Fewer than a third of the cells in the federal manufacturing-energy tables are ordinary published numbers
A cell-by-cell account of what the EIA Manufacturing Energy Consumption Survey 2022 and EPA eGRID2023 publish about industrial fuel, steam raising and cogeneration, with every unpublished cell retained.
Keywords: MECS; eGRID; industrial energy statistics; data suppression; cogeneration; grid emission factors
Abstract
Federal statistics are the usual basis for estimating the scale of industrial fuel use, steam raising and cogeneration in the United States. We read every cell of 5 tables of the EIA Manufacturing Energy Consumption Survey 2022, retaining each cell's status marker, together with 322 emission-rate and grid-loss records from EPA eGRID2023. Of the 23,409 survey cells, 7,013 (30 percent) are ordinary published numbers; the remainder are zeros, values rounded below the reporting unit, cells not applicable or not available, and 557 cells withheld for sampling error or confidentiality. Of 145 subtotals computed from published industry rows, 133 include at least one cell without a published value. Consequently, sums of published cells are not totals: in the South Census region they fall 422 TBtu (5.1 percent) short of the published total, while in the West they exceed it. Across the 27 eGRID subregions, annual CO2e emission rates span a factor of 6.4, and California's CAMX subregion ranks 2 of 27 from the cleanest. Analyses based on these sources should carry cells without published values as unknowns and recognize that results computed on the CAMX grid do not transfer to the median subregion.
1. Introduction
The Manufacturing Energy Consumption Survey (MECS) [1] and the Emissions and Generation Resource Integrated Database (eGRID) [2] are the principal public sources on industrial energy use and electricity emissions in the United States. Both are widely used to scale site-level results to sectors and regions. MECS, however, suppresses values that fail sampling-error or confidentiality criteria, and eGRID reports annual inventory rates that differ greatly between subregions.
This study documents what these sources contain cell by cell, how suppression affects sums, and where the California grid lies in the national distribution of emission rates. It reports published statistics only; nothing is imputed, modeled or estimated.
2. Methods
2.1 Sources and extraction
All cells were extracted from MECS 2022 Tables 3.1, 3.2, 5.2, 11.1 and 11.3. From eGRID2023 Revision 2, only the subregion and state emission rates and the grid gross loss values were extracted, 322 records in total; plant, unit and generation tables were not read. Each cell was classified by its status marker as a published number, zero, rounded below the reporting unit, not applicable, not available, withheld for sampling error, or withheld for confidentiality. Extraction was checked against predeclared anchor cells.
2.2 Categories and units
Fuel, feedstock, net electricity, onsite generation and cogeneration are kept as separate categories, following the survey; cogeneration is a component of onsite generation and is never added to it. Fuel quantities remain in their native units, barrels, cubic feet and short tons, because the tables do not state a heating-value basis for conversion.
2.3 Aggregate analysis
For each Census region, the published industry rows of Table 3.2 were summed and compared with the published regional total. The difference reflects cells without published values and independent rounding.
3. Results
3.1 Cell status
Of the 23,409 survey cells, 7,013 are ordinary published numbers, 6,285 are zero and 5,261 are rounded to a small value below the reporting unit; 3,645 are not applicable and 648 not available; 497 are withheld because the sampling error is too large and 60 to protect respondent confidentiality. All 322 extracted eGRID records are published numbers, bringing the total to 7,335 of 23,731 (Figure 1). Withheld cells are few, but they occur in the industry rows most often needed for sector analysis.
3.2 Sums of published cells are not totals
Table 1. Total fuel consumption by Census region, MECS 2022 Table 3.2.
| Region | Published total, TBtu | Sum of published industries, TBtu | Difference, TBtu | Industries without a published value |
|---|---|---|---|---|
| Northeast | 1,274 | 1,271 | +3 | 2 |
| Midwest | 4,213 | 4,212 | +1 | 1 |
| South | 8,306 | 7,884 | +422 | 4 |
| West | 1,948 | 1,949 | -1 | 2 |
In the South the published industry cells account for all but 422 TBtu, 5.1 percent of the regional total, which lies in 4 industries without a published value, withheld or rounded below the reporting unit (Table 1, Figure 2). In the West the published cells sum to 1,949 TBtu against a published total of 1,948: the subtotal exceeds the total. Because independently rounded cells can sum to more than their rounded total, a subtotal of published cells is not a lower bound, and the sign of the suppressed remainder is not fixed by the source.
3.3 Grid emission rates
Across the 27 eGRID subregions, the annual total-output CO2e rate ranges from 242.8 to 1,548.5 lb/MWh, a factor of 6.4, with a median of 867.7 lb/MWh. The CAMX subregion, at 429.983 lb/MWh, ranks 2 of 27 from the cleanest; 25 subregions have higher rates (Figure 3). Grid gross loss ranges from 4.1 to 4.4 percent across the interconnections.
4. Discussion
MECS establishes the scale of industrial fuel use and cogeneration by region and industry, but not to the last unit. Suppressed cells should be carried as unknowns rather than zeros, and published subtotals should not be treated as bounds on totals.
The spread of grid emission rates matters for any study of industrial electricity. A comparison between onsite generation and grid import, or between a gas boiler and an electric one, can change sign between subregions. A result computed on the CAMX average applies to one of the cleanest grids in the country and does not transfer to the median subregion. These are annual inventory rates, appropriate for emissions inventories [3]; they do not describe what an additional megawatt-hour displaces in a given hour [4].
5. Limitations
- Only 5 survey tables and one inventory were read; feedstock energy is not in these tables and is left unpopulated.
- Census regions, eGRID subregions and states are different geographies and are not joined.
- Extraction was checked against a small sample of anchor cells; layout changes elsewhere would not be detected.
- The atlas reports published statistics and estimates no savings.
6. Conclusions
Fewer than a third of the cells in the federal manufacturing-energy tables examined are ordinary published numbers, and sums of published cells are not totals. California's grid lies near the clean end of the national range of emission rates. Both facts constrain how site-level results can be scaled or compared.
References
- U.S. Energy Information Administration. Manufacturing Energy Consumption Survey 2022: Data Tables. Washington, DC.
- U.S. Environmental Protection Agency. Emissions and Generation Resource Integrated Database, eGRID2023 Revision 2. Washington, DC.
- World Resources Institute and World Business Council for Sustainable Development. GHG Protocol Scope 2 Guidance. Washington, DC, 2015.
- Siler-Evans K, Azevedo IL, Morgan MG. Marginal emissions factors for the U.S. electricity system. Environmental Science and Technology, 46(9):4742-4748, 2012.