Gender Gaps in Urban Mobility

The study investigates whether and how daily urban mobility differs by gender and what social and infrastructural factors are associated with these differences. In many cities, women’s mobility is constrained by safety concerns, transport design, and gendered roles such as caregiving and household responsibilities. Traditional mobility surveys are limited in scale and frequency, and gender-disaggregated data are often lacking, hampering evidence-based, gender-responsive planning. Using large-scale mobile phone data for Santiago, Chile, the authors aim to quantify gender disparities in mobility and relate them to socio-demographic indicators and transport access, thereby informing policies to reduce gender gaps in urban mobility.

Prior work across social sciences and urban studies shows mobility is gendered, with women typically traveling shorter distances, chaining trips, spending more time traveling, and relying more on public transport and taxis than men. Traditional studies have largely used small-sample surveys and regression models, limiting generalizability and temporal updates. Over the last decade, mobile phone call detail records (CDRs) have enabled fine-grained mobility analyses and have been validated against survey-based origin–destination patterns and travel demand models. While some studies touched on gender differences using CDRs, evidence remains limited and often peripheral or small-scale. The present study extends this literature by applying gender-disaggregated CDR analysis at metropolitan scale and linking outcomes to socio-demographic context and urban affordances.

Data sources and study area: The study covers the Santiago Metropolitan Region (SMR), Chile, focusing on the urban area. Anonymized CDRs (calls only) from May–July 2016 (91 days) were analyzed: 2,148,132,995 events. User filtering retained phone numbers with: only 1 registered line; at least one call/day on average (≥91 calls total); identifiable home location; and more than two distinct visited locations. Final cohort: 418,624 users (51% female). Socioeconomic segment (GSE) available for 315,844 users.

Anonymization and privacy: Phone numbers were hashed (SHA-3) by the operator. Analyses were conducted on operator systems; only aggregated outputs were exported. Locations were spatially aggregated by rounding BTS coordinates to two decimal places, merging antennas into 726 regularly spaced tower clusters (~1 km resolution). Results were suppressed for cells with fewer than three unique users.

Spatial framework and home detection: More than 1300 BTS were clustered into 726 grid cells via coordinate rounding. User home was assigned as the most visited cell between 19:00–08:00 across the 91 days (time-constrained home detection), a method validated in prior work.

Socio-demographic and transport data: Census 2017 (INE) provided comuna-level indicators: employment and education gender ratios; general fertility rate; household typologies (couples, extended, family, single parent, single person). GSE categories (AB, C1a, C1b, C2, C3, D, E) were summarized into upper (AB, C1a, C1b, C2) vs lower (C3, D, E) groups; a GSE ratio was defined as the ratio of population below vs above a reference household income (~1M CLP/year). GTFS stop coordinates identified cells with access to public transport (Transantiago). Private vehicle access was proxied via census-based car ownership thresholds (>1 car per 4 residents vs lower).

Points of Interest (POIs): POIs were obtained from OpenStreetMap (amenity-tag nodes with ≥50 instances; 28 types) plus layers for malls (OSM polygons), metro stops (Observatorio de Ciudades), colectivos (shared taxi routes), and two synthetic layers: tower clusters (BTS locations) and a uniform grid. Total of 33 POI types.

Mobility metrics: For each user, four metrics were computed over the 3-month window: (i) N_l, number of distinct visited cells; (ii) \u007eN_l, number of ‘core’ locations covering 80% of calling activity; (iii) S, Shannon entropy of visit probability distribution across locations; (iv) r_g, radius of gyration quantifying travel distance scale. Metrics were then averaged by gender and by spatial unit (cell or comuna). Gender ratios at location l were defined as R_x = x_F/x_M for metric x.

POI visitation imbalance: For each POI type k, kernel density estimation (Gaussian kernel) produced a spatial POI density ρ_k(x), evaluated across varying bandwidth d (from sub-km to city scale). For user u with visited locations L_u, POI exposure ρ_k^u was the average of ρ_k(x) over L_u. Gender-aggregated POI densities p_k^F and p_k^M were computed by averaging ρ_k^u across female and male users, respectively. The gender density ratio r_k = p_k^F/p_k^M was analyzed as a function of d. Significance was assessed against two spatial null models: (1) spatial sampling—sampling N_k towers proportional to visit probability and perturbing locations (100 realizations to form 95% CI), and (2) spatial perturbation—±0.01 degree offsets applied to each original POI’s lat/long.

Statistical analysis: Estimation statistics quantified effect sizes and 95% confidence intervals for gender differences. Semi-partial Pearson correlations related comuna-level gender ratios (R_S, R_N) to socio-demographic indicators, controlling for gender differences in calling activity and sex ratio by comuna. Additional robustness checks included restricting to high-activity users, down-sampling men’s calls, temporal stability over 9-day windows, and excluding home/work and near-home visits in POI analyses.

• Citywide gender gap in mobility:

  • Men visited 30% more unique locations than women over 3 months: mean N_l: men 37.05 vs women 28.48; ΔN_l = 8.57 (95% CI [8.42, 8.72]).
  • Core locations (80% activity): mean \u007eN_l: men 6.92 vs women 4.79; Δ\u007eN_l = 2.13 (95% CI [2.09, 2.16]); a 45% increase for men over women.
  • Mobility diversity: Shannon entropy S lower for women by ΔS = 0.26 (95% CI [0.26, 0.27]), indicating women concentrate activity more in top locations. Rank-frequency shows women have higher visit probabilities at their top 1–2 locations; men dominate from rank ≥3.
  • Travel distance scale: women’s average radius of gyration is 1.09 km shorter than men’s (95% CI [1.07, 1.12]), implying more localized movement.
  • Robustness: Differences persist when restricting to high-activity users (gender differences increase) and after down-sampling men’s calls by up to 50%. Mobility metrics are temporally stable over 9-day windows; hourly call distributions by gender are similar.

• Socioeconomic gradients and spatial patterns:

  • Gender gaps widen with lower socioeconomic status (GSE): effect sizes increase from wealthiest (ABC1) to most deprived (E) groups across metrics (e.g., Δ\u007eN_l grows from 1.66 to ~2.50; ΔN_l from 7.28 to ~10.22; ΔS from 0.19 to ~0.31; Δr_g from 0.83 to ~1.36).
  • Comuna-level gender ratios R_S and R_N are strongly negatively correlated with the (log) GSE ratio (poorer areas): r = -0.59 and -0.53 (both p < 0.001). Wealthier comunas approach gender parity (R ≈ 1).

• Associations with census indicators (semi-partial Pearson, controlling for activity and distribution):

  • Positive correlates (higher values associated with smaller gender gap): employment gender ratio (R_S 0.51***; R_N 0.37**), HDI (0.42**, 0.37**), couples households (0.55***, 0.50***), single person households (0.56***, 0.44**).
  • Negative correlates (higher values associated with larger gender gap): general fertility rate (-0.53***, -0.40**), extended households (-0.61***, -0.57***), single parent households (-0.32*, -0.32*). Education gender ratios show weak/non-significant relationships.

• Transport access:

  • Public transport (GTFS stops) increases mobility for both genders but does not close the gap: women +0.76 locations (95% CI [0.51, 1.00]); men +1.39 (95% CI [1.05, 1.72]).
  • Among GTFS-accessible cells, socioeconomic inequalities persist: women in lowest-income quartile visit 1.53 fewer locations than those in highest quartile (95% CI [1.29, 1.78]); for men, the difference is <1 location (95% CI [0.51, 1.00]).
  • Higher car ownership correlates with higher N_l for both genders; socioeconomic differences are smaller for men than women in high car-ownership areas.

• POI-specific gendered visitation:

  • Significant female-skewed visitation near hospitals, malls, and taxi stands across intermediate spatial scales (around log10 d ≈ -2.5, ~300 m bandwidth), with r_k > 1 and exceeding null-model confidence intervals. Results are robust to alternative assumptions and to excluding home/work and near-home visits. Minimal spatial correlations among the most imbalanced POI layers.

The findings demonstrate substantial, systematic gender disparities in urban mobility in Santiago: women visit fewer distinct places, concentrate activity in fewer locations, and travel shorter distances than men. These disparities persist after controlling for calling behavior and across time windows, indicating they reflect genuine differences in mobility rather than data artifacts. Spatial and socioeconomic analyses show that the gender mobility gap is tightly linked to income and employment: wealthier areas approach gender parity, and higher employment equality corresponds to smaller mobility gaps. Household structures and fertility further align with larger gaps, consistent with caregiving and domestic responsibilities constraining women’s mobility. Transport access boosts mobility for all but is less equalizing for women, especially in lower-income areas, suggesting access alone is insufficient without addressing safety, affordability, time constraints, and design features. POI analyses reveal that gender differences extend to types of places visited, with women more frequently near hospitals, malls, and taxi stands, potentially reflecting caregiving roles and safety- or convenience-driven modal choices. Overall, the results answer the research questions by quantifying the gender mobility gap, mapping its spatial heterogeneity, and linking it to socio-demographic and transport factors, offering actionable insights for gender-responsive urban and transport policies.

This study shows that large-scale, gender-disaggregated CDRs can quantify gender gaps in urban mobility and relate them to socio-economic context and transport access. In Santiago, women’s mobility is more localized, less diverse, and involves fewer places than men’s, with gaps widening in lower-income areas. Employment equality and higher human development align with smaller gaps, while higher fertility and larger/extended households align with larger gaps. Public transport availability increases mobility for both genders but does not eliminate inequalities, particularly for low-income women. POI-based analyses indicate women’s trajectories more often include areas near hospitals, malls, and taxi stands, suggesting differentiated mobility needs and responsibilities. These insights can inform gender-responsive transport and urban planning. Future research should test generalizability across cities and rural settings, integrate higher-frequency mobile data (e.g., XDR), incorporate age and additional demographics, and further examine causal mechanisms and policy interventions.

Key limitations include potential sample biases (operator market share; user demographics not fully representative), reliance on calling activity to infer movement (activity varies by age/gender; age not available), and possible mismatches between subscriber-recorded sex and actual user. Spatial resolution is limited by tower clustering (≈1 km). The study cannot confirm trip purposes or actual engagement with nearby POIs; POIs are proxies for location characteristics. Although multiple robustness checks were performed and null models employed, unobserved confounders may remain. Privacy constraints limited access to raw individual-level data and variables such as age; results are based on aggregated analyses. Findings are observational and do not establish causality.