TL;DR
This paper introduces a novel data-driven approach to identify "mobility of care" travel patterns, which are disproportionately undertaken by women, using transit smart card data. By combining origin-destination inference, name-based gender inference, and geospatial analysis, the study reveals that women are significantly more likely to engage in these care-related trips, offering valuable insights for transit agencies to better understand and address gender equity in their services.
What Problem Does This Paper Address?
Existing research consistently shows gender-based differences in public transit usage, largely attributed to women's disproportionate share of household and caregiving responsibilities, a concept termed "mobility of care." While previous studies relied on surveys and qualitative data, these methods are time-consuming and expensive for large-scale analysis. This paper addresses the gap in understanding these gendered travel behaviors at a granular, individual level using readily available automated data. The challenge is to move beyond self-reported data to observe actual travel behavior and quantify the extent to which women engage in mobility of care, which is often hidden within broader trip categories like shopping or escorting.
Methods
The authors developed a novel data-driven workflow that integrates several key components:
- Origin-Destination and Transfer (ODX) Inference: This technique reconstructs individual trip details (boarding, alighting, routes, etc.) from raw smart card tap data.
- Name-Based Gender Inference: Using a commercial API (Genderize.io) and the US Social Security Baby Names database, the first names associated with registered SmarTrip cards are used to infer a likely gender. This is a crucial step to overcome the lack of explicit demographic data in fare card systems.
- Geospatial Analysis: This involves identifying "Places of Interest" (POIs) such as daycares, schools, and grocery stores. A route-direction-based algorithm is used to find the nearest transit stops to these POIs, allowing for the identification of trips potentially related to mobility of care.
- Accompaniment Analysis: This mode-agnostic approach identifies patterns of two SmarTrip cards tapping in close succession, particularly when one card is associated with a student, senior, or disabled fare product, suggesting a caregiving or accompanying role.
Datasets
The study utilizes data from the Washington Metropolitan Area Transit Authority (WMATA) for the first quarter of 2019. This includes:
- Origin-Destination and Transfer (ODX) data: Millions of individual journey stages and millions of unique registered SmarTrip cards.
- Registered SmarTrip Card data: Includes basic contact information like first name, used for gender inference.
- DC Open Data repository: Used to identify the geographical locations of daycares, schools, and grocery stores.
Key Insights and Ideas
- Data-Driven Mobility of Care: The study demonstrates that automated transit data, when combined with sophisticated inference techniques, can effectively identify and quantify "mobility of care" travel patterns, moving beyond traditional survey methods.
- Gendered Travel Patterns are Real and Measurable: The research provides strong evidence that women are significantly more likely to engage in mobility of care trips, even when controlling for other factors and using a balanced sample.
- Accompaniment as a Proxy for Care: The accompaniment analysis highlights that specific fare products and tap-in patterns can reliably indicate caregiving roles, offering a complementary method to geospatial analysis.
- Policy Implications for Equity: The findings underscore that seemingly gender-neutral transit policies can have unintended impacts on women. Understanding these patterns is crucial for designing more equitable transit services.
Key Results
- Gender Parity Deviation: In an equally sampled subset of users, women showed a 10%-15% higher likelihood of conducting mobility of care trips compared to men, particularly at stops near daycares, schools, and grocery stores.
- Accompaniment Patterns: Women were found to be significantly more likely to exhibit consistent accompaniment patterns with riders who are children, elderly, or people with disabilities.
- Daycare and School Trips: An average of 10% more women than men trip-chained at stops near daycares during AM drop-off times. Similar trends were observed for school stops.
- Weekend Travel: Overall, significantly fewer women were observed on the system on weekends, with a decrease of approximately 10%.
- Accompaniment Analysis: Approximately 25% of trips identified through accompaniment patterns were also classified as mobility of care trips via the geospatial analysis.
Limitations and Future Work
- Registered User Bias: The study acknowledges that the registered SmarTrip card data may over-represent full-time employees with typical commute patterns, potentially skewing results. Future work should aim to capture a wider array of individuals, including those who do not register their cards.
- Name-Based Gender Inference Limitations: While validated, name-based gender inference can have lower accuracy for certain demographic groups (e.g., Black and Asian individuals, women). The authors emphasize that this analysis does not seek to perpetuate stereotypes and that gender identity is complex.
- Inferring Intent: The study infers mobility of care based on proximity to POIs and accompaniment patterns, but it cannot definitively confirm the intent of every trip.
- Future Research: The authors suggest further research to capture a wider array of individuals, explore the impact of income and employment status, and investigate how transit agencies can leverage these insights to mitigate biases and improve services for all users. The relevance of these findings is amplified in the post-COVID era, where transit agencies need to consider all trip purposes for financial sustainability.
Summary updated