ENERGY INTENSIVE MANUFACTURING INDUSTRIES AND GHG EMISSIONS

Abstract

This paper analyzes changes in U.S. energy-intensive, trade-exposed (EITE) manufacturing over the past decade, through the lens of previously proposed climate policy measures. The American Clean Energy and Security Act of 2009 defined measures and thresholds for EITE eligibility and proposed compensatory allowances designed to reduce negative competitive impacts to domestic industry and to prevent emissions leakage. We undertook a retrospective analysis of the 2009 eligibility criteria, using the same methods with more recent data to examine trends over the 2004–2017 period. We find that energy intensity, emissions intensity, output, and emissions have fluctuated with economic conditions, and defining measures and thresholds that remain informative is challenging. Had ACES been enacted as written and not revised, the number of sectors qualifying for rebates would have decreased from 39 to 26, after adjustment for the changes in North American Industry Classification System definitions. Emissions from the eligible sectors fell 26% across the three periods of analysis, while emissions from manufacturing as a whole fell 5%. We decompose the changes in emissions into scale and intensity measures based on a hybrid measure derived from Grossman and Krueger [(1993). Environmental impacts of a North American free trade agreement. In The US-Mexico Free Trade Agreement, PM Garber (ed.). Cambridge, MA: MIT Press] and Kaya and Yokoburi [(1997). Environment, Energy, and Economy: Strategies for Sustainability. Tokyo: United Nations University Press]. As an alternative, we perform the same analyses using the EPA’s Greenhouse Gas Reporting Program data. These data, not available when ACES was written, offer annual greenhouse gas estimates for facilities that emit more than 25,000 tons CO₂e annually. Finally, we draw some recommendations for future policy including (1) using measures that make price level adjustments straightforward or unnecessary, (2) keeping EITE policy focused on a small group of industries to minimize sectoral reclassification problems, (3) identifying industries prone to emissions leakage rather than just changes in output and (4) consider spatial heterogeneity of emissions and trade patterns.

1. Introduction

Domestic regulations can have global economic and greenhouse gas (GHG) emissions consequences. The pollution havens hypothesis (Copeland and Taylor, 2004; Levinson and Taylor, 2008) suggests that environmental regulations put domestic producers at a competitive disadvantage in global markets, resulting in the relocation of manufacturing capacity to countries with limited regulation. In the case of GHG emissions, this can undercut the effectiveness of the regulation if emissions leakage results in redistribution rather than reduction of global pollutants. Policymakers and academics, recognizing these issues, have proposed various kinds of rebates, exemptions, border taxes and other measures to address energy-intensive, trade-exposed (EITE) competitiveness issues and emissions leakage. Canada, the largest U.S. trading partner, has implemented emissions taxes with rebates based on emission intensity benchmarks for EITE industries with emissions over 50,000 tons CO₂e (Kaufman et al., 2020). Differing levels of national climate ambition under the Paris climate agreement have led to a resurgence of interest in border carbon adjustments to protect EITE industry (Prag, 2020). For instance, the EU announced its Border Carbon Adjustment Mechanism in June 2021. President Biden’s 2021 Trade Agenda includes “consideration of border carbon adjustments” … “as appropriate, and consistent with domestic approaches to reduce U.S. greenhouse gas emissions” as a means of avoiding emissions leakage (Executive Office of the President, 2021).

Much of this recent work on EITE competitiveness and emissions leakage is based implicitly on the definitions, criteria and policy proposals contained in the American Clean Energy and Security Act of 2009 (ACES), cap and trade legislation which passed the U.S. House of Representatives.² ACES popularized the term “energy intensive, trade-exposed” (and its acronym “EITE”), and provided specific criteria for determining EITE sector eligibility including threshold values for energy intensity, trade exposure and GHG emissions intensity.³

The purpose of this paper is to contribute to a fuller discussion of policy measures for addressing emissions leakage and competitiveness impacts through a retrospective analysis of EITE sectors. We reproduce the EITE eligibility analysis from ACES, using the same methods with more recent data to examine trends in energy intensity and emissions over the 2004–2017 period. We also decompose the causal factors of change in those measures. As an alternative, we estimate GHG emissions measures using the EPA Greenhouse Gas Reporting Program (GHGRP) data. These data, not available when ACES was written, offer annual GHG estimates for most sectors of the economy. The paper proceeds as follows. Section 2 reviews provisions of ACES and selected follow-on literature. Section 3 presents methods for establishing EITE eligibility and calculating emissions. Results are presented in Sec. 4. In Sec. 5, we compare data from the GHGRP. A discussion of the implications for future policy efforts concludes.

2. EITE Analysis and the American Clean Energy and Security Act

As a cap and trade proposal, ACES created GHG emission allowances and a set of rules governing their distribution, trading and use. Under ACES, all covered entities would have to remit one “allowance” per ton of GHG emitted. Allowances were to be issued annually by EPA and the total number of allowances would decline over time to achieve a target reduction of 83% below 2005 levels by 2050. Allowances could be bought, sold, or banked to promote an economically efficient balance of mitigation activity across the economy over time. ACES included an elaborate formula for the initial distribution of allowances across groups and sectors, designed to address distributional concerns, incentivize technologies and behaviors. To address competitiveness issues, section 782 reserved 15% of allowances to compensate EITE industries through output-based allocations. The intent of this provision was to subsidize the output of firms that both use a significant amount of energy (and would thus face increased costs associated with allowances), and/or face significant competition from international trade (and thus risk losing market share to imports from countries without similarly stringent climate policy). Section 764 of ACES directed EPA to publish a list of eligible industrial sectors and provided references to specific data sources, measures and thresholds. To be included, an industrial sector must be in a six-digit classification of North American Industry Classification System (NAICS) that has energy intensity of at least 5% and trade intensity of 15%, or an energy/GHG intensity of at least 20% (details below).^4,5 After two years and every four years thereafter, ACES required that EPA would update that list using the same criteria.⁶

In 2009–2010, EPA participated in an interagency analysis of competitiveness impacts of ACES, in response to a request from Senators Bayh, Specter, Stabenow, McCaskill and Brown (Interagency, 2009).⁷ Economists from the Departments of Commerce, Treasury, Energy and EPA conducted a comprehensive analysis of eligibility and impact.

In support of the interagency effort, EPA analyzed GHG emissions and economic data according to criteria established in ACES and developed a list of 46 sectors that were “presumptively eligible” for output-based allocation under ACES as well as their corresponding emissions levels. The eligibility analysis used the best publicly available data available at that time, representing 2004–2007, using a combination of the Annual Survey of Manufacturers (ASM) and the Economic Census (EC) (U.S. Department of Commerce, 2021), as well as the Manufacturing Energy Consumption Survey (MECS) (U.S. Department of Energy, 2021), the Greenhouse Gas Emissions Inventory (Environmental Protection Agency 2020) as well as other agency reports. The interagency group found that output-based allocations could eliminate almost all cost impacts as well as the resulting changes in net imports and associated emission leakage. ACES expired with the end of the 111th Congress without a corresponding Senate bill. The structure of the U.S. economy has continued to evolve, with changes in fuel prices (both relative and absolute), state, regional and foreign country climate policies. Yet, the same issues face policymakers now as then: How can climate policy both reduce emissions, yet not trigger emissions leakage (or accelerate the loss of market share in the U.S. manufacturing sector)? How do we identify industries where leakage is likely to occur and design measures that can stand the test of time against a backdrop of transformational change? To better understand such unintended consequences, economic models and data developments have connected theory with more detailed data on sector level production and emissions.

Agan and Look (2014) updated the EPA industry analysis and examined various alternative thresholds. Agan and Look advocated using a supply chain measure of carbon which expands eligibility to 182 sectors. Aldy and Pizer (2015) used an econometric approach with historical data to estimate the effects of fuel price changes on industrial output, and find that energy-intensive manufacturing industries are more likely to experience decreases in production and increases in net imports than less intensive industries.⁸ They also make a strong case that the correct measure for emissions leakage is net imports, not reduced domestic EITE output. EPA (2016) placed the interagency results in the context of other research (notably Böhringer et al., 2012) finding that domestic output effects as measured by a percentage change in output from EITE industries associated with a stylized climate policy, range from 1.5% to 6.5%.

Recent work has focused on disaggregation — by sector and even within sector. Disaggregation of energy-intensive sectors is crucial for measuring competitive impacts (Alexeeva-Talebi et al., 2012; Caron, 2012). Targeting the industrial sector, Gonzales et al. (2020) found a great amount of variance across firms within a sector, and the largest 10 polluters have a disproportionate share of total sector emissions. Heterogenous-firm trade models of the form in Melitz (2003) have been applied to carbon leakage questions, finding that the relationship between emissions intensity and production is critically important, determining whether a carbon tax increases global emissions or results in an additional source of pollution reductions (Balistreri and Rutherford, 2012; Kreickemeier and Richter, 2014; Altaghlibi and d’Autume, 2018). (Dechezleprêtre and Sato, 2017) provided an excellent review of nearly 40 years of research that, overall, has supported the findings of Jaffe et al. (1995) that environmental regulation does not have a large effect on competitiveness. The authors conclude that the concerns about competitiveness expressed in public policy debates are disproportionate to the magnitude of research findings on the subject.

3. Method and Data for Estimating Energy Intensity and Emissions

In this section, we summarize criteria for EITE industry following EPA (2010). In all cases, we use data at the NAICS-6 level of industry specificity. For consistency, we applied concordances to translate all the data to NAICS 2017 sector definitions.⁹ We include three time periods in this analysis. Data from 2004 to 2007 form period 1. We extend this with two additional periods, 2009–2012 and 2014–2017. Each four-year period consists of three years of data from the ASM followed by one year of data from the EC,¹⁰ plus additional sources as described in what follows.

3.1. Energy intensity

ACES established that energy-intensive sectors were those having energy intensity greater than 5%, defined as the ratio of fuel purchases to the total value of shipments. This was the first of two criteria used in ACES to determine eligibility for free allocation of emissions allowances. We calculate energy intensity from the ASM (EI ASM) for each sector i as

$$EIAS{M}_i=\frac{Avg(P{F}_i+P{E}_i)}{AvgV{S}_i}$$ (1)

where PF_i is fuels purchased by sector i, PE_i is purchased electricity by sector i and VS_i is value of shipments by sector i. All amounts are in dollars. To reduce sensitivity to year-to-year variability, we use three-year averages of data from the ASM.¹¹ For those sectors that were not represented at the six-digit level in the ASM (i.e., because they were aggregated into broader sectors), we used EC data for energy expenditures and value of shipments to determine energy intensity.

$$EIE{C}_i=\frac{(P{F}_i+P{E}_i)}{V{S}_i}$$ (2)

The calculations excluded any year for which any of the above data were withheld. Data sources are shown in Table 1. ACES partially addressed the possibility of a windfall by using current period output (not historical output) and by using industry average emissions rates (McMackin, 2009).

Table 1.

Data summary.

Data source	Measures	Period 1 (2004–2007)	Period 2 (2009–2012)	Period 3 (2014–2017)
Eligibility criteria
ASM	Cost of fuels, electricity; value of shipments	2004–2006	2009–2011	2014–2016
EC	Cost of fuels, electricity; value of shipments	2007	2012	2017
Emissions calculation
ASM	Cost of fuels, purchased electricity (kW h)	2006	2011	2016
EC	Cost of fuels, purchased electricity (kW h)	2007	2012	2017
IO matrix	Fuel shares	2007	2012	2012
MECS	Fuel prices	2006	2010	2014
U.S. GHG Inventory	Process emissions	2007	2012	2017

Notes: 2017 IO tables are not yet available. BEA plans to release updated estimates based on the 2017 EC (including 2017 detailed IO tables) in the fall of 2023. We used 2012 values to represent both period 2 and period 3.

3.2. Trade exposure

Trade-exposed industries are defined in ACES as those that have total trade, defined as imports plus exports of more than 15% of total domestic supply, defined as the value of shipments plus imports.

$$T{I}_i=\frac{{\mathit{\text{Imports}}}_i+{\mathit{\text{Exports}}}_i}{\mathit{\text{Value}}of{\mathit{\text{Shipments}}}_i+{\mathit{\text{Imports}}}_i}>0.15$$ (3)

3.3. GHG emissions

Following EPA (2010), two methods were used to estimate GHG emissions, depending on data availability. For the industries represented at the NAICS-6 level of detail in the MECS (EIA, 2021), we estimated emissions directly from reported fuel use. These data were available for 40–45 industries including most presumptively eligible sectors. For those sectors where detailed estimates were not provided in MECS, we used an alternate method described below. Nonfuel use is excluded from this analysis. GHG emissions from sector i (GHG_i) are the sum of emissions from direct fuel use (Edirec_i), indirect emissions from electricity use (Eelec_i) and process emissions (Eproc_i).

$$GH{G}_i={\mathit{\text{Edirec}}}_i+{\mathit{\text{Eelec}}}_i+{\mathit{\text{Eproc}}}_i$$ (4)

where Edirec_i denotes the direct fuel use emissions from sector i based on MECS data, calculated as the product of fuel use (F_ik) and an emissions factor (EF_k), summed over fuel types k[k ¼ (petroleum,gas,coal)].

$${\mathit{\text{Edirec}}}_i=\sum _k{F}_{ik}*E{F}_k$$ (5)

Indirect emissions Eelec_i from sector i are calculated

$${\mathit{\text{Eelec}}}_i=E_i*E{F}_E$$ (6)

where E_i is the quantity of electricity used in sector i purchased (kW h) from ASM data, or EC data where six-digit ASM data were not available. EF_E is the average U.S. CO₂ emissions factor from the Department of Energy (1405 lb CO₂/MW h), to convert to CO₂ emissions.¹² The analysis does not include non-CO₂ combustion emissions, which are likely to be very small.

Process emissions (Eproc_i) were included by matching the sources in the U.S. GHG Inventory of Emissions and Sinks (EPA, 2020) with NAICS sectors. The process emissions in the analysis include all GHGs, not just CO₂.

Alternate Method

For sectors not included in MECS we used an input–output (IO) approach to estimate direct fuel use emissions (Edirec^A_i). Emissions were calculated using bottom-up energy expenditure and electricity use data from the ASM and the EC.¹³

$$\mathit{\text{EDire}}{c}_i^A=\sum _k\frac{P{F}_i*A_{ik}}{P_k}*E{F}_k$$ (7)

Where Edirec^A_i is the direct fuel use emissions from sector i, PF_i is the value of purchased fuels in sector i, defined as total purchased fuel costs from the ASM, or 2007 EC data where six-digit ASM data were not available. A_ik is the share of fuel k[k = (petroleum,gas,coal)] given by the IO use table. Finally, P_k is the price of fuel k, taken from MECS price data for that fuel at the corresponding three-digit NAICS.

Data sources are summarized in Table 1. The EC is released every five years (years ending in 2 and 7). The national IO data follow the census schedule with a lag. MECS is produced every four years. The ASM is produced every year (except census years) and the GHG Inventory is annual. Generally, we follow EPA (2010) by using three years of the ASM that precede an EC year, together with the closest MECS and IO data.

4. Results

4.1. Trade intensity results

We calculated trade intensity as in Eq. (3), but it proved to be of no consequence. Every sector that was energy-intensive by Eqs. (1) or (2), also met this threshold for trade intensiveness.¹⁴ We conclude that trade exposure, as defined in ACES is ubiquitous among energy-intensive industries.

Figure 1 gives a high-level sense of the sectors at play in this space and a sense of their size, energy intensity and trade intensity. Bubble size shows the relative changes in value of shipments, and the transparency shows the time periods with solid color representing period 3 fading to transparent circles for period 1. The concept of energy intensity relates to the intensity effect while the value of shipments captures the scale effect (Grossman and Krueger, 1993).

Figure 1.

Change in key EITE metrics: trade intensity, energy intensity and output (aggregated sectors).

Notes: Top EITE sectors aggregated to four-digit NAICS. Pesticide and fertilizer manufacturing is not shown. Threshold values of energy intensity < 5% and trade intensity < 15% are shown, all sectors would qualify.

Most of the aggregated EITE sectors with the exception of glass manufacturing and pesticide and fertilizer manufacturing (not shown) declined in energy intensity and increased in trade intensity, moving from lower-right to upper-left. Glass manufacturing and pesticide and fertilizer manufacturing increased slightly in energy intensity and declined slightly in trade intensity. Additionally, all the aggregated sectors increased in value of shipments, with the exceptions of glass manufacturing, lime and gypsum manufacturing and aluminum.

4.2. Energy intensity results and the number of “presumptively eligible” industries

Figure 2 provides summary of the change in eligibility and emissions, showing the number of EITE sectors in each period as well as the emissions from EITE sectors relative to all of manufacturing. In the 2004–2007 period, there were 39 industries that met the ACES threshold. The number of EITE sectors fell to 34 in 2009–2012 and fell further to 26 in 2014–2017.

Figure 2.

Comparison of GHG emissions from EI (> 5%) and emissions from all sectors.

Notes: EPA(2010) included 16 additional presumptively eligible sectors in the first period, 13 were sectors that have been combined with other sectors in NAICS revisions, one whose energy intensity declined between 2002 and 2007, and two that were included by specific language in ACES. EITE sector emissions declined more in percentage terms than did manufacturing as a whole, even accounting for eligibility. Overall, manufacturing emissions became less concentrated in EITE sectors. Data in Table S1.

We further analyzed the period 3 results and found that the 5% threshold included in ACES would have to be lowered to 4% to retain the same number of EITE sectors as 2004–2007, and reduced to 1% to retain every EITE sector from period 1. While ACES tasked EPA with updating the data and calculation, updating the threshold was not anticipated. These changes track the aggregate effect of declines in the relative costs of energy, fuel switching, efficiency gains or individual industry effects. Discussion of these causal factors is in what follows.

Although the criterion for designation as an energy-intensive industry is unitless, it is nonetheless affected by relative prices over time. For example, in 2020, natural gas prices are about 77% lower than they were at their peak in 2008 (EIA, 2020). This fact has much to do with the changes in the number of energy-intensive industries reported in Fig. 2. Table 2 shows average nominal values of the value of shipments, costs of fuels and cost of electricity as reported in the EC of 2007, 2012 and 2017. No deflator has been applied, as it would simply cancel in the intensity calculation.

Table 2.

Change in key energy intensity measures 2007–2017 ($1000, nominal).

	2007	2012	2017	Annual percentage change (%) 2007–2017
Value of shipments	14,746,795	12,366,185	18,070,281	2.1%
Cost of fuels	649,714	342,153	372,107	5:4%
Cost of electricity	428,113	399,130	234,522	5:8%
Average energy intensity	7.3%	6.0%	3.4%	*

Notes: Average values for 39 sectors ranked by energy intensity, data from EC. Nominal values.

Table 2 shows that the value of shipments increased in nominal terms at an average annual rate of 2.1% while the cost of fuels and electricity purchased fell by 5–6% per year. The bottom row shows that average energy intensity for the top 39 sectors ranked by energy intensity was 7.3% in 2007 and just 3.4% in 2017. The energy intensity measure defined in ACES is heavily influenced by relative prices, which complicates its use over time as a measure for defining energy intensity.

In terms of emissions, Fig. 2 (above) shows the GHG emissions associated with EITE industries has declined both in absolute terms and as share of overall manufacturing. In period 1, EITE sectors emitted 515 MMTCO₂e, 75% of the overall manufacturing sectors total emissions of 683 MMTCO₂e. By period 3, the EITE sectors emissions had declined 26% to 383 MMTCO₂e, or 59% of the overall manufacturing sectors total emissions of 650 MMTCO₂e. Overall, the manufacturing industries reduced emissions 5% between period 1 and period 3.

The areas in Fig. 2 marked by diagonal lines represent the GHG emissions of sectors that lost EITE status (five sectors in period 2, 13 in period 3). Had these sectors remained EITE, they would have partially offset, but not eliminated the reduction in concentration of industrial emissions. Within this group, emissions were down 9% in 2017, while non-EI manufacturing sectors emissions rose 8%. In period 3, the emissions of EITE sectors plus these “dropped” sectors would have totaled 467 MMTCO₂e, 71% of total manufacturing emissions. Thus, manufacturing emissions became less concentrated in EITE sectors, even controlling for eligibility.

How do these estimates compare with the official U.S. GHG Inventory? For comparison, Table 3 shows data for fossil fuel combustion CO₂. The inventory has higher emissions from industry overall because it includes the emissions from intermediate purchases of electricity by the industrial sector but also emissions from agriculture, petroleum and natural gas processing and waste which are excluded from our analysis. We find similar trends but slightly higher emissions overall from the U.S. GHG Inventory for the industrial sector, which shows a 7% decline over the same period.

Table 3.

U.S. GHG Inventory industrial sector emissions.

	2007	2012	2017
Total industrial sector fossil fuel combustion CO2 (MMT)	816	759	750
Change from 2007 (%)	—	7	8

Source: EPA (2020).

An alternative energy intensity measure could be calculated as a ratio of physical quantities, BTUs per ton of output, etc., a measure that would be much more robust to economic change, but less comparable across industries. It may be preferable under these circumstances to simply establish the list of EITE sectors directly, by appeal to another data source, or a base year. In the rest of this paper, we follow this approach, defining the group of EITE sectors as the 39 sectors identified in period 1.

4.3. Decomposing emissions results

It is helpful to think of emissions as the product of scale and intensity effects.¹⁵

$$GH{G}_i=V{S}_i\cdot \frac{{\sum }_i\left({\mathit{\text{Edirec}}}_i+{\mathit{\text{Eelec}}}_i+{\mathit{\text{Eproc}}}_i\right)}{V{S}_i}$$ (8)

$$GH{G}_i={\mathit{\text{VS}}}_i\cdot {\mathit{\text{EmissionsIntensity}}}_i,$$ (9)

We can apply this decomposition to examine the effects of scale and emissions intensity over time on emissions changes for EITE sectors. First, we examine scale and intensity to verify that they are not correlated. Figure 3 shows emissions intensity and total GHG emissions across the three periods and shows that the two measures are not strongly correlated.

Figure 3.

Emissions intensity and total GHG emissions across three periods.

Note: “Presumptively eligible” sectors consistent with EPA (2010).

We can apply this decomposition to examine the effects of scale and emissions intensity over time on emissions changes for EITE sectors. First, we examine scale and intensity to verify that they are not correlated. Figure 3 shows emissions intensity and total GHG emissions across the three periods and shows that the two measures are not strongly correlated.

GHG emissions factors for the top nine industry groups ranked by GHG emissions across the three periods are shown in Fig. 4 and Tables 4–6. In Fig. 4, the width of the bars represents value of output in millions of dollars, the height represents the emissions intensity, therefore, the area of the rectangle represents total emissions. The data show that the top emitters can be divided into two groups: those like cement with high emissions intensities and those like iron and steel with lower intensities that produce emissions through higher volume of sales. Iron and steel is the largest emitting sector in all three periods with emissions of 92.1 MMTCO₂e in period 1, falling slightly to 86.8 MMTCO₂e in period, while the value of iron and steel output rose from $84.6B to $92.4B. Thus, GHG emissions per unit output in iron and steel fell from 1.1 MMT/$B to 0.9 MMT/$B. Cement manufacturing is the second-largest emitting sector at 76.9 MMTCO₂e in period 1, falling to 63.8 MMTCO₂e in period 3. The value of cement manufacturing output fell from $9.7B in period 1 to $8.1B in period 2. GHG emissions per unit output in cement manufacturing was 7.9 MMT/$B in period 1 and returned to that same value in period 3, after rising slightly in period 2.

Figure 4.

GHG emissions, GHG intensity and value of output across three periods (top nine sectors ranked by total GHG emissions).

Note: Excludes petroleum refineries per H.R. 2454.

Table 4.

GHG emissions calculation — period 1 (2004–Industry group 2007).

Industry Group	GHG emissions (MMTCO2e)		Value of output ($B)		GHG/output (MTCO2e/$M)
Iron and steel mills and ferroalloy manufacturing	92.1		84.6		1.1
Cement manufacturing	76.9		9.7		7.9
All other basic organic chemical manufacturing	46.8		80.5		0.6
Paper (except newsprint) mills	38.3		7.9		0.8
Industrial gas manufacturing	36.5		24.7		4.6
Other basic inorganic chemical manufacturing	31.8		21.7		1.3
Plastics material and resin manufacturing	30.2		4.3		0.4
Paperboard mills	29.9		1.4		1.4
Nitrogenous fertilizer manufacturing	23.8		9.3		6.9
Total “presumptively eligible”	515.2		519.2		0.9
Total industry	683.2		4880.2		0.1

Table 6.

GHG emissions calculation — period 3 (2014–2017)

Industry Group	GHG emissions (MMTCO2e)	Value of output ($B)	GHG/output (MTCO2e/$M)
Iron and steel mills and ferroalloy manufacturing	86.8	92.4	0.9
Cement manufacturing	63.8	8.1	7.9
All other basic organic chemical manufacturing	42.6	81.3	0.5
Nitrogenous fertilizer manufacturing	32.8	7.8	4.2
Ethyl alcohol manufacturing	32.8	31.8	1.0
Plastics material and resin manufacturing	30.0	87.9	0.3
Paper (except newsprint) mills	27.1	44.5	0.6
Other basic inorganic chemical manufacturing	25.0	30.4	0.8
Paperboard mills	23.1	29.7	0.8
Total “presumptively eligible”	382.8	341.3	1.1
Total industry	649.9	5580.0	1.1

Across the three periods of analysis, iron and steel and cement and other basic chemicals are consistently the top three emitting sectors. Some other sectors like nitrogenous fertilizer reduced their emissions intensity but increased their output and emissions rank overall. Paper, on the other hand, reduced emissions intensity but also reduced output, and reduced its emissions rank.

5. The Greenhouse Gas Reporting Program Data

ACES specified what was, at the time, the best available data. In this section, we introduce and compare a new data source, EPA’s GHGRP. The GHGRP provides emissions data consistent with the U.S. GHG Inventory, accessible from the facility level up to the national scale. We show that it provides similar emission results to the methods based on MECS, for more sectors with annual estimates at finer geographic scales. The additional fidelity of the data should be an important feature for competitiveness analysis in the future. EPA developed the GHGRP using its existing authorities under the Clean Air Act. GHGRP covers all sectors of the economy including both direct emitters as well as suppliers. Facilities and suppliers across 41 source categories report GHG data to EPA annually. Facilities are generally required to submit annual reports if GHG emissions from covered sources exceed 25,000 metric tons CO₂e per year or supply of certain products would result in over 25,000 metric tons CO₂e of GHG emissions if those products were released, combusted or oxidized. For certain “all-in” direct emitting source categories no threshold applies and all sources report.¹⁶

GHGRP receives approximately 8000 reports every year, with reporting beginning as early as 2010 for some sources. Approximately 1000 of these reports are from suppliers of fossil fuels and industrial gases or facilities injecting CO₂ underground. Total reported annual direct emissions are about 3 billion metric tons CO₂e, which is about 50% of total U.S. GHG emissions. Additional GHGs are accounted for by data collected from the supplier sectors. In total, data covering 85–90% of U.S. GHG emissions are reported.

Methodologies for calculating GHG emissions under GHGRP were developed through a notice and comment regulatory process using domestic and international references, including Intergovernmental Panel for Climate Change, WRI and California Air Resources Board AB32 guidelines, Both stationary combustion and process emissions data are reported with methods that include direct measurement e.g. using continuous emissions monitoring systems (CEMS), site-specific emissions factors and mass-balance depending on source category and facility characteristics. All GHGRP data are collected electronically and verified by EPA using a combination of electronic and manual checks.

5.1. Methodology — Using the GHGRP

The GHGRP reports emissions annually at the facility level by primary six-digit NAICS code. Using the emissions data is straightforward. CO₂ is reported, as is methane, N₂O and other GHGs, both directly from combustion and from processes. We used GHGRP data in Eqs. (7) and (8) with the same value of shipments data from the EC and ASM used above to calculate emissions intensities for the EITE sectors. Analogous to Fig. 3, Fig. 5 and Tables 7 and 8 shows the top sectors from the GHGRP, by period.

Figure 5.

GHG emissions, GHG intensity and value of output by period using GHGRP data (top nine sectors ranked by total GHG emissions).

Notes: GHGRP data are available annually from 2011 on 2012 and 2017 shown as representative.

Table 7.

GHGRP GHG emissions calculation — period 2 (2009–2012).

Industry group	GHG emissions (MMTCO2e)	Value of output ($B)	GHG/output (MTCO2e/$M)
Iron and steel mills and ferroalloy manufacturing	80.0	91.0	0.9
Cement manufacturing	64.4	5.4	11.9
Industrial gas manufacturing	44.7	34.2	6.1
Other basic inorganic chemical manufacturing	40.1	23.9	1.3
Plastics material and resin manufacturing	35.1	7.5	0.5
All other basic organic chemical manufacturing	33.1	75.7	1.0
Paper (except newsprint) mills	32.7	31.2	1.4
Nitrogenous fertilizer manufacturing	27.2	26.3	3.6
Lime manufacturing	25.3	2.1	12.2
Total “presumptively eligible”	517.9	445.8	1.2
Total industry		4904.2

Table 8.

GHGRP GHG emissions calculation — period 3 (2014–2017).

Industry group	GHG emissions (MMTCO2e)	Value of output ($B)	GHG/output (MTCO2e/$M)
Iron and steel mills and ferroalloy manufacturing	97.8	92.4	1.1
Cement manufacturing	71.6	8.1	8.9
All other basic organic chemical manufacturing	48.2	81.3	6.3
Nitrogenous fertilizer manufacturing	39.3	7.8	0.5
Ethyl alcohol manufacturing	37.2	31.8	4.8
Plastics material and resin manufacturing	33.6	87.9	0.4
Paper (except newsprint) mills	28.5	44.5	0.6
Other basic inorganic chemical manufacturing	27.9	30.4	0.9
Paperboard mills	25.3	29.7	0.8
Total “presumptively eligible”	539.4	341.3	1.6
Total industry		5580.0

The GHGRP data in Tables 7 and 8 show great similarities to the data in Tables 4–6 estimated from MECS and other sources in the rankings of top sectors and sector estimates. The GHGRP data show a higher total for the group of “presumptively eligible” sectors (539 MMTCO₂e in 2017 versus 382 MMTCO₂e), despite the 25,000 ton threshold. This may be because of the more comprehensive measures of non-CO₂ GHGs in the GHGRP.

6. Discussion — Lessons for Climate Policy

In this section, we distill several lessons for the next round of competitiveness policies. First, we acknowledge that eligibility for any special status under climate policy will be sought after and hard to remove once conferred. Retrospective analyses such as this one may have some value to policymakers. We found issues in implementing the ACES definitions over time because of differences in the embodied relative prices. The “alternate method”, in ACES, relying on expenditures and IO table values from different years was particularly problematic. At minimum, deflators should be specified, but in practice this is difficult because different sectors face different prices for energy. Another alternative involves using a benchmark value of nonmonetary measures of energy against nonmonetary measures of output. Such a system is used in California’s EITE provisions and in Canada’s output-based performance standards. A third option might be just naming a list of EITE sectors together with a petition method for changes.

A second source of challenges involved the changing economic sector classification schemes. In particular, the 2012 NAICS revision had a significant impact on the list of eligible sectors. Revisions are inevitable and valuable to keep pace with economic transformation. We found little change in the core group (top nine) sectors, however, so keeping a program small is probably easier in the long term.

A third issue surrounds the trade exposure of EITE industries. We found the trade exposure measure to be uninformative, as all the energy-intensive sectors also qualified. A bigger problem is its focus on trade volumes rather than emissions leakage. At the time when ACES was being proposed and analyzed, it would have been a unilateral action on the part of the U.S.. While there have been no climate regulations passed by the U.S. Congress, other countries including major U.S. trading partners have implemented climate polices. Protecting EITE industries requires an understanding of whether exporters would really face a competitive disadvantage. EITE industries that are primarily focused on trading with Canada do not, and other jurisdictions in the world have also adopted carbon pricing at various levels — notably the European Union. Future proposals should identify which industries would be truly affected and subject to significant emissions leakage based on their international trading patterns (Aldy and Pizer, 2015).

Lastly, future policymakers should consider eligibility criteria that are spatially explicit. ACES achieved one dimension of precision by specifying the highest-resolution sectoral definitions — NAICS-6 — be used, but then obscured regional differences by using national estimates. This obscures much industry heterogeneity. In the decade since ACES, economists have improved state and regional economics accounts. EPA is poised to release its first ever state-level GHG inventory, and data sources such as the GHGRP provide detailed facility level emissions. Economic models and data support a more disaggregated approach to EITE eligibility criteria. The authors’ plans for future work include exploring the spatial dimensions of competitiveness.

7. Conclusion

This paper provides a retrospective analysis of the EITE industry definitions in ACES, which prescribed measures, thresholds and data sources for determining eligibility for compensatory allowances designed to reduce competitive impacts and emissions leakage. We reproduce the 2009 eligibility criteria, using the same methods with more recent data to examine trends over the 2004–2017 period. We find that energy intensity, emissions intensity, output and emissions fluctuate with economic conditions, and defining measures and thresholds, as attempted in ACES, is challenging. Had ACES been enacted as written and not revised, the number of sectors qualifying for rebates would have decreased from 39 to 26, after adjustment for the changes in NAICS definitions. Emissions from the eligible sectors fell 26% across the three periods of analysis, while emissions from manufacturing fell 5%. We decompose the changes in emissions into scale and intensity measures based on a hybrid measure derived from Grossman and Krueger (1993) and Kaya and Yokoburi (1997). As an alternative, we perform the same emissions decomposition using data from the EPA GHGRP, finding it similar in results but available in annual reports and various scales down to the facility level. Finally, we draw some recommendations for future policy including (1) using measures that make price level adjustments straightforward or unnecessary, (2) keeping EITE policy focused on a small group to minimize sectoral reclassification problems, (3) identifying industries prone to emissions leakage rather than just trade volumes and (4) consider spatial heterogeneity of emissions and trade patterns. At the time of writing, Both the EU and the USA stand poised to enact measures to protect EITE industries and reduce emissions leakage. We hope these ideas can contribute to a more robust policy debate.

Table 5.

GHG emissions calculation — period 2 (2009–2012).

Industry group	GHG emissions (MMTCO2e)	Value of output ($B)	GHG/output (MTCO2e/$M)
Iron and steel mills and ferroalloy manufacturing	90.0	91.0	1.0
Cement manufacturing	54.5	5.4	10.0
All other basic organic chemical manufacturing	38.7	34.2	1.1
Paper (except newsprint) mills	31.3	23.9	1.3
Nitrogenous fertilizer manufacturing	29.4	7.5	3.9
Plastics material and resin manufacturing	28.6	75.7	0.4
Other basic inorganic chemical manufacturing	27.9	31.2	0.9
Paperboard mills	24.6	26.3	3.05
Lime manufacturing	22.6	2.1	10.9
Total “presumptively eligible”	429.0	445.8	1.0
Total industry	626.7	4904.2	1.3