How does walkability of an area via Walk Score affect access to BART? The algorithm that Walk Score is based on takes into account many area factors. An area is given points based on the number of amenities and how far away they are from an area. There are many limitations to the Walk Score value; Walk Score does not take into account sidewalk locations, crime, or weather. Additionally, Walk Score does not differentiate between amenities—a liquor store is rated the same as a grocery store.
There is a huge range of how people access BART stations: the combined walk and bicycle mode share ranges from 6.5% at North Concord/Martinez station to 97.2% at Powell Street station. There is also a relationship between walk/bike mode share, yet it is only a moderate predictor of the mode share (R-squared = 0.41). Whether or not Walk Score is greater than or less than 93 has a higher predictive value, with an R-squared value of 0.55. Both statistical tests are significant at a 99% confidence level. The combined effect of the dummy variable for Walk Score and the raw Walk Score value is highest, with an R-squared of 0.59.
Walk Score is based solely on the number of amenities (such as grocery stores) and distance those amenities are from the specified location. For such a simple index, this model shows how amenities is a moderate predictor of Walk/Bike mode share. However, there are several stations—specifically Walnut Creek, Orinda, and San Bruno—where Walk Score was a very poor measure of how individuals are accessing a station. These stations might be good areas to focus on if transportation planners want to improve non-motorized mode share. It is possible that large parking lots (between 1800-3500 spaces each) surrounding the station are contributing to lower-than-expected walk/bicycle mode shares.