Which politicians are the best stock traders? Why most rankings are wrong

Leaderboards of Congress's best-performing traders circulate constantly and almost all of them are methodologically broken. Here are the six errors that inflate them, and how a defensible ranking works.

2026-08-05 · 12 min read · By Insider Option

"This congressman returned 240% last year" is one of the most reliable pieces of content on financial social media. It is also, in almost every instance, measuring something other than what it claims.

This post is about why these rankings are so consistently wrong, and what a defensible one would require. It is less exciting than a leaderboard, but it will stop you being fooled by one — including, if we get this wrong, by ours.

The core problem: the data cannot support a return figure

A percentage return needs two things: the value of each position, and the total portfolio it sits in. Congressional Periodic Transaction Reports provide neither.

Amounts come as bands — "$100,001–$250,000" — so a single trade's size is uncertain by a factor of 2.5, and top-bracket trades ("over $50,000,000") are unbounded. There is no total portfolio figure anywhere in the filings. And whole asset classes are exempt: mutual funds, ETFs, and Treasuries are not reported the way individual stocks are, so a member's disclosed stock trades may be a small and unrepresentative slice of their actual wealth.

Every leaderboard resolves this by assuming. Most assume the midpoint of each band and treat disclosed trades as the whole portfolio. Both assumptions are wrong in ways that do not average out — and neither is usually disclosed.

The six errors

1. Transaction-date entry

The most consequential error. Computing returns from the date the member traded credits them with up to 45 days of price movement that was not observable to anyone else. It answers "how did the member do", which is a fine question, but it is then presented as if it were a strategy you could have followed.

A ranking meant to inform your decisions must use the disclosure date. The two figures often differ dramatically, and the gap is largest for exactly the trades that make headlines — the well-timed ones.

2. No minimum sample

A member who made three trades, one of which was a small-cap that tripled, will top any unfiltered leaderboard. This is not skill, it is variance, and with hundreds of members trading it is guaranteed to happen every single year to someone.

Any ranking without a minimum trade count — twenty is a reasonable floor, and even that is thin — is measuring luck and reporting it as ability.

3. Wrong benchmark

Congressional disclosed holdings skew toward large-cap technology. Benchmarking that against the S&P 500 during a period when technology outperformed hands you apparent alpha that is entirely a sector bet.

The fix is either a factor-matched benchmark or, at minimum, reporting results against several indices so readers can see how much of the result is composition. A ranking that does not name its benchmark is not a measurement at all.

4. Attribution errors

Three distinct problems, all from the owner and comments fields that most datasets discard:

• Spousal trades file under the member's name. When a member's spouse is a professional investor, ranking the member as a trader is simply mislabelled.
• Managed accounts. Some members disclose that trades were made by an adviser with full discretion, without their knowledge. Including those in a ranking of political trading acumen is incoherent.
• Dependent children's accounts — same issue, smaller scale.

5. Survivorship and selection

Rankings built from "notable trades" are selected on the outcome. So are the screenshots that circulate. A member's record must include every disclosed trade in the period, including the ones that went to zero, or the number is meaningless.

Related: members who leave Congress often vanish from datasets, which systematically removes whoever was there during the periods being averaged.

6. Multiple comparisons

This is the subtle one, and it invalidates more analysis in this field than anything else. Search across ~535 members, several years, multiple holding periods, and multiple benchmarks, and you are running thousands of implicit tests. Some will produce spectacular results by chance. Reporting the maximum of a large number of noisy estimates as a discovery is not a finding — it is the definition of data mining.

The only real defence is out-of-sample testing: identify your top performers in one period, then check whether they persist in a later one they had no part in selecting. In our experience this is where most apparent Congressional trading skill disappears.

What a defensible ranking looks like

If you are evaluating someone's leaderboard, or building one, this is the bar:

1. Disclosure-date entries, with the filing timestamp respected.
2. A stated, appropriate benchmark — ideally several.
3. A minimum trade count, stated, before a member is ranked at all.
4. Explicit handling of spousal, dependent, and managed-account trades — either excluded or reported separately, never silently merged.
5. Confidence intervals that reflect the amount bands. Report the range the disclosures actually permit, not a false point estimate.
6. Complete trade coverage in the period, with parse failures disclosed rather than dropped.
7. Out-of-sample persistence, or an explicit acknowledgement that the ranking is descriptive of the past and not predictive.

The more useful questions

Given that a trustworthy return ranking may simply not be constructible from this data, there are better things to ask of it — questions the filings can answer:

• Who trades in industries their committee oversees? This is answerable from public committee assignments and sector classification, requires no return calculation, and is arguably the question the STOCK Act exists to let you ask.
• Who trades near legislative or briefing events? A timing question, not a performance one, and far more robust.
• Who files late, and how often? Directly measurable from the filing-versus-transaction date gap. A reasonable proxy for how seriously a member takes disclosure.
• Who trades individual stocks at all? Many members hold only funds. That is itself informative about how they approach the conflict.

These are all accountability questions rather than alpha questions, and they are the ones this dataset is genuinely well suited to.

Our own position

We publish performance data on our strategies pages, so we are subject to everything above. Our commitments: entries are dated from disclosure, not transaction; the benchmark is named; and our methodology page documents the assumptions, including where the approach has underperformed.

If any tracker — ours included — shows you a leaderboard without telling you its benchmark, its entry-date convention, and its minimum sample, treat the number as marketing. The broader question of whether any of this is actionable is in does copying Congressional trades actually work?

This post is for informational purposes only and is not investment advice. Past performance does not predict future results. See our disclaimer.