Tracking migrants is a challenging task that policymakers, researchers and humanitarian organizations have long struggled to achieve. The data they use is often outdated, incomplete and inconsistent and this hampers efforts to understand movement and respond effectively to global crises and economic changes.
Now that looks set to change thanks to the work of Guanghua Chi at Facebook-owner Meta in Menlo Park, and colleagues, who have used anonymized data from three billion Facebook users to estimate monthly migration flows across 181 countries.
“Our estimates closely match high-quality measures of migration where available but can be produced nearly worldwide and with less delay than alternative methods,” they say.
Global Movement
The problem the team addresses is simple to state but hard to solve: to measure the long-term movement of people accurately, in near-real time, on a global scale. Traditional methods rely on censuses or administrative records, which are often collected irregularly and inconsistently across countries. Some recent innovations have tapped into cell phone data or social media posts, but these efforts are usually geographically limited. The Facebook-based approach circumvents these limitations, providing a new standard for measuring human movement.
Their method uses the United Nations' recommended definition of migration, counting only those who settle for at least a year in a new country. The team built their estimates by first predicting each user's home country based on signals like self-reported location and IP addresses. They then detected long-term changes in residence, applying a segment-based algorithm designed to minimize noise and maximize alignment with standard migration definitions. After identifying migration events, they aggregated them monthly and weighted the data to reflect population-level flows.
Of course, Facebook users aren't perfectly representative of the global population. Wealthier individuals, for example, are often more likely to use Facebook and more likely to migrate, particularly in developing nations.
To address this, the researchers developed a weighting system that corrects for these imbalances. Their selection model accounts for country-level Facebook usage rates and income levels by adjusting the raw counts to better reflect population-level migration flows. Finally, they add a small amount of statistical noise to protect individual user privacy while preserving the overall trends.
This adjustment ensures the resulting migration figures mirror real-world flows rather than the quirks of social media usage. "Our estimates improve substantially on existing figures, allowing us to estimate monthly migration flows between 181 countries by drawing on data from more than 3 billion individuals," say the researchers.
To validate their approach, Chi and co compared their estimates with high-quality official statistics from various countries and regions. When benchmarking against New Zealand’s immigration statistics, for example, their model achieved a near-perfect match with a correlation of 0.98.
Their data reveal fascinating trends. In 2022, they estimate that 39.1 million people migrated internationally among the 181 countries included in their study, about 0.63% of the population sampled. The United States led the world in net migration gains, with 3.92 million (841,200 emigrants versus 4,109,400 immigrants. While Ukraine suffered the largest net loss of 2.34 million (66,600 immigrants v 2,402,100 emigrants).
Global migration plummeted by 64% during the COVID-19 pandemic, reflecting widespread travel restrictions, before rebounding by 2022 to rates 24% higher than pre-pandemic levels.
Crises Capture
The data also vividly captures the human consequences of geopolitical crises. Following Russia's invasion in February 2022, emigration from Ukraine surged tenfold compared to pre-war levels, with an estimated 2.3 million people settling elsewhere for at least a year by December 2022. The primary destinations identified – Poland, Germany, Czech Republic, the US, and UK – align closely with UNHCR figures, though the team notes their stricter migration definition results in lower counts than figures based purely on refugee registrations or temporary protection statuses.
Similar spikes were observed following Hong Kong's passage of a contentious security law in 2020 (migration to the UK increased fifteenfold shortly after) and the 2021 coup in Myanmar. “We observe that crises can lead to dramatic changes in migration," say the researchers, highlighting the utility of their dataset for tracking real-time responses to global events.
Interestingly, the team found that migration patterns reflect economic disparities and social networks. Wealthier countries attract a disproportionate share of migrants, while migration between neighboring countries is far more common than between distant ones. "Distance is a strong predictor of the rate of migration between countries," they note, emphasizing that proximity, economic opportunity, and existing diaspora communities all shape global flows.
Migration also tends to follow a steplike pattern with migrants from poor countries moving to middle income countries and migrants from there moving to wealthy countries.
That’s important work with numerous applications. Real-time, high-resolution migration data could transform fields ranging from economics and sociology to disaster response and urban planning. Governments should also be able to tailor immigration policies better, humanitarian organizations could respond faster to crises, and researchers could build more accurate models of human mobility.
To that end, the team are releasing their dataset publicly through the Humanitarian Data Exchange, ensuring broad access for future work. This is an important example of the kind of transparency that other global tech companies should aim to repeat. There is clearly much low hanging fruit in the enormous global databases tech companies are compiling. How they release it, if at all, should be part of an important public debate.
Ref: Measuring Global Migration Flows using Online Data: arxiv.org/abs/2504.11691
The New York Times has also published an interactive tool for visualizing this data.
