UC Irvine

The recent growth in bikeshare systems has received an enthusiastic response from the research community interested in understanding the factors that influence bikeshare demand. Many research efforts have modeled spatial interactions using a distance-based weight matrix. However, when biking for utilitarian reasons, biking time may be equally, if not more, relevant. To this end, this thesis explores the demand for bikesharing in downtown Los Angeles by contrasting two spatial SARAR models with the same explanatory variables but different weight matrices: one has a distance-based weight matrix, and the other a time-based weight matrix. To the best of my knowledge, this work is the first to contrast these models and to analyze the demand for Los Angeles’ bike sharing program. Explanatory variables in my spatial models include socio-demographic, land use and transport characteristics in the proximity of bikeshare stations. The results show that (i) incorporating spatial interactions (spatial lag and spatial errors) is an important feature of bikesharing demand; and, (ii) that models with time-based weight matrices perform better than similar models with distance-based weight matrices for the models I considered. Finally, my models show that bikesharing is seldom used as a mode for the ‘last mile’ travel to access transit in downtown Los Angeles.