What Congress disclosed, and what copying it would have returned

Under the STOCK Act, a member has 30 to 45 days to report a trade. In practice the median gap between the trade and its disclosure is around 26 days, and roughly one filing in ten arrives late. A headline that says a member made 60% on a position is describing a return that started at a price no copier could reach.

That is why the ranking above is a copy-returnrather than the member’s own return, and why the two are different numbers, usually worse, sometimes much worse. It is the only version of the question that means anything to somebody deciding whether to follow along.

A backtest tells you how someone’s past disclosures would have treated you. The half that matters going forward is not missing the next one, and a filing is only useful in the days after it appears, which is exactly when nobody is watching the House Clerk’s website.

So the ranking and the alerts are one product rather than two. Follow any member in Pulse and you get an email as soon as a new filing of theirs is parsed: the trades in it, both dates, and how long they sat on it: “traded Jul 2, disclosed Aug 9, 38 days later”. The same disclosure date the backtest would have bought on is the one that triggers the email, so what you get alerted to is the thing you just tested.

A periodic transaction report gives the member, the ticker, whether it was a purchase or a sale, the trade date, the filing date, and an amount band: “$1,001–$15,000” and so on, up to “over $1M”. The law never requires an exact figure. Nobody can tell you what a member actually put in, and any site that quotes a precise position size is filling in a blank the filing left empty. Trades by spouses and dependent children are covered too, and are included here.

Because sizes are unknowable, the ranking sizes every position the same way, so the comparison between members is about which trades they disclosed, not about a position size invented for them.

Each member’s figure simulates one concrete, followable strategy: your money sits in an S&P 500 fund by default, and every time the member disclosed a buy, a slice of the portfolio moves from the index into that stock at the first opening price after the filing became public. The position is sold when the member discloses selling, or after one year, whichever comes first, and the money goes back into the index.

How big that slice is depends on how hard theywent in. A filing discloses a band, not a number, but a $1–5M purchase and a $1,001 purchase are plainly not the same conviction. So each buy is ranked against that member’s own earlier filings and sized between 1% and 20% of the portfolio, so their biggest-ever bet gets the top of that range, their smallest gets the bottom. Only filings they had already disclosed at the time count, so nothing is sized using information that did not exist yet.

Because idle money holds the index rather than nothing, the “vs SPY” column answers the only question that matters: did adding this member’s picks to an index portfolio beat just holding the index? A ranking that lets cash sit uninvested mostly measures the sitting; this one measures the picks.

The cost of sizing by conviction, stated because it is the kind of thing that gets left out: an account that commits more can run out of money and turn a later disclosure away. Where that happens to more than a tenth of a member’s entries, the count is printed above their figures. What survives a refusal is not a random sample of what they filed. Flat 1% never refused anything, but it also left the account almost entirely in the index, which flattered every “vs SPY” toward zero.

Almost all of it is read straight from the primary source: the House Clerk’s disclosure archive and the Senate’s electronic filing system, refreshed daily, going back to 2012, the year the STOCK Act made these reports a requirement. Electronically filed reports are parsed; the ones filed on paper are scans, and a scan is skipped and counted, coverage here is a number we can show you rather than a silent gap.

The exception, stated because it is the kind of thing sites leave out: a small share of older filings exist only as those paper scans, and for those we fall back to two public volunteer-transcribed datasets. That is roughly 2% of filings, every one of them tagged “mirror” in the feed, and our own read of a document always wins where we have one.

A member filing a purchase tells you what they bought and when it became public; it tells you nothing about why, and nothing about what happens next. Members leave office, stop trading, or file in late batches. The numbers here ignore taxes and any cost beyond the modelled fees.