Industrial production (IP) measures the volume of output produced by a defined set of industrial sectors. It gives a quant researcher a timely view of activity in factories, mines and utilities, useful for forecasting cyclical demand or testing macroeconomic exposures. Its coverage is narrower than gross domestic product; rising IP does not establish that the whole economy grew at the same rate.
What the US measure includes
Definitions depend on the publisher and country. The US Federal Reserve's monthly G.17 index covers manufacturing, mining, and electric and gas utilities. Its manufacturing definition also retains some historically industrial activities, including logging and certain publishing industries. Construction and most services are outside this measure. It measures real output, rather than sales revenue in dollars. The Federal Reserve's G.17 overview explains its scope and role in reading the business cycle.
The Fed builds industry estimates from physical production where suitable, supplemented by input measures such as production-worker hours and, for some industries, price-adjusted nominal output. It combines them using a chain-type Fisher index with value-added weights. This is more informative than adding tonnes, barrels and worker hours or taking an unweighted average of sector growth rates. For exact sector contributions, use the publisher's contribution data. See the current industrial-production explanatory notes.
An index level is not a growth rate
A volume index expresses output relative to a reference period. In the methodology checked for this article, the Fed uses 2017 as its reference year. Always retain the base label supplied with the downloaded series. A reference value of 100 is a normalisation, rather than full industrial capacity or a currency amount.
Consider fictional consecutive monthly observations on the same basis, both seasonally adjusted: an output index rises from 100 to 102. Its monthly growth is
The index rises two points; the growth rate is 2%. Points and percentages coincide numerically here only because the starting value is 100. This is a month-on-month change, without annualisation. A pure rebasing multiplies every level by the same positive constant, leaving growth unchanged. Revisions that change the underlying estimates or weights can change growth.
Output and capacity utilisation answer different questions
Capacity utilisation measures output relative to estimated sustainable maximum output. For compatible US output and capacity indexes, utilisation is their ratio expressed as a percentage. Capacity means a realistic operating schedule allowing normal downtime and sufficient inputs, rather than a plant's theoretical maximum. The Fed's capacity notes explain the distinction.
Fictional seasonally adjusted monthly example. Output and capacity indexes share a compatible scale; capacity grows alongside output, leaving utilisation unchanged.
Python makes the two denominators explicit:
output = [100.0, 102.0]
capacity = [125.0, 127.5]
growth = output[1] / output[0] - 1
utilisation = [q / c for q, c in zip(output, capacity)]
print(f"Monthly output growth: {growth:.1%}")
print(f"Utilisation: {utilisation[0]:.1%} -> {utilisation[1]:.1%}")
# Monthly output growth: 2.0%
# Utilisation: 80.0% -> 80.0%Use the information available at the forecast date
Seasonal adjustment removes estimated recurring seasonal effects; it does not remove every temporary disruption. Check industry detail before attributing a monthly move to broad demand. The Fed seasonally adjusts individual series and aggregates the adjusted components. Published percentage changes use unrounded indexes, so rounded displayed levels may not reproduce them exactly.
IP estimates are preliminary and revised as information arrives, with broader annual revisions as well. A historical observation month and its publication date are different dates. In a backtest, retain the release timestamp, seasonal-adjustment basis and data vintage; using subsequently revised history as if it were known earlier creates look-ahead bias. The G.17 release archive separates monthly releases from annual revisions.
For an advanced application, Real-time macro forecasting requires sign-in and membership. The capstone forecasts US IP growth using information available at each forecast origin and separates first-release outcomes from revised targets.
Sources
- Federal Reserve — About G.17: coverage, reference year and index construction.
- Federal Reserve — Industrial production explanatory notes: source data, weighting, adjustment, revisions and rounding.
- Federal Reserve — Capacity utilisation explanatory notes: capacity definition and utilisation calculation.
- Federal Reserve — G.17 releases and archive: publication schedule and monthly/annual vintages.