EIA Petrochemical Feedstock & Refinery Olefin Production Data

It is the federal measurement of one specific flow: naphtha that US refineries and blenders produce net for sale as petrochemical feedstock, as opposed to naphtha that stays in the fuels pool. That split matters because feedstock naphtha is the principal light-liquid input to American steam cracking, and its availability is a first-order driver of ethylene and bulk olefin production economics.

The U.S. Energy Information Administration publishes the family under its petroleum refining statistics, compiled from the Petroleum Supply Monthly and the agency's weekly and monthly refinery surveys. One product, many cuts of geography: a United States total (series key MNFRPUS2), five Petroleum Administration for Defense Districts, and the refining districts nested inside PADD 2 and PADD 3 - Indiana-Illinois-Kentucky, Minnesota-Wisconsin-Dakotas and Oklahoma-Kansas-Missouri in the Midwest; Texas Inland, Texas Gulf Coast, Louisiana Gulf Coast and North Louisiana-Arkansas on the Gulf. A discontinued Appalachian No. 1 district rounds out the set. Each series carries its own MNFRP-prefixed key, which is what makes district-level joins mechanical rather than manual.

One row per series per period, values in thousand barrels per day or thousand barrels depending on the view. The national annual series reads 126 thousand barrels per day for 2022, 131 for 2023, 123 for 2024 and 142 for 2025 - a nineteen-unit jump between 2024 and 2025, roughly fifteen percent more naphtha leaving refinery gates for crackers in a single year. That kind of move is exactly what feedstock-demand models exist to catch.

Regional rows arrive identically shaped. Midwest (PADD 2) printed 13 thousand barrels per day in 2024 against the national 123 - the Gulf Coast carries most of the feedstock weight, and the district columns make that concentration visible rather than anecdotal. Cells a district did not report arrive legend-flagged with the publisher's -, --, NA and W markers, never silently zero-filled, so summing across districts cannot manufacture volume that does not exist.

Sourcekey is the spine - the MNFRP-family identifier that distinguishes the US total from each PADD and district series, and the natural join key against any procurement, trading or model table you bolt on. Date carries the reporting period, end-of-period dated on annual rows. The production value sits in the unit the view selects: thousand barrels for cumulative volume, thousand barrels per day for rate.

Two dimensional fields complete the row. Product is constant here - naphtha for petrochemical feedstock use, product code EPPPN - and exists so this family can sit beside sibling product series without ambiguity. Area records the layout orientation the publisher offered, product-by-area or area-by-product, which sounds cosmetic until a reshaped pivot quietly drops which cut a figure came from. Delivered rows normalize region into its own field so PADD-level filtering is a WHERE clause, not spreadsheet surgery.

Geographically: the United States total, then PADDs 1 through 5 - East Coast, Midwest, Gulf Coast, Rocky Mountain, West Coast - then the districts inside them, with the Gulf Coast split finest because that is where the crackers concentrate. Temporally, the national series reaches from 1983 through 2025, forty-three annual observations, with monthly views layered underneath. Regional series generally begin in 1993; some run shorter, and Appalachian No. 1 is discontinued, which is why every delivered table carries its series status rather than padding gaps with zeros.

Granularity is deliberately bounded: monthly and annual totals per product per geography. There is no facility level, no company attribution, no plant throughput anywhere in the family - questions at that altitude belong to companion sources such as the EIA Refinery Capacity Report, which pairs cleanly against these volumes.

Through Datadory, on your terms: API, files, or landed directly in your warehouse. Daily, weekly, or hourly - the cadence is your call, not the publisher's. Rows arrive normalized with region, product, unit and frequency as explicit fields, legend-flagged where the source withholds, and documented against the field dictionary above so the first join works on day one.

Feedstock and margin analysts read the national and Gulf Coast series as the supply side of US ethylene economics - when net feedstock production moves fifteen percent in a year, cracker utilization and derivative margins move with it. Energy market researchers track the PADD splits to see where naphtha demand competes with gasoline blending inside the refinery slate.

Corporate development teams use the district detail when siting or valuing Gulf Coast and Midwest petrochemical assets, pairing feedstock availability against the capacity report. Procurement and supply-chain planners at naphtha consumers benchmark their own purchases against measured national and regional flows. Journalists covering the American petrochemical build-out cite the annual prints because they are federal measurements, not trade estimates. Quant desks fold the series into chemical-equity factor models as a hard supply signal.

Investors and quants get a federal, monthly-resolution supply input for chemicals coverage. Competitive intelligence and product teams benchmark whether feedstock conditions in a district support the expansion story a competitor is telling. Market researchers and consultants anchor petrochemical demand studies in measured flows rather than consultant estimates. Data scientists and ML engineers inherit a clean panel - series key, period, geography, value - ready for feature pipelines. Journalists and academics get citable federal numbers with four decades of history.

Because the raw family is a lattice of parallel series views - product-by-area, area-by-product, four unit-and-frequency combinations per layout - and reconstructing one honest table from it is easy to get subtly wrong. Datadory hands you the normalized version: one row per series per period, region and unit explicit, withheld cells flagged, district status documented. Name the PADDs, districts and periods you need and a sample comes back shaped like your model expects, with companion sources - the wider petroleum portal, the refinery capacity report - joinable on the same geographic keys.

The cards below sit closest to this dataset in our catalog: the broader petroleum portal that frames these series, the capacity report that supplies the denominator, property references for the molecules themselves, and the vertical ranking that shows where this family lands among commodity chemicals sources.