🔎 Main Claim:

The standard approach to measuring party valence in spatial models—using estimated party intercepts from discrete choice voting models—does not deliver unique or reliable rankings in fully specified models. This instability stems from core properties of discrete choice frameworks and the particular role that party intercepts play in those models.

📋 How Measurement Is Usually Done:

• Discrete choice models are the empirical workhorse for spatial theories of elections.
• In the classic spatial model, voter choice depends only on spatial proximity to parties.
• Neo-Downsian extensions add voter-level nonpolicy attributes (sociodemographics) into utility functions.
• Schofield’s Valence Model further inserts party valences via party intercepts in voter utility functions.
• Empirical practice orders these estimated party intercepts to create a valence ranking, and this ranking is then used to predict equilibrium party locations.

🇩🇪 Evidence From German Survey Data:

• A simple illustrative example using mass election surveys from Germany shows how the valence ranking can flip depending on arbitrary coding choices.
• The demonstration highlights that the same fully specified discrete choice model can yield different intercept orderings under innocuous coding variations.

⚠️ Why It Matters:

• Because party intercepts in these models are sensitive to coding decisions and to inherent characteristics of discrete choice specifications, they do not provide a unique, interpretable measure of valence.
• As a result, representing party valence solely with those constants and drawing substantive inferences from the resulting rankings is not defensible.

This finding calls for reconsideration of how valence is operationalized and suggests caution in using intercept-based valence rankings to analyze spatial competition or predict equilibrium outcomes.