) - Policy indicators such as feed-in tariffs and CO2 schemes have a positive relationship with the share of solar, wind and biomass in electricity generation. - These policies do not appear to have a discernible impact on total electricity generation per capita. - EPS associations with fossil fuel shares: - Relationships with annual change of coal share tend to be negative (though not always statistically significant). - Effects on the share of natural gas are ambiguous; natural gas can complement intermittent renewables due to dispatchability. - Interpretation caveat: results are associations rather than strict causal estimates; instrumental variable evidence in the literature provides mixed guidance, with some studies suggesting OLS may underestimate the effect of feed-in tariffs. --- ### Employment effects of EPS — methodology, samples, and main findings ### Estimating approach and data - Basic specification: augmented labor demand equation at firm level (employment n_i,j,t), controlling for lags of employment, firm, country, and year fixed effects. - Interaction of interest: EPS_j,t × d_i, where d_i = 1 if CO2 emissions = "high", 0 otherwise. - Specification 1: interaction uses firm CO2 emissions dummy (requires firms reporting at least 3 instances of CO2 emissions). - Specification 2: interaction uses sector dummy to proxy CO2 emission intensity (expands sample coverage). - Estimation methodology: panel GMM following Van Reenen (1997). - Controls X_i,j,t include: - Firm-level average annual employee wages - Real capital stock (sum of building and machinery capital stock deflated by building and machinery price indices from Penn World Tables) - Rental rate of capital (proxied by price of capital services from Penn World Tables) - Output gap - Variable treatment: all variables in logs except EPS indicator which enters in levels. ### Samples and summary statistics - Firm-level data source: Worldscope database. - Sample sizes: - Specification 1 (firm CO2 dummy): 670 firms, 30 countries, 2000–15. - Specification 2 (sector dummies): 5,305 firms, 31 countries, 2000–15. - EPS variables: OECD indices on scale 0 to 6 for 32 OECD and emerging market countries, 1990–2015. - Selected summary statistics (Sample 1): - Aggregate EPS: Mean 2.5, Std. Dev. 0.9, Min 0.4, Max 4.1. - Market EPS: Mean 1.9, Std. Dev. 1.0, Min 0.04, Max 4.0. - EPS: CO2 tax: Mean 0.1, Std. Dev. 0.9, Min 0.06, Max 6.0. - Log employees: Mean 9.6, Std. Dev. 1.5, Min 2.8, Max 13.4. - Selected summary statistics (Sample 2): - Aggregate EPS: Mean 2.2, Std. Dev. 0.9, Min 0.4, Max 4.1. - Market EPS: Mean 1.7, Std. Dev. 0.9, Min 0.04, Max 4.0. - Log employees: Mean 7.8, Std. Dev. 1.9, Min 0.0, Max 13.4. ### Main findings (Specification 1: interaction with firm CO2 emissions) - High-emission firms experience negative employment effects when EPS tightens; low-emission firms experience employment increases (effect for low-emission firms not always significant). - Estimated semi-elasticities: - A 1-standard-deviation tightening in the EPS indicator would lower employment in high-carbon firms by 5 percent. - The same tightening would raise employment in low-carbon firms by 2.6 percent. - Market EPS and carbon taxation specifications yield qualitatively similar patterns; some carbon taxation estimates are not significant. - Non-market EPS: - A 1-standard-deviation tightening in non-market EPS would lower employment in high-emission firms by 5 percent and raise employment in low-emission firms by 4.4 percent. ### Cyclical sensitivity and robustness - Interaction of output gap with EPS shows: - Market EPS can have a positive employment effect when the output gap is negative (severe contractionary conditions). - Average marginal effects of market EPS on employment: modestly positive under severe contractionary conditions; turn negative during normal/expansionary periods. - Robustness: - Results robust to changing high/low emission threshold (e.g., 75th percentile). - Results robust to excluding firms with fiscal year ending before December (signs retained; significance weaker). - Selection concerns on CO2 reporting discussed; reporting rates vary by country and sector but patterns suggest selection bias unlikely to invalidate findings. --- ### Sectoral heterogeneity, net employment impacts, and medium-term dynamics (Annex Table 3.3.4 and related) ### Sectoral specification and key coefficients (selected) - Six sectors used as proxies for emission intensity: fossil fuel industries; high-emission manufacturing (food, metals and minerals, chemicals, paper and packaging); services; construction; transport; other (low-emission) manufacturing. - Selected lag and control coefficients: - Log N t-1: 0.573*** (Annex Table 3.3.2, col 1); 0.543*** (Annex Table 3.3.4, col 1). - Log capital stock: 0.276** (Annex Table 3.3.2, col 1); 0.186*** (Annex Table 3.3.4, col 1). - Log wages: −0.260** (Annex Table 3.3.2, col 1); −0.218*** (Annex Table 3.3.4, col 1). - Log r: 0.308** (Annex Table 3.3.2, col 1); 0.241*** (Annex Table 3.3.4, col 1). - Sector × EPS interaction coefficients (selected): - Construction × EPS: −0.0836*** (col 1); −0.0693** (col 2); −0.0717*** (col 5). - High CO2 industries × EPS: mixed but some specifications show −0.0346*. - Services × EPS: 0.0174 (col 1); 0.0225* (col 2). - Fossil fuels × EPS: mixed signs across specifications; tax policies vs trading schemes differ. - Utilities × EPS: −0.0226 to −0.0266 in some specs. - Joint p-values for sector interactions reported across columns: 0.05, 0.06, 0.65, 0.05, 0.05, 0.07, 0.03, 0.07, 0.10. ### Net employment impact (preferred specifications) - Based on preferred specifications including output gap (columns 6 and 9): - Net loss of between 500-600 thousand jobs in the sample for aggregate EPS tightening by 1 standard deviation (about 1 percent of total employment in the sample). - For tightening market EPS by 1 standard deviation: - Small net job increase between 13-34 thousand jobs (estimates in columns 7 and 10). ### Medium-term dynamics and business-cycle dependence - Medium-term regressions show short-term effects of aggregate, market, and non-market EPS tend to reverse and fade away over the medium term. - Short-term effects summary: - Tightening environmental policies associated with reallocation of jobs: employment rises in low-CO2 firms and falls in high-CO2 firms (when emissions proxied by sector). - Overall impact depends on policy mix: past changes suggest short-term negative effects for overall EPS (likely driven by negative effects of non-market policies); greater reliance on market-based policies may yield positive overall effects. - Short-term effects are "quite modest" and tend to fade in the medium term. - Business-cycle dependence: - Impact of tightening market-based policies depends on business cycle: under highly contractionary conditions the impact may be positive; under normal/expansionary periods the effect is negative. --- ### G‑Cubed model inputs, Net Zero assumptions, and carbon removal potentials ### G‑Cubed model version and structure - Model version: GGG20v154 (10 regions and 20 sectors). - Database updated to GTAP10 and latest IMF, World Bank, OECD, UN, and US EIA data. - Regions: 10 region codes including AUS, CHN, EUW, IND, JPN, OPC, OEC, ROW, RUS, USA. - 20 sectors listed (1 Electricity delivery; 2 Gas extraction and utilities; …; 20 Other generation). - Model features include full accounting for stocks and flows of assets, Henderson-McKibbin-Taylor monetary rules, nominal wage stickiness, sector-specific hiring rules, heterogeneous households and firms, and an exogenous budget deficit fiscal rule with lump-sum tax adjustments. ### Net Zero by 2050 and carbon removal assumptions - Carbon removal potentials (Fuss (2018) summary): - Authors take average of ranges and sum to 13.8Gt CO2. - Conservative assumption: 75% of 13.8 Gt achievable by 2050 → about 10Gt CO2 per year. - 10Gt CO2 per year corresponds to about 20% of global CO2 emissions in the baseline in 2050. - Regional emission reduction assumption: - All regions reduce emissions by 80% by 2050 relative to 2018 except OPC, which remains at 2018 level by 2050. - Carbon tax growth-rate assumptions: - Main assumption: constant growth rate of 7 percent for carbon taxes over 2019-2050, then solve tax rates to achieve emissions targets by 2050 after accounting for other policy layers. - Comparison scenario: growth rate of 5 percent for carbon taxes to achieve same targets by 2050. ### Annex Table 3.4.3 — Global Carbon Removal Potentials in 2050 (Gigaton CO2 ranges) - Afforestation and reforestation: 0.5–3.6 - BECCS: 0.5–5 - Biochar: 0.5–2 - Enhanced weathering: 2–4 - DACCS: 0.5–5 - Soil carbon sequestration: Up to 5 ### Policy package and scenario design (selected) - Solar and wind output are subsidized at a rate of 80% for all regions since 2019. - Economy-wide productivity improvements: - Driven by avoided damages from climate change due to all other policies. - Imposed equally on all sectors except electricity generation. - Carbon tax revenue transfers: - 25% of carbon tax revenues are transferred to households as compensatory transfers. - Participation scenarios: - Largest 5 countries scenario: USA, EUW, CHN, IND, JPN participate (policy shocks use aggregate scenario shocks without re-solving carbon taxes). - Advanced economies scenario: USA, EUW, JPN, AUS, OEC participate (carbon taxes not re-solved). - Implementation timing: - Policy shocks applied from 2019 onward; simulation presented as starting in 2021 for presentation. ### Energy sector model structure (equations and R&D) - Energy produced by CES technology using N fuels; imported conventional oil indexed first. - Energy production formula preserved as in source (equation 3.5.1 structure). - Fuel-specific technologies and R&D cost functions preserved: - R&D cost form r_{i,t}(x_{i,j,t}/x̅_{i,j,t}) = ε_{i,j}(1−χ_{i,j,t})^{η−1}(x_{i,j,t}/x̅_{i,j,t})^{η/(η−1)} with η>1. - Carbon intensity R&D cost r_{i,t}(g_{i,j,t}/g̅_{i,j,t}) = θ_{i,j}(1−χ_{i,j,t})^{η−1}(g_{i,j,t}/g̅_{i,j,t})^{η/(η−1)}. - Endogenous technical change: input-saving and emissions-reducing R&D, convex research costs (η>1), aggregate market-size effect amplifies policy impact. ### Calibration and parameter values (selected) - Fuel types: seven fuels including international oil; domestic oil; natural gas; coal; hydroelectric; nuclear; renewables. - Selected calibrated parameters: - ρ (Intra-fuel CES parameter): 0.67 (Common). - η (R&D returns, common): 10. - σ (Aggregate CES parameter, common): −3. - Elasticities and targets: - Intra-fuel elasticity of substitution set to 3. - Elasticity of substitution of energy and capital-labor bundle set to 0.25. - Returns to R&D (η = 10) consistent with innovation response reducing required carbon tax by 20 percent in 20 years in Fried (2018) comparison. - Labor and productivity assumptions: - Labor force growth from ILO forecasts until 2030, converging to 0.2% annual growth by 2070. - Long-run labor productivity growth assumed 1.3% per year. --- ### CarMMa, intermittency, and electricity transition implications (Annex Figure 3.6.1 and model details) ### Electricity mixes and empirical context - Electricity and heating share of global CO2: - Emissions from electricity generation and heating: roughly 40 percent of total global CO2 emissions in 2018. - Regional per-capita emissions (electricity and heating, 2017): - European Union: 2,176 kg per year. - United States: 5,592 kg per year. - China: 3,312 kg per year. - Coal and gas shares: - EU: coal share over 20 percent in many places. - US: coal and gas together make up over 60 percent of generation. - China: coal share almost 70 percent. ### Technology readiness, costs, and feasibility - Renewables (wind and solar PV) identified as most promising politically acceptable carbon-neutral technology; key drawback is intermittency. - Hydraulic power limited by geography; biomass competes for land; nuclear scalable but politically constrained; CCS costly and unlikely to fall substantially in near term. - Mitigation policies more effective now due to rapid declines in renewable costs. ### Intermittency, storage, and backup requirements - Load factors: - Onshore wind: 35 percent. - Offshore wind: 45 percent. - Solar PV: about 25 percent. - Storage capacity: - Global hydro-pump power capacity: 158 GW. - In US, EU, China this hydro-pump capacity represents about 30 minutes of power consumption. - Managing solar intermittency would require about 18 hours of electricity storage. - Managing wind intermittency would require about 72 hours of electricity storage. - Interim solution: intermittency must be compensated by flexible sources (natural gas and hydro most common); coal retrofits possible but slow emission decline. ### CarMMa intermittency module and calibration - Any intermittent generation capacity required to be paired with dispatchable backup capacity idle most of the time but covering shortfalls. - Overcapacity and curtailment allowed; generation duration curves approximated by power function with intermittency parameter γ (load factor = 1/(1+γ)). - Intermittency parameter examples: offshore wind between 1 and 1.5; onshore wind between 2 and 3. - Cost intuition: variable cost savings from renewables exceed installation costs up to peak output equal to desired output L; beyond that curtailment increases rapidly. - CarMMa model features: - DSGE tailored to electricity transition; four electricity technologies (coal, natural gas, renewables, nuclear plus hydro). - Renewables paired with cost-efficient backup (hydro and natural gas common; coal with flexibility retrofits third). - Calibration reproduces national accounts and IEA shares for 2018. ### Selected sector shares (Percent of GDP) reproduced from source - United States / European Union / China: - Electricity: 1.9 3.0 2.3 - Manufacturing: 0.4 0.9 1.2 - Services: 0.5 0.7 0.6 - Consumption: 1.0 1.4 0.5 --- ### Electricity generation, carbon pricing scenarios, and macro implications (Annex Table 3.6.1 and related) ### Transmission of a carbon price into electricity prices and macro - Carbon price partially absorbed by mining sector; reduces fuel demand and lowers coal and gas prices in short run, dampening pass-through into electricity prices. - Rebalancing toward low-carbon technologies lowers average carbon intensity, so a given carbon price produces smaller electricity price increase. - Strength of rebalancing depends on availability of natural gas as backup. ### Scenario: 50USD carbon price phased in over 10 years (United States, European Union, China) - Policy design: Carbon price 50USD phased in over 10 years; carbon tax revenues returned to households as transfers. - Electricity mix adjustments: - Coal share declines in all regions; in United States the share falls below 10 percent given abundant natural gas. - Average 40-year lifetime of coal plants slows transition via gradual depreciation. - EU and China: coal use alongside gas as backup mitigates coal share decline. - Electricity price impacts after ten years (cumulative increase relative to baseline): - European Union: 10 percent - United-states: 20 percent - China: 30 percent - Macroeconomic effects after ten years (average annual growth reductions relative to baseline): - United-States: about 0.1 percentage point - European Union: about 0.1 percentage point - China: about 0.3 percentage point - Electricity-sector CO2 emission reductions after ten years (relative to baseline): - European Union: about 30 percent - United States: about 35 percent - China: about 38 percent - Absolute declines in electricity-related emissions after ten years: - United States: 745 megatons - European Union: 390 megatons - China: 1919 megatons ### Carbon tax with a macro package (United States and European Union) - Macro package components: - Frontloaded subsidies for investment in renewables, financed by public debt in first five years. - Accommodative monetary policy in short run. - Subsidy rates: - United States: initial subsidies 60 percent of investment costs in renewables, declining to 30 percent after five years. - European Union: initial rate 40 percent, declining to 20 percent after five years. - Effects: - Subsidies boost short-term investment and accelerate transition, lowering average carbon intensity. - Lower carbon intensity dampens electricity price and GDP impacts. - Macro package compensates short-run output decline, mitigates long-run decline, and amplifies emissions reductions. ### Carbon tax with macro package plus additional nuclear and gas (China) - Policy mix: 50 USD carbon tax phased in + expansion of nuclear power + improved availability of natural gas as backup. - Effects: - Additional measures cut output costs by roughly a half and amplify emissions decline by about 50 percent (relative to isolated 50 USD carbon price scenario). - Additional nuclear capacity increases electricity supply immediately and crowds out coal. - Subsidies for natural gas generation support flexible backup capacity, further reducing coal share. ### Policy implications and options to minimize financial damages - Existing coal fleet concern: - Many coal plants young: 60 percent are 20 years or younger, complicating rapid retirement. - IEA: existing coal power plants represent globally more than $1 trillion unrecovered capital investment. - Model simulations show dramatic declines in value (Tobin’s Q) of coal power plants under transition scenarios. - Options to minimize financial damages: - Retrofitting and repurposing younger, more efficient plants to be compatible with climate targets (e.g., CCUS or biomass co-firing). - Operating repurposed plants at lower utilization to provide flexibility for renewables. ### Distributional effects and revenue recycling (50 USD per tCO2 carbon tax) - Key channels increasing income inequality: - Low-income households spend larger share on high-energy intensive goods. - Carbon taxes can reduce wages and job opportunities of unskilled low-income workers; skill premium increases. - Model framework: multi-sector heterogeneous agent model with skilled and unskilled households; elasticities specified for calibration. - Revenue recycling cases analyzed: - (i) Low-energy government spending - (ii) Universal cash-transfers - (iii) Targeted cash-transfers to bottom two quintiles - (iv) Feebates (subsidy to clean energy consumption) --- ### Distributional impact of a 50 USD per tCO2 tax (Annex Table 3.7.1 — percentage change with respect to baseline) ### United States — Percentage change with respect to baseline - Gini coefficient: - Low-energy government spending: 0.35 - Universal cash-transfers: −1.35 - Targeted cash-transfers: −2.24 - Feebates: −0.28 - Skill premium: - Low-energy government spending: 1.72 - Universal cash-transfers: 1.10 - Targeted cash-transfers: 0.14 - Feebates: −0.41 ### China — Percentage change with respect to baseline - Gini coefficient: - Low-energy government spending: 0.12 - Universal cash-transfers: −3.81 - Targeted cash-transfers: −4.52 - Feebates: −0.27 - Skill premium: - Low-energy government spending: 2.52 - Universal cash-transfers: 1.53 - Targeted cash-transfers: 0.24 - Feebates: −1.51 ### Interpretation - Without compensatory measures, carbon taxes increase income inequality (Gini rises). - Universal or targeted transfers can reduce inequality; targeted transfers to bottom two quintiles are most effective. - Feebates mitigate price impacts on bottom-income households and can reduce the skill premium by boosting unskilled labor demand in clean energy production. *Source: IMF staff calculations.* ### Annex Table 3.1.1. Aggregate Effect of Environmental Policy on Clean Innovation ### Annex Table 3.1.1. Aggregate Effect of Environmental Policy on Clean Innovation ### Overview and identification strategy - Dependent variable in the main regressions: CCM energy patent families / CCM energy patent applications (climate change mitigating). - All regressions include country and year fixed effects. T-statistics reported in parentheses. - EPS = environmental policy stringency; CCM = climate change mitigating; tech stock = patent stock in specific technology; all tech = total patenting in all technologies; bb = billions of barrels; ETCR = energy, transport and communication regulation. - Year fixed effects from the baseline regression were regressed on the country-specific EPS indicator and oil prices to assess the contribution of these global factors to the time trend in innovation. ### Contribution of EPS and oil prices to global innovation trend - Change in the EPS indicator is responsible for 37 percent of the increase between 1990 and 2010 in global innovation captured by year fixed effects. - Oil prices account for about twice as much: contribution from the increase in oil prices is about double the EPS contribution. - Individual contributions to variation in year fixed effects: - Environmental tightening (EPS): 30 percent. - Oil prices: 46 percent. - Joint contribution of EPS and oil prices: 61 percent. - Note: indirect, second-round effects through a larger knowledge stock are acknowledged but considered of second order. ### Effect of individual environmental policies on clean innovation (Annex Table 3.1.2) - Regression framework: country and year fixed effects; dependent variable: CCM Energy patent applications. - Key coefficient estimates (selected): - Log tech stock t-1: - Column (1): 0.551*** (14.20) - Column (2): 0.531*** (13.72) - Column (3): 0.596*** (11.29) - Column (4): 0.543*** (12.32) - Column (5): 0.508*** (13.22) - Column (6): 0.519*** (10.48) - Log all tech t-1: - Column (1): 0.492*** (9.28) - Column (2): 0.506*** (10.40) - Column (3): 0.438*** (7.16) - Column (4): 0.448*** (10.23) - Column (5): 0.593*** (12.66) - Column (6): 0.525*** (9.71) - CO2 taxes: - 0.0105 (0.47) and -0.016 (0.55) reported (insignificant) - Trading schemes: - 0.0333** (2.03) and 0.0320*** (2.79) - Feed-in tariffs: - 0.0278*** (3.10) and 0.0207* (1.67) - Emission limits: - 0.0511** (2.29) and 0.0388* (1.65) - R&D subsidies: - 0.0693*** (5.41) and 0.0616*** (3.16) - Number of observations: 788, 785, 788, 788, 788, 785 (by column). - Statistical significance legend: *** p<0.01, ** p<0.05, * p<0.1. - Main interpretation: - Both non-market policies (emission limits, R&D subsidies) and market policies (trading schemes, feed-in tariffs) made a statistically significant contribution to clean innovation. - Carbon taxes show an insignificant effect, plausibly due to limited use: only slightly more than 10 percent of countries used carbon taxes in 2015, while other policy tools were used by 60-100 percent of countries in the sample. ### Electricity sector: clean, gray, and dirty innovation (Annex Table 3.1.3) - Focus: electricity-related patent families classified as Clean, Gray (environmental improvements to dirty tech), Dirty, and Total. - Key coefficient estimates (selected): - EPS t-1: - Clean (column 1): 0.0688*** (0.02) - Gray (column 2): 0.151*** (0.04) - Dirty (column 3): -0.0405** (0.02) - Total (column 4): 0.175*** (0.04) - Log all tech t-1: - Clean: 0.0522 (0.10) - Gray: -0.338*** (0.10) - Dirty: 0.0543** (0.02) - Total: 0.450*** (0.07) - Log knowledge stock specific to the type of electricity t-1: - Clean: -0.0255 (0.07) - Gray: 0.531*** (0.07) - Dirty: 0.0391 (0.30) - Log knowledge stock for all types of electricity t-1: - Clean: -0.00121 (0.17) - Gray: 0.132 (0.21) - Dirty: -0.103 (0.30) - Total: 0.581*** (0.10) - Log oil and gas reserves t-1: - Clean: -0.129** (0.05) - Gray: -0.0843 (0.06) - Dirty: 0.0398 (0.02) - Number of observations by specification: 738, 743, 743, 781. - Statistical significance legend: *** p<0.01, ** p<0.05, * p<0.1. - Main interpretation: - EPS tightening increases the relative share of clean innovation and increases gray innovation even more. - Dirty innovation declined (negative coefficient), and overall electricity innovation (Total) increased; the decline in dirty innovation was more than offset by increases in clean and gray innovation. ### Renewable electricity share and drivers (models summarized from Annex Table 3.2.1) - Dependent variable of main analysis: annual change in the share of non-hydro renewable energy in electricity generation (∆y_i,t), using IEA (2019) data for 1990-2017. - EPS index used: OECD Environmental Policy Stringency (scale 0 to 6) covering 32 OECD and emerging market countries between 1990-2015 and including instruments such as taxes on NOx/SO2/PM, CO2 and SO2 trading schemes, renewable/green certificates, white certificates, feed-in tariffs, emission limits, and government R&D for renewable energy. - Main findings (Model 1): - Average tightening of market-based environmental policies between 1990 and 2010 can explain a 0.38 percentage point increase in the share of renewable electricity generation per year. - Contextualization of 0.38 percentage points: - Equivalent to 29 percent of the average model-implied increase in the share of renewable electricity in 2014. - Equivalent to 55 percent of the actual increase of 0.69 percentage points in the sample of 32 countries in 2014. - Model 2 (disaggregating market-based policies: feed-in tariffs, taxes, trading schemes): - Feed-in tariffs and trading schemes are statistically and quantitatively significant. - A one standard deviation tightening of: - Feed-in tariffs increases renewable share by 0.118 percentage points per year. - Trading schemes increases renewable share by 0.183 percentage points per year. - Illustrative case: Germany between 1997 and 2007 scored 4 or higher on OECD feed-in tariff variable; based on Model 2, such a policy for a decade would add 2.5 percent to its share of renewable electricity, with a cumulative indirect effect through higher share level adding another 7.5 percent over the same decade. - Model 3 (separating trading schemes into CO2 trading, green certificates, white certificates): - Green certificates are the only trading-scheme subtype with statistically significant evidence. - A one standard deviation change in the green certificates variable (controlling for other policies) increases the share of renewable electricity generation by 0.116 percentage points per year. - CO2 trading schemes show no significant effect, attributed to limited sample variation and relatively weak average policy strength compared to green certificates and feed-in tariffs. - Robustness (Models 4–6): - Coefficients on environmental policy variables are broadly insensitive to additional controls. - Statistically significant negative effects found for income and the share of nuclear power in electricity generation. - Electricity market deregulation: - The average de-regulation of electricity markets in OECD countries between 1990 and 2010 supported an annual increase of 0.38 percentage point in the share of renewable electricity generation. - One standard deviation increase in deregulation corresponds to a 0.223 percentage point increase in renewable share per year. ### Electricity mix and electricity generation per capita (Annex Table 3.2.2) - Main pattern: - Policy indicators such as feed-in tariffs and CO2 schemes have a positive relationship with the share of solar, wind and biomass in electricity generation. - These policies do not appear to have a discernible impact on total electricity generation per capita. - EPS variable associations with fossil fuel shares: - Relationships with annual change of coal share tend to be negative (though not always statistically significant at conventional levels). - Effects on the share of natural gas are ambiguous; natural gas can complement intermittent renewables due to dispatchability. - Interpretation caveat: - Results are associations rather than strict causal estimates; instrumental variable evidence in the literature provides mixed guidance, with some studies suggesting OLS may underestimate the effect of feed-in tariffs. ### Synthesis of evidence - Environmental policy tightening has a statistically and economically significant effect on clean energy innovation. - In the electricity sector, policy tightening: - Shifts innovation away from dirty toward clean and gray technologies. - Yields a net positive effect on electricity innovation overall. - Market-based support policies (feed-in tariffs, green certificates) and electricity market deregulation are particularly effective in increasing the renewable electricity share. - Carbon taxes show an insignificant effect on clean innovation in the sample, likely reflecting limited adoption. *Source: IMF staff calculations.* ### Annex Table 3.2.1. Main Results (1990–2014) ### Annex Table 3.2.1. Main Results (1990–2014) ### Key regression results (electricity share of renewables) - Dependent variable: Δ(electricity share of renewables). - Solar wind and biomass share MA t-1 coefficients: - Column (1): 0.0852** (standard error 0.0304) - Column (2): 0.0791* (0.0361) - Column (3): 0.0826* (0.0365) - Column (4): 0.0808* (0.0361) - Column (5): 0.0741+ (0.0410) - Column (6): 0.0799+ (0.0419) - Policy variables (selected coefficients and reported standard errors): - Market EPS: 0.231* (0.0974) and 0.249* (0.101) - Non-market EPS: -0.124 (0.106), -0.147 (0.113), -0.146 (0.113), -0.109 (0.122), -0.140 (0.131), -0.138 (0.130) - Market EPS taxes: 0.0842 (0.151), 0.0860 (0.147), 0.132 (0.163), 0.133 (0.155) - Market EPS feed-in tariff: 0.0621* (0.0287), 0.0625* (0.0290), 0.0658* (0.0305), 0.0666* (0.0307) - Market EPS trading: 0.166** (0.0543), 0.174** (0.0622) - Market EPS trading green certificates: 0.0918* (0.0373), 0.108** (0.0322) - Market EPS trading CO2: 0.0392 (0.0360), 0.0298 (0.0388) - Market EPS trading white certificates: 0.0561 (0.102), 0.0491 (0.101) - Log electricity PMR t-1: - Column (1): -0.440** (0.159) - Column (2): -0.441** (0.149) - Column (3): -0.445** (0.141) - Column (4): -0.563** (0.197) - Column (5): -0.557** (0.184) - Column (6): -0.568** (0.174) - Controls (selected reporting as in source; standard errors for controls not reported): - Log GDP per capita t-1: 0.0607, 0.0320, 0.113, -0.951+, -0.996*, -0.898+ - Short-term interest rate t-1: 0.000731, 0.00247, 0.00176 - Log crude oil price t-1: -0.511, -0.504, -0.542 - Proven oil reserves per capita t-1: -72.98+, -41.57, -47.24 - Proven natural gas reserves per capita t-1: 847.8, 719.7, 890.8 - Hydropower share t-1: -0.00900, -0.0104, -0.0122 - Nuclear share t-1: -0.00909, -0.0105, -0.00976 - Sample and fit: - Number of observations: 652, 652, 652, 558, 558, 558 - Number of countries: 32, 32, 32, 28, 28, 28 - R2: 0.28 across columns - Notes: - Robust standard errors clustered at the country level in parentheses, not reported for controls. - Variables are in logarithmic scale. Constant included, but not reported. - EPS = Environmental policy stringency; MA = moving average; PMR = product market regulation. - Significance codes: *** p<0.001, ** p<0.01, * p<0.05, + p<0.1. ### Annex Table 3.2.2. Electricity Mix (1990–2014) — selected coefficients - Dependent variables covered: Δ(electricity share of renewables), Δ(electricity share of coal), Δ(electricity share of natural gas), Δ(electricity generation per capita). - Solar wind and biomass share MA t-1: - 0.0826* (0.0365) and 0.0799+ (0.0419) for renewables specifications. - Coal electricity share MA t-1: - -0.0930*** (0.0225) and -0.0821** (0.0293) - Natural gas electricity share MA t-1: - -0.0878*** (0.0172) and -0.0972*** (0.0233) - Electricity generation per capita MA t-1: - -0.281+ (0.164) and -0.319+ (0.166) - Selected policy variable coefficients (with standard errors): - Market EPS taxes: 0.0860 (0.147), 0.133 (0.155), -0.0885 (0.194), -0.0771 (0.260), -0.298 (0.226), -0.237 (0.284), 0.04070 (0.107), 0.0358 (0.105) - Market EPS feed-in tariff: 0.0625* (0.0290), 0.0666* (0.0307), -0.0574 (0.0764), -0.0532 (0.0821), 0.1060 (0.0940), 0.130 (0.108), -0.00233 (0.0139), -0.00160 (0.0138) - Market EPS trading green certificates: 0.0918* (0.0373), 0.108** (0.0322), 0.00752 (0.0915), -0.05000 (0.0838), 0.00692 (0.119), -0.0101 (0.135), -0.00158 (0.0233), -0.00843 (0.0228) - Market EPS trading CO2: 0.0392 (0.0360), 0.0298 (0.0388), -0.153 (0.0989), -0.0391 (0.0911), -0.0560 (0.0684), -0.0214 (0.0698), -0.0297 (0.0237), -0.0411 (0.0264) - Market EPS trading white certificates: 0.0561 (0.102), 0.0491 (0.101), 0.238 (0.221), 0.233 (0.226), -0.376 (0.274), -0.414 (0.328), -0.0586 (0.0367), -0.0607 (0.0430) - Non-market EPS: -0.146 (0.113), -0.138 (0.130), -0.330 (0.195), -0.350 (0.234), 0.173 (0.213), 0.321 (0.221), 0.03350 (0.0672), 0.0671 (0.0854) - Log electricity PMR t-1 in these specifications: - -0.445** (0.141), -0.568** (0.174) for renewables; 0.389 (0.359), 0.503 (0.530) for coal; 0.617 (0.633), 0.117 (0.735) for natural gas; 0.01880 (0.0974), 0.0114 (0.122) for generation per capita. - Controls and other reported statistics appear in the table as provided (Log GDP per capita, short-term interest rate, log crude oil price, proven reserves, hydropower share, nuclear share). - Number of observations and countries by dependent variable: - Observations: 652, 558, 652, 558, 652, 558, 652, 558 (matching columns 1–8) - Number of countries: 32, 28, 32, 28, 32, 28, 32, 28 - R2 by column: 0.28, 0.28, 0.10, 0.17, 0.20, 0.25, 0.10, 0.11 - Notes: - Robust standard errors clustered at the country level in parentheses, not reported for controls. - Variables are in logarithmic scale. Constant included, but not reported. - EPS = Environmental policy stringency; MA = moving average; PMR = product market regulation. - Significance codes: *** p<0.001, ** p<0.01, * p<0.05, + p<0.1. --- ### Impact of environmental policies on employment — methodology and samples ### Estimating approach - Basic specification: augmented labor demand equation (firm-level employment n i,j,t), controlling for standard determinants (including lags of employment), firm, country, and year fixed effects. - Interaction of interest: EPS j,t x d_i, where d_i = 1 if CO2 emissions = "high", 0 otherwise. - Specification 1: interaction uses firm CO2 emissions dummy (requires firms reporting at least 3 instances of CO2 emissions for inclusion). - Specification 2: interaction uses sector dummy to proxy CO2 emission intensity (expands sample coverage). - Estimation methodology: panel GMM following Van Reenen (1997). - Controls X i,j,t include: - Firm-level average annual employee wages - Real capital stock (sum of building and machinery capital stock deflated by building and machinery price indices from Penn World Tables) - Rental rate of capital (proxied by price of capital services from Penn World Tables) - Output gap - Variable treatment: - All variables are expressed in logs, except the EPS indicator which enters in levels. - Firm-level employment regressed on its own lags and the vector of controls. ### Samples and data sources - Firm-level data source: Worldscope database. - Sample sizes: - Specification 1 (interaction with high/low CO2): 670 firms, from 30 countries over 2000–15. - Specification 2 (interaction with sector dummies): 5,305 firms, covering 31 countries over 2000–15. - EPS variables: from OECD, measured as indices on a scale from 0 to 6, covering 32 OECD and emerging market countries between 1990–2015 for various policy indicators. - Rental rate of capital: Penn World Tables (price of capital services as proxy). - Output gap controls: IMF WEO database. - CO2 emissions measure: thousand tons; firms coded as high-emission if emissions-to-employees ratio exceeds the median for the country-year in any year reported (time-invariant in that framework). ### Summary statistics (selected, as reported) - Sample 1 (interaction with high/low CO2) — Variable: Mean, Std. Dev., Min, Max - Aggregate EPS: 2.5, 0.9, 0.4, 4.1 - Market EPS: 1.9, 1.0, 0.04, 4.0 - EPS: CO2 tax: 0.1, 0.9, 0.06, 6.0 - Non-market EPS: 3.1, 1.2, 0.6, 5.5 - Log employees: 9.6, 1.5, 2.8, 13.4 - Log capital stock: 14.5, 1.8, 8.3, 19.6 - Log wage: 3.9, 0.8, -3.4, 11.5 - Log r: 0.0, 0.1, -0.8, 0.6 - Output gap: 0.02, 2.2, -15.4, 8.9 - Log CO2 emissions: 8.9, 10.0, -1.3, 12.3 - Sample 2 (interaction with sector dummies) — selected: - Aggregate EPS: 2.2, 0.9, 0.4, 4.1 - Market EPS: 1.7, 0.9, 0.04, 4.0 - EPS: CO2 tax: 0.1, 0.8, 0.06, 6.0 - Non-market EPS: 2.7, 1.3, 0.6, 5.5 - Log employees: 7.8, 1.9, 0.0, 13.4 - Log capital stock: 12.2, 2.2, 3.4, 19.5 - Log wage: 3.2, 1.2, -3.8, 11.5 - Log r: 0.0, 0.1, -0.8, 0.6 - Output gap: -0.2, 2.0, -15.4, 8.9 --- ### Main findings on employment effects of EPS - Specification 1 (interaction with firm CO2 emissions): - Column 1: baseline labor demand equation with 2 lags of employment; standard determinants are highly statistically significant with expected signs. - Column 2: aggregate EPS x CO2-emissions dummy interaction is significant: - High-emission firms experience negative employment effects when EPS tightens. - Low-emission firms experience an increase in employment (effect not significant for low-emission firms). - Estimated semi-elasticities (as reported): - A 1-standard-deviation tightening in the EPS indicator would lower employment in high-carbon firms by 5 percent. - The same tightening would raise employment in low-carbon firms by 2.6 percent. - Column 3 (market EPS): coefficients suggest decline in employment in high-emission firms and increase in low-emission firms (qualitatively similar). - Column 4 (carbon taxation): qualitatively similar results, though not significant. - Column 5 (non-market EPS): both non-market EPS and its interaction with emission intensity are significant: - A 1-standard-deviation tightening in the non-market EPS would lower employment in high-emission firms by 5 percent. - The same tightening would raise employment in low-emission firms by 4.4 percent. - Cyclical sensitivity (columns 6–9): - Interaction of output gap with EPS included. - Output gap term has expected sign; only significant in regression considering market EPS. - Market EPS can have a positive employment effect when the output gap is negative (severe contractionary conditions). - Average marginal effects of market EPS on employment: modestly positive under severe contractionary conditions; turn negative during normal/expansionary periods. - Possible explanation: tighter EPS may raise inflation and, under severe contraction, lower real rate of interest (e.g., if policy rates at zero-lower bound), stimulating demand. ### Robustness and selection considerations - Results robust to setting high/low emission intensity threshold at different percentile (e.g., 75th percentile) — signs broadly robust though significance may vary. - Results broadly robust to excluding firms with fiscal year ending before December of the given year (signs retained; statistical significance weaker due to reduced sample). - Selection concerns on CO2 reporting discussed: - Firms from fossil fuel industries (26%), transport industries (24%), and utilities (34%) have higher proportions reporting at least one year of CO2 emissions relative to sample average (21%). - Reporting rates in China, India, Indonesia are much lower (3%, 4%, 5% respectively) compared to sample average of 32%; frequency of "high-emission" firms among reporters is similar to sample average. - These patterns suggest selection bias is unlikely to invalidate findings. --- *Source: IMF staff calculations. Annex Tables and accompanying text as provided.* ### Annex Table 3.3.4 shows results from replacing the CO2 emission dummy in Specification ### Annex Table 3.3.4 shows results from replacing the CO2 emission dummy in Specification ### Specification and sectoral definitions - Six sectors included in the baseline as proxies for emission intensity: fossil fuel industries; high-emission manufacturing industries (food, metals and minerals, chemicals, paper and packaging); services; construction; transport; other (low-emission) manufacturing industries. - The construction sector includes residential, commercial, and industrial construction and is noted as using high-emission inputs such as cement. - Regressions include panel and year fixed effects. Wages, capital, and rental rate are GMM-instrumented with lags. - Sample sizes (Annex Table 3.3.4): Observations 28,122 (column 1) and up to 26,072 (columns 9–10); Number of firms 5,579 (column 1) and up to 5,384 (columns 9–10). ### Key regression findings and coefficients (selected) - Aggregate and EPS-type indicators used: Aggregate EPS, Market EPS, Carbon taxes, Non-market EPS (per table notes). - Lag and control coefficients (from Annex Table 3.3.2 and 3.3.4 excerpts): - Log N t-1: 0.573*** (Annex Table 3.3.2, col 1); 0.543*** (Annex Table 3.3.4, col 1). - Log N t-2: −0.142*** (Annex Table 3.3.2, col 1); 0.0189 (Annex Table 3.3.4, col 1). - Log capital stock: 0.276** (Annex Table 3.3.2, col 1); 0.186*** (Annex Table 3.3.4, col 1). - Log wages: −0.260** (Annex Table 3.3.2, col 1); −0.218*** (Annex Table 3.3.4, col 1). - Log r: 0.308** (Annex Table 3.3.2, col 1); 0.241*** (Annex Table 3.3.4, col 1). - Sector × EPS interaction coefficients (Annex Table 3.3.4, selected): - Construction × EPS: −0.0836*** (col 1); −0.0693** (col 2); −0.0717*** (col 5); significance preserved across columns. - High CO2 industries × EPS: −0.0228 (col 1); −0.0257 (col 2); −0.0346* (one column); several columns show significance at * or ** levels. - Services × EPS: 0.0174 (col 1); 0.0225* (col 2). - Fossil fuels × EPS: mixed signs across specifications (see interpretation below). - Utilities × EPS: −0.0226 (one column); −0.0266 (another column). - Joint significance of the sector interactions: reported joint p-values such as 0.05, 0.06, 0.65, 0.05, 0.05, 0.07, 0.03, 0.07, 0.10 across columns (Annex Table 3.3.4). ### Interpretation of sectoral and EPS-type heterogeneity - Aggregate EPS (column 2, Annex Table 3.3.4): each sector interaction has expected sign; construction interaction individually significant and negative. - The negative construction effect may reflect higher costs of high-emission inputs (e.g., cement) and reduced activity in high-emission sectors. - Market EPS (column 3): services interaction positive and statistically significant at the 10 percent level; fossil fuel industries interaction positive in aggregate market EPS (note: subcomponents differ—tax policies versus trading schemes). - Footnote interpretation: interaction with fossil fuels is negative for tax policies but positive for trading schemes. - Non-market EPS (column 5): pattern similar to market EPS overall, but interaction with fossil fuel sector is negative (column 5). - Including output gap (columns 6 and 7) as control: - Output gap coefficient positive and significant (example: 0.00806*** in Annex Table 3.3.4). - Upon introducing the output gap, the coefficient on high-emission manufacturing becomes significant (t-statistic rises from 1.2 to 1.9 in the relevant comparison). - Utility category (robustness exercise, columns 9–10): - Broader utility category (fossil fuel utilities and multiline utilities) shows negative coefficient on utilities interaction as expected. - Interactions remain jointly significant at the 10 percent level in these regressions. ### Net employment impact (aggregate estimates) - Based on preferred specifications that include controls for the output gap (estimates in columns 6 and 9): - Net loss of between 500-600 thousand jobs in the sample for aggregate EPS tightening by 1 standard deviation (about 1 percent of total employment in the sample). - For tightening market EPS by 1 standard deviation: - Small net job increase between 13-34 thousand jobs (based on estimates in columns 7 and 10; Annex Figure 3.3.2 referenced). ### Medium-term dynamics and business-cycle dependence - Medium-term regressions (excluding interactions with CO2 emissions) show that: - Short-term effects of aggregate EPS, market EPS, and non-market EPS tend to reverse and fade away over the medium term (Annex Figure 3.3.3 referenced). - Short-term effects summary: - Tightening of environmental policies associated with reallocation of jobs: employment rises in low-CO2 emissions firms and falls in high-CO2 emissions firms. - Effects suggest labor reallocation from high carbon-intensive sectors to low carbon-intensive ones when emissions are proxied by sector. - Overall impact depends on the mix of policies: past policy changes suggest short-term effects are negative for overall EPS (likely driven by negative effects of non-market policies); greater reliance on market-based policies may yield positive overall effects. - Short-term effects are described as "quite modest" whether net positive or net negative, and tend to fade over the medium term. - Business-cycle dependence: - Impact of tightening market-based policies depends on the state of the business cycle: under highly contractionary conditions the impact may be positive (other things equal), whereas under normal or expansionary periods the effect is negative. ### Modeling context (G‑Cubed model inputs and assumptions) - Model version for this project: GGG20v154 (10 regions and 20 sectors). - Key model changes for this project: - Database updated to GTAP10 and latest IMF, World Bank, OECD, UN, and US EIA data. - Gas extraction and gas utilities merged into one gas sector. - New construction sector added. - Capacity for implementing government infrastructure investment following Calderon and others (2015); green infrastructure projects incorporated. - Regions covered: 10 region codes including AUS, CHN, EUW, IND, JPN, OPC, OEC, ROW, RUS, USA. - Sectors (Annex Table 3.4.2): 1 Electricity delivery; 2 Gas extraction and utilities; 3 Petroleum refining; 4 Coal mining; 5 Crude oil extraction; 6 Construction; 7 Other mining; 8 Agriculture and forestry; 9 Durable goods; 10 Nondurable goods; 11 Transportation; 12 Services; 13 Coal generation; 14 Natural gas generation; 15 Petroleum generation; 16 Nuclear generation; 17 Wind generation; 18 Solar generation; 19 Hydroelectric generation; 20 Other generation. - Model features highlighted: - Full accounting for stocks and flows of physical and financial assets; intertemporal budget constraints. - Central banks set short-term nominal interest rates via Henderson-McKibbin-Taylor monetary rules. - Nominal wage stickiness and sector-specific hiring rules; unemployment arises in short run. - Rigidities: wage stickiness, limited foresight, sector-specific investment adjustment costs, monetary and fiscal rules. - Heterogeneous households and firms: mixture of forward-looking and rule-of-thumb agents. - Fiscal rule in this version: exogenous budget deficit with lump-sum taxes on households adjusted to ensure fiscal sustainability. ### Net Zero by 2050 and carbon removal assumptions - Carbon removal potentials (from Fuss (2018) summary in the text): - Authors take the average of the range for each technology and sum them up: 13.8Gt CO2. - Conservative assumption: 75% of 13.8 Gt achievable by 2050 → about 10Gt CO2 per year. - 10Gt CO2 per year corresponds to about 20% of global CO2 emissions in the baseline in 2050. - Regional emission reduction assumption: - All regions in the model reduce their emissions by 80% by 2050 relative to 2018 except OPC, which remains at the same level of 2018 by 2050. - Carbon tax growth-rate assumptions in main results: - Main assumption: constant growth rate of 7 percent for carbon taxes over 2019-2050, then solve tax rates to achieve emissions targets by 2050 after accounting for other policy layers. - Comparison scenario: growth rate of 5 percent for carbon taxes to achieve the same targets by 2050. *Source: IMF staff calculations (Annex Table 3.3.4 and surrounding text).* ### Annex Table 3.4.3. Global Carbon Removal Potentials in 2050 ### Annex Table 3.4.3. Global Carbon Removal Potentials in 2050 ### Carbon removal technology potentials (Gigaton CO2) - Afforestation and reforestation: 0.5–3.6 - BECCS: 0.5–5 - Biochar: 0.5–2 - Enhanced weathering: 2–4 - DACCS: 0.5–5 - Soil carbon sequestration: Up to 5 ### Policy package and scenario design - Solar and wind output are subsidized at a rate of 80% for all regions since 2019. - Economy-wide productivity improvements: - Driven by avoided damages from climate change due to all other policies in the package. - Imposed equally on all sectors except electricity generation. - Calculated using an extension of the Hassler and others (2020) integrated assessment model (see Annex 3.5). - Carbon tax revenue transfers: - 25% of carbon tax revenues are transferred to households as compensatory transfers. - The fraction 25% is based on analysis in the section How to build inclusion (and Annex 3.7). - Policy package composition: carbon taxes, green investment, green subsidy, avoided damages, and the carbon tax revenue transfer. - Participation scenarios: - Largest 5 countries (economic union) scenario: USA, EUW, CHN, IND, JPN participate. Policy shocks for these five use the aggregate scenario shocks without re-solving carbon taxes. - Advanced economies scenario: USA, EUW, JPN, AUS, OEC participate. Carbon taxes are not re-solved in this scenario. - Implementation timing: - Given 2018 is the last observed year, policy shocks are applied from 2019 onward. - Simulation is presented as starting in 2021 for presentational purposes. ### Energy sector model structure - Energy is produced by a CES technology using N fuels; imported conventional oil is indexed first. - Energy production in region j at time t: - E_{j,t} = [ sum_i λ_{i,j} e_{i,j,t}^{(ρ-1)/ρ} ]^{ρ/(ρ-1)} (equation 3.5.1 structure preserved) - where g_{i,j,t} (usage of fuel i in region j), λ_{i,j} (production weight), and ρ (intra-fuel elasticity of substitution) are key elements. - Fuel-specific technologies: - Initial technology level: x̅_{i,j,t} (productivity: units of final good spent to produce one unit of input i). - Firms can improve technology to x_{i,j,t} at cost r_{i,t}(x_{i,j,t}/x̅_{i,j,t}) = ε_{i,j}(1−χ_{i,j,t})^{η−1}(x_{i,j,t}/x̅_{i,j,t})^{η/(η−1)}, with η>1 and ε_{i,j} region-specific. - Carbon intensity technology: - Initial carbon intensity g̅_{i,j,t}, improvable via research at cost r_{i,t}(g_{i,j,t}/g̅_{i,j,t}) = θ_{i,j}(1−χ_{i,j,t})^{η−1}(g_{i,j,t}/g̅_{i,j,t})^{η/(η−1)}, with θ_{i,j} region-specific. - Cost of production net of research includes carbon tax τ_{i,j,t} and research costs; energy production exhibits a downward-sloping average cost curve, implying a natural monopoly regulated so the monopoly supplier makes zero profits (price cap at average cost). ### R&D, endogenous technical change, and market-size effects - Firms conduct fuel-specific R&D that lowers input cost or carbon intensity; R&D spending increases when the market for a fuel expands. - Endogenous technical change has two forms: input-saving and emissions-reducing. - Cost of research is convex (η>1); returns to scale in R&D improve as η declines. - This framework allows: - An R&D composition effect: composition of energy bundle responds to relationship between elasticity of substitution and returns to scale in R&D. - An aggregate R&D effect: total research increases with total energy demand (aggregate market size effect), amplifying policy impact via aggregate energy price channels. - The model extends Hassler and others (2020) by: - More general R&D cost function allowing returns to scale governed by η. - Allowing for aggregate market size effect (total research increases with total energy demand). - Allowing non-unit price elasticity in final energy demand and extending fuel sources to match IEA data. ### Macro structure and climate feedbacks - Aggregate production: - Y_{j,t} = φ_j(T_{t−1}) [ (1−ν_j)(A_{j,t} L_{j,t})^{(1−α_j)/ (1−α_j)} K_{j,t}^{α_j/(1−α_j)} + ν_j E_{j,t}^{(1−σ)/ (1−σ)} ]^{(1/(1−σ))} - where ν_j is energy share, α_j capital share, A_{j,t} labor productivity, L_{j,t} labor, K_{j,t} capital, and σ is the elasticity of substitution of energy in final production. - Climate damages: - Region-specific productivity scaling φ_j(T_t) is quadratic in global temperature: φ_j(T) = 1 − φ_{j,0} + φ_{j,1} T + φ_{j,2} T^2. - Atmospheric and ocean temperatures follow a linear energy budget model derived from RICE/DICE; global emissions drive temperatures, which reduce regional productivities. - Emissions accounting: - Region j emissions: m_{j,t} = sum_i g_{i,j,t} e_{i,j,t} - Global emissions: M_t = sum_j m_{j,t} - Stock of global CO2 S_t aggregates past emissions with decay: S_t = sum_{τ=0} (1−d_τ) M_{t−τ}, where 1−d_τ = ψ_0 + (1−ψ_0) ψ^{τ} (1−ψ)^{τ} formulation implies a permanent fraction ψ_0 remains in atmosphere. - International features: - Conventional oil produced at zero cost by an oil-producing region managing fixed reserves (monopoly rents); unconventional (fracked) oil produced domestically. - Oil price equilibrates global oil demand and OPEC supply; oil price is main international price linkage. - International diffusion of ideas via constant catch-up: x̅_{i,j,t+1} = ω_j x̅_{i,j,t} + (1−ω_j) max_k x̅_{i,k,t}, similarly for g̅_{i,j,t+1} with min. ### Calibration and key parameter values - Model regions and fuels: - Ten regions (including OPEC producing only oil for international trade). - Seven fuel types: international oil; domestic oil; natural gas; coal; hydroelectric; nuclear; renewables. - Selected calibrated parameter values (Annex Table 3.5.1 highlights): - ρ (Intra-fuel CES parameter): 0.67 (Common) — Papageorgiou et al. (2013) - η (R&D returns, common): 10 — Fried (2018) dynamics - σ (Aggregate CES parameter, common): -3 — Consistent with Annex 3.6 - Calibration targets and choices: - CES fuel weights λ_{i,j}: Regional IEA fuel shares - ε_{i,j}: Regional R&D cost shift — target regional fuel prices - θ_{i,j}: Carbon intensity R&D cost shift — target IEA carbon intensities - Energy output share ν_j: Regional IEA energy shares - Capital share α_j: Regional WEO capital shares - Intra-fuel elasticity of substitution set to 3, consistent with Papageorgiou and others (2013). - Elasticity of substitution of energy and the capital-labor bundle set to 0.25 (consistent with Annex 3.6). - Returns to R&D (η = 10): implied by Fried (2018) estimate that innovation response reduces carbon tax required by 20 percent to meet a given reduction in emissions in 20 years; after accounting for model differences, implies returns to scale in R&D of around 0.11 (η of 10). - Other calibration details: - Initial capital and labor productivity set to match average regional weights in global GDP and emissions during 2010-2019. - Labor force growth from ILO forecasts until 2030, converging smoothly to annual growth rate of 0.2% by 2070 (UN population projections). - Long-run labor productivity growth assumed 1.3% per year, with catch-up growth in India, China, and RoW in short term. - Energy share parameter ν_j assumed to decrease at around 0.7 percent annually to capture energy efficiency trends. ### Model solution and state variables - State variables include region-specific labor productivity, capital, fuel-specific technologies x̅_{i,j,t} and g̅_{i,j,t}, global stock of emissions, and their permanent share. - With nine productive regions and six improvable fuels this yields 128 state variables (strong simplifying assumptions justified). - Comparison and fit: - Annex Figure 3.5.1 compares calibrated model results to data averaged across 2010-2019; model matches level and distribution of output, emissions, energy usage, and prices across regions. - Climate module calibration: - Uses standard parameter values from Nordhaus (2010), Golosov and others (2014), and Hassler and others (2020). - Baseline climate damages from Nordhaus (2010) with alternative specification using Burke, Hsiang, and Miguel (2015). ### Key statistics and sectoral context (from supplied text) - Emissions from electricity generation and heating: - Amounted to roughly 40 percent of total global carbon dioxide (CO2) emissions in 2018. - Electricity-sector emissions characteristics: - Coal emits about one kg of CO2 per kWh and causes roughly 30 percent of global CO2 emissions by itself. - Natural gas emits about 400 grams of CO2 per kWh; it is less polluting than coal but still too high for a low-carbon economy except to provide flexibility for renewables. - IEA assessment cited: - Growth of low-carbon electricity sources up to 2040 under stated policies is estimated to be half as large as what is needed to meet the UN’s Sustainable Development Goals and to cut emissions in line with the objectives of the Paris Agreement. *Source: annexch3 - Annex Table 3.4.3. Global Carbon Removal Potentials in 2050 (PDF chapter content provided).* ### Annex Figure 3.6.1 shows the electricity ### Annex Figure 3.6.1 — electricity mix and implications for an immediate transition ### Key empirical findings on electricity mixes and emissions - In all regions, the share of coal and natural gas is described as "unsustainably high" and, absent a dramatic rebalancing, "electricity generation will be a key driver of irreversible climate damage." - European Union: - Annual per-capita CO2 emissions from electricity and heating: 2,176 kg (as of 2017). - Coal still has a share of over 20 percent and is in many places backed by subsidies that delay the required transition. - United States: - Annual per-capita CO2 emissions from electricity and heating: 5,592 kg (2017). - Per-capita emissions are about 2.5 times higher than in the EU, reflecting a per-capita electricity consumption about twice as large combined with a more polluting electricity mix in which coal and gas together make up over 60 percent of the generation. - China: - Electricity consumption per capita is about 2/3 as large as in the EU. - Coal share is "almost 70 percent", elevating annual per-capita emissions to 3,312 kg (2017). ### Technology readiness, costs, and feasibility - Feasibility: - Feasibility has improved dramatically over the last decade due to rapid declines in prices for key renewable technologies that are expected to continue, making renewables economically viable to replace coal-fired electricity at scale. - The improved competitiveness of renewables makes mitigation policies more effective now than a decade ago. - Technology-specific notes: - Hydraulic power: geographic requirements limit site availability and cause broader environmental damages. - Biomass: can generate electricity but production could compete with other uses of land. - Nuclear power: carbon-neutral and scalable but generally low popularity due to nuclear disasters, waste management concerns, and political considerations; actuarial argument presented that nuclear's low-probability, geographically limited damages are likely dwarfed by certain, global, irreversible damages from climate change. - Renewables (wind and solar PV): identified as the most promising and politically acceptable carbon-neutral technology; key drawback is intermittency. - Carbon capture and storage (CSS): has potential to reduce emissions from coal but deployment is prevented by high costs not predicted to fall substantially in the near term. - Policy implication: - Analysis focuses on renewables as the technology to bring about an electricity transition because it is in principle scalable and, relative to nuclear power, politically less controversial. ### Intermittency, storage, and backup requirements - Generation duration curves and load factors: - Generation duration curves show electricity output from wind ordered by wind regimes; data normalized by average electricity output. - Load factor (ratio between average generation and peak generation): - Onshore wind: 35 percent. - Offshore wind: 45 percent. - Solar PV: about 25 percent. - Storage capacity and current limitations: - Global hydro-pump power capacity (IHA estimate): 158 GW. - In the United States, the European Union and China, this hydro-pump capacity would represent about 30 minutes of power consumption. - Managing solar intermittency would require about 18 hours of electricity storage. - Managing wind intermittency would require about 72 hours of electricity storage. - Chemical storage (batteries or hydrogen) remain too expensive for large deployment; other technologies based on heat or gravity are not mature yet. - Interim solution: - With electricity storage not yet economically viable at large scale, intermittency of renewables must be compensated by other flexible electricity sources in the grid (natural gas and hydro are most commonly used). - Flexibility retrofits could potentially allow coal power plants to serve as a backup, but that would entail a slower decline in emissions. ### Modelling intermittency and backup in CarMMa - Intermittency module assumptions: - Under the assumption that demand flexibility and electricity storage are not yet sufficiently mature, any intermittent generation capacity is required to be paired with a dispatchable backup capacity that is idle most of the time but can cover power shortfalls of intermittent technologies. - The backup covers, at any point in time, the difference between intermittent power generation and the desired output. - The framework allows for "overcapacity," i.e., installation of renewable capacity that at peak output produces electricity above demand that must be curtailed. - Functional forms and parameters (preserved as in source): - Generation duration curves can be well approximated by the power function: - 퐸=푝 ఊ - where 훾 is a parameter measuring the degree of intermittency (the corresponding load factor is given by 1/(1+훾)). - Intermittency parameter examples: - Offshore wind: between 1 and 1.5. - Onshore wind: between 2 and 3. - Solar: different intermittency profile (no production at night) but similar daytime pattern to wind. - Cost notation (preserved): - Variable costs of the backup: 퐶௩. - Fixed costs of renewables: 퐶௙௥. - Fixed costs of the backup: 퐶௙௕. - Optimal backup output (as stated in source): - 퐵= 훾 1+훾 ൬ 퐶 ௩ 퐶 ௙௥ ൰ ିଵ ଵାఊ - Intuition from cost-minimization: - Deploying renewables operating at zero variable costs lowers overall variable costs by substituting for costly backup generation. - Variable cost savings from expanding renewables exceed installation costs up to the point where peak renewable output equals desired output L. - Expanding renewables beyond that point leads to curtailment; curtailment rises at an increasing pace due to the shape of the generation duration curve. - At the cost-minimizing ratio between renewable and backup capacities, curtailment reduces the variable cost savings from additional renewables such that they equal the fixed costs, determining optimal backup output B. ### CarMMa model structure and calibration - Model overview: - The Carbon Mitigation Macro Model (CarMMa) is a Dynamic Structural General Equilibrium (DSGE) model tailored to analyze how governments can trigger an electricity transition and its macroeconomic implications. - CarMMa builds largely on the IMF’s workhorse model GIMF and inherits a detailed description of households, firms, a detailed fiscal sector and monetary policy, as well as a menu of real and nominal rigidities. - CarMMa is currently a closed-economy model. - Electricity sector in CarMMa: - Encompasses four technologies: coal, natural gas, renewables, and nuclear power plus hydro. - Coal and natural gas fuels are mined in two specific mining sectors. - Nuclear and hydropower have negligible carbon emissions and close to zero marginal cost. - Hydro and nuclear capacities are exogenous (limited hydropower sites and political constraints on nuclear); nuclear capacity expansion is subject to a time-to-build constraint. - Renewables are paired with a backup capacity in a cost-efficient manner; most common backups are hydropower and natural gas, with coal (with flexibility retrofits) third according to a merit-order model. - For the US model, natural gas is assumed to be the only backup; for China and the European Union, both natural gas and coal are used as backups (to capture shortage of natural gas in those regions). - Electricity generations compete on a commodity market for electricity and are treated as very close substitutes (equally dispatchable since intermittency from renewables is compensated by backup). - Electricity is used as an intermediate input in production of manufacturing goods and services and also enters the final consumption good. - Calibration and accounts: - Structure of GDP by sector, by expenditure, and by income reproduces national accounts in 2018 and the most recent input-output tables. - The share of different electricity-generation technologies and their emissions (excluding installation and dismantlement emissions) reproduce data from the IEA. ### Selected sector shares (Percent of GDP) — reproduced from source - United States / European Union / China - Electricity: 1.9 3.0 2.3 - Manufacturing: 0.4 0.9 1.2 - Services: 0.5 0.7 0.6 - Consumption: 1.0 1.4 0.5 *Source: annexch3 - Annex Figure 3.6.1 shows the electricity (annexch3.pdf).* ### Annex Table 3.6.1. Electricity Generation and Use ### Annex Table 3.6.1. Electricity Generation and Use ### Transmission of a carbon price into electricity prices - A carbon price is partially absorbed by the mining sector, cushioning the rise in fuel costs for coal- and gas-based electricity producers. - A carbon price reduces fuel demand and thereby lowers coal and gas prices in the short run, preventing a one-for-one increase in electricity producers' fuel costs. - Market competition further dampens the pass-through of higher fuel costs into electricity prices. - The carbon price increases marginal fuel costs for coal (and gas) generation but has no impact on marginal costs of other technologies. - Declines in investment (due to expected permanently reduced profitability) provide room to absorb higher costs. - Rebalancing of the electricity mix toward low-carbon technologies lowers the average carbon intensity, so a given carbon price produces a smaller electricity price increase. - The strength of the rebalancing effect depends on the availability of natural gas as backup; scarcity of gas weakens emissions declines for a given surge in renewables. - The intermittency problem is proxied in the model by using an intermittency parameter 훾 lower than observed. ### Transmission of higher electricity prices into the macroeconomy - Electricity raises labor productivity via operation of machines/buildings; higher electricity prices do not materially change technical coefficients. - Firms have limited substitution possibilities away from electricity; instead they reduce demand for capital and labor. - Given high price elasticity of capital supply and low price elasticity for labor in general equilibrium, higher electricity prices reduce investment and capital accumulation and lower real wages. - Spillovers to non-electricity sectors are determined by: - the elasticity of substitution between electricity and other factors, set to 0.3; - the share of electricity in each sector. - Macroeconomic impact of a carbon price in the electricity sector depends on: - the initial share of coal in the electricity mix; - the portion of the carbon burden absorbed by the mining sector; - the availability of low-carbon backup technologies; - the investment share in the economy. ### Scenario: 50USD carbon price phased in over 10 years (United States, European Union, China) - Policy design: - Carbon price: 50USD, phased in over 10 years. - Carbon tax revenues are returned to households as transfers. - Electricity mix adjustments: - Coal share declines in all regions; in the United States the share falls below 10 percent given abundant natural gas. - Assumption of average 40-year lifetime of a coal power plant slows the transition via gradual depreciation of coal capital stock. - In the European Union and China, coal use alongside gas as backup mitigates coal share decline. - Electricity price impacts after ten years (cumulative increase relative to baseline): - European Union: 10 percent - United-states: 20 percent - China: 30 percent - Macroeconomic effects: - Investment declines due to shrinking coal sectors and limited substitution away from more costly electricity. - Consumption declines after an initial uptick from higher dividend payouts associated with less investment spending. - GDP declines gradually over ten years (relative to baseline), implying average annual growth reductions of: - about 0.1 percentage point in the United-States - about 0.1 percentage point in the European Union - about 0.3 percentage point in China - Larger decline in China explained by: - larger share of coal in the electricity mix amplifying electricity price rise; - high share of investment in the economy amplifying the effect of reduced investment on aggregate demand and output. - The carbon price dampens investment and works toward rebalancing the economy toward a larger consumption share. - Electricity-sector CO2 emission reductions after ten years (relative to baseline): - European Union: about 30 percent - United States: about 35 percent - China: about 38 percent - Absolute declines in electricity-related emissions after ten years: - United States: 745 megatons - European Union: 390 megatons - China: 1919 megatons ### Carbon tax with a macro package (United States and European Union) - Macro package components: - Frontloaded subsidies for investment in renewables, financed by public debt in the first five years. - Accommodative monetary policy in the short run. - Subsidy rates: - United States: initial subsidies amount to 60 percent of investment costs in renewables, declining to 30 percent after five years. - European Union: initial rate 40 percent, declining to 20 percent after five years. - Effects: - Subsidies boost short-term investment and accelerate the electricity transition, lowering average carbon intensity. - Lower carbon intensity dampens the impact of the carbon price on the electricity price and GDP. - The macro package compensates the short-run output decline, mitigates long-run output decline, and amplifies emissions reductions. - The macro package is more effective when paired with the carbon price because expected higher long-run renewable market share increases subsidy impact. ### Carbon tax with macro package plus additional nuclear and gas (China) - Policy mix: 50 USD carbon tax phased in, expansion of nuclear power, and improved availability of natural gas as backup. - Effects: - Additional measures cut output costs by roughly a half and amplify the decline in emissions by about 50 percent (relative to the isolated 50 USD carbon price scenario). - Deployment of additional nuclear capacity immediately increases electricity supply and crowds out coal-based producers, reducing emissions. - Subsidies for natural gas generation support new flexible backup capacity for renewables, splitting backup needs between coal and gas and further reducing coal share. - Additional nuclear and gas measures partially offset the electricity price increase from the carbon price via direct supply increases and by reducing average carbon intensity. ### Policy implications beyond model analysis - Existing coal fleet concern: - Many existing coal power plants are young (60 percent are 20 years or younger), complicating rapid retirement policies. - Continued operation of existing fleet could generate enough emissions to jeopardize sustainable development targets. - The IEA estimates that existing coal power plants represent globally more than $1 trillion unrecovered capital investment. - Model simulations show dramatic declines in the value of coal power plants (Tobin’s Q) under transition scenarios, raising financial loss concerns for owners, including governments. - Options to minimize financial damages: - Retrofitting and repurposing younger, more efficient plants to be compatible with climate targets (e.g., installing CCUS or biomass co-firing). - Repurposed plants could operate at lower utilization to provide flexibility for intermittent renewables. ### Distributional effects and revenue recycling (carbon tax of 50 USD per ton of CO2) - Key channels increasing income inequality: - Low-income households spend a larger share of income on high-energy intensive goods. - Carbon taxes can reduce wages and job opportunities of unskilled low-income workers, who are more likely to work in high-energy intensive sectors. - Unskilled workers' wages fall more than skilled workers' wages; the skill premium increases as carbon tax reduces demand for high-energy intensive goods. - Model framework: - Multi-sector heterogeneous agent model (small open economy) with four goods: high-energy intensive good, low-energy intensive good, dirty energy, clean energy; and two household types: skilled and unskilled. - Two key model features: non-homothetic preferences and varying skilled/unskilled labor intensity across sectors. - Elasticities used in calibration: - Elasticity of substitution between electricity and other factors: 0.3 (stated earlier for spillovers). - Elasticity of substitution between clean and dirty energy: 3. - Elasticity of substitution between energy and the capital-labor composite: 0.25. - Elasticity of substitution between skilled and unskilled labor: 2. - Carbon tax simulated: constant 50 USD per ton of CO2. - Revenue recycling cases analyzed (for distributional impact): - (i) Carbon tax revenue used to finance government spending on low-energy intensive goods. - (ii) Revenue used to finance a universal cash-transfers program. - (iii) Revenue used to finance a targeted cash-transfers program for the bottom two quintiles. - (iv) Revenue used to fund a subsidy to clean energy consumption ("feebates"). - Main distributional results: - Without compensatory measures, carbon taxes increase income inequality measured by the Gini coefficient (Annex Table 3.7.1). - Bottom-income households are impacted more due to higher shares of energy-intensive consumption and larger relative wage/job impacts. - Using revenue for cash transfers: - Universal or targeted transfers raise consumption of unskilled households and can reduce income inequality to levels below the baseline; targeted transfers to bottom two quintiles are more effective. - Feebates: - Subsidizing clean energy consumption mitigates price impacts on bottom-income households. - Production of clean energy is more labor-intensive in unskilled labor than dirty energy; feebates boost unskilled labor demand, mitigating the skill premium increase and reducing income inequality. *Source: annexch3 - Annex Table 3.6.1. Electricity Generation and Use* ### Annex Table 3.7.1. The Distributional Impact of Carbon Tax and Mitigation Measures ### Annex Table 3.7.1. The Distributional Impact of Carbon Tax and Mitigation Measures ### Policy scenarios modeled - 50 US dollar per tCO2 tax on carbon where the revenue is used to finance: - Low-energy government spending - Universal cash-transfers - Targeted cash-transfers to the bottom two quintiles of the income distribution - Feebates (a subsidy to the consumption of clean energy) ### United States — Percentage change with respect to baseline - Gini coefficient: - Low-energy government spending: 0.35 - Universal cash-transfers: -1.35 - Targeted cash-transfers: -2.24 - Feebates: -0.28 - Skill premium (ratio of wages of workers with more than high school education over wages of workers with at most high school education): - Low-energy government spending: 1.72 - Universal cash-transfers: 1.10 - Targeted cash-transfers: 0.14 - Feebates: -0.41 ### China — Percentage change with respect to baseline - Gini coefficient: - Low-energy government spending: 0.12 - Universal cash-transfers: -3.81 - Targeted cash-transfers: -4.52 - Feebates: -0.27 - Skill premium (ratio of wages of workers with more than high school education over wages of workers with at most high school education): - Low-energy government spending: 2.52 - Universal cash-transfers: 1.53 - Targeted cash-transfers: 0.24 - Feebates: -1.51 ### Model and measurement note - Source: IMF staff calculations. - The Table shows the result of the multi-sector heterogeneous agent model simulation of a 50 US dollar per tCO2 tax on carbon where the revenue is used to finance the four measures listed above. - The table shows the percentage change with respect to the baseline of the Gini coefficient and the skill premium, measured as the ratio of wages of workers with more than high school education (skilled) over the wages of workers with at most high school education (unskilled). *Source: IMF staff calculations.* --- _Source: https://www.imf.org/-/media/files/publications/weo/2020/october/english/annexch3.pdf_