docs: caveat SVD spatial positions reflect voting patterns, not policy content

SVD axes capture agreement structure — centrists 'moving left' means
voting patterns diverged from right-wing, not that parties changed
ideology. 'Acceptance without conversion' is a behavioral claim.
Documented as best-practice learning.

docs/solutions/best-practices/svd-spatial-positions-voting-patterns-not-policy-content.md

@@ -0,0 +1,42 @@
+---
+module: analysis
+problem_type: best_practice
+date: 2026-05-25
+tags:
+  - svd
+  - overton-window
+  - voting-patterns
+  - policy-content
+  - spatial-analysis
+category: best-practices
+---
+
+# SVD Positions Reflect Voting Patterns, Not Policy Content
+
+## Context
+
+When analyzing political shifts with SVD/PCA party positions, it is tempting to interpret spatial movement as ideological change — "centrists moved left" sounds like centrist parties became more progressive. This interpretation is incorrect and dangerously misleading.
+
+## Guidance
+
+**SVD axes capture agreement structure, not policy positions.** A party's coordinate on an SVD axis reflects whom it votes with and against, not what policies it advocates. This was already documented in `svd-labels-voting-patterns-not-semantics.md` for axis labeling, but the same principle applies to spatial drift analysis.
+
+**What SVD movement actually means:**
+- Centrists moving left on an axis means centrist voting patterns became *more distinct* from right-wing voting patterns
+- It does NOT mean centrist parties adopted more left-wing policy positions
+- It does NOT mean the motions being voted on became more left-wing or right-wing in content
+
+**Why this matters for Overton window analysis:**
+- "Acceptance without conversion" (centrists vote more with right-wing while spatially diverging) is a claim about *voting behavior*, not stated ideology
+- A right-wing motion can have high centrist support while still being in a "right-wing" cluster if centrist parties split on it
+- Topics with cross-partisan agreement (defense, nuclear energy) show minimal spatial separation regardless of how radical the motion text is
+
+## When to Apply
+
+- Before interpreting any SVD/PCA spatial drift as "ideological shift"
+- When presenting SVD movement findings to non-technical audiences
+- When cross-referencing voting-pattern results with content-based analysis (LLM extremity scores, mechanism classification)
+
+## Prevention
+
+Always pair SVD drift findings with a caveat: "SVD positions reflect voting patterns, not policy content. We measure behavioral shifts in voting coalitions, not ideological repositioning of party platforms."