Who Wants What? Identification of Transport Policy Preference Classes by Latent Class Regression
Friederike Beck, Isabella Waldorf, Klaus Bogenberger, Allister LoderAmbitious transport policies are necessary to mitigate current, undesired transport system outcomes and effects. However, in times of rising societal polarization, it is unclear whether there is political support for such measures. Based on a representative survey with over 1,700 respondents, the latent policy preference classes that can be identified in German society are examined. Which groups prefer regulatory interventions such as pricing over infrastructural improvements is assessed. Based on the covariates, the results are interpreted as political cleavages in the transport policy domain. Using latent class regression on best/worst rankings of policies across six areas, four distinct preference classes are identified: (1) Public Transport Advocates (26%); (2) Long-Distance Optimizers (15%); (3) Car-Oriented Pragmatists (43%); and (4) Multimodal Interventionists (17%). Of interest, not all groups consistently favor pull measures such as infrastructure improvements over push measures such as a carbon dioxide tax. This suggests that preferences cannot be explained by self-interest alone. The results show the influence of sociodemographics, spatial residential context, and travel behavior on policy preference classes. To reduce dimensionality, principal component analysis is applied to posterior-weighted covariate averages. The results suggest two main conflict lines: (1) a sociodemographic axis (lower to higher resources); and (2) a travel behavior axis (green and multimodal to car-centered mobility). The travel behavior axis is only relevant among those with higher socioeconomic status, suggesting that behavior reflects preferences only if individuals have the resources to choose. The findings contribute to the political economy of transport by highlighting which cleavages shape policy preferences. The results can inform future land use and transport modeling.