Dockless bikeshare systems show potential for replacing traditional dock-based systems, primarily by offering greater flexibility for bike returns. However, many cities in the US currently regulate the maximum number of bikes a dockless system can deploy due to bicycle management issues. Despite inventory management challenges, dockless systems offer two main advantages over dock-based systems: a lower (sometimes zero) membership fee, and being free-range (or, at least free-range within designated service areas). Moreover, these two advantages may help to solve existing access barriers for disadvantaged populations. To date, much of the research on micro-mobility options has focused on addressing equity issues in dock-based systems. We have limited knowledge of whether, and the extent to which dockless systems might help mitigate barriers to bikeshare for disadvantaged populations. Using San Francisco and Los Angeles as case studies, because both cities have both dock-based and dockless systems running concurrently, we quantify bikeshare service levels for communities of concern (CoCs) by analyzing the spatial distribution of service areas, available bikes and bike idle times, trip data, and rebalancing among the dock-based and dockless systems. We find that dockless systems can provide greater availability of bikes for CoCs than for other communities, attracting more trip demand in these communities because of a larger service area and frequent bike rebalancing practices. More importantly, we notice that the existence of electric bikes helps mitigate the bikeshare usage gap between CoCs and other tracts. Besides the data analyses for bikeshare trips, we also study the spatial distribution of online suggested station locations and find that the participants’ desired destinations for work/school purposes have not been covered to the same extent in CoCs as in other communities. Our results provide policy insights to local municipalities on how to properly regulate and develop dockless bikeshare systems to improve mobility equity.

There are three sets of data collected for this study: 1) communities of concern; 2) bikeshare data; 3) online suggestion data. We introduce the collection process for three datasets separately.

Communities of concern: There is not a unified definition of disadvantaged populations or underserved communities across different regions in the US. Our previous study defines disadvantaged communities based on income, percentage of minority populations, and vehicle ownership (Qian and Niemeier 2019). In San Francisco, the MTC provides its own definition of disadvantaged communities for the transportation field. In a recent report evaluating dockless bikeshare in San Francisco, the SFMTA uses the MTC definition when conducting equity analyses (MTC 2018). The MTC terms and identifies disadvantaged populations as “Communities of Concern (CoCs)” based on the 2012-2016 American Community Survey (ACS) 5-year tract-level data.

In Los Angeles, since different areas have their customized disadvantaged populations, we adopt a different definition of CoCs in Los Angeles, which is provided by the Southern California Association of Governments (SCAG) from its Plan Performance Environmental Justice Technical Report. In the report, SCAG investigated all Census Designated Places (CDPs) and City of Los Angeles Community Planning Areas (CPAs) and selected regions that fall in the top 33% of all communities in SCAG region for having the highest concentration of minority populations and low-income households (Southern California Association of Governments 2020). A person is classified as “minority” if the individual self-identified as one of the minority groups in the census (Table 4). SCAG performed poverty classification according to the income guidelines outlined by the U.S. Department of Health & Human Services.

Bikeshare data: The analyses require data from both bikeshare systems. However, as would be true in cities across the U.S., the data from the two bikeshare systems in San Francisco/Los Angeles are not in the same format. Currently, Motivate and B-cycle operate most of the dock-based bikeshare systems in the US. Among all of these dock-based systems, Motivate operates Citi Bike (New York), Divvy (Chicago), Capital Bike Share (Washington DC), Ford GoBike (Bay Area), Biki (Honolulu), and Bluebikes (Greater Boston) which together contributed over 80% of all dock-based bikeshare trips in 2018 (NACTO 2019). All of the dock-based bikeshare systems operated by Motivate and B-cycle provide trip data, including information about trip start day and time, end day and time, start station, end station, bike id, and rider type (annual member or day pass user). For the members’ trips, the database also includes the riders’ gender and year of birth. However, the operators do not provide information on bike availability or rebalancing activities. We also found that these limitations exist for all dock-based bikeshare systems in the US because, currently, all operators only provide trip-based data, not operational data.

On the other hand, there is no trip data for dockless bikeshare systems; companies have not shared the data in this form because of privacy concerns or commercial advantages. However, they provide information through the General Bikeshare Feed Specification (GBFS), which is an open data standard for bikeshare. The GBFS provides real-time bike information (including bike id, location, battery level, and service status), and the number of available bikes in available hubs in a city. Unfortunately, the standard does not provide the bike id when the bike is at a hub. Additionally, the real-time bike data does not include any user data. Currently, many cities have required dockless bikeshare companies (e.g., JUMP, Bird, Lime, Lyft, Skip, Spin) to share real-time data in GBFS format. If a dockless bikeshare company provides data as required, it will be information in this format and available through an application programming interface (API). We developed a web-scraping (web data extraction) tool for the systematic and continuous collection of the real-time information from GBFS (e.g., JUMP Bike). Despite its limitations, the GBFS is very useful, and we developed a robust framework based on reasonable assumptions to infer other bikeshare data (e.g., bike availability and rebalancing operations) to support our analyses.

For the study, we use historical bikeshare trip data provided by Ford GoBike; their database includes all bikeshare trips between 2013 and 2019. JUMP Bike (dockless), as mentioned, does not directly provide historical bikeshare trip data, thus, we use the web-scraping tool to gather minute-by-minute data from January to March 2019. We use this three-month sample data because, by March 2019, JUMP Bike had already been operating for over one year; thus, users were familiar with the service. Moreover, although there are some declines in bike ridership in San Francisco at the end of the year, ridership (based on data from bike counters) does not significantly fluctuate throughout the year (T. Winters 2017). For example, during 2018 the average number of bike counts (at the available bike counters) between January and March was 15,385 per month, which is 93% of the overall monthly average of 16,533. We found similar trends when analyzing the monthly trip numbers for Ford GoBike in San Francisco. The average number of monthly dock-based bikeshare trips between January and March was 210,598 in 2019, which is 3% above the average monthly usage (204,063 trips). As these findings were an indication that the three-month sample collected was representative, the analyses compare both systems within this period (January to March of 2019).

Online suggestion data: The bikeshare operation companies (Baywheels in San Francisco or Metro Bike Share in Los Angeles) develop public online portals where users can suggest potential bikeshare station locations and comment on existing ones (Metro Bike Share 2021; Baywheels 2021). In this portal, users placed a dot (suggested location) on the maps of San Francisco and Los Angeles, providing a detailed description of their reasons (e.g., home, work/school, shopping, and fun) for choosing the locations, and their home zip codes. We applied the web scraping technology to download those suggestion data from these online portals. In San Francisco, we collected the historical information in the portal until the end of 2020. However, the data contained a number of duplicate locations by the same users and test data, which were removed. In the end, there is a total of 721 records of new bikeshare station suggestions available in San Francisco. In Los Angeles, after removing duplicated data, there are 2,354 suggestions of new bikeshare stations.

Please refer to the data method section