More sources and unconventional data can make it better to predict and calculate refugee flows and researchers at Malmö University have helped develop a model that can be applied at both national and local levels.

The starting point for the research has been to understand the factors that trigger migration flows to Europe, such as economic conditions, conflicts, and climate factors. These drivers are then fed into a statistical model to predict changes in the near future.

Now we have a basis to better predict future patterns and to make more evidence-based decisions.

Haodong Qi

Based on traditional population data from Eurostat and World Development Indicators, the researchers’ HuMingBird project also incorporates more unconventional data into the model. These include satellite monitoring of the climate or data from the Uppsala Conflict Data Program (UCDP), and a research programme that collects data on organised violence. Another source is Google Trends, that is, online search trends that may show a particular interest in, for example, Sweden in Iraq, Syria or any other country.

“We find correlations between trends in online searches and migration flows,” says Haodong Qi , a researcher at the Department of Global Policy Studies.

According to Qi, this data fills a gap that official statistics do not cover. Climate factors that come into play include floods, droughts, and harvesting conditions.

“There are two types of climate factors: sudden disasters and slower scenarios," says Qi.

Together with a research colleague in Brussels, Qi has developed the Flow Specific Tempro Gravity model, which provides a better basis for future calculations. In particular, they are currently studying how climate affects the migration of people from individual districts in Somalia to the EU.

“The older models could not predict the refugee flows in 2016. Now we have a basis to better predict future patterns and to make more evidence-based decisions.”

The mechanisms behind migration are dynamic. Qi emphasises that there are other refined models already in use, and that what we can say now is based on past events. If new patterns emerge, the model needs to be updated.

“Of course, we can't say 100 per cent that it will be a certain way. We develop tools for evidence-based decision-making, but then it has to be used in the real world, and then politicians have to act and agree on appropriate action," adds Qi.

Text: Magnus Jando & Adrian Grist