Assessing Refugees’ Integration via Spatio-temporal Similarities of Mobility and Calling Behaviors

Antonio L. Alfeo, Mario G. C. A. Cimino, Bruno Lepri, Alex S. Pentland, Gigliola Vaglini

IEEE Transactions on Computational Social Systems, Volume 6, Issue 4 (2019), Pages 726-738


This paper analyzes the conditions that can contribute to the integration of Syrian refugees in Turkey by analyzing Call Details Record (CDR) datasets including calls from refugees and locals in Turkey throughout 2017. The authors propose the following set of metrics to assess the social integration of refugees:

  • Refugee’s Interaction Level: percentage of calls made by a given refugee to a local in a given period of time, an indicator of the refugee’s social connections in the local community.
  • Refugee’s Calling Regularity: number of calls made by a person in a given hour of the day during a given period of time. The similarity between the calling patterns of locals and refugees (dependent on daily routines) may be considered as a proxy for integration.
  • Refugee’s Mobility Similarity: similarity between refugees’ and locals’ mobility patterns.
  • Residential Inclusion by District: the coexistence of resident locals and refugees in a given district in a given month, based on the assumption that most calls during the night and early morning are made from a person’s residence.
  • District Attractiveness: based on the percentage of resident refugees who do not leave the district in that month, i.e. continue to reside in the district the following month.
  • District Cost of Living: average rent cost per square meter in a given district, as an indicator of the cost of living for that district.

The authors employ a novel computational technique (Computational Stigmergy) to analyze spatio-temporal patterns. They focus their analysis on the cities of Istanbul, Ankara and Izmir, which have the larger density of antennas (granularity of mobility patterns) and larger calling activity made by refugees. They find:

  • Interaction level and calling regularity are positively correlated, suggesting that refugees that exhibit greater interaction with locals may have similar daily routines.
  • Districts in Istanbul with higher residential inclusion may also be characterized by higher similarity between locals’ and refugees’ routines. This suggests that a minimum number of refugees per area is required for triggering integration.
  • District attractiveness and cost of living are significantly and inversely correlated, suggesting that districts with a lower cost of living are more attractive to refugees.
  • Calling regularity of refugees living in a district is correlated with the cost of living in that district. This suggests that calling regularity may be used as a proxy for daily routine similarity and for the economic capacity of refugees (i.e. the ability to meet a certain cost of living), and may therefore indicate the employment of refugees. This result is confirmed by evaluating the correlation of the cost of living with the distances between calling patterns of locals and refugees; as the distance increases, the district’s cost of living decreases.
  • Variables with the strongest correlation with calling regularity are district attractiveness and cost of living, suggesting that economic capacity (e.g. being employed) and a long-term residence provide the greatest contribution to integration.
  • Mobility similarity is correlated with interaction level. The more refugees have interactions with locals, the more they share urban spaces with locals.
  • Social tension (drawn from news reports of violent brawls, clashes and terrorist attacks etc.) affects the behavior of refugees by reducing the amount of shared urban space with locals (i.e. lowering the mobility similarity after the event). Moreover, in terms of calls made toward locals, the social tension event has a greater effect on the group of refugees with lower interaction levels.

The authors conclude that mobility similarity and calling regularity have great potential as measures of social integration, since they are: (i) correlated with the amount of interaction between refugees and locals; (ii) calling regularity is an effective proxy for refugee’s economic capacity, implying refugee’s employment; (iii) mobility similarity is affected by social tension events; and (iv) the behavior of less integrated refugees appears to be significantly more affected by social tensions. The authors acknowledge that findings may be limited by the representativeness of a behavioral model based on call data (i.e. excludes other communication and messaging platforms).




Syria | Turkey