7 September 2026

What changed

If you have used this site before, the numbers moved this week. Most members are in a different band than they were, and almost nobody's score went up. That is a change in how the score is calculated, not a change in how anyone voted.

The score measures one thing now, not two

The Independence Score used to be the average of two numbers: how often a member broke with their own party on contested votes, and how often they broke with bipartisan consensus. Those two do not share a scale. Consensus deviation runs about six times larger — a median of 3.5% against 0.6% — so the average tracked consensus deviation almost exactly and said very little about party discipline, which is the thing this site is about.

The score is now cohesion-weighted partisan deviation alone: how often a member votes against their own party's majority, weighted by how united that party was on each vote. Consensus deviation has not gone away — it sits in its own column, beside the score instead of inside it.

The practical effect is that the typical score fell by roughly three quarters.

 BeforeAfter
Median member2.71%0.59%
Most independent member34.0%29.9%

The bands were rescaled, and most labels are new

The old cut points were set for the blended score. Against the new one they left about 90% of Congress in the bottom two bands, which is not a scale so much as a pile. The new cuts were checked against the 117th, 118th and 119th Congresses so that every band is occupied in each of them and the top band stays genuinely rare.

ScoreLabelMembers
under 0.5%Mindless Drone244
0.5 – 1.5%Bobblehead156
1.5 – 4%Squeaky Wheel86
4 – 8%Free Agent39
8 – 20%Heretic13
20% and upLone Wolf5

76% of members carry a different label than they did last week. Almost all of that is the two changes above, not new voting.

Consensus deviation shows its denominator

The two chambers do not hold comparable numbers of bipartisan votes — the House's are mostly suspension-of-the-rules bills and it holds roughly three times as many as the Senate, whose bipartisan votes are mostly procedural. One Senate defection therefore moves that rate about three times as far. The column now reads 62% of 67 rather than 62%, and sorting it groups the chambers instead of ranking them against each other.

Two columns are gone

Party Unity and Consensus Votes have been removed from the table. Party unity is unchanged and still published in data.json as party_unity_pct, and each member's page still carries it. The consensus denominator now travels beside its own rate, so a separate column for it was showing the same number twice.

The page is laid out differently

The table and its filters are at the top now, and the explanations are underneath it rather than above. There is a new chart showing where everyone lands on the scale. The heap at the left-hand end of it is not a rendering fault: most of Congress almost never breaks ranks, and that is the point the chart exists to make.

If you read the data files

data.json now carries a tiers table and each member has an independence_tier id in place of independence_label. Four fields nothing consumed were dropped: independence_label, weighted_partisan_deviation_pct (identical to the score), consensus_loyalty_pct (the inverse of consensus deviation) and district.

← Back to the index