1. Introduction
Nationally Determined Contributions (NDCs) lay out the actions countries intend to take to limit greenhouse gas emissions (GHGs) under the Paris Agreement at the 2015 Conference of Parties to the UN Framework Convention on Climate Change. A very large body of analytical work indicates that putting some kind of price on GHG emissions—“carbon pricing”—can be an effective tool for GHG mitigation, although practical experience with carbon pricing is still evolving [
1]. The NDCs of more than 90 countries refer to carbon pricing in one form or the other [
2]. Carbon pricing also can yield collateral benefits including air pollution reduction and efficiency-increasing fiscal restructuring. So far, however, carbon pricing has been applied almost exclusively in high- and middle-income countries.
This paper addresses that gap by examining the impacts of a carbon tax applied to petroleum fuels and kerosene in Ethiopia. The core objective is to get a quantitative estimate of the environmental, economic, and distributional impacts of carbon taxes in Ethiopia. Implementing carbon pricing policy, of course, requires a through political economy analysis and development effective institutional approaches that work for context in consideration. Our study should therefore be taken as a first but crucial step for exploring the potential for carbon pricing in a developing-economy context.
Over the past decade, the government of Ethiopia has prioritized low-carbon growth and poverty reduction. (Ethiopia also has taken several actions to better manage adverse impacts and risks of climate change and variability [
3,
4]) For example, fossil fuel subsidies have been removed, with both economic benefits and limited social impacts, since studies show that fossil fuel subsidies benefit mainly the richer segments of the population [
5]. Ethiopia has laid out ambitious GHG mitigation commitments in its NDC (The text of Ethiopia’s NDC can be accessed at
https://www4.unfccc.int/sites/ndcstaging/PublishedDocuments/Ethiopia%20First/INDC-Ethiopia-100615.pdf). Those commitments in turn are based on its Climate Resilient Green Economy (CRGE) Strategy [
6].
Figure 1 summarizes the approach to national GHG mitigation in the CRGE Strategy. The country aims to limit its net annual greenhouse gas (GHG) emissions in 2030 to 145 Mt CO2e or lower, which would constitute a 255 Mt CO2e (64 percent) reduction from projected BAU emissions in that year. Ethiopia’s NDC commitments are
conditional, in the sense that they depend on receipt of technological and financial support. Ethiopia has estimated the need for US
$7.5 billion in expenditures annually until 2030 on initiatives that will contribute to GHG emissions reductions while also supporting economic advancement [
6].
To assess the various impacts of carbon pricing in Ethiopia, we use a Computable General Equilibrium (CGE) model to analyze various policy scenarios using a carbon tax. We focus on scenarios that apply the carbon tax to petroleum fuels and kerosene, for reasons we explain below. More specifically, we analyze the implication of a price that starts at $5 per ton of CO2 in 2018 and rises to $30 per ton in 2030. It is worth to note that direct emissions from electricity production in Ethiopia are minimal because of reliance on hydroelectricity. There are also efforts underway to introduce intermittent sources of renewable electricity, such as solar, whose costs have been declining significantly over time.
The policy scenarios differ in terms of what is done with the revenue generated through the carbon tax. Previous research shows that a carbon tax can raise considerable, and that the economic impacts of a carbon tax depend significantly on what is done with that revenue. In particular, “recycling” revenue by cutting other taxes can soften negative effects of the carbon tax on overall economic activity and on the distribution of income. Accordingly, we study the differences in impacts of the carbon tax with revenue recycling through a uniform sales tax reduction, reduction of labor income tax, reduction of business income tax, direct transfer back to households, and use by government to reduce debt.
One obvious question is the rationale for focusing on a carbon tax applied only to petroleum fuels and kerosene. GHG emissions from livestock and from land use and forest change, including unsustainable harvesting of fuelwood, represent 93 percent of Ethiopia’s national emissions (see
Figure 2 below). Petroleum fuels and kerosene, on the other hand, account for less than 6.5 percent of national emissions.
A carbon tax to fossil fuels is easy to design and implement, based on the amounts of GHG emitted per unit of energy of the different fuels. Carbon taxes also are relatively easy to implement, since most countries (including Ethiopia) already apply excise taxes to energy. Moreover, fuel taxes are less prone to problems of tax evasion than other forms of taxes [
7,
8]. In contrast, pricing carbon emissions that occur as a consequence of land-use change and agriculture is more difficult technically and politically, so we do not consider such options in this paper.
In principle, it would be possible to track forest removal on different patches of land and levy a fee on the landholder based on an estimate of the resulting emissions. In practice this would be difficult to enforce, especially for land clearing in the informal sector on land without clear title. Similarly, it would be possible to tax extracted timber to the extent that the resulting loss of carbon sequestration was not offset by replanting. However, calculating the net emissions from timber harvesting is quite complicated, especially if different land holdings have different rates of harvest and replanting. For this reason, the focus in mitigating GHG emissions from land clearing and forestry has been on “payment for environmental services” initiatives like REDD, and in reducing other tax distortions that give advantages to land clearing for agriculture [
9]. Finally, while it would be possible to tax food products based on relative GHG emissions (including methane emissions from beef cattle, and emissions based on fertilizing practices), this too would give rise to technical and political problems. Nevertheless, it is clear that for many countries in addition to Ethiopia, there is need for additional well-functioning incentive policies to curb emissions from sources other than fossil fuels. This remains an important topic for further research.
One other important caveat concerns the CGE model used in this study for evaluating different carbon tax policy options. A key finding in our analysis is that the impacts of the assumed carbon tax on GDP and its growth are negative, though quite small, across the policy scenarios. The model we use is very much representative of current practice for this kind of analysis. Nevertheless, there are some important limitations of the model for capturing fully the impacts of a carbon tax. Specifically, the model does not explicitly differentiate between employment and output from informal and formal sectors of the economy. The model also does not include the possibility of lasting structural unemployment for part of the labor force.
Incorporating these factors can in some cases reduce or even eliminate the negative GDP effects of a carbon tax. However, the methods for accomplishing this are still very much a work in progress. We return to these points in the concluding section of the paper.
In
Section 2 of the paper we provide additional information about the Ethiopian economy and its sources of GHG emissions.
Section 3 through
Section 5 describe the CGE model, the policy scenarios, and the results of the simulation analyses. Concluding remarks are in
Section 6.
5. Modeling Results
5.1. Impact on Emissions
Figure 4 and
Figure 5 present summary results of changes in emissions across the different scenarios. The figures show that, as expected, emission reductions increase over time as the price on carbon increases. Depending on the recycling scenario, the reduction ranges from 1.1 million tons to 1.5 million tons relative to BAU in 2030, when the carbon tax reaches its maximum value. The differences across scenarios are due to the differences in growth of the economy due to the different revenue recycling mechanisms considered, which in turn influence the demands for fossil fuels.
Comparing across the scenarios, the least emission reduction is achieved in the sales tax (STAX) reduction scenario. This is because its effect on the general economy results in the least growth impact. Direct tax reduction (DTAX), on the other hand, leads to more significant reductions in emissions.
Similar comparisons follow when we compare the resulting reductions in emissions from all emission sources. These will be larger than the reductions in emissions in fossil fuels, because the changes in fuel prices will have effects on the quantities of other emissions-producing activities in the economy. By 2030, as compared to the base case, the STAX reduction scenario will lead to a reduction of total emission by 1.7 million tons while the DTAX reduction scenario will lead to a reduction of total emission by about 3.8 million tons (see
Figure 6 and
Figure 7).
5.2. Impacts on the Economy
The assumed time path of carbon prices has a fairly modest effect on the growth of GDP (
Figure 8 and
Figure 9). In BAU, GDP is much higher in 2030 than in 2015 given the assumed growth rates for the economy. The same is true in the policy scenarios; the different tax and recycling combinations only slow growth by a little over the study period. GDP in 2030 is about 1 percent lower compared to BAU in almost all scenarios. Average GDP growth is from 0.38 to 0.52 percentage points lower with the carbon pricing than under BAU, depending on the chosen scenario. This is a small decrease in the growth rate of 7.6 percent per year in BAU.
A cut in either sales or corporate tax leads to smaller negative effects on GDP compared to the other policy scenarios. The reason is that lower sales tax reduces cost of goods and services and stimulates demand across income groups. Lower business tax reduces the cost of capital which encourages investment and therefore employment and production. Both tax reductions end up reducing existing distortions in the economy. That, in turn, will lessen the amount of emission reduction from the carbon tax.
The scenario in which personal income tax is reduced leads to a larger decrease in the growth of economic activity. The reason is that professional and educated labor—the types of workers paying the personal income tax—is assumed to be fully employed. Reducing income tax therefore increases neither labor supply nor economic activity.
5.3. Fiscal Impacts
Because the government account should balance, and government expenditures are treated as fixed in the analysis, the government recycles the entire change in revenues in each of the five policy scenarios. To assess the amount of revenue raised through carbon pricing in each scenario, we first calculate the amount of fuel imported each year. We then multiply the fuel import by the emission factor (i.e., emission per unit of fuel) to arrive at a figure for total emissions, and then use the assumed carbon tax rate to calculate total carbon tax revenue.
The revenue generated in 2030 ranges from
$786 million under the direct tax scenario to
$798 million under the sales tax reduction scenario (
Figure 10). The differences in revenue among scenarios are due to the differences in economic growth which, in turn, affect the demand for fuel.
5.4. Distributional Impacts
Although the impact of the carbon tax on overall economic growth is one indication of the economic consequences of the policy, a much clearer understanding of the full implications for economic well-being comes from looking at the impacts across heterogeneous households based on factors such as their location (urban, rural) and their abilities (skilled, unskilled). Households receive income from wages (employment), returns on land for which they have the right of use, returns on capital they own, and income transfers. The implication of a carbon tax on a household’s welfare, therefore, depends on how it affects each of these sources of household income.
In this subsection, we highlight the heterogeneous impacts on labor demand (employment) and household consumption.
Figure 11 and
Figure 12 provide the difference in employment trend of low-skilled individuals located in the urban and rural sector. The carbon tax on fuels increases the price of goods reliant on fuel-using transport. As a result, it shifts demand away from those goods towards others. Sectors that are transport and thus fuel dependent include the service and manufacturing sectors. Both sectors are overwhelmingly concentrated in urban areas. As a result, there will be a decrease in urban unemployment (
Figure 11) among low-education workers. Because we assume full employment for high-skill individuals, they adjust to the carbon tax through a wage decrease rather than an employment decrease.
The agriculture sector, on the other hand, does not rely as heavily on transportation. Therefore, it benefits from an increase in demand relative to other goods. This, in turn, increases the employment potential of rural unskilled labor compared to their urban counterparts.
Figure 12 indicates that there is almost no effect on employment across scenarios in rural areas compared to the base case scenario. (Note that in the base case scenario, employment of low-skilled individuals in urban areas is assumed to be 4,186,322 in 2018 and 6,290,527 in 2030. Similarly, employment of low-skilled individuals in rural areas is assumed to be 38,516,966 in 2018 and 54,301,218 in 2030.)
The tax has differing impacts on the consumption of the lowest quintile households in urban and rural areas. In urban areas (
Figure 13), the main source of income for poor households is employment income (wages). As the sectors that poor households work in (services and industry) are affected more by the carbon tax, their income decreases due to either increased unemployment or a decrease in wage. The implication for economic welfare depends on the scenario: revenue recycling under STAX and CORPOR leads to smaller decline in consumption than others because there is less impact on the economy. TRANS, on the other hand, limits the decline in consumption by transferring resources to poorer households.
In rural areas, the ‘almost non-existent’ effect of the carbon tax on agriculture means that employment is not really affected (
Figure 14). In addition, the tax slightly increases the return to land owned by rural households. The combination of these two effects implies that the impact of the carbon tax on the consumption of poor rural households is minimal. It is worth noting that, ex ante, one would expect that the poor will fare better under the transfer scenario. However, the result shows the opposite. There are a couple of reasons for this. First, the transfer scenario does not restrict transfers to poor households. Since a large share of transfers are made to urban/richer households, it limits the extent to which the aggregate transfer can reduce poverty and inequality. Second, the slowdown in economic activity under the transfer scenario introduces unemployment/wage reduction that affects welfare directly. Another way to understand the result under the transfer scenario is that one needs to restrict the transfer to poor households in order to achieve a meaningful reduction in poverty and inequality.
6. Concluding Remarks
Over 90 percent of Ethiopia’s GHG emissions comes from sources other than fossil fuels. This is because (a) energy use in general is relatively low, including in transport, although it is expected to grow as the economy expands; (b) most of Ethiopia’s electricity comes from hydro; and (c) biomass remains the overwhelming choice of energy source for cooking. Consequently, application of carbon pricing to fossil fuels use in Ethiopia necessarily will have a somewhat limited effect on total GHG emissions.
Nevertheless, since motorization increases as incomes rise, the carbon price on fuels can contribute to mitigating a “lock-in” of high levels of individual vehicle use and high demand for road expansion by limiting vehicle use and creating support for expanding well-performing public transit and smaller, efficient cars in more densely populated areas. The carbon price on fossil fuels limits fuel use and associated GHG emissions cost effectively compared to what might result from a patchwork of different regulatory standards on various emissions sources. It creates incentives for increasing energy efficiency including in transport choices. By moving away from carbon, Ethiopia may benefit in the long run from energy efficiency and relatively cheaper sources of energy, providing a competitive advantage over those countries that did not make the transition and have locked-in inefficient technologies. In addition, revenue-neutral reductions in other taxes can be made to spur investment and productivity gains.
Regarding economic impacts, GDP continues to grow substantially, albeit at a modestly lower rate. The direct effect of the carbon price is likely to be felt more by higher-income households since they are more intensive consumers of fossil fuels. Indirect effects through adjustments in the economy to higher fuel prices can lead to modestly slower growth for the urban poor, but the size of that effect will depend on how carbon revenues are recycled. Because the rural poor are not intensive users of fossil fuels, and the economic adjustments to carbon pricing may have limited impact on the agriculture sector, the rural poor are likely to face little impact from it. As we have shown, some portion of carbon tax revenues can be used to soften impacts on the poor.
A significant impact for Ethiopia of applying a carbon price to fossil fuels is that substantial revenues can be generated—up to $800 million per year by 2030. Part of the revenue could be used to finance other carbon mitigation activities, with a focus on those activities that have high societal co-benefits that strengthen the rationale for such public expenditures. Three such types of expenditures seem to stand out:
Increase reforestation activities over and beyond what might be financed internationally through the country’s REDD program. Investments in forest recovery can provide important ecosystem benefits, including soil and watershed protection and habitat for valued species, as well as expended carbon storage. This use of revenues is well aligned with the pillars of the CRGE [
6].
Provide technology-neutral subsidies to increase affordability of improved cookstoves that use biomass fuel more efficiently or not at all, thereby reducing time spent collecting fuelwood and the substantial adverse health impacts of indoor smoke, especially for women and girls. Even if affordable alternatives to fuelwood for cooking take time to scale up in rural areas, cookstoves with improved fuel efficiency and improved ventilation can generate some improvement of indoor air quality, while reducing the level of unsustainable fuelwood harvest and the production of black carbon. Increasing access to cleaner cooking also is a pillar of CRGE [
6].
Find ways to increase the efficiency of fossil fuel use in urban transport, thereby slowing its growth and the corresponding increase in GHG emissions. This could be done through investment in more fuel-efficient and less-polluting multiple-rider transit vehicles in urban areas, thereby mitigating another major public health challenge. It would also be beneficial to use a portion of carbon pricing revenues to increase oversight of fuel quality.
As noted, GHG emissions from fossil fuel combustion are only a limited part of the story in Ethiopia. Other significant emissions sources include land clearing for expanding agriculture, and unsustainable consumption of wood-based fuels for cooking and heating. To provide a complete picture of how carbon pricing and other policies can manage Ethiopia’s emissions in accordance with its national commitments, it is necessary to go beyond the scope of this study to consider what combinations of pricing and other measures can be effective in limiting emissions from the agriculture and forest sectors, including fuelwood in the latter. Using price-based measures to restrict land clearing and fuelwood use is challenging, though not impossible. Further research can address land taxes that encourage land rehabilitation, reforestation, and maintenance of natural forests, as well as land clearing taxes. Another fruitful avenue for investigation is how changes in current industrial development and import substitution policies could lower the cost of access to more energy-efficient and less-polluting technology and capital.
Finally, the model does not explicitly differentiate between employment and output from informal and formal sectors of the economy. The model also does not include the possibility of lasting structural unemployment for part of the labor force. Both of these considerations can substantially affect the conclusions of a carbon tax impact analysis.
Although the importance of these issues has been noted for some time, the practical capability to deal with them using larger-scale CGE models is still developing. Markandya, González-Eguino, and Escapa [
12] developed a CGE model for Spain including informal employment. Mirhosseini, Mahmoudi, and Valokolaie [
13] use a variant of that model for Iran. Bento, Jacobsen, and Liu [
14] apply a much more aggregated model with informal employment and different types of energy usage in the informal sector to China and India. These studies indicate that a significant informal sector can lead to lower adverse impacts on the economy of carbon pricing. However, all three of the studies use static, long-run equilibrium models that by their nature cannot capture the effects of ongoing adjustment over time to a potentially changing carbon tax.
Employment impacts of a carbon tax depend on the types and extent of pre-existing distortions in labor markets, how the carbon tax may alter the sectoral content of GDP, and how recycling of carbon tax revenues may affect the after-tax cost of hiring for firms. Markandya [
15] provides a very useful review of these considerations. Labor market distortions can be the result of wage rigidity, labor market segmentation, or search costs [
16,
17,
18,
19]. As with the inclusion of informality, methods for including labor market structural rigidities in CGE models are still evolving. Relaxing the limitations noted here with a more advanced CGE model would represent a valuable direction for follow-up work.