Socioeconomic disparities in mobility behavior during the COVID-19 pandemic in developing countries

Abstract

Mobile phone data have played a key role in quantifying human mobility during the COVID-19 pandemic. Existing studies on mobility patterns have primarily focused on regional aggregates in high-income countries, obfuscating the accentuated impact of the pandemic on the most vulnerable populations. Leveraging geolocation data from mobile-phone users and population census for 6 middle-income countries across 3 continents between March and December 2020, we uncovered common disparities in the behavioral response to the pandemic across socioeconomic groups. Users living in low-wealth neighborhoods were less likely to respond by self-isolating, relocating to rural areas, or refraining from commuting to work. The gap in the behavioral responses between socioeconomic groups persisted during the entire observation period. Among users living in low-wealth neighborhoods, those who commute to work in high-wealth neighborhoods pre-pandemic were particularly at risk of experiencing economic stress, facing both the reduction in economic activity in the high-wealth neighborhood and being more likely to be affected by public transport closures due to their longer commute distances. While confinement policies were predominantly country-wide, these results suggest that, when data to identify vulnerable individuals are not readily available, GPS-based analytics could help design targeted place-based policies to aid the most vulnerable.

Figures (5)

• Figure 1: Inferring location type and time use from GPS trajectories.(a) Trajectory of a hypothetical mobile phone user over one day. A stop event (dotted circle) characterizes a location where a user spends at least 5 minutes within a 25-meter distance. The color of a stop event corresponds to the time of the day when the event occurs, and lines connect two consecutive stop events. (b) Stop events are spatially clustered together to form a stop location. (c) Stop locations are then labeled as home, workplace, or other based on how frequently they are visited during the observation period and the hours of the day at which a visit occurs. (d) Final structure of the user location data, which consists of a non-continuous time series of labeled stop events. The background image shows a portion of Mexico City where blocks are colored by the level of a wealth index constructed from census data, illustrating the granularity of our data in urban areas.
• Figure 2: Change in the share of users self-isolating at home by socioeconomic group. Each panel shows the relative change in the share of active users staying at home over the entire course of a day relative to the pre-pandemic period for the six countries studied, conditioning on the wealth of their primary home administrative unit. Shaded areas indicate the standard errors of the mean computed pulling together the self-isolating share of all administrative units of a single country. In countries where many administrative units are home to users, the standard error becomes very small due to the higher number of geographical areas included in the computation.
• Figure 3: Net share of urban users relocating to rural areas by socioeconomic group. Each panel shows the percentage change in the difference between the number of users relocating from an urban to a rural area and those moving from rural to urban, for different socioeconomic groups in the six countries under study. Results are normalized to remove pre-pandemic relocation flows. We also report the stringency index of containment policies in each country over time (green line). The values of the stringency index of containment policies are reported on the y-axis on the right of each panel. Across all countries, users whose primary home is located in a high-wealth neighborhood (red line) were more likely to relocate to a rural area than those living in a low-wealth neighborhood (blue line).
• Figure 4: Change in the fraction of users not commuting by socioeconomic group. Focusing on the 26% of users with a work location during the observation period, we illustrate the percentage change in the number of users who are not commuting, conditioning on their wealth group classification. Users in the high-wealth group (solid red line) were more likely to stop commuting than those in the low-wealth group (solid light-blue line). We then restrict to users from the low-wealth group and measure their changes in commuting patterns conditioning on the wealth of their workplace. Users living in low-wealth neighborhoods who used to work in high-wealth neighborhoods pre-pandemic (dashed violet line) were more likely to stop commuting than those who used to work in low-wealth neighborhoods (dashed grey-blue line). The shaded region highlights the standard errors of the mean computed pulling together the share of users not commuting to their workplace from all administrative units of a single country. The green line shows the stringency of containment policies in the corresponding countries over time. The values of the stringency index of containment policies are reported on the y-axis on the right of each panel.
• Figure 5: Policy restrictions and commuting patterns across socioeconomic groups. a) We present the estimated coefficients for policy restrictions, $c_{i,n}$, modeling the propensity of users from different socioeconomic groups to suspend commuting. Errorbars report the $95\%$ confidence intervals for the estimated coefficient values. b-d) Distributions of the distance between home and workplace across socioeconomic groups, b) comparing users living in low- versus high-wealth neighborhoods, then c) focusing on users living in high-wealth neighborhoods and comparing those working in low- versus high-wealth neighborhoods, and d) focusing on users living in low-wealth neighborhoods and comparing those working in low- versus high-wealth neighborhoods.