��ɫ��Ƶ

The Department of Politics and Public Administration kicks off its weekly research seminar series with a talk from Philip Warncke (UL Psych/Freie Universität Berlin) entitled “Boundless but Bundled: Modelling Quasi-infinite Dimensions in Ideological Space”. The seminar will take place on Wednesday, September 24, 14:00 in the ERB006 (Engineering Research Building). 

All are welcome to attend! The abstract for Philip's seminar can be found below. Boundless but Bundled: Modelling Quasi-infinite Dimensions in Ideological Space Ideological summary scales derived from policy position items are prevalent in political psychology and behavioral research. However, past scholarly practice shows little to no consensus on how many such scales (i.e., ideological dimensions) researchers should consider to adequately capture the main political dividing lines among mass publics. Using extensive statistical sensitivity analyses based on ANES and CCES data, we demonstrate that the optimal number of latent ideological dimensions grows indefinitely as additional issue position items are included. Rather than increasing measurement precision, these additional items instead likely increase the uncertainty about which underlying construct they are intended to measure in the first place. 

At the same time, nearly all latent ideological factors detectable within policy position data are strongly and positively correlated with each other, suggesting that distinct ideological sub-dimensions ultimately stem from a more abstract, unifying parent concept. We propose a Bayesian hierarchical latent variable modeling framework that reconciles the boundless yet correlated dimensional nature of mass ideology. The proposed model estimates ideology as a higher-level expression of correlated, lower-level building blocks, enabling researchers to simultaneously assess whether external predictors—such as income or gender—consistently relate to specific ideological sub-dimensions (e.g., economic or socio-cultural ideology) or instead to a generalized, unidimensional representation thereof. Our results underscore the potential of this approach, offering insights into the unique features of different ideological sub-factors and their overarching parent dimension.