What Separates 137.0 From 113.7 Bcf/d, and Which Number the Federal Reserve Uses

On October 2, 2026 the U.S. Energy Information Administration published an item headed “U.S. natural gas production reached a record high in July 2026” and put the record at 137 billion cubic feet per day. That is gross withdrawals: the full well stream as it leaves the ground. The same July column in EIA's own production table also carries a marketed production figure and a dry production figure, and both are smaller. Divided by July's 31 days they come to 125.2 Bcf/d and 113.7 Bcf/d. The distance from the headline to the consumer-grade volume is 23.3 Bcf/d, which is 17.0 percent of the record.

Three line items in that table are defined as the deductions that bridge the first two volumes. Each prints NA. The difference they are supposed to explain is 366,894 million cubic feet, recoverable only by subtraction. That record was announced in EIA Today in Energy, U.S. natural gas production reached a record high in July 2026. The question underneath it is which of the three volumes the Federal Reserve's industrial production index consumes. The answer is the middle one.

Three Volumes for One July

The national table, EIA, Natural Gas Gross Withdrawals and Production, is headed “Volumes in Million Cubic Feet” and carries a release date of September 30, 2026. The displayed columns run Feb-26 through Jul-26, with no January 2026 column. Reading down the Jul-26 column: Gross Withdrawals 4,247,306; Marketed Production 3,880,412; NGPL Production, Gaseous Equivalent 355,515; Dry Production 3,524,896.

Dividing by 31 days gives 137.0, 125.2 and 113.7 Bcf/d, and the first ties the table to the press item: 4,247,306 over 31 is 137,009.9, the announced 137 Bcf/d. The two gaps are 366,894 million cubic feet between gross withdrawals and marketed production and 355,516 between marketed and dry, or 11.8 Bcf/d and 11.5 Bcf/d. Together, 722,410 million cubic feet: 23.3 Bcf/d, 17.0 percent of gross withdrawals, split 8.6 and 8.4 percent.

One detail belongs here, because it gets over-read. Marketed production less dry production is 355,516, while the published NGPL Production, Gaseous Equivalent cell says 355,515. The 1 million cubic feet residual is rounding, not a disagreement between tables.

The Three Deductions That Print NA

The EIA Natural Gas Monthly glossary defines the middle volume this way: “Gross withdrawals minus natural gas used for repressuring, quantities vented and flared, and less nonhydrocarbon gases removed in treating or processing operations.” It then adds an inclusion that is easy to skip: “Includes all quantities of natural gas used in field and processing operations.” So the definition names exactly three deductions and one thing that stays in. Gas burned to run a compressor at a processing plant is not deducted; gas burned in a flare is.

The table has a row for each deduction, in this order: Repressuring, Vented and Flared, Nonhydrocarbon Gases Removed. In each of the six displayed months, February 2026 through July 2026, each of those rows prints NA, which the table's own legend expands as Not Available. The same holds for From Gas Wells, From Oil Wells, From Shale Gas Wells and From Coalbed Wells. A reader who wants to know how much of the 366,894 was flared cannot get it here.

EIA, Definitions, Sources and Explanatory Notes for these tables does answer why, and twice. It records that for 2011, where state repressuring, vented and flared, nonhydrocarbon or lease fuel figures were unavailable, the agency estimated them from each state's historic ratio, and that from 2012 forward non-marketed components are reported as NA if not available directly from states. It also records that from 2017, state monthly marketed production is estimated from gross withdrawals using historical relationships. Both notes are written at state level and neither describes the national figure, though the table itself flags production series data from 2021 forward as estimates. So the 366,894 is a residual between two published national totals, the second resting on state estimates whose per-state composition this piece has not opened. The notes are on the page. The components are not on the table.

Six Reductions, Two Groups, One Identity With Two Names

The glossary's entry for the bottom volume is longer, and it does not reuse the three-item list. It opens with “The process of producing consumer-grade natural gas” and then says: “Natural gas withdrawn from reservoirs is reduced by volumes used at the production (lease) site and by processing losses.” Each group is then enumerated. The lease-site group: “Volumes used at the production site include the volume returned to reservoirs in cycling, repressuring of oil reservoirs, and conservation operations as well as natural gas vented and flared.” The processing group: “Processing losses include nonhydrocarbon gases (for example, water vapor, carbon dioxide, helium, hydrogen sulfide, and nitrogen) removed from the natural gas stream as well as natural gas converted to liquid form, such as lease condensate and plant liquids.”

Count them. Four reductions at the lease site, two in processing, six in all, against three in the marketed production entry. Cycling and conservation operations appear in one definition and not the other. The lists count different spans, three deductions for one step and six for the whole path, and lease condensate is in neither total because gross withdrawals is itself defined to exclude it: “Full well stream volume, including all natural gas plant liquid and nonhydrocarbon gases, but excluding lease condensate.” But a reader who learned the three-item list and then met the six-item list has two maps of one wedge, and only one mentions cycling.

There is also an exclusion that decides what the series is not: “We do not consider volumes of dry natural gas withdrawn from natural gas storage reservoirs as part of production.” In a month when storage is drawn down hard, the gas moving through the pipeline system exceeds the production series by the withdrawal, and nothing in the production table shows it. That boundary ran through Record Gas Storage Faces January, and it is the same boundary that makes liquefaction fuel a separate accounting question, taken up in Count Liquefaction Fuel Before Turning.

The sharpest duplication is in the closing identity, which appears in two agency wordings for the same subtraction. The glossary ends: “Dry natural gas production equals marketed production minus natural gas plant liquids production.” The Definitions, Sources and Explanatory Notes page ends: “Dry natural gas production equals marketed production less extraction loss.” The table column that carries the quantity is headed neither way; it is headed NGPL Production, Gaseous Equivalent. Three names, one cell, 355,515. The glossary's NGPL production entry closes the loop by saying the extraction of those constituents is “sometimes referred to as extraction loss”, and that it is “Usually reported in barrels or gallons, but may be reported in cubic feet to compare with dry natural gas volumes.” The gaseous equivalent in the table is that comparison column, not a measured gas volume.

The same explanatory page says it outright: “Flared and vented gas is also considered production.” It then draws the contrast itself, noting that this differs from marketed production, which excludes flared and vented gas. So the volume marketed production deducts is, one definition earlier, production. Whether flared gas counts depends on which row a reader is standing on.

The wedge is not a constant. Running the same subtraction down the six displayed columns and dividing by gross withdrawals gives 17.8 percent in February 2026, 18.2 in March, 18.0 in April, 17.7 in May, 17.1 in June and 17.0 in July. A month that sets a gross withdrawals record is therefore also the month whose stripped-out share is lowest of the six. Both are true of one column.

Which Volume Travels to the Federal Reserve

The three volumes do not travel together. Federal Reserve Board, G.17 Source and Description Table 1.25 covers NAICS 211, oil and gas extraction, and was released on November 24, 2025. The row labelled Natural gas is keyed to NAICS 211130pt, its series type is given as Prod, a monthly product series, and its measurement basis is stated as “Billion cubic feet, marketed production”, with the source line naming the Department of Energy.

That fixes the basis, not the vintage. The index takes the middle volume for natural gas, marketed production in billion cubic feet: 3,880,412 million cubic feet in July 2026, 125.2 Bcf/d. Table 1.25 was released before that column existed and cannot show whether the index now in print has taken it. The basis is not the 137.0 Bcf/d the record was announced at, and not the 113.7 Bcf/d of consumer-grade gas. It is also, by the notes above, the volume whose three bridging components print NA and whose state-level values are estimated from gross withdrawals by historical relationship. A reader who hears record production and then checks the index is comparing a number to a series built on one 11.8 Bcf/d smaller.

Two details on that page are worth keeping. The three crude oil rows, the Propane row and the Liquefied petroleum row carry an asterisk whose footnote reads “Auxiliary index.” The Natural gas row does not. And propane and liquefied petroleum are measured in barrels at gas-processing plants, so the liquids EIA removes to get from 3,880,412 to 3,524,896 reappear on the same page as barrel-denominated series, both of them asterisked. The 355,515 is not lost to the page. It changes units and rows.

The Industry Code Has Two Numbered Branches

The code that the Federal Reserve keys the natural gas series to is worth reading in full, because it decides where the wedge lives. In Bureau of Labor Statistics, 2017 NAICS Industry Definitions for 2111, published for its own survey respondents, the entry for 211130 Natural Gas Extraction reads: “This industry comprises establishments primarily engaged in (1) the exploration, development and/or the production of natural gas from wells in which the hydrocarbons will initially flow or can be produced using normal or enhanced drilling and extraction techniques or (2) the recovery of liquid hydrocarbons from oil and gas field gases. Establishments primarily engaged in sulfur recovery from natural gas are included in this industry.”

Two numbered branches, and the second one is the wedge. The recovery of liquid hydrocarbons from field gases is the activity that produces the 355,515, and it sits inside the same industry code as the drilling in branch one. Sulfur recovery is pulled in by the following sentence. So most of the 23.3 Bcf/d does not leave the industry when it leaves the gas stream.

One piece of it does leave. The same entry carries four cross-references, and the last one reads: “Recovering helium from natural gas--are classified in Industry 325120, Industrial Gas Manufacturing.” Helium is one of the five nonhydrocarbon gases the dry production definition names, alongside water vapor, carbon dioxide, hydrogen sulfide and nitrogen. The other three cross-references send contract gas field services to 21311, petrochemical manufacturing to 325110 and refining to 324110. The boundary EIA draws as a processing loss and the boundary the industry classification draws are not the same, and helium is where they part.

Concrete Framework

Five checks, in the order that catches the most errors soonest.

First, name the row before quoting the number. Gross Withdrawals, Marketed Production and Dry Production are three rows in one column, and 137.0, 125.2 and 113.7 are the same month. A citation that says only production has not said anything yet.

Second, divide by the days. The monthly table is in million cubic feet for the month, so monthly totals and daily rates rank months differently. February 2026 has 28 days: its 3,770,940 is 134.7 Bcf/d against March's 4,195,088 at 135.3. The monthly total rises 11.2 percent, the daily rate 0.5 percent. Month length alone moves the apparent trend.

Third, look for NA before trusting a decomposition. If the components of a published difference print NA, the difference is a residual. The 366,894 is a residual; the 355,515 is a printed cell.

Fourth, read the identity in both wordings. Minus natural gas plant liquids production and less extraction loss describe the same subtraction, and the column header uses a third phrase. Searching a methodology document for one of the three phrases will miss the other two.

Fifth, ask which volume the downstream user took. For the industrial production index the answer is marketed production in billion cubic feet. For a pipeline balance it is dry production plus storage withdrawals, and the storage part is excluded from production by definition. The useful question is not which number is right, but which row the user read.

What to Watch Next Week

The table carries a Next Release Date of October 30, 2026. Three things on it are worth checking.

Whether the Repressuring, Vented and Flared and Nonhydrocarbon Gases Removed rows stay at NA for the July 2026 column, or fill in. If they fill in, the 366,894 becomes a sum and the three pieces can be sized for the first time.

Whether the wedge share continues down. It has fallen in four of the five steps charted above, from 18.2 percent to 17.0 percent. Next is the August 2026 column, and a reversal would mean the composition of the well stream, not the drilling rate, moved.

Whether the record holds in revision. The explanatory notes describe monthly figures as “considered preliminary prior to the publication of the Natural Gas Annual (NGA)”, and describe monthly extraction loss as revised after that annual, with final monthly values allocating the annual total in proportion to each month's marketed production. So the printed 355,515 is not, in final form, an independent monthly measurement, and the dry production resting on it inherits the allocation. The 113.7 Bcf/d is the most derived of the three volumes and the most likely to move.

What Would Invalidate This

The reading above rests on one vintage of one table, and several things would undo parts of it.

If a later vintage populates the three NA rows and their sum does not come to 366,894, then the marketed production figure for July 2026 was revised at the same time, and the gap arithmetic in this piece applies only to the September 30, 2026 vintage. If the Natural Gas Annual reallocates extraction loss and the July 2026 cell moves off 355,515, the 113.7 Bcf/d moves with it and the 17.0 percent does too.

Four things were not verified and should not be read as settled. The full source string for the Natural gas row in Table 1.25 came back from retrieval with what looks like legend text interleaved into a vendor name, so only the measurement basis and the Department of Energy attribution are relied on here. The state-level components behind the national Repressuring, Vented and Flared and Nonhydrocarbon Gases Removed rows were not opened, so this piece cannot say which states report them and which are estimated. The 211130 wording is the 2017 NAICS vintage as BLS publishes it to respondents, and a later vintage could word the two branches differently. And whether the index now in print carries this July 2026 marketed production figure, rather than an earlier vintage of it, was not checked.

Two things would not invalidate anything. The 1 million cubic foot residual noted above is rounding and carries no information. And EIA's reporting of records for marketed production and dry production in the same month does not conflict with any of the above. The point is not that one volume rose and another fell. It is that the three volumes are 137.0, 125.2 and 113.7, that the Federal Reserve takes the middle one, and that the rows which would explain the first step print NA.