The STOCK Act has a reputation as the law that made congressional trading transparent, and I will admit I used the data it produces for years before I actually read the text. It is short, and once you know what it does and does not require, most of the quirks in political trading data stop looking like accidents. The delays, the vague dollar ranges, the scanned handwriting, the filings that show up months late, all of it traces back to specific choices in the law.
The short version of the history is that Congress passed the Stop Trading on Congressional Knowledge Act in 2012, largely because a television investigation had made trading by members a public embarrassment. The law does two big things. It states explicitly that members of Congress and federal employees owe a duty not to trade on nonpublic information they learn through their jobs, which had been a genuinely murky legal question before. And it forces relatively fast public reporting of securities trades that previously surfaced only in an annual filing, sometimes a year or more after the fact.
Who has to file, and which trades count
The reporting net is wider than most trackers show. Members of Congress file, and so do their senior staff, the President and Vice President, and senior executive branch officials. A later law extended similar transaction reporting to federal judges. In practice nearly all the attention lands on the 535 members, partly because their filings get aggregated into clean public feeds and partly because committee assignments give their trades an obvious information angle. A senator on an armed services committee buying a defense contractor draws a level of scrutiny that a mid-level agency official never will.
Coverage also extends past the filer personally. Trades by a spouse or dependent child are reportable, and historically a meaningful share of the interesting filings have been spouse trades. Each transaction record marks whether the owner was the member, the spouse, or a joint account, and that field deserves more respect than it gets, because a spouse who runs an active trading operation produces a very different kind of signal from a member quietly rebalancing a retirement account.
The trigger is any purchase, sale, or exchange of stocks, bonds, commodity futures, or other securities where the amount involved exceeds $1,000. Two carve-outs shape the resulting data more than anything else. First, diversified mutual funds and similar widely held investment vehicles are exempt from the fast reporting requirement, so a member who holds nothing but index funds barely appears in the transaction feed at all. Assets parked in a qualified blind trust do not generate reports either, though genuinely blind trusts are rarer than the press releases about them suggest. Second, amounts are disclosed in ranges rather than exact figures. The smallest bucket runs from $1,001 to $15,000 and the buckets widen from there, so a filing gives you direction and rough scale, never actual size.
The 45-day window is a ceiling, and plenty of filers treat it that way
Here is the actual rule, because it gets misquoted constantly. A covered person must file a periodic transaction report within 30 days of becoming aware of a transaction, and in no case later than 45 days after the transaction date. Forty-five days is the outer limit, not the expectation, but a lot of filings land near it anyway. Add the time it takes for a report to be processed and posted to the public disclosure systems, and the trade you are reading about can easily be six or seven weeks old before you could possibly act on it.
That delay is the single most important property of the dataset, because by the time a filing becomes public, whatever the member knew when they traded is stale, and every other person watching these feeds sees the disclosure at the same moment you do. The useful questions are whether the market has fully digested a disclosure that just posted, and whether the pattern across many filers tells you something no individual filing does.
The two-date structure also creates the classic backtest mistake in this niche. Every disclosure record carries a transaction date and a filing date. If you build a strategy that enters on the transaction date, you are trading on information that was not public yet, and the backtest will look spectacular for reasons that have nothing to do with anything you could have done. Any rule you test has to key off the filing date, and ideally off the timestamp when the filing actually became retrievable, because those can differ too.
Annual reports, the $200 fine, and what weak enforcement does to the data
The periodic reports sit on top of an older system. Since the late 1970s, senior federal officials have filed annual financial disclosure reports covering assets, liabilities, outside income, and positions held, again in broad ranges. Those annual reports typically arrive in the spring covering the prior calendar year, with extensions available. The STOCK Act left that layer in place and added the fast transaction layer on top, so a serious tracker uses both. Annual reports give you the holdings picture, periodic reports give you the flow.
Then there is enforcement, which is where the law goes soft. The standard penalty for a late periodic transaction report is $200, and the ethics committees can waive even that. Members have filed months late, in some cases more than a year late, and the typical consequence has been the fee and an unflattering news cycle. Prosecutions under the insider trading provisions have been essentially nonexistent. Congress also quietly amended the law in 2013 to kill the planned searchable, downloadable database for staff filings, which is why staff data remains so much harder to work with than member data.
You can be cynical about all this, and I am, somewhat. But if you work with the data, the weak penalty has a specific practical consequence that matters more than the outrage. Cheap noncompliance makes the dataset ragged. Late filings, amendments that correct earlier records weeks after the fact, paper filings digitized badly, tickers entered wrong, options positions described three different ways by three different offices. Any pipeline consuming this data needs to expect corrections landing long after the original record, because the law gives filers no strong reason to get it right the first time.
How I actually read a filing
Roughly this checklist runs in my head whenever a new disclosure crosses the feed.
- Check the gap between transaction date and filing date first. A trade reported within days reads differently from one reported at day 44, and a chronically late filer's reports carry almost no timing information at all.
- Skip fund transactions. The exempt-fund carve-out means single-name stock and options trades are where whatever signal exists actually lives.
- Read the owner field, then ask whether this person trades constantly or almost never. A first purchase in years from a quiet filer is more interesting than the fortieth trade of the month from an active one.
- Map the ticker against committee assignments. A health subcommittee member trading a pharma name deserves a closer look than the same member trading a broad industrial.
- Look for clusters. One member selling a sector is noise, while several members from relevant committees selling the same sector inside a short window is historically where the interesting cases have surfaced.
None of this turns congressional disclosures into a standalone strategy. The reporting lag and the range-based sizing mean you can rarely mirror a specific trade profitably in any mechanical way. Where the data earns its keep is context and prioritization, a reason to look harder at a name, a flag that people with better information than you were moving a particular direction weeks ago. We built the political trading feed in Blockcircle to normalize exactly this mess, filing dates versus trade dates, amendments, owner fields, committee mappings, because the raw filings punish anyone who consumes them casually.
If you want a feel for the raw material, open the House or Senate disclosure portal and read a handful of periodic transaction reports yourself. Ten minutes with the actual documents, the ranges, the vague asset descriptions, the occasional scanned handwriting, will teach you more about what this data can and cannot support than any summary, including this one.