Big data in the time of COVID-19
The COVID-19 pandemic has highlighted the value of high-frequency, localized data in inferring the economic impact of shocks to inform decision making. This includes the use of publicly available nightlights, air quality, and aggregated location data. Since tracking mobility is crucial during a pandemic, researchers have worked with governments and mobile network operators to leverage Call Detail Record (CDR) data, aggregated and anonymized cell-phone data. CDR data is used to understand migration patterns, create poverty maps, and estimate household’s economic characteristics. However, CDR data raises privacy concerns, requiring strict protocols restricting access and ensuring confidentiality. In developing countries, the lack of capacity and data access policies can constrain the creation of durable data pipelines linking data availability, analysis, and policymaking.
This blog outlines how we overcame these constraints to build a durable CDR data pipeline in The Gambia,[1] and used this data to derive insights into the effects of the COVID-19 outbreak on population dynamics. We worked in partnership with the national regulator Public Utilities Regulatory Authority (PURA) and The Gambia Bureau of Statistics (GBOS). As part of its oversight activities, PURA collects aggregated indicators from telecom activities to monitor service quality. With technical support from the World Bank and the University of Tokyo, PURA has worked with mobile phone operators to expand the list of indicators routinely collected for the purpose of mobility analysis and store them in a secure on-site server. A first set of these indicators was collected between March and May 2020, offering a snapshot of changes in mobility during the COVID-19 lockdown. The data analyzed is from two of the four telecom operators, together constituting 70% of market share, and includes 2 billion individual anonymized call records.
Stylized fact 1: CDR data are a valid instrument to monitor population density and dynamics
A necessary validation when using CDR data is whether it is representative of the population. To test this assumption, we plot the known population density for each district against the density of unique phone users as defined by their International Mobile Equipment Identity (IMEI) prior to the confinement order (March 22, 2020).[2] We confirm a highly correlated relationship in both significance and magnitude, with population density as computed by WorldPop and in the most recent population census.[3] This validates the assumption that IMEI is a valid proxy for population density, and that tracking shifts in IMEI can therefore offer insights into short-term and long-term population movement dynamics.
Validation of CDR data against known population density
Source: Author’s calculations based on CDR data from The Gambia, March to May 2020.
Stylized fact 2: CDR data reveal a shift in population movement
CDR data allows us to track when unique cell-phone users travel across administrative boundaries, quantifying mobility as the inflow and outflow. Upon the imposition of a nationwide lockdown on March 22nd, we observe a substantial out-migration from the densely populated coastal urban areas into rural areas inland as people leave the capital city region. This movement intensifies during the week of April 19-25th, which coincides with the beginning of Ramadan, likely because people visit relatives in rural areas. This pattern of out-migration continues until mid-May, when we start seeing people gradually trickle back into the capital region as Ramadan ends and economic activities resume. Findings show how population movement dynamics shift on a weekly basis.
Changes in weekly population inflow across time and districts
Source: Author’s calculations based on CDR data from The Gambia, March to May 2020. Note: Green: greater inflow relative to baseline; Lila: lower inflow relative to baseline.
Stylized fact 3: CDR data can be linked to spatially aggregated indicators from survey data
Did these disruptions differentially affect different populations? While we cannot match aggregated CDR data to household level well-being, we can match movement patterns with district- or ward-level information on poverty from the national household survey (2015). When we do so, we notice that the wards with the highest proportion of poor households see the largest disruption to their mobility in terms of both inflow and outflow relative to wards with the lower proportion of poor households. Mobility dropped by as much as 25%, particularly during Ramadan period, with some small spikes before and after Ramadan, and then recovered gradually. This result is intriguing as it suggests that the poorest wards suffered the most from the lockdown, perhaps because they rely on economic activities that were most affected by social distancing policies (informal trade, markets, etc.).
Changes in weekly population inflow across wards with highest and lowest poverty rates
Source: Author’s calculations based on CDR data from The Gambia, March to May 2020.
Smart containment: Data-driven decision making in a resource constrained environment
These insights highlight that real-time data and analysis are valuable, but only when produced in close collaboration with counterparts. In a context of constrained capacity, the population movement patterns outlined under stylized fact 2 could inform targeted testing and tracing initiatives, by concentrating efforts in areas of high mobility. When a full lockdown is not possible given the economic costs, this can also inform where constraints on mobility should be enforced given higher risks of transmission. Further insights can be gained by supplementing with additional data on communities and households. Stylized fact 3 highlights that the lockdown disproportionally affected poorer districts, and relief and recovery efforts should therefore aim to address these inequities. These are but a few of the policy relevant insights CDR data can deliver.
Rather than aim for shortcuts, it is important to invest in a strong relationship with the relevant regulators in order to ensure sufficient buy-in. The process of collecting this data was as informative as the results. Lessons learned include:
- Build consensus among all stakeholders and use strategic alliances/champions embedded in country to foster ownership and sustainability.
- Address technical challenges and invest in the necessary capacity and data systems up front.
- Work within institutional parameters, accessing data collected by the private sector through the national regulator to ensure compliance and encourage system building.
Building on these early successes, the team intends to expand the scope and scale of its collaboration. This includes dissemination to key stakeholders, as-well as capacity building to host and process large monthly CDR datasets. In terms of analysis, overlaying CDR data with existing surveys can deliver valuable insights. Given the importance of migration to the economy, mobility data can be used to better understand where migrants are moving to, both internally and abroad. For example, early results suggest areas with a high percentage of international calls also have a large concentration of families with international migrant workers.
This blog benefited from guidance from Tara Vishwanath under the Global Solution Group “Welfare Implications of Climate Change, Fragility and Conflict Risks” in the Poverty and Equity GP in the World Bank. This project received support from the Trust Fund for Statistical Capacity Building III (TFSCB-III) and the World Bank DEC-DIME Group, and was implemented in close collaboration with PURA, GBoS, and the University of Tokyo. TFSCB-III is supported by the United Kingdom’s Foreign, Commonwealth & Development Office, the Department of Foreign Affairs and Trade of Ireland, and the Governments of Canada and Korea.
[1] The Gambia is a West African country of 2.4 million people, where tourism and remittances from abroad are among the key drivers of the economy Gambia is on the west Atlantic coast, surrounded by Senegal. The capital city region at the mouth of the Gambia river encompasses Banjul City and the Kanifing region, with tourist resorts strung southwards along the coast. Inland is largely rural, its economy driven by agriculture and often dependent on the flow of domestic and international remittances. These disparities in access to services and opportunities have led to high levels of internal migration, especially among the young who left rural areas to look for better jobs in the capital city region.
[2] Deville et al. (2014) showed that a strong log-log relationship between the density of unique users in a cell tower’s catchment area and population density can validate the use of CDR data as proxy for population movement (Deville, Pierre, et al. "Dynamic population mapping using mobile phone data." Proceedings of the National Academy of Sciences 111.45 (2014): 15888-15893.)
[3] The R2 is much higher for the WorldPop model, perhaps because as a more recent estimate based on geospatial indicators, World Pop better captures population movements between 2013 and 2020 and census data is no longer representative. This suggests the potential to use CDR data to update sampling frames and inform policy interventions which are sensitive to population numbers.
Join the Conversation