Does U.S. Export Global Warming? Coal Trade and the Shale Gas Boom Christopher R. Knittel

Konstantinos Metaxoglou

Anson Soderbery

Andre Trindade

∗

This version: 07/31/2017

Abstract The U.S. shale gas boom in recent years generated vast amounts of cheap natural gas, which reduced demand for coal in the domestic electric power sector. At the same time, it contributed to an increase in the U.S. exports of coal. Although the domestic environmental implications of coal displacement have been well-studied and documented, the global environmental implications of the surge in coal exports have not received as much attention. In this paper, we examine the impact on global carbon emissions of the shale gas boom by analyzing whether the increased U.S. coal exports have contributed to an increase in imported coal consumption around the world such that the reduction in domestic carbon emissions is offset by an increase in carbon emissions elsewhere. Our workhorse is a structural model that links the domestic coal market to the international coal market employing techniques from industrial organization and international trade. We use our structural model to simulate counterfactual international coal trade flows in the absence of the shale gas boom. The trade flows are then converted into potential carbon dioxide (CO2 ) emissions. In the absence of the U.S. shale gas boom, consumer welfare associated with coal imports is higher by 34 billion dollars, after accounting for environmental damages due to emissions. This represents roughly a 4% of the dollar value of world coal trade. CO2 emissions are lower by approximately 0.4%. Keywords: Coal, Emissions, International Trade, Shale Gas Boom. JEL codes: F18, L13, Q53.

∗

Knittel: Sloan School of Management, MIT, and NBER, [email protected]. Metaxoglou: Carleton University, [email protected]. Soderbery: Department of Economics, Purdue University, asoderbe@ purdue.edu. Trindade: FGV/ EPGE, Brazilian School of Economics and Finance, [email protected]

1

Introduction

In this paper, we examine the effects of the change in a country’s consumption of fossil fuels on the environment worldwide via trade flows. Our work is motivated by the change in the mix of fossil fuels consumed by the U.S. electric power sector. This exogenous change was triggered by the dramatic drop in the price of natural gas in the aftermath of what has become known as the “Shale Gas Boom” due to new developments in hydraulic fracturing and horizontal drilling (Figure 1). Although the domestic environmental implications of the U.S. shale gas boom have been well-studied, to the best of our knowledge, the global environmental implications have not—the paper aims to fill this void. The downward pressure on the price of U.S. coal due to lower domestic demand by the electric power sector—which has historically accounted for more than 80% coal consumption— coupled with the inability of the U.S. to export cheap natural gas in large scale, made U.S. coal an attractive option for coal-importing countries, leading to a large increase of U.S. coal exports. In 2009Q1, U.S. exported 4.2 million metric tons of steam coal for electricity generation while in 2012Q2 it exported almost four times as much. The lower domestic demand by the electric power sector has been attributed, to a large extent, to the dramatic drop in the price of gas, which, as a result, has become an increasingly closer coal substitute for electricity generation in recent years. In June of 2008, the average monthly price of gas paid by U.S. power plants was $12/MMBtu, while that for coal was around $2/MMBtu. By April of 2012, the two prices were almost at parity with the vast amounts of cheap natural gas that flooded north America being the primary driver of this big change in the relative price of the two fuels.1 The changing landscape in the U.S. electric power sector due to the shale gas boom, has a two-pronged effect on the trade flows of coal around the world. First, there is a decrease in the domestic demand for coal. Second, there is an increase in export supply of U.S. coal because domestic producers are looking for alternative markets to sell their product at competitive prices. Translating these domestic comparative statics, to global comparative statics that pertain to flows of coal around the world is ultimately an empirical question and the answer depends on export supply and import demand elasticities, whose magnitude is 1 The widespread coal-to-gas switching throughout the industry, for which we should not also discount contemporaneous environmental policy, and its implications for emissions, are by now well documented. See Cullen and Mansur (2014), Knittel et al. (2015), and Linn et al. (2014), among others. Hausman and Kellogg (2015) provide an in-depth analysis of the economic and environmental impacts of the shale revolution. See also Jackson et al. (2014) on the environmental costs and benefits of fracking. Bartik et al. (2017) offer a very informative primer on hydraulic fracturing.

2

determined by several factors. For example, the U.S. export supply elasticity is affected by the ability of domestic coal producers to ship coal outside the country, which in turn depends on the current infrastructure of major railroads and ports in the Eastern seaboard that have historically served European markets with metallurgical coal from the Appalachian region.2 The same point can be made for the mine, rail, and port utilization in Indonesia, and Australia, which are the world’s largest coal producers. At the same time, the import demand elasticity for U.S. coal in major consuming regions such as Western Europe, China, and India, depends on the availability, or lack of, close substitutes. Most importantly, the implications of an increase in exports of U.S. coal for global emissions associated with coal trade are ambiguous and depend on the exports’ effect on trade flows. Our empirical approach to assess the implications of the shale gas boom on coal trade flows builds on an econometric model with an international and a domestic component.3 The first component draws from the literature on international trade. Following Soderbery (2016), we estimate the link between U.S. exports and the global market for coal focusing on the mechanism through which the U.S. gas market affects U.S. coal production and exports. In particular, our trade model allows for upward sloping export supply curves, which is an important difference from the standard gravity models that assume perfectly elastic export supply curves, in a partial equilibrium framework. Assuming that export supply curves are subject to shocks (shifts), we treat the U.S. shale gas boom as a shock to the U.S. coal exports. We then construct counterfactual coal trade flows in the absence of the boom, which we model as a negative shock to the U.S. export supply of coal. We depart from the trade models with upward sloping export supply in Broda and Weinstein (2006) and Feenstra (1994) (FBW, henceforth) allowing export supply elasticities to exhibit heterogeneity across importers, goods, and exporters. We do so because, although homogeneous import demand elasticities find empirical support in the trade data, homogeneous export supply elasticities do not (Soderbery (2015)). In our case, the imported good is one of three types of coal: anthracite, bituminous, and other. Following the standard approach in the literature, a variety is defined by the country of supply for a particular good (Armington (1969)).4 While the FBW approach is better suited than gravity models for our analysis, their assumptions of homogeneous export supply of coal across exporters within 2

Port infrastructure on the Pacific coast is also very important for U.S. coal producers, especially those in the Western region, for accessing the Asian market. 3 We recognize that both econometric and computable general equilibrium (CGE) models have their advantages and disadvantages; see Kirkpatrick and Scrieciu (2008) for an informative discussion. 4 For example, U.S. bituminous coal is a different variety from Australian bituminous coal.

3

an importing country is overly restrictive. Allowing for this heterogeneity is crucial in our case because the shock to the model in our counterfactual scenario starts from one particular export supply curve (from the U.S.) and then propagates to the rest of the world. The second component of our econometric model is a stylized industrial organization (I.O.) model that links U.S. coal production to the domestic price of gas. The international trade model allows us to estimate import demand and export supply elasticities while the I.O. model provides the link between the international market for coal and the U.S. price for gas through the U.S. export supply curve. Having established this link, we calculate counterfactual world coal trade flows removing the drop in the U.S. price of gas caused by the shale gas boom. Using information on the heat and carbon dioxide content of coal we translate these trade flows into potential emissions to estimate the global environmental impact of the shale gas boom. For the first component of our model we use the nonlinear SUR estimator in Soderbery (2016) and UN COMTRADE data between 1990–2014 for close to 140 (190) importing (exporting) countries to estimate import demand and export elasticities, as well as shocks to the export supply curve for U.S. coal. We then utilize the first-order conditions of our I.O. model to link these shocks to the domestic price of gas in the U.S.. Estimating the relationship between the price of gas in the U.S. and Europe for 1990–2006, we construct counterfactual U.S. gas prices for 2007–2014. Our assumption is that these counterfactual prices are the ones that would prevail but for shale gas boom. The counterfactual U.S. gas prices allow us to construct counterfactual shocks to the U.S. export supply that translate into counterfactual coal trade flows. We present detailed results regarding our counterfactual analysis for 35 countries that account for more than 90% of global coal imports and exports during the period of interest. The same group of countries also accounts for more than 90% of imports of U.S. coal. In the absence of the shale gas boom during 2007–2014, the quantity (metric tons) of coal traded is approximately 0.4% lower than the actual amount and the price of coal (USD/metric ton) is 1.4% higher. The dollar value of trade is 1% higher in the absence of the boom. In the case of major importers of U.S. coal, we see a drop of 46% in quantity and a 14% increase in price. For the same group of countries, the dollar value of trade is 39% lower. The counterfactual prices, quantities and dollars for major coal importers and exporters lie somewhere between -0.397% (quantity for major exporters) and 1.4% (price for major exporters) relative to the actual values. Using information on the heat, carbon dioxide (CO2 ), and sulfur dioxide (SO2 ) content of coal, we are able to compare actual and counterfactual CO2 and SO2 emissions associated 4

with the consumption of imported coal. We also calculate potential environmental damages from CO2 emissions assuming a social cost of carbon of $37/metric ton. Based on our calculations, CO2 emissions are about 0.4% lower and SO2 emissions are 0.3% lower in the absence of the shale gas boom. We find that consumer welfare associated with coal imports taking into account potential environmental damages is 34 billion USD higher in the absence of the boom. To put this number into perspective, the value of actual (counterfactual) world coal trade for 2007–2014 is 826.5 (835) billion USD. We conclude our analysis by providing a back-of-the-envelope calculation of the change in emissions associated with seaborne trade of U.S. coal. To the best of our knowledge, this is the first paper to systematically examine and assess the global environmental effects of U.S. coal exports bringing together the international trade, I.O, and environmental economics literature.5 In related work, Wolak (2016) uses a spatial equilibrium model to assess how the U.S. shale gas boom impacts global coal market outcomes accounting for coal-to-gas switching in the electricity sector in the U.S. and Europe, the potential for China to exercise buyer power, and the impact of increasing the western U.S. coal export port capacity. Although Wolak’s model allows for substitution between fuels (e.g., coal and gas) for some parts of the world and we don’t, it lacks the flexibility of our model, which allows for different export and import elasticities. This flexibility is crucial for our counterfactual because we consider a shock to the export supply curve of a single country. Another important difference between our model and Wolak’s model is that his model is mostly calibrated while all of our parameters are estimated. The paper also contributes to an emerging literature on the interplay between environmental economics and international trade with the work by Eyer (2014) being the most closely related to our paper. Eyer estimates the effect of domestic natural gas prices on U.S. coal exports and finds that a 1% increase in the domestic price of natural gas leads to a 2.2% decrease in U.S. coal exports. According to his findings, approximately 75% of displaced U.S. steam coal will be shipped abroad. Arezki et al. (2016) find that U.S. energy intensive manufacturing sectors benefited from the reduced gas prices due to the shale gas boom. A back of the envelope calculation suggests that energy intensive manufacturing exports increased by $101 billion in 2012 due to the boom. Shapiro (2016) finds that the benefits 5

According to Afsah and Salcito (2014) had U.S. steam coal exports in each year between 2007 and 2012 been 20.7 million short tons, which is the annual average for 2000–2007, the counterfactual U.S. steam coal exports would have been 82.7 million short tons. Between 2007 and 2012, the actual U.S. exports of steam coal were 207 million short tons. The implied additional CO2 emissions due to increased exports following the shale gas boom are approximately 149 million tons (see Exhibit 3). In the same paper, the authors show that coal-to-gas switching in the U.S. electric power sector led to a decrease in CO2 emission of 86 million tons. Hence, there is a net increase in CO2 emissions of 149-86=63 million tons.

5

of international trade exceed environmental costs due to CO2 emissions by two orders of magnitude. While proposed regional carbon taxes on shipping-related CO2 emissions would increase global welfare and increase the implementing region’s GDP, they would also harm poor countries (see also Cristea et al. (2013)).6 The remainder of the paper is organized as follows. In Section 2, we provide a background on U.S. production and coal exports, as well as on international coal trade, which the reader familiar with the industry may want to skip. Section 3 first describes the model of international trade and then the model of the U.S. domestic coal market. The empirical findings are reported in Section 4 and the results of the counterfactual trade flows in the absence of the U.S. shale gas boom are presented in Section 5. We finally conclude. All tables and figures are provided after the main text. In the Appendix, we provide some additional background on U.S. coal exports and world trade of bituminous coal as well as some descriptive statistics for coal imports and exports and additional estimation results.

2

Background

2.1

U.S. Coal

Domestic Production: The U.S. has vast amounts of energy in coal fields that spread across its Appalachian, Interior, and Western regions. The Powder River Basin (PRB) alone contains one of the largest sources of energy on the planet with over 200 billion short tons of coal in place, which is equivalent to more than 3,616 quadrillion Btu (quads).7 According to figures from the World Energy Council for 2011, the U.S. accounts for 28% of global recoverable coal reserves followed by Russia (18%) and China (13%) noting that 10 countries account from more than 92% of global reserves.8 Coal is an organic rock that contains 40%–90% carbon by weight and it is classified into four types (ranks), based on the amount of heat it produces and, for coking or metallurgical 6

The effect of trade on the environment is theoretically ambiguous. The race-to-the-bottom hypothesis (negative effect) competes against the gains-from-trade hypothesis (positive effect). For example, Frankel and Rose (2005) find that trade tends to reduce three measures of air pollution; in particular, sulfur dioxide and nitrogen dioxide. According to the authors, while results for other environmental measures are not as encouraging, there is little evidence that trade has a detrimental effect on the environment. See Kirkpatrick and Scrieciu (2008) for a summary of the literature findings based on both econometric and CGE models. 7 The second largest energy reserve, the North Dome-South Pars natural gas field, has 1,228 quads in reserve. Ghawar, the world’s largest oil field in Saudi Arabia has 418 quads in reserve (Considine (2013)). 8 Each of the remaining countries—Australia, India, Germany, Ukraine, Kazakhstan, Indonesia, and Serbia—accounts for less than 10%. See Figure A.1.

6

coal, its agglomerating (“caking”) properties.9 Lignite is the lowest coal rank. It is a brown coal and it is used almost exclusively as fuel for steam electric power generation with a heat content of 9–17 MMBtu per ton. It is mainly produced in North Dakota and Texas. Subbituminous coal, the second type of brown coal, is also used in electric power generation and has a heat content of 17–24 MMBtu. It is produced in vast amounts in the PRB. Bituminous coal, one of the two hard coals, produced in the Appalachian region and the Midwest, has a content of 21–30 MMBtu. It can be used as steam coal in electricity generation, as well as metallurgical coal in steel production. Finally, anthracite, the second of the hard coals, is the highest coal rank with heat content of 22–28 MMBtu. It is extracted in the U.S. only in northeast Pennsylvania. Between 1994 and 2015, bituminous and sub-bituminous coal have accounted for 93% of annual US production (tons), while anthracite has accounted for less than 1% (EIA, Annual Coal Review). Exports: Coal consumption by the U.S. electric power sector during 2004–2008 was close to 1 billion short tons, its highest levels since 1992. By 2012, it fell to 824 million short tons because of the drop in gas prices, the slowdown of the economy due to the Great Recession, and a series of regional and federal environmental regulations aiming to curb coal-related emissions. This contraction of the domestic market was accompanied by the surge in exports of U.S. coal exports documented in Figure 1 attracting increased attention in the popular press.10 To give the reader an idea about the magnitude of the increase in coal exports, in 2008, the U.S. exported 242 thousand short tons of coal (steam plus metallurgical) to China. In 2012, these exports were 10 million short tons. As a result, the exports’ share in production increased from 5.3% to 12.5% (Figure 2).11

2.2

International Trade

According to the EIA international energy statistics, world coal consumption increased from around 5 billion metric tons in 1990 to more than 7.5 billion metric tons by 2012 (Figure 3, 9 Coking coal refers to bituminous coal suitable for making coke used as a fuel and as a reducing agent in smelting iron ore. 10 Andrew Revkin of The New York Times was writing that the “U.S. Push to Export Dirty Fossil Fuels Parallels Past Action on Tobacco,” in February, 2014. According to Ernst and Young (2013), there were $16.6 billion of gross value added (GVA) in the U.S. Economy by U.S. coal exports in 2011 with Virginia and West Virginia alone accounting for roughly a third. 11 Based on data from the EIA and Department of Commerce. Between 2007 and 2012, the share of bituminous coal in U.S. exports increased from 64% to 84%, while the share of other coal decreased from 35% to 15% noting that U.S. coal production dropped from 1,147 million short tons in 2007 to 1,016 million short tons in 2012 (EIA, International Energy Statistics). Section A.1 provides some additional information regarding the split between metallurgical and steam coal of U.S. exports, as well as the customs districts from which U.S. coal is shipped.

7

panel (a)).12 During this time, coal trade increased from 400 million metric tons to more than 1.2 billion (panel (b)) with seaborne trade accounting for about 85% of all trade in the last 25 years.13 Historically, two regions, Europe (Atlantic Market) and Asia (Pacific Market) have played a key role in coal trade following different trends in recent years as we discuss below. Overall, less than 40 countries account for more than 90% of total exports, total imports, and imports of U.S. coal during this period (Table A.2). Australia, Indonesia, the U.S., Russia, Colombia, and South Africa are the top exporters, with the first two accounting for more than half of all exports after 2010. Overall, the countries listed in panel (c) of Figure 3 accounted for more than 80% of all coal exports during 1990–2012. Australia, Indonesia, Russia, and the U.S. account for about 70% of total coal exports (tons) for 1990–2014 (Table A.5). Ten countries accounted for more than 2/3 of annual world coal imports during 1990–2014 (panel (d)). Japan’s share of world imports fell from around 50% in 1990 to close to 20% in 2014. Korea’s share remained relatively stable around 10%, while China’s share was close to 20% for 2010–2014. India’s share increased from less than 10% in 2010 to about 20% in 2014, while none of the remaining countries has accounted for more than 5% during the same period (Table A.3). Canada, Japan, Brazil, Italy, and Great Britain, accounted for half of the imports of U.S. coal during 1990–2014 period. (Table A.4). Setting aside the vast energy needs of China and India in recent years, a series of events have also contributed to an increase in the demand for coal worldwide, which in turn also contributed to the increase in U.S. coal exports. The European Union (E.U.) Emissions Trading System essentially collapsed by early 2006 leading to a dramatic drop in the CO2 permit prices. The Arab Spring began in December 2010 in Tunisia disrupting the E.U. natural gas markets that have historically relied on gas originating in Africa (e.g., Algeria, Egypt, Nigeria). Japanese demand for coal and natural gas increased in March of 2011 due to the Fukushima nuclear accident. The Bowen Basin in Australia, which accounts for close to a third of global metallurgical coal production was hit by floods in December of 2011.14 More recently, in May 2012, Germany announced that it would retire all its nuclear capacity by 2022 increasing Germany’s demand for alternative sources of energy. In early 2014, Russia, one of the E.U. largest suppliers of energy, invaded Ukraine causing major gas supply disruptions in the E.U. market. This is because Ukraine is the main corridor of 12

We use ISO Alpha 3 country codes to identify countries in various tables and figures. See Table A.1. Based on annual figures from the IEA Coal Information 2014 (see Table 3.1). Sea freight rates are an important component of cost accounting for up to 40–50% of CIF prices (Ritschel and Schiffer (2007)). The majority of international steam coal shipping utilizes freight rates for Capesize vessels (Zaklan et al. (2012)). See Schernikau (2010) for a very informative discussion of the economics of the international coal trade. 14 https://goo.gl/TZ1GMK. 13

8

Russian natural gas to the E.U. Figure 4 shows the annual time series of the quantity (million metric tons), value (billion USD), and price (USD/metric ton) of UN COMTRADE import data for the three types of coal used in estimating our international trade model: anthracite, bituminous, and other coal. Consistent with our earlier discussion, there is an upward trend in both quantities and dollars across all three types of coal with the bituminous coal capturing the lion’s share—more than 70% during the entire period. Between 1990 and 2000, coal prices decreased from around $60 per metric ton to almost $40. Between 2000 and 2011, prices for bituminous coal increased by a factor of 3 reaching $160 per ton in 2011 after a brief drop in 2009–2010 due to the most recent recession. Section A.2 in the Online Appendix provides information regarding the primary destinations (sources) of bituminous coal for major exporters (importers). Figure 5 shows that import prices paid for coal (USD/metric ton) in China, India, Japan, and Korea, are highly comparable, and follow the same pattern over time, especially prior to 2005 (panel (a)). The import prices paid in major European markets, such as Germany, Great Britain, the Netherlands, and Spain, also track each other closely, and are comparable to those in Asia (panel (b)). In general, the price spread between the European and Asian markets is small. Import prices for coal originating from major producing countries, such as the US, Australia, Indonesia, Russia, Colombia, Canada, and South Africa, track each other closely, which is consistent with spatial arbitrage and competitive supply (panels (c) and (d)). Some signs of divergence in prices, however, seem to have emerged post 2008.15

3

Model

Our trade model is designed to quantify competitive equilibrium responses to exporter shocks market by market. In what follows, we link a structural model of the U.S. markets for coal and gas to global markets. To do so, we show how to micro-found the export supply curves using a flexible model of domestic production of coal. Our model allows us to establish a structural link between the U.S. domestic markets of coal and gas along with how shocks in these domestic markets affect U.S. exports of coal around the world. 15

There is plenty of evidence that the world market for steam coal is integrated (Li et al. (2010)), which is in sharp contrast with the world market for natural gas, where three distinct markets have been developed over time: Asian market, European market, and U.S. market (Knittel et al. (2016)).

9

3.1

International Trade

We maintain common assumptions from new trade theory in a model that is amenable to structural estimation. Once estimated, the model allows us to quantify the welfare implications of CO2 emissions associated with coal imports due to the effect of U.S. gas prices on the global coal trade. We bring the model to the data following common functional forms and the estimation strategy in Soderbery (2016). To introduce some notation, we use I to denote the importing country, g to denote the imported good, and v to denote the variety of the imported good. The total number of goods imported by country I is GI and the total number of varieties is GIv . Goods are defined by their COMTRADE HS6 code and their varieties are determined by their country of origin. For our purpose, a good is one of three types of coal: anthracite, bituminous, and other. Additionally, U.S. bituminous coal imported in Japan is a different variety from Australian bituminous coal imported in Japan due to physical characteristics, such as calorific value, sulfur content, ash content, moisture, etc.16 We consider a representative consumer in importing country I with constant elasticity of substitution preferences (CES) for variety v of coal type g. The representative consumer aggregates consumption of coal varieties via Cobb-Douglas preferences.17 These underlying assumptions give rise to the following utility function at time t: I

Gt Y I I Ut = (QIgt )αgt

(1)

g=1 16

We use the following the HS6 codes 270111 (anthracite, pulverized or not, not agglomerated), 270112 (bituminous, pulverized or not, not agglomerated), 270119 (other coal, except anthracite or bituminous, pulverized or not, not agglomerated). See http://comtrade.un.org/db/mr/rfCommoditiesList.aspx? px=H1&cc=2701. As an example, if Japan imports all three types of coal from the U.S. and Australia only, GI = 3 and GIv = 6. 17 We assume separability in the utility over the composite domestic (d) and imported goods along the I I QG I t lines of UtI = (QIdt )αdt g=1 (QIgt )αgt . This assumption allows us to focus on prices and the consumption of imported goods for estimation and relax the constraint imposed by the lack of data, primarily on prices, for domestic coal. Note that with the exception of Germany and Great Britain, the average annual import/consumption ratio is either large (at least 70%, Brazil) or small (at most 11%, Russia) for 1990–2014, which means that the domestic and import markets are segmented. On a secondary note, our model does not allow for substitution between coal and gas, which is relevant for the electric power sector. Wolak (2016) makes a strong case that such substitution is only possible in North America and western Europe because of the availability of gas supplied by pipelines and the current gas-fired generation mix in the short- and medium-term. Hence, by ignoring the substitutability between domestic and imported coal, as well as between gas and imported coal, our elasticity demand estimates may be somewhat inflated for some western European countries and Canada.

10



GIvt

QIgt ≡ 

X

1 I

I ) (bIgv ) σg (qgvt

I −1 σg I σg



I σg I −1 σg



,

(2)

v=1

where QIgt is the CES aggregate consumption of imported coal varieties assuming GIvt varieties in total with GIt being the total number of goods. Additionally, σgI > 1 is the elasticity of substitution across coal types and bIgv captures variety-specific tastes. For example, bIgv may capture the fact that coal of type g originating in country v is better suited for the steel industry or the electric power sector due to its coking properties and its sulfur content, respectively. Because of the Cobb-Douglas preferences across coal types, the expenditure for I coal type g accounts for αgt of the total expenditure associated with coal purchases. Although we model preferences similar to Shapiro (2016), our approach generally departs from his. Shapiro focuses on emissions due to the transport of a wide range of goods while, we focus on emissions associated with coal trade alone. Hence, we are interested in structurally estimating demand and supply in the world market for coal and the welfare effects from changes in the consumption of imported coal. Notably, assuming utility is log separable across goods, we can focus on the market for coal in importing countries holding other trade constant, without loss of generality. We model the international market for coal following Soderbery (2016) and estimate import demand and export supply elasticities allowing for substantial heterogeneity. The import demand for coal of type g implied by (1) is I

I

I I I I σg −1 qgvt = αgt bgv (pIgvt )−σg (Pgt ) ,

 I Pgt ≡



I

Gvt X

bIgv pIgvt


1−σgI

(3)

1 I 1−σg



(4)

v=1 I where pIgvt is the delivered price and Pgt is the CES price index. We combine import demand with a flexible export supply specification to facilitate structural estimation. We assume monopolistic competition among exporters with export supply curves that are variety- and exporter-specific as in Armington (1969) and upward sloping with a constant inverse export I supply elasticity ωgv : I

I I pIgvt = exp(ηgvt )(qgvt )ωgv .

11

(5)

I We also allow for unobservable variety-specific supply shocks ηgvt to facilitate estimation. These shocks will be the channel through which changes in U.S. gas prices have partial equilibrium effects on the world flows of coal. The U.S. shale gas boom serves as a positive shock to the U.S. coal export supply curve. Upon estimating the import demand (σgI ) and I inverse export supply (ωgv ) elasticities, we can calculate the demand and supply shocks using (3) and (5). We then use the firms’ profit-maximizing first-order conditions from the I.O. model to link U.S. gas production to world coal trade through these shocks.

3.2

Domestic Market

We now sketch a stylized model for the U.S. domestic production of coal, which will allow us to establish a link between the U.S. coal export supply shock and the domestic price of gas. We consider a representative firm that extracts coal for sale in the international (f ) and c domestic (d) markets at time t with (pcf t , qfc t ) and (pcdt , qdt ) being the corresponding prices and quantities. Consistent with the assumption of monopolistic competition in exports of the trade model, the firm is a price-taker in the foreign market but faces a downward-sloping residual demand curve in the domestic market. The domestic inverse demand for coal is a function of the domestic gas price, pgdt , and an additional demand shifter to account for additional factors driving the demand for coal, wdt , such as fossil-fuel generation by electric power plants. Assuming linearity, we write c n c pcd (qdt , pgdt , wdt ; θ) = θ0 + θ1 qdt + θ2 pgdt + θ3 wdt ,

(6)

where θ ≡ (θ0 , θ1 , θ2 , θ3 ). The motivation for the domestic inverse demand curve stems from the fact that electric power plants account for the vast majority of coal consumption and natural gas is the closest substitute for coal during the period that is relevant in our analysis. It also resembles the demand equation for coal in the U.S. in Wolak (2016) with the caveat that he uses a log linear specification. The hypothetical representative firm first decides how much coal to sell in the domestic market. Subsequently, the firm decides how much to sell in the foreign coal market. Although arbitrage is not possible, the two markets are related through production costs: c αd c αf c c C(qdt , qfc t ; γ) = β0 qdt + β1 (qdt ) (qf t ) ,

(7)

where γ ≡ (β0 , β1 , αd , αf )0 . The parameters af and ad , associated with the marginal costs, 12

introduce convexity assuming af > 1 and ad > 1. The interpretation for the functional form in (7) is that extracting coal for the domestic market makes it more costly to extract coal for the foreign market. It captures the salient feature of the mining costs since extracting additional amounts of coal implies higher marginal costs. In the absence of the foreign market, extraction to serve the domestic market is done at a constant marginal cost β0 . Furthermore, production for the foreign market has a marginal cost, which is increasing in the quantity for the domestic market. Based on the assumptions above, the firm’s profitmaximization problem is as follows: c c c + pcf t qfc t − C(qdt , qfc t ; γ). max pcdt (qdt , pgdt , wdt ; θ)qdt

(8)

c ,q c qdt ft

Given the sequential nature of the problem, we proceed via backward induction starting with the foreign market, where marginal-cost pricing implies: c αd c αf −1 pcf t = β1 αf (qdt ) (qf t )

qfc t

 =

pcf t c αd β1 αf (qdt )

(9)

 α 1−1 f

.

(10)

We then move to the profit-maximization problem for the domestic market c c c c c max pcdt (qdt , pgdt , wdt ; θ)qdt + pcf t qfc t (qdt ) − C(qdt , qfc t (qdt , pcf t ; z); γ), c

(11)

qdt

where z ≡ (af , ad , β1 ) and pcf t is exogenous. The implied first-order condition that provides the optimal amount of domestic coal production is given by: θ3 wdt − β0 + θ0 +

θ2 pgdt

+

c 2θ1 qdt

ad (β1 − 1) + −1 + af



pcf t af β 1



af −1+af

c (qdt )

1−ad −af −1+af

=0

(12)

In the special case of β1 = 1, which does not compromise the most important feature of the c assumed cost function, we have the following linear equation to solve for qdt : c θ3 wdt − β0 + θ0 + θ2 pgdt + 2θ1 qdt = 0,

(13)

which implies c qdt = H(pgdt , wdt ; θ, γ) ≡

β0 − θ0 − θ2 pgdt − θ3 wdt 2θ1

13

(14)

Given the nature of the profit-maximization problem, knowing the optimal level of domestic production allows us to infer production for the foreign market: qfc t = G(pgdt , wdt , pcf t ; θ, γ).

(15)

Recall that the export supply curve is given by I

I I pIgvt = exp(ηgvt )(qgvt )ωgvt

(16)

Using (9), we establish a link between the domestic and foreign markets using the following I qgvt = qfc t

(17)

I ωgv = αf − 1

(18)

I exp(ηgvt ) = β1 αf (H(pgdt , wdt ; θ, γ))αd

4 4.1

(19)

Empirical Analysis International Trade

We now discuss the results of the structural estimations of import demand and inverse export supply elasticities following Soderbery (2016). This estimator leverages time series variation in prices and quantities within import markets and across export markets. We obtain consistent estimates of the supply and demand elasticities for every exported variety of coal in every importing country via nonlinear SUR (NLSUR). Similar to Feenstra (1994) and Broda and Weinstein (2006), the key identifying assumption in Soderbery (2016) is that once the econometrician controls for good and time effects by first- and reference-country differencing the data, the variety-level errors entering the system of demand and supply equations are uncorrelated. Feenstra’s estimator, which entails 2SLS estimation using variety (country of origin) fixed effects as instruments assuming panel data for different varieties in a given market (importing country), cannot accommodate heterogeneity in export supply elasticities. Soderbery’s estimator, on the other hand, allows us to identify heterogeneous export supply elasticities. It does so by combining the standard system of demand and supply equations for importing countries from Feenstra’s estimator with a system of demand and supply equations for exporters (“exporter system”). The estimator requires that the variety-level errors entering

14

the exporter system are also uncorrelated and it invokes a destination-country differencing.18 The alternative to our NLSUR approach would be to establish demand and supply IVs for every importer-exporter-good combination in the data, which given our model and data is not practically feasible. The only data required for our NLSUR estimation are bilateral trade flows associated with country pairs for the three types of coal, which are readily available from the UN COMTRADE data for 1990–2014. Our estimates are based on data at the HS6 level for 143 importing and 194 exporting countries. Although not all countries trade coal with each other, there are close to 5,650 inverse export supply elasticities and 400 import demand elasticities to be estimated. Recall that the former exhibit variation by origin (exporting I country) and coal type for each importing country (ωgv ) while the latter exhibit variation I 19 by importing country and coal type (σg ) only. To alleviate the computational burden due to the high-dimension of the parameter space and the highly nonlinear nature of the NLSUR optimization problem in hand, we assume countries in the same region have identical supply technologies as in Soderbery (2016) with some adjustments. In particular, major exporting countries are excluded from the regional aggregation.20 Although this is a restrictive assumption, it still allows for heterogeneity in our estimates. Importantly, due to the weighting scheme implemented with the NLSUR estimation, the export supply elasticity for a particular region will be influenced the most by the data on the region’s largest exporter. Applying the estimator requires imports from at least two countries that both export to at least one other destination for a minimum of three periods. Imposing such a requirement translates to utilizing observations associated with 96% of all trade flows (quantity and value) for the three types of coal. I Table 1 presents summary statistics of inverse export supply (ωgv ) and import demand 18

For a succinct illustration of Feenstra’s estimator see Section 2.3 in Soderbery (2015). The issue with Feenstra’s estimator in the case of heterogeneity in export supply elasticities is shown in equations (5) and (6) of Soderbery (2016). Equations (8) and (9) in Soderbery (2016) provide the additional system of demand and supply equations for exporters. Equations (10) and (11) are the NLSUR equations. Note that we apply the Broda and Weinstein (2006) weighting scheme in the NLSUR estimation as in Soderbery (2016) to address measurement error in prices since trade data record unit values. 19 For example, although we estimate a different inverse export supply elasticity for U.S. and Australian bituminous coal for Japan, we estimate a single import demand elasticity for bituminous coal. During 1990– 2014, there were 4–5 varieties of bituminous coal, from different exporting countries that were shipped to an importing country, on average, each year. The average number of varieties of anthracite and other coal are very similar. 20 Table A.6 and the associated note provides information regarding the aggregation discussed here. An implication of our aggregation is that, for example, Mongolia and Vietnam, which are the 11 and 12 largest exporters accounting for a combined 2.4% of total exports during the period we analyze (see Table A.5), have the same export supply elasticities for bituminous coal because they all belong to the region we define as Asia (ASA).

15

(σgI ) elasticities for bituminous coal for major importers excluding inverse export elasticities larger than 20, which corresponds to 2.8% of the roughly 1,100 elasticities in total. We also provide some information about the size of the importing country in terms of GDP and its imports of bituminous coal in dollars and tons.21 Tables A.12 and A.14 in the on-line Appendix provide the same statistics for anthracite and other coal. The standard deviation of the inverse export elasticities highlight the degree of heterogeneity in the curvature of the inverse export supply curves of the exporters serving a particular importer. Due to this heterogeneity, it is difficult to report a single meaningful statistic that pertains to inference. Conversely, in the case of the import demand elasticities, which exhibit variation by importer only for a given coal type, we do report standard errors. Table 2 and on-line Appendix tables A.13 and A.15 provide summary statistics for the inverse export supply elasticities for major exporters along with some information on the size of the exporter similar to the one provided in Table 1. Before discussing our elasticity estimates in detail, it is important to keep in mind that the inverse export supply elasticity, ω, is a measure of importer buyer power. Since ω governs the degree of pass-through of a shock to delivered prices, a large ω implies a high degree of importer buyer power because there is low pass through of any price changes for more inelastic export supply curves. For the largest importer in our sample, Japan, the median ω is 0.26 implying an export elasticity (1/ω) of around 4, such that a 1% increase in the price of bituminous coal leads to a 4% increase in bituminous coal exports to Japan. Among major exporters, the U.S. (Russia) is the country that is the least (most) exposed to Japan’s buyer power with ω values of 0.10 and 0.91, respectively. U.S. and Russia account for a similar fraction, around 4%, of Japan’s imports (tons) between 1990 and 2014. Australia and Indonesia together account for almost 3/4 of Japan’s imports with ω values of 0.259 and 0.299, respectively. India, which is the smallest importer of bituminous coal has a median ω of 0.45 with an implied export elasticity of 2.2. Among major exporters, Indonesia’s ω for bituminous coal exports to India is 0.45 with its counterparts for Australia and the U.S. being very similar (0.47). Indonesia and Australia account for almost 3/4 of India’s imports and the U.S. accounts for around 5%. For China, the third largest importer of bituminous coal, the median ω is 0.05 implying an export elasticity of 20, which means that China has the least buyer power among major importers. Among countries exporting bituminous coal to China, the smallest ω values are 21

To economize on notation in this section, we will use ω and σ to refer to the two elasticities in the remainder of our discussion.

16

for Australia, Indonesia, Kazakhstan, and Mongolia, with values of around 0.05 while the largest ω value, 4.2, is for South Africa. Since Australia, Indonesia, and Mongolia, collectively account for about 3/4 of China’s bituminous coal imports (tons), a plausible explanation of China’s limited buyer power against these countries despite its sheer size is its reliance on imports from these countries. The ω values for other major producers exporting to China are the following: Russia (0.10), Colombia (0.27), and the U.S. (0.51). Although Russia accounts for the rather notable 8% of China’s bituminous coal imports that makes China rather reliant on Russian coal, Colombia and the U.S. account for less than 1%. These patterns in inverse export elasticities are consistent with the findings in Soderbery (2016). According to Table 2, among the largest exporters, Kazakhstan, Poland, and Russia are the countries that are least exposed to importer buyer power with median ω values of 0.13, 0.14, and 0.17, respectively. In the middle of the pack, we see Indonesia, Colombia, U.S, and Australia, with ω values in the 0.26–0.37 range. South Africa, China, and Canada are the exporters most prone to importer buyer power with ω values between 0.50 and 0.98. In the case of the import demand elasticities reported in the rightmost column of Table 1, we see values between 2.26, with a standard error of 2.18 for Russia, and 5.98, with a standard error of 1.66, for Brazil. With the exception of Russia, all demand elasticity estimates are highly significant. The average import demand elasticity across all importers for bituminous coal is 3.6, while that for other coal and anthracite is 3.7 and 3.4. A natural question to ask is how our estimates compare to others in the literature. The median import coal demand elasticity for HS4 2701 in Soderbery (2016), who uses COMMTRADE data for 1991–2007, is 2.9, which is highly comparable with our estimates in Table 1. The mean is 3.1 and the standard deviation is 0.7. His median inverse export elasticity is 0.05 with a mean of 0.40 and a standard deviation of 0.92. Again, these numbers are comparable to the ones we report in Tables 1 and 2. In Broda et al. (2006) (BGW), the median import coal demand elasticity for HS3 270 is 2.9. The mean is 9.2 and the standard deviation is almost 22.7. Although their median estimate is comparable to our estimates in Table 1, we have to keep in mind the different levels of aggregation and different time coverage. BGW use COMTRADE data for 1994–2003 and they don’t aggregate various countries into regions as we and Soderbery (2016) do. The inverse coal export elasticity estimates in Broda et al. (2008) (BLW) for HS4 2701 excluding values above 20 (as in our case) have an average of 0.16 calculated. BLW also use COMTRADE data for 1994–2003 and the inverse export supply elasticities exhibit variation only by exporting country. Once again, the level of aggregation and the different time period make a direct comparison between our estimates and theirs

17

difficult.22

4.2

Domestic Market

The domestic production model generates an equation that relates the estimated export I ), to the U.S. price of gas in (19). In a fully structural model, the supply shock, exp(b ηgvt functional form for H(·) in (14) depends on the functional form of the inverse domestic demand, the production costs, as well as the assumption regarding the model of competition of U.S. coal producers as we discussed in Section 3.2. For the purpose of our counterfactuals, and aiming to allow some flexibility in this important relationship for our counterfactuals, we estimate via OLS the following model: I ηbgvt

= h(·) + ugvt = µIg +

G X

µg pgdt + ugvt ,

(20)

g=1 I is the variety-specific shock to the inverse export supply estimated using our where ηbgvt I I ln(qgvt ), and h(·) is the logarithmic bgv ≡ ln(pIgvt ) − ω model for international trade with ηbgvt transformation of H(·). Furthermore, µIg is an importer-by-coal type fixed effect, pgdt is the U.S. price of gas for which we use an annual average of the Henry Hub benchmark, and µg allows for a slope coefficient that is coal-type specific noting that the annual frequency is I . Furthermore, we expect positive slope due to the COMTRADE data used to obtain ηbgvt coefficients, such that an increase in the U.S. price of gas shift the U.S. export supply curve leftward.

A potential concern about the model in (20) is that we don’t control for U.S. environmental policy for the electric power sector, which is positively correlated with the U.S. price of gas, and would be included in wdt in (19). The positive correlation is due to coal-to-gas switching given that gas is cleaner than coal. Furthermore, the correlation should be fairly strong because the electric power sector has accounted for 25% of the annual U.S. gas consumption, on average, between 1990 and 2014 (EIA, Monthly Energy Review). The dependent variable is the intercept of a constant elasticity inverse export supply curve, which is expected to be negatively correlated with the U.S. environmental policy because, all else equal, a more aggressive environmental policy implies a shift to the right of the inverse export supply curve. However, this relationship is expected to be weak given the long list of factors affecting the 22

Importantly, the average inverse export elasticity we report here from BLW is based on 5 observations. The summary statistics reported for BGW and BLW are based on publicly available files from David Weinstein’s website at: http://www.columbia.edu/~dew35/TradeElasticities/TradeElasticities.html.

18

international market for coal. Therefore, there is a downward, and potentially small, bias in our estimates for the effect of the gas price. The regression in (20) has an R-squared of 0.81, which is not surprising given the importerby-coal type fixed effects, and the slope coefficients have the proper signs. The interactions of the gas price with the bituminous and other coal fixed effects are significant at 5% using bootstrapped standard errors with 1,000 repetitions to account for the estimation error in the supply shocks. The coefficient of the interaction with the anthracite fixed effect is significant at 7% and is about half the size of the other two slope coefficients, which are of similar magnitude.23

5 5.1

Counterfactual Analysis Overview

The counterfactual analysis is based on calculating worldwide trade flows for coal but-for the decrease in the U.S. price of gas due to the shale gas boom. We assess the implications of the decrease in the price of gas by comparing actual and counterfactual values of economic variables of interest, such as prices, quantities, dollar sales, and consumer welfare. In addition, we compare the actual and counterfactual CO2 and SO2 content of trade flows based on the physical characteristics for coal traded around the world. For the remainder of our discussion here, the reader should keep in mind that we abstract from the effects of changes in the consumption of imported coal on the consumption of domestic coal, as well as any substitutability between coal and gas, in importing countries. The underlying reasoning of the counterfactual exercise is straightforward. First, in the absence of the shale gas boom, the gas price in the U.S. is higher. Second, the counterfactual demand for gas (coal) in the U.S. electric power sector is lower (higher) than the actual demand. This is the case because coal and gas are closer substitutes for electric power plants when gas prices are lower even accounting for the fact that it takes a larger amount of heat (MMBtu) generated by using coal than by using gas to generate the same amount of electricity.24 Finally, the increased U.S. domestic demand for coal is served by the domestic supply and plays the role of a negative shock to the U.S. coal export supply curve. 23

Regarding the flexibility of the specification in (20), we experimented with higher-degree polynomials, but nonlinear transformations of the gas price did not seem to matter. 24 Coal-fired electric generating units have higher heat rates (consumption-over-generation) ratios that can be as high as 1.5 times the heat rates of gas-fired units.

19

Our counterfactual essentially studies how the positive shock to the U.S. domestic supply of gas due to the shale gas boom affected the international trade of coal. Additionally, our trade model allows for U.S exports to displace (or be displaced by) exports from other countries in I ) each destination. Having estimated the relationship between the export supply shocks (ηgvt g and the U.S. price of gas (pdt ) in Section 4.2, we can compute counterfactual export supply shocks and simulate the counterfactual trade flows using (19) and the counterfactual U.S. price of gas, pgdt,CF , which we calculate using two alternative approaches. In particular, using pgdt and pget to denote the U.S Henry Hub and the Europe import border gas prices from the World Bank Pink Sheets for 1990–2006, we first calculate counterfactual prices using the following: pgdt,CF1

 p g , dt = λb + λb pg , 0 1 et

t = 1990, ...2006

(21)

t = 2007, ...2014.

We also calculate counterfactual prices using: pgdt,CF 2

 p g , dt = λ b + pg , 0

et

t = 1990, ...2006

(22)

t = 2007, ...2014.

where λb0 and λb1 in (22) are the OLS estimates from the following regression pgdt,CF1 = λ0 + λ1 pget + ut ,

t = 1990, ...2006.

(23)

The second regression should generate identical counterfactual prices if the slope coefficient in the first is not statistically different from 1. As Figure 6 shows, the two counterfactual b1 U.S. gas prices are very similar, which is due to the fact that the confidence interval for λ contains 1. The most notable difference between the actual and counterfactual prices occurs in 2011 and 2012 with the counterfactual prices being almost three times as high as the actual price. The annual price increase, in the absence of the shale gas boom, is somewhere between 136% and 154%, on average, depending on the counterfactual price series. For the counterfactual analysis that follows, we use pgdt,CF2 .25 To calculate the counterfactual global coal trade equilibrium, we first need to calculate the changes in U.S. exports to every importing country and then calculate how competing 25

We also experimented with a specification that included an Asian gas benchmark price, the price of liquefied natural gas in Japan from the World Bank Pink Sheets. Given the substantially higher Asian prices during this period, the counterfactual prices are much higher (up to 9-fold increase) than the ones reported here.

20

exporters respond to changes in the prices and quantities of U.S. coal exports. The trade model from Section 3.1, provides estimates of the import demand (σgI ) and inverse export I ) elasticities. Given our estimates, prices and quantities of coal are driven by the supply (ωgv I and bIgv , respectively, along with the structure export supply and import demand shocks ηgvt I of the import market, captured by the price index (Pgt ). Table 3 and the on-line Appendix tables A.10 and A.11 provide summary statistics for the exponentiated actual and counterfactual supply shocks for bituminous coal, other coal, and anthracite by major importer of U.S. coal. Consistent with the comparative statics discussed earlier, the counterfactual supply shocks are generally higher than the actual ones, such that the counterfactual U.S. exports are smaller than the actual U.S. exports at all price levels. For the remainder of our discussion, and the calculations that follow, we assume that the import demand shocks bIgv are not affected by the counterfactual prices. The first economic variable of interest is the change in the price index for coal imports implied by the change in the U.S. inverse export supply curve, which is derived from the trade model: I ∆ln(Pgt )=

1 ∆η Igt 1+ω Igt

∆η Igt ≡ η Igt,CF − η Igt ,

(24) (25)

where ω Igt and η Igt are quantity-weighted harmonic means of the inverse export supply elasticities and shocks using the actual quantities. We see immediately that the magnitude of the change in the price index will depend on the importance of the change in the U.S. export supply shock in the market overall. With the counterfactual price index in hand, we calculate counterfactual prices and quantities for every exporter and importer using the following differences: ∆ln(pIgvt ) =

I ωgv (σgI − 1) 1 I I ∆η + ∆ln(Pgt ) gvt I I I I 1 + σg ωgv 1 + σg ωgv

(26)

I ∆ln(qgvt )=

−σgI (σgI − 1) I I ∆η + ∆ln(Pgt ) gvt I I 1 + σgI ωgv 1 + σgI ωgv

(27)

I I I ∆ηgvt ≡ ηgvt,CF − ηgvt .

(28)

Non-U.S. exports are only affected by changes in the price index in each importing country I because ∆ηgvt = 0 for non-U.S. coal. U.S. exports are affected by both the shifts in the export supply curve and the resulting impact on the price index. With our counterfactual changes in prices and quantities in each importing country, we can calculate changes in consumer welfare associated with the consumption of imported coal. 21

Given our utility in (1), we can derive the percentage change in welfare in importing country I using: I

∆ln(UtI )

≡

I ln(Ut,CF )

−

ln(UtI )

=

Gt X

I I αgt ∆ln(Pgt ).

(29)

g=1 I We calculate αgt using the share of coal type g in each importer’s total spending on coal imports based on the Cobb-Douglas assumption. From the duality of the CES utility, we can conveniently write the welfare associated with coal imports as the inverse of the CES price index.

5.2

Economic Outcomes

Tables 4–7 show the actual and counterfactual coal dollars, quantities, and prices, as well as the implied percentage change, for different levels of aggregation. Figures 7–9 contain time series of actual and counterfactual values for the same economic variables. The comparison of actual and counterfactual outcomes is limited to the period 2007–2014 and the difference is due to the increase in the U.S. domestic price of gas in the absence of the shale gas boom. As shown in Table 4, we find that in the absence of the shale gas boom, the quantity is 0.35% higher, the dollar value is 1% higher and prices are 1.4% higher for all coal trade flows. These small effects for the world trade flows are not surprising given that the U.S. has not accounted for more than 10% of total exports for the last 15 years. The year with the smallest change in quantity is 2013 (0.14%), while the year with the largest change is 2008 (1.14%). We see the smallest change in dollars in 2009 (0.04%) and the largest in 2012 (3.08%). The smallest change in import price (USD/metric ton) is in 2007 (0.31%), while the largest is in 2012 (2.2%). In the case of major importers of U.S. coal, we see a dollar value that is 38.5% lower in the absence of the shale gas boom, a quantity that is 46% higher, and a price that is 14% higher (Table 5) for U.S.coal. The breakdown by country shows counterfactual figures that are lower than actual figures by as much as 80% for dollar value (Japan) and 82% in quantity (Japan), and counterfactuals that are higher by up to 71% for price (Chile). The average change in dollar value (quantity) is -35% (-46%), while the average change in price 29%. France and Mexico are the countries with the smallest drop in quantity, of 4% and 1%, respectively. The same countries also experience the smallest price changes, about 2%. Morocco is the only country that experiences no changes when we compare actual and counterfactual outcomes.

22

For major importers, the counterfactual dollar value is 0.9% higher, the quantity is 0.3% lower, and the price is 1.25% higher (Table 6) in the absence of the boom. The largest dollar value change is that for Germany (8.5%) followed by that for the Netherlands (3.2%). We see the largest drop in quantity for Brazil (-12.4%) and Canada (-9.1%). The largest increase in quantity is for Ukraine (3%) and Germany (3.3%). Brazil experiences the largest increase in price (12.7%) followed by Canada (10%). For the remaining major importers, we see higher prices by no more than 7% (6.8% for Turkey). Finally, for major exporters, we estimate the dollar value is 1% higher in the absence of the shale gas boom and quantity is 0.4% lower (Table 7), while price is 1.4% higher. For the U.S., quantity is 45% lower while Poland, Colombia, and the Czech Republic, all experience higher exports that range between 7.4% and 10% in the absence of the shale gas boom. The price increase for the U.S. (15%) is the largest with a distant second that for Czech Republic of 3.3%. Other major exporters such as Australia, Indonesia, Russia, and South Africa, experience an increase in their quantities between 2% and 9% and an increase in prices that does not exceed 2.5%.

5.3

Environmental Outcomes

In order to identify the carbon dioxide (CO2 ) and sulfur dioxide (SO2 ) content for the coal trade flows, we need the calorific value (Btu/lb) and sulfur content (percent)—henceforth, specifications—of the various types of coal traded around the world. Ideally, we would like to know the calorific value and sulfur content of anthracite, bituminous, and other coal for each of the exporting countries for 2007–2014, which is a rather demanding task. In what follows, we considered three alternative sources of coal specifications. The first is the annual heat content reported for U.S. coal exports by the EIA. The second is the Platts October 2016 Coal Methodology and Specifications Guide. Platts provides the specifications of standardized coal contracts shipped from (delivered to) major exporting (importing) countries. For example, Platts provides specifications for coal shipped (FOB) from Newcastle, Australia, or Richards Bay, South Africa, under standardized contracts. The same information is available for coal delivered (CIF/CFR) to Japan, Korea, or the AmsterdamRotterdam-Antwerp (ARA) trading hub, which serves major Western European markets such as France, Belgium, Germany, Spain, and the Netherlands. Note that the heat and sulfur content from Platts does not exhibit time variation.26 Our third source is annual 26

See http://www.platts.com/methodology-specifications/coal noting that multiple contracts may pertain to particular exporting or importing country in which case we use an average of the calorific value and sulfur content. Similar information is available from Argus Coal Daily International

23

information for country-specific calorific values in the IEA Coal Information and the Key World Statistics as discussed below. Panel (a) of Figure A.2 shows the heat content (MMBtu/metric ton) for each year in our sample. The heat content for U.S. coal exports is readily available from the EIA. In the case of Platts, we report a quantity-weighted average of heat content for coal originating in the major producing countries listed in the Platts column of the on-line Appendix Table A.16, as well as for coal imported by the major importing countries in the on-line Appendix Table A.17. Additionally, we calculated a quantity-weighted heat content using average calorific values for bituminous coal reported in the IEA Coal Information for the producing countries in Platts. A problem with this calculation is that we cannot distinguish between domestic and imported coal, although the former should dominate the latter given that these countries are major producers. Overall, depending on the source, the heat content is approximately 20–25 MMBtu/metric ton with the lower bound dictated by the exporting countries for which standardized Platts contracts are available. We also report a quantity-weighted heat content using the calorific values reported in the IEA Key World statistics for the producing countries listed in the IEA column of the same table. The country-specific calorific values between 2002 and 2014 are provided in panel (b). There are two distinct features in this figure. First, there is very little variation in the calorific values for a given country across time. Second, there is tiering in calorific values. For example, Kazakhstan and Indonesia consistently produce coal with lower heat content relative to the remaining countries. South Africa, Poland, and Russia are in the middle of the pack while Australia and U.S. appear in the top, which is not surprising given that both countries are the biggest exporters of metallurgical coal. In panel (c) of Figure A.2, we report a quantity-weighted SO2 content (lbs./MMBtu) using sulfur-content information from Platts. Panel (d) of the same figure shows that using Platts information for heat and sulfur content for major exporters, we capture more than 90% of all coal flows during this period, while using the same information for major importers, we capture on average 70% of all coal flows. With the calorific value of coal in hand, the calculation of CO2 content of coal trade flows is straightforward given that there are 211 lbs. of CO2 per MMBtu of coal. The calculation of SO2 lbs per MMBtu of coal is also straightforward once the sulfur content is known. For example, assuming a caloric value of 12,000 Btu/lb, and a sulfur content of 3%, the at http://www.argusmedia.com/coal/argus-coal-daily-international/ and globalCoal at https:// www.globalcoal.com/Brochureware/standardtradingcontract/specifications/.

24

SO2 content of coal is (0.03 × 2)/0.012 = 5.0 lbs./MMBtu.27 Although more detailed calculations allowing for variation in the heat content by year and country are available, assuming 211 lbs. of CO2 per MMBtu of coal and 21 MMBtu per metric ton of coal, the implied actual and counterfactual CO2 content (million metric tons) of all coal trade flows is 14,874 and 14,821, respectively (Table 8). This is a decrease of roughly 0.4%. At a social cost of CO2 (SCC) of $37 per metric ton, the potential environmental damages from emissions due to combustion of imported coal are $550 and $548 billions for the actual and counterfactual, respectively. Hence, damages are about $2 billion in the absence of the shale gas boom. Using 1.3 lbs. of SO2 per MMBtu of coal and 21 MMBtu per metric ton of coal, the implied actual and counterfactual SO2 content for all trade flows are 91.6 and 91.3 million tons, respectively. This represents and SO2 content that is around 0.3% lower in the absence of the shale gas boom. Finally, we calculate the change in welfare based on counterfactual coal trade flows in Table 9. There are two components in our welfare calculation. The first component (product) is the change in surplus due to the consumption of imported coal, which we calculate as the dollar value associated with actual trade flows times the percentage change in prices implied by (29) for each importing country and each year between 2007 and 2014. The second pertains to harmful CO2 emissions (damages) associated with the combustion of imported coal evaluated using the SCC. The damages related to potential CO2 emissions are almost 2 billion dollars less in the absence of the shale gas boom as already discussed, while the surplus from coal consumption is almost 32 billion dollars higher. As a result, welfare in the absence of the boom is 34 billion dollars higher than in the actual scenario. To put this figure in perspective, the actual and counterfactual dollar values of coal trade are 826.5 and 835 billion dollars. We see higher emissions-related damages only in 2012 and 2013 for a total of about 800 million dollars due to higher counterfactual quantities. The largest fraction of this increase is offset by the decrease of 680 million dollars in 2014. Almost 70% of the higher surplus from coal consumption between 2007 and 2014 is due to the higher surplus between 2011 and 2014. 27 Note that 2 is the atomic mass of sulfur dioxide divided by the atomic mass of sulfur. The denominator is due to the fact that there are 106 Btu in a MMBtu.

25

6

U.S. Seaborne Coal Trade Emissions

Finally, we assess the implications of the shale gas boom for CO2 emissions associated with seaborne exports of U.S. coal. We note that Canada and Mexico are the only potential importers of US coal that are accessible by land so we also provide calculations excluding these two countries. Based on the results discussed above, in the absence of the shale gas boom, more U.S. coal would be consumed domestically and less would be exported. As a result, there would be a reduction CO2 emissions associated with seaborne coal trade due to a reduction in the consumption of fuel by marine vessels as we discuss below. Overall, the implications for seaborne emissions deserve attention given that 85% of all coal trade is seaborne and long-range seaborne trade is the only option for two of the largest coal exporters in the world, Australia and Indonesia. For example, for a Panamax vessel with a speed of 14.5 knots/hour, it takes roughly 11 (13) days to get from Newcastle in Australia to Quangzhou in South China (Quinhuangdao in North China) given the distance of 3,885 (4,522) nautical miles. The same ship would take 10 days to get from Baltimore in the U.S. to Rotterdam in the Netherlands. Given the distance traveled and the high CO2 content of residual fuel oil used to propel marine vessels there is a potential for a substantial effect on emissions.28 Our calculations, which are admittedly approximate, are as follows. Residual fuel oil (RFO) has the second highest CO2 emissions factor (after coal) among fossil fuels (173.7 lbs./MMBtu). An annual time series of RFO consumption by marine vessels is available from EIA. The RFO MMBtu content is also available from EIA. Hence, we have an annual series of vessel-related O,total RFO emissions (emissRF ). We want to allocate the total vessel RFO fuel consumption t to coal exports and we do so in two steps. In the first step, we compute the annual share of U.S. waterborne freight that is associated with exports using data from the U.S. Department of Transportation (sft re,exp ). In the second step, we compute the annual (weight) share of coal in U.S. exports from the Census export database maintained by Peter Schott at Yale (scoal,exp ). Hence, we have: t O,coal emissRF t

exports

coal|exp

O,total = emissRF × sft re,exp × st t

28

,

(30)

The following vessel types are used in coal transportation: handysize 40–45,000 deadweight tonnage (DWT), Panamax: 60–80,000 DWT and Capesize 80,000 DWT.http://www.worldcoal.org. For vessel speeds, see: http://hudsonshipping.com/?q=node/95. For distance calculations, see http://www. sea-distances.org/.

26

which allows us to calculate counterfactual emissions due to RFO combustion as follows O,coal ∆emissRF t coal|exp

where sCF,t

exports

coal|exp

O,total = emissRF × sft re,exp × (st t

coal|exp

− sCF,t

)

(31)

is the counterfactual share of coal in US exports.

Figure 10 shows the actual and counterfactual CO2 emissions related to U.S. coal marine coal|exp trade. When we exclude Canada and Mexico, we adjust st but not sft re,exp for the actual calculations due to data availability. The actual seaborne CO2 emissions for 2007–2014 are 15.8 (13.3) million metric tons, while the counterfactual are 8 (5) million when including (excluding) Canada and Mexico. At a SCC of $37/ton, the implied total SCC in billion USD is 584 (0.493) for the actual and 298 (182) for the counterfactual emissions. According to our calculations, there were 896 (412) million metric tons of actual (counterfactual) CO2 emissions associated with imports of U.S. coal excluding Canada and Mexico. If we include the two major trading partners, there were 1,091 (588) million metric tons of actual (counterfactual) CO2 emissions associated with imports of U.S. coal. Hence, seaborne emissions account somewhere between 1.2% and 1.5% of coal-trade related emissions. These bounds are based on counterfactual and actual emissions excluding Canada and Mexico.

7

Conclusion

The paper analyzes the impact of the U.S. shale gas boom on global carbon emissions associated with international coal trade flows. In particular, we analyze whether the increase in U.S. coal exports following the boom has contributed to an increase in coal imports around the world such that the reduction in domestic carbon emissions due to coal-to-gas switching is offset by an increase in carbon emissions elsewhere. We build a structural model that links the domestic to the international coal market employing techniques from industrial organization (I.O.) and international trade. Recently developed techniques in international trade allow us to estimate a large number of heterogeneous inverse export supply and import demand elasticities that play a key role in our analysis. The first-order conditions of a stylized I.O. model for the U.S. domestic coal market allows us to link shocks in the U.S. inverse export supply curve to the domestic gas price. We construct counterfactual U.S. gas prices for 2007–2014 using a simple linear regressions that links the gas price in the U.S. to the gas price in Europe using data for 1990–2006. We use our structural model to simulate counterfactual international coal trade flows in the

27

absence of the shale gas boom. We then convert trade flows into potential carbon dioxide (CO2 ) emissions using information on the heat and CO2 content of coal. According to our findings, the shale gas boom had a substantial impact on major importers of U.S. coal, which experience a drop of 46% in quantity and an increase of 14% price in the absence of the shale gas boom. The effect of the shale gas boom on world trade flows is a decrease of -0.4% in quantity and an increase of 1.4% in price. In the absence of the U.S. shale gas boom, consumer welfare—after accounting for environmental damages due to CO2 emissions from the combustion of imported coal—increases by close to 34 billion dollars.

References Afsah, S., and K. Salcito. 2014. “US Coal Exports Erode All CO2 Savings from Shale Gas.” CO2 Scorecard. Arezki, R., T. Fetzer, and F. Pisch. 2016. “On the Comparative Advantage of U.S. Manufacturing: Evidence from the Shale Gas Revolution.” CEP Discussion Paper No 1454. Armington, P.S. 1969. “A Theory of Demand for Products Distinguished by Place of Production.” Staff Papers-International Monetary Fund, 16(1): 159–178. Bartik, A W., J. Currie, M. Greenstone, and C.R. Knittel. 2017. “The Local Economic Consequences of Hydraulic Fracturing.” Working Paper. Broda, C., N. Lim ao, and D.E. Weinstein. 2008. “Optimal Tariffs and Market Power: The Evidence.” American Economic Review, 98(5): 2032–2065. Broda, C., J. Greenfield, and D.E. Weinstein. 2006. “From Groundnuts to Globalization: A Structural Estimate of Growth and Trade.” NBER Working Paper 12512. Broda, C., and D.E. Weinstein. 2006. “Globalization and the Gains from Variety.” Quarterly Journal of Economics, 121(2): 541–85. Considine, T.J. 2013. “Powder River Basin Coal: Powering America.” Natural Resources, 4 514–533. Cristea, A., D. Hummels, L. Puzzello, and M. Avetisyan. 2013. “Trade and the greenhouse gas emissions from international freight transport.” Journal of Environmental Economics and Management, 65 153–173. 28

Cullen, J.A., and E.T. Mansur. 2014. “Inferring Carbon Abatement Costs in Electricity Markets: A Revealed Preference Approach Using the Shale Revolution.” NBER Working Paper 20795. Ernst, and Young. 2013. “U.S. Coal Exports: National and State Economic Contributions.” Prepared for the National Mining Association. Eyer, J. 2014. “The Effect of the Shale Gas Boom on Greenhouse Gas Emissions: Accounting for Coal Exports.” Working Paper. Feenstra, R.C. 1994. “New Product Varieties and the Measurement of International Prices.” American Economic Review, 84(1): 157–177. Frankel, J.A., and A.K. Rose. 2005. “Is Trade Good or Bad for the Environment? Sorting out the Causality.” Review of Economics and Statistics, 87(1): 85–91. Hausman, C., and R. Kellogg. 2015. “Welfare and Distributional Implications of Shale Gas.” NBER Working Paper 21115. Jackson, R.B., A. Vengosh, J.W. Carrey, R.J. Davies, T.H. Darrah, F. O’Sullivan, and G. Petron. 2014. “The Environmental Costs and Benefits of Fracking.” Annual Review of Environmental and Resource Economics(39): 329–362. Kirkpatrick, C., and S. Scrieciu. 2008. “Is Trade Liberalization Bad for the Environment? A review of the economic evidence.” Journal of Environmental Planning and Management, 51(4): 497–510. Knittel, C.R., K. Metaxoglou, and A. Trindade. 2015. “Natural Gas Prices and Coal Displacement: Evidence from Electricity Markets.” NBER Working Paper 21627. Knittel, C.R., K. Metaxoglou, and A. Trindade. 2016. “Are we fracked? The impact of falling gas prices and the implications for coal-to-gas switching and carbon emissions.” Oxford Review of Economic Policy, 32(2): 241–259. Li, R., R. Joyeux, and R.D. Ripple. 2010. “International Steam Coal Market Integration.” Energy Journal, 31(3): 181–202.

29

Linn, J., L. Muehlenbachs, and Y. Wang. 2014. “How Do Natural Gas Prices Affect Electricity Consumers and the Environment.” Resources for the Future Discussion Paper 14-19. Ritschel, W., and H-W. Schiffer. 2007. “Wolrd Market for Hard Coal.” RWE Power. Schernikau, L. 2010. Economics of the International Coal Trade.: Springer. Shapiro, J. 2016. “Trade Costs, CO2, and the Environment.” American Economic Journal: Economic Policy, 8(4): 220–254. Soderbery, A. 2015. “Estimating Import Demand and Supply Elasticities: Analysis and Implications.” Journal of International Economics, 96(1): 1–17. Soderbery, A. 2016. “Trade Elasticities, Heterogeneity, and Optimal Tariffs.” Working Paper. Wolak, F.A. 2016. “Assessing the Impact of the Diffusion of Shale Oil and Gas Technology on the Global Coal Market.” Working Paper. Zaklan, A., A. Cullman, A. Neumann, and C. von Hirschhausen. 2012. “The globalization of steam coal markets and the role of logistics: An empirical analysis.” Energy Economics, 34 105–116.

30

Tables and Figures

31

Table 1: Inverse export supply and import demand elasticities: Bituminous coal, major importers

Imports Importer Coal

Inverse Export Supply

Import Demand

GDP

Value

Quantity

Mean

Median

Std. Dev

Est

S.E.

JPN

BIT

4.340

294.588

3559.822

0.294

0.259

0.149

4.355

0.000

KOR

BIT

0.888 129.183

1672.996

0.519

0.365

0.356

4.274

0.074

CHN

BIT

2.668

91.282

887.908

0.290

0.049

0.962

3.344

0.435

DEU

BIT

2.907

46.744

544.566

0.520

0.416

0.470

4.487

0.201

GBR

BIT

2.345

39.004

440.822

0.632

0.113

0.789

4.625

1.751

ITA

BIT

1.845

32.338

334.061

0.498

0.496

0.062

2.536

0.205

NLD

BIT

0.658

19.711

285.105

0.145

0.157

0.041

2.424

0.018

ESP

BIT

1.224

10.370

151.077

0.092

0.098

0.095

2.755

0.023

FRA

BIT

2.231

12.165

141.650

1.178

0.277

1.239

3.280

0.317

BRA

BIT

1.068

15.817

127.505

0.769

0.450

1.580

5.977

1.656

RUS

BIT

0.987

2.292

18.221

2.772

0.132

4.991

2.262

2.811

IND

BIT

0.906

1.196

14.971

0.368

0.445

0.194

2.697

0.000

Note: GDP value for 2006 in current USD (trillion). Import values are in billion USD and quantities are in million metric tons for 1990–2014. All statistics are quantity-weighted using UN COMTRADE data. The import demand elasticity exhibits no variation by importer and coal type.

Table 2: Inverse export supply elasticities: Bituminous coal, major exporters

Exports Exporter Coal

Inverse Export Supply

GDP

Value

Quantity

Mean Median

Std. Dev

AUS

BIT

1.231

339.954

3792.082

0.299

0.259

0.311

IDN

BIT

1.295

95.083

1355.568

0.235

0.262

0.167

RUS

BIT

0.919

89.034

956.550

0.732

0.910

0.825

USA

BIT

0.610

88.154

829.846

0.628

0.164

1.826

CAN

BIT

1.644

70.903

702.655

0.589

0.349

0.451

COL

BIT

1.098

42.487

591.635

0.276

0.276

0.195

CHN

BIT

1.585

32.394

585.873

0.382

0.275

0.156

ZAF

BIT

1.154

35.099

567.991

0.892

0.189

1.250

POL

BIT

0.707

13.494

216.960

0.244

0.139

0.154

KAZ

BIT

0.821

2.167

20.397

0.142

0.132

0.022

Note: GDP value for 2006 in current USD (trillion). Import values are in billion USD and quantities are in million metric tons for 1990–2014. All statistics are quantity-weighted using UN COMTRADE data. The import demand elasticity exhibits no variation by importer and coal type.

32

Table 3: Supply shocks for major importers of U.S coal: bituminous coal

Actual Country Coal Imports

Counterfactual

Mean

Median

Std.Dev.

Mean

Median

Std.Dev.

JPN

BIT

209.334

10.396

6.390

8.032

16.569

6.390

21.123

BRA

BIT

113.558

0.035

0.036

0.010

0.066

0.062

0.033

ITA

BIT

100.914

0.000

0.000

0.000

0.001

0.001

0.001

GBR

BIT

110.167

76.838

81.086

29.481

139.910

159.566

92.451

DEU

BIT

146.424

0.002

0.002

0.001

0.003

0.003

0.001

NLD

BIT

82.972

2.339

1.446

1.286

3.778

1.446

3.404

KOR

BIT

90.516

0.000

0.000

0.000

0.001

0.001

0.001

ESP

BIT

41.699

8.828

6.647

5.682

13.594

6.647

17.346

TUR

BIT

57.224

0.001

0.000

0.000

0.001

0.001

0.001

BEL

BIT

31.865

2.320

1.542

1.528

4.017

1.542

4.680

CHN

BIT

98.376

0.002

0.001

0.010

0.003

0.003

0.010

IND

BIT

6.972

0.013

0.007

0.082

0.027

0.014

0.161

PRT

BIT

27.062

7.595

6.037

2.960

10.962

6.037

9.472

UKR

BIT

28.273

25.814

26.170

7.066

57.270

51.715

17.333

CHL

BIT

28.501

0.006

0.005

0.002

0.013

0.014

0.003

Note: All statistics are quantity-weighted using actual quantities from the UN COMTRADE data. Imports are in million metric tons.

Table 4: Counterfactual analysis: economic outcomes, world 2007–2014

Coal Value

Coal Quantity

% change

Coal Price

Year

actual

CF

actual

CF

% change

actual

CF

2007

57.536

57.431

-0.183

757.457

753.738

-0.491

75.960

76.195

% change 0.310

2008

97.802

97.955

0.156

752.626

744.007

-1.145

129.948

131.658

1.316

2009

79.809

79.837

0.035

746.000

739.788

-0.833

106.983

107.919

0.875

2010

94.144

93.908

-0.251

806.681

800.177

-0.806

116.705

117.359

0.560

2011

139.864 141.249

0.990

940.417

937.477

-0.313

148.726

150.669

1.307

2012

134.643 138.787

3.078

1081.249

1090.478

0.853

124.525

127.272

2.206

2013

118.824 121.512

2.262

1173.095

1174.722

0.139

101.291

103.439

2.120

2014

103.909 104.282

0.359

1143.131

1133.943

-0.804

90.898

91.964

1.172

TOTAL

826.531 834.960

1.020

7400.659

7374.329

-0.356

111.683

113.225

1.380

Note: Values in billion USD, quantities are in million metric tons, and prices in USD/metric ton. CF stands for counterfactual.

33

Table 5: Counterfactual analysis: economic outcomes, major importers of U.S. coal 2007–2014

Coal Value Country

Actual

CF

Coal Quantity

Coal Price

% Change

Actual

CF

% Change

Actual

CF

CAN

5.997

5.997

-0.003

81.668

72.525

-11.196

73.434

82.690

% Change 12.604

DEU

7.625

4.646

-39.068

58.437

29.068

-50.258

130.488

159.842

22.495

BRA

8.930

5.088

-43.027

55.707

22.635

-59.369

160.307

224.782

40.220

GBR

6.342

1.728

-72.754

54.615

14.367

-73.694

116.129

120.277

3.572

ITA

5.557

3.965

-28.653

37.043

18.936

-48.882

150.007

209.370

39.573

CHN

4.512

2.374

-47.380

32.173

13.026

-59.514

140.228

182.253

29.969

JPN

6.067

1.234

-79.663

30.302

5.501

-81.847

200.232

224.327

12.034

NLD

3.760

2.610

-30.601

28.501

14.243

-50.026

131.932

183.214

38.870

KOR

4.840

2.391

-50.596

28.384

11.098

-60.900

170.515

215.449

26.352

TUR

3.509

2.439

-30.497

21.536

10.925

-49.274

162.954

223.273

37.016

IND

3.902

3.868

-0.886

20.862

20.395

-2.241

187.057

189.650

1.386

FRA

2.524

2.524

-0.019

16.995

16.989

-0.036

148.530

148.556

0.018

MEX

1.673

1.672

-0.037

15.152

15.150

-0.012

110.424

110.396

-0.025

UKR

3.431

1.554

-54.718

15.032

4.104

-72.700

228.256

378.603

65.868

CHL

1.180

0.760

-35.604

12.580

4.733

-62.373

93.810

160.547

71.141

MAR

1.179

1.179

0.000

11.122

11.122

0.000

105.966

105.966

0.000

BEL

1.650

0.964

-41.544

10.683

4.315

-59.607

154.421

223.474

44.717

ESP

1.088

0.431

-60.347

6.681

1.828

-72.645

162.793

235.981

44.957

PRT

0.507

0.244

-51.978

5.286

1.561

-70.460

95.977

156.027

62.566

74.274 45.667

-38.516

542.758 292.519

-46.105

136.846

156.117

14.082

TOTAL

Note: Values in billion USD, quantities are in million metric tons, and prices in USD/metric ton based on imports of U.S. coal only. CF stands for counterfactual. Entries are sorted in descending order of actual quantity.

34

Table 6: Counterfactual analysis: economic outcomes, major importers 2007–2014

Coal Value Country

Actual

CF

Coal Quantity

% Change

Actual

CF

Coal Price

% Change

JPN

193.597 193.452

-0.075

1464.862 1465.394

0.036

CHN

124.406 126.385

1.591

1291.275 1306.543

1.182

KOR

100.453 100.725

0.270

921.565

921.803

0.007

815.282

815.213

IND

91.508

91.514

Actual

CF

132.161 132.014 96.344

% Change -0.111

96.733

0.404

0.026

109.003 109.269

0.244

-0.008

112.241 112.257

0.015

DEU

40.869

44.331

8.472

334.144

345.176

3.302

122.308 128.430

5.005

GBR

35.465

35.035

-1.212

305.789

292.087

-4.481

115.978 119.947

3.422

ITA

24.395

24.801

1.664

184.565

176.700

-4.261

132.177 140.358

6.189

113.256 118.519

NLD

18.065

18.651

3.241

159.508

157.365

-1.343

THA

9.564

9.562

-0.014

153.336

153.337

0.001

62.370

62.361

-0.015

RUS

4.601

4.601

-0.000

151.194

151.194

-0.000

30.429

30.429

0.000

148.434 167.309

12.716

BRA

21.354

21.076

-1.302

143.864

125.973

-12.436

MYS

12.808

12.808

-0.001

136.041

136.041

0.000

94.147

94.146

79.334

4.646

-0.001

USA

10.276

10.276

-0.000

129.523

129.523

0.000

79.334

0.000

TUR

16.257

16.562

1.881

113.463

108.222

-4.619

143.277 153.040

6.814

UKR

15.108

15.527

2.770

102.352

105.472

3.048

147.610 147.212

-0.270

CAN

7.631

7.631

0.002

100.090

90.950

-9.131

76.237

83.899

10.051

HKG

7.322

7.322

-0.000

97.891

97.891

-0.000

74.793

74.793

-0.000

FRA

11.512

11.512

0.004

83.526

83.523

-0.004

137.826 137.836

0.007

ESP

7.621

7.802

2.378

68.641

70.552

2.784

111.025 110.586

-0.395

ISR

3.890

3.890

-0.000

48.836

48.836

-0.000

756.700 763.462

0.894

6805.746 6781.795

-0.352

TOTAL

79.653

79.653

0.000

111.185 112.575

1.250

Note: Values in billion USD, quantities are in million metric tons, and prices in USD/metric ton. CF stands for counterfactual. Entries are sorted in descending order of actual quantity.

35

Table 7: Counterfactual analysis: economic outcomes, major exporters 2007–2014

Coal Value Country

Actual

CF

Coal Quantity

% Change

Actual

CF

Coal Price

% Change

AUS

289.239 302.318

4.522

2092.253 2156.323

3.062

IDN

148.537 151.823

2.213

1825.252 1853.640

1.555

Actual

CF

138.243 140.201 81.379

81.905

% Change 1.416 0.647

RUS

90.480

98.465

8.825

791.248

843.699

6.629

114.351 116.706

2.060

USA

82.155

51.608

-37.183

581.746

315.057

-45.843

141.222 163.805

15.991

COL

45.126

51.134

13.314

476.619

526.778

10.524

97.069

2.525

ZAF

39.384

40.773

3.526

387.468

396.217

2.258

101.644 102.905

1.240

CAN

45.294

49.356

8.966

274.290

290.316

5.843

165.134 170.006

2.951

VNM

13.285

13.288

0.027

173.135

173.157

0.013

76.730

76.741

0.014

KAZ

4.897

5.062

3.365

160.722

161.475

0.469

30.470

31.348

2.883

CHN

17.301

17.462

0.931

134.671

135.405

0.545

128.465 128.959

0.384

MNG

7.170

7.560

5.449

108.697

113.821

4.714

65.960

66.423

0.701

PRK

5.887

5.887

0.000

69.227

69.227

-0.000

85.043

85.043

0.000

UKR

5.831

6.006

2.999

65.880

67.269

2.108

88.517

89.289

0.872

POL

6.906

7.633

10.523

54.883

58.958

7.425

125.833 129.463

2.885

CZE

4.886

5.598

14.578

30.584

33.934

10.952

159.742 164.962

3.268

VEN

2.484

2.727

9.793

22.953

24.494

6.711

108.201 111.326

2.888

CHE

1.764

1.764

0.000

20.301

20.301

-0.000

NLD

1.700

1.832

7.749

15.228

15.942

NZL

2.795

2.830

1.263

15.073

15.245

815.120 823.125

0.982

7300.230 7271.257

TOTAL

94.679

86.880

86.880

0.000

4.688

111.662 114.927

2.924

1.139

185.405 185.632

0.122

-0.397

111.657 113.203

1.384

Note: Values in billion USD, quantities are in million metric tons, and prices in USD/metric ton. CF stands for counterfactual. Entries are sorted in descending order of actual quantity.

36

Table 8: Counterfactual analysis: environmental outcomes, world 2007–2014

CO2 Emissions

CO2 Social Cost

SO2 Emissions Actual

Year

Actual

CF

Actual

CF

CF

2007

1522.391

1514.915

56.328

56.052

9.380

9.334

2008

1512.681

1495.358

55.969

55.328

9.320

9.213

2009

1499.363

1486.878

55.476

55.014

9.238

9.161

2010

1621.325

1608.251

59.989

59.505

9.989

9.909

2011

1890.116

1884.208

69.934

69.716

11.645 11.609

2012

2173.171

2191.718

80.407

81.094

13.389 13.503

2013

2357.770

2361.039

87.237

87.358

14.527 14.547

2014

2297.546

2279.077

85.009

84.326

14.155 14.042

TOTAL

14874.363

14821.444

550.351

548.393

91.643 91.317

Note: Emissions associated with coal imports in million metric tons. The social cost is measured in billion USD assuming $37 per metric ton.

Table 9: Counterfactual analysis: accounting for environmental damages, world 2007–2014

Coal Value

Coal Quantity

Damages

∆ Welfare (CF-actual)

Year

actual

CF

actual

CF

actual

CF

Damages Product

Total

2007

57.536

57.431

757.457

753.738

56.328

56.052

-0.277

0.398

0.674

2008

97.802

97.955

752.626

744.007

55.969

55.328

-0.641

2.738

3.379

2009

79.809

79.837

746.000

739.788

55.476

55.014

-0.462

1.535

1.997

2010

94.144

93.908

806.681

800.177

59.989

59.505

-0.484

1.996

2.480

2011

139.864

141.249

940.417

937.477

69.934

69.716

-0.219

6.676

6.895

2012

134.643

138.787

1081.249 1090.478

80.407

81.094

0.686

8.874

8.188

2013

118.824

121.512

1173.095 1174.722

87.237

87.358

0.121

6.487

6.366

2014

103.909

104.282

1143.131 1133.943

85.009

84.326

-0.683

3.268

3.952

TOTAL

826.531

834.960

7400.659 7374.329

550.351

548.393

-1.958

31.973

33.931

Note: values, damages, and consumer welfare in billion USD, quantities are in million metric tons, and prices in USD/metric ton. CF stands for counterfactual. The value and quantity panels are identical to the ones in Table 4. The Damages column in the Welfare Change (∆ Welfare) panel refers to environmental damages due to the CO2 content of coal assuming 211 lbs. per MMBtu of coal 21 MMBtu per metric ton of coal, and a social cost of CO2 of $37/metric ton of coal. The product column in the ∆ Welfare panel refers to change in consumer welfare associated with the consumption of imported coal.

37

Figure 1: U.S. coal exports and domestic price of natural gas 20

steam coal exports Henry Hub gas price

18

12

16 14

10

12 10

8

8

$/MMBtu

million metric tons tons

14

6

6 4

4

2 0

2 2002

2004

2006

2008

2010

2012

2014

Note: the quarterly gas price reported in this figure is an average of EIA monthly prices for the Henry Hub benchmark. The quarterly exports of coal are from the EIA International Energy Statistics.

Figure 2: U.S. coal production, consumption, and exports

1,450 1,400

14

Prodcution Consumption Exports %

12

1,350 million short tons

10 1,300 1,250

8

1,200 6

1,150 1,100

4

1,050 1,000

percent of U.S. production

1,500

2

950 900

0 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

Note: Production and consumption based on data from the EIA monthly coal production and international energy statistics, respectively. Exports based on EIA and Census data.

38

Figure 3: Coal markets overview 0.17

7,500

0.16 7,000

million metric tons

0.15 6,500

0.14 0.13

6,000

0.12 5,500 0.11 5,000 0.10 4,500

0.09 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 2012

(a) Consumption 0.60

USA

CAN

COL

CHN

AUS

IDN

ZAF

RUS

POL

0.60

0.50 share of world imports

share of world exports

0.50

(b) Trade/Consumption Ratio

0.40

0.30

0.20

KOR

CHN

IND

BRA

NLD

ITA

GBR

DEU

0.40

0.30

0.20

0.10

0.10

0.05

0.05

0.00

JPN

0.00 1990

1995

2000

2005

2010

2015

(c) Major exporters

1990

1995

2000

2005

2010

2015

(d) Major importers

Note: panels (a) and (b) are based on data from the EIA International Energy Statistics. Panels (c) and (d) are based on UN COMTRADE import data.

39

Figure 4: World trade by type of coal 1,400

160

ANT BIT OTH ALL

1,200

140 120

billion USD

1,000 million metric tons

ANT BIT OTH ALL

800 600

100 80 60

400

40

200

20

0

0 1990

1995

2000

2005

2010

2015

1990

1995

100

100

90

90

80

80

70

70

60 ANT BIT OTH

50

2005

2010

2015

2005

2010

2015

(b) value

% of metric tons

% of metric tons

(a) quantity

2000

40

60

40

30

30

20

20

10

10

0

ANT BIT OTH

50

0 1990

1995

2000

2005

2010

2015

1990

1995

(c) quantity, % 160

2000

(d) value, % ANT BIT OTH ALL

140

USD/metric ton

120 100 80 60 40 20 0 1990

1995

2000

2005

2010

2015

(e) price

Note: based on import file for HS6 codes: 270111 (anthracite (ANT)), 270112 (bituminous (BIT)), 270119 (other (OTH)).

40

Figure 5: Coal prices per metric ton 240

200

200

180

180

160

160

140 120 100

140 120 100

80

80

60

60

40

40

20

20

0

0 1990

1995

2000

2005

2010

2015

1990

(a) coal prices for importers: Asia 240

200

1995

2000

2005

2010

2015

(b) coal prices for importers: Europe 240

USA CAN COL

220

CHN AUS IDN ZAF RUS POL

220 200

180

180

160

160

USD/metric ton

USD/metric ton

NLD ITA GBR DEU ESP

220

USD/metric ton

USD/metric ton

240

CHN IND JPN KOR

220

140 120 100

140 120 100

80

80

60

60

40

40

20

20

0

0 1990

1995

2000

2005

2010

2015

(c) coal prices for exporters: Americas

1990

1995

2000

2005

2010

2015

(d) coal prices exporters: Other

Note: Based on UN COMTRADE import prices. Panel (a) shows the import price for coal paid by various Asian countries. Panel (b) shows the import price for coal paid by various European countries. Panel (c) shows the import price for coal originating in the Americas. Panel (d) shows the import price for coal originating in other parts of the world.

41

Figure 6: Counterfactual analysis: U.S. gas price 2007–2014 16

US Henry Hub US Henry Hub Cf1

14

US Henry Hub Cf2 Europe border import

USD/MMBtu

12 10 8 6 4 2 0 1990

1995

2000

2005

2010

2015

Note: The annual average of the U.S. Henry Hub, the Europe border import, and the Japan LNG gas prices are from the World Bank Pink Sheets. The maroon and green lines are the counterfactual Henry Hub gas prices under the two alternative calculations discussed in Section 5.

42

Figure 7: Counterfactual analysis: economic outcomes, major importers of US coal 2007–2014 220

110

actual counterfactual

200

90 Metric tons (millions)

180 USD per metric ton

actual counterfactual

100

160 140 120 100

80 70 60 50 40

80

30

60

20

40 1990

1995

2000

2005

2010

2015

1990

1995

(a) price 16

2000

2005

2010

2015

(b) quantity 55

actual counterfactual

value quantity price

50

14

45 40

10

% change

USD (billions)

12

8 6

35 30 25 20 15

4

10 2

5 1990

1995

2000

2005

2010

2015

(c) dollars

2007

2008

2009

2010

2011

2012

(d) % change (abs.value)

43

2013

2014

Figure 8: Counterfactual analysis: economic outcomes, major coal importers 2007–2014 160

1100

actual counterfactual

actual counterfactual

1000

140

Metric tons (millions)

USD per metric ton

900 120

100

80

800 700 600 500 400

60 300 40

200 1990

1995

2000

2005

2010

2015

1990

1995

(a) price 3.0

2010

2015

value quantity price

2.5

100

2.0

80 % change

USD (billions)

2005

(b) quantity

actual counterfactual

120

2000

60

1.5

40

1.0

20

0.5

0

0.0 1990

1995

2000

2005

2010

2015

(c) dollars

2007

2008

2009

2010

2011

2012

(d) % change (abs.value)

44

2013

2014

Figure 9: Counterfactual analysis: economic outcomes, major coal exporters 2007–2014 160

1200

actual counterfactual

actual counterfactual

1100

140 1000 Metric tons (millions)

USD per metric ton

120 100 80 60

900 800 700 600 500 400

40

300 200

20 1990

1995

2000

2005

2010

2015

1990

1995

(a) price 140

2000

2005

2010

2015

(b) quantity 3.0

actual counterfactual

120

value quantity price

2.5

100 % change

USD (billions)

2.0 80 60

1.5

1.0 40 0.5

20 0

0.0 1990

1995

2000

2005

2010

2015

(c) dollars

2007

2008

2009

2010

2011

2012

(d) % change (abs.value)

45

2013

2014

Figure 10: Counterfactual analysis: environmental outcomes, CO2 emissions due to U.S. seaborne coal trade 0.24

actual actual excl. CAN & MEX counterfactual counterfactual excl. CAN & MEX

0.22 0.20 0.18 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 2007

2008

2009

2010

2011

2012

2013

2014

2013

2014

(a) share of coal in total exports 3.5

actual actual excl. CAN & MEX counterfactual counterfactual excl. CAN & MEX

3.0

million metric tons

2.5 2.0 1.5 1.0 0.5 0.0 2007

2008

2009

2010

2011

(b) CO2 emissions

46

2012

A A.1

Appendix Additional information on U.S. coal exports

Regarding the export split between metallurgical and steam coal, the share of metallurgical coal in annual U.S. exports to China was between 68% and 86% for 2009–2012. Metallurgical coal accounted for around 73% of U.S. exports to Italy and Spain, and close to 50% of U.S. exports to Germany (49%) and the Netherlands (53%) during 2007–2012. The share of metallurgical coal was 38% for the UK. During the same time, metallurgical coal accounted for 84% of U.S. exports to India and 73% of U.S. exports to China. The share of metallurgical coal in U.S. exports to Japan and Korea was 88% and 54%, respectively. The vast majority of metallurgical coal was exported to China from Baltimore and Norfolk during 2007–2012, due to their proximity to the Appalachian region. The same three ports accounted for the vast majority of total (steam plus metallurgical) exports to India and Europe. Most of the steam coal exported to China was shipped from New Orleans, Seattle, or Los Angeles. New Orleans is close to the barges on the Mississippi river moving steam coal and shipping routes to Europe and South America. Seattle and Los Angeles are among the closest ports to the Western region. The largest fraction of (metallurgical plus steam) coal to Europe was shipped from one of the three Eastern ports or Houston. Between 2002 and 2014, Norfolk, Baltimore, and New Orleans accounted for about 86% of total coal exports, 67% of steam coal exports and close to 93% of metallurgical coal exports.

A.2

World trade of bituminous coal

To give the reader an idea about the primary destinations of bituminous coal that accounts for most of world coal trade for major exporters, Japan (57%) and Korea (17%) account for about 3/4 of Australia’s exports with China being a distant third at 9.6%. Indonesia’s export split is similar to Australia’s: Japan (32%) and Korea (25%) account for 60% of its exports. Russia’s export destinations are rather diverse, which should not be surprising given its geographic spread and the fact that transportation costs are an important factor for coal trade: Japan (17%), Great Britain (13%), Ukraine (13%), Korea (11%), Turkey (9%). Canada (21%) and Japan (12%) account for 1/3 of U.S. exports, with roughly another quarter accounted for by Italy (8%), Great Britain (7%), Germany (6%), and the Netherlands (6%). As for other major exporters of bituminous coal, About a third of Colombia’s exports are to the U.S. (28%), and another 22% is roughly equally split between Germany (12%) and Great 47

Britain (10%). China (15%) and Hong Kong (11)% account for about another 1/4. Seven European countries account for 64% of South Africa’ exports, with Japan (9%) and Korea (7%) accounting for another 16%. Close to 80% of Kazakhstan’s exports of bituminous coal, which is of primary interest in our empirical analysis, is to Russia (55%) and Ukraine (26%) during 1990–2014. Ten European countries account for 90% of Poland’s export with Germany alone accounting for 41%. Regarding the primary sources of bituminous coal for major importers, Australia (61%) and Indonesia (12%) together account for almost 3/4 of Japan’s imports of bituminous coal. Australia (40%) and Indonesia (20%) also account for most of Korea’s imports with China (16%) and Canada (10%) accounting for roughly 25%. Australia (41%), Indonesia (22%) and Mongolia (12%) collectively account for about 3/4 of China’s bituminous coal imports (tons). Germany import bituminous coal from a rather diverse set of countries: South Africa (19%), Poland (16.5%), Russia (14%), Colombia (13%), U.S. (13% and Australia (12%). Russia provides more than a quarter of Great Britain’s bituminous coal (27%) while Australia (18%), U.S. (17%), and Colombia (13%), accounting collectively account for about half of the country’s imports.

48

Table A.1: ISO Alpha-3 country codes

Code

Country

AUS

AUSTRALIA

BEL

BELGIUM

BRA

BRAZIL

CAN

CANADA

CHE

SWITZERLAND

CHL

CHILE

CHN

CHINA

COL

COLOMBIA

CZE

CZECH REPUBLIC

DEU

GERMANY

ESP

SPAIN

FRA

FRANCE

GBR

UNITED KINGDOM

HKG

HONG KONG

HUN

HUNGARY

IDN

INDONESIA

IND

INDIA

ISR

ISRAEL

ITA

ITALY

JPN

JAPAN

KAZ

KAZAKHSTAN

KOR

SOUTH KOREA

MAR

MOROCCO

MEX

MEXICO

MNG

MONGOLIA

MYS

MALAYSIA

NLD

NETHERLANDS

NZL

NEW ZEALAND

POL

POLAND

PRK

NORTH KOREA

PRT

PORTUGAL

RUS

RUSSIAN FEDERATION

THA

THAILAND

TUR

TURKEY

UKR

UKRAINE

USA

UNITED STATES

VEN

VENEZUELA

VNM

VIET NAM

ZAF

SOUTH AFRICA

49

Table A.2: Major countries

% Coal Quantity exports imports 99.13

94.02

% Coal Value

imports USA

exports

94.82

99.19

imports imports USA 94.50

93.99

Note: Based on UN COMTRADE data for 1990–2014. The table shows the percentage of total exports, total imports, and imports of U.S. coal that 35 major countries account for. For example, the 35 countries we consider account for 99.13% of total exports. Quantities are in million metric tons and values are in billion USD.

50

Table A.3: List of major countries, sorted by total coal imports (tons)

Coal Quantity %

Coal Value %

Country

Exports

Imports

JPN

0.0111

25.0691

Imports USA Exports Imports 9.9318

0.0121

KOR

0.0084

11.9404

4.7300

0.0121

11.6492

5.4954

CHN

4.8564

9.2107

2.2726

3.7243

10.5327

3.4924

IND

0.0484

7.1772

1.6598

0.0419

9.0150

3.3336

25.8091

Imports USA 10.2482

DEU

0.0823

4.7630

5.7217

0.1358

5.1503

7.2129

GBR

0.1317

4.4861

6.7320

0.1619

4.6963

6.8576

NLD

0.2696

2.9286

5.2655

0.2680

2.6106

4.9294

ITA

0.0114

2.8471

6.9603

0.0144

3.1795

7.4102

RUS

8.4389

2.5252

0.3251

9.6917

0.5771

0.8357

BRA

0.0065

2.3996

9.7133

0.0080

2.7657

10.9245

CAN

4.8682

2.1948

21.0420

6.0597

1.3940

11.2455

USA

9.2217

2.1934

0.0000

10.4921

1.6472

0.0000

ESP

0.0673

1.8958

3.6811

0.1032

1.3919

2.6636

FRA

0.0344

1.8896

3.8489

0.0534

1.9485

3.8245

TUR

0.0021

1.7784

3.1747

0.0038

2.1739

3.9561

THA

0.0024

1.4299

0.0044

0.0030

0.9712

0.0057

HKG

0.0102

1.4269

0.0000

0.0068

0.9310

0.0001

MYS

0.0188

1.3450

0.0144

0.0216

1.2783

0.0075

UKR

0.6925

1.2442

1.0864

0.6523

1.6306

2.6799

ISR

0.0196

1.0974

0.1633

0.0133

1.2429

0.0885

BEL

0.1349

0.8504

2.3137

0.1550

0.8462

2.2277

CHL

0.0180

0.7638

0.9771

0.0129

0.4731

0.9750

PRT

0.0015

0.7219

1.4970

0.0014

0.5420

1.0103

POL

2.2394

0.4643

0.2131

1.7278

0.5450

0.5386

MEX

0.0060

0.3643

1.7338

0.0144

0.3938

1.6457

MAR

0.0001

0.2666

1.1151

0.0001

0.2335

1.0771

HUN

0.0057

0.1970

0.4317

0.0047

0.2655

0.9256

CZE

0.6484

0.1929

0.0425

0.7175

0.1797

0.1072

ZAF

7.3667

0.1484

0.1297

6.0006

0.2281

0.2307

IDN

17.0008

0.0513

0.0023

14.5532

0.0603

0.0021

NZL

0.2417

0.0409

0.0000

0.3495

0.0262

0.0001

KAZ

2.7125

0.0284

0.0000

0.6253

0.0170

0.0000

CHE

0.2192

0.0259

0.0004

0.2378

0.0302

0.0004

VEN

0.7000

0.0235

0.0336

0.5434

0.0268

0.0405

VNM

1.7033

0.0184

0.0000

1.3930

0.0240

0.0000

AUS

29.2627

0.0137

0.0014

34.0939

0.0160

0.0007

MNG

0.7521

0.0020

0.0000

0.5892

0.0007

0.0000

COL

5.8012

0.0000

0.0000

5.3523

0.0000

0.0002

PRK

0.5510

0.5241

Note: based on UN COMTRADE data for HS6 codes 270111, 270112, 270119 for 1990–2014.

51

Table A.4: List of major countries, sorted by imports of US coal (tons)

Coal Quantity % Country

Exports

Imports

CAN

4.8682

2.1948

Coal Value %

Imports USA Exports Imports 21.0420

Imports USA

6.0597

1.3940

11.2455

JPN

0.0111

25.0691

9.9318

0.0121

25.8091

10.2482

BRA

0.0065

2.3996

9.7133

0.0080

2.7657

10.9245

ITA

0.0114

2.8471

6.9603

0.0144

3.1795

7.4102

GBR

0.1317

4.4861

6.7320

0.1619

4.6963

6.8576

DEU

0.0823

4.7630

5.7217

0.1358

5.1503

7.2129

NLD

0.2696

2.9286

5.2655

0.2680

2.6106

4.9294

KOR

0.0084

11.9404

4.7300

0.0121

11.6492

5.4954

FRA

0.0344

1.8896

3.8489

0.0534

1.9485

3.8245

ESP

0.0673

1.8958

3.6811

0.1032

1.3919

2.6636

TUR

0.0021

1.7784

3.1747

0.0038

2.1739

3.9561

BEL

0.1349

0.8504

2.3137

0.1550

0.8462

2.2277

CHN

4.8564

9.2107

2.2726

3.7243

10.5327

3.4924

MEX

0.0060

0.3643

1.7338

0.0144

0.3938

1.6457

IND

0.0484

7.1772

1.6598

0.0419

9.0150

3.3336

PRT

0.0015

0.7219

1.4970

0.0014

0.5420

1.0103

MAR

0.0001

0.2666

1.1151

0.0001

0.2335

1.0771

UKR

0.6925

1.2442

1.0864

0.6523

1.6306

2.6799

CHL

0.0180

0.7638

0.9771

0.0129

0.4731

0.9750

HUN

0.0057

0.1970

0.4317

0.0047

0.2655

0.9256

RUS

8.4389

2.5252

0.3251

9.6917

0.5771

0.8357

POL

2.2394

0.4643

0.2131

1.7278

0.5450

0.5386

ISR

0.0196

1.0974

0.1633

0.0133

1.2429

0.0885

ZAF

7.3667

0.1484

0.1297

6.0006

0.2281

0.2307

CZE

0.6484

0.1929

0.0425

0.7175

0.1797

0.1072

VEN

0.7000

0.0235

0.0336

0.5434

0.0268

0.0405

MYS

0.0188

1.3450

0.0144

0.0216

1.2783

0.0075

THA

0.0024

1.4299

0.0044

0.0030

0.9712

0.0057

IDN

17.0008

0.0513

0.0023

14.5532

0.0603

0.0021

AUS

29.2627

0.0137

0.0014

34.0939

0.0160

0.0007

CHE

0.2192

0.0259

0.0004

0.2378

0.0302

0.0004

HKG

0.0102

1.4269

0.0000

0.0068

0.9310

0.0001

NZL

0.2417

0.0409

0.0000

0.3495

0.0262

0.0001

COL

5.8012

0.0000

0.0000

5.3523

0.0000

0.0002

KAZ

2.7125

0.0284

0.0000

0.6253

0.0170

0.0000

VNM

1.7033

0.0184

0.0000

1.3930

0.0240

0.0000

MNG

0.7521

0.0020

0.0000

0.5892

0.0007

0.0000

PRK

0.5510

USA

9.2217

1.6472

0.0000

0.5241 2.1934

0.0000

10.4921

Note: based on UN COMTRADE data for HS6 codes 270111, 270112, 270119 for 1990–2014.

52

Table A.5: List of major countries, sorted by total coal exports (tons)

Coal Quantity % Country

Exports

Imports

AUS

29.2627

0.0137

Coal Value %

Imports USA Exports Imports 0.0014

34.0939

0.0160

Imports USA 0.0007

IDN

17.0008

0.0513

0.0023

14.5532

0.0603

0.0021

USA

9.2217

2.1934

0.0000

10.4921

1.6472

0.0000

RUS

8.4389

2.5252

0.3251

9.6917

0.5771

0.8357

ZAF

7.3667

0.1484

0.1297

6.0006

0.2281

0.2307

COL

5.8012

0.0000

0.0000

5.3523

0.0000

0.0002

CAN

4.8682

2.1948

21.0420

6.0597

1.3940

11.2455

CHN

4.8564

9.2107

2.2726

3.7243

10.5327

3.4924

KAZ

2.7125

0.0284

0.0000

0.6253

0.0170

0.0000

POL

2.2394

0.4643

0.2131

1.7278

0.5450

0.5386

VNM

1.7033

0.0184

0.0000

1.3930

0.0240

0.0000

MNG

0.7521

0.0020

0.0000

0.5892

0.0007

0.0000

VEN

0.7000

0.0235

0.0336

0.5434

0.0268

0.0405

UKR

0.6925

1.2442

1.0864

0.6523

1.6306

2.6799

0.1929

0.0425

0.7175

0.1797

0.1072

2.6106

4.9294

CZE

0.6484

PRK

0.5510

NLD

0.2696

2.9286

5.2655

0.5241 0.2680

NZL

0.2417

0.0409

0.0000

0.3495

0.0262

0.0001

CHE

0.2192

0.0259

0.0004

0.2378

0.0302

0.0004

BEL

0.1349

0.8504

2.3137

0.1550

0.8462

2.2277

GBR

0.1317

4.4861

6.7320

0.1619

4.6963

6.8576

DEU

0.0823

4.7630

5.7217

0.1358

5.1503

7.2129

ESP

0.0673

1.8958

3.6811

0.1032

1.3919

2.6636

IND

0.0484

7.1772

1.6598

0.0419

9.0150

3.3336

FRA

0.0344

1.8896

3.8489

0.0534

1.9485

3.8245

ISR

0.0196

1.0974

0.1633

0.0133

1.2429

0.0885

MYS

0.0188

1.3450

0.0144

0.0216

1.2783

0.0075

CHL

0.0180

0.7638

0.9771

0.0129

0.4731

0.9750

ITA

0.0114

2.8471

6.9603

0.0144

3.1795

7.4102

JPN

0.0111

25.0691

9.9318

0.0121

25.8091

10.2482

HKG

0.0102

1.4269

0.0000

0.0068

0.9310

0.0001

KOR

0.0084

11.9404

4.7300

0.0121

11.6492

5.4954

BRA

0.0065

2.3996

9.7133

0.0080

2.7657

10.9245

MEX

0.0060

0.3643

1.7338

0.0144

0.3938

1.6457

HUN

0.0057

0.1970

0.4317

0.0047

0.2655

0.9256

THA

0.0024

1.4299

0.0044

0.0030

0.9712

0.0057

TUR

0.0021

1.7784

3.1747

0.0038

2.1739

3.9561

PRT

0.0015

0.7219

1.4970

0.0014

0.5420

1.0103

MAR

0.0001

0.2666

1.1151

0.0001

0.2335

1.0771

Note: based on UN COMTRADE data for HS6 codes 270111, 270112, 270119 for 1990–2014. The table reads as follows: Australia (AUS) accounts for 29.3% of total exports in terms of tons and 53 accounts for 25.07% of all imports in terms 34% of total exports in terms of USD. Japan (JPN) tons and for 25.81% in terms of USD.

Table A.6: Regions and trade: coal exports 1990–2014

Country/

County

BIT

ANT

Region

Count

Quantity

AUS

1

5062.613

397.790

13.245

0.657

IDN

1

2109.747

108.964

17.785

RUS

1

1330.593

86.543

USA

1

1248.570

ZAF

1

CHN

Value Quantity

OTH Value

Quantity

Value

0.634

1188.287

59.371

122.146

8.888

27.251

1.019

112.521

9.699

0.836

99.284

3.785

1031.450

56.518

11.971

0.931

22.199

0.993

1

840.725

39.888

91.756

7.495

10.008

0.381

COL

1

831.062

45.535

0.154

0.014

0.272

0.033

CAN

1

770.804

68.295

0.165

0.006

0.533

0.036

POL

1

82.117

5.016

1.147

0.070

229.267

10.069

KAZ

1

15.638

1.385

1.071

0.015

262.598

4.409

GBR

1

11.031

1.306

4.681

0.581

0.589

0.049

FRA

1

1.026

0.104

0.413

0.046

1.847

0.204

MEX

1

0.575

0.076

0.024

0.007

0.000

0.000

DEU

1

0.252

0.059

1.102

0.306

19.516

2.699

IND

1

0.230

0.012

0.033

0.003

33.040

1.529

ITA

1

0.196

0.020

0.249

0.057

0.235

0.026

BRA

1

0.078

0.011

0.019

0.004

0.000

0.000

JPN

1

0.036

0.011

0.005

0.003

0.026

0.010

NWU

13

145.629

11.507

16.217

1.999

44.953

3.189

SEU

17

69.202

5.695

55.256

4.477

59.400

3.526

SAM

10

64.159

2.452

1.005

0.108

3.887

0.095

ASA

30

38.519

2.079

51.351

3.907

38.984

1.747

OCE

2

25.043

1.399

0.001

0.000

0.044

0.007

AFR

22

1.735

0.308

6.558

0.258

0.929

0.071

CAR

8

0.000

0.000

0.000

0.000

0.000

0.000

Note: Trade volume for bituminous coal (BIT), anthracite (ANT), and other coal (OTH). Quantities in million metric tons and values in billion USD. The lower part of the table contains information for the following regions: Asia (ASA), Southern/Eastern Europe (SEU), Northern/Western Europe (NWU), South America (SAM), Africa (AFR), Oceania (OCE), and the Caribbean (CAR). See Table A2 in Soderbery (2016) for the assignment of countries to regions subject to the following changes: SAM excludes COL, ASA excludes IDN and KAZ, SEU excludes POL, AFR excludes ZAF.

54

Table A.7: Regions and trade: coal imports 1990–2014

Country/

County

BIT

ANT

OTH

Region

Count

Quantity

Value

Quantity

Value

Quantity

Value

JPN

1

3576.286

295.541

107.062

10.203

166.068

12.088

KOR

1

1686.981

130.139

95.539

10.342

50.955

2.975

CHN

1

1109.503

103.000

302.404

21.186

221.514

16.985

DEU

1

556.147

47.682

37.365

3.813

137.861

11.929

GBR

1

473.020

40.677

17.553

1.451

198.275

15.706

ITA

1

335.594

32.462

6.641

0.653

94.940

6.040

NLD

1

290.584

20.187

8.949

0.666

150.157

11.296

USA

1

264.489

15.647

4.958

0.398

67.361

4.240

CAN

1

254.590

13.695

9.836

0.964

72.597

2.508

FRA

1

251.957

20.484

32.806

3.051

5.383

0.461

ESP

1

153.489

10.600

13.085

0.966

124.534

5.575

BRA

1

130.485

16.057

26.494

2.004

211.485

15.997

MEX

1

55.047

4.762

0.625

0.078

0.271

0.010

RUS

1

18.461

2.329

20.528

0.354

348.753

4.424

IND

1

15.006

1.200

7.304

1.310

1079.769

108.508

AUS

1

0.097

0.008

0.941

0.121

1.064

0.068

SEU

19

452.839

44.237

60.228

4.624

130.210

7.917

ASA

33

418.794

34.898

27.457

2.850

642.867

47.776

NWU

14

292.393

21.894

29.080

3.676

127.222

7.298

SAM

15

119.844

10.537

1.073

0.140

53.430

0.808

AFR

30

7.321

0.849

24.850

0.551

100.820

7.669

OCE

3

6.219

0.544

0.553

0.084

5.473

0.250

CAR

11

5.670

0.624

0.228

0.033

0.043

0.005

Note: Trade volume for bituminous coal (BIT), anthracite (ANT), and other coal (OTH). Quantities in million metric tons and values in billion USD. The lower part of the table contains information for the following regions: Asia (ASA), Southern/Eastern Europe (SEU), Northern/Western Europe (NWU), South America (SAM), Africa (AFR), Oceania (OCE), and the Caribbean (CAR). See Table A2 in Soderbery (2016) for the assignment of countries to regions subject to the following changes: ASA excludes KOR, NWU excludes NLD, and SEU excludes ESP.

55

Table A.8: Asia Pacific and Atlantic thermal coal assessments I

Contract

Cal.Value I Cal. Value II Sulfur %

Ash %

Moisture %

Vol. Matter %

CFR Guangzhou

3,600-4,000

NAR

Max 1

Max 10

Max 40

N/A

CFR Guangzhou

4,500-5,900

NAR

Max 1

Max 12

Max 30

N/A

CFR Guangzhou

5,300-5,700

NAR

Max 1

Max 23

Max 18

Max 40

CFR India East

3,600-4,000

GAR

Max 0.6

Max 8

Max 41

N/A

CFR India East

4,000-4,400

GAR

Max 1

Max 8

Max 41

N/A

CFR India East

4,800-5,200

GAR

Max 1

Max 8

Max 41

N/A

CFR India East

5,300-5,700

NAR

Max 1

Max 8

Max 41

N/A

CFR India East

N/A

N/A

Max 1

Max 8

Max 41

N/A

CFR India West

3,600-4,000

GAR

Max 0.6

Max 8

Max 41

N/A

CFR India West

4,000-4,400

GAR

Max 1

Max 8

Max 41

N/A

CFR India West

4,800-5,200

GAR

Max 1

Max 8

Max 41

N/A

CFR India West

5,300-5,700

NAR

Max 1

Max 8

Max 41

N/A

CFR India West

N/A

N/A

Max 1

Max 8

Max 41

N/A

CFR South China

5,300-5,700

NAR

Max 1

Max 23

Max 18

Max 40

CIF ARA

5,800-6,100

NAR

Max 1

Max 16

Max 14

N/A

CIF Japan

5,850-6,250

NAR

Max 1

Max 14

Max 15

Max 30

CIF Korea

5,850-6,250

NAR

Max 1

Max 17

Max 15

Max 30

CIF Turkey

5,850-6,300

NAR

0.5-1

6-15

15-Oct

N/A

Note: The assessments are from the Platts Coal Methodology and Specifications Guide for October 2016. The difference between net- and gross-as-received energy values is the latent heat of the water vapor, which lowers the effective calorific value of the coal. Mositure and Ash reduces net calorific value.

56

Table A.9: Asia Pacific and Atlantic thermal coal assessments II

Contract

Cal.Value I Cal. Value II

Sulfur %

Ash %

Moisture %

Vol. Matter %

FOB ARA Barge

5,800-6,100

NAR

Max 1

Max 16

Max 14

N/A

FOB Colombia

5,750-6,100

NAR

Max 0.9

Max 12

Max 15

N/A

FOB Gladstone

6,300-6,700

GAR

Max 0.6

Max 12

Max 10

Max 30

FOB Kalimantan

3,600-4,000

GAR

Max 0.6

Max 9

Max 41

N/A

FOB Kalimantan

4,000-4,400

GAR

Max 1

Max 10

Max 40

N/A

FOB Kalimantan

4,800-5,200

GAR

Max 1

Max 12

Max 30

N/A

FOB Kalimantan

5,700-6,100

GAR

Max 1

Max 15

Max 20

N/A

FOB Kalimantan

5,700-6,100

GAR

Max 1

Max 15

Max 20

N/A

FOB Newcastle

5,300-5,700

NAR

Max 0.75

17-23

Max 15

N/A

FOB Newcastle

6,100-6,500

GAR

Max 0.75

Max 14

Max 15

27-35

FOB Newcastle

6,100-6,500

GAR

Max 0.75

Max 15

Max 15

27-35

FOB Newcastle

6,100-6,500

GAR

Max 0.75

Max 16

Max 15

27-35

FOB Newcastle

6,100-6,500

GAR

Max 0.75

Max 17

Max 15

27-35

FOB Poland

5,800-6,100

NAR

Max 0.8

Max 16

Max 14

N/A

FOB Qinhuangdao

4,800-5,200

NAR

Max 1

Max 25

Max 18

Max 40

FOB Qinhuangdao

5,300-5,700

NAR

Max 1

Max 20

Max 18

Max 40

FOB Qinhuangdao

5,300-5,700

NAR

Max 1

Max 14

Max 12

Max 30

FOB Qinhuangdao

6,100-6,300

GAR

Max 0.8

Max 10

Max 12

Max 30

FOB Richards Bay

5,300-5,700

NAR

Max 1

17-23

Max 13

N/A

FOB Richards Bay

5,800-6,100

NAR

Max 1

Max 16

Max 12

N/A

FOB Russia Baltic

5,800-6,100

NAR

Max 1

Max 16

Max 14

N/A

FOB Russia Pacific

6,200-6,400

GAR

Max 0.4

Max 15

Max 14

Max 30

Note: The assessments are from the Platts Coal Methodology and Specifications Guide for October 2016. The difference between net- and gross-as-received energy values is the latent heat of the water vapor, which lowers the effective calorific value of the coal. Mositure and Ash reduces net calorific value.

57

Table A.10: Supply shocks for major importers of U.S coal: other

Actual

Counterfactual

Country

Coal

Imports

Mean

Median

Std.Dev.

Mean

Median

Std.Dev.

CAN

OTH

144.962

0.540

0.437

6.899

0.740

0.553

6.917

BRA

OTH

97.697

10.487

7.657

6.141

14.260

7.657

16.292

ITA

OTH

29.108

0.003

0.002

0.261

0.004

0.002

0.526

FRA

OTH

0.504

61.147

55.678

12.765

71.108

55.678

42.939

MEX

OTH

4.538

0.012

0.002

0.285

0.017

0.002

0.347

CHL

OTH

4.082

78.801

80.886

5.001

202.528

187.314

25.712

Note: All statistics are quantity-weighted using actual quantities from the UN COMTRADE data. Imports are in million metric tons.

Table A.11: Supply shocks for major importers of U.S coal: anthracite

Actual

Counterfactual

Country

Coal

Imports

Mean

Median

Std.Dev.

Mean

Median Std.Dev.

ITA

ANT

1.202

57.046

60.187

8.084

57.046

60.187

8.084

GBR

ANT

6.302

0.240

0.170

1.259

0.248

0.170

1.789

NLD

ANT

0.433

5.997

5.361

3.314

7.273

5.361

5.307

KOR

ANT

1.584

0.400

0.411

0.333

0.590

0.622

0.426

ESP

ANT

0.106

3.682

2.869

214.204

5.332

4.088

214.275

TUR

ANT

0.085

2.081

1.108

5.466

2.233

1.108

6.715

BEL

ANT

7.517

3.218

3.510

1.189

4.950

5.313

2.339

CHN

ANT

0.068

12.886

5.023

41.785

18.151

8.348

47.142

MEX

ANT

0.013

0.000

0.000

0.000

0.000

0.000

0.000

IND

ANT

0.002

2.687

2.196

2.155

3.393

3.324

2.295

CHL

ANT

0.112

0.069

0.011

0.188

0.094

0.011

0.233

Note: All statistics are quantity-weighted using actual quantities from the UN COMTRADE data. Imports are in million metric tons.

58

Table A.12: Inverse export supply and import demand elasticities: Other coal, major importers

Imports Importer

Coal

Inverse Export Supply

Import Demand

GDP

Value

Quantity

Mean

Median

Std. Dev

Est

S.E.

RUS

OTH 0.987

4.366

348.099

0.007

0.007

0.001

3.622

0.091

BRA

OTH

1.068

9.468

124.118

0.108

0.094

0.103

3.249

0.000

DEU

OTH

2.907

7.417

89.097

0.464

0.096

0.648

3.844

0.150

ITA

OTH 1.845

4.570

72.253

0.322

0.483

0.252

3.340

0.143

JPN

OTH

4.340

2.620

31.774

0.065

0.017

0.075

4.140

0.088

FRA

OTH

2.231

0.108

1.800

0.440

0.414

0.450

4.158

0.000

Note: GDP value for 2006 in current USD (trillion). Import values are in billion USD and quantities are in million metric tons for 1990–2014. All statistics are quantity-weighted using UN COMTRADE data. The import demand elasticity exhibits no variation by importer and coal type.

Table A.13: Inverse export supply elasticities: Other coal, major exporters

Exports

Inverse Export Supply

Exporter

Coal

GDP

Value

Quantity

Mean

Median

Std. Dev

KAZ

OTH

1.416

4.340

347.463

0.007

0.007

0.008

USA

OTH 0.707

9.401

156.443

0.149

0.094

0.102

IDN

OTH

4.525

3.354

93.175

0.298

0.353

0.464

ZAF

OTH 3.751

2.736

50.687

0.089

0.096

0.122

COL

OTH 5.036

3.261

47.867

0.482

0.485

0.018

RUS

OTH 5.573

4.064

42.587

0.067

0.000

0.120

POL

OTH

3.706

2.234

32.154

0.667

0.903

0.352

CAN

OTH

5.114

2.061

29.557

0.040

0.000

0.100

CHN

OTH

2.664

1.541

16.639

0.017

0.017

0.000

Note: GDP value for 2006 in current USD (trillion). Import values are in billion USD and quantities are in million metric tons for 1990–2014. All statistics are quantity-weighted using UN COMTRADE data. The import demand elasticity exhibits no variation by importer and coal type.

59

Table A.14: Inverse export supply and import demand elasticities: Anthracite, major importers

Imports Importer

Coal

Inverse Export Supply

Import Demand

GDP

Value

Quantity

Mean

Median

Std. Dev

Est

S.E.

RUS

OTH 0.987

4.366

348.099

0.007

0.007

0.001

3.622

0.091

BRA

OTH

1.068

9.468

124.118

0.108

0.094

0.103

3.249

0.000

DEU

OTH

2.907

7.417

89.097

0.464

0.096

0.648

3.844

0.150

ITA

OTH 1.845

4.570

72.253

0.322

0.483

0.252

3.340

0.143

JPN

OTH

4.340

2.620

31.774

0.065

0.017

0.075

4.140

0.088

FRA

OTH

2.231

0.108

1.800

0.440

0.414

0.450

4.158

0.000

Note: GDP value for 2006 in current USD (trillion). Import values are in billion USD and quantities are in million metric tons for 1990–2014. All statistics are quantity-weighted using UN COMTRADE data. The import demand elasticity exhibits no variation by importer and coal type.

Table A.15: Inverse export supply elasticities: Anthracite, major exporters

Exports Exporter

Coal

Inverse Export Supply

GDP

Value

Quantity Mean Median Std. Dev

CHN

ANT 1.066

9.223

101.129

1.541

0.401

2.342

RUS

ANT

1.043 10.978

100.644

0.415

0.450

0.186

AUS

ANT

1.596

8.111

67.971

0.205

0.024

0.770

ZAF

ANT

1.230

2.682

34.775

1.124

0.626

1.089

KAZ

ANT

0.495

0.139

17.984

0.359

0.359

0.009

USA

ANT

0.721

1.773

15.818

3.435

2.556

4.090

POL

ANT

0.681

0.314

3.560

0.510

0.450

0.149

IDN

ANT

1.395

0.134

2.010

0.283

0.287

0.053

CAN

ANT

4.229

0.023

0.346

0.592

0.759

0.352

COL

ANT

0.184

0.009

0.071

0.268

0.325

0.108

Note: GDP value for 2006 in current USD (trillion). Import values are in billion USD and quantities are in million metric tons for 1990–2014. All statistics are quantity-weighted using UN COMTRADE data. The import demand elasticity exhibits no variation by importer and coal type.

60

Table A.16: coal heat and sulfur dioxide content: major exporters

Heat

SO2

Country

MMBtu/metric ton

lbs./MMBtu

AUS

24.12

1.18

CAN

19.84

1.11

CHN

22.02

1.52

COL

23.81

1.39

IDN

19.27

1.52

IND

15.08

0.88

POL

23.81

1.30

RUS

24.41

0.73

USA

23.44

2.44

ZAF

22.82

1.55

Note: The heat content is an average of the heat content from Platts FOB contracts for coal originating in the countries listed in the leftmost column with the exception of the U.S. for which we report an average of the annual heat rates for coal exports reported by the EIA. The SO2 content is an average of the sulfur content for Platts FOB contracts.

Table A.17: coal heat and sulfur dioxide content: major importers

Heat

SO2

Country

MMBtu/metric ton

lbs./MMBtu

ARA

23.81

1.30

CHN

19.12

1.64

IND

20.17

1.59

JPN

24.13

1.10

KOR

24.13

1.10

TUR

23.81

1.48

Note: The heat content is an average of the heat content for coal in Platts CFR/CIF contracts for coal delivered in the countries listed in the leftmost column with the exception The SO2 content is an average sulfur content for Platts CFR/CIF contracts. ARA refers to the Platts contracts for the Amsterdam-Roterdam-Antwerp (ARA) hub, which we use for Western European Countries (NLD, DEU, GBR, ITA).

61

62

Note: Reserves in million ktoe from World Energy Council for 2011.

Figure A.1: Global recoverable coal reserves

Figure A.2: Coal specifications 28 27 26

35

25

MMBtu/metric ton

MMBtu/metric ton

40

Platts Exports IEA Key World IEA Coal Info BIT USA Exports Platts Imports

24 23 22 21

30

AUS

CHN

COL

DEU

IDN

IND

KAZ

POL

RUS

UKR

USA

ZAF

25

20

15

20 10 1990

1995

2000

2005

2010

2015

2002

(a) Heat content: MMBtu/metric ton 1.6

2004

2006

2008

2010

2012

2014

(b) Heat content: MMBtu/metric ton 100

Platts Exports Platts Imports

95

1.5

90

1.4

80 percent

lbs./MMBtu

85

1.3

75 70 65

1.2

60 55

1.1

Platts Exports Platts Imports

50 1990

1995

2000

2005

2010

(c) SO2 content: lbs./MMbtu

2015

1990

1995

2000

2005

2010

2015

(d) Share of coal flows with heat and sulfur content information

Note: See Section 5.3 for additional details.

63