Built and run by one person.
What Trump Has Been Buying, Selling and Holding. The Filings, Extracted and Searchable.

What Trump Has Been Buying, Selling and Holding. The Filings, Extracted and Searchable.

Published: August 24, 2026

If you want to see what Trump has been buying, selling and holding, the filings are on my site, extracted and searchable. Trades, assets, income, liabilities, gifts.

Two kinds of filing sit in there. The trade reports come in within 30 to 45 days of each trade and cover securities only, 15 filings and 5,368 trades so far.

The split is not even. For nine months it was mostly muni and corporate bonds, the earlier 14 filings have 1,726 trades between them.

Then one filing on 14 May came in with 3,642 trades, almost all equities. That May one you can go through trade by trade. The earlier 14 are in there as per-filing summaries.

And the annual disclosure, the 927 page one, which covers the whole year, 28,007 records including 21,285 trades.

You can use it two ways. Search and filter in the tool, roll the trades up by company, see them plotted against the S&P 500. Or download the lot as CSV or JSON and run your own cuts.

Free, no sign-up.

tigzig.com ➜ TREMOR ➜ Trump Trades.

Compliance note

Not investment advice. Educational analysis of public documents. Past Trump trades are no guide to future returns. Consult a qualified investment advisor. I hold positions in some of the tickers that show up in the data.

On how the data was pulled

The two filings were not the same job at all.

The trade reports arrive as scans. Seemed like someone ran the paper through an office copier and OCRs it, so you are working off a picture of a page. I ran several extraction methods over it and took the value the methods agreed on, with a fourth independent method redoing a sample of pages, and then a separate checker pass I kept away from. It found things. I fixed them.

The annual is the opposite. It came out as a proper digital document rather than a scan, so extraction is character exact. Two unrelated parsing engines, reconciled row by row, and I only kept the dataset because the two agreed on every field. After that, blind AI vision rereads on sampled pages, with any disagreement settled against a high resolution image of the printed page.

Aggregates were checked against independently reported figures as well, down to the exact dollar on the big income lines.

And even after all the checking, an extraction error is still possible. I cannot rule it out. Every filing on the tool carries a link to the original OGE document, so if a number looks off, you can go and read it against the source.

A few caveats