The extent to which traveler preferences and built environment characteristics influence travel behavior are difficult to disentangle because they are so intricately tied to the residential location decision. This paper treats these two issues jointly by focusing on the spatial location of past rail infrastructure in different representative eras as an indicator that is currently omitted from auto ownership models. Past rail has influenced the current built environment, and is likely related to household preferences for automobility. We find that shorter distances to past rail have lower levels of auto ownership in Boston area. This finding is supported by models incorporating instrumental variables (IV) to address the endogeneity issue. The results are robust to the controls of demographic and current BE variables used and the specification of the spatial structure of the model. The implication is that transportation infrastructure has lasting effects on travel behavior beyond the immediate use value. Excluding these long-term effects in current models can skew decision-making.