The State-level impacts of the introduction of a carbon tax in the United States

Authors:

Jon Stenning, Cambridge Econometrics

Hector Pollitt, Cambridge Econometrics

Unnada Chewpreecha, Cambridge Econometrics

Lead contact:

Jon Stenning; Cambridge Econometrics; Covent Garden, Cambridge, CB1 2HT, UNITED KINGDOM;

Tel:+44 1223 533138; Email: js@camecon.com.

Abstract This paper explores the potential role for carbon prices in the decarbonization of the US economy, at

a national and state level. It explores the effectiveness of carbon pricing policy, and the trade-off

between realizing reductions in GHG emissions and the economic and political feasibility of policy.

We apply a new state-level macroeconomic simulation model (E3-US) to explore the impacts of the

imposition of different levels of carbon tax, initially within a single state, and then across the US as a

whole, setting out at each stage the emissions reductions that might be achieved and the

macroeconomic implications, including both direct impacts (price impacts on energy consumers) and

indirect/induced effects (through supply chains, and the impacts of changing prices on consumption

of various goods/services) including rebound effects, where the recycling of carbon tax revenues

back into the economy can lead to additional energy demand and emissions. The aim is to

demonstrate the potential impacts of carbon taxation policy in a non-optimised economy.

Context Carbon taxes are often presented as a favoured policy for tackling climate change (Kaufman 2018).

Such a policy internalises (through the price mechanism) the externality of the contribution of

greenhouse gas (GHG) emissions to climate change. The High-Level Commission on Carbon Prices

(Stiglitz and Stern 2017) concludes that imposing a cost on carbon is the most economically efficient

way to reduce greenhouse gas emissions to keep global temperature increases within the targets of

the Paris agreement. The advantages of a carbon tax are that it can have a relatively simple design

(although, as the Washington State example shows, revenue recycling design is a key component),

the price certainty provided and that, in theory, there is a straightforward relationship between the

level of the tax and the quantity of emissions reduction achieved. However, many of these

conclusions are based upon a neoclassical view of economics, whereby agents in markets operate

‘rationally’, with perfect information, and therefore markets themselves operate in an optimal

fashion. While assessments have been carried out of the potential macroeconomic impacts of

carbon taxes in the US (see, for example, (Diamond and Zodrow 2018)), these have typically utilised

general equilibrium approaches, which are based upon similar assumptions as to the rationality of

agents and the optimality of markets.

In this paper, we instead apply an econometric energy-environment-economy model, built upon a

post-Keynesian framework, to assess the macroeconomic impacts of carbon taxes. Neoclassical

economics discusses carbon taxes from the cost perspective; however, there is a substantial body of

evidence that environmental taxation can have a positive impact on economic growth and

competitiveness (European Environment Agency 2012) (The Ex’tax Project 2016) (OECD 2017). This

can potentially bolster the arguments in favour of the introduction of a carbon tax; if such a policy

can increase employment, and raise household incomes, alongside reducing emissions and reducing

other taxes (as with I-732) or promoting a better natural environment (I-1631), then it may prove

more attractive to the population.

In November 2016, Washington Initiative 732 (I-732) was rejected in a state-wide ballot across

Washington State. The initiative sough to introduce a carbon tax on the use of fossil fuels, starting at

$15 per tonne in 2017, rising to $25/t in 2018 and from there increasing steadily to $100/t (all in

2016 $, i.e. adjusted for inflation). The revenues from the carbon tax were to be used to cut taxes –

reducing state sales tax by 1%, funding the working families tax rebate (a tax credit for low-income

households) and reducing taxes on (manufacturing) businesses. This method of revenue-recycling;

using them to cut existing taxes, and therefore not increase overall government revenues; was

deemed to be the most politically acceptable, in the sense that it does not increase the overall scale

of government activities, while also reducing the tax burden on the lowest earning households.

Despite the tax being set at a relatively low level (compared to those often shown to be required to

meet broader climate change goals), designed in a way which would minimise impacts on low-

income households, and in a way that ensured that there was no increase in the role of the state in

the economy, the initiative was defeated by 59.3% to 40.7%. A subsequent alternative measure (I-

1631), which sought to address some of the perceived shortcomings of I-732 – by reducing the level

of the tax, and using the revenues to invest in clean technologies and environmental projects – was

also defeated, in November 2018, by 56.3% to 43.7%. It is therefore clear that these measures lie on

the unacceptable side on the question of political feasibility, at least at the time that the question

was put to voters.

The defeat of these measures can be seen as something of a litmus test for carbon taxes in the

current political climate – and their defeat suggests that the introduction of such measures in the

near future is going to be extremely challenging. Washington State was identified by the Carbon Tax

Center as a ‘promising’ state for the deployment of carbon tax initiatives, alongside seven others

(Bauman and Komanoff 2017). Despite this, such policy measures continue to be explored by state

actors; for example ten states continue to operate within the Regional Greenhouse Gas Initiative

(RGGI). A key question for future policy in this area therefore is what level of carbon taxation can

feasibly be approved by voters (either directly, as was the case in Washington State, or indirectly,

whereby an elected figure will only support a carbon tax policy that will not lead to them being

ousted at the next election), and what are the emissions reduction that such a policy could achieve;

most fundamentally, is the emissions reduction worth the political capital that would have to be

expended to realise such a policy?

Method The modelling in this paper has been done using a new energy-environment-economy model, E3-US.

E3-US is a macroeconomic model designed to assess energy-economy linkages within the US at the

state level. The model follows the approach developed over many years in Cambridge Econometrics’

global E3ME model. The model is designed to capture the energy, economic and environmental

impacts of potential policies. It is a macro-econometric simulation model, meaning that it is based

upon a series of econometric equations estimated based upon historical relationships, built within a

national accounts framework. The model captures the impact of policy within the economy or

energy system, and feedbacks between these systems allow the assessment of both direct effects

(e.g. changes in energy demand as a result of price changes) and indirect effects (e.g. the

employment and value added impacts from changing energy demand, and how those impact

consumer incomes/expenditures and effects cascade through the state/national/global economy). It

has been designed in such a way that different carbon tax rates can be applied to different groups of

energy users, and so that revenues collected from such policies can be used for different purposes,

including reducing other tax rates or funding energy efficiency programs.

Figure 1 shows how the three components (modules) of the model - energy, environment and

economy - fit together. Each component is shown in its own box. Each data set has been constructed

by statistical offices to conform with accounting conventions. For each state’s economy the

exogenous factors are economic policies (including tax rates, growth in government expenditures,

interest rates and exchange rates). For the energy system, the outside factors are the world oil

prices and energy policy (including regulation of the energy industries). For the environment

component, exogenous factors include policies such as reduction in CO2 emissions tax. The linkages

between the components of the model are shown explicitly by the arrows that indicate which values

are transmitted between components.

The economy module provides measures of economic activity and general price levels to the energy

module; the energy module provides measures of emissions of the main air pollutants to the

environment module. The energy module provides detailed price levels for energy carriers

distinguished in the economy module and the overall price of energy as well as energy use in the

economy.

Figure 1 E3 linkages in the E3-US model

Treatment of inter-state and international trade

In a sub-national model, trade represents a major issue in assessing regional economic impacts.

Demand in each state can be met either by production within that state, production in another state

in the US, or production in another country.

The approach can be summarized as:

• econometric estimation of state’s sectoral international import demand

• econometric estimation of state’s sectoral international export demand

• trade between states is estimated using production shares (export) and domestic demand shares (import)

The labor markets

Treatment of the labor market is an area that distinguishes E3-US from other macroeconomic

models. E3-US includes econometric equation sets for employment, wage rates and participation

rates. The first two of these are disaggregated by economic sector while participation rates are

disaggregated by gender.

The labor force is determined by multiplying labor market participation rates by population.

Unemployment (including both voluntary and involuntary unemployment) is determined by taking

the difference between the labor force and employment. This is typically a key variable of interest

for policy makers.

The role of technology

Technological progress plays an important role in the E3-US model, affecting all three E’s: economy,

energy and environment. The model’s endogenous technical progress indicators (TPIs), a function of

R&D and gross investment, appear in ten of E3-US’s econometric equation sets including trade, the

labor market and prices. Investment and R&D in new technologies also appears in the E3-US’s

energy demand equations to capture energy savings technologies as well as pollution abatement

equipment. In addition, E3-US also captures low carbon technologies in the power sector through

the FTT power sector model (see below).

The modelling of the power sector

The power sector in E3-US is modelled at three interconnecting areas:

• The Eastern Interconnection

• The Western Interconnection

• The Electric Reliability Council of Texas (ERCOT) covers most of Texas.

The treatment of the power sector in E3-US is based on a novel framework for the dynamic selection

and diffusion of innovations, initially developed by J.-F. Mercure (Mercure, 2012), called FTT:Power

(Future Technology Transformations for the Power sector). It uses a decision-making core for

investors wanting to build new electrical capacity, facing several options. The model is based on

theories of technology diffusion, with rates of diffusion affected by relative market shares and

technology prices. The detailed technology representation allows for a range of policy options,

including:

• Feed-in-Tariffs

• Investment subsidies for renewables

• Public sector construction

• Forced phase-out of old technologies

Many of the policies are characterized by long lag times due to the lifetimes of the plants that are

built. However, the model can show rapid transitions as technologies gain market penetration,

reinforced by cost reductions that result from learning rates.

The decision-making core takes place by pairwise levelized cost (LCOE) comparisons, conceptually

equivalent to a binary logit model, parameterized by measured technology cost distributions. Costs

include reductions originating from learning curves. The diffusion of technology follows a set of

coupled non-linear differential equations, sometimes called ‘Lotka-Volterra’ or ‘replicator dynamics’,

which represent the better ability of larger or well-established industries to capture the market, and

the life expectancy of technologies. Due to learning-by-doing and increasing returns to adoption, it

results in path-dependent technology scenarios that arise from electricity sector policies.

FTT:Power determines a technology mix by each interconnecting area given a scenario of detailed

electricity policy: carbon prices, subsidies, feed-in tariffs and regulations by technology. Changes in

the power technology mix result in changes of production costs, reflected in the price of electricity.

The model takes electricity demand from E3-US at state level and feeds back a price, fuel use and

investment for replacements and new generators.

Scenario design Three separate scenarios are considered in this analysis;

1. a stylized version of the Washington State’s carbon tax as proposed in I-732; a carbon tax starting at $25 per tonne of CO2 in 2020 and increasing by 3.5% per year in real terms to reach a price of around $70/tCO2 in 2050. The tax is introduced in Washington State only; no other actors (either States within the US, or countries outside the US) introduce a change to policy alongside Washington State.

2. a more ambitious policy than I-732; the carbon tax starts at around $25 tonne of CO2 in 2020, and rapidly increases to over $300 and $2,000 per tonne of CO2 in 2030 and 2050 respectively. The tax is introduced only in Washington State.

3. A US-wide carbon tax; an ambitious carbon tax (starting at $25 per tonne of CO2 in 2020 and increasing to $2,000 per tonne of CO2 as in the ambitious Washington State scenario) deployed across the whole of the United States.

Findings The first question that we sought to set out what could be the macroeconomic impacts of a carbon

tax policy within a specific state. To assess this, a stylized version of I-732 was modelled. The key

finding is that (unsurprisingly), the macroeconomic impacts depend upon whether (and how)

revenues are recycled. With no revenue recycling (i.e. additional government revenues are used to

patch holes in the state balance sheet), the impact on GDP is small and negative (see Figure 3).

However, if all revenue is recycled, then these negative impacts are largely mitigated. We assume

that revenues are recycled broadly in line with the methods proposed for I-732; through a reduction

in sales taxes and income taxes. In this scenario, GDP in Washington State is broadly unaffected

compared to baseline (the difference is less than 0.01% in 2050).

Figure 2 The basic structure of FTT:Power

However, another major finding in this scenario is that the impact on CO2 emissions is also relatively

minor. While GHG emissions have been falling in the state since 2000, they remain above 1990

levels, and it does not seem likely that the State’s aim to reduce emissions in 2020 to be in line

emissions in 1990 will be achieved. The carbon tax, as modelled, certainly does little to help with the

achievement of later goals; emissions are between 1-1.2% lower in each year to 2050, depending

upon whether revenues are recycled (since recycling revenues increases economic activity, and

therefore associated emissions). This demonstrates that a carbon tax policy that is more ambitious

than the (rejected) I-732 will be required to contribute substantially to emissions reductions in

Washington State.

In the second scenario considered, just such a policy was modelled; a more ambitious carbon tax for

Washington State, starting at €334/tCO2 in 2020. Unsurprisingly, the impacts upon GDP are more

pronounced – at least in the case of no revenue recycling. In this scenario, GDP is more than 2.5%

lower in 2050 than the baseline (where no such policy is introduced). Once revenue recycling is

added (split, as in the previous scenario, between sales and income tax reductions), the GDP impact

is again small and negative.

-0.20%

-0.15%

-0.10%

-0.05%

0.00%

0.05%

2015 2020 2025 2030 2035 2040 2045 2050

% d

iffe

ren

ce f

rom

bas

elin

e

$25-$70/tCO2 tax with revenue recycling

$25-$70/tCO2 tax without revenue recycling

Source: E3-US, Cambridge Econometrics

Figure 3 GDP impacts of I-732 style carbon tax in Washington State only

The economic impacts are broadly mirrored in the emissions reductions observed; with no revenue

recycling, the carbon price leads to a substantial volume of emissions being removed from the

economy; emissions are 10% lower than the baseline in 2050. On the face of it, the reduction in

emissions is surprisingly small, given the carbon price; the key driver of this result is emissions from

power generation, which only fall by around 3% by 2050. There are two factors which contribute to

this outcome; first, fuel switching in other parts of the economy (as final use of oil becomes

prohibitively expensive) leads to higher demand for electricity in this scenario. Second, in early years

the carbon price is not particularly high, and leads to a technology switch away from coal and

towards gas, but not towards more expensive low- or zero-carbon technologies. By the time carbon

prices reach high levels, through 2040 and beyond, the gas generation capacity is still within it’s

operational lifespan, and because there was not early take-up of low carbon technologies they

remain relatively expensive and cannot diffuse at speed within the State.

In other sectors, greater reductions in emissions are achieved; there is some reduction in emissions

from transport, reflecting the deployment of more efficient and less polluting technologies as well as

changes in demand for transport as a result of price changes. There are much more substantial

reductions in emissions from industry, commerce and households – emissions from these are 35-

50% lower than the baseline by 2050 (see Figure 5).

-3.00%

-2.50%

-2.00%

-1.50%

-1.00%

-0.50%

0.00%

2015 2020 2025 2030 2035 2040 2045 2050

% d

iffe

ren

ce f

rom

bas

elin

e

$25-$2000/tCO2 tax with revenue recycling

$25-$2000/tCO2 tax without revenue recycling

Figure 4 GDP impacts of an ambitious carbon tax in Washington State only

Source: E3-US, Cambridge Econometrics

Figure 5 CO2 emissions impact of an ambitious carbon tax in Washington State in 2050

When revenue recycling is introduced, emissions reductions are slightly lower suggesting that

emissions reductions is less effective with higher economic activity. Policy makers should therefore

take this into consideration when designing a carbon tax scheme.

Both of the scenarios above show potential impacts from unilateral impacts from a single state.

However, it is clear that a state acting in this way is likely to damage its own productivity relative to

other states, by introducing an increase in costs that is not mirrored in neighbouring states (and

therefore opening up the potential for carbon leakage, where emissions are not reduced but simply

displaced to areas outside of the state). This can be tackled through increasing the geographical

coverage of the carbon tax, for example to cover the whole of the Western Interconnection, or

indeed the whole of the US. However, such an ambitious deployment is well beyond the realm of

what is realistically possible, given current attitudes across the US (Bauman and Komanoff 2017).

In the final scenario considered in this analysis, we consider just such a case; an ambitious US-wide

carbon tax. In this scenario, the impacts on Washington State’s economy are much more

pronounced; with no revenue recycling, GDP is 8% lower by 2050, while with revenue recycling

added GDP is instead around 8% higher by 2050 in the scenario compared to the baseline (see

figure).

-60%

-50%

-40%

-30%

-20%

-10%

0%

Power Industry Transport Households Services Total

% d

iffe

ren

ce f

rom

bas

elin

e

$25-$2000/tCO2 tax with revenue recycling $25-$2000/tCO2 tax without revenue recycling

Source: E3-US, Cambridge Econometrics

Figure 6 GDP impacts of an ambitious US-wide carbon tax in Washington State

This GDP outcome reflects primarily the switch that is taking place in the power generation sector

across the US; the imposition of the carbon tax makes end use of coal and oil within power

generation, particularly expensive, and essentially uncompetitive with alternative generation

technologies. This provides a benefit to Washington State, where these is little coal-fired generation

but substantial capacity fired by natural gas. Due to the interconnected nature of grids within the

Western Interconnection, the coal-fired electricity that is lost in states across the grid is replaced by

additional electricity generated by gas in Washington State – essentially, Washington State exports

more electricity than in the baseline. Since this electricity still attracts a carbon price (it is primarily

delivered via gas, rather than zero-carbon technologies), without revenue recycling more money is

taken out of the economy in this scenario compared to a tax introduced in Washington State only,

and therefore the negative economic impacts are larger. Conversely, when this money is re-injected

into the state economy, substantial positive economic impacts are observed, as a result of the

increased demand for exported electricity, and more generally an improvement in competitiveness

across the state (reflecting the relatively low carbon-intensity of economic activity).

-10%

-8%

-6%

-4%

-2%

0%

2%

4%

6%

8%

10%

2015 2020 2025 2030 2035 2040 2045 2050

% d

iffe

ren

ce f

rom

bas

elin

e

$25-$2000/tCO2 tax with revenue recycling ALL STATES

$25-$2000/tCO2 tax without revenue recycling ALL STATES

Source: E3-US, Cambridge Econometrics

Figure 7 CO2 emissions impact of an ambitious US-wide carbon tax in Washington State in 2050

Within the state, this leads to some counterintuitive results; once revenue-recycling is accounted

for, emissions actually increase compared to baseline (see figure) – as a result of the increase in

electricity generated for export to other states. At the national level, the effects are clearer;

emissions are reduced by around 15% in the US by 2050, even accounting for the impacts of revenue

balancing (see Figure). At the same time, substantial improvements in GDP are observed; at the US

level, GDP is almost 8% higher in 2050 when all revenues are recycled through tax cuts.

-60%

-50%

-40%

-30%

-20%

-10%

0%

10%

20%

Power Industry Transport Households Services Total

$25-$2000/tCO2 tax with revenue recycling ALL STATES

$25-$2000/tCO2 tax without revenue recycling ALL STATES

Source: E3-US, Cambridge Econometrics

Figure 8 National CO2 emissions impacts of an ambitious US-wide carbon tax

This scenario very clearly demonstrates the importance of the rebound effect upon the impacts that

derive from the carbon tax policy; by using revenues accrued through the tax to bolster the spending

power of US consumers (via tax cuts), additional demand is being created in the economy, and that

demand is met through production processes which increase carbon emissions.

In summary, this modelling highlights that carbon taxes can deliver economic benefits across the

economy under the right circumstances, alongside emissions reductions – but that there is an

intrinsic trade-off between the scale of the economic and environmental gains that can be made.

Conclusions Given the political and social constraints in the US currently, it seems unlikely that any state will be

able to introduce carbon tax legislation that will have a material impact upon GHG emissions. The I-

732 legislation that was proposed (and voted down) in Washington State in 2016 would have

reduced emissions by only around 1% per year, according to our modelling. This makes it clear that

any policy aimed at more stringent reductions would struggle to gain acceptance.

However, our modelling has shown that a more ambitious policy could achieve substantial emissions

reductions, primarily by encouraging a switch in power generation from coal to gas, and through

reducing energy consumption from industry and households. The specific design of policy matters,

however; revenues from the imposition of the tax should be redistributed – although this paper does

not seek to explore the impact of different methods of revenue recycling. The coverage of the tax

also matters; the wider the geographical jurisdiction of policy, the greater the opportunity for

carbon emission reductions.

-25%

-20%

-15%

-10%

-5%

0%

2015 2020 2025 2030 2035 2040 2045 2050

% d

iffe

ren

ce f

rom

bas

elin

e

$25-$2000/tCO2 tax with revenue recycling ALL STATES

$25-$2000/tCO2 tax without revenue recycling ALL STATES

Source: E3-US, Cambridge Econometrics

Perhaps most importantly, this analysis finds that even the most ambitious scenarios in this

modelling do not achieve emissions reductions in line with the stated aims of most local, state or

federal actors; this highlights that carbon taxes are only part of the solution to reducing GHG

emissions in the US, and a balanced portfolio of policies (likely including measures encouraging

energy efficiency, and measures to encourage the take-up of specific technologies such as electric

vehicles and renewable electricity generation) will be required to move the US onto a path towards

becoming a low-carbon economy; see for example (Mercure, et al. 2018) (Knobloch, et al. 2018).

Some of these findings (primarily the potential for positive macroeconomic outcomes) run counter

to the established narrative (for example, (Diamond and Zodrow 2018)); this is primarily due to the

underlying view of markets and agents which underpin the different modelling approaches (i.e.

optimising agents versus econometrically-estimated behaviours). However, the factors identified as

key determinants of outcomes are in common with the existing literature.

This paper is the first application of the new state level energy-environment-economy model, E3-US.

This research has highlighted some key areas for further research, including the impact of different

methods of revenue distribution and the sectoral coverage of the tax. A further key unanswered

question is how to promote greater decarbonisation of power generation; the most ambitious

deployment of carbon tax modelled in this paper led primarily to a switch from coal-fired to gas-fired

generation (i.e. a switch from high-carbon to lower-carbon generation), not the widespread

adoption of zero-carbon technologies.

Further suggestions from readers for additional research are strongly encouraged.

References Bauman, Yoram, and Charles Komanoff. 2017. Opportunities for Carbon Taxes at the State Level.

Carbon Tax Center.

Diamond, John W, and George R Zodrow. 2018. The Effects of Carbon Tax Policies on the US Economy

and the Welfare of Households. New York: Columbia SIPA Center on Global Energy Policy.

European Environment Agency. 2012. “Environmental tax reform: increasing individual incomes and

boosting innovation.” Accessed April 28, 2019.

https://www.eea.europa.eu/highlights/environmental-tax-reform-increasing-individual.

Kaufman, Noah. 2018. “A comparison of the bipartisan energy innovation and carbon dividend act

with other carbon tax proposals.” Working Paper.

Knobloch, Florian, Hector Pollitt, Unnada Chewpreecha, Vassilis Daioglu, and Jean-Francois Mercure.

2018. “Simulating the deep decarbonisation of residential heating for limiting global

warming to 1.5°C.” Energy Efficiency 12 (2): 521-550.

Mercure, Jean-Francois, Aileen Lam, Sophie Billington, and Hector Pollitt. 2018. “Integrated

assessment modelling as a positive science: private passenger road transport policies to

meet a climate target will below 2°C.” Climate Change 151 (2): 109-129.

OECD. 2017. “Environmental Fiscal Reform: Progress, Prospects and Pitfalls.” Accessed April 28,

2019. https://www.oecd.org/tax/tax-policy/environmental-fiscal-reform-G7-environment-

ministerial-meeting-june-2017.pdf.

Roberts, David. 2016. “The left vs. a carbon tax.” 8 November. Accessed August 12, 2019.

https://www.vox.com/2016/10/18/13012394/i-732-carbon-tax-washington.

—. 2018. “Washington votes no on a carbon tax — again.” 6 November. Accessed August 12, 2019.

https://www.vox.com/energy-and-environment/2018/9/28/17899804/washington-1631-

results-carbon-fee-green-new-deal.

Stiglitz, J. E, and N Stern. 2017. Report of the High-Level Commission on Carbon Prices. World Bank

Group, Carbon Pricing Leadership Coalition. Accessed March 10, 2019.

https://static1.squarespace.com/static/54ff9c5ce4b0a53decccfb4c/t/59244eed17bffc0ac25

6cf16/1495551740633/CarbonPricing_Final_May29.pdf.

The Ex’tax Project. 2016. “New era. New plan. Europe. A fiscal strategy for an inclusive, circular

economy.” Accessed April 28, 2019. http://www.neweranewplan.com/wp-

content/uploads/2016/12/New-Era-New-Plan-Europe-Extax-Report-DEF.compressed.pdf.