Examining the nonlinear effects of neighborhood housing + transportation affordability on shared dockless e-scooter trips using machine learning approach

Despite the growing popularity and benefits of shared dockless e-scooters, there is controversy over whether this is an affordable travel mode for everyone. This paper explores the nonlinear relationship between shared dockless e-scooters and the location affordability of neighborhoods. By analyzing shared dockless e-scooter trip data collected between April 2019 and March 2020 from 1,886 census block groups in Los Angeles, we used a random forest model to investigate this nonlinear relationship. The variable importance plot revealed that economic variables (cost versus income) have greater explanatory power than other variables. In the partial dependency plots, neighborhoods spending more than 35% of their income on housing costs were more inclined to use e-scooters. On the other hand, when transportation represents more than 9% of household income, the e-scooter trip density decreases. Location affordability appears to be serving as a proxy for compactness, with compact areas having higher housing costs and lower transportation costs. The two together are lower in compact areas. The market for e-scooters is thus higher in compact areas where there are many potential users, trips are shorter, and users have more discretionary income due to lower h + t costs. The results of this study highlight the importance of location-specific planning in promoting the effective use of shared dockless e-scooters as a sustainable and active transportation mode beyond simply focusing on costs and incentive programs.