On the Spillover Effects of CO2 Taxation on the Emissions of other Air Pollutants

In this paper, we compare and contrast the environmental, macroeconomic and distributive effects of CO2 taxation with the effects of taxing a variety of air pollutants at their external costs. We do so using a multi-sector and multi-household dynamic computable general equilibrium model of the Portuguese economy. We find that a carbon tax of 114 euros per ton of CO2 is necessary to achieve the IPCC 2030 targets. It does so, however, at a high macroeconomic and distributional cost. In turn, the macroeconomic and distributional effects of taxing different pollutants at their external costs in line both qualitatively and quantitatively with the effects of the CO2 taxation. In absolute terms, however, better environmental results in terms of GHG and air pollutants emissions are achieved through the level of CO2 taxation necessary to achieve the IPCC targets than through direct taxation of such emissions at their external costs. Ultimately, the benefits of complementing the CO2 taxation with the taxation of other air pollutants at their external costs does not seem significant from either efficiency, fairness, or environmental perspectives to justify the practical complexity of considering it.


Introduction
The purpose of this paper is to identify the environmental, macroeconomic and distributional effects of carbon taxation and of the taxation of a multiplicity of air pollution at their external costs. The practical objective is to determine whether the use of a myriad of policy instruments to correct air pollution externalities is necessary in the presence of the carbon taxation necessary to achieve Intergovernmental because as argued above the emissions of many of these pollutants are connected and in practical terms because the political environment is not particularly conducive to the introduction on multiple taxes and/or fees. This raises the question of identifying the effects of an overarching policy to reach the IPCC goals through proper pricing of carbon emissions on the emissions of the co-pollutants and the other greenhouse gases. Specifically, the question is to determine how much taxing carbon emissions at a level necessary to achieve IPCC goals affects the other emissions and how it compares with taxing such emissions at their own external costs.
In this paper, we compare the environmental, macroeconomic and distributive effects of a CO2 tax with the effects of taxing a variety of air pollutants at their external costs. To do so, we use the most recent version of the DGEP, the dynamic general equilibrium model of the Portuguese economy. Previous versions of this model have been used to address energy and climate policy issues (see Pereira & Pereira, 2014a, 2014b, 2017a, 2017b, 2017c, 2018and Pereira et al., 2016). This model has a detailed description of the tax system and a fine differentiation of consumer and producer goods, particularly those with a focus on energy products. We consider twenty-two sectors spanning the all spectrum of economic activity. Household heterogeneity in income and consumption patterns is captured by differentiating among five household groups based on income levels.
From a methodological perspective, this paper builds upon a vast computable general equilibrium literature. General equilibrium models have been extensively used in energy studies. For general surveys see Bhattacharyya (1996) and Bergman (2005) and for a discussion of the merits and concerns with this approach see Sbordone et al. (2010) and Blanchard (2016). Our model follows in the tradition of the early models developed by Borges and Goulder (1984) and Ballard, Fullerton, Shoven and Whalley (2009) while in its specifics is more directly linked to the recent contributions of Goulder and Hafstead (2013), Bhattarai et al. (2016), Tran and Wende (2017), and Annicchiarico et al. (2017).
In turn, from a conceptual perspective, this paper builds upon a well-established literature on co-pollutants and the co-benefits of environmental policies. Parry (2015) and Coady et al. (2018), provide overall reviews of the conceptual issues for the design of fiscal policies to address the external costs of energy use. Fullerton and Karney (2018) and Ambec and Coria (2018) Stranlund and Son (2019) provide conceptual discussions of the co-benefits of policies to address GHG emissions and local air pollutant emissions under different situations. Finally, Fichtner et al (2003), Jiang et al. (2103), and sector-specific private capital stock variables, as well as their respective shadow prices/co-state variables. In addition, the evolution of the stocks of public debt and of the foreign debt act as resource constraints in the overall economy.
This system of nonlinear first order difference equations is solved numerically using the GAMS (General Algebraic Modeling System) software and the MINOS (Modular In-Core Nonlinear Optimization Solver) solver. MINOS uses a reduced gradient algorithm generalized by means of a projected Lagrangian approach to solve mathematical programs with nonlinear constraints, which employs linear approximations for the nonlinear constraints and adds a Lagrangian and penalty term to the objective to compensate for approximation error. This series of sub-problems is then solved using a quasi-Newton algorithm to select a search direction and step length.
The calibration of the dynamic general equilibrium model of the Portuguese economy is designed to replicate, as its most fundamental base case, a stylized steady state path for the Portuguese economy.
We define the steady-state growth path as an intertemporal equilibrium trajectory in which all the flow and stock variables grow at the same rate while market and shadow prices are constant. Specifically, the steady state path is defined by the trends and information contained in the data set. In the absence of any policy changes, or any other exogenous changes, the model implementation will just replicate into the future such stylized economic trends.
We calibrate the dynamic general equilibrium model with data for the period 2005-2015 and stock values for 2015. In fact, rather than focusing on a single year of data, we use a ten-year interval. This roughly captures an entire business cycle thereby avoiding contaminating the calibrated model with business cycle effects. Although more recent data was available for most economic indicators, data on a variety of energy indicators has only been validated for Portugal through 2015 at the time calibration.
To guarantee the existence of a steady state for the dynamic general equilibrium model there are three types of calibration restrictions. First, calibration determines the value of critical production parameters, such as adjustment costs and depreciation rates, given the initial capital stocks. These stocks, in turn, are determined by assuming that the observed levels of investment of the respective type are such that the ratios of capital to GDP do not change in the steady state. Second, the need for constant public debt and foreign debt to GDP ratios implies that the steady-state budget deficit and the current account deficit are a fraction of the respective stocks of debt equal to the steady-state growth rate. Finally, the exogenous variables, such as public transfers or international transfers, have to grow at the steady-state growth rate.

Reference Scenario
The reference scenario serves as a basis for evaluating the impact of policies that follow. The reference scenario embodies several assumptions regarding climate policy and technological progress, which are superimposed on the steady state trajectory used in the calibration of the model. The main climate policy considerations present in our reference scenario are first, that a tax of 6.85 Euro/tCO 2 persists at this level through 2050 and second that the major coal fired power plants in Portugal cease operations at the end of their useful life and no additional coal capacity is installed. Power has two major coal fired power plants, one in Sines and one in Pego. The plant in Sines is scheduled to close in 2035 and the plant in Pego in 2040. Third, we assume that fossil fuel prices follow forecasts developed by the International Energy Agency (2018).
Given this reference scenario, counterfactual simulations allow us to identify marginal effects of any policy or exogenous change, as deviations from this reference scenario.

Greenhouse Gases
We incorporate in the model GHG emissions considered within the common reporting framework of the IPCC framework (see, for example, IPCC, 2019) and which represent the whole universe of GHG pollutants in Portugal: Carbon Dioxide (CO 2 ); Methane (CH 4 ); Nitrous Oxide (N 2 O); Hydrofluorocarbons (HFC); Perfluorocarbons (PFC); and Sulfur Hexafluoride (SF 6 ). See Figure 1.

Emissions by Gas
Emissions by Activity Of the GHG considered, carbon dioxide, and in a small part methane, are directly related to the combustion of fossil fuels. In turn, the bulk of emissions from methane and remaining GHG derive mostly from agriculture and a variety of industrial processes.

Air Pollutants
In turn, we incorporate in the model the air pollutants considered within the National Emission Ceiling

Figure 2. Air Pollutants in 2015
These air pollutants are induced by the combustion of fossil fuels, either directly as is the case of nitrogen oxide and sulfur dioxide or indirectly by road transportation activities such as particulate matter, volatile organic matter and carbon monoxide. These are the relevant co-pollutants when we consider policies designed to reduce carbon dioxide emissions.

On the Modelling of the Different Emissions
We model emissions of the different GHG and air pollutants in two different ways. For emissions that are generated by fossil fuel combustion, i.e., the co-pollutants with carbon dioxide, we model emissions as direct function of the amount of the fossil fuel used in the corresponding activities. For emissions that are induced by agriculture of industrial processes we modelled them as a fixed function of the output of each of the different production sector or activities.
From a conceptual perspective, for fossil fuel based emissions, carbon dioxide and its co-pollutants, we capture the following three effects of the different policies: effects due to fossil fuel switching; effects due to changes in the level of economic activity; and effects due to changes in the composition of economic activity.
For process-based emissions, we capture only the two following effects of policies: effects due to changes in the level of economic activity; and effects due to changes in the composition of economic activity. Accordingly, in this work, the effects of the different policies on process-based emissions are underestimated by the amount of process switching the policies may generate.
It should be noted that, given the focus and level of aggregation of the analysis, we implicitly assume  that the different co-pollutants are complements with carbon dioxide. Although there is a debate in the literature on whether one should observe complementary of substitution among co-pollutants our approach is consistent with the arguments and evidence in Fullerton and Karney (2018) to the effect that under the most plausible parameter specifications emissions of CO 2 and co-pollutants are complements.

Benefits Table Database (BeTa) for Air Pollutants
Of the air pollutants considered above we consider taxation of sulphur dioxide, oxides of nitrogen, particulate matter, volatile organic compounds and carbon monoxideall in some way related to combustion or closely related activities -at their external costs.  As one can observe in Table 1, the external costs of the different pollutants for Portugal are in general substantially below the EU-15 average. This is due to differences in purchasing power vis-à -vis the other countries and to the fact that some of measured externalities depend critically on standards of living, population density, etc.

Simulation Results
We start by analyzing the environmental, macroeconomic, and distributional effects of a CO 2 tax of the magnitude necessary to reach IPCC 2018 goal of a 45% reduction in CO 2 emissions by 2030 relative to the 2010 levels. Then, we consider the corresponding effects of taxing the different air pollutants at their external costs. We present the simulation results in Tables 2-7.

On the Effects of CO2 Taxation
The magnitude of the carbon tax necessary to reach IPCC 2018 CO 2 reduction goals is 114 euros per ton of CO 2 . This tax generates tax revenues that are approximately 1.85% of the GDP.

Effects on Energy Markets and Emissions
The introduction of this CO 2 tax leads to an increase in energy prices of 13.91% and to a decrease of energy demand by 12.40%. The price of domestic electricity generation itself increases by 12.59%, which leads to a 10.17% decrease in domestic electricity production and a 12.81% increase in electricity imports. Overall electricity demand declines by 9.80%. Accordingly, the share of electricity in final energy demand increases by 2.97%.
The introduction of the CO 2 tax leads to a reduction in CO 2 emissions of 36.02% which represents 53.8% of the 2010 levels. The CO 2 tax induces significant reductions in other GHG emissions, in particular CH 4 and in N 2 O emissions, which decline by 25.29% and 30.73%. It induces smaller reductions for emissions of HFC, PFC, and SF 6 . The CO 2 tax leads also to significant reductions of emissions of air pollutants. This is true particularly for emissions of NO x , SO 2 , CO, and PM, which decline by 37.22%, 43.13%, 51.08%, and 71.71%, respectively and less so for emissions of VOC and NH 3 .

Macroeconomic and Distributional Effects
The macroeconomic effects of the CO 2 tax are naturally adverse. GDP declines by 5.21% linked directly on the supply side to the reduction in investment by 1.33% and of employment by 2.71% and on the demand side by a reduction in private consumption of 1.21%. The CPI increases by 2.32%. In turn, foreign debt increases by 3.70% with increased reliance of relatively cheaper foreign goods.
Finally, there is by construction a reduction of 12.66% in the public debt.
The industries that are the most adversely affected in terms of their output are petroleum refining and electricity generation as expected as well as rubber, basic metals, equipment, and transportation as well as textiles, wood and chemicals. These are all internationally traded goods.
Overall, there is an aggregate welfare loss of 1.34%. Across the different income groups, this loss is felt in a regressive manner. Indeed, the lowest income group suffers a loss of 1.85% while the highest income group loses just 1.02%. Accordingly, the factor of regressivity is 1.8.

On the Effects of Taxing other Pollutants at their External Costs
In counterfactual simulation CF2, we consider the results of taxing air pollutants at their external costs as detailed in Table 1. The corresponding tax revenues are 0.67% of the GDP and therefore about 36% of the CO 2 tax revenues considered in CF1.

Effects on Energy Markets and Emissions
The effects on the energy market essentially mirror the effects induced by the CO 2 tax. Quantitatively, they are in line with the relative magnitude of the two policies. Qualitatively, there are no significant changes in the observed patterns of results.
In turn, CO 2 emissions decrease by 21.38%, which means that they reach 73.2% of the 2010 levels. is a substantial cross effect on CO2 emissions coming from the reduction in economic activity but also from the fact that that the pollutants being taxed are directly or indirectly related to the combustion of fossil fuels.
The cross effects on emissions of other GHG are in line with the relative magnitude of the two policies except for N2O, in which case the reduction is now 15.50% or about 50% of what observed under the CO2 tax.
In turn, reductions in air pollutants are enhanced greatly under the direct taxation of their external costs.
The largest reductions occur with emissions NO x , SO 2 , CO, and PM, which decline by 25.45%, 31.57%, 34.14%, and 55.69%, respectively and less so for emissions of VOC and NH 3 .
Overall, with an overall tax levy just over one third of the CF1

Economic and Distributional Effects
The macroeconomic effects under CF2 are, broadly speaking, about one-third of the effects observed under CF1. Therefore, they are in line with the relative magnitude of the two policies. Qualitatively, there are no changes.
The sectors affected under CF2 are essentially the same as under CF1 although there are some small differences in the relative importance of the outputs reductions across sectors compared to CF1.
identify the relevance of the environmental spillovers of CO2 taxation.
We can summarize our simulation results as follows. A carbon tax of 114 euros per ton imposed on top of the current energy taxation is enough to achieve the IPCC 2030 targets as well as significant reductions in other GHG emissions as well as emissions of air pollutants. It does so, however, at a high macroeconomic and distributional cost. The macroeconomic and distributional effects of taxing different pollutants at their external costs are closely aligned with the effects of carbon taxation. They show the same qualitative patterns and the different in magnitude is in line with the relative magnitude of the two policies. Yet, under the taxation of different air pollutants at their external costs, CO 2 , N2O, NOx, SO2, CO, and PM emissions decline much more than proportionally vis-à -vis the relative magnitude of the two policies. Still, such policy is not enough to generate the desired reductions in CO 2 emissions. More importantly, however, in absolute terms better environmental results in terms of GHG emissions and the air pollutants are achieved through CO 2 taxation than through direct taxation of such emissions at their external costs.
The results pertaining the introduction of other GHG gases and the different air pollutants raise the question of the environmental relevance of independent taxation of the different air pollutants in addition to CO2 taxation. That is, it questions the relevance of using multiple tax instruments to achieve reductions in different emissions that are linked through technological and economic conditions. Ultimately, the benefits of complementing the taxation of carbon dioxide with the taxation of other air pollutants at their external costs does not seem significant from either efficiency, fairness or environmental perspectives to justify the complexity of considering them. Indeed, a greater reduction in the emissions of all GHG and of all air pollutants is achieved simply by using a CO2 tax to achieve the IPCC CO2 emissions targets.
These results and recommendations are fully consistent with recent evidence in the literature. For example, Muller (2012) and Crago and Stranlund (2015) show that co-benefits of GHG policies can be significant in magnitude and argue that it is not socially beneficial that climate policies should be tailored to reflect these local air pollution co-benefits. In turn, Brunel and Johnson (2019) local pollution policies are unlikely to be of the magnitude necessary to address greenhouse gas targets. We add the macroeconomic and distributional dimension to the issue to suggest that the policy focus should be on developing an adequate carbon tax and counting on its spillovers to achieve the desired reductions in the emissions of air pollutants.
This research opens the door to a few critical follow-ups from a practical environmental policy perspective. In this work, we assume that the revenues from carbon taxation are not recycled, i.e., they revert to the general government budget. There is, however, plenty of evidence that careful recycling of such revenues is necessary if the adverse macroeconomic and distributional effects of carbon taxation are to be avoided. (See, for example, Marron and Toder (2014), Jorgenson et al (2015), and Kirchner et al (2019)). Naturally, different recycling strategies have different macroeconomic and distributional effects and therefore different potential for rebound effects in terms of the use of the different fossil www.scholink.org/ojs/index.php/jepf Journal of Economics and Public Finance Vol. 7, No. 3, 2021 fuels and the corresponding emissions of CO2 and co-pollutants. On the flip side Parry et al (2015) highlight the relevance of recycling mechanisms in the presence of co-pollutants to increase the co-benefits of carbon policies.
Finally, and although this is an energy policy paper applied to the Portuguese economy and its policy implications directly relevant for the Portuguese case, its interest is far from parochial. The quest for decarbonization is universal. The existence of significant challenges in terms of air pollution widespread. The concerns over the macroeconomic and distributional effects of environmental policies and the quest for parsimony in the choice of instruments unavoidable if there is some hope of meaningful policies ever being adopted.