Industrial Production vs. Capacity Utilization: G.17 Explained

The Federal Reserve publishes the industrial production (IP) index and capacity utilization in the same monthly G.17 release because the measures describe related but different parts of industrial activity. IP tracks the level of real output relative to a reference period. Capacity utilization compares that output with an estimate of how much the covered industries can sustainably produce. One is an output index; the other is a ratio expressed as a percentage. The [Federal Reserve's G.17 overview]({source:fedG17About}) defines the two measures and their common industry coverage.

The distinction matters when both output and productive capacity change. If output rises, the IP index can increase even if capacity expands more quickly and utilization declines. If output falls while capacity contracts by a larger proportion, utilization can rise. Neither movement is automatically an error. Each answers a different question: how did real output change, and how much of estimated sustainable capacity was being used?

The labels can also mislead. A production index of 103 is not 103 percent of capacity, and a utilization rate of 80 percent does not mean output fell 20 percent. To interpret a G.17 headline, first identify the index or rate, its reference period, whether it is seasonally adjusted, and which industry aggregate it describes.

The IP index measures real output in manufacturing, mining, and electric and gas utilities. Its reference period is 2017 in the September 2026 release: the total index is scaled so that average output in that year equals 100. An index level of 103 therefore means measured output is about 3 percent above the 2017 reference-year average for that series. It does not mean output grew 3 percent in the latest month, and it does not say factories are operating at 103 percent of capacity. The Fed's [industrial production notes]({source:fedG17IndustrialProductionNotes}) explain the output concept and index construction.

The index is designed to track production volume rather than the dollar value of sales. The Fed draws on physical output measures where suitable, such as quantities of products, and on indicators of production inputs when direct monthly output measures are unavailable. Production-worker hours, for example, can help estimate output for some industries. Certain series use other methods, so IP is a constructed measure informed by multiple sources rather than a monthly census of every item produced.

Industry series are combined using a Fisher index approach, with weights related to industries' value added. As a result, a one-percent change in a large industry can affect total IP more than the same percentage change in a small industry. The total can also conceal different paths among manufacturing, mining, and utilities. An IP index is not the same as nominal industrial sales, a broad measure of all U.S. production, or GDP. It focuses on a defined industrial sector and aims to remove price change from output measurement.

The reference year is a scaling convention, not a claim that the year was typical or that 100 is a target. It can change when the Fed rebases or revises the indexes. As of September 27, 2026, the release used 2017=100 and announced a planned annual revision that would set the new reference year to 2022. When comparing a series across releases, read the reference-year label and revision notice rather than assuming the index base never changes. The [current G.17 release]({source:fedG17CurrentRelease}) carries the reference-year and revision information.

For an industry or aggregate, the rate is calculated as seasonally adjusted output divided by the corresponding capacity index, multiplied by 100:

Capacity utilization = seasonally adjusted output index / capacity index × 100

The numerator represents output; the denominator represents estimated sustainable capacity. The denominator is not a count of installed machines or a plant's absolute engineering limit. The Fed defines capacity as the greatest level of output an establishment can sustain within a realistic work schedule, after normal downtime and assuming adequate inputs are available to use the capital in place. See the [capacity-utilization notes]({source:fedG17CapacityUtilizationNotes}).

That definition makes utilization an estimate, not a direct monthly meter reading of every production line. The Fed uses a mix of capacity data, including physical-capacity information for some industries and survey information for many manufacturers, together with methods for industries where direct observations are limited. The capacity indexes are built from periodic source information and updated over time; the monthly rate combines them with monthly production estimates. A rate of 90 percent therefore means that the measured output index is 90 percent of the associated estimated capacity index under the Fed's definitions.

An aggregate utilization rate also reflects how component industries are weighted. It is not necessarily the simple average of each industry's rate. Two industries can have the same utilization percentage but very different output weights. A rising total rate can coexist with slack in some industries, while a high rate in a small industry may have little effect on the aggregate. Always check whether the number is for total industry, manufacturing, mining, utilities, or a narrower group.

The Fed defines the industrial sector for G.17 as manufacturing, mining, and electric and gas utilities. Manufacturing follows the North American Industry Classification System with some historically included activities: logging and certain newspaper, periodical, book, and directory publishing are included, while exclusive Internet publishing is excluded. Mining and utilities also have defined industry boundaries. The [current release notes]({source:fedG17CurrentRelease}) summarize these classifications.

Construction itself is not one of the three industrial industries in total IP. The G.17 release does show market-group indexes such as construction supplies, which are products used in construction. That label describes a group of goods, not an index of construction projects or the entire construction industry. Market groups organize products and materials by type or use; industry groups classify the establishments producing them. Confusing these two views can make the release appear to cover more of the economy than it does.

Services, agriculture, and most construction activity are outside the industrial-sector total. This does not make them unimportant to the economy; it means the IP index is not meant to represent all production. For a broader production measure, use national accounts such as GDP and be explicit about the different frequency, coverage, and construction of those statistics. Related guides on real GDP growth rates and nominal versus real GDP address those separate measures.

Consider a fully hypothetical two-period example. In Period A, suppose the seasonally adjusted IP index is 108 and the related capacity index is 120. Utilization is 108 ÷ 120 × 100 = 90 percent. In Period B, suppose output rises to 110 while estimated capacity rises to 125. Utilization is then 110 ÷ 125 × 100 = 88 percent.

The production index increased by (110 ÷ 108 − 1) × 100, or about 1.85 percent. Yet capacity utilization fell from 90 to 88 percent, a decline of 2 percentage points. These statements can both be true because output rose more slowly than the capacity estimate. The capacity index grew about 4.17 percent, compared with the IP index's approximately 1.85 percent.

The percentage-point description is important. Moving from 90 percent to 88 percent is a 2 percentage-point decrease. Relative to the original 90 percent rate, it is a decrease of about 2.22 percent. In a report, say which calculation you mean. Do not subtract the utilization-rate movement from the output-growth rate as if they were changes in one series.

The figures are classroom arithmetic, not Federal Reserve data, a forecast, or evidence about a particular industry. In an actual release, use the matching seasonally adjusted output and capacity series, and remember that published values may be rounded. The Fed calculates percentage changes from unrounded indexes, so recomputing a result from displayed rounded levels may not reproduce the printed figure exactly.

<!-- learn:illustration -->

Industrial output has recurring seasonal patterns. Utilities, for example, can respond to seasonal demand, while factory production can be affected by holidays, model-year schedules, and planned shutdowns. Seasonal adjustment attempts to remove predictable calendar patterns so a monthly change is easier to compare with the preceding month. The Fed seasonally adjusts individual series and builds seasonally adjusted aggregates from them; its [industrial production notes]({source:fedG17IndustrialProductionNotes}) describe the adjustment process.

The G.17 capacity-utilization rate uses seasonally adjusted output divided by the corresponding capacity index. When comparing monthly IP levels or changes, use a consistently adjusted series and read the table heading. A seasonally adjusted month-to-month change answers a different question from a not-seasonally-adjusted year-over-year comparison. Do not pair a seasonally adjusted numerator with a not-seasonally-adjusted series chosen from another table and assume the ratio is comparable.

Seasonal adjustment does not remove every short-lived disruption or explain why output changed. A strike, severe weather event, supply interruption, or unusual shutdown may affect activity in a way that differs from typical seasonal patterns. It is useful to check industry detail and the release commentary before treating a single monthly move as a broad turning point. Percentage changes should also be read with the series' frequency and comparison period in view.

The first estimate for a month is generally released around the middle of the following month and is marked preliminary. More source information arrives later, so the Fed revises an estimate during each of the next five monthly releases. It may also update indexes, seasonal factors, weights, classifications, and methods during an annual or major revision. The [monthly-procedures summary]({source:fedG17MonthlyProcedures}) describes the production workflow, while the [revision-history documentation]({source:fedG17RevisionHistory}) shows how initial and later estimates are recorded.

This means an economic statement should identify its data vintage when timing matters. A contemporaneous account can use the figure available on the release date; a current historical chart may contain later revisions. Those are different information sets. If assessing what analysts knew at the time, do not silently replace the original estimate with today's revised value. If comparing long historical periods, use a consistent current vintage and note that the Fed may have made broader revisions.

The current release labels estimates as preliminary or revised and posts notices when broader updates are scheduled. Its September 2026 notice, for example, planned an annual revision for November 24, 2026, including new source data, seasonal factors, and a change in reference year. That is a dated example of how the series can change, not a permanent schedule or a reason to compare index levels without checking their bases. The [current release and notice]({source:fedG17CurrentRelease}) are the place to verify the vintage and current base year.

Use IP to describe the change in real output within the industries covered by G.17. Use capacity utilization to describe output relative to estimated sustainable capacity. Pairing them can help distinguish output momentum from how fully available capacity is being used. For example, output could rise while utilization falls if capacity expands faster; utilization could rise while output is flat if the capacity estimate falls.

Neither measure alone gives a complete reading of the economy. A high utilization rate is not a universal threshold for inflation or a guarantee that production must fall. A decline is not proof of recession, and IP does not include every service or construction activity. Compare the release with its history, industry detail, other indicators, and the question being asked. The G.17 release itself supplies historical averages and industry breakdowns, but those comparisons should be made on a consistent basis.

Before quoting a number, record five items: the series and industry group, the index or rate, the start and end dates, seasonal adjustment, and the release vintage. For index levels, include the reference-year base. For utilization changes, use percentage points when describing the difference between two rates. If the question is about broad economic production or consumer prices, choose a measure designed for that question; the guide to [producer and consumer price indexes](/learn/ppi-vs-cpi-producer-prices-vs-consumer-inflation-explained) covers a separate measurement family.