Scanned trades

Some members file their stock-trade reports on paper. The House posts those only as scanned page images: no text, nothing to search or add up. We read every page, so these 43,948 trades exist here as data, and, as far as we know, nowhere else.

Verified, or read by machine

Every trade found on a paper report's pages. Once a human has verified a trade against the scan, it's settled. Until then, the reader's own confidence shows: a trade it read surely counts, one with a minor doubt counts and is marked not yet reviewed, and one it doubted more is held back, uncounted, until a human reviews it.

• Verified by a human
• Read surely
• Some doubt, not yet reviewed
• Held back until reviewed

What we had to estimate

A paper form can leave a box unmarked, and a scan can leave a date or an owner code past reading. Rather than drop the trade, we estimate that one field from the rest of the page or the report, and say so on the trade. 2,699 of the 43,948 trades counted (6.1%) carry an estimate; a trade can carry more than one.

By member: the trades with any field estimated, most first.

7,239 trades (16%) sit on pages printed too small or scanned too poorly for our reader. Those were transcribed by a vision model, twice over and independently, and a trade counts only where the two readings agree.

Whose they are

The members whose paper reports these are, by whether each trade is verified or still read by machine alone.

When they traded

By the trade's date.

What they traded

The assets traded most on paper reports, by how many trades (one asset's spellings joined); hover for the purchases and sales.

• Purchases
• Sales

In whose account

Most of these trades are a spouse's: members must report them all the same. 5 were filed with no date and 92 with no amount ticked.

The largest

The 100 largest trades read from paper, by the top of their band; 97 of them verified against the scan by a human.

To see a report, search efdsearch.house.gov for the member's name and the date it was filed. Something read wrong? Tell us.