We Backtested Every Member of Congress. Exactly One Beat the S&P 500

Congress trading is now its own cottage industry. Copy-trading accounts, newsletters, even ETFs, and the premise behind all of them is that members of Congress know something. Maybe they do. The question nobody seems to run properly is whether you could have made money from it, buying when the filing became public, which is the only moment you could have acted.

So we ran it on all of them.

146 members. 22,001 disclosed buys. Entry at the first opening price after each filing went public. Same rules for everyone, no cherry-picking, no hindsight about who turned out to be interesting.

The result

Read on its own, that table says a bit over a third of Congress beat the market. That reading is wrong, and the reason is the row the table doesn't have.

The number that changes the answer

The simulation puts 1.0% of the account into each disclosed buy and holds the idle remainder in SPY, so nobody is penalised for sitting in cash. The consequence is easy to miss: a member who trades rarely never gets much of your money into their picks. The rest of the account is just an index fund.

The median member on this board had 3.9% of the account in their own positions.

So the beat count is mostly measuring how little was at stake:

82 of the 146 ranked members (56.2% of the board) held under 5% of the account in their own picks. Those portfolios were index funds with a tilt, they finished within half a point of the index either way, and they supply 36 of the 56 "wins".

Take them out and the pattern reverses cleanly. Of the 26 members whose copied book was ever meaningfully invested (15% of the account or more) 4 beat the S&P 500. The median one trailed it by 13.3 points.

That is the finding: the more of your money a member's disclosures would actually have put to work, the worse you did.

The exception, which is real

This is not "Congress can't pick stocks". A handful clearly could, and they did it with real money on the table, the top of the board is not made of index-huggers:

The best of them beat the index by +67.4 points. The problem is not that outperformance is absent. It is that picking the right member in advance is the entire game, and the board offers no evidence that this year's winner is next year's.

The single most famous name in congressional trading is not among them. More on that below.

Is this just a data artifact?

Fair question, and we asked it first, because a previous version of this leaderboard was an artifact and we had to throw it away.

Some members disclose so many trades at once that a simulated account can't fund all of them, and the ones it skips are dropped. If the skipping is uneven, the ranking measures truncation rather than skill. So we re-ran the whole thing restricted to members whose simulation actually filled at least 80% of their disclosed buys:

Near-identical, and if anything the well-covered group looks slightly worse. The conclusion survives the filter, which is what makes it worth publishing.

For the record, this run priced 90.0% of all disclosed buys (19,809 of 22,001). The rest are tickers with no available price history, mostly delisted names and one-off obscurities.

Why 146 and not 303

Congress has far more members than this, and 303 appear in our filings. Three rules cut it down, all deliberately:

• At least three disclosed buys. 114 members fall out here. Ranking someone on one or two trades is ranking a coin flip.
• At least three years of history. Another 34 fall out. This one we added after an earlier version of this article, because the board was publishing a member who had one day of history, eight buys that became public the day before the board was computed, showing +0.03% against the index. Enough trades in no time at all is not a track record.
• At least three positions the simulation could actually fill. Another 8. Disclosed buys and copied positions are not the same number: a member can disclose five and have one priced, and one fill is not a strategy.

Three years was picked by measuring what each threshold costs, not by taste: one year admitted members whose entire record was a few weeks, and five removed a further chunk of the field without changing which members came out on top. Three spans at least one full market cycle, which one year does not.

All three exist because the same mistake kept appearing on a different axis: enough of one thing, none of another.

Worth stating plainly: every one of those thresholds makes our own headline worse. Fewer members beat the index, and the median trails further. That is what tells you the filters aren't there to flatter the result.

What the numbers are, and what they aren't

This matters more than the headline, so it goes above the fold rather than in the footer.

These numbers are the second version, and the first one was wrong.

An earlier draft of this article reported far higher returns: Nancy Pelosi at +340%, Josh Gottheimer at +514% and "one of only 13 members who beat the market". It also carried a confident explanation for why the simulated drawdowns were so small: the portfolio was mostly in cash, so it couldn't fall far.

Both were wrong, and they were the same bug. One planned entry whose ticker had a dead price series, a company renamed years earlier, leaving a stub, blocked every entry queued behind it. Accounts froze for years, then executed the entire backlog at original prices with current cash, booking years of appreciation risk-free. The flat equity curve I read as "mostly in cash" was an account that had stopped trading.

With it fixed, the equity curves behave like ordinary equity strategies rather than frozen accounts. We are leaving this paragraph in rather than quietly restating the figures, because a page that publishes returns should show its corrections.

And these are the third version, because the second was wrong too, in the opposite direction.

The second version sized every position at 5% of the account and left uninvested money flat. That is a punishing assumption: a member who trades rarely spent most of a decade in dead cash while the index compounded, so the simulation charged them for the market's return rather than measuring their picks. It produced the headline "exactly one member of 153 beat the S&P 500", which this article published.

Fixing it (smaller positions, idle cash held in SPY) moved the count to 56 of 146. We nearly published that as "over a third of Congress beat the market", and it would have been just as misleading, because the same change dropped the median member's invested share to 3.9%. The first version measured cash drag. The second measured SPY. Only splitting the board by exposure measures the members.

The lesson we'd rather not have learned twice: a backtest's headline can be an artefact of a parameter nobody thinks of as a parameter.

The win rates are flattering by construction. A position is only scored win-or-lose once the member sells it and the simulation mirrors that exit. Positions nobody ever closed stay open and never count. Since people are famously more willing to sell winners than losers, a high win rate describes the trades a member chose to close, not their judgement.

Options are modelled as shares. A filing gives a ticker, a date and a dollar band. It gives no strike and no expiry, so an option purchase cannot literally be copied. We buy the underlying instead. Across the whole corpus that's a rounding error: 1.6% of buys, but for two well-known members it isn't: Nancy Pelosi at 36.5% and Tommy Tuberville at 25.1%. For those two, a meaningful slice of the copied book is a leveraged bet flattened into a plain share purchase. It reflects what a reader could actually have done. It does not reflect what the member did.

Amounts are bands, not sizes. Every position is equal-weighted at 1.0% of the account because the filings genuinely do not say how much was bought. This is also why the Invested column exists: an equal weight applied to a member who files twice a year leaves the account almost entirely in SPY, and the resulting "excess return" is then a fact about the sizing rule rather than about the member.

Why the lag is the whole game

Members have 30 to 45 days to report. The median gap between trade and disclosure is 28 days, and one in five transactions breaks the 45-day limit entirely, 8% surface more than a year later.

So a headline like "member X made 60% on this stock" is describing a return that began at a price you were never offered. By the time the document is public, the first move has usually happened. Every number on this page starts from the filing date for that reason.

The three famous names rank in order of how little they'd have invested

This is the clearest illustration of the whole problem, so it gets its own section.

Nancy Pelosi ranks 30th of 146, ahead of the index by +6.0 points over 11.7 years. That sounds like a vindication and it is not one. Her copied account held 3.9% in her own picks; the other 96% was SPY. In annual terms she compounded at 14.0% against the index's 13.9%, an edge of roughly a tenth of a point a year, on a portfolio that was almost entirely the index. It is indistinguishable from noise, and it is a much weaker claim than her reputation.

The two members who did commit the account both finished behind. Josh Gottheimer, at 66.4% invested across 1,088 copied positions, is 135th. Tommy Tuberville, at 51.7%, is 119th.

Earlier versions of this article had Pelosi 16th and 15.9 points behind, then 96th. Both were engine artefacts, the fill-queue bug, then the cash-drag sizing. Ranked among members who actually deploy capital, she has too little exposure to rank meaningfully at all.

Two caveats specific to her, both stated above and both load-bearing: 30.7% of her disclosed buys were options modelled here as shares, and 110 of the 128 transactions on her filings belong to her spouse. The filings are joint; the trading is not hers alone.

So should you copy Congress?

On this evidence: the more seriously you copy a member, the worse it has gone. Among the 26 members whose disclosures would have put a meaningful share of your account to work, 4 beat the index and the median trailed by 13.3 points. The members who "beat the market" are overwhelmingly the ones who would barely have invested it, which is not a strategy, it's an index fund with extra steps and a worse tax bill.

Whether following a specific member is worth it is a different question, and the honest answer is that picking the right one in advance is the entire problem. Real outperformance exists on this board, the best member beat the index by +67.4 points with a genuinely invested account, but they are identifiable only in hindsight. Nothing here says it will be the same member next year.

What the data does support, less excitingly: congressional filings are a source of ideas with a timestamp, not a signal with an edge. Treated that way, as a screen to investigate rather than a trade to mirror, the lag matters less and the absence of an edge matters less too.

Run it yourself

The board, every member's fill count beside their return, and the option to combine several members into one strategy: congress trading backtester.

The fine print

Simulation runs on daily closes from the first open after each filing became public, 1.0% of equity per position with idle cash held in SPY, mirroring the member's own disclosed exits. It runs on adjusted closes, so dividends are reinvested; it excludes taxes and market impact, and charges no commission. It can only price still-listed or historically-covered tickers (90.0% of disclosed buys here). Past performance predicts nothing, this is educational analysis and not investment advice, and none of it is a claim about any individual's conduct, only about what a mechanical copy of public filings would have returned.