Paper mill locations database: every US mill, delivered as rows

It is a table where each row is one physical mill - name, street address, county, coordinates, corporate owner, operating status - built to be joined against, not a brochure page built to be read. That distinction decides whether a mill list can power supplier verification, territory routing or an emissions screen, and most of what circulates under this keyword cannot survive the join.

The strongest mill-level spine in Datadory's Paper Products slice is the US EPA Envirofacts facility database, scored 8 of 10 on our rubric. Envirofacts integrates roughly twenty EPA program systems - TRI for toxic-release facilities, RCRAInfo for hazardous-waste handlers, ICIS-AIR for permitted emitters, SDWIS for public water systems, SEMS for Superfund sites - stitched together by the Facility Registry Service so one physical site can be followed across programs. A verified query against its TRI facility register for names containing 'PAPER' returned 353 real paper-mill records, each arriving with 48 columns of identity, location, ownership and status.

Within Datadory's catalog of 1,744 datasets across 159 viable industries this is one of seven primary Paper Products records, and it is the only primary organized around physical sites. The rest describe trade flows, energy consumption, company profiles or country output - context that becomes far more useful once every row has an address to land on.

Three records carry the mill-level work, ranked by how directly they answer a locations query:

1. US EPA Envirofacts facility database. The who-and-where layer: one record per reporting facility across roughly twenty program systems, with TRI reporting years from 1987 in the same tables. Its 353 'PAPER' matches make it a usable universe definition for American mills - specialty plants like Kanzaki in western Massachusetts, recycled-board operations like Howard Paper Mills in Dayton, Ohio, converted-products makers like American Eagle in central Pennsylvania.

2. US EPA Toxics Release Inventory (paper manufacturing sector). Also scored 8: one row per facility x chemical x reporting year for NAICS 322, covering reporting years 1987 through 2024. Every quantity is pounds, split by air, water and land route, with off-site transfers, production-related waste managed and the parent company carried on each row - the what-they-released record behind every address.

3. PaperIndex paper industry directory. The commercial counterparty layer beyond the US border: roughly 7,125 listed suppliers across 8,456 categories claimed by the platform, with just over 6,600 individual profiles resolved in the August 2026 research pass - mills, merchants, exporters, converters, waste-paper traders, chemicals suppliers and machinery vendors, with buyers cited from 188 countries.

Together they cover who exists, where it stands, what it reported releasing and who sells into the chain - which is more than most mill-database searches expect to find in one place.

Three of the 353 paper-mill records, resolved live during research:

tri_facility_id : 01082KNZKS20COM
facility_name   : KANZAKI SPECIALTY PAPERS INC.
address         : 20 CUMMINGS ST, WARE, MA 01082
registry_id     : 110009336931
coordinates     : 421502 / 721446  (DMS packed into integers)

tri_facility_id : 45401HWRDP115CO
facility_name   : HOWARD PAPER MILLS INC
address         : 115 COLUMBIA ST, DAYTON, OH 45401
registry_id     : 110002113766

tri_facility_id : 16686WSTVC1600P
facility_name   : AMERICAN EAGLE PAPER MILLS
address         : 1600 PENNSYLVANIA AVE, TYRONE, PA 16686
registry_id     : 110000332186

Twelve field groups ride on every record, and four structural facts decide whether the rows model cleanly. First, identifiers stack: tri_facility_id keys release reporting, epa_registry_id links the site across EPA programs, and frs_id appears where populated - joins run on these, never on fuzzy name matching. Second, coordinates come in two forms: the fac_* pair packs degrees-minutes-seconds into integers (421502 is a position, not a latitude), while the pref_* pair holds best-available decimal degrees for distance work. Third, ownership is first-class: parent company name, DUNS number and a standardised parent field roll plants up to corporate groups. Fourth, closure is explicit - the fac_closed_ind flag keeps shut mills queryable as history rather than letting them vanish from the universe.

The TRI slice extends the same mill into chemistry: chemical name with CAS number, reporting year, on-site releases split by route, off-site transfers, waste managed, pollution-prevention quantities and parent owner - eleven field groups keyed one row per facility x chemical x year, so a single plant becomes a time series rather than a static entry.

Four workflows consume these rows directly, and each fails quietly on a ZIP-code approximation:

• Supplier verification for packaging buyers. Confirm a converter's mill actually exists at the stated address, with coordinates, registry ID and closure flag attached - the fastest way to retire a risky line item on a supplier form.

• ESG and exposure screening. Join TRI's facility-by-chemical reports (1987-2024) onto the register through shared identifiers, then roll facility-level discharge up to holding companies via the parent fields - one coherent footprint per group instead of three reconciliations.

• Territory building for vendors. Segment the 353-mill universe by state, county FIPS and EPA region to route equipment, chemical and service coverage; the directory adds named contacts and revenue bands for the call sheet itself.

• Site screening and due diligence. Coordinates plus the closure flag answer both halves of a property question: what operates within a radius today, and what operated there before. Insurers price environmental liability the same way, at mill rather than ZIP resolution.

Quant researchers add a fifth use: release trajectories as risk signals. A facility-chemical pair whose reported quantities climb for three consecutive reporting years is a measurable exposure with a paper trail - factor input that only exists when geography and chemistry share one schema.

Envirofacts ends at the US border, and the PaperIndex paper industry directory is the slice's company-level record beyond it. Each listing resolves to one structured profile carrying eleven field groups: identity, location block, membership tier, founding year, legal structure, supply-chain role, overview text, products, self-reported revenue band, named contact with title, and full postal address. Member-since dates run back to November 2006, which makes platform tenure a measurable field rather than a guess.

Two cautions keep the record honest. Supplier facts are self-reported by members - a revenue band is a screening signal, not an audited financial - and supplier gravity sits firmly in India and wider Asia while buyer registration spans 188 countries, so a European or North American sourcing question deserves an explicit scope check before anyone builds on the counts.

Trading Economics — Paper Production by Country answers a different question entirely: output tonnage, not coordinates. One row per listing economy - country, latest annual figure, multi-year historical series, last-observed stamp - grouped under World, Europe, America, Asia, Africa, Oceania and G20 views, normalized into a four-field table. It is the fastest route from nothing to a defensible cross-country production picture, and it joins to the mill register on nothing more exotic than a country name.

An address list earns its keep once something flows through it, and three companions supply the flows.

UN Comtrade international trade statistics - the slice's top score at quality 9 - carries pulp of wood (HS Chapter 47) and paper and paperboard (Chapter 48) as reporter x partner x year x HS-code x flow cells, annual series from 1962 onward with monthly detail for recent periods, filed by around 200 economies representing more than 99 percent of world merchandise trade. One verified 2022 cell - China exporting Chapter 48 paper to a single partner - carries a value of 122,326 US dollars; the same row shape repeats across every pairing since the 1960s.

OEC — Paper, paperboard and articles trade profile mirrors the bilateral view from CEPII's BACI cube for 1995-2024, cut to HS six-digit depth: the two paper-side sections account for about $1.23 trillion in pulp and recovered fibre against roughly $4.77 trillion in converted articles - just under $6 trillion of cumulative cross-border trade. Filter one grade, newsprint, to country-year resolution and 3,930 rows come back.

US EIA Manufacturing Energy Consumption Survey (paper) owns the input side for the United States: eight quadrennial editions, 1994 through 2022 with the latest finalised in 2026, reporting NAICS 322 alongside its Paper Mills (3221) and Paperboard Mills (3222) children across eight fuel columns in trillion Btu by census region, with value-of-shipments and employment-size cuts. Establishments stay anonymous by design - which is precisely why mills disclose accurate fuel bills - so the survey pairs with the facility register rather than replacing it. Stack releases over fuel and intensity ratios become arithmetic instead of hand-waves.

Because the registers are public and the working database still isn't. The mill spine arrives as twelve field groups with DMS coordinates already converted to decimal degrees, identifier stacks intact and closure flags preserved; the toxics panel arrives route-split rather than pre-totalled; the directory arrives as typed profile rows with contacts and addresses rather than rendered pages. Every delivery ships with the field dictionary and sample rows for validation, so the schema you test is the schema you ship against.

Delivery follows your pipeline, not the other way around: API, files, or straight into your warehouse - daily, weekly, or hourly, your call. Name the states, NAICS codes, chemicals, parent companies or supplier categories you care about and the sample goes out scoped to them, sized to test against your own join keys before anything recurring starts.

Start small and prove the seam: a packaging buyer might take Ohio and Pennsylvania mills with parents attached; an ESG desk might take one chemical family across all 38 reporting years for its covered facilities. Either way the test worth running is your own join, not a screenshot.

This post is one thread of the Paper Products cluster. For the full seven-primary landscape start at the paper products data hub, then open the best paper products datasets ranking to see why Comtrade, MECS, Envirofacts and TRI hold the top scores.

For record-level detail behind this article, read US EPA Envirofacts facility database for the mill register and US EPA Toxics Release Inventory (paper manufacturing sector) for the chemistry behind each address. The border-versus-fence-line framing is walked field by field in UN Comtrade versus Envirofacts, and provenance for the federal records lives on the US EPA source profile.