Refugee Mobility: Evidence from Phone Data in Turkey

Michel Beine, Luisito Bertinell, Rana Comertpay, Anastasia Litina, Jean-Francois Maystad, and Benteng Zou

Paper presented at the Research Conference on Forced Displacement 2020, co-organized by the Joint Data Center on Forced Displacement.


This paper examines the mobility of refugees across provinces in Turkey as a measure of their social integration. The analysis is based on call detail records from the Data for Refugees Turkey (D4R) challenge, combined with socioeconomic data at the province level (nightlight density from satellite data as a measure of income) and indices for various types of news (leadership change, boycotts, violent protests, economic aid, humanitarian aid, and asylum grants). The authors employ a standard gravity model to empirically estimate determinants of refugee movements—compared to non-refugee movements—including: province characteristics; distances between provinces; income levels; network effects; refugee-specific determinants such as the presence of refugee camps and the intensity of phone call interaction among refugees; and certain types of news events.

Key findings:

  • Non-refugees move further and more frequently than refugees. Given that refugees are mostly free to move within Turkey (in some provinces restrictions may apply but they are not strictly enforced), the authors posit that the imperfect integration of refugees is the main reason for reduced mobility.
  • Low income levels in the province are a push factor for refugees. Refugees tend to leave poorer provinces but may not be able to reach wealthier provinces (possibly due to restrictions on refugees).
  • Distance between provinces has a negative impact on mobility.
  • Number of refugee calls (a measure of network effects) has a positive impact on mobility. Refugees tend to move to regions where other migrants have already settled.
  • Refugees tend to leave provinces with an ongoing rally for leadership change. This may capture political instability and pre-election rhetoric that might be directed against refugees. However, this does not act as a pull factor.
  • Higher incidence of boycotts is associated with lower mobility. Provinces with a higher incidence of boycott-related news could be more active in political and humanitarian spheres, which may encourage immigrants to settle. However, this does not act as a pull factor.
  • Refugees tend to move to provinces with more economic aid and leave as economic aid decreases. Humanitarian aid, however, does not have any effect on mobility.
  • News of asylum grants acts as a pull factor for refugee movements.
  • The presence of a refugee camp in the origin or destination province does not have any effect on mobility.

Overall, the authors conclude that the standard gravity determinants apply such as distance, income levels, as well as network effects. Policy interventions associated with political stability, the granting of asylum, and economic aid also matter.