An indirect latent informational conformity social influence choice model: Formulation and case study

•Latent class discrete choice model of informational conformity developed.•Equilibrium properties derived and compared to direct-benefit conformity models.•Two-stage control function approach developed for endogeneity in latent class models.•Distributions of social equilibria derived in case study on bike ownership.

The current state-of-the-art in social influence models of travel behavior is conformity models with direct benefit social influence effects; indirect effects have seen limited development. This paper presents a latent class discrete choice model of an indirect informational conformity hypothesis. Class membership depends on the proportion of group members who adopt a behavior. Membership into the “more informed” class causes taste variation in those individuals thus making adoption more attractive. Equilibrium properties are derived for the informational conformity model showing the possibility of multiple equilibria but under different conditions than the direct-benefit formulations. Social influence elasticity is computed for both models types and non-linear elasticity behavior is represented. Additionally, a two-stage control function is developed to obtain consistent parameter estimates in the presence of an endogenous class membership model covariate that is correlated with choice utility unobservables. The modeling framework is applied in a case study on social influence for bicycle ownership in the United States. Results showed that “more informed” households had a greater chance of owning a bike due to taste variation. These households were less sensitive to smaller home footprints and limited incomes. The behavioral hypothesis of positive preference change due to information transfer was confirmed. Observed ownership share closely matched predicted local-level equilibrium in some metropolitan areas, but the model was unable to fully achieve the expected prediction rates within confidence intervals. The elasticity of social influence was found to range locally from about 0.5% to 1.0%.

Graphical abstractDownload high-res image (88KB)Download full-size image