Research Data for Oil-Relevant Presidential Posts and WTI Tail and Jump Activity
Description
Version 3 adds the research data underlying the manuscript “Oil-Relevant Trump Truth Social Posts Mark Immediate Tail and Jump Activity in WTI Futures” to the executable analysis code released in Version 2. Package 01 contains raw and annotated minute-level CME/Databento and Binance market data and the corresponding analysis panels. Package 02 contains the complete public-research-API exports of Donald J. Trump posts from trump.fm, post identifiers, timestamps, public URLs, text hashes, semantic audit records, the independent Kimi comparator audit, and the event crosswalk for the blinded human-reviewed 305-post oil sample. Package 03 contains source data, coefficients, contrasts, robustness outputs, and the 21 July 2026 comparator sensitivity results. Duplicate byte-identical files are stored once and mapped in each archive; all files have SHA-256 checksums. Potential credential-like strings in text files were redacted only from the upload copy. LSEG/Reuters records, headlines, story identifiers, and LSEG/Reuters-derived tables are excluded. The author confirmed permission to upload the included CME/Databento historical records. Third-party market and social-media records remain subject to their source-specific terms and are not relicensed by the repository-level Creative Commons licence.
Files
Steps to reproduce
1. Download all three Version 3 ZIP archives, UPLOAD_PACKAGE_MANIFEST.csv, and RIGHTS_SOURCES_AND_EXCLUSIONS.md. 2. Verify each archive against the SHA-256 value in the upload manifest. 3. Extract each archive and verify individual files against its internal FILE_MANIFEST_SHA256.csv. 4. Read README_PACKAGE.md and DUPLICATE_PATH_CROSSWALK.csv in each archive. 5. Use the raw and annotated minute data and regression panels in Package 01 as the market inputs. 6. Use Package 02 for the complete public-API post archive, provenance, machine audits, Kimi comparator audit, and frozen human-label crosswalk; the main treatment group is identified by analysis_post_group_primary_305. 7. Use Package 03 for source tables, model coefficients, contrasts, robustness outputs, and comparator sensitivity results. 8. Run the executable analysis scripts released in Version 2. LSEG/Reuters-dependent diagnostics remain outside the public reproduction route.