Playbook
How to track congressional stock trades (STOCK Act)
“Congress is buying” travels fast on financial Twitter. The data is real, the disclosures are public domain, and the mechanism is credible — but the framing is usually wrong. The average member of Congress does not beat the market. The lag between the trade and the filing is 30 to 45 days. The dollar amounts are buckets, not precision. Used carefully, the STOCK Act feed is a legitimate input to research. Used carelessly, it is noise dressed up as alpha.
The short version
- The average member of Congress does not beat a broad index (NBER w26975). Do not treat “Congress” as a monolith with an edge.
- The conditional exception is leadership and members on powerful, relevant committees whose jurisdiction overlaps with the sector being traded.
- The disclosure lag is 30–45 days — you are following a trade weeks after the fact, not front-running it. Anchor recency analysis on the publication date, not the trade date.
- Reported values are bucket midpoints, not exact dollars. Use them only to rank relative size.
What the evidence says
The claim that members of Congress systematically outperform the market due to privileged information is intuitive and commercially appealing — but the peer-reviewed evidence does not support it as a blanket statement. NBER working paper w26975 established the base rate clearly: the average member shows approximately zero outperformance over a broad index. This is the number to hold in mind before reading any headline about “Congress buys X.”
The mechanism behind the conditional exception is specific: information asymmetry from committee work, not broad foreknowledge. A member sitting on a committee with jurisdiction over defense procurement, healthcare reimbursement, or technology regulation may encounter material non-public developments through their committee role before those developments are public. That is the hypothesis — and even then the documented outperformance is concentrated in the subset where committee jurisdiction and the stock sector directly overlap.
The upshot: condition on power and relevance before reading anything into a trade. A backbench member buying a consumer staple tells you almost nothing. A senior member of a powerful industry-adjacent committee buying into that same sector is a different data point — though still not a guaranteed edge, and still 30–45 days stale by the time you see it.
How to read the disclosures honestly
Condition on power first
The single most useful filter on this dataset is committee membership. Not all disclosures carry equal weight. Before treating a trade as signal, check whether the member serves on a committee with jurisdiction over the industry in question. A trade in a sector with no obvious overlap to the member’s committee assignments has no particular information advantage behind it and should be weighted accordingly.
Leadership positions amplify this further. Members in leadership roles encounter broader legislative intelligence than rank-and-file counterparts. The practical heuristic: weight trades from leadership and relevant committee members; discount everything else toward the no-edge baseline.
Anchor on the publication date, not the trade date
The STOCK Act requires disclosure within 45 days of the transaction date. That window can run anywhere from a few days to the full 45. When you see a filing, the trade may be weeks old — but the publication is today’s event. The publication date is the moment when the information becomes materially public, and that is the date you should anchor recency analysis on. Looking at trades sorted by publication date gives you the current news flow. Looking at trades sorted by trade date tells you about a historical pattern but mixes what’s fresh with what’s been public for a month.
This is follow-the-herd-after-the-fact, not front-running. By definition you cannot act before the trade is disclosed. Any framing that implies you are acting on the same information the lawmaker had — or at the same time — is wrong.
Treat reported value as a bucket midpoint
The STOCK Act does not require exact dollar amounts. Members report in bands: $1,001–$15,000; $15,001–$50,000; $50,001–$100,000; and so on. The standard analytical practice is to use the midpoint of the band as a proxy value — roughly $8,000 for the first band, $32,500 for the second. This is useful for ranking the relative size of trades, but the margin of error around any individual trade is wide. Never quote a midpoint-derived figure as a precise dollar amount, and never use it to compute exact notional positions.
Treat value as an ordinal rank signal: this member committed notably more than average; this one bought at minimal scale. That is the limit of the precision the data supports.
Mind asset class and foreign tickers
Disclosures cover more than equities. The data includes municipal bonds, private fund interests, corporate bonds, and other instruments. If your interest is purely in tradable US equities, filter to asset_class = equity to avoid treating a muni bond purchase or a private fund subscription as a stock signal. Foreign-listed securities may carry a ticker in the disclosure that has no canonical US-listed equivalent — those rows are best treated as informational context rather than directly actionable.
Each member in the data carries a stable Bioguide ID, the official congressional identifier, which allows clean longitudinal tracking of a member’s full trading history even across seat changes.
How to track it with an agent
Once your agent is connected to ClawTerminal, the congressional trades data is directly queryable without writing SQL. The four tools you need map directly onto the filters above:
| Step | Tool | What it does |
|---|---|---|
| Recent disclosure flow | recent_congress_trades | Trades published recently; filter by ticker, issuer name, or asset class; anchors on publication date |
| Top buyers by sector | top_congressional_buys | Aggregate buying across members; asset_class filter keeps it to equities; sortable by notional midpoint |
| Top sellers by sector | top_congressional_sells | Same as above for sell-side; note the sell signal is weaker than the buy signal |
| Per-member history | track_lawmaker | Full trade history for one member, enriched with asset class and ticker context |
| Member profile | congress_member_profile | Lifetime trade summary per member: top tickers, asset-class breakdown, total disclosed activity; uses Bioguide ID |
A plain-English prompt you can paste to a connected agent:
“Show me the most recently published congressional equity trades in the technology and healthcare sectors. Filter to leadership and senior committee members where possible. For any name appearing in more than one trade, pull the member’s full profile and flag how their committee assignments relate to the sector they are trading. Summarize the notional midpoints for relative sizing but do not quote them as precise figures.”
You can also set a standing watch on a specific lawmaker or ticker using ClawTerminal Watchlists. A watch fires when a new STOCK Act disclosure touches the name you care about — you get the notification on publication date, which is the earliest actionable moment in this dataset.
Congress trades are a corroborating signal, not a primary thesis. Pair them with the corporate insider view and the 13F institutional picture for a more complete read. A lawmaker buying a name that also has cluster insider buying and rising institutional accumulation is a more interesting data point than any one signal in isolation.
What not to conclude
- Do not assume front-running. The trade happened 30–45 days before you see the disclosure. The information advantage, if any existed, has largely decayed by publication date.
- Do not treat the average member as having an edge. The base rate from NBER w26975 is approximately zero for the typical member. A headline that says “Congress is buying X” without conditioning on who is buying means very little.
- Do not quote dollar amounts as precise figures. The $8,000 or $32,500 midpoint is a bucket approximation. The actual trade could have been anywhere within the disclosed range.
- Do not extrapolate from a single member’s run. One member trading one sector well over one quarter is anecdote. It is not proof of a repeatable edge. Sample sizes matter; verify with the full history before drawing conclusions.
- Do not conflate asset classes. A muni bond purchase is not an equity bet. Filter by
asset_classbefore treating any disclosure as a stock signal.
The STOCK Act dataset is one of the more unusual inputs in public markets research — direct visibility into disclosed transactions by the people who write the rules. That is genuinely interesting and worth monitoring. The discipline is in the framing: condition on power, anchor on publication date, rank by buckets, and validate with sample sizes. Applied that way, it is a legitimate corroborating signal. Applied as a headline, it is mostly noise.