Donald J. Trump

Read by Open Cabinet, not by us. The filings of the President of the United States are scanned images, and we read nothing out of them ourselves; every line below is read out of one of them by a third party. The dataset page says what we measured.

These transactions were read by someone else. The filings behind them are scanned images: the U.S. Office of Government Ethics publishes the filings of the President of the United States as pictures of paper, and we measured that there is no text layer in them worth reading — not even the one the source lays under its own scans. The lines below come from Open Cabinet, which reads those documents and publishes the result as a downloadable dataset. Every band and every estimate here is theirs, and every line links to the document they read it from. The two return figures are ours, not theirs — we compute them on our own price cache — but the ticker each one rests on is usually their inference from the description, at their top confidence grade and only where they call the instrument common stock; on 1 of these lines it is ours instead, read from the same description where they left none. Their most recent export is dated October 29, 2026; we show that date and not the day we downloaded it, because a fresh download of an old export is not fresh data. We check each amount band against the eleven bands the form itself carries and drop any line that does not land on one exactly (our reading of their file, version 1.3.0). We have not verified any line against the document. That check would mean reading the scans ourselves, and the measurement that started all of this is exactly that we cannot.

Taken by date filed. These lines come from the filing dated October 8, 2026, over 1 filing day, and cover trades of Aug 31, 2026 to Aug 31, 2026.

No ticker column, on purpose. The form has no ticker column, so every ticker in this reading is something the reader inferred from the description. We remeasured that on the export of September 11, 2026: of the President’s 8,940 lines, 10 carry a ticker in the column we read, and 6,471 carry one in a separate column of theirs with their own confidence grade next to it. Since September 13, 2026 we use that second column — top grade only, and only where they call the instrument common stock — as the key to a price series, and that is what the two return figures rest on. It is left off every other line: 2,044 of those 8,940 lines are municipal bonds, corporate notes, preferred shares or ETFs rather than common stock, and a further 486 are shares they could not grade. Since September 13, 2026 some of those 486 carry a ticker we established ourselves, from the description, with the evidence recorded per description — the next paragraph says how many of the tickers on screen are theirs and how many are ours.

Two returns per line, and the second one is the point. Half of these lines were filed more than 65 days after the trade (minimum 16, maximum 301). The first figure is therefore a description of what the share did, not a return anyone outside the filer could have taken: on the trade date nobody else knew. That is the whole reason the second figure stands next to it. Both figures run to the same day — the most recent close in our price cache. Each figure carries its own sign, and the color only repeats that sign: this is a description of what the share did, not a verdict.

4 of 5 lines carry a return. The other 1 does not, and here is why: 1 — the source did not confirm a ticker for this line at its top grade, and we have not established one ourselves. A share price on a municipal bond would be a wrong figure rather than a missing one, so those lines are counted here instead of left blank. Of the tickers behind those figures, 3 are Open Cabinet’s and 1 is ours, read from the description with the evidence recorded per description.