Predicting The Break: How Nations Can Get Ahead Of The Next Refugee Crisis

The flood of people into Europe took leaders by surprise. But with better forecasting systems, maybe the next crisis can be averted.

Predicting The Break: How Nations Can Get Ahead Of The Next Refugee Crisis
[Illustration: Made by Radio]

Europe’s leaders were so caught off guard by the refugee crisis when it first erupted in 2014 that the German city of Cologne–overwhelmed by the number of asylum seekers that November–bought a luxury tourist hotel for $7 million to house some of them. It would only get worse. The whole of Europe, in fact, was shell-shocked (and who wouldn’t be at the sight of Aylan Kurdi?).


The big question now, for governments, migrations researchers, and analysts, is: Can we do better next time?

Refugee flows are more complex than predicting the weather, but today it is possible to write programs that predict almost anything, from crime hotspots to dating matches to stock market movements. So predicting the movements of people–and particularly the moment when refugees leave or “break” from their homes and which direction they are going–shouldn’t be beyond the bounds of possibility.

In several respects, the crisis was predictable, and Europe should have been better prepared. The unchecked wars in Syria and Iraq were bound to produce large displacements of people, and, if experts had been looking, they contained clues and patterns of what would follow.

If they had been more proactive, analysts say, Europe’s leaders could have forestalled some of the fallout. Early warning of when refugees and migrants move, how many move, and where they might go, could have allowed aid agencies and governments to put resources where they were needed and avert an exodus before it starts. By sending more aid to Jordan and Turkey, for example, far fewer refugees might have made the dangerous trip across the Mediterranean or turned to traffickers to make it to Europe. Better information could have allowed for earlier aid, coordination among governments and the NGO community, and more sensible border policies, or even establishing a “safe zone” (an untried option in Syria).

Though aid agencies and NGOs have had early warning systems for conflicts and famines before now (the UN started calling for them in the late 1970s), they haven’t had the capacity to predict large displacements of migrants, according to Lisa Singh, a computer scientist at Georgetown University who is working on a predictive model that will do exactly this. The project Singh is helping develop is new for two reasons: It crunches large amounts of data from disparate sources, and two, it aims to have forecasting capability.

The search for trigger points

Singh’s research group works with Georgetown social scientists to process a database of 700 million newspaper articles, tweets, and survey data looking for clues and patterns in refugee flights. Their work combines historical data (the newspaper articles), “perception data” or how people feel about the political and economic climate (the Twitter data), and on-the-ground surveys of thousands of refugees from Syria and Iraq.


“The interviews allow us to assess the underlying economic, social, political, demographic, and environmental drivers of displacement and prepare a detailed timeline of events that have triggered actual movements,” says Susan Martin, Singh’s colleague, who oversaw the surveys.

By searching for “trigger points,” their model aims to offer an early warning of when refugees could be about to flee and a decision-assistance system for governments and aid organizations. Yet it’s not only important to understand when people move, but also how many people are moving and where they are going. The current rate of forced migration is high–1 out of every 113 people on earth, according to the UN refugee agency, UNHCR–but it could get worse because of climate change and pressure on resources in poor places. Environmental pressures, like droughts and flood, compound the destructive forces of tribal, civic, and inter-country conflict.

Major storms and flooding already cause more than 20 million people to leave their homes every year, according to the Internal Displacement Monitoring Center. And several reports predict greater displacement as climate change continues to create more–and more devastating–natural disasters. Fifty to 200 million people could be displaced by the impacts of global warming by 2050, according to forecasts.

To help predict the refugee flows from climate change, the Flowminder Foundation, in Sweden, is using mobile phone data to map migration movements. “There’s really no other way than to use mobile phones. Satellites can see whether a person is in a field, but not where that person is moving and the direction he’s going in,” says executive director Linus Bengtsson.

Real-time data after a disaster

Flowminder uses anonymized data from phone companies to track people from cell tower to cell tower, indicating large people movements. This allows movements after many disasters to become relatively predictable, Bengtsson says. After the 2010 Haiti earthquake, for example, Flowminder showed that many people who fled Port au Prince went in the same direction as they’d gone the previous Christmas–presumably to take refuge with their families. Separate research, published recently in Nature, showed that the number of people migrating between two places is highly determined by the distance involved, the size of population living between the two points, and the socioeconomic level of the people migrating.

“Based on these factors, and enough mobile and household data, it’s possible to come up with quite accurate numbers of how many people are going between those locations,” Bengtsson says. “In the event of a disaster, we have been able to respond with the close to real-time data in Haiti and Nepal.”


He’s cautious not to overstate the accuracy of the predictions. But it’s clear, a paper on the project says, that “people’s movements are highly influenced by their historic behavior and their social bonds . . . even after the most severe disasters in history.” Together with satellite imagery analysis, and perhaps drones and social media, it should be possible to track a lot about refugee flows in real time.

Of course, mobile-based analysis has flaws at the moment. If people have more than one phone it skews the results. And it’s hard to know if the people with phones are the ones you want to follow. There could more vulnerable people, including the elderly, the very young, and the very poorest, who don’t have phones, and are therefore missing from the data.

But this kind of data could be particularly useful for internal displacements within a country, within the range of single national cellular systems. Mobile phones are less useful for cross-border migration, Bengtsson says, where she would have to work with more than one provider to get the necessary data.

These are just two examples of projects in the works that use computer systems to improve human migration forecasts and reduce their human toll. There are many other efforts, from a USAID project that aims to predict when atrocities are going to occur to an effort to use algorithms to resettle refugees more efficiently by matching people to the best destination.

To Rana Novack, founder of the Refugee Admissions Network Alliance, it’s important to import the expertise of the business sector to help improve predictions. If, for example, Netflix is able to crunch a billion bytes of viewing data to come up with suitable movie choices, then it should be possible to build a system for migrant data. As an employee at IBM, she is now developing a forecasting tool with engineers there.

“The IT community must work hand in hand with government agencies and aid organizations around the world to help manage the existing refugee crisis and make sure that next time–-and there will be a next time–we are prepared to do things better,” she wrote in an article for Wired last year.


The Georgetown project, involving seven academic institutes, has had funding from the National Science Foundation, the Canadian government, and Livermore National Labs. Singh hopes the system will be ready in three to five years, though she doesn’t think the final product should be a blinking red light sort of alert system. She envisions it as a free decision-support tool that could help governments and aid groups spread resources like food and shelter most efficiently. Flowminder, too, is now working with the International Organization for Migration and aid groups to scale up its work.

With better systems in place, Europe could have been more prepared for refugees. Instead, governments were caught on the back foot by the flows, and the result was bad all-around: chaos in the Mediterranean, overstretched coastguards, and millions of dollars spent. If officials could have acted earlier and more decisively, they may have been able to do more to protect millions of people. Now the most important goal is to do a better job of forecasting, and hopefully preventing, the next crisis.

Have something to say about this article? You can email us and let us know. If it’s interesting and thoughtful, we may publish your response.

About the author

Ben Schiller is a New York staff writer for Fast Company. Previously, he edited a European management magazine and was a reporter in San Francisco, Prague, and Brussels.