Historical Calibration Method — Applied to Q1 (Federal Coalition Durability)

Purpose: A reusable method for deciding how far back to look when answering a judgment question, and how to convert historical base rates into a forward-looking indicator. Worked example below uses Q1 from the Germany governance judgment scaffold.

The calibration principle

“How far back is reasonable” isn’t one answer for a whole question — it depends on what you’re trying to extract from the history. Depth should be matched in layers. Going further back doesn’t automatically add value; it can dilute a real signal with irrelevant regime context.

LayerDepthWhat it’s forWhat it’s not for
Current cycle 2025–now The live, high-resolution mechanism: this coalition’s actual margin, actual tensions, actual defections Base rates — one data point isn’t a pattern
Comparative era ~2005–now (20 yrs) Base rates: how German coalitions actually form, strain, and end in the current fragmented-party-system era Anything before this era assumes a party system (stable two-and-a-half-party CDU/SPD/FDP) that no longer exists
Institutional base rate 1949–now (77 yrs) How rare/common the formal removal mechanisms actually are — calibrates whether “confidence vote risk” is a real near-term threat Predicting when — this layer gives frequency, not timing
Design intent 1918–1949 (Weimar) Why the guardrails (constructive no-confidence, 5% threshold) exist at all — useful for judging whether current stress resembles what they were built to prevent Not a data source — don’t pull “base rates” from Weimar; the system was deliberately redesigned to not behave like that one

Rule of thumb: the era where the current party system actually operates is where the real analytical yield is. Older layers are for framing and for calibrating rarity, not for pattern-matching.

For Q1, the productive layer is 2005–now. Pre-1949/Weimar is one paragraph of framing, not a data table.

Worked example: Q1 traced back to 1949

GovernmentEndedHow it actually ended
Erhard (1963–66) Coalition collapse FDP walked out over a budget dispute
Brandt (1969–74) Resignation Spy scandal — survived a constructive no-confidence attempt in 1972
Schmidt (1974–82) Removed Only successful constructive no-confidence vote in FRG history — but it happened because the FDP switched sides mid-term, not because the opposition won a vote against an intact coalition
Schröder (1998–2005) Early election Schröder deliberately lost a confidence vote after a state election defeat, to force his own early election
Merkel (2005–21) Normal term / voluntary exit 16 years, four coalitions, no collapse — closest thing to a stable baseline
Scholz (2021–24) Coalition collapse Scholz fired his FDP finance minister; FDP exited; early election followed

The pattern that falls out

In 77 years there has been exactly one successful constructive no-confidence vote — and even that was really a coalition-partner defection dressed up as a formal mechanism. The dominant real-world failure mode for German coalitions is not “the Bundestag votes out the chancellor.” It is:

  1. A junior coalition partner walking (FDP in 1966, FDP in 1982, FDP in 2024), or
  2. The chancellor engineering their own exit (Schröder 2005).

The formal no-confidence vote is almost never the actual mechanism — it’s usually the paperwork after the real decision (a partner leaving) has already happened.

What this changes for the forward-looking question

The original Q1 sub-question was: “Is there a realistic path to an early confidence vote?” The base rate says that’s the wrong thing to watch — floor votes are the tail, not the trunk, of the historical pattern.

Reframed question: “What is the SPD’s exit threshold, and how close is it?”

This is more answerable and more trackable in real time than modeling a floor vote. It also tells you where to look:

Net effect on the research plan: shift effort away from legal/procedural research on confidence-vote mechanics, and toward tracking SPD internal signals against the historical exit pattern (policy dispute + declining poll position + a visible off-ramp).

Generalizing the method to the other scaffold questions

This file is a companion methodology note to germany_judgment_scaffold_reworked.html. Apply the same layered-lookback structure to each judgment question before researching it, rather than defaulting to a single fixed lookback window across all seven.