Do congressional stock trades actually beat the market?
Raw, they look excellent. Measured properly, they don't.
That is the entire finding, and both halves are true at once — which is why this question produces such wildly different headlines depending on who is answering it and what they are selling.
Here are the numbers from our own dataset: 4,791 disclosed buys by the most active traders in Congress, each scored at three checkpoints after its purchase date.
The impressive version
| After | Average return | Win rate |
|---|---|---|
| 1 week | +0.29% | 54.9% |
| 1 month | +1.60% | 56.0% |
| 3 months | +5.15% | 59.3% |
Up 5.15% in three months, winning nearly 60% of the time. Annualise that naively and you have a headline. This is the number most coverage of congressional trading is built on, and it is not wrong — it is just not an answer to the question people think it answers.
The version that accounts for what they bought
Every one of those buys sat in a sector. Sectors move. A stock that rose 8% while its own sector rose 9% did not do well, and crediting the buyer with +8% describes the market, not the decision.
So we subtract it: for each buy, the same window's move in that stock's own sector benchmark ETF. Same trade, same dates, one fewer confound.
| After | vs. sector benchmark | Win rate vs. sector |
|---|---|---|
| 1 week | −0.16% | 47.5% |
| 1 month | −0.58% | 45.7% |
| 3 months | −0.32% | 45.8% |
The +5.15% becomes −0.32%. The 59.3% win rate becomes 45.8% — worse than a coin flip against the sector they bought into.
Nothing was cherry-picked between those two tables. They are the same 4,791 trades, measured over the same windows. The only difference is whether the sector's own movement is credited to the politician.
Why the gap is so large
Because these are overwhelmingly purchases in sectors that went up.
The dataset spans a period of broad equity gains, concentrated in technology. Technology is the single largest sector in the trades here. A portfolio of large-cap US equities bought at more or less any point in that window shows a strong raw return, and so does a portfolio assembled by anyone else — including one assembled at random.
That is what "beating the market" is supposed to isolate, and it is exactly what the raw figure fails to isolate.
The part that surprised us
If congressional trading carried an information advantage, you would expect it to be concentrated where the access is: trades in sectors the member's own committee oversees.
We flag those. Notable means the stock's sector falls under a committee the member sits on. Cluster is stronger — colleagues on the same committee trading the same stock within 30 days. Then everything else.
| Signal | Trades | Raw return (3M) | vs. sector |
|---|---|---|---|
| Cluster | 129 | +4.28% | −1.56% |
| Notable | 693 | +2.52% | −2.82% |
| Aligned (Mega-cap) | 476 | +5.46% | −0.09% |
| Unaligned | 3,376 | +5.68% | +0.21% |
The committee-aligned trades performed worst. Notable buys trail their sector by 2.82%; ordinary unflagged buys beat theirs by 0.21%. A spread of 3.02 percentage points, in the opposite direction to the theory.
This is the finding that cuts against the premise of this site. We built the committee-overlap signal expecting to measure an edge, and measured a deficit. It leads because it is what the data says, not because it is the better story.
What this does and does not prove
It does not prove nobody trades on inside information. A single well-informed trade could be enormously profitable and vanish inside an average of 4,791. Averages describe populations, not individuals.
It does not clear anyone. Poor average performance is not evidence of propriety, and the conflict-of-interest question is untouched by it — a member voting on defence appropriations while holding defence stock has a conflict whether or not the position made money.
It does undermine the copy-trading pitch. If the aggregate has no edge against sector benchmarks, a product built on mechanically following disclosures has no edge to harvest either — before fees, and before the 30–45 day reporting lag.
Caveats we would want applied to anyone else's numbers
- Coverage. This is the most active disclosed traders, not all 535 members.
- Sector assignment is a judgement. A conglomerate genuinely sits in several. Every stock's assigned sector is published.
- Benchmarks are one ETF per sector. A reasonable alternative choice would move the figures somewhat, though not by three percentage points.
- Amounts are bands, prices are closes. Neither is disclosed exactly. Why.
Check it yourself
Every figure above comes from two CSV files we publish in full: the trades and the scored outcomes, with the per-trade returns and benchmark comparisons included. If you think the benchmarking is wrong, recompute it. Download the data · Full method · The live comparison