Across 30,000+ STOCK Act trades, congressional purchases lag the market on average — and high-conviction sells were once the most accurate signal we tracked. A July 2026 recalibration cut that claim down to a coin-flip win rate on a much larger sample — but the +4.2% average alpha survived. Here is the thesis, what changed, and why we show both.
Congressional buys are a media genre unto themselves — a representative picks up a chip stock and it's a headline by morning. Sells get the opposite treatment: written off as noise. Taxes, rebalancing, tuition, an ethics divestment — there are a dozen boring reasons to sell, so the assumption is that a sale tells you nothing.
We've leaned on that assumption ourselves. In Does Congress Beat the Market? we deliberately excluded sales, calling them liquidity-driven rather than information-driven. On average, across all sales, that's a fair simplification.
But “on average” buries the most interesting thing in the data. Filter for conviction, and the exits start to look a lot smarter than the entrances.
Start with the baseline. Buy every stock Congress bought — all 10,428 purchases — and you'd have trailed the S&P 500 by 2.40% over the following 180 days, with a win rate of just 43%. As a group, Congress is a below-average buyer.
Now flip to the sell side. Isolate high-conviction sells — the ones the conviction model flags on size, speed of disclosure, and relevance — and the picture inverts.
High conviction isn't every sale. It's the subset the conviction score elevates — meaningful position sizes, fast disclosure rather than a 45-day-late filing, and committee or sector relevance. Routine, small, slow-filed sales don't qualify.
A “win” for a sell signal is scored the opposite of a buy: the signal is right when the stock subsequently underperforms the market. At 50.3%, the current cohort is right about half the time — what keeps it valuable is the payoff asymmetry. The weakness it catches is larger than what its misses give back, which is how a coin-flip hit rate still nets +4.20% average alpha over six months.
This is the part we can't prove, only reason about — so treat it as hypothesis, not accusation.
None of that implies wrongdoing by any individual. It's a statistical pattern in public filings — but it's a pattern worth watching.
See every strategy side by side — buys and the lone sell signal — on the Research page, watch new high-conviction sells land in the Signal Log, and rank the members behind them on the Leaderboard.
High-conviction sell signals, conviction-scored trades, and real-time STOCK Act alerts — updated daily from the official filings.