How Investors & Quant Researchers Use Paper Products Data

OEC — Paper, paperboard and articles trade profile (quality 7) is the cross-check and the competitive cut. Its Tesseract API serves the verified cube trade_i_baci_a_92 - HS6 REV. 1992, 1995 to 2024, built on CEPII's BACI harmonisation of UN-reported customs data - with drilldowns by Year, Exporter Country and Importer Country against Trade Value, plus a separate REV. 2022 cube covering 2022-2024. In that cube the paper and paperboard HS2 section accounts for roughly 4.8 trillion USD of cumulative trade value, and queries descend to product lines such as HS4=104801 newsprint flows, within a ladder running from 104701 Mechanical Wood Pulp to 104808 Corrugated Paper.

Running both is standard practice on this desk: when the UN-reported and BACI-harmonised values diverge, classification differences or revision timing are usually why, and the divergence is a data-quality flag rather than a tradable signal. One documentation gap to log - oec.world returned HTTP 403 to automated research clients, leaving OEC's own attribution requirements unstated, though BACI itself is freely available for research with CEPII attribution.

Pulp and paper is among the most energy-intensive manufacturing sectors, which is why US EIA Manufacturing Energy Consumption Survey (paper) carries the slice's only relevance-3 rating. MECS is the quadrennial national sample survey EIA describes as the only source of national estimates of manufacturing energy characteristics, spanning eight waves from 1994 to 2022, with the 2022 edition reaching final release on 18 March 2026 after staged preliminary results from March 2025.

The structural caveat goes straight into model code: MECS arrives as survey waves, not a continuous monthly series. Eight observations across 28 calendar years means the effective sample is far smaller than the span implies, so between-wave interpolation needs explicit assumptions and every figure in a memo should name its vintage.

US EPA Envirofacts facility database (quality 8) is the join layer. Its REST service requires no authentication and no key, following the documented pattern data.epa.gov/dmapservice/[program].[table]/[column]/[operator]/[value] with operators such as equals, beginsWith, contains, like and in, and returning JSON by default alongside CSV, Excel, HTML, JSONP, Parquet, PDF and XML. A verified query for TRI facility names containing 'PAPER' returns 353 mill records with names, addresses, coordinates and EPA registry identifiers, each joinable across roughly 20 program systems including RCRAInfo, ICIS-AIR, SDWIS and FRS. The documented execution cap is 15 minutes per request, so batch facility lookups.

Identity resolution is the practical trap: TRI, Envirofacts and PaperIndex name the same mills differently, so match on location-plus-name rather than exact strings before joining facility tables into a panel.