Tutz, G. & Mauerer, I. Heterogeneity in General Multinomial Choice Models.

arXiv preprint. arXiv:2011.02688

Different voters behave differently, different governments make different decisions, or different organizations are ruled differently. Many research questions important to political scientists concern choice behavior, which involves dealing with nominal-scale dependent variables. Drawing on the principle of maximum random utility, we propose a flexible and general heterogeneous multinomial logit model for studying differences in choice behavior. The model systematically accounts for heterogeneity that is not captured by classical models, indicates the strength of heterogeneity, and permits examining which explanatory variables cause heterogeneity. As the proposed approach allows incorporating theoretical expectations about heterogeneity into the analysis of nominal dependent variables, it can be applied to a wide range of research problems. Our empirical example uses data on multiparty elections to demonstrate the benefits of the model in the study of heterogeneity in spatial voting.

Mauerer, I. & Schneider, M. Uncertainty in Issue Placements and Spatial Voting.

LMU Munich, Department of Statistics: Technical Reports, Nr. 226.  

doi:10.5282/ubm/epub.68451

Empirical applications of spatial voting approaches frequently rely on ordinal policy scales to measure the policy preferences of voters and their perceptions about party or candidate platforms. Even though it is well known that these placements are affected by uncertainty, only a few empirical voter choice models incorporate uncertainty into the choice rule. In this manuscript, we develop a two-stage approach to further the understanding of how uncertainty impacts on spatial issue voting. First, we model survey responses to ordinal policy scales where specific response styles capture the uncertainty structure in issue placements. At the second stage, we model voter choice and use the placements adjusted for the detected uncertainty as predictors in calculating spatial proximity. We apply the approach to the 2016 US presidential election and study voter preferences and perceptions of the two major candidate platforms on the traditional liberal-conservative scale and three specific issues. Our approach gives insights into how voters attribute issue positions and spatial voting behavior, and performs better than a voter choice model without accounting for uncertainty measured by AIC.

Mauerer, I. & Walter, A. S. Taking Ballots Seriously: Heterogeneous Ballot Compositions and Vote Choice. POLMETH Paper Archive

In partially-contested multiparty elections, voters are confronted with different party choices, depending on their constituency. We present a computationally straightforward modeling approach that systematically integrates heterogeneous ballot compositions, which classical models neglect, into the voter utility functions. We illustrate the benefits of the approach in studying British spatial voting behavior, where previous studies tend to simplify the actual choice situation by modeling a single ballot composition, thereby ignoring a substantial part of the electorate. Using 2015 British Election Study data, we simultaneously consider up to seven parties, spread across eight unique ballots, and provide a fully-specified vote model. The results show that both spatial and tactical considerations depend on which party voters evaluate. Whereas spatial proximity substantially impacts voting for the large parties, we uncover the reversed pattern for tactical  considerations. These party-specific effects are not found when neglecting ballot composition heterogeneity.