The distributional effects of emissions taxation in Brazil and their implications for climate policy

    The Distributional Effects of Emissions Taxation In Brazil and Their Implications for Climate Policy Lucio Flavio da Silva Freitas, Luiz Carlos de Santana Ribeiro, Kˆenia Barreiro de Souza, Geoffrey John Dennis Hewings PII: DOI: Reference:

S0140-9883(16)30193-1 doi: 10.1016/j.eneco.2016.07.021 ENEECO 3398

To appear in:

Energy Economics

Received date: Revised date: Accepted date:

22 April 2015 20 June 2016 24 July 2016

Please cite this article as: da Silva Freitas, Lucio Flavio, de Santana Ribeiro, Luiz Carlos, de Souza, Kˆenia Barreiro, Hewings, Geoffrey John Dennis, The Distributional Effects of Emissions Taxation In Brazil and Their Implications for Climate Policy, Energy Economics (2016), doi: 10.1016/j.eneco.2016.07.021

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT THE DISTRIBUTIONAL EFFECTS OF EMISSIONS TAXATION IN BRAZIL AND THEIR IMPLICATIONS FOR CLIMATE

T

POLICY

SC R

a

IP

Lucio Flavio da Silva Freitasa; (Family name: Silva Freitas).

City University of São Caetano do Sul. Address: Av. Goiás, 3.400, São Caetano do Sul - SP, Brazil. Zip code:

MA

NU

09550-051. Email: [email protected]

D

Luiz Carlos de Santana Ribeirob; (Family name: Santana Ribeiro).

b

TE

Assistant Professor at Federal University of Sergipe - Brazil. Address: Av. Marechal Rondon, Jardim Rosa

CE P

Elze– São Cristóvão/SE –Brazil. Zip code 49100-000. Email: [email protected]

c

AC

Kênia Barreiro de Souzac (Family nam: Souza).

Researcher at Federal University of Minas Gerais - Brazil. Address: Av. Antônio Carlos, 6627, Belo

Horizonte, MG – Brazil. Zip code 31270-901. Email: [email protected].

Geoffrey John Dennis Hewingsd; (Family name: Hewings)

d

Regional Economics Applications Laboratory. Address: 607 S. Mattews, Room 318, M/C 151. Urbana, IL –

United States. Email: [email protected].

ACCEPTED MANUSCRIPT 2

THE DISTRIBUTIONAL EFFECTS OF EMISSIONS TAXATION IN BRAZIL AND THEIR IMPLICATIONS FOR CLIMATE POLICY

T

ABSTRACT

MA

NU

SC R

IP

The emission of greenhouse gases (GHG) generated by human activity is a major cause of global warming and climate change. There is considerable debate about the choice of the best mechanism to reduce emissions under a climate policy. The aim of this paper is to measure the impact of a policy of taxing GHG emissions on the Brazilian economy as a whole and on different household groups based on income levels in 2009. The following databases were used: Supply and Use Tables, Household Budget Survey, National Household Sample Survey and emissions data from the Brazilian Ministry of Science and Technology and Innovation. A price system from a national input-output model that incorporates the intensity of GHG emissions is used, as well as a consumption vector broken down into ten representative households with different income levels. The main results indicate that this taxation system was slightly regressive and had a small negative impact on output. There were, however, significant emissions reductions.

D

Keywords: Emissions; taxation; income distribution; input-output.

1 Introduction

CE P

TE

JEL CODES: C67; Q52; Q54.

Greenhouse gas (GHG) emissions generated by human activity are one of the main causes of global

AC

warming and climate changes. Magalhães and Domingues (2013) argue that the rise in average temperature observed since the mid-twentieth century has been largely caused by the concentration of GHG in the atmosphere. According to these authors, Brazil is very vulnerable to the negative impacts of climate change. In the last fifty years, this country has experienced an approximate rise of 0.7º C in mean annual temperatures, which is higher than the most optimistic estimates of the global average rise of 0.64ºC. Therefore, climate policies should be elaborated to address the prospect of continued increases. Developing countries have been under pressure from the international community regarding the creation of climate policies due in part to their rapid GDP and emissions' growth (Rong, 2010). It is expected that CO2 emissions in these countries will generate more than half of global emissions in 2030, even though, in per capita terms, developed countries still occupy the first positions (Bosetti and Buchner, 2009; IEA, 2009).

ACCEPTED MANUSCRIPT 3

According to the World Input-Output Database (WIOD), Brazil, China, India and Russia together contributed 39% to global GHG emissions in 2009. It is important to highlight that Chinese emissions alone accounted for 24% the same year. On the other hand, Brazilian emissions were

T

approximately 2.4% (Genty et al., 2012; Timmer et al., 2015).

IP

The Brazilian Nationally Determined Contributions (NDC), presented at 2015 United Nations Climate Change Conference, held in Paris, is designed to reduce Brazilian GHG emissions by 43%

SC R

below 2005 levels in 2030. It aims to move towards zero illegal deforestation by 2030 and compensate for greenhouse gas emissions from the legal suppression of vegetation by 2030 (BRAZIL, 2015, p. 7). While NDCs promote deforestation control, they also aim to maintain the

NU

share of renewable sources in the energy mix, and promote the restoration and reforestation of 12 million hectares of degraded land (BRAZIL, 2015).

MA

From 2005 to 2012, Brazil reduced its total emissions by 12%. Nevertheless, total GHG emissions excluding land-use change and forestry have increased 18% (WRI, 2016), while per capita GPD rose by 17%. Brazilian climate policy must go beyond the control of deforestation to avoid an

D

unsustainable pattern of development. Otherwise, in the long run Brazilian emissions could reach a

TE

similar level of developed countries today.

NDCs are committed to maintaining Brazilian per capita emissions close to the global per capita

CE P

carbon budget, around 1.4 ton/per capita, to avoid the most dangerous climatic scenarios (BRAZIL, 2015). This proposal, combined with an annual economic growth of 3% from 2012 to 2050, requires an economy that is eight times less intensive in producing GHG emissions with zero net

AC

emissions from land use change and forestry. There is still considerable debate regarding the choice of the best mechanism to reduce emissions as part of this climate policy. Among points of contention, we can highlight: market mechanisms, subsidies, taxes, government regulations, carbon trade, carbon tax and cap and trade. However, a topic rarely addressed in the literature, especially in the Brazilian economy, is the impact that the adoption of such a policy would have on the economy, industries and emissions (Magalhães and Domingues, 2013). More specifically, what would the distributional impact of an emissions taxing policy in Brazil especially given the pronounced levels of income inequality that exist in the country? The distributive issue associated with GHG emissions charges was discussed by Symons et al. (1994). For these authors, even a neutral tax reform would bring significant emission reductions in the United Kingdom, but with adverse distributional effects. Tiezzi (2005), on the other hand,

ACCEPTED MANUSCRIPT 4

discussed the charge introduced in Italy in 1999. The regressivity impact of this tax, however, has not been confirmed. In developing countries, these kinds of studies are rare. For instance, Gonzalez (2012) tested

T

different ways to generate revenue through a tax in Mexico. Subsidies on food caused a more

IP

progressive distribution while a compensatory reduction in taxes on manufactured goods had a regressive outcome. In Russia, the best result, both in terms of environmental and economic

SC R

efficiency, was obtained from taxing labor (Orlov and Grethe, 2012). Given the oligopoly structure of the energy market, carbon taxation implies production decreases and induces mark-ups on some energy-intensive sectors. It is important to highlight that emissions from fossil fuel consumption in

NU

these two countries are high.

In Brazil, Magalhães and Domingues (2013) used a Computable General Equilibrium (CGE) model

MA

to endogenously determine the carbon prices necessary to achieve different emission targets. In the short-run, the tax was regressive and the carbon price higher. Tourinho et al. (2003) simulated charges on CO2 from the burning of fossil fuels using a Brazilian environmental CGE model. As

D

expected, the model showed the displacement of resources in GHG intensive industries to less

TE

intensive sectors. As a result, total investment has increased. Both papers found a small and negative impact on household income, output and emission levels. In turn, Gurgel and Paltsev

CE P

(2014), using a dynamic CGE model, evaluated the impact of alternative policies to achieve voluntary targets recently adopted by Brazil and concluded, among other things, that direct emissions reduction from deforestation is the most cost-effective option. In the present paper, the

the results.

AC

idea is to test outcomes using an alternative model to the CGE, an input-output model, and compare

To sum up, the impacts of taxes vary according to the productive and demand structure in each country; therefore, they are not readily generalizable. For Brazil, the literature suggests a regressive effect for emissions tax, generated because the poorest households are the ones with the highest emissions coefficient per dollar of expenditure. Given the heterogeneity of emissions among different industries, multisectoral models such as input-output (hereafter, IO) and CGE are very suitable for measurement of climate policies impacts. Unlike CGE modeling, IO models are easier to operationalize. According to Rose (1995, p. 297): “The sectoral scheme of an IO table facilitates data collection, and its matrix representation facilitates data organization. The simplicity and transparency of this table are strengths rather than weaknesses.” The traditional formulation of IO models, however, treats the economy by considering only the demand side. To measure the impact of a taxing policy using this method, the

ACCEPTED MANUSCRIPT 5

supply-side IO models must be used. To do this, the literature suggests the price or Ghosh InputOutput models (Ghosh, 1958; Leontief, 1941; 1966; Miller and Blair, 2009). Taking into consideration the supply-side perspective, this paper aims to measure the impact of

T

GHG emissions taxing policy on the Brazilian economy as a whole and on households with

IP

different income levels. We developed a pricing system from a national input-output matrix that

representative households with different income levels.

SC R

incorporates the intensity of GHG emissions, as well as a consumption vector broken down into ten

In contrast to previous studies, Brazilian emissions were broken down into more sectors to provide a closer look at the different consumption patterns as well as calculations of the short-run effects on

NU

household income. This approach, when treated from a distributive perspective, becomes more relevant for Brazil as historically there are high levels of income inequality among individuals and

MA

regions. The income inequality among individuals is one of the highest in the world (Baer, 2007) although, it has decreased in the last fifteen years, as pointed out by Ferreira (2004), Hoffman

D

(2006) and Silveira Neto and Azzoni (2011; 2012).

The next section describes the methodological procedures followed. The third section shows the

TE

database and the descriptive statistics, followed by the main results and discussion. The last section

AC

2 Method

CE P

presents the main findings and policy directions.

The basic equation of the IO model, according to Miller and Blair (2009), can be expressed by equation (1):

xZ f

(1)

Where x is the total output, Z is the intermediate input matrix, and f is the final demand. Then, the technical coefficient matrix is given by:

A  Zxˆ 1

(2)

where each A  [aij ] shows the amount of input i used as an intermediate input in the output of industry j. Therefore, Leontief model's solution can be represented by equation 3.

x  ( I  A) 1 f

(3)

ACCEPTED MANUSCRIPT 6

where (I-A)-1 is the total impact or Leontief Inverse matrix. Emissions intensity (e) was calculated as the product of an emissions coefficient (m), i.e., the ratio between emissions of each economic sector and the total impact matrix on the economy: (4)

IP

T

e  m( I  A) 1

The dimension of vector e is 56x1 and represents emissions released during the production chain of

SC R

final goods.

Traditionally, in the input-output literature, two price models have been presented: the Ghosh (1958) and Leontief price models (1941, 1946).1 In this paper, we adopted the latter one, which

NU

assumes that variations in production costs are converted into price increases. Thus, price  x´ is

MA

equal to the sum of inputs cost to the value added v  components.

x´ i Axˆ  v

(5)

TE

D

1 Post-multiplying equation (5) by xˆ , yields:

i´ i A  vc

(6)



 p  L' v c

(7)

AC

CE P

1 If, L  ( I  A) and vc  vxˆ 1 , and also called i´ p , the price index for the base year is given by:

If a tax on the amount of emitted CO2 equivalent were charged in the productive sectors, the tax vector (T) would be defined as:

T ´ exˆ

(8)

1 Given    T ´ xˆ , and  the rate per ton of CO2 equivalent, R$50.2 Finally, the adjusted prices

vector ( ~ p ) is: ~ p  L' (v   )  ~ p

1

(9)

It is important to highlight, however, according to Miller and Blair (2009), that both models produce the same results. For different interpretations of Ghosh model, see Dietzenbacher (1997), Oosterhaven (1996) and Mesnard (2009). 2 The carbon price was chosen to allow comparison to the Brazilian literature. For instance, Magalhães and Domingues (2013) estimated an endogenous carbon price to meet a 5% CO2 emissions reductions in 2030, in that year the carbon price was estimated to be around R$ 50. Silva and Gurgel (2012) considered exogenous US$ 20 per tonne of carbon; the same price was simulated by Tourinho et al (2003). Considering the exchange rate and the inflation, US$ 20 is near R$ 50 in 2009.

ACCEPTED MANUSCRIPT 7

Following Gemechu et al. (2012), if the monetary values of sectoral output are held constant, before and after tax, then the sectoral output becomes:

 p 0 1 xj  ~ xj pj

T

(10)

The impact on the price index (  ) is given by:

(11)

NU

Where x1 is the vector of sectoral production after tax.

SC R

e1  mx1

IP

Total emissions after tax were calculated as:

56

   ~p j j

MA

j 1

(12)

D

where  j is the share which industry j production represents in the total output. The government

TE

revenue from the new tax was estimated as:

R  mx1

CE P

(13)

Assuming households maximize their utilities using a Leontief function, and their income and savings are unchanged, none of each representative household could afford the same basket of goods. Therefore, using price changes derived from the model, it is possible to calculate the income

AC

variation necessary to compensate households for the loss of welfare. Formally, household welfare change ( wk ) for decile k is the following:

 wk  ( cik * p j )  ( cik * ~ pj) i

i

(14)

where cik is the quantity consumed by decile k from industry i. Knowing that Wik are total payments made by industry i to the labor related to the k deciles, and assuming industries use of labor follows a constant share of production, the effect on labor income can be calculated as:

Wik 

Wik x j xj

(15)

ACCEPTED MANUSCRIPT 8

where x j  x j  x j , i.e., the sectoral change in production after tax. 1

0

On the other hand, the effect on total income is straightforward:

Wik  Wik  Wik   Wik

T

i

(16)

SC R

IP

i

3 Database

The input-output matrix used was calculated at basic prices from the Supply and Use Tables (SUT)

NU

of Brazilian Institute of Geography and Statistics (IBGE) base for the year 2009, according to the procedures described in Guilhoto and Sesso Filho (2005) using the hypothesis of "industry-based"

MA

technology (Miller and Blair, 2009).

To construct the emissions vector, the following gases were taken into account: carbon dioxide (CO2), methane (CH4) and nitrous oxide (N2O) measured in carbon equivalents. The data were

D

taken from Estimativas anuais de emissões de gases do efeito estufa no Brasil3 (MCTI, 2013).

CE P

directly to global warming.

TE

These pollutants together constitute the so-called greenhouse gases4 or GHG, which contribute

When deforestation slows, Brazilian emissions will start to follow economic cycles. In the Amazon, deforestation has fallen from 27,772 km2 to its lowest level of 4,656 km2, between 2004 and 2012 (INPE, 2014). Therefore, the Brazilian Panel on Climate Change estimates that emissions will rise

AC

again after 2021 due to growth in the energy and agribusiness sectors (PBMC, 2014). The estimates for Brazilian GHG emissions measured in CO2 equivalent between 1990 and 2010 are reproduced in Figure 1.

While land-use change and forestry (hereafter, LULUCF) emissions have changed widely, one can observe a continuous and stable growth in other GHGs releases in the atmosphere, increasing by 77% for the whole period. Given the erratic and seemingly detached behavior from the economic cycle, as well as the growing importance of other factors in the total emissions, the simulation did not include LULUCF.

3

Annual estimates of greenhouse gases emissions in Brazil (own translation). The steps to reconcile the surveyed sectors and the sectors in the input-output matrix are described in Annex 1. 4 Despite the Sulfur Hexafluoride (SF6), Chlorofluorocarbons (CFCs) and Hydrofluorocarbons (HFCs) being also considered greenhouse gases, according to Genty et al. (2012), they have a small impact on global warming.

ACCEPTED MANUSCRIPT 9

The household consumption breakdown into different income deciles was done using from data of Household Budget Survey (POF), while labor incomes were broken down according to data from the National Household Sample Survey (PNAD), both released by IBGE for 2009. For both surveys, household per capita income was used to split data into ten deciles. In order to be

T

consistent with the IO data, only the shares of consumption and labor income for each household

IP

were used to disaggregate consumption and labor income vectors, respectively. Data obtained from

SC R

PNAD shows clearly the income concentration in Brazil, as one can observe in table 1.

In 2009, the first decile had an average household income of R$ 61.67 per month, i.e., 10% of

NU

Brazilians received the equivalent of nearly 30 American dollars5 per person. This reaches R$ 2,684.09 for the 10% richest people. With regard to labor income, it corresponds to around 73% of

MA

total income on average, with all deciles having similar shares. Using the combination of income and emissions data, it is possible to observe the significant variation in emission levels among

D

household groups in 2009 as displayed in figure 2.

TE



The decrease in the emissions coefficient per dollar reflects changes in consumption patterns as

CE P

income rises. For that year (2009), emissions per household remained relatively stable until the higher income deciles when the consumption scale effect is more prominent. Each GHG series considered showed similar behaviors.

AC

As the concentration of consumer spending is higher than the concentration of emissions, a tax on GHG-intensive items would have a regressive effect on welfare as measured by consumption expenditure. The per capita household consumption ratio between the two highest income deciles and the two lower was 13.21. In the case of total emissions, the ratio was 4.09.

4 Results and Discussions The results reveal that the taxation policy is capable of achieving its main goal: to mitigate emissions. It was estimated that there would be a significant drop in the level of emissions, an approximate 9.1% reduction for the whole economy. The government's estimated revenue from the new tax reached a high value (2009) of R$ 37 billion, and the decrease in production was 1.54%. At

5

Using the average exchange rate for 2009, when one U.S. dollar was equivalent to 1.99 Brazilian reals.

ACCEPTED MANUSCRIPT 10

the sectoral level, emissions variations are heterogeneous, as shown in table 2 for the 21 most polluting sectors.

T

The most intensive CO2 sector is Livestock and fishing, accounting for both the largest absolute and

IP

percentage changes of emissions as a result of the introduction of the tax; a similar outcome was

SC R

found for Transport. The underlying hypothesis of a linear production function implies that any change in emissions leads to a decline in sectoral output, which was weighted by rising prices; the emission coefficients remained constant. Therefore, the variation in the amount of GHG released

NU

into the atmosphere is inversely related to the price change.

In the simulation exercise conducted here, this negative impact reached R$ 84.4 billion, or a 1.54% reduction in the total Brazilian production. Figure 3 indicates the sectors with the largest declines.

MA

The sectors displayed in figure 3 together account for 79.4% of the total effect. It can be noted that they include services closely related to household consumption, such as Transport, storage and mail,

D

Trade and Lodging and food services. Some other sectors with high emissions (as shown in table 2)

TE

also appear: Food and beverage, Agriculture, Livestock and fishing and Oil refining, for instance.

CE P

Even though the policy achieves its goal of reducing emissions with a relative small drop in production, sectoral changes in production suggests that sectors directly related to household consumption are the most affected. Therefore, given that a significant proportion of the spending of

AC

lower income households is on food and transport items, the result of the tax has the greatest impact on the consumption of lower level income households, and thus the distributional effects become even more important.

Implementing the taxation policy, ceteris paribus, implies purchasing power losses for consumers, which can firstly be measured by a general price index. The total estimated effect on the general price index is relatively small, 1.01%; nevertheless, Agriculture, food and beverage prices, and some CO2-intensive industries vary the most. According to the assumptions made about household consumption, the compensatory variation measures the income needed for each household with the same utility level after tax and associated price changes. Figure 4 shows the compensatory variation related to household consumption. Regarding the lower income deciles, the tax losses noted earlier generate a 3.09% decline in

ACCEPTED MANUSCRIPT 11

welfare, reaching 1.18% for the richest household. This clear regressive pattern is related to the differences in consumption and income among households. While, at the lower end of the income distribution, items such as food and transport account for a larger proportion of household spending, for upper income households, services are the major component. Consequently, as the most affected

SC R



IP

T

sectors are those related to basic consumption, the most affected households are the poorest.

Results for labor and total household income exhibit a similar pattern. Figure 5 shows the labor income and total income effects over the ten income deciles. Similar to the compensatory variation results (figure 4), the negative effect over income is greater at the beginning of the distribution. For

NU

the first decile, the decline is 3.72% for labor income, compared to 1.02% for the highest decile. The impact is clearly smoothed when looking at the total income, with a more homogeneous

MA

distribution across the households. This behavior can be partially explained by the fact that income from other sources other than labor is basically retirement pensions or social programs.6

D

We also calculate the Gini index7 related to the household income before and after tax. With the from 0.562 to 0.564.8

TE

taxation policy, we can see a marginal worsening in the inequality as the Gini coefficient increased

CE P

The results obtained here are similar to those found in the previous literature. For instance, similar assumptions on household consumption were used in Wier et al. (2001), Wier et al. (2005) and Kerkhof et al. (2008). Wier et al. (2001) evaluated the relationship between the consumption

AC

pattern of household types and their CO2 requirements. Through the combination of input-output tables energy flow matrices, CO2 emissions factors, and a national consumer survey, the authors found that national differences in population density and climate cause differences in the contribution to CO2 emissions. Wier et al. (2005) found that CO2 taxes imposed on household’s energy consumption, as well as on industrial production, have undesirable distributional effects. The authors used a static input-output system combined with national consumer statistics for the Danish economy and also assumed that

6

It is worth mentioning that the "Bolsa família" program was created in 2004. Broadly speaking, the program goal was to transfer money to poor or extreme poor households. 7

G  1

k  n 1

(X k 0

k 1

 X k )(Yk 1  Yk ) . Where G is the Gini coefficient; X and Y are the cumulative proportion of

the variables" population" and "income", respectively. 8 The index was calculated using PNAD/2009 data on household income. For the index after tax, household income from labor was updated according to sectoral changes observed in the exercise. Therefore, it is worth nothing that the change in inequality here only captures changes in labor income between sectors, but cannot address eventual wage changes within sectors.

ACCEPTED MANUSCRIPT 12

the tax burden was fully imposed on consumer prices. Likewise, Kerkhof et al. (2008) showed that a GHG tax would be less regressive than a CO2 tax in the Netherlands. The authors combined consumer statistics with the GHG and CO2 intensities obtained from an input-output model.

T

On the other hand, using detailed demand systems combined with input-output data, Cornwell and

IP

Creedy (1996) and Labandeira and Labeaga (1999) and Kerkhof et al (2009) found similar results as well. Cornwell and Creedy (1996) simulated the inequality consequences of carbon taxation in

SC R

Australia. The authors allowed for substitution on consumption and on production processes. The model was calibrated to generate a CO2 emission reduction of the same magnitude proposed by the Toronto target, i.e., 20% of 1988 levels by 2005. They showed that most of the emissions reduction

NU

was due to substitution in production, and that the carbon tax would make the total tax revenue less progressive.

MA

In Labandeira and Labeaga (1999), the estimation of a demand system provided the consumers response to the new set of prices and the distributional effects of a taxation imposed on fossil fuels products. The tax burden was fully transmitted to consumers. Instead of taking the tax rate as

D

endogenously determined by an explicit target, the authors used the estimated damage per tonne of

TE

carbon, as valued by Fankhauser (1994). The results are a smaller tax rate than the one from an endogenous method, a modest environmental effect in short run, and non-regressively taxation.

CE P

Also, Kerkhof et al. (2009) evaluated the relationships between household expenditures and different environmental impact categories (climate change, acidification, eutrophication and smog

expenditures.

AC

formation). The household CO2 emissions increase less than proportionally with increasing

In comparison to the above literature, the model presented in section 2 incorporates taxation on GHG (or multigas taxation); the fully transfer of the tax burden to consumers; emissions reductions exclusively due to consumption changes; no price substitution effect on consumption or in production processes; an exogenous tax rate; and a combination of national consumer data to an environmentally augmented input-output price model. The regressive aspect of a tax on pollutant release was considered by Seroa da Motta (2002), who estimated the environmental pressure exerted by income brackets in Brazil. In the specific case of GHG, Silva and Gurgel (2012) claim that the tax impact on GDP is small in the long run. For Tourinho et al. (2003), the charge on fossil fuel CO2 emissions in 1998 generated small impacts for macroeconomic variables, even considering three different scenarios where the price of C0 2 tonnes varies between $3, $10 and $20. The most dramatic changes were in sectoral investment with the smaller emission-intensive sectors benefitting the most, as expected by the policy goals. However,

ACCEPTED MANUSCRIPT 13

the distributive aspects of emissions taxation was not dealt with in the studies of Seroa da Motta (2002), Tourinho et al. (2003) and Silva and Gurgel (2012). Magalhães and Domingues (2013) introduce the first results for Brazil involving both the impacts of

T

charges on GHG emissions and their distributive impacts. The authors concluded that even if the

IP

households are directly compensated with the resources from taxation, the final results are still regressive, or slightly progressive if the transfer is made directly to the poorest households. The

SC R

estimated cost for a 10% decrease in emissions was 1.26% of GDP through 2030. The result was claimed to be related to the energy matrix that is not very intensive on fossil fuels, and the high levels of emissions for food production are deeply related to poor households.

NU

In sum, the exercise indicates that: i) as the income concentration is greater than emissions concentration, the tax shows regressive initial effects, measured by a compensatory variation in

MA

household consumption; ii) when production changes, and consequently factor payments are taken into account, the regressive impacts are smaller, and they almost disappear when considering total income and not just income from work; iii) the effect on the price level and production was small,

D

because the highest tax was on livestock that this accounts for a small portion in terms of gross

TE

production value; iv) for the lowest income deciles, the largest portion of agriculture wages on overall income generated a regressive aspect in terms of total income.9

CE P

Therefore, there is a convergence of findings in the literature about the small change in GDP, and the regressive aspect of a tax on emissions. However, in this paper, we report an important reduction in GHG emissions concomitant with a not-so-large change in terms of income

AC

distribution, giving support for the policy application even in the short-run. Nevertheless, other aspects not considered here could possibly generate a different result. These include the implications on sectoral competitiveness, or the optimal tax value and its dynamic aspects. Other limitations are related to the method used. Production, emissions and labor coefficients were constant in the simulation, and it was assumed that households react to price changes by retaining fixed expenditure shares in their budget constraint.

9

For example, for the first decile 36.75% of labor income comes from activities related to Agriculture, silviculture and forestry.

ACCEPTED MANUSCRIPT 14

5 Conclusion and Policy Implications This paper has calculated the distributive impact on Brazilian households of a tax on GHG. In the short-term, there is a significant reduction in emissions, but an uneven distribution of its burden.

T

The poorest households are more affected in terms of their welfare. Because these households are

IP

also those who will suffer the deeper consequences of global warming, such a tax would require compensatory measures. Moreover, if GHG emissions are taken as a proxy for environmental

SC R

pressure, then there is already intense inequity; the households who are more exposed to higher environmental risks are those who contribute less to environmental degradation. In the near future, the main drivers of Brazilian emissions will be food production and energy

NU

consumption. On the supply side, there are promising initiatives for reducing GHG emissions in agriculture, on the demand side, consumption of food tends to decrease as a proportion of

MA

household spending and household emissions. Energy demand however, is expected to grow faster than the increase in renewable energy sources due to investment in pre-salt oil reserves and the increasing costs of new hydropower plants. Despite its slightly regressive character, a tax on GHG

D

emissions could stimulate a more sustainable path for the energy sector.

TE

The effects of a GHG emissions tax on welfare goes far beyond the short-term changes in the price index and real income. For instance, it involves normative issues such as the preservation of the

CE P

wellbeing of future generations because GHG released into the atmosphere today must have impacts on global warming over a broad time horizon. Brazilian NDCs reinforce the most cost effective policies to mitigate GHG emissions, mainly the control of deforestation. A new set of

AC

policies should be drawn up for the long run. A carbon tax should be combined with compensatory measures on income distribution. Further investigation should also observe the impact in the long-term, considering the recursive effects as well as compensatory measures, on income distribution and its implications. The model presented does not incorporate any changes in the technological matrix in response to the relative price modifications. Nevertheless, the focus was on household expenditure and labor profiles. In comparison to other studies, a more disaggregated matrix seems to soften the regressive impact of a carbon tax. In the midterm, reorientation towards less intensive sectors could also favor higher wage sectors. It is important to highlight that consumption and income effects do not interact in the model, therefore all estimates are first-round effects. In the real world, when income falls, household consumption also declines, and with reduction in consumption companies have to adjust their production creating a negative cycle. These are general equilibrium effects not taken into account

ACCEPTED MANUSCRIPT 15

here. This study could also include the flows between households and government, which is usual in the Social Accounting Matrix (SAM) and CGE model. This improvement will allow us to

T

identify the compensatory effects from the government.

IP

References

SC R

Baer, W. Brazilian economy: growth and development (6th Revd E). Boulder, CO: Lynne Rienner, 2007. Bosetti, V., Buchner, B. Data envelopment analysis of different climate policy scenarios. Ecological Economics 68(5): 1340 –1354, 2009.

MA

NU

BRAZIL. Ministério do Meio Ambiente (MMA). Informação adicional sobre a INDC. Apenas para fins de esclarecimento. Brasília, 2015, available at http://mma.gov.br/index.php/comunicacao/agencia-informma?view=blog&id=1163 , in 29/09/2015.Cornwell, A., Creedy, J., 1996. Carbon taxation, prices and inequality in Australia. Fiscal Studies 17, 21-38.

D

Dietzenbacher, E. In vindication of the Ghosh model: a reinterpretation as a price model. Journal of Regional Science 37(4): 629-651, 1997.

TE

Ferreira, P. Regional policy in Brazil: a review. Retrieved February 3, 2016, from http://www.fgv.br/professor/ferreira/RegionalPolicyFerreira.pdf. 2004.

CE P

Gemechu, E. D., Butnar, I., Llop, M. and Castells, F. Economic and environmental effects of the CO2 taxation: an input-output analysis for Spain. Working Paper, 24, 2012. Available at: .

AC

Genty, A., Arto, I. and Neuwahl, F. Final database of environmental satellite accounts: technical report on their compilation. WIOD documentation, 2012. Available at: . Ghosh, A. Input-Output approach in an allocation system. Economica New Series 25(97): 58-64, 1958. Gonzalez, F., Distributional effects of carbon taxes: The case of Mexico. Energy Economics 34: 2102-2115, 2012. Guilhoto, J. J. M., Sesso Filho, U. Estimação da matriz insumo-produto a partir de dados preliminares das contas nacionais. Economia Aplicada 9(2): 277-299, 2005. Gurgel, A. C., Paltsev, S. Costs of reducing GHG emissions in Brazil. Climate Policy 14(2): 209223, 2014. Hoffman, R. Transferências de renda e a redução da desigualdade no Brasil e cinco regiões entre 1997 e 2004. Econômica (Rio de Janeiro) 8, 55-81, 2006. INPE – Instituto Brasileiro de Pesquisas Espaciais. Programa de Monitoramento da Floresta Amazônica Brasileira. Available at: .

ACCEPTED MANUSCRIPT 16

International Energy Agency - IEA. CO2 Emissions from Fuel Combustion 2009. Edition IEA, Paris, 2009. Kerkhof, A. C., Moll, H. C., Drissen, E., & Wilting, H. C. (2008). Taxation of multiple greenhouse gases and the effects on income distribution: A case study of the Netherlands. Ecological Economics, 67(2), 318-326.

IP

T

Kerkhof, A. C., Nonhebel, S., & Moll, H. C. (2009). Relating the environmental impact of consumption to household expenditures: An input-output analysis. Ecological Economics, 68(4), 1160-1170.

SC R

Labandeira, X., Labeaga, J.M., 1999. Combining input output analysis and microsimulation to assess the effects of carbon taxation on Spanish households. Fiscal Studies 20, 305-320

NU

Leontief, W. The Structure of the American Economy, 1919-1939: An Empirical Application of Equilibrium Analysis. Oxford University Press: New York, 1941. Leontief, W. Input output economics. New York: Oxford University Press, 1966.

MA

Magalhães, A. S., Domingues, E. P. Economia de baixo carbono no Brasil: alternativas de políticas e custos de redução de emissões de gases de efeito estufa. Texto para discussão, n. 491, CEDEPLAR/UFMG, Belo Horizonte, agosto de 2013. Mesnard, L. Is the Ghosh model interesting, Journal of Regional Science 49(2): 361-372, 2009.

TE

D

Miller, R. E.; Blair, P. D. Input-output analysis: foundations and extensions. New York: Cambridge University Press. Second edition, 2009.

CE P

Oosterhaven, J. Leontief versus Ghoshian Price and Quantity Models. Southern Economic Journal 62: 750–759, 1996. Orlov, A., Grethe, H. Carbon taxation and market structure: a CGE analysis for Russia. Energy Policy 51: 696-707, 2012. de

Mudanças

Climáticas.

Available

at:

AC

PBMC – Painel Brasileiro .

Rong, F. Understanding developing country stances on post-2012 climate change negotiations: comparative analysis of Brazil, China, India, Mexico, and South Africa. Energy Policy 38(8): 45824591, 2010. Rose, A. Input-output economics and computable general equilibrium models. Structural Change and Economic Dynamics 6: 295-304, 1995. Seroa da Motta, R. Padrão de consumo, distribuição de renda e meio ambiente no Brasil. Texto para discussão 856, IPEA, Rio de Janeiro, 2002. Silva, J. G.; Gurgel, A. C. Impactos de impostos às emissões de carbono na economia brasileira. In: XXXVIII Encontro Nacional de Economia, 2010, Salvador. XXXVIII Encontro Nacional de Economia da ANPEC, 2010. Silva, J. G.; Gurgel, A. C. Impactos econômicos de cenários de políticas climáticas no Brasil. Pesquisa e Planejamento Econômico, vol. 42, n 1, april, 2012. Silveira Neto, R. D. M., & Azzoni, C. R. Non-spatial government policies and regional income inequality in Brazil. Regional Studies, 45(4), 453–461, 2011.

ACCEPTED MANUSCRIPT 17

Silveira-Neto, R. M., & Azzoni, C. R. Social policy as regional policy: Market and nonmarket factors determining regional inequality. Journal of Regional Science, 52(3), 433–450, 2012. Symons, E., Proops, J., Gay P. Carbon Taxes, Consumer Demand and Carbon Dioxide Emissions: A Simulation Analysis for the UK. Fiscal Studies 15(2): 19−43, 1994.

IP

T

Tiezzi, S. The welfare effects and the distributive impact of carbon taxation on Italian households. Energy Policy 33: 1597–1612, 2005.

SC R

Timmer, M. P., Dietzenbacher, E., Los, B., Stehrer, R. and de Vries, G. J. An illustrated user guide to the World Input–Output Database: the case of global automotive production. Review of International Economics., 23: 575–605, 2015.

NU

Tourinho, O. A. F., Seroa da Motta, R., Alves, Y., Uma Aplicação Ambiental de um Modelo de Equilíbrio Geral. Texto para Discussão, nº 976, IPEA – Instituto de Pesquisa Econômica Aplicada, Rio de Janeiro, 2003.

MA

Wier, M., Lenzen, M., Munksgaard, J., & Smed, S. (2001). Effects of household consumption patterns on CO2 requirements. Economic Systems Research, 13(3), 259-274. Wier, M., Birr-Pedersen,K., Jacobsen, H. K., & Klok, J. (2005). Are CO2 taxes regressive? Evidence from the Danish experience. Ecological Economics, 52(2), 239-2

D

ANNEX 1 – MATCHING GHG EMISSIONS TO INPUT-OUTPUT SECTORS

TE

The second Brazilian communication of the National Inventory Report shows the greenhouse gas emissions from 1990 to 2010. The data were presented according to both methods suggested by

CE P

IPCC. The Top Down, when the total supply of goods whose consumption causes GHG is taken as a reference for emissions. And the Bottom Up, the emissions are set to economic sectors following their consumption of these goods.

AC

The correspondence between GHG sources to input-output sectors is shown in Figure A1. If an inventoried source matched more than one economic sector, then its emissions were distributed accordingly to sectoral share of production or consumption. All the same, the emissions were made equivalent to intermediate consumption or sectoral production taken from Use and Make Tables. Hence the CO2 emissions from fossil fuel burning by the energy source were set through the sectors: “Oil and natural gas”, “Oil refining and coke” and “Production and distribution of electricity”, according to their fossil fuel consumption in the Top Down inventory. Matching the input-output sectors to the National Inventory’s sources of GHG was made possible by the National Economic Activities Classification (CNAE). CNAE codes are the reference for both, the Brazilian Energy Balance and the National Accounting System. The inventoried sources had their PRODLIST codes matched to CNAE codes.

ACCEPTED MANUSCRIPT 18

Figure A1: Correspondence between GHG sources to input-output sectors (continues...) Inventory sectors

MIP sectors Oil and natural gas Oil refining and coke

Energetic subsector

Alcohol Production and distribution of electricity Cement

Pig iron and steel

Manufacture of steel and derivatives

Iron alloys

Manufacture of steel and derivatives Iron ore

SC R

Mining

IP

T

Cement

Other mining and quarrying Metallurgy of non-ferrous metals

Non-ferrous

Metal products Alcohol

Chemicals

Chemical

NU

Manufacture of resin and elastomers Pharmaceutical products Agrochemicals Paints, varnishes, enamels and lacquers Diverse chemical products and mixtures Food and Beverage

Textiles

Textiles

Pulp and paper products Ceramics

Pulp and paper products

D

Food and Beverage

TE

Fossil fuel burning

Other products of non-metallic minerals Smoking products Vestment goods and acessories Leather goods and footwear Wood products Newspapers, magazines and discs Rubber and plastic goods Metal products

Other industries

AC

CE P

ENERGY

Industry subsector

MA

Perfumaria, higiene e limpeza

Machinery and equipment, including maintenance and repairs Machinery for office and computer equipment Electronic materials and comunication equipment Automobiles, station wagons and pick-ups Trucks and buses Parts and acessories for automotive vehicles Furniture and products from diverse industries Other transport equipments Medical and hospital rquipment, measurement and optical

Transport subsector Agriculture subsetor Trade subsector

Transportat, storage and mail Agriculture, forestry, extractive Livestock and fishing Trade Public education

Public Subsector

Public heath Public administration and social security

Fugitive emissiona

Source: Author’s own, 2014.

Coal mining

Other mining and quarrying

Petroleum extraction and transportation

Oil and natural gas

ACCEPTED MANUSCRIPT 19

Figure A1: Correspondence between GHG sources to input-output sectors (conclusion) Inventory sectors Mineral products

MIP sectors

Cement production (clínquer)

Cement

Lime production

Other products of non-metallic minerals

T

Manufacture of resin and elastomers Pharmaceutical products

Chemical industry

Production of ammonia, adipic acid; nitric acid; other chemicals

IP

Agrochemicals

Perfumes, hygiene and cleaning Paints, varnishes, enamels and lacquers

SC R

INDUSTRIAL PROCESSES

Chemicals

Diverse chemical products and mixtures Agriculture, forestry, extractive Other mining and quarrying Food and Beverage Textiles

Enteric fermentation

Burning of agricultural waste Agricultural soils Animals on pasture

MA

Livestock and fishing

Rice cultivation

Agriculture, forestry, extractive Agriculture, forestry, extractive Livestock and fishing Agriculture, forestry, extractive

Animal waste Agricultural wastes

Livestock and fishing

Organic soils

Agriculture, forestry, extractive

D

AC

Source: Author’s own, 2014.

Agriculture, forestry, extractive

Synthetic fertilizers

CE P

WASTE

Metallurgy of non-ferrous metals Livestock and fishing

Treatment of animal wastes

Direct emissions

NU

Manufacture of steel and derivatives

Aluminum production

TE

AGRICULTURE AND LIVESTOCK

Metallurgy industry

Pig-iron and steel

Agriculture, forestry, extractive Production and distribution of electricity

ACCEPTED MANUSCRIPT 20

Figure 1 – CO2 equivalent emissions by source, in Gg, between 1990 and 2010. 3000000.00 2500000.00

T

2000000.00 1500000.00

IP

1000000.00 500000.00

ENERGY

INDUSTRIAL PROCESS

AGRIBUSINESS

LULUCF NET EMISSIONS

NU

Source: MCT Data, 2013, p.12.

WASTE

SC R

-

MA

Figure 2 - Household CO2 equivalent Emissions in 2009. 0.25

4.00 3.00

CE P

0.1

0

5.00

TE

0.15

0.05

6.00

D

0.2

7.00

2.00

1.00 -

AC

DECILE DECILE DECILE DECILE DECILE DECILE DECILE DECILE DECILE DECILE 1 2 3 4 5 6 7 8 9 10

Emissions coefficient in consumption

Emissions by DMCL

*On the left axis, for the series of consumption emission coefficients the values are in Gg per million U$. On the right axis, for emissions by households, values in 100 Gg. Source: Author’s own, 2014.

ACCEPTED MANUSCRIPT 21

Figure 3: Household consumption change effect Food and beverage Livestock and fishing Transport, storage and mail

T

Agriculture, forestry and extractive Manufacture of steel and derivatives

IP

Oil refining and coke

-17,000 -15,000 -13,000 -11,000

-9,000

-7,000

NU

SC R

Construction

-5,000

-3,000

Lodging and food services Production and distribution of electricity Oil and natural gas Trade Cement Chemicals

-1,000

1,000

MA

Value of production in RS million

D

Source: Author’s, 2014.

CE P

TE

Figure 4 – Welfare losses measured by compensatory variation

0.00%

-1.00% -1.50% -2.00%

AC

-0.50%

-1.18% -1.45% -1.83% -1.75%

-2.50% -2.50% -3.00%

-2.79% -3.09%

-3.50% Source: Author’s Own, 2014.

-2.36%

-2.22%

-2.10%

ACCEPTED MANUSCRIPT 22

Figure 5: Effects on labor income and total income Decile 1 Decile 2 Decile 3 Decile 4 Decile 5 Decile 6 Decile 7 Decile 8 Decile 9 Decile 10

0.00 -0.50

-1.91 -2.14

-2.17 -2.36

-3.00

-2.44 -2.94

-4.00

-3.72

MA

Labor income

CE P

TE

D

Source: Author’s Own, 2014.

AC

-1.78

NU

-3.50

-1.45

-1.61

-1.69

-1.77

-1.11

IP

-1.45

-2.00 -2.50

-1.25

SC R

Income variation (%)

-1.18

-1.50

T

-0.73

-1.00

Total income

-0.95 -1.02

-1.29

ACCEPTED MANUSCRIPT 23

Table 1 - Descriptive statistics of the income data per month by representative household (in real R$) Household per capital income Std. Dev

Min

Max

Mean

61.67

32.31

0

106

39.13

Decile 2

137.16

16.71

107

165

101.53

Decile 3

201.47

21.06

166

232

146.43

Decile 4

266.73

21.01

233

300

Decile 5

337.08

21.72

301

375

Decile 6

426.87

29.40

376

465

Decile 7

529.77

40.86

466

600

Decile 8

696.46

56.99

601

Decile 9

1,004.92

135.82

801

Decile 10

2,684.09

2,515.47

1,294

631.20

1,083.89

0

AC

CE P

TE

D

Source: Authors’ Own based on PNAD data (2009).

Max

0

900

192.92

0

1,300

246.86

0

2,000

211.29

312.81

0

2,200

255.99

348.91

0

2,500

282.98

390.33

0

3,500

409.76

491.48

0

3,890

800

532.33

626.29

0

5,000

1,293

740.90

901.66

0

9,000

94,669

1,920.25

3,266.91

0

150,000

94,669

461.25

1,238.536

0

150,000

SC R

NU

MA

Total

Min

104.10

IP

Decile 1

Std. Dev

T

Mean

Household labor per capita income

ACCEPTED MANUSCRIPT 24

After Tax

Variation

Livestock and fishing

339,829.00

285,011.59

-0.16

Transport, storage and postal mail

140,911.19

136,435.67

-0.03

Agriculture, forestry, extractive

100,126.36

96,630.97

-0.03

Manufacture of steel and derivatives

58,654.91

55,655.17

Oil refining and coke

32,650.38

IP

-0.05

31,916.18

-0.02

Cement

28,402.67

25,068.37

-0.12

Other mining and quarrying

21,581.18

20,243.87

-0.06

Oil and natural gas

19,362.45

18,997.09

-0.02

Production and distribution of electricity

17,120.65

16,962.35

-0.01

12,671.26

12,397.48

-0.02

12,084.05

11,719.51

-0.03

6,281.45

6,126.41

-0.02

5,404.61

5,154.71

-0.05

4,488.48

4,417.48

-0.02

3,530.32

3,485.18

-0.01

2,918.34

2,838.49

-0.03

2,100.35

2,093.98

0.00

Paints, varnishes, enamels and lacquers

1,859.99

1,830.30

-0.02

Public administration and social security

1,586.84

1,583.55

0.00

Construction

1,533.02

1,515.42

-0.01

Textiles

1,311.12

1,297.85

-0.01

Total for 21 sectors

814,408.61

741,381.63

-0.09

Total

822,069.73

748,978.31

-0.09

Chemicals

Metallurgy of non-ferrous metals Food and beverage Pulp and paper products Iron ore Alcohol

AC

CE P

TE

Trade

Source: Author’s Own, 2014.

SC R

MA

Other products of non-metallic minerals

NU

Sectors

T

Before Tax

D

Table 2 - Total CO2 equivalent emissions by sector before and after tax, in Gg.

ACCEPTED MANUSCRIPT 25

HIGHLIGHTS We measure the distributive impact of a taxation policy of GHG emissions in Brazil.



Poorest households have the highest expenditure emissions coefficient.



The poorest households are most affected by both reducing consumption and labor income.



Overall, the taxation was slightly regressive and had a small negative impact on output.

AC

CE P

TE

D

MA

NU

SC R

IP

T

