Abstract

Mobile phone data (MPD), including call detail records (CDR), can help Senegal close persistent gaps in its statistical system by adding frequency, geographic detail and a repeatable data pipeline. They complement official statistics and do not replace them. This abstract draws on the World Bank proposal to the Global Data Facility Window and on international benchmarks.

The proposal's central argument concerns timing. The census gives a detailed but static snapshot once a decade, which quickly becomes outdated in fast-growing, urbanizing areas. Dakar, Thiès and Diourbel hold almost half of Senegal's population (22%, 13% and 12%). MPD would supply higher-frequency information on internal, incoming and outgoing movements. These would be summarized as origin-destination matrices and tracked against seasonal, economic and cultural events. The National Statistics Agency (ANSD) would be the primary beneficiary. CETUD, the Dakar urban transport regulator, would be the end user.

Senegal has several strengths:

Sonatel/Orange holds about 60% of the mobile market, with near-universal network coverage.
ANSD has a record of statistical innovation, from the Listen to Senegal call-center panel to the 2023 GUISSTANN pilot, which predicts socioeconomic indicators from CDR.
A 2024 African Development Bank diagnosis recommended a Data Innovation Lab at ANSD.
Calibration against the 5th census and household surveys addresses the main validity risk.
The proposal also acknowledges constraints: limited access to MPD, a regulatory framework not yet aligned with big data, and low user capacity. Much of the grant's value is therefore institutional. It would deliver an MoU between ANSD and the operators, a sustainable data pipeline and technical skills.

International benchmarks support the approach. The Gambia is the closest precedent. Its CDR pipeline was built with the regulator and the statistics bureau. Phone-user counts by district tracked census population patterns, and weekly movements responded to the 2020 lockdown and Ramadan. Mobility could also be linked to ward-level poverty. In Sri Lanka, CDR supplemented the census with high-frequency population distribution. Research by Blumenstock and co-authors showed that CDR-based poverty estimates can approximate survey results faster and more cheaply. Togo's Novissi program used machine learning on CDR to prioritize the poorest people. Senegal's proposal dropped welfare targeting from its main component, so it remains a possible later phase.

Important limits remain. CDR covers only phone users, so bias by income, gender, age, location, multi-SIM use and customers of other operators must be tested. Big data works best when ground-truthed against surveys. Sonatel will share processed results and not raw data, and stronger data protection rules may complicate access. This leaves open how to audit the anonymization, reproduce the methods and scale beyond one operator. Durable, country-owned arrangements are needed, and the link to the HISWACA-SOP1 project is intended to provide them.

To strengthen the case, the proposal could:

List the specific missing indicators and the decisions CETUD cannot yet make.
Add Gambia-style validation against census density, by region and demographic group.
Position the use case as a step toward updated sampling frames and later poverty estimation.
Explain how single-operator data will be reweighted or complemented.