When committees lack complete voting records, analyzing decision records becomes essential for understanding institutional outcomes. However, this study reveals significant limitations associated with relying solely on such data rather than full votes.

The paper introduces a novel Bayesian structural model designed specifically for decision-record analysis. This approach constructs an exact likelihood function adaptable to diverse institutional contexts while addressing key methodological challenges related to identification and inference.

Data & Methods: Decision records from US state supreme courts (abortion rulings) and UN Security Council deployment votes are analyzed using this Bayesian framework, complemented by a Gibbs sampler for data-augmented posterior density estimation.

* Key Finding 1: Standard decision-record methods systematically introduce bias when compared to voting record analysis.

* Key Finding 2: The proposed model provides more accurate and less biased estimates across various institutional settings.

Why It Matters: This research demonstrates how seemingly minor data differences can fundamentally alter our understanding of committee processes, offering clearer insights into decision-making mechanisms.