Bilateral Trade Data as Reported by Country

This dataset provides bilateral trade flows at the country-country-product-year level. Each record contains both reported and estimated export and import values between trading partner pairs for specific products and years. Data is available in both HS92 and SITC Revision 2 product classifications. Dataset Structure The data includes the following variables: country_id: Numeric identifier for the exporting country country_iso3_code: Three-letter ISO code for the exporting country (e.g., AFG, USA) partner_country_id: Numeric identifier for the importing country partner_iso3_code: Three-letter ISO code for the importing country (e.g., DZA, AND) product_id: Internal product identifier used across datasets product_hs92_code (or product_sitc2_code): Six-digit product code in HS92 or four-digit code in SITC Rev. 2 year: Year of trade flow export_value_reported: Trade value as reported by the exporting country in current USD (thousands), FOB basis import_value_reported: Trade value as reported by the importing country in current USD (thousands), CIF basis export_value: Estimated export value in current USD (thousands), using reliability-weighted methodology import_value: Estimated import value in current USD (thousands), using reliability-weighted methodology Available Classifications The repository includes bilateral trade data in two product classification systems: HS92 (Harmonized System 1992): Six-digit product codes, covering 1995-present SITC Rev. 2 (Standard International Trade Classification Revision 2): Four-digit product codes, covering earlier time periods File Organization Data files are organized by classification system and time period (e.g., hs92_country_country_product_year_6_1995_1999). Files are stored in Parquet format and split into multiple parts for efficient processing of large datasets. Methodology International trade flows are recorded twice: once by exporters and once by importers. In practice, these records often differ, are incomplete, or are reported using different product classifications. To address these issues, the Growth Lab developed a standardized, data-driven process to produce consistent and comparable bilateral trade estimates. Our methodology combines two core steps: reconciling partner reports and harmonizing product classifications over time. Reconciling Exporter and Importer Reports We first adjust reported values to ensure comparability between exporter-reported (FOB) and importer-reported (CIF) data. We then assess the reliability of each country's reporting based on its consistency across trading partners. Using these reliability scores, we combine exporter and importer reports into a single estimated trade value that places greater weight on more reliable sources. Both the original reported values and the reconciled estimated values are included in the dataset. Related Resources The data and methodology are described in detail in our peer-reviewed publication: Bustos et al. (2026), Tackling Discrepancies in Trade Data: The Harvard Growth Lab International Trade Datasets, Scientific Data. For a visual explanation of the Growth Lab's trade data methodology, please visit our companion website. This bilateral trade dataset forms the core data infrastructure underlying the Atlas of Economic Complexity.