Correcting temporal bias in mobility data using time-use surveys

Abstract

GPS mobility data is a valuable source of behavioral measurement which is subject to systematic biases including the over- or under-representation of demographic groups, and variations in the quality of location sampling across time. In this paper, we address the challenge of temporal bias in mobility data, which can skew the representation of mobility behaviors due to the event-based nature of location data sampling. We use the American Time Use Survey (ATUS) to assess the accuracy of a place-based measure of economic segregation drawn from large-scale mobility data across 11 U.S. cities. We show that comparisons with high quality time use surveys such as the ATUS can validate behavioral insights from mobility data, while quantifying uncertainty and highlighting areas of relative instability in analytical findings. We also propose a temporal re-weighting method that can complement existing bias-mitigation techniques to improve the accuracy of conclusions drawn from GPS-based mobility data.

Figures (10)

• Figure 1: Temporal differences between ATUS and mobility data reveal temporal bias in mobility data-based time-use estimates.a) Mobility data captures visitation by individuals in different income quartiles to the same POI category. b) Temporal bias indicates different quantities of time recorded by mobility data and ATUS for a specific POI category (shaded areas). The difference between Mobility Data and ATUS reported time indicates over- or under-representation by mobility data for individuals in a given income group. Panel b shows visitation by high income (Q4) individuals to Food & Coffee POIs in Boston.
• Figure 2: Comparison of total time spent per hour shows temporal differences between income quartiles in ATUS and mobility data between 6 a.m. and 11 p.m.a) Cumulative time spent by hour for 2,502 ATUS respondents from 11 cities, by income quartile. b) Cumulative time captured by hour for each income quartile from mobility data users in Boston (for results in other cities, see Supplementary Figures \ref{['fig:figs1']}--\ref{['fig:figs3']}). c) Ratio between ATUS and mobility data by hour for each income quartile. The red dotted line is at the mean. Hours above the mean line show when visit type is under-represented in mobility data while hours below the mean show over-representation. Hours without a ratio occur when no ATUS respondents from that quartile recorded spending time at that POI category within that hour.
• Figure 3: ATUS / MD ratio and minutes per quartile result in smaller or larger shifts in segregation after re-weighting. Two Food & Coffee POIs in Boston at 12 p.m. which exhibit different changes in segregation after re-weighting.
• Figure 4: Average income segregation decreases for Food establishments but increases for Grocery stores and Gyms. Comparing average income segregation by metro-area before and after re-weighting with ATUS / MD ratios for each POI category. Significance codes: 0 '***' 0.001 '**' 0.01 '*' 0.05
• Figure 5: Magnitude of difference between income segregation pre- and post- ATUS-weighting throughout the day. Hourly comparison of original and re-weighted income segregation, averaged across POIs from all 11 study cities.