Signal validation Does each signal pay after disclosure?
For every congressional buy we note which of six signals applied, using only what was public on its filing date. Then we measure what a follower could have earned: the excess return vs the S&P 500 from the filing date, 30 and 90 days later. Same method for every signal, published whatever it shows.
1 signal did better and 0 did worse than the average buy. The average congressional buy returned -0.3% vs the S&P over 90 days (95% CI -0.7 to +0.2, n=11,063). A signal only counts as standing out when that average falls outside its interval, widened for testing 6 signals at once.
| Signal | Avg 30d | Avg 90d | 95% CI (90d) | Hit rate (90d) | Buys | Result |
|---|---|---|---|---|---|---|
| Committee overlap The member sits on a committee that oversees the stock's sector. |
+1.2% | +0.4% | -1.2…+2.0 | 43.5% | 775 | no measurable edge |
| Bipartisan cluster At least 3 members, from both parties, bought the stock within 30 days (all already disclosed). |
+1.3% | +3.1% | +1.0…+5.3 | 52.2% | 709 | did better |
| Cross-chamber cluster At least 3 members, from both the House and the Senate, bought within 30 days (all already disclosed). |
+1.4% | +2.8% | +0.4…+5.2 | 56.7% | 328 | no measurable edge |
| Legislative timing A bill touching the stock's sector was introduced or acted on within 10 days of the trade. |
+0.0% | -0.4% | -1.3…+0.5 | 43.8% | 3,651 | no measurable edge |
| Unusual size The amount is unusually large or out of character for this member. |
+0.2% | -0.9% | -1.5…-0.3 | 42.7% | 4,200 | no measurable edge |
| Rare trader From a member with very few trades on record (2 or fewer). |
-17.6% | -10.8% | -22.0…+0.3 | 0.0% | 2 | too few cases |
| All congressional buys (baseline) | +0.1% | -0.3% | -0.7…+0.2 | 44.7% | 11,063 |
How to read this
Excess return is the stock's return minus the S&P 500's over the same days, starting on the public filing date (what a follower could act on, not the member's own trade date). Hit rate is the share of buys that beat the index over 90 days. A trade can carry several signals, so the rows overlap.
Why the intervals are wide: buys of the same stock in the same month move together, so we treat each stock-month as one unit (a clustered interval). Treating them as independent would make every signal look more certain than it is. With six signals tested, one can look good by chance; that is why "stands out" uses an interval widened for six tests.
Covers buys filed since 2021-10-25 with a full 30- or 90-day window. Committee assignments are the current ones, a small simplification for older trades. No costs or taxes. This describes the past; it is not a forecast or advice.
See also: score bands vs returns · track record · conviction lens · for quants