Skip to content
MEP Watch

Brandmauer on fire

Every European Parliament vote where the EPP broke its own rule against allying with the far right. Methodology

Rollcalls flagged
Reports
Far right won

By plenary day

Flagged rollcalls

Click a row to open the vote
DateVoteAsked byPositionEPP cohesionForAgainstAbst.

Methodology

Detecting Cordon Sanitaire Breaks in European Parliament Votes

"Brandmauer" (literally "firewall") is the German term used for the cordon sanitaire (political convention by which mainstream parties refuse to cooperate with or lend a majority to the far right).

The core problem: collusion is not directly observable

"Collusion" or "alliance" implies intent: a deliberate political choice by group leadership to vote together, coordinate messaging, or trade support. Roll-call data doesn't contain intent. We only see the outcome: who voted for, against, or abstained on a given text. Two groups can end up voting identically for entirely unrelated reasons; shared ideology on a specific issue, tactical calculation, or pure coincidence on a non-controversial file.

Because of this, we deliberately don't try to answer "did the EPP and the far right plan this together?" That question would require internal party communications, floor speeches, or coalition negotiations that aren't in the data. Instead, we ask a narrower, purely behavioral question: on how many votes did the outcome look like a Brandmauer break, regardless of why?

We measure the pattern, not the direction of the alliance

A second thing the query deliberately does not do is attribute agency. It doesn't try to determine whether "the EPP moved toward ECR/PfE" or "ECR/PfE moved toward the EPP." Both are consistent with the same recorded outcome: they voted together. Assigning direction of movement would require a baseline of each group's "normal" position and a claim about who shifted, which is a separate (and much more interpretively loaded) analysis. Here we flag the convergence itself and let the report title/context inform interpretation case by case.

Exclude broad-consensus votes

If we only looked for "ECR + PfE + EPP voted together," we'd catch an enormous number of votes that have nothing to do with any political realignment. A large share of European Parliament roll calls are broad-consensus votes: technical files, non-controversial resolutions, procedural matters, or texts where almost every group (including the far right) ends up voting yes (or no) simply because there's nothing to contest. In those cases, the EPP isn't "voting with" ECR and PfE in any meaningful political sense; everyone, including S&D and Renew, is voting the same way because the substance is likely uncontroversial.

If we counted those, the result list would be dominated by mundane agenda items rather than genuine political fault lines, and it would say nothing about the Brandmauer.

So the query adds a condition that at least one of the two anchor centrist/pro-European groups — S&D or Renew — voted differently from the ECR–PfE–EPP bloc. This is what turns "everyone agreed" into "the EPP sided with the far right right against (part of) the traditional pro-European coalition." It's the presence of this split with S&D/Renew that gives the convergence political meaning: it isolates cases where the vote was contested enough that the pro-European center was not unanimous, and the EPP chose the ECR/PfE side of that split rather than the S&D/Renew side.

What "breaking" means here, precisely

Put together, a row in the result represents a roll call where:

  • ECR, PfE, and EPP had clear (non-null) majority positions and they matched each other, and
  • Either S&D or Renew (or both) had a clear majority position that diverged from that ECR–PfE–EPP position.

This is a proxy for "the traditional Brandmauer alignment (center-left + center-right + liberals against the far right) did not hold, and the EPP ended up on the far right's side of a real division." It's a statistical signature of the pattern, not a causal or intentional claim.

Known limitations

  • No use of internal group cohesion. A "majority" position with a 51/49 internal split counts identically to a unanimous one.
  • No magnitude of divergence. We only know S&D or Renew's majority differed — not by how much (e.g., 90% vs. 55% of the group dissenting).
  • No analysis of the ESN in practice, if ECR and PfE votes together, the ESN is voting with them too.
  • No analysis of the S&D/Greens in practice, some parties or MEPs might vote with the far right, but not the majority of their group
  • This is descriptive, not diagnostic. The list of flagged votes is a starting point for qualitative review (reading what each amendment is actually about), not a finished claim about political strategy or intent.
  • this isn't a campaign tool (yet). we have a plan to let you use this data directly on your website to let your supporters take action and contact their MEPs