UN Comtrade paper exports: who shipped which grade to whom

"UN Comtrade paper exports" names a very specific object: bilateral merchandise flows, compiled from national customs submissions and harmonised under the Harmonized System, for the two chapters that carry the paper value chain. HS Chapter 47 covers pulp of wood or of other fibrous cellulosic material, recovered paper included; HS Chapter 48 covers paper and paperboard together with the articles made from them. Codes run from the two-digit chapter down to individual six-digit lines - newsprint, kraft liner, cartonboard - and each line is observable at any reporter-by-partner pairing in either flow direction, with re-export and re-import codes sitting alongside the main export and import flows.

Datadory structures exactly those observations as typed rows under UN Comtrade international trade statistics, the anchor record of its paper products shelf: 9 out of 10 on the Datadory rubric, the highest mark among the slice's seven primary datasets and one shared by just 534 of the 1,744 records in the catalog. The wider cube moves tens of millions of cells a year across all commodities; Chapters 47 and 48 are the paper cut of that total, and they load through precisely the same schema as everything else - which is why a sixty-year series and last month's flow sit side by side in one table rather than two.

Two cataloged records report these flows, and they separate cleanly on one axis: fidelity versus agreement.

UN Comtrade international trade statistics is the fidelity play. Every cell keeps what the reporting customs administration filed - the dollar value, the net weight, the supplementary quantity with its unit and estimation flag, the customs procedure and the mode of transport - and annual series reach back to 1962, varying only by which classification revision a reporter applied when. Nothing else in the slice carries pre-1995 paper history at all.

OEC - Paper, paperboard and articles trade profile is the agreement play. Built on CEPII's BACI reconstruction of the same underlying flows, it resolves the disagreements between reporters into one usable figure per exporter-importer-product-year, at six-digit depth from 1995 through 2024, scoring 7 out of 10. Analysts usually end up needing both - which is why the two arrive with joinable keys rather than competing definitions.

Cells exactly as the record returns them - from the verified chapter 48 pull for one reporter-year:

# One bilateral cell - China reporting, 2022, chapter line
refYear             : 2022         classificationCode : H6   (HS 2022)
reporterCode        : 156          (China)
partnerCode         : 132          flowCode : X export
cmdCode             : 48           customsCode : C00
primaryValue        : 1,189,670    USD, fob on export rows
netWgt              : 0            kg - weight not supplied by the reporter

# Two neighbouring partner cells from the same pull, identical column shape
partnerCode 748     : primaryValue 122,326 USD   customsCode C00
partnerCode 678     : primaryValue 62,158  USD   customsCode C00

# Columns that ride on every cell regardless of commodity
qty / altQty        : supplementary quantities, each with qtyUnitCode and an isQtyEstimated flag
motCode             : mode of transport for the crossing

That single reporter-year resolved to 210 partner-level cells, and three things in the shape settle questions buyers usually ask twice. Valuation rides on direction: export rows populate primaryValue fob while import rows carry cif, so a mirror comparison is a decision about freight treatment rather than a puzzle. Estimation lives in flags rather than footnotes - isQtyEstimated separates a figure the reporter supplied from one filled in upstream, and the estimation flags deserve reading before any arithmetic. And the zero-weight cell above is the honest case, not a defect: some reporter-years simply declare value without mass, which makes value-only analysis the correct option there and unit-value division impossible.

Every field above is a documented definition with a verified example behind it, not an inference from column naming. Get a sample cut to your own reporters, partners, years and chapter lines before anything else is discussed.

• Time - annual cells from 1962 onward, depending on when a reporter adopted a given classification revision; recent periods exist at monthly resolution, so intra-year sight is available where the question needs it. Six decades of who-shipped-what-to-whom is the longest continuous paper record anywhere in the catalog.
• Geography - reporters span the UN membership and beyond, and every reporting economy also appears as a partner in everyone else's filings, so the bilateral matrix closes on itself rather than trailing off after the top twenty traders.
• Granularity - reporter x partner x period x HS code x flow, with the commodity axis cutting at chapter, heading or six-digit line. Nothing exists below the pair: no consignee, no port, no counterparty company. Treat the pair as the floor and plan around it.

Stitching is the discipline that separates citable series from bar-chart filler. Each cell stamps the Harmonized System revision in force - classificationCode H6 marks HS 2022 in the verified pull - so a sixty-year panel crosses several revision boundaries, and a series built without pinning vintages quietly misstates years at each seam. Date-stamp every figure with its revision and reference year, and blended multi-record cuts get labelled with their mix rather than covered by one label.

The two trade records disagree usefully, and the disagreement has a shape worth knowing before quoting either. The same shipment enters the record twice - the exporter's fob declaration and the importer's cif declaration, separated by freight and insurance - so bilateral totals differ by source side before anyone makes an error. The OEC paper profile, built on CEPII's BACI harmonisation, exists precisely to resolve that: one agreed value per exporter-importer-HS6-year, 1995 through 2024, with seven classification revisions stitched into a continuous panel and a separate current-revision cube covering 2022-2024.

The working split: the customs-reported series when a question needs weights, customs procedure or anything before 1995; the harmonised cube when it needs comparable cross-country values after 1995.

A trade cell counts what crossed a border. It says nothing about who milled the fibre, how much energy the mill burned, or which company sold it - and four neighboring records close exactly those gaps.

Production. Trading Economics - Paper Production by Country gives one annual production indicator per country with a latest print and multi-year history - a ranking layer at 4 out of 10, lighter than anything else here. Grade-level American output belongs to AF&PA statistics & resources: forty-plus report products across nine categories from containerboard to recovered fibre. European context comes from CEPI Key Statistics, whose latest edition tracks 77.4 million tonnes of paper and board across nineteen member countries back to 1991.

Facilities. US EPA Envirofacts facility database resolves the American mill estate - 353 facility records matching 'PAPER', each fan-out across roughly twenty program systems with registry identifiers and coordinates attached. US EPA Toxics Release Inventory adds what each mill released: facility x chemical x reporting year for NAICS 322 across thirty-eight years, 1987 through 2024, split by air, water, land and off-site transfer with parent-company names on the row.

Energy. US EIA Manufacturing Energy Consumption Survey (paper) benchmarks the sector every four years, 1994 through 2022, returning NAICS 322, paper mills and paperboard mills measured across the same eight fuel columns with census-region cuts alongside.

Fibre and counterparties. Upstream, FAOSTAT Forestry - Forest Production and Trade measures production and trade of pulpwood and paper grades across 104 items and 285 areas from 1961. Downstream, the PaperIndex paper industry directory lists roughly 6,633 supplier profiles - manufacturers, converters, waste-paper traders, machinery vendors - with buyers cited from 188 countries.

Stacked, the layers turn "China exported $1.19 million of chapter 48 goods to one partner" into a question about which mill regions, which fuel mix and which named suppliers sat behind it.

Market researchers and consultants size end markets from official prints: chapter totals and bilateral dependencies anchor the trade chapter of a market entry deck, with the production and association records supplying the domestic denominators. Their page: paper products data for market researchers.

Investors and quant researchers build factors on dollar-precise bilateral values - export momentum, import penetration, pulp-versus-articles mix, exporter concentration across cycles - with sixty-plus annual points per pair making long-window backtests possible. Their page: paper products data for investors and quants.

Competitive intelligence and product teams watch grade-line shifts between named country pairs months before they surface in a competitor's earnings call, and pair the flow read with facility registers to see who could serve the demand. Their page: paper products data for competitive intel and product teams.

Sales and growth teams work the same matrix from the account side: which destinations import which grades, then which named mills and converters to call. Their page: paper products data for sales and growth teams.

Data scientists and journalists inherit estimation flags, revision stamps and citation-shaped rows - the difference between a model feature you can defend and one you cannot. The workflow collection lives on data scientists use cases; the common jobs are market sizing, demand forecasting, competitor tracking and citation-grade research.

Start with the anchor record itself: UN Comtrade international trade statistics, sampled to your reporters and years before anything is committed, with OEC - Paper, paperboard and articles trade profile as the harmonised counterpart. For the fence-line-versus-border contrast, the Comtrade versus Envirofacts head-to-head scores the two approaches field by field.

This page is one thread of a wider map. The paper products data guide walks the whole slice - trade, production, energy, emissions and directories - the best paper products datasets ranking lines up all nineteen pooled records, and the paper products data hub indexes every record with its coverage statement.