Environmental impact assessment for climate change policy with the simulation-based integrated assessment model E3ME-FTT-GENIE

• Climate policies can have far-reaching, unintended socio-economic effects across sectors and actors, including risks to energy access, employment, food, water, and financial stability.Source: §1.1 · p. 2
• Impact assessment must evaluate detailed policy baskets and their interactions rather than recommend policies that appear optimal in an expansive policy space.Source: §1.1 · pp. 2–3
• Total energy-system cost, consumption loss, and GDP loss are insufficient indicators as policy-makers seek information on wider impacts, including macroeconomic effects of financing technological change.Source: §1.1 · p. 3
• The paper introduces E3ME-FTT-GENIE1 to assess environmental policies using a detailed policy definition and explore associated macro-financial issues.Source: §1.1 · p. 3

2.1. Pervasive property: fundamental uncertainty means no equilibrium

• Fundamental uncertainty means agents cannot know all decision outcomes or their probabilities, so behavior cannot be reduced to optimization; spare capacity and unused resources may result.Source: §2.1 · pp. 3–4

2.2. There is no optimality in policy-making

• Policy-making has no guaranteed optimum: policies must be assessed for effectiveness, side effects, and political and legal feasibility, with feasible options tested across key indicators.Source: §2.2 · p. 4

2.3. Path-dependence and the need for simulation models

• Policy can influence technological change, but diffusion is path-dependent and uncertain; because outcomes depend on past conditions, the paper argues that simulation is needed to model them.Source: §2.3 · pp. 4–5

3.1. Overview of the integrated assessment simulation model

• E3ME-FTT-GENIE1 links a macroeconomic simulation, FTT technology-diffusion modules, resource modules, and GENIE climate modelling to assess policy outcomes.Source: §3.1 · pp. 5–6
• The model uses a “what if” approach: users select real-world policy instruments and observe outcomes rather than optimize policy scenarios.Source: §3.1 · p. 6

3.2. The macroeconomic model E3ME

• E3ME represents 59 regions and 43 sectors per region, using bilateral trade equations, input-output linkages, and data from 1970–2015 to simulate 2016–2050.Source: §3.2 · p. 6
• Because E3ME allows spare resources, it can sometimes predict positive economic and social benefits from climate-mitigation technology policies, including possible double dividends.Source: §3.2 · p. 7
• In E3ME, econometric equations determine consumption, investment, and imports, while output follows the supply-demand identity; investment and R&D contribute to technological progress and price declines.Source: §3.2 · p. 7

• Technology adoption is shaped by network effects and path dependence, producing S-shaped diffusion and momentum as established technologies gain market share [55].Source: §3.3 · pp. 7–8
• FTT represents replacement as gradual transfers between technology market shares, using the Lotka-Volterra competition equation [58–60].Source: §3.3 · p. 8
• Binary logits represent heterogeneous agents’ pairwise preferences, but substitutions also depend on technology availability and the Lotka-Volterra dynamics.Source: §3.3 · pp. 8–9
• FTT calibrates initial diffusion to historical adoption rates, and preference diversity makes significant diffusion require cost differences to exceed perceived-cost diversity.Source: §3.3 · p. 9

3.4. A dynamical fossil fuel resource depletion model

• Fossil-fuel extraction is distributed across cost ranges, with marginal cost setting commodity price and reduced extraction under stable demand raising marginal cost as reserves deplete.Supply and price are iteratively aligned with E3ME-FTT demand; the resulting fuel price changes are proportionally scaled in E3ME.Source: Figure 3 · p. 9, §3.4 · p. 10

3.5. Limits to renewables

• Renewable resource potentials are represented as cost-supply curves, so greater use raises marginal cost and can make less-suitable sites less attractive to investors.The resource database covers examples including land suitable for wind, hydroelectric sites and geothermal areas.Source: §3.5 · pp. 10–11

3.6. Climate model GENIE

• GENIE-1 simulates atmospheric CO2 and climate from emissions, land-use change and other forcing, using 86-member ensembles that vary 28 parameters.The economy-to-climate coupling is soft-linked; the model currently lacks land-use/land-use change emissions modelling, increasing uncertainty in results.Source: §3.6 · pp. 11–13, Figure 4 · p. 12

3.7. Policy instruments in E3ME-FTT-GENIE1

• Policies are represented as sector-wide or economy-wide instruments in E3ME, or as instruments for specific technologies or applications in FTT, spanning incentives, regulations, procurement and monetary tools.The taxonomy is less detailed than qualitative policy or legal assessments, but is described as sufficient to represent diverse instruments and their interactions for policy-makers.Source: §3.7 · p. 13

3.1. Coupling of the models to one-another

• Macroeconomic and technology systems are dynamically integrated and solved iteratively, with feedbacks that include electricity prices and demand, investment, fuel use, and public revenues and expenditures.This coupling also includes the fossil-fuel depletion model and FTT modules for power, road transport and household heating; GENIE is soft-linked.Source: §3.1 · p. 14

4.2. Technology diffusion for reaching 2°C

• In the 2°C scenario, renewables and electric vehicles diffuse faster after successive waves of more efficient vehicles, changing fuel use in sectors responsible for over 60% of CO2 emissions.Alternative emissions extrapolations yield 75% probability for linear, 70% for exponential, and 68% for exponential decay to 0.5 GtC residual emissions.Source: §4.2 · pp. 16–17, Figure 5 · p. 16

4.1. Scenario for >70% chance 2°C

• The combined sectoral contributions reach a greater than 70% probability of not exceeding 2°C, which the authors consider consistent with the Paris Agreement.The example basket achieves greater than 66% probability and is not presented as the only route; BECCS is not dominant, even though that means higher overall system cost.Source: §4.1 · p. 15, Figure 5 · p. 16, §4.2 · p. 17

4.3. The economics of a 2°C scenario in a simulation model

• Investment and building activity can raise employment, household income, consumption, and GDP despite higher prices, while fossil-fuel exporters can lose more from declining exports than low-carbon investment generates.Fast-growing economies receive intensified investment; fossil-fuel exporters face stranded assets, while fossil-fuel importers benefit moderately from reduced fuel expenditures.Source: §4.3 · p. 18

4.4. Comparison to other energy models

• FTT's history-dependent diffusion, endogenous learning, and stability constraints yield baseline hybrid and electric-vehicle diffusion and link renewables with flexible power systems.This simulation-based approach contrasts with the system-wide cost or utility optimization used by most other models.Source: §4.4 · pp. 18–19

4.5. Sources of variations in macroeconomic impacts of climate policies

• The pace of technological change, rather than overall ambition, determines most E3ME economic impacts; stranded fossil-fuel assets can cause substantial losses in exporting regions.Unlike standard equilibrium models, E3ME-FTT relates macroeconomic impacts to transition pace and how quickly the economy absorbs change.Source: §4.5 · pp. 19–20

4.6. Deconstructing a basket of policies for 2°C

• 2.7°C is the maximum warming increase reported when any one country pulls out; carbon pricing is the most important instrument, but removing multiple policies has nonlinear effects.The analysis uses over 50 simulations, and other policies remain important for changing technology composition outside the power sector.Source: §4.6 · p. 20

4.7. Sensitivity analyses: technology uncertainty

• For cost changes of 20%, technology shares change by at most 20% by 2050.The authors note that uncertainty accumulates over simulation time, so long-term outcomes can shift substantially with small input-parameter changes.Source: §4.7 · p. 20

4.8. Summary for policy-making

• Policy interactions prevent expressing most instruments as carbon-price equivalents; transport procurement can strengthen taxes and subsidies, while power-generation regulations change carbon-price effectiveness.Because multiple policy frameworks can reach similar emissions outcomes, the model is better suited to detailed impact assessment than agenda setting or identifying an optimum.Source: §4.8 · pp. 20–21

5. Conclusion: blueprint for a new role for integrated assessment models post-Paris

• Implementing the Paris Agreement requires trial and error, while policy-makers have limited time and need detailed assessment of policy components to build support for legislation.The passage links assessment of a broader strategy to sufficient public and political support for turning it into law.Source: §5 · p. 21
• The integrated model simulates the economy, technology, and climate system with detailed policy instruments to assess complex low-carbon policy baskets and their ability to meet climate targets.The authors describe it as a first-of-a-kind approach.Source: §5 · p. 21
• Policy analysis should account for uncertainty and context-specific instrument details; the authors offer this model as a blueprint for analyzing low-carbon policy impacts worldwide.They argue simplified storylines may mislead and policy-making cannot be reduced to pricing-only strategies.Source: §5 · p. 21

1.1. Basic model information

• E3ME uses yearly error-correction steps, while FTT uses shorter path-dependent steps; the models iterate until values stabilize, and GENIE1 is soft-linked.The coupled E3ME-FTT iteration typically requires 50–100 iterations; dynamic coupling with GENIE1 is identified as needed for strong economy-climate feedbacks such as Amazon deforestation.Source: §1.1 · p. 26

1.2. Sensitivity analysis and uncertainty propagation across E3ME-FTT

• Technology shares in 2050 change relatively little under parameter shifts of about 20% for costs or 5% for discount and learning rates, with smaller changes in the 2°C scenario than the baseline.The analysis varies technology costs, discount rates, learning rates, nonpecuniary costs, and combustion-vehicle fuel efficiency; the authors describe the resulting changes as within reasonable bounds.Source: §1.2 · p. 27
• Technology-parameter variations change cumulated GDP and employment by about 1% or less; USA GDP is more affected than China or the EU, primarily through fossil-fuel production losses.GDP and employment changes are cumulated between 2017 and 2050.Source: §1.2 · p. 27

1.3. Varying the basket of policies

• The policy-basket experiment removes each policy-instrument group from each regional group across more than 50 simulations, with peak warming evaluated at 80% probability.The study covers broad policy and country groups; a deeper analysis is outside its scope.Source: §1.3 · p. 27, Figure 1 · p. 29

1.4. Econometrics in E3ME

• E3ME specifies equations using a levels relationship to test for cointegration, followed where present by a dynamic first-difference error-correction regression with one lag per variable.Equations are estimated individually using instrumental variables, principally because many relationships are simultaneous.Source: §1.4 · p. 29

1.5. Calibration procedure

• E3ME calibration matches a published baseline, applying the same scaling factors across baseline and scenario runs so their influence largely cancels in comparisons.Baseline values can still matter for absolute targets or when the baseline is biased; sensitivity tests usually find limited influence when assumptions remain within reasonable bounds.Source: §1.5 · pp. 30–31

1.6. Forecasting performance of E3ME

• An earlier E3ME version reproduced historical economic and emissions data relatively successfully for 1970–2007, despite missing data in some regions and periods.The displayed forecasts are compared with actual historical data and were produced without calibration.Source: §1.6 · p. 32, Figure 4 · p. 32

GDP identity

• GDP combines consumer expenditure, investment, government expenditure, and exports, less imports; it is observed by the user rather than used in the model.At the sectoral level, output includes intermediate inputs, whereas value added does not.Source: GDP identity · p. 33

Calculation of output

• Output is determined from demand, including final consumption, investment, government spending, exports, and imports.The variables are defined by region and sector.Source: Calculation of output · p. 33

Balancing supply and demand

• Production is set equal to demand to balance supply and demand, with demand adjusted separately if supply constraints prevent a match.The identity equates industry output with product output.Source: Balancing supply and demand · p. 33

Calculating value added

• Value added is output less material input costs, which include unit intermediate costs and unit tax costs.Value added comprises wages, company profits, and production taxes.Source: Calculating value added · p. 33

Consumer prices

• Consumer prices combine converted industry prices with indirect taxes on consumer products.The conversion matrix maps industry production to consumer products.Source: Consumer prices · p. 34

The consumer price index

• Inflation is the annual change in the aggregate consumer price index, calculated as expenditure-weighted consumer prices relative to total consumer expenditure.The index is normalized to 2005 = 1.0.Source: The consumer price index · p. 34

Calculating nominal incomes

• Nominal household income sums non-wage income, wage income, and benefit payments, less direct taxes and employee social contributions.Non-wage income includes property and other asset rents and remittances; it may be held as a fixed differential to wage income.Source: Calculating nominal incomes · p. 34

Calculating real incomes

• Real household income is obtained by converting nominal income using the consumer price deflator, with exchange rates also included in the stated measure.Source: Calculating real incomes · p. 34

Disaggregated Final Consumption

• Final consumption by product depends on disposable income, product prices, real interest rates, price-level changes, and demographic dependency ratios.Total consumer expenditure also depends on unemployment and cumulated investment in dwellings.Source: §1.8 · pp. 34, 37

Energy demand

• Energy demand and fuel shares depend on sectoral output, energy prices, R&D spending, and investment in energy-using equipment.Source: §1.8 · p. 35

Investment

• Industrial investment depends on output, the investment-to-production price ratio, real labor costs, real long-run interest rates, and spare capacity.Source: §1.8 · p. 35

Prices

• Industry prices are represented by product prices, while domestic sales prices, import prices, and export prices depend on costs, competing or world prices, exchange rates, and technology investment.Domestic prices also include spare capacity; export prices include competing export prices.Source: §1.8 · pp. 35–36

Trade

• Imports from inside and outside a trading region depend on total supply, import and domestic prices, exchange rates, technology investment, trade-bloc progress, and spare capacity.Source: §1.8 · p. 36

Production and labor

• Industry employment and hours depend on output, wages or standard hours, technology investment, skills, and production capacity utilization.Industrial capacity also depends on recent output growth, technology investment, and working-age population.Source: §1.8 · p. 37

The labour market and household incomes

• Labor-force size is participation rates multiplied by population, while real household income combines wages, benefits, and other income net of taxes and employee social-security contributions.The retained-wages identity expresses wages relative to wages after income tax and social-security deductions.Source: The labour market and wages · p. 38, Determining household income · p. 39

Output and demand

• Real GDP and industry output are defined as demand components summed together, less imports, under the assumption that supply adjusts to meet demand.The identities include consumption, investment, government spending, exports, and imports.Source: Determining household income · p. 39

Prices and costs

• Industry prices are simplified as domestic-consumption product prices, while unit costs combine material, labor, and taxation costs.Domestic-consumption prices are derived from domestic output and trade values and volumes.Source: Price formation · p. 40, Sectoral unit costs · p. 40

Investment and sectoral aggregates

• Investment uses separate accumulation functions for ICT and other investment where data permit; otherwise, total investment is used.The identities also aggregate services and non-services production separately and sum R&D expenditure globally for vehicles and machinery.Source: Services and non-services production · pp. 41–42, Technology indices · p. 42, Global R&D for spillover effects · p. 42

Discrete choice model preferences matrix

• Pairwise vehicle-cost comparisons define a logistic discrete-choice preference matrix, with regulations determining whether a technology can be sold and Fij + Fji = 1.When TREG = 1, preferences depend on mean perceived costs and their standard deviations; otherwise Fij = 0 and Fji = 1.Source: Main FTT:Transport Identity… · p. 46

Fleet size and transport generation

• The fleet combines vehicle sales over the past 20 years with survival rates; technology shares allocate it, while service demand and use parameters determine utilization and generation.Transport service generation is the product of vehicle fleet capacity and fleet use factors.Source: Main FTT:Transport Identity… · p. 47

Transport costs and learning

• Discounted vehicle prices, fuel, operation and maintenance, and taxes determine perceived technology costs; their standard deviations inform cost dispersion, and learning lowers prices as cumulative sales grow.The cost measures include registration, fuel, and road taxes; learning-by-doing links price reductions to global cumulative vehicle sales by technology.Source: Main FTT:Transport Identity… · p. 47

Fuel use and global cumulative capacity

• For liquid-fuel technologies, fuel use is linked to transport service generation and fuel-economy factors, with the biofuel mandate applying to liquid fuel served at the pump.Global cumulative vehicle sales incorporate changes in fleet capacity and technology lifetimes.Source: Main FTT:Transport Identity… · p. 48

FTT:Transport policy inputs and exogenous values

• Exogenous transport inputs include technology characteristics, registration and fuel-economy taxes, fuel and road taxes, vehicle-sale regulations, biofuel mandates, and average vehicle prices.Source: Main FTT:Transport Identity… · p. 48

FTT:Heat technology market shares (non-linear differential equation)

• Heating technology shares change through pairwise preferences based on perceived costs, with regular and premature-replacement choices represented separately.The share-change equation also incorporates technology lifetimes, behavioural assumptions, and payback thresholds.Source: §1.1 · pp. 48–49

Heat generation from shares and demand

• Heat generation equals each technology’s share multiplied by total heat-service demand, and heating capacity is derived using its capacity factor.Source: §1.1 · p. 49

Mean levelised cost for generating heat by technology

• Levelised, marginal, and payback-based perceived heating costs include different combinations of capital, fuel, operation and maintenance costs, taxes, subsidies, and feed-in tariffs.The payback-based cost includes investment cost adjusted by capacity factor and behavioural payback-time threshold.Source: §1.1 · p. 49

Global cumulative capacity

• Investment costs decline with cumulative installed capacity through a learning relationship, while cumulative capacity reflects additions and technology lifetimes.Source: §1.1 · p. 50

Final energy demand for heat generation

• Fuel use for heat generation is calculated from technology-level heat generation and fuel-use factors.Source: §1.1 · p. 50

FTT:Heat policy inputs and exogenous values

• The model’s policy inputs include capital subsidies, feed-in tariffs, regulations, fuel taxes, technology characteristics, and demand for heating services.Source: FTT:Heat policy inputs and exogenous… · pp. 50–51

1.2. Natural resource database and depletion algorithm

• The fossil-fuel depletion calculation iterates until global fuel use (ERRY) and marginal cost (MERC) stabilize.It solves for MERC using a discrete integral over production cost C.Source: §1.2 · p. 51
• Global fossil-fuel supply (ERRY) is calculated from coal, heavy oil, crude oil, middle distillates, and natural-gas use.The listed fuel-use quantities are summed across fuel users.Source: §1.2 · p. 51
• The renewable-resources use algorithm is based on cost-supply curves, with marginal production cost represented as a function of electricity generation and resources left in the ground.The passage identifies these inputs as MEWG and BCSC.Source: §1.2 · p. 51
• The resource database covers non-renewable and renewable resources in 59 regions.Reserve-to-production ratios are listed as an input in the FTT:Transport policy inputs.Source: FTT:Transport policy inputs and… · p. 51