What we measured
We set out to answer three questions. How promptly do members of the House report their trades, measured against the deadline in the law? Has that changed since 2014? And what does published research say about whether members’ trades beat the market?
The first two come from our own analysis of the House’s records. For the third we report other researchers’ findings. We did not compute returns ourselves (see “What we didn’t do,” below).
The records
Members of the House report trades on periodic transaction reports, which the Clerk of the House publishes at disclosures-clerk.house.gov. On November 4, 2026, we downloaded the Clerk’s yearly indexes of filings for 2014 to 2026, and every report they list, 8,431 in all. The 2012 index lists no periodic transaction reports and the 2013 index eight, so our series starts in 2014.
- Electronic reports (5,979) carry a text layer. We read every trade from them, one row per trade.
- Scanned paper reports (2,452) are images with no text layer. We list them and count their pages, but we don’t read them, so their trades are not in our figures.
To check our reading, we counted each report’s trades a second way, from a different text extraction, and compared the counts. They matched on every electronic report. The signing date printed on each electronic report also matched the Clerk’s filing date on every one. We also compared the dates on 14 reports, ten of them chosen at random from 2014 to 2026, with the PDFs, and every one matched.
Older forms differ from today’s. Their text turns some capital letters into lower case, some of their dates lack leading zeros, and they carried no asset codes until 2018. Our scripts handle each difference, and report any line they can’t place rather than guessing at it. There were 13 such lines in all, and the scripts’ reports list each one.
When a report is late
The law gives members 30 days from learning of a trade to report it, and in no case more than 45 days from the trade itself (5 U.S.C. 13105(l)). So for each trade:
- The deadline is the earlier of the notification date plus 30 days and the trade date plus 45 days.
- The filing date is the date in the Clerk’s index.
- A trade is late if the report was filed after the deadline. We don’t move deadlines that fall on weekends, because electronic reports can be filed on any day.
- ”More than 30 days late” is the point at which the $200 late fee applies (5 U.S.C. 13106(d)). The House Ethics Committee’s instruction guide gives the same reading in a worked example.
We leave some lines out of the lateness figures:
- Amended and deleted lines. An amendment’s filing date says nothing about when the trade was first reported.
- Trades made before the member took office.
- One line whose status couldn’t be read.
Where a member gives a notification date earlier than the trade date, which is almost certainly a typing error, the 45-day limit decides.
The lead’s figures cover the reports in the Clerk’s 2025 and 2026 indexes, filed from January 1, 2025, to November 3, 2026. The figures by year count reports by the calendar year they were filed, so a report filed in January 2026 counts in 2026 whichever index lists it.
Who filed
Each filer in the Clerk’s index is matched to a member in the public-domain congress-legislators dataset by state, district, term dates and surname. That gives each member a stable ID and the name they use publicly, rather than the name as filed (“Scott Scott Franklin” in the index is Scott Franklin). Names of current members follow the Clerk’s member list. Every report matched except two filed by candidates, not members.
What kind of asset
For the counts by asset type, we classify each trade as a stock, an option, a fund or ETF, a bond or government security, or other.
- From 2018: we use the asset code on the form.
- Before 2018: we use the ticker and the words in the asset name.
The two methods aren’t strictly comparable. From 2018, for example, many filers code exchange-traded funds as stocks. So the story makes no claim that depends on how the mix of assets changed.
The bills and the research
For the bills, we used the Library of Congress records of actions and cosponsors, and the story links to the same records on Congress.gov. The Senate roll call on H.R. 7008 comes from senate.gov.
The research on returns is reported from the papers themselves: each figure comes with the table or page it is taken from, in studies.csv and chen_sacerdote_table1.csv. Our charts redraw numbers printed in the papers. Two of the papers are working papers that have not been peer reviewed, and the story says which.
What we didn’t do
- We didn’t estimate members’ returns. That takes daily stock prices for thousands of companies over more than a decade. The price sources we found either limit their free data to personal use or charge for a license to publish results built from it.
- We didn’t read the scanned paper reports. Some of the most active traders file on paper, so any count based on electronic reports, including ours, leaves them out.
Limits
- Notification dates come from the members. A wrong date can make a trade look late, or on time, when it isn’t.
- Values are reported in bands, such as “$15,001 - $50,000,” not amounts.
- We can’t tell whether a late fee was paid or waived. The House doesn’t publish either.
- Some forms contain obvious errors, such as trade dates years after the filing. We keep them as filed and leave the analysis to the deadline rule.
Rerunning our work
The scripts in code.zip download the records, read them and produce every file here. readme.md lists every column and gives the steps. The last script checks that the files reproduce the figures in the story, and stops if they don’t.
The code is under the MIT license. Our data files are under Creative Commons Attribution 4.0, so you’re free to reuse them with credit to The Countersign. The reports themselves are public records.
If you find an error, write to corrections@thecountersign.org.
Downloads
| File | What it is | Size |
|---|---|---|
| filings.csv | Every periodic transaction report in the House Clerk’s 2014 to 2026 indexes: 8,431 rows, one per report, with a link to the original. | 1.1 MB |
| transactions.csv | Every trade read from an electronic report, 2014 to 2026: 57,657 rows, with its deadline and whether it was late. | 12.3 MB |
| late_by_member_2025_2026.csv | Late trades by member on reports in the 2025 and 2026 indexes: the figures behind the story’s lead. | 4 KB |
| by_year.csv | Reports, trades and late shares by year filed, 2014 to 2026. | 1 KB |
| members_2014_2026.csv | Reports, trades and late trades by member, 2014 to 2026. | 21 KB |
| studies.csv | The studies of members’ returns described in the story, with the page each figure comes from. | 2 KB |
| chen_sacerdote_table1.csv | The one-year results from Chen and Sacerdote’s Table 1, as charted in the story. | 1 KB |
| readme.md | What each file and column means, how we built them, and how to rerun everything. | 7 KB |
| data.zip | All of the data files above and the readme, in one download. | 1.5 MB |
| code.zip | The Python scripts that download the reports, read them and produce every figure, with the readme and licences. | 42 KB |
SHA-256 checksums
To confirm a download is the file we published, compare its SHA-256 checksum with this list.
ebe12d63449faf0d171595dc083f6b84f06e147757573cb6d731fc0925e0c0e8filings.csv03ede8bbf34f764d592eb61bc67adb16bf26b96bbcccfa98a93c674da3f83dfdtransactions.csv3e9d114f57e368331d78b61f07851aa873ec83ae06abe7b09940baa3ac00bf14late_by_member_2025_2026.csvb47798bf126c719f2b2299b21739534bd9f86cf0b77239b1f4d65bd7353e6969by_year.csv21b9889cec30873cdd3e8568f8b737a39895ab4068eb431200d360ece3113ffcmembers_2014_2026.csvbd180d1d7d94f83a5b5c71143881d8b0b6fd838d9c0d773c6845741cc5fb7fb5studies.csv9754b61c48600048e07933e9418a9e2313e62c613622547d13fa19418cbd057cchen_sacerdote_table1.csv3aad0d02495c3162b6d421224bc9d3f053be556264bf00c20b769c6efab7d368readme.mdc0054e2f92de8841468040879d8a314fa8fd7d1f7bdb2dbe7c4e2295b2c17ccadata.zipf58edecf87de85e95347e2f737df502c88d715717e55d60064c86cc6d0304d82code.zip