Distributional impacts of carbon pricing in developing Asia

Distributional impacts of carbon pricing in developing Asia

Jan C. Steckel, Ira I. Dorband, Lorenzo Montrone, Hauke Ward, Leonard Missbach, Fabian Hafner, Michael Jakob & Sebastian Renner
Nature Sustainability (2021)
Published: 23 September 2021
Mercator Research Institute on Global Commons and Climate Change, Berlin, Germany
Department Economics of Climate Change, Technische Universität Berlin, Berlin, Germany
https://www.nature.com/articles/s41893-021-00758-8

Abstract

Understanding who would be affected in which way by carbon pricing is pivotal for effective and socially equitable policy design, addressing climate change and reducing inequality. This paper focuses on eight key countries in developing Asia (Bangladesh, India, Indonesia, Pakistan, Philippines, Thailand, Turkey and Vietnam). By combining national household surveys with input–output data, we compare the distributional effects of four carbon pricing design options, including a globally harmonized carbon price, a national carbon price and sectoral carbon prices in the power and transport sectors, respectively. Our analysis reveals a substantial degree of variation regarding who would be affected across policy designs and countries. Looking into national carbon pricing as the most favourable policy option from an economic point of view, we find that differences in distributional outcomes are generally more pronounced within income groups than across income groups. These differences are mainly driven by households’ energy use patterns, which vary across countries. Equally recycling revenues back to all citizens would overcompensate the burden of a carbon price for the poorest households in all countries.

Main

To achieve the Paris goals, climate policies will increasingly need to be implemented in low- and middle-income countries. A growing number of countries consider carbon pricing as a means to achieve their emissions targets, including in Asia. The region is important with respect to global climate mitigation efforts as its high energy demand growth has mainly been fuelled by carbon-intensive coal in recent years. Across the globe, >500 GW of coal power is currently under construction or planned, most of it (>89%), located in developing Asia. Once built the planned coal capacity would seriously jeopardize the Paris climate targets. Even though it is too early to know how COVID-19 will affect further investment plans, investments in coal can be expected to continue if recovery packages are not increasingly targeted at green investments.

Economy-wide carbon pricing is generally seen as the economically most efficient policy to reduce greenhouse gas emissions. A sufficiently high carbon price would substantially reduce incentives to invest in new coal-fired power plants and could make clean forms of electricity generation, such as renewables, economically competitive. Public support for carbon pricing depends on its distributional effect, that is which parts of the population are affected in which way by the policy. Therefore, policies that directly increase the prices of fossil fuels can be contentious, as demonstrated by protests in France in late 2018 (after an increase of carbon taxes on fuels) as well as in Ecuador in late 2019 (following proposals by the government to cut fossil fuel subsidies).

In this paper, we contribute to the understanding of distributional implications of carbon pricing in eight Asian countries which are among those that currently invest most heavily in coal, namely Bangladesh, India, Indonesia, Pakistan, Philippines, Thailand, Turkey and Vietnam. They host 16% of the current global coal capacity and 37% of the global coal pipeline (73% of the pipeline outside China); that is, what is currently under construction and planned (Extended Data Table 1). It is hence highly unlikely that international climate policy will succeed without those countries undertaking efforts to reduce their emissions and avoid future lock-ins.

At the same time, these countries face substantial challenges to promote human development objectives in line with the agenda of the Sustainable Development Goals (SDGs). Some previous studies have analysed the relationship between climate change mitigation (SDG13) and socio-economic development, in particular energy access (SDG7) and the health benefits of reducing ambient air pollution due to fossil fuel combustion (SDG3). Our study, by contrast, focuses on the distribution of income (SDG10) in a cross-country perspective on the basis of detailed accounts of household carbon footprints derived on the basis of a unified methodology. It hence differs from previous studies that have either looked at individual countries or used a cross-country approach based on highly aggregated data. Some comparable analyses have been undertaken for Latin America as well as the EU but to our knowledge not for Asia. Our paper extends the geographical coverage and provides a detailed analysis of horizontal equity and the distributional effect of different approaches to carbon pricing that were hitherto restricted to single-country studies.

Our analyses answer three questions. First, we compare the distributional impacts of different design options for carbon pricing, including an international harmonized carbon price, national economy-wide carbon pricing as well as sectoral carbon prices that only apply to the electricity or transport sector, respectively. Second, we compare the vertical and horizontal distributional effects (across and within income groups) of a national carbon price of US$40 per tCO2 (which is regarded as the lower bound required to achieve the targets of the Paris Agreement). Note that we focus on a national carbon price as the most efficient option from an economic theory point of view. Third, we analyse why the distributional incidences of national carbon prices differ across countries as a result of differences in spending patterns for different consumption categories.

There is a large literature on distributional implications of climate- and energy-policies across income groups, using different methods and datasets (see Ohlendorf et al.¹ for a review). Previous studies have, for instance, assessed the distributional impacts of fuel taxes, fossil fuel subsidy reform and carbon pricing from a cross-country perspective. Recent literature has also taken into account ‘horizontal’ distributional effects within income groups. Our paper provides a cross-country, cross-instrument analysis with highly detailed and representative household data that examines distributional implications across as well as within income groups and sheds light on underlying drivers of variation.

To understand how different household types would be impacted, we perform microsimulations on the basis of representative household survey data and multiregional input–output (MRIO) data. Survey data are collected from the countries’ statistical agencies, harmonized and matched with an environmentally extended MRIO model (Methods and Supplementary Information provide details). Combining both datasets allows us to identify the carbon footprint of households. To assess the distributional impacts of a carbon price, we derive the additional expenditure they would require to maintain their initial consumption. Throughout this analysis we proxy household income by expenditures. Our analysis accounts for the effects of higher costs for direct energy use as well as rising prices for goods and services that use energy as inputs in their production (Supplementary Fig.10 gives a decomposition of those effects). Despite some methodological limitations regarding the matching of household surveys to MRIO data and issues related to data availability, our results appear to be robust under a broad set of alternative specifications (Supplementary Information).

Comparing different design options

From a first best point of view, a globally homogenized carbon price would be economically optimal. Arguably, given the political realities, this is unlikely to unfold anytime soon. Yet, some countries have proposed national economy-wide carbon pricing schemes or schemes that target specific sectors. In China, for example, the recently launched emission trading scheme is focusing on the power sector. We hence compare the distributional effects of four different design options for carbon pricing: an internationally harmonized carbon price, a national carbon price, a carbon price that focuses on the electricity sector and a carbon price on liquid fuels, covering mainly the transport sector. We assign households to quintiles on the basis of total per capita expenditures and normalize the median distributional incidences by quintile to the first one (the poorest 20% of the population) to make results comparable across countries and instruments. The shape of distributional incidences therefore is independent of the absolute magnitude—and hence of the level of the carbon price. Instead, it emphasizes whether richer households are more or less strongly affected than poor households (relative to their income).

why there is only one label on the first quintile showing the incidence while there are four instruments(four lines) in the figure?

Figure 1 reveals a large heterogeneity of distributional effects across countries and policies. Pakistan and the Philippines are the only two countries in which all designs would lead to progressive outcomes. In Bangladesh and Indonesia, all design options but carbon pricing only for liquid fuels would be progressive. Generally, a carbon price on liquid fuels seems to follow slightly different distributional patterns than other design options; also in Turkey and India (where it is the only progressive policy) and in Pakistan (where it is substantially more progressive than other policies). In Thailand, where most policies would be regressive, we find a hump-backed shape with the highest impacts on middle class households. A national economy-wide carbon price would be progressive in five countries, neutral in one and mixed or regressive in two (Thailand and Turkey). Hence, whether a certain policy is progressive or regressive, as well as the ranking of different policy instruments with regard to their distributional effects, depends on the specific country context.

Dots refer to the average incidence in each household quintile for: (1) a global carbon price (red), (2) a national carbon price (light blue), (3) an electricity sector carbon price (green), and (4) a liquid fuel carbon price (purple) normalized to the average incidence of the first quintile. A value of, for example, 1.2 would imply a 20% higher median effect on households of expenditure quintile i relative to the median effect of the first expenditure quintile. Note that no revenue recycling is assumed (Fig. 5 and Supplementary Table 18 show the effects of revenue recycling). Label at first quintile shows the median incidence for a national carbon price (Supplementary Table 10). Bars display the amount of covered CO2 emissions for each instrument, expressed as a share of global emissions embedded in national consumption. Labels display total levels of global emissions embedded in national consumption.

The analysis in Fig. 1 also allows to measure differences between instruments in terms of their progressivity and the share of households’ carbon footprints that would be covered. For example, while a tax on liquid fuels would be highly progressive in India, it would only cover <20% of the country’s emissions.

Absolute effects of a national carbon price

Arguably, it is also important how households are affected in absolute terms, which is particularly relevant when concerned about the political acceptability of a price reform. Absolute effects vary considerably between the poorest quintiles across countries. In Bangladesh, Pakistan and the Philippines, a US$40 per tCO2 national carbon price would impact poorest households by ~1%; in India and Thailand, it would be >4% of their expenditures.

Figure 2 zooms into the absolute effects of a national US$40 carbon price. Note that Supplementary Tables 15–17 also indicate results for the other instruments discussed above, including international carbon price, transport sector carbon price and electricity sector carbon price. Due to the linearity of the input–output system, these effects can be generalized for different carbon prices by proportional scaling. Despite progressive or neutral distributional effects for most of the countries in the sample, the absolute effects of a carbon price of this magnitude would be substantial. Median values for the poorest quintile are identical to the values indicated in Fig. 1. For example, in India, poor households would need to increase their expenditures by on average 4.5% to maintain their current consumption patterns, while 25% of poor households would even be affected by >5%. The major driving factor for these large welfare impacts in India is the relatively carbon-intensive agricultural sector, mostly regarding the production of rice, wheat, grains, fruits and other crops (Supplementary Table 8), in combination with high expenditure shares for food by Indian households.

Each smoothed density curve refers to the distribution of cost burden as percentage of household expenditures for a US$40 per tCO2 carbon price (x axis). The y axis displays the share of households within each quintile. Curves are fitted over binned incidence levels with Δx=0.1%. Colours and line styles refer to expenditure quintiles and the dots correspond to median values in each quintile. Area in grey displays ΔV, the difference in medians between the most- and least-affected expenditure quintile at the median (Table 1). When any given expenditure quintile is more affected, its curve is skewed more to the right². Households within expenditure quintiles are more heterogeneously affected, if curves are more widespread. Cumulative densities sum up to 100%.

The distributions in Fig. 2 already indicate that the horizontal effect (within quintiles) is more disperse than the vertical effect between quintiles. Table 1 offers a detailed comparison of vertical and horizontal distributional effects. It compares the spread (in percentage points) between the most- and the least-affected quintile to each quintiles’ spread between the 20th and the 80th percentile. Note that we also apply additional measures to quantify the horizontal equity effects in Supplementary Tables 12–14. The difference between the median incidence of the most- and the least-affected quintile (ΔV) is particularly small in India and Indonesia (both ~0.4%), while for the other countries it is slightly larger. However, these differences are small compared to the variation within quintiles. For example, India and Indonesia exhibit 1.4 and a 2.6 percentage points differences, respectively, between the most and the least affected 20% in the first quintile (ΔH1). Compared to vertical effects, the horizontal effects by quintile are 3.7–5.3 times larger for India and 4.8–6.1 times larger for Indonesia (comparison column, Table 1). Thailand and Turkey also display large discrepancies between vertical and horizontal effects. While the exact values on comparing vertical to horizontal effects are subject to the specific method chosen to calculate horizontal effects, the key result (that horizontal effects are more pronounced than vertical ones) also holds when applying different methodologies (Supplementary Tables 12–14).