Bayesian Aldrich-McKelvey scaling offers an improved approach to measuring ideological preferences and perceptions.
Methodology:
This method addresses differential item functioning (DIF) in issue scale data, correcting biases in both self-placement and stimuli placement along a latent policy dimension. It provides a Bayesian implementation of classical maximum likelihood methods.
Key Findings:
Unlike classical approaches, this technique reveals higher levels of polarization among contemporary American voters than previously estimated using traditional scaling methods applied to ANES or CCES data.
Citizens demonstrate surprisingly accurate understanding of senators' and Senate candidates' ideological positions when measured with Bayesian scaling.
Policy Implications:
These findings suggest that classical survey methods may undercount ideological divides in the electorate, potentially affecting how political scientists understand public opinion dynamics and campaign strategies relying on perceived polarization.
The analysis uses data from two significant election surveys: American National Election Studies (2004-2012) and Cooperative Congressional Election Study.
