Party positioning space: a comprehensive framework for understanding multidimensional policy-making dynamics

Existing research in political science – specifically party politics – overwhelmingly analyze political parties from a voter-centric perspective. As a result, parties are often portrayed as unitary actors with a few salient policy dimensions that they campaign on for each election. In reality, however, parties constantly navigate both inter-party and intra-party conflicts while spreading their interests across a far broader range of policy dimensions. Therefore, to better understand how parties negotiate within and between themselves to reach agreements on key policy issues, I argue for a shift from a voter-centric to a party-centric perspective. This shift requires a comprehensive theoretical and methodological framework which considers party positions as being multidimensional, stochastic, and fluid over time. Existing frameworks either do not support party-centric view or lack a proper method for scaling stochastic and multidimensional party positions. This cumulative dissertation addresses such a gap by introducing the Party Positioning Framework, developed and implemented through three independent research articles. The first paper presents ContextScale, a sentiment-based approach to scaling multidimensional party positions at the sequence level. Unlike existing scaling methods, ContextScale can distinguish between conflictual and non-conflictual situations, generating a multidimensional positioning space across all policy topics for each party. The second paper improves ContextScale’s cross-domain classification performances by adopting a domain adaptation pipeline which significantly mitigates performance degredation when transferring across political domains without additional fine-tuning. Finally, the third paper applies ContextScale to the case of Germany, highlighting the complex strategies parties employ in their position-taking across stages of the policy making process (from manifestos to parliamentary debates). Empirical findings from the second and third papers demonstrate that adopting a party-centric perspective offers valuable insights into party politics and party competition.