Appendix H                
Description of the Commission’s Survey Methodology

 

The U.S. Trade Representative (Trade Representative) requested the U.S. International Trade Commission (Commission) conduct a survey of firms with facilities producing covered steel and aluminum products (covered products) in the United States to collect information pertinent to the estimation of greenhouse gas (GHG) emissions intensities by product category in 2022. The request asked that the survey collect data from these firms to the extent the data were not already publicly available via reporting to the U.S. Environmental Protection Agency’s Greenhouse Gas Reporting Program (GHGRP) or other sources. To address this request, the Commission issued questionnaires to companies with facilities producing covered steel and aluminum products. Using the data collected via the questionnaire in combination with data from external sources, the Commission estimated GHG emissions intensities at the product-category level. This appendix discusses the Commission’s approach to compiling the survey population, drafting its questionnaire, collecting data, analyzing questionnaire responses, and presenting findings in this report.

Survey Process

The survey process for this investigation consisted of three major steps. First, the Commission compiled the survey population. To identify the companies and their associated U.S. facilities that likely produced covered products in 2022, the Commission used various steel and aluminum association membership lists and other sources.

Second, the Commission developed questionnaires and collected data from companies and facilities. To facilitate the data collection process and to reduce the burden on facilities, the Commission conducted a two-part data collection. In the first part, the Commission sent a company-level questionnaire to companies that were identified as possibly producing covered products in 2022. After companies submitted the company-level questionnaire, the Commission initiated the second part of the data collection, sending facility-level questionnaires to each individual facility indicated in the company-level response. The facility-level questionnaire collected information on the production quantity of covered products, fuel usage, inputs, and sources of those inputs in the production process.

Finally, the Commission developed estimation calculations that combined information from facility-level questionnaire responses with other publicly available data, including data from the GHGRP and Emissions and Generation Resource Integrated Database (eGRID), to estimate production-weighted national averages and highest measures of emissions intensities of covered products in 2022 at the product-category level as described in chapter 3 of the report.

Survey Population Development

The survey population for this investigation is composed of companies and their associated facilities that produced covered products in 2022. Because this list of companies is not readily available from any one source, the Commission used a variety of sources to generate a comprehensive list of possible producers of covered products. The Commission then verified the companies on this list for inclusion in or exclusion from the survey population as described below.

Industry association lists were a major source used by the Commission to identify relevant companies and facilities. For steel, these included the Association for Iron & Steel Technology directory, Steel Manufacturers Association membership list, the Specialty Steel Industry of North America membership list, the American Wire Producers Association membership list, information from pipe and tube market research firm Preston Pipe, and the Committee on Pipe & Tube Imports membership list. For aluminum, industry association lists included those from the Aluminum Association, Forging Industry Association, American Foundry Society, and North American Die Casting Association. The Commission received several of these lists (which may not be publicly available) directly from industry contacts.

Additional sources the Commission used to build its comprehensive list of companies producing covered products included the EPA’s National Emissions Inventory and the GHGRP.686F[683] All companies with facilities reporting GHG emissions to the GHGRP in 2022 under subparts F (aluminum production) and Q (iron and steel production) were included in the population.687F[684]

The Commission closely reviewed companies and their associated facilities from each of these sources to determine their inclusion in the population. Cumulatively, the initial list of companies that possibly produced covered steel and aluminum product categories included more than 1,700 companies.

The Commission sent a prenotification letter to all companies on this initial list in January 2024. The letter asked companies to confirm whether they owned at least one facility that produced covered products in the United States in 2022 and to verify their company-level contact information. The letter also instructed companies to contact the Commission if none of their facilities produced any covered products in 2022 so that they could be removed from the list after confirmation of their claim. For companies that did not respond to the prenotification letter, the Commission used proprietary search tools to find potential company contacts and conducted further research to determine if these companies could be removed from the population.688F[685]

The final count of companies that the Commission identified as eligible for the company-level questionnaire totaled 948. The Commission was aware that this number would likely change after data collection began because some companies included in the population had not yet been verified as producers of covered products.

Questionnaire Development

The initial step of the questionnaire development involved extensive desk research. The Commission spoke to industry experts to understand the data requirements for estimating emissions intensities by product category. This outreach was extremely helpful to the Commission in identifying data that were already available and data that needed to be collected via the questionnaire.

As the U.S. Office of Management and Budget (OMB) guidelines require, the Commission held a 60-day public comment period during which the public could provide feedback on its questionnaires.[686] The Commission posted the questionnaires for this investigation for public comment on its website from November 7, 2023 to January 6, 2024.[687] Additionally, the Commission conducted field testing and cognitive interviews during the public comment period.[688] The participants in field testing and cognitive interviews included up to nine organizations each. The participating organizations were steel and aluminum producers of various sizes and using different production processes, and associations representing such steel and aluminum producers. This process allowed the Commission to receive feedback on the publicly posted draft questionnaires from all types of steel and aluminum producers.

After incorporating comments received during the public comment period and feedback from field testing and cognitive interviews, the Commission submitted its proposed data collection package$—$which included the questionnaires$—$to the OMB for approval in February 2024. The Commission then conducted extensive internal testing of the final questionnaires’ online interface to ensure the smooth operation of online data collection and to ensure displayed questions were tailored to a specific respondent’s operations. The Commission began data collection for this investigation in April 2024, shortly after receiving OMB approval.

Data Collection

The 948 companies in the survey population received both an email and a letter containing instructions for completing the company-level questionnaire, which was due two weeks after the data collection period started in early April 2024. In the company-level questionnaire, companies received a prepopulated list of their facilities to confirm production of covered products in 2022 and were given the opportunity to add or remove any facilities from this list, depending on their production of covered products during that year.692F[689] Respondents were also asked to provide contact information for personnel with whom the Commission could correspond at each identified facility. After submission of the company-level questionnaire to the Commission, the identified facility-level contacts were automatically emailed instructions for completing the facility-level questionnaire, which was due 60 days from the date the company-level questionnaire completion instructions were emailed.693F[690] Companies and facilities received extensions to questionnaire submission deadlines upon request$—$to the degree the investigation’s schedule allowed$—$to accommodate internal delays.

To gather complete and accurate data inputs for estimating emissions intensity, the facility-level questionnaire collected a large amount of data. In some cases, respondents may have been unfamiliar with specific information (e.g., energy usage) requested in the questionnaire. The Commission hosted two webinars during the data collection period to explain the survey process, provide more detailed information on how to complete the questionnaire, and answer respondents’ questions. One webinar specific to steel producers was held on April 23, 2024, and another specific to aluminum producers was held on April 26, 2024. Both webinars were well attended and included active question-and-answer sessions. The Commission updated the investigation’s Frequently Asked Questions (FAQs) document to reflect these questions and answers so that the information disseminated during these events was accessible to all potential respondents.

Response Rates

Of the 948 companies that received a company-level questionnaire, 329 were exempted from responding because none of their facilities produced covered products in 2022. During the course of data collection, an additional 33 companies were identified as owners of facilities that produced covered products in 2022. In those 33 cases, either the owner had purchased a facility that produced covered products from another company in the population; or, because of its corporate structure, the owner requested that a company listed in the population as one company be divided into two or more companies for questionnaire reporting purposes.

After all adjustments were made, 652 companies remained in the company-level survey population and 538 submitted completed questionnaires. Using the equation H.1 below, the Commission calculated 82.5 percent as the overall response rate for the company-level questionnaire.[691] The company-level response rate for steel companies was 89.6 percent; the response rate for aluminum companies was 80.0 percent (table H.1).

$\begin{array}{c}Company response rate=\frac{Total questionnaire responses received}{Total questionnaires sent-Companies exempted+Additional companies} \left(H.1\right)\end{array}$

Table H.1 Company questionnaire response activity

In number of company-level questionnaires.

Activity	Steel companies	Aluminum companies	Total companies
Total questionnaires sent	349	627	948
Companies exempted	147	186	329
Additional companies	19	15	33
Adjusted company population	221	456	652
Total questionnaire responses received	198	365	538

Source: USITC, Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Company-Level, 2024.

Notes: Adjusted survey population is calculated as total questionnaires sent minus exempted plus additional companies by column. Steel companies and aluminum companies do not sum to total companies because 28 companies owned facilities that produced both covered steel and aluminum products.

The facility-level questionnaire was sent only to facilities of the companies that responded to the company-level questionnaire and reported that the facility produced covered products in 2022. If a company did not respond to the company-level questionnaire, facilities of those companies did not receive facility-level questionnaires; therefore, those facilities were not considered in the response rate calculation. The company-level questionnaire respondents owned 1,030 facilities that produced covered products in 2022. All these facilities received the facility-level questionnaire, and 54 were subsequently granted exemptions, resulting in 976 eligible facilities.695F[692]

Of 976 facilities, 913 responded to the facility-level questionnaire, resulting in an overall response rate of 93.5 percent. The facility response rates were calculated by dividing the total responses received by the total questionnaires sent minus the facilities exempted (equation H.2). The response rates for facilities producing covered steel products and for those producing covered aluminum products were comparable at 93.1 percent and 93.9 percent, respectively (table H.2).696F[693]

$\begin{array}{c}Facility response rate= \frac{Total responses received}{Total questionnaires sent-Exempted facilites} \left(H.2\right)\end{array}$

Responses from 35 facilities were deemed unusable because the data provided were insufficient and could not be verified despite outreach attempts by the Commission (see Data Cleaning section below). The usable response rate is calculated after removing the total unusable facility responses from the numerator (equation H.3).

$\begin{array}{c}Facility usable response rate= \frac{Total responses received-Total unusable}{Total questionnaires sent-Exempted facilities}\left(H.3\right)\end{array}$

After removing unusable questionnaire responses, the overall facility usable response rate was 90.0 percent. The usable response rate for facilities producing covered steel products was 89.1 percent; the rate for facilities producing covered aluminum products was 90.5 percent.

Table H.2 Facility questionnaire response activity

In number of facility-level questionnaires.

Activity	Steel facilities	Aluminum facilities	Total facilities
Total questionnaires sent	502	541	1,030
Exempted facilities	53	1	54
Adjusted facility population	449	540	976
Total unusable questionnaire responses	18	18	35
Total questionnaire responses received	418	507	913

Source: USITC, Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Facility-Level, 2024.

Notes: Adjusted survey population for facilities is calculated by removing the exempted facilities from the total number of questionnaires sent. Steel and aluminum facilities responses do not sum to the total facilities across because 12 facilities produced both steel and aluminum covered products in 2022.

Analysis of Responses

Data Cleaning

Given the complexity and volume of data requested in the questionnaires, the Commission reviewed each submitted questionnaire to ensure respondents had properly reported all required data for the calculations. In cases where data were missing, appeared inconsistent, or were found to be extreme values (see the Outlier Analysis section below), the Commission attempted to verify or revise the submitted data by contacting the respondent via phone, video conference, or email, depending on the level of complexity and detail needed. A substantial number of initial responses needed to be revised. Using experience reviewing questionnaires and industry knowledge, the Commission made simple corrections to data (e.g., obvious typographical mistakes, unit errors, common terminology misunderstandings, etc.). Because certain corrections to responses in the questionnaires were complex, many questionnaires were reopened to allow the facility to make appropriate corrections.

Nonresponse Adjustment

The high response rates for the questionnaire meant that the nonresponse rates were low for this survey. When there is evidence of nonresponse bias and adequate auxiliary data are available for all nonrespondents, it is possible to improve some survey estimates through a process of nonresponse adjustment. That adjustment was not possible in the Commission’s survey because auxiliary data were not available.697F[694]

The Commission reviewed the list of survey nonrespondents to confirm that all major presumed producers of covered steel and aluminum products had submitted a questionnaire. Upon further assessment, the Commission found that the risk of bias to emissions intensity estimates given the low nonresponse rates was minimal, for three main reasons. First, almost all nonrespondents were producers of downstream steel and aluminum products. Despite this nonresponse, the volume of responses received in these product categories was sufficient to produce reliable emissions intensity estimates for these downstream products. Second, nonrespondents were not expected to have significantly different emissions intensity profiles than respondents within the same product category (i.e., nonrespondents’ emissions intensity estimates were not expected to fall consistently on the highest or lowest end of the range).698F[695] Third, given the survey’s very high response rates, remaining nonrespondents did not comprise large enough shares of U.S. production within any product category to have an outsized impact on any national estimates.

Response Coverage

Beyond the high response rate, the Commission is also confident that the production output of facility-level respondents to the questionnaire comprises the vast majority of U.S. production in covered steel and aluminum product categories in 2022. When comparing the 2022 total production collected in responses to the Commission’s questionnaire to that from external data sources, the survey captured nearly 100 percent of production for almost all product categories (tables H.3 and H.4).

The Commission sought to include all producers in the industry, which includes many smaller producers that are often excluded from national estimates produced by external data providers. For some categories, such as non-seamless tubular steel and secondary aluminum, the production totals captured through the survey were much higher than the production totals gathered from external data sources.

Table H.3 Commission steel production totals compared to external data steel production totals

In 1,000 metric tons (mt).

Product category	USITC production totals	External data production totals
Semifinished steel	85,329	81,392
Stainless steel semifinished steel	2,625	2,017
Carbon and alloy semifinished steel	82,704	79,518
Hot-rolled flat steel	56,813	54,531
Hot-worked long steel	27,646	26,025
Seamless tubular steel	3,292	1,861
Non-seamless tubular steel	6,416	1,822
Coated flat steel	19,928	19,950
Rebar	9,662	8,657

Sources: USITC, Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Facility-Level, 2024, responses to question 2.1.1; AISI, “Pig Iron and Raw Steel Production,” accessed January 31, 2023; worldsteel, Steel Statistical Yearbook 2023, accessed September 21, 2024.

Note: With the exception of semifinished steel, product categories shown above are combined totals for stainless and carbon and alloy steel types.

For secondary aluminum, the production totals estimated by external public data sources are smaller because those data sources typically do not include production for captive consumption9F(table H.4).[696] The production totals collected in the Commission’s survey include captive consumption, which accounted for 56.8 percent of total production.700F[697] Other production totals provided by external data sources may be missing data on captive production or likely reflect small survey sample sizes.

Table H.4 Commission aluminum production totals compared to external data aluminum production totals

In 1,000 metric tons (mt).

Product category	USITC production totals	External data production totals
Primary unwrought	877	861
Secondary unwrought	9,693	4,701
Wrought	8,598	9,311

Sources: USITC, Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Facility-Level, 2024, section 2. Primary aluminum 2022 production total from USGS, Mineral Commodity Summaries 2022: Aluminum, January 2022. Secondary aluminum 2022 production from the Aluminum Association, email message to USITC staff, September 18, 2024. Wrought 2022 production total from LSEG Metals Research, “World Metals Statistics Yearbook 2022,” 2023.

Note: The wrought external production total excludes castings, forgings, and wire production.

Computational Methods

Calculation of the Average Emissions Intensity Estimates

As requested in the Trade Representative’s letter, the Commission is providing the estimated production-weighted national average emissions intensity by product category in this report. The concept of a production-weighted average is most clearly explained when first described at the facility level. First, the Commission uses responses to the questionnaire and external data, e.g., from the GHGRP and eGRID, as inputs to the calculation equations$—$described in detail in the Commission’s calculation methodology$—$to compute facility-level emissions by scope for each covered product produced at that facility.701F[698] Next, scope 1, 2, and 3 emissions are added to compute the total emissions for each product produced in that facility. Finally, total emissions are divided by the production volume in metric tons of that product at that facility to produce facility-specific product-level emissions intensity estimates (equation H.4).

Estimated facility-specific emissions intensity for product $y$:

$\begin{array}{c}{\widehat{x}}_{iy} =\frac{{{\displaystyle \sum }}_{j=1}^3{e}_{ijy}}{p_{iy}} \left(H.4\right)\end{array}$

$e_{ijy}$ denotes estimated emissions in metric tons for product $y$ for facility $i$ and scope $j$; $p_{iy}$ denotes the total production in metric tons of product $y$ at facility $i$.

Emissions intensity estimates from each facility are then weighted by the production in metric tons of product $y$ and summed across facilities producing product $y$ to estimate the production-weighted national average emissions intensity for product $y$.

Estimated production-weighted national average emissions intensity:

$\begin{array}{c}{\overline{x}}_y=\frac{{{\displaystyle \sum }}_{i=1}^n{\widehat{x}}_{iy}*p_{iy}}{{{\displaystyle \sum }}_{i=1}^n{p}_{iy}} \left(H.5\right)\end{array}$

Mathematically, equation H.5 simplifies to equation H.6 below when ${\widehat{x}}_{iy}$ is replaced with equation H.4. This is the equation the Commission used in its estimation of average emissions intensities.

$\begin{array}{c}{\overline{x}}_y=\frac{{{\displaystyle \sum }}_{i=1}^n{{\displaystyle \sum }}_{j=1}^3{e}_{ijy}}{{{\displaystyle \sum }}_{i=1}^n{p}_{iy}} \left(H.6\right)\end{array}$

$e_{ijy}$ denotes emissions for each product category for facility $i$ and scope $j$; $p_{iy}$ denotes the total production of product $y$ for facility $i$; and $n$ denotes the total number of facilities producing product $y$.

Calculation of the Highest Emissions Intensity Estimates

The Trade Representative’s request letter asked that the Commission estimate the highest emissions intensities by product category. To estimate the highest emissions intensities without disclosing confidential business information of facilities or companies, the report presents the production-weighted average of a set production share of the highest emissions-intensive facilities. This is computed for 10 percent (i.e., 90$–$100th percentile range) of the total production of a particular product category using the steps below.

Facility-specific emissions intensity estimates are calculated using equation H.4 for each steel and aluminum product category. Next, facilities are arranged in descending order of the emissions intensity estimates, and cumulative production shares are calculated for that product category. Facilities are included until 10 percent of production is captured from the top end of the emissions intensity estimate distribution.702F[699] Finally, production-weighted average emissions intensity estimates are calculated for the 90$–$100th percentile range using equation (H.6) over the facilities included.

Production-weighted averages have also been calculated for the 50$–$100th, 60$–$100th, 70$–$100th, and 80$–$100th percentile ranges (i.e., the most emissions-intensive facilities representing 50 percent, 40 percent, 30 percent, and 20 percent of all U.S. production in a product category, respectively) using these steps and presented in appendix I.

For product categories that do not qualify for publication in this report under the data disclosure rules (see Data Disclosure Review section below) at the 90$–$100th percentile range, the highest measure will be shown at the percentile range representing the narrowest percentile range of highest emissions intensities estimates as possible while protecting confidential business information. For example, if the highest emissions intensities estimate for a product category at the 90$–$100th percentile range does not meet the requirements of the Commission data disclosure rules but meets the requirements at 80$–$100th percentile range, the report will display the 80$–$100th percentile range.

Precision of the Estimates

The standard error is a measure of the precision of the sample mean.703F[700] The Commission estimated the standard error using the Taylor series linearization method because the emissions intensities estimator is a non-linear estimator computed from the estimator of totals.704F[701] The standard errors calculated using the Taylor series method were all small compared to the mean estimate. As a result, the relative standard errors (standard error divided by the mean) were less than 5 percent for all emissions intensities estimates for steel and aluminum product categories. The low relative standard errors show that the estimates presented in the report have high levels of precision.

Measures of Dispersion

When analyzing estimates of averages (i.e., means), it is important to consider how the underlying data are distributed. A measure of dispersion indicates how the data are distributed around a measure of central tendency. Production-weighted standard deviation was calculated on emissions intensities estimates to provide insight into the spread of the data (tables H.5$–$H.7).

The standard deviation, equation H.7, measures the amount of variation of the emissions intensity estimates around its mean for product $y$. Large standard deviations indicate the data are spread over a wider range than if the standard deviation were small, relative to the mean.705F706F[702] Because the national average emissions intensity estimate is production weighted, the standard deviation is also production weighted.

$\begin{array}{c}{s}_y=\sqrt{\frac{{{\displaystyle \sum }}_{i=1}^n{p}_{iy}\left({\widehat{x}}_{iy}-{\overline{x}}_y\right)}{{{\displaystyle \sum }}_{i=1}^n{p}_{iy}}} \left(H.7\right)\end{array}$

Table H.5 Carbon and alloy steels: measures of dispersion by product category and subcategory

In number of facilities and metric tons of carbon dioxide equivalent per metric ton of steel produced (mt CO₂e/mt steel).

Product category and subcategory	Facilities (number)	Average emissions intensity (mt CO2e/mt steel)	Standard deviation (mt CO2e/mt steel)
Semifinished	87	1.02	0.60
Slab	30	1.35	0.52
Ingot	12	0.61	0.59
All other semifinished	51	0.50	0.22
Flat	79	1.83	0.72
Hot-rolled flat	47	1.59	0.62
Plate	27	1.41	0.69
All other hot-rolled flat	36	1.61	0.60
Cold-rolled flat	41	1.91	0.70
Coated flat	45	2.17	0.80
Long	160	0.75	0.45
Hot-worked long	70	0.67	0.32
Rebar	32	0.54	0.13
Wire rod	14	0.94	0.52
Heavy structural shapes	15	0.67	0.22
All other hot-worked long	33	0.74	0.36
Cold-formed long	99	1.25	0.70
Wire	72	1.48	0.71
All other cold-formed long	34	0.89	0.50
Tubular	114	1.50	0.50
Seamless tubular	21	1.09	0.18
Seamless oil country tubular goods	14	1.08	0.16
All other seamless tubular	10	1.23	0.29
Non-seamless tubular	97	1.71	0.49
Non-seamless oil country tubular goods	13	1.52	0.51
All other non-seamless tubular	88	1.74	0.48

Source: USITC, Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Facility-Level, 2024, responses to question 1.2.3. USITC estimates based on its calculation methodology, see appendix E.

 

Table H.6 Stainless steel: measures of dispersion by product category and subcategory

In number of facilities and metric tons of carbon dioxide equivalent per metric ton of steel produced (mt CO₂e/mt steel). d.s. = data are suppressed to protect confidentiality.

Product category and subcategory	Facilities (number)	Average emissions intensity (mt CO2e/mt steel)	Standard deviation (mt CO2e/mt steel)
Stainless steel	92	2.78	0.75
Semifinished	17	2.23	0.74
Slab	7	2.16	0.47
Ingot	11	2.85	1.65
All other semifinished	d.s.	d.s.	d.s.
Hot-rolled flat	14	2.31	0.50
Cold-rolled flat	15	3.08	0.32
Hot-worked long	14	2.93	1.29
Cold-formed long	22	3.55	1.03
Wire	16	4.55	1.31
All other cold-formed long	10	3.34	0.81
Seamless tubular	10	4.07	1.91
Non-seamless tubular	21	3.16	0.71

Source: USITC, Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Facility-Level, 2024, responses to question 1.2.3. USITC estimates based on its calculation methodology, see appendixes E and H.

 

Table H.7 Measures of dispersion by aluminum product category

In number of facilities and metric tons of carbon dioxide equivalent per metric ton of aluminum produced (mt CO₂e/mt aluminum).

Product category	Facilities (number)	Average emissions intensity (mt CO2e/mt aluminum)	Standard deviation (mt CO2e/mt aluminum)
Unwrought	108	3.46	4.75
Primary unwrought aluminum	6	14.52	7.37
Secondary unwrought aluminum	102	2.46	2.76
Wrought	417	6.23	4.78
Bars, rods, and profiles	126	8.35	5.93
Wire	22	8.35	4.51
Plates, sheet, and strip	36	4.97	3.70
Foil	8	8.66	2.02
Tubes, pipes, and tube or pipe fittings	42	8.21	4.11
Castings	200	6.00	6.39
Forgings	29	5.00	3.26

Source: USITC, Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Facility-Level, 2024, section 1.2.3. USITC estimates based on its calculation methodology, see appendixes E and H.

 

Outlier Analysis

Estimates of averages and totals can be sensitive to the presence of unusual or extreme values (i.e., outliers) for a variable. The two primary categories of outliers included are reporting errors and unique units. The goal of outlier analysis is to identify extreme values and confirm them as unique units or correct them if they are reporting errors.707F[703]

Possible outliers in facility-specific product-level GHG emissions intensity estimates were identified and further assessed for validity and data quality. To identify these potential outliers, a modified Z-score was calculated at the product category level.708F[704] A modified Z-score (equation H.8) was used rather than a Z-score because the modified Z-score uses the median of the observations, which is more resistant to unusual observations than a mean that is used to calculate a Z-score. Additionally, emissions intensities estimate data did not follow a normal distribution for every product category, which made modified Z-scores more suitable for the outlier analysis for all product categories because they also do not assume a normal distribution. The modified Z-score is computed as follows for a product category:

$\begin{array}{c}{M}_{iy}=\frac{0.6745\left({\widehat{x}}_{iy}-{\stackrel{}{x}}_y\right)}{MAD}\left(H.8\right)\end{array}$

${\stackrel{}{x}}_y$ is the median GHG emissions intensity estimate for all facilities that produce product category $y$, and $MAD=media{n}_i\left\{\left|x_i-\tilde{x}\right|\right\}$ is the median of the absolute deviations about the median GHG emissions intensity.

Any values of $M_i>3.5$ were identified as possible outliers and scrutinized closely to determine if they were reporting errors or correctly reported extreme values.709F[705] Respondents with reporting errors were contacted for corrections (see the Data Cleaning section above).

Significance Testing and p-value

In instances where the intensities estimates were compared between product categories or between different methods in the sensitivity analyses, the Commission performed a statistical significance testing using a t-test.

A probability value, commonly known as p-value, is a statistical measurement used to validate a hypothesis against observed data. As p-values are generally used in this report to compare two groups (e.g., emissions intensity estimates between different types of products, or emissions intensity estimates for the same product using different calculation methodologies), the hypothesis is that there is no difference between the two groups or the two methods in case of sensitivity analyses. The smaller the p-value, the stronger the evidence that there is a difference between the means of two groups being compared.[706] A p-value less than 0.05 is considered statistically significant. Use of the term “significantly” in the text also indicates statistical significance between the compared groups.

Data Disclosure Review

The Trade Representative’s letter requested that the Commission not include any confidential business information in its report. The Commission has designated the information provided in response to its questionnaires as confidential business information unless such information is otherwise available to the public. Therefore, the Commission is obligated to withhold or suppress any data that would reveal a company’s or facility’s information. A comprehensive disclosure review was conducted for all survey results presented in this report. Data were suppressed to protect any data that were determined to be sensitive to a disclosure of information. Data such as production and emissions intensity estimates were determined to be sensitive and were subject to disclosure controls.

Estimates using survey data presented in the report were either calculated using only questionnaire responses (e.g., production totals) or questionnaire responses combined with external data in complex calculations (e.g., emissions intensity). Estimates based solely on questionnaire responses were determined to be sensitive to a disclosure of information if they failed either of two rules: the threshold rule or the dominance rule. Estimate disclosure failed the threshold rule if the estimate comprised data from fewer than a set minimum of companies and facilities. Estimate disclosure failed the dominance rule if the data from a small number of companies or facilities dominated the estimate, which could allow a data user to estimate any respondent’s data too closely. Estimates based on the Commission’s calculation methodology follow the threshold rule only; the data were transformed substantially so that no data user could infer an individual facility’s estimate.

 

Bibliography

Altman, Douglas, and J Martin Bland. “Standard Deviations and Standard Errors.” The British Medical Journal 331, no. 903 (October 15, 2005). https://doi.org/10.1136/bmj.331.7521.903.

American Iron and Steel Institute (AISI). “AIS-7 Pig Iron and Raw Steel Production, December 2022.” Accessed January 31, 2023. https://www.steel.org/.

Beaumonth, Jean-François, and Louis-Paul Rivest. Handbook of Statistics, Vol. 29A, digital printing. Edited by Danny Pfeffermann, and Calyampudi Radhakrishna Rao. Sample Surveys: Design, Methods and Applications. Amsterdam: Elsevier, 2010. https://www.sciencedirect.com/handbook/handbook-of-statistics/vol/29/part/PA.

Beers, Brian. “P-Value: What It Is, How to Calculate It, and Why It Matters.” Investopedia, 2024. https://www.investopedia.com/terms/p/p-value.asp.

Iglewicz, Boris, and David C. Hoaglin. How to Detect and Handle Outliers, Vol. 16. The ASQC Basic References in Quality Control: Statistical Techniques. Milwaukee, Wisconsin: ASQC Quality Press, 2004. https://hwbdocs.env.nm.gov/Los%20Alamos%20National%20Labs/TA%2054/11587.pdf.

London Stock Exhange Group (LSEG). “WBMS, World Metal Statistics Yearbook 2022.” LSEG Metals Research, 2023. https://www.lseg.com/en/data-analytics/financial-data/commodities-data/metals-data/world-bureau-of-metal-statistics-wbms.

SAS Institute Inc. “Taylor Series Variance Estimation.” SAS Institute Inc. Accessed October 30, 2024. https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.4/statug/statug_surveyfreq_details19.htm#:~:text=The%20Taylor%20series%20linearization%20method,formed%20from%20estimators%20of%20totals.

Sunstein, Cass R. “Information Collection Under the Paperwork Reduction Act.” Executive Office of the President. Office of Management and Budget. Memorandum for the heads of executive departments and agencies, and independent regulatory agencies, April 7, 2010. https://www.whitehouse.gov/wp-content/uploads/legacy_drupal_files/omb/assets/inforeg/PRAPrimer_04072010.pdf.

Upton, Graham, and Ian Cook. Understanding Statistics. OUP Oxford, 1996. https://archive.org/details/understandingsta0000upto/page/n3/mode/2up.

U.S. Environmental Protection Agency (EPA). “National Emissions Inventory (NEI).” Other Policies and Guidance, June 2, 2015. https://www.epa.gov/air-emissions-inventories/national-emissions-inventory-nei.

U.S. Environmental Protection Agency (EPA). “Find and Use GHGRP Data.” Data and Tools, July 30, 2021. https://www.epa.gov/ghgreporting/find-and-use-ghgrp-data.

U.S. Geological Survey (USGS). Mineral Commodity Summaries 2022: Aluminum. USGS, January 2022. https://pubs.usgs.gov/periodicals/mcs2022/mcs2022-aluminum.pdf .

U.S. International Trade Commission (USITC). Aluminum: Competitive Conditions Affecting the U.S. Industry. USITC Publication 4703, June 2017. https://www.usitc.gov/publications/332/pub4703.pdf.

U.S. International Trade Commission (USITC). Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Company-Level. 2024. https://www.usitc.gov/research_and_analysis/documents/sa_emissions/emissions_company-level_questionnaire.pdf.

U.S. International Trade Commission (USITC). Greenhouse Gas (GHG) Emissions Intensities Questionnaire: Facility-Level. 2024. https://www.usitc.gov/research_and_analysis/documents/sa_emissions/emissions_facility-level_questionnaire_aluminum_and_steel_combined.pdf.

World Steel Association (worldsteel). Steel Statistical Yearbook 2023. Accessed September 21, 2024. https://worldsteel.org/publications/bookshop/ssy_subscription-2023/.