Every few weeks a screenshot goes viral: a member of Congress bought a stock, the stock went up a lot, and the implication is that you could have ridden along. The screenshot is usually accurate. What it leaves out is when you could have known — and that single detail is the difference between a strategy and a story.
We pulled every congressional stock disclosure we could get and kept the ones carrying both a trade date and a filing date: 40,082 transactions from 5,601 filings by 318 members, going back to 2012. (That's the machine-readable slice — about a quarter of all congressional filings are scanned images nobody can parse. The 9,877 filings quoted there is every document we've ingested, most of which disclose no tradeable transaction at all; 5,601 is the subset carrying one with both dates on it.) Then we measured the only number that matters for anyone trying to follow along — the gap between the day a trade happened and the day it became public.
The rule, and how often it holds
Under the STOCK Act, a member has to report a transaction within 30 days of learning about it, and in no case later than 45 days after it happened. So 45 days is the outer bound. Here's the reality:
| days from trade to disclosure | |
|---|---|
| Median | 28 |
| 75th percentile | 41 |
| 90th percentile | 302 |
Half of all trades surface within a month, which is roughly what the law intends. The 90th percentile is where it falls apart: 302 days.
That is not a rounding artifact. Measured against the statute's 45-day ceiling:
- 19.8% of transactions — about one in five — were reported more than 45 days after they happened.
- 8.0% were reported more than a year later.
The penalty for filing late is a $200 fee.
Why "the average is 28 days" is the wrong way to read this
A median of 28 days sounds manageable. It's also misleading, because the distribution has a long right tail and the tail is not random — it clusters in bulk catch-up filings. A member who hasn't filed in a while submits one report covering many months of activity at once, and every trade in it lands on the same day, some of them very stale.
56 separate filings disclose at least one trade from more than 600 days earlier. Those aren't parsing errors — we checked them against the source documents.
The record is genuinely hard to believe. A filing submitted in June 2025 disclosed 236 transactions, the oldest dated 8 May 2015 — a gap of 3,698 days. Just over ten years.
It isn't alone. Twenty-one filings carry a gap longer than 880 days, and one submitted in August 2026 reported 165 transactions reaching back to March 2024. Everything in documents like these was, on the day it became public, information about a portfolio that had long since moved on.
For anyone copying trades, the tail is the whole problem. You don't get to skip the stale filings, because you don't know a filing is stale until you read it — and by then you've already learned about a position the member may have opened, ridden, and closed while you had no idea it existed.
What the lag does to a copied trade
This is why every backtest on this site starts from the disclosure date, not the trade date.
It's the less flattering way to run the numbers and it's the only honest one. When a member buys at $100 and the stock is at $128 by the time the filing is public, a "+60% trade" is a headline about a price you were never offered. Your entry is $128. Your return is whatever happens after that.
The gap cuts both ways, which is the part people miss:
- On a winner, you buy in after the initial move. You get the remainder, not the whole thing.
- On a loser, you buy in after the initial drop — which sometimes means you get a better entry than the member did.
- On a position they've already sold, you're buying something they no longer own, and you won't find out for another month.
That third case is not hypothetical. With a median 28-day lag on both the buy and the sell, a member can complete a full round trip before either side of it is public.
Is it getting better?
Disclosure volume has grown — 5,401 transactions dated 2025 against 2,731 in 2015 — but volume isn't compliance. The 45-day breach rate has stayed stubbornly close to one in five across the whole period. Reporting is a manual process built on PDFs, handled by staff, with a $200 penalty attached. Nothing in that setup produces urgency.
So is following Congress useless?
Not useless — just much less magical than the screenshots suggest, and only worth doing if you measure it properly. The lag doesn't destroy the signal; it changes what the signal is. You are not front-running anyone. You are acting on public information about what a specific person, who may or may not know something, did about a month ago.
Whether that's worth anything is an empirical question, and it has an answer. We ran the copy-trade simulation on every member's disclosed buys, entering at the first available price after each filing went public — what copying Congress actually returned.
The fine print
Disclosure amounts are reported as bands ($1,001 – $15,000), never exact sizes, so any position-size assumption is an assumption. Filings are PDFs of varying quality, and some are handwritten scans; we parse what's machine-readable and count what we couldn't. Trade and filing dates are as filed — including the occasional obvious typo, which we exclude rather than trust. This is educational analysis, not investment advice.