Eight parties
Parties
Party by party: how closely each one’s actual votes match what its own election manifesto and party programme say it should have done.
Alignment
Who votes with whom?
| V | S | MP | C | L | KD | M | SD | |
|---|---|---|---|---|---|---|---|---|
| V | – | 41% | 55% | 34% | 42% | 42% | 42% | 29% |
| S | 41% | – | 44% | 39% | 50% | 49% | 50% | 36% |
| MP | 55% | 44% | – | 39% | 37% | 37% | 37% | 27% |
| C | 34% | 39% | 39% | – | 50% | 50% | 50% | 42% |
| L | 42% | 50% | 37% | 50% | – | 100% | 100% | 79% |
| KD | 42% | 49% | 37% | 50% | 100% | – | 100% | 79% |
| M | 42% | 50% | 37% | 50% | 100% | 100% | – | 79% |
| SD | 29% | 36% | 27% | 42% | 79% | 79% | 79% | – |
Share of divisions where two parties took the same position (Ja/Nej/Avstår), whole term.
The coalition, or the party programme?
Every division pits the committee's proposal (Ja) against a reservation (Nej). Two rival explanations of how a party votes are set against each other: the coalition line — vote with the committee, knowing nothing about the party — and the party programme, which the AI agent reads without ever being told who authored the reservation. The gap between them is the measurement.
What explains the party's vote?
| Coalition line | Party programme | What the programme adds | n | ||
|---|---|---|---|---|---|
| KD | 99% | 21% | −78.3 pp | 1182 | |
| L | 99% | 23% | −76.3 pp | 1182 | |
| M | 99% | 42% | −57.1 pp | 1183 | |
| SD | 71% | 42% | −29.3 pp | 1183 | |
| C | 50% | 37% | −12.4 pp | 1183 | |
| S | 55% | 47% | −8.2 pp | 1183 | |
| V | 49% | 48% | −0.9 pp | 1182 | |
| MP | 41% | 44% | +3.2 pp | 1182 |
Coalition line = the share of votes hit by the rule "always vote Ja, with the committee". Party programme = how often the party's own plan points to the vote it actually cast. The difference is what the plan adds: zero means it explains nothing beyond coalition position.
M · KD · L — This is now measured without bloc documents: M, KD and L are read against their own manifestos and party programmes, not against the Tidö agreement. That the coalition line accounts for nearly 99% of their votes while their own plans often point elsewhere is therefore not an error in the measurement — it is the measurement. The difference is what the coalition deal carries, not the party programme.
When the coalition outweighs a party’s own documents
Divisions where the party’s own documents explicitly imply voting against the committee — yet it voted Ja anyway. Split by whether the party moved the reservation itself.
| Own reservation | Another party's reservation | |||
|---|---|---|---|---|
| strict | loose | strict | loose | |
| KD | – – | – – | 101/101 100% | 150/150 100% |
| L | – – | – – | 182/182 100% | 230/230 100% |
| M | – – | – – | 101/101 100% | 136/136 100% |
| SD | 0/54 0% | 0/54 0% | 90/120 75% | 102/137 74% |
| C | 0/81 0% | 0/90 0% | 42/101 42% | 49/113 43% |
| S | 0/93 0% | 0/99 0% | 26/47 55% | 33/58 57% |
| V | 0/105 0% | 0/110 0% | 78/122 64% | 82/129 64% |
| MP | 0/132 0% | 0/137 0% | 55/135 41% | 60/145 41% |
Where a party moved the reservation itself, it never abandoned it — 0% for every party, at both thresholds. Procedure is decisive where procedure applies. The right-hand column is therefore the real test.
The divisions (842)
The individual divisions are listed on each party's own page, alongside the rest of that party's figures.
What this does not show
A flag means a document-grounded reading of the party’s programme pointed the other way from its vote. It cannot by itself distinguish "coalition discipline overrode the programme" from "the AI agent misread the programme". That parties never abandon their own reservation (0% in every case below) suggests the measurement captures something real — but individual rows should be read as questions, not verdicts.