Segmenting fare-evaders by tandem clustering and logistic regression models

In this study, a tandem clustering is applied on data collected by an Italian public transport company. Three clusters of evader passengers are discovered. Next, for each cluster, the influence of significant determinants in evaluating the chance of being a frequent fare evader is shown by logistic regression models. Members of Cluster 1 are a small segment of choice-workers, who seldom evade fares significantly. Members of Cluster 2 represent a big segment of captive students, who often evade the fare. Members of Cluster 3 are a medium segment of captive unemployed, who always evade the fare. The logistic regression models show that attributes related to the situational factors are significant, and honesty is the common variable that significantly affects the chance of being a frequent fare evader among segments. These outcomes are relevant and useful for both research and practice. Indeed, this paper contributes to the empirical understanding of the determinants of being a frequent fare evader for segments a posteriori selected. Moreover, it helps PTCs to better understand how some segments differ from each other.