Comparing Twitter and LODES Data for Detecting Commuter Mobility Patterns

Transportation policy decisions have historically relied on understanding commuting patterns, the journey-to-work. While scholars often acknowledge the value inherent in examining commuting patterns holistically by considering both work and non-work trips, the difficulty of assembling data about non-work trips has restricted their inclusion in regional/large scale analyses. The overarching goal of this research project is to examine the feasibility of using geo-referenced tweets and Open Street Map (OSM) as proxy measures to understand and explain a robust range of commuting behaviors and spatio-temporal movement patterns.

Specifically, the research team uses LEHD Origin Destination Employment Statistics (LODES), a dataset created by the US Census Bureau to develop a baseline of movement patterns in the nine-county San Francisco Bay Area region. The LODES dataset provides spatial distributions of workers' employment and residential locations and the relation between the two at the Census block-level. Working with our research partners at the University of Salzburg and Hunter College, CUNY, we extract commuter patterns from social media data (in this case, geo-referenced tweets). We use semantic clustering methods (interpreting the text of the tweets themselves) to identify the purpose of the trip. We compare Census-derived data with social media-derived data in order to identify spatial movement clusters. In other words, the research team uses social media data to explain movement patterns that fall outside of conventional journey-to-work trips. The incorporation of address-level points-of-interest data derived from OSM provides an additional opportunity to characterize the trip purpose. In addition to the locational references, the social media data also contains date/time stamps that allow for a spatio-temporal analysis (weekday-weekend, seasonal) of movement patterns. The unique shutdown as a result of COVID-19 allows the research team an additional opportunity to examine non-work trips.

This research project investigates the benefits and limitations of using continuously and publicly available volunteered geographic information (tweets and OSM) to complement official census-derived data to conduct data-driven transportation policy research.