Time Profile of U.S. Neighborhoods: Datasets of Time Use at Social Infrastructure Places

Yan Wang, Ziyi GuoarXiv 2025 · DOI · arXiv

This paper introduces a first-of-its-kind national dataset that tracks how much time Americans spend at "social infrastructure" places—the parks, cafes, libraries, and churches that form the backbone of community life. By leveraging anonymized mobile device data from 2019 to 2024, researchers have created a high-resolution map of human engagement that bridges the gap between urban planning and public health.

TL;DR

The authors developed a scalable framework and dataset for Social-infrastructure Time Use (STU) across the continental U.S. at multiple geographic scales. By analyzing mobile foot traffic, they reveal how neighborhood characteristics, urban-rural divides, and socio-demographics influence the "time wealth" or "time poverty" of communities.

Addressing the Spatially-Resolved Data Gap

Urban researchers have long been able to map accessibility—how far you live from a park or a grocery store. However, they have struggled to measure actual engagement—how much time people actually spend there.

Existing tools like the American Time Use Survey (ATUS) provide deep insights into what people do, but they lack the spatial resolution to tell us where these activities happen at a neighborhood level. Conversely, standard GPS datasets often focus on "where" but ignore the "how long." This paper addresses this gap, allowing researchers to see if people in specific neighborhoods are actually utilizing the social infrastructure available to them.

Methodology and Categorization

The researchers utilized anonymized, aggregated foot traffic data from Advan, covering roughly 32 million mobile devices (about 15% of the U.S. phone-using population). A critical step involved mapping these mobile "pings" to specific types of social infrastructure based on NAICS sectors.

Mapping of activity categories to NAICS sectors for social infrastructure identification.

Key methodological innovations include:

  • Bucketed Dwell Time Processing: The raw data categorizes visits into time "buckets" (e.g., 11–20 minutes). The authors used the median of these buckets to estimate total time spent.
  • Normalization for Panel Bias: The authors normalized the data by neighborhood panel sizes to ensure metrics represent "average time per user" rather than raw volume.
  • Multi-Dimensional Metrics: Four primary measures were introduced: Per User STU (weekly time per resident), Per Visit STU (duration of a single visit), Shannon’s Diversity Index (variety of activities), and the Gini Index (inequality of social time distribution).

The STU Dataset

The study integrates several massive data sources, including Advan Weekly Patterns (2019–2024), the American Community Survey (ACS) for socio-demographics, and the ATUS for validation. The resulting STU Dataset provides weekly time-use data for 73,642 census tracts, 3,242 counties, and 939 metropolitan areas.

Key Insights: Diversity, Equity, and Third Places

  • Log-Normal Distribution: Neighborhood time use follows a log-normal distribution; while most neighborhoods have baseline engagement, a "long tail" shows exceptionally high interaction.
  • Rising Engagement: Median time use increased from 2019 to 2023, suggesting a growing trend in utilizing social infrastructure post-pandemic.
  • The Diversity-Equity Link: Regions with a more diverse set of social infrastructure types (arts, sports, dining, etc.) tend to have a more equitable distribution of time use among residents.
  • "Third Places": The data confirms that grocery stores and dining establishments function as vital "third places"—locations that are neither home nor work but are essential for social friction and community cohesion.

The analysis revealed significant disparities in how social time is spent across the country, highlighting an urban-rural divide and socio-demographic correlations.

Spatial distribution of Per User STU across the United States, Florida, and Jacksonville.
  • Urban-Rural Divide: Urban neighborhoods showed significantly higher Per User STU compared to rural areas, highlighting a "social time" disparity.
  • Socio-Demographic Correlations: Neighborhoods with higher percentages of bachelor’s degrees showed a strong positive correlation (0.30) with time spent in sports and exercise. Conversely, higher percentages of Black residents correlated with lower total Per User STU in leisure activities, consistent with historical data regarding racial disparities in "time wealth."
  • Validation: The mobile-derived data showed over 90% alignment with official population figures at the county and state levels, proving the dataset's robustness.

Limitations and Future Directions

While powerful, the dataset has limitations. Anonymized data makes it difficult to distinguish between visitors and employees (e.g., a barista vs. a customer). Additionally, mobile panels may underrepresent the elderly and very low-income households. Future work aims to link this "social time" data to health outcomes like longevity and mental health to explore its potential in public health prescriptions.

Summary updated

BibTeX
@misc{wang2025time,
  author = {Yan Wang and Ziyi Guo},
  title = {{Time Profile of U.S. Neighborhoods: Datasets of Time Use at Social Infrastructure Places}},
  year = {2025},
  eprint = {2508.13295},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2508.13295},
}
Figure