Capitol Alpha? Do Politicians Beat the Market? Congress Stock Trading, 2012–2026

Open with the question in the title: is there alpha in Washington’s trades? Over fourteen years and seventy thousand trades, the answer is: rarely, slowly, and mostly in the Senate — and most of what looks like skill is style. Anyone can tell you what Congress bought. The institutional question is what you could have known, and on which day. That is what this database is built around: a point-in-time timestamp on every row.

Howard Marks’ distinction. First-level thinking is the headline everyone already has. Second-level thinking asks what was actually knowable and tradable, and what is already priced by the crowd. The disclosure lag is the whole game: a typical trade becomes public about four weeks after it happened.

Roughly a billion dollars of retail capital already mirrors congressional disclosures, ten times more than two years ago. When that much money buys the same disclosed names on the same day, the disclosure itself becomes the event — which is a second-level signal in its own right.

The party split behaves like growth versus value. Strip out sector and size and the copy trade looks like a tech tilt with a fee. A buyer who wants real alpha needs to separate the style exposure from any information content — which requires point-in-time data and disclosure-date returns.

The early studies were trade-weighted, and four senators made almost half the Senate trades. Equal-weight them and the edge shrinks or vanishes. Since the STOCK Act, the broad samples show nothing — and our own fourteen-year test, with factors removed, finds zero alpha for Congress as a whole. So Congress as a whole is not the signal. That is exactly why a buyer needs the granular, member-level, point-in-time record.

This is the second-level finding. The signal is conditional: leaders, and committee members trading within their jurisdiction. Those conditions are joins — member identity, committee assignments, tenure — on top of a point-in-time trade record. Note the leadership result rests on a small sample; buyers should test it themselves, which is what the database lets them do.

Two careful papers disagree. Lazzaretto finds a post-disclosure drift that looks like copy-trading pressure, not insider information. Pyun finds no difference once the benchmark and same-day filings are handled carefully. For an institution, the answer depends on its own universe, costs and timing — so the asset is the clean record with both dates on every row.

The scoreboard on the full history. Against the S&P 500, every branch is behind at disclosure and the House is behind at a year, minus 2.2 percent. Against the equal-weight index — the fair benchmark for a stock picker — the House is flat and the Senate is ahead by 2.8 percent at a year. The President’s purchases lost 2.7 percent against the S&P 500 and 3.5 against equal weight before the public saw them.

This is the slide that shows why second-level thinking matters. Against a cap-weighted index, the President’s sells look brilliant and his buys look terrible — both significant. Re-run against an equal-weighted basket of the same stocks and both effects vanish. April 2026 was a mega-cap-led rally: the index rose 4 percent while the typical stock fell 2 percent, so anything that is not mega-cap tech looks like bad buys and good sells. Any provider who benchmarks congressional or presidential trades only against SPY will hand you this artifact as alpha. FinzData publishes both benchmarks.

Every purchase since 2012, measured at every horizon from the trade to four years out. Against the cap-weighted S&P 500, Congress loses, and the House loses badly — almost fourteen points over four years. Against the equal-weight index, the House is flat and the Senate wins, but slowly: nothing before disclosure, then 2.8 percent at a year and 7.7 at four. The President’s April 2026 purchases lost 3.5 percent between trade and disclosure.

This is the whole time path, day by day, for every purchase since 2012. The House line sits on zero for four years: no edge at any horizon. The Senate line drifts up slowly — about 3 percent by one year, 11 percent by four — but the band is wide, and there is no jump before disclosure. Whatever senators know, it pays over years, not weeks.

This is the test institutional allocators should ask for. We hold every disclosed purchase in a live portfolio and regress its daily returns on market, size, value and momentum. After those four factors, Congress as a whole has zero alpha: what looked like an edge over the equal-weight index is a steady small-cap and value tilt. The Senate keeps about 1.8 percent a year, but with a t of 1.2 it is not distinguishable from luck. Event-study t-statistics overstate significance because purchases cluster in time; the calendar-time portfolio is the honest benchmark.

The investable version: buy each purchase the day it is disclosed and hold a year. Since 2014 a Senate copy book compounded at 12.5 percent a year against 11.8 for the S&P 500 — but with more volatility and a deeper drawdown, so risk-adjusted it is a tie. The House book trails the index by more than two points a year, in line with the equal-weight index it resembles. Copying Congress wholesale is a small-cap fund with extra turnover; the value is in the filters on the next slides.

Republicans versus Democrats against the equal-weight index. Democrats come out ahead, but by about a point a year — significant from 90 days to one year and plus 3 percent at four. Republicans are flat or slightly behind, with House Republicans about two points behind at two years. Two Senate exceptions: Senate Republicans gain 2.7 percent at a year, and the small Senate Democrat book compounds to 13 percent at four years. The next slide shows why the gap exists.

Democrats’ purchases drift ahead over four years and Republicans’ drift behind, a gap of about seven points. The right panel explains it. Both parties buy smaller companies than the index, equally. But Republicans tilt firmly to value stocks and Democrats do not — and value lagged growth for most of this decade. Once style is removed, both parties’ alpha is the same, about minus 0.4 percent a year. The party label is a style signal, not a skill signal.

Grouping purchases by the presidential term in which they were made. There is no permanent party edge. Under Trump’s first term, Democrats bought better, about 3 points ahead at a year while Republicans lagged. Under Obama’s second term, Republicans lagged badly — over 5 points at two years. Under Biden, neither side stood out at a year. Under Trump’s second term it flips: Republicans’ purchases lead, plus 5.2 percent at a year.

By term, the winning side flips: Democrats under Trump’s first term, Republicans under Trump’s second — plus 5.2 percent at a year. By year, Congress beat the equal-weight index in 2017, 2019 and 2023 to 2025 and lost badly on 2021 purchases, minus 5.8 percent, the year speculative stocks peaked. Against the S&P 500, Congress lost in every year but 2020. A durable signal has to be conditioned on regime, not party.

Two filters separate signal from noise. Size: the routine $1,000 to $50,000 purchases that make up nine in ten trades show nothing, while purchases of $50,000 and up beat equal weight by 2.6 to 6.6 percent over a year. Speed: purchases disclosed within 15 days gain 2.7 percent, while late filings — after the 45-day deadline — lose 2 to 3 percent. Late filing is not hiding winners. A product that ranks disclosures by size and filing speed is more useful than one that copies everything.

Member by member, over the full history. Per trade, fourteen members beat the equal-weight index with statistical significance, where chance would give about three. But look at the dates: the biggest winners, Mark Green, Don Beyer and Marjorie Taylor Greene, did most of their buying in the 2020 to 2021 rebound, when small speculative stocks soared, and purchases bunch together in time, which inflates per-trade t-statistics. Hold each member’s purchases as a portfolio and remove size, value and momentum, and only three remain significant — Beyer, Jerry Moran and Bill Keating. On the other side, Virginia Foxx lagged consistently for a decade under both tests.

The cleanest test of skill per member. We hold each member’s purchases as a portfolio and strip out the market, company size, value and momentum. The cloud centres on a beta of 1.0 and an alpha of zero: most of Washington is buying the market with a small-cap tilt. Only 3 members show significant positive alpha — Don Beyer, Jerry Moran and Bill Keating — and 3 significant negative, roughly what chance alone produces. Names that look brilliant per trade, such as Mark Green, Marjorie Taylor Greene and Josh Gottheimer, lose significance once style is removed.

The chambers are structurally different. Half of House trades are in the member’s own account; in the Senate, three quarters run through a spouse or a joint account, and senators are more likely to trade over $50,000. Both chambers disclose in under a month at the median, and about one trade in six arrives after the 45-day deadline. The Senate’s purchases are the ones that beat equal weight over a year.

The President’s disclosures run through the Office of Government Ethics on Form 278-T, not through Congress, and they are scanned PDFs. The White House says outside managers run the accounts. The volume is the story: more than 3,700 trades in one quarter, more than our entire 50-member congressional sample over nine months, with individual positions in companies that have business before the administration. For institutions this is a flow and positioning dataset in its own right — and it is the next coverage FinzData is adding.

Four applications. First, filter rather than copy: big tickets, fast filers and a short list of members with factor-robust records, tested from the disclosure date. Second, the dependable effect is flow: retail copy capital and a very large presidential account trade disclosed names, which a fund can position around. Third, factor hygiene: across fourteen years Congress’s apparent edge is size and value exposure, with zero alpha after factors. Fourth, governance and policy risk across all three branches.

Late filings cut both ways. They are look-ahead poison in any backtest dated by trade date — a trade disclosed two years late looks tradable two years too early. And they are a behavioural signal: late-filed purchases underperform by two to three percent over a year. Lateness spikes in particular years, such as 2021, when catch-up filings arrived in bulk. FinzData flags every late row against the 45-day rule.

Each of these is a common way a congressional-trades backtest looks better than reality. The product decisions on the right are the reason the database exists.

These numbers come from our own build against the official House Clerk and Senate eFD sources, back to 2012. Every row links back to its source document, carries the date it became public, and keeps every amended version.

The same point-in-time record serves several desks. Quants test conditional signals; risk teams anticipate crowded flow; policy teams map exposure; AI agents query it directly through MCP.

This is the diligence checklist an institutional data team runs. The differentiator is not the trades themselves — those are public — but whether the history is point in time, revision-tracked and complete.

Pricing is designed to fall below discretionary-spend thresholds, so a research team can buy without a procurement cycle. The complete two-chamber database is $395 one-time, and the data is theirs to keep.

Reform is a tailwind for a point-in-time vendor. If a ban passes, disclosed buys stop, sale notices arrive in advance, and the historical archive becomes the only place to study the old regime. If it does not pass, the status quo continues.

Close on the trial. The free three months lets a research team run a disclosure-date event study on its own universe before buying. The history to 2012 has shipped; the roadmap items are in build now.