TiRADE

About

How to use

1. Download the CSV. It is the whole dataset: one row per stated rate, UTF-8, comma-separated, with a header row. It opens directly in Excel, R (read.csv), Python (pandas.read_csv), or Stata (import delimited).
2. One row is one numeric rate. A source that states several rates (“25%; 50%”) appears once per rate: rate_role says how the row relates to its source cell and n_rates_in_source how many rows the cell was split into. Where an action was recorded without a stated number — a framework deal, a threat carrying no rate, a per-item dollar duty — rate_value is blank and rate_role is tbd; every blank rate_value is a tbd row, so filtering them out is one test.
3. Filter to what you need. source_type selects the channel and event_type separates Threatened, Announced, and Implemented. countries and sectors are lists separated by “; ”, so match on the part rather than the whole cell. Rows with rate_role = scheduled are future step-ups; include or exclude them deliberately.
4. Times are US Eastern wall-clock, the convention US markets trade on; datetime_et combines date and time_eastern. For a UTC instant, localise datetime_et to America/New_York. Where the source gives no time — most Federal Register instruments without a public signing event — time_eastern is blank and datetime_et is the bare date; a parser will read that as midnight, which is an artifact, not a claim about the hour. For intraday work, keep only rows with a time_eastern.
5. Verify any row from its excerpt and link. ref_id is the stable key to the source document and is safe to join across versions. event_id groups records of the same action across channels but is re-keyed on every release, so never join across versions on it.
6. For trade exposure, iso_matched and sector_codes_matched give the OECD ICIO cells the row covers, USA_imp_affected_musd the value of US imports in those cells, and stateful_direction whether the action raised or lowered the prevailing rate. Rows with no geography, such as policy-defined groups, are left unmapped; match_status says why.

Key columns

The remaining columns are alternative event keys (event_id_strict, overlap_type, event_multisource_w1), the timezone label tz and the source medium and social-media metadata (medium, platform, deleted, engagement_fav_reblog_reply), and intermediate ICIO fields (idc, idp, match_status, tariff_action, prior_rate, pp_decrease, and the stateful_* cell counts).

Download

Release v1.3 is a fixed snapshot; its file never changes. To cite this exact release, link to its permanent URL.

Citation

@misc{tirade2026,
  author       = {Kalemli-Özcan, Şebnem and Soylu, Can and Yildirim, Muhammed A.},
  title        = {{TiRADE}: A Database of Tariff Rates in Announcements, Declarations, and Events},
  year         = {2026},
  version      = {v1.3},
  howpublished = {Global Linkages Lab, Brown University},
  url          = {https://www.globallinkages.org/datasets/tirade/v1.3},
  note         = {Data cutoff 13 September 2026}
}

Versions

Updated weekly on Mondays: each week's new tariff statements are screened, reviewed, and released as the next version.