Congressional Trading: Signal vs Noise
Members of Congress may trade with an information edge. The disclosure you can read is a degraded copy of it — and by the time it prints, it is mostly measuring its own audience.
The naive version of the congressional trading thesis is that legislators trade on privileged knowledge and copying them captures it. The sophisticated investor has usually moved past that — and landed on a subtler error: believing that even if the edge is real, the disclosure feed is a usable proxy for it. It is not, and the reasons are structural. A member of Congress may well trade with an information advantage. The filing you can read is a degraded copy of that trade — degraded on three separate dimensions at once — and the degradation is severe enough that the feed's observable market impact has more to do with who is watching it than with what it contains.
Here is the framing we use. A disclosure is not a position; it is a report about a position, filtered through time, size-bucketing, and hands. Time: the governing disclosure regime gives members up to forty-five days to report a transaction, so the trade you are reading may be weeks old — an eternity relative to the half-life of most legislative information. Size: amounts are reported in broad ranges rather than exact figures, so a token purchase and a conviction-weighted one can print identically. Hands: the report does not reliably distinguish a trade the member chose from a trade a spouse, a dependent, or a third-party manager executed in an account the member may not even direct. Each filter destroys information on its own. Compounded, they mean the disclosure feed preserves the existence of trades while shredding almost everything that made them informative — timing, conviction, and intent.
What the filing actually tells you
Work through what survives the filters, because something does. The filing tells you that a household connected to legislative power transacted in a name, in a size band, within a trailing window. It does not tell you why, it does not tell you the entry price relative to the price you would pay on the disclosure date, and it will not tell you about the exit until another filing, equally lagged, eventually prints. For a portfolio owner the immediate consequence is that the feed cannot support the use it is most commonly put to — trade replication. Replicating a position weeks after it was established, at a price that has already absorbed whatever short-lived information motivated it, is not copying the trade; it is buying the residue of the trade. And because the disclosure regime reports entries more visibly than the reasoning and exits around them, a replication strategy systematically holds positions its supposed source may have already abandoned.
The deeper problem is selection. Most reported transactions are the ordinary financial life of wealthy households: diversification, managed-account rebalancing, tax management, a financial adviser running a model portfolio that happens to include a client with a Senate seat. Legislative information, when it exists, is a thin layer on top of a thick base of noise — and the filing format gives you almost no purchase for separating the layers. Any framework that treats the feed as homogeneous is therefore averaging a possible signal with a dominant noise floor, which is how the aggregate feed can look impressive in anecdote and unremarkable in aggregate.
The discretion-jurisdiction-urgency filter
The separation problem is not hopeless; it is just that the informative subset is small and has a recognizable shape. We screen congressional disclosures through three questions, and a filing has to pass all three before it earns a place in the information column rather than the attention column.
Discretion: did the member plausibly place this trade? A transaction in a self-directed account, inconsistent with the account's prior pattern, is a decision. A transaction inside a managed account that trades dozens of names on a schedule is an adviser's model, and the member's job is incidental to it. Jurisdiction: does the member's committee assignment touch the name's regulatory or fiscal fate? Legislative information is narrow — it concerns bills, appropriations, investigations, and briefings that flow through specific committees. A trade adjacent to that flow is categorically different from a trade in a name the member has no institutional window into. Urgency: does the structure of the trade imply the information is perishable? A first-time position, unusual size for that filer, a departure from the household's diversified pattern, or an options structure with an expiry — these are the signatures of someone acting on something time-bound. A recurring purchase of a broad index vehicle is the signature of someone saving for retirement.
Notice what the filter implies about base rates: the overwhelming majority of filings fail at least one leg, and the feed's few genuinely interesting prints are precisely the ones the forty-five-day lag damages most, because urgency and staleness are enemies. The filter does not manufacture an edge. It tells you where the edge could live, and it tells you that most of what the dashboards surface is not it.
The reflexive turn: when the audience becomes the signal
The most important development in this data category is not regulatory; it is behavioral, and it changed what the feed measures. Once congressional disclosures were aggregated into dashboards, tracker accounts, and copy-trading products with large retail followings, the disclosure event itself became a demand event. A prominent member's reported purchase can now move a mid-cap name on the day it prints — not because the market learned anything about the company, but because a synchronized cohort of followers buys the disclosure. This is reflexivity in its cleanest form: the signal's popularity generates the price action that appears to validate the signal.
For a serious investor this inverts the object of study. The tradable content of a widely-followed congressional disclosure is mostly the crowd it summons — a short-horizon attention flow with no fundamental anchor, which builds and decays on the cadence of social amplification rather than the cadence of the underlying business. That has two practical consequences. First, entering on the disclosure means paying an attention premium layered on top of a stale entry price: two costs, stacked. Second, the feed becomes genuinely useful in a different column of the research process — as a crowding and attention input. Knowing that a name has just been injected into the copy-trading complex tells you something real about who its marginal holder is about to become and how fragile that holding cohort is. That is worth knowing. It is simply not the thing the dashboard claims to be selling.
The strongest case against this framework
The serious objection cites the research literature: studies of congressional trading have, in some periods and samples, found returns consistent with an information advantage — particularly in earlier eras with laxer disclosure. If the members have edge, the objection runs, then even a degraded copy of their trades should retain some of it, and dismissing the feed throws away a positive-expectancy signal because it is inconvenient to extract.
We accept the premise and reject the inference, on two grounds. First, the empirical record is genuinely mixed: results are sensitive to the sample period, the weighting scheme, and which chamber and era is studied, and later work using stricter methods has tended to find weaker effects than the headline early studies. A finding that fragile should be held with calibrated confidence, not treated as an established anomaly. Second — and this is the decisive point — every study measuring an advantage measures it from the trade date. The copyable residue begins at the disclosure date, which arrives up to forty-five days later, after the perishable component has largely expired and, in the modern era, after the attention flow has repriced the entry. An edge at the trade date is fully compatible with nothing at the disclosure date. The objection, examined closely, is an argument about whether members should be allowed to trade — a governance question we take seriously — rather than an argument that the public feed is investable.
There is a second, quieter objection worth naming: that the filter framework is itself becoming public knowledge, and committee-adjacent, self-directed, urgent-looking trades will attract the most copying and therefore the most attention distortion. This is correct, and it is self-limiting in the usual way — the more reliably a disclosure subset is copied, the more of its return is pulled forward into the disclosure-day repricing, and the less remains for the copier. Signals embedded in public filings decay toward attention events. That is the general law this piece is one instance of.
How to apply this framework
- Reclassify the feed before you read it. Route congressional disclosures into two columns: the rare filings that pass discretion, jurisdiction, and urgency, and everything else. The first column is a weak, lagged information input worth a research note. The second is an attention input, and reading it as information is the category error this framework exists to prevent.
- Price the two lags separately. Before acting on any filing, ask what has happened to the name since the trade date (the information lag) and since the disclosure date (the attention repricing). If the answer to either is a material move, the filing's content is already in the price you would pay.
- Use the feed as a crowding sensor for names you already own. A widely-amplified disclosure in one of your holdings tells you a retail attention cohort is arriving — a change in the marginal holder and in short-horizon volatility, worth folding into position sizing even though it says nothing about the business.
- Watch absence as well as presence. The disclosure regime shows entries loudly and exits late. Never infer a standing endorsement from an old filing: the position may be gone, and the follow-up report confirming that will arrive on the same forty-five-day delay the entry did.