The Defection Index

Party discipline is the quiet machinery of Congress. Most of what a legislature does is settled before a vote is called, and by the time members are on the floor, the outcome is usually known. What a roll call still records is who was willing to be counted on the other side of their own party.

The defection index measures exactly that, and nothing more. For every recorded vote, we determine where each party's majority landed, then ask how often an individual member voted the other way. A member who votes with their party on every question has a defection rate of zero. A member who splits from their party on one question in three has a rate of roughly a third. That is the whole calculation.

What the number is measuring

It is worth being precise about the comparison, because "voting against your party" can mean several different things and only one of them is captured here.

The benchmark is not a party platform, a leadership endorsement, or a whip count. It is simply the position that most of a member's co-partisans took on that specific question. This has a useful property: it needs no interpretation of what the party "wanted," and it adjusts automatically when a party is itself divided. On a question where a party splits nearly evenly, almost nobody registers as defecting in a meaningful sense, because the majority position is barely a majority at all.

It also has a limitation. A high rate does not tell you why someone broke ranks. It cannot distinguish a member voting their district from one voting their conscience, or from one objecting to a bill's procedural vehicle rather than its substance. A libertarian-leaning member who opposes spending bills on principle and a member from a marginal seat who avoids a politically costly vote will look identical here. The number describes a pattern of behaviour; it does not explain it.

Why most rates are low

The headline figures make the central point about the modern Congress more clearly than any commentary could: the typical member almost never breaks ranks. Party-line voting is not the exception that gets reported, it is the default that goes unreported.

This is partly structural. Leadership in both chambers controls which questions reach the floor, and a bill that would split a party often simply never gets a vote. The roll calls that do occur are therefore a filtered sample, skewed toward questions where the outcome is already secure. Measuring dissent on the floor measures the dissent that leadership was willing to allow into public view.

That filtering matters when reading the rankings below. The members at the top are not necessarily the most independent thinkers in Congress. They are the ones whose independence shows up in the specific, curated set of questions that reached a recorded vote.

The other end of the distribution

The most party-loyal members are less discussed than the mavericks, but they are the larger story. A member with a near-zero rate across hundreds of questions is not necessarily a follower; safe-seat members, members whose views genuinely align with their party's centre of gravity, and members in leadership all cluster here for different reasons.

Leadership positions are worth noting specifically. Members who hold or aspire to leadership have institutional reasons to vote with their party that have little to do with the merits of any individual bill, and their low rates should be read in that light.

Comparing across parties

Aggregate rates by party invite an obvious question: is one party more disciplined than the other? The honest answer is that this measure cannot settle it, because the two parties are not voting on the same questions from the same position.

The majority party controls the floor. Its members are voting on bills their own leadership chose to advance, which tends to produce agreement. The minority party is voting on someone else's agenda, which produces a different mix of unified opposition and occasional cross-over. A gap between the parties in the table below reflects those different positions at least as much as it reflects any difference in internal discipline.

How to use this

Treat the rate as a starting question rather than a verdict. A member near the top is worth looking at more closely: their individual voting record will show which questions they broke on, and that pattern is usually far more informative than the summary figure. A member near the bottom is telling you something about the alignment between their seat and their party, not necessarily about their independence of mind.

Independents are excluded from these rankings entirely. Only three sit in Congress, so "the Independent majority position" would frequently mean one member agreeing with themselves — a comparison with no meaning. Non-voting delegates are excluded for a related reason: they cannot vote on final passage, so their records are not comparable with those of voting members. Both exclusions, and every other rule applied here, are set out in the methodology.