This is the ninth in this years series of job market posts.
In 2019, there were more than 25 million refugees worldwide (WMR, 2020). Many of them face material insecurity and lack access to basic services such as healthcare and education. To overcome these challenges, fostering refugees' integration in the host society is essential. Social integration, understood as refugees’ ability to interact with locals, is crucial as it directly influences their economic and cultural integration into their new communities.
In my job market paper, I explore how refugee networks influence social integration in the host community. I study this question in the context of Turkey, a country that has been profoundly impacted by the arrival of more than 3.5 million Syrian refugees since 2011. Using a rich dataset on the mobile phone communications of Syrian refugees living in the country, I document a clear pattern. In places with a larger refugee population, refugees make significantly more phone calls with locals.
You may have thought that as a refugee, the more you lived with members of your home community, the higher the risks of non-integration. Previous research has indeed shown that such a threat exists, as communicating with individuals who share the same background is often easier (Lazear 1999; Danzer and Yaman 2013). However, social scientists have also argued that co-refugee networks could provide invaluable help to their members by sharing information on local customs and culture and creating new opportunities to meet locals (Portes 1995; Nannestad et al. 2008; Pendakur and Mata 2012).
Recently, one of the major Turkish telecom companies has made available mobile phone data from more than 1 million customers to researchers (Salah et al. 2018). A key aspect of this database is the existence of a refugee tag, assigned to customers with ID numbers given to refugees and foreigners in Turkey, who registered with Syrian passports, or used special tariffs reserved for refugees. A total of more than 180,000 customers are tagged as refugees. Importantly, for each call involving a refugee, the researcher can observe whether the receiver of the call is also a refugee or not. Other researchers have used this database to gain essential insights into refugees' living conditions in Turkey (Salah et al. 2019).
I use this rich database to construct measures of refugee presence and social integration at the village and neighborhood level. Indeed, the spatial granularity of mobile phone data has two main advantages. First, it allows me to observe refugees' exact location down to the cell tower level. Second, it allows me to directly observe how frequently refugees interact with other refugees and locals.
While the richness of this database opens exciting venues of research, it also comes at a cost. First, to protect the anonymity of the telecom company’s customers, I do not observe their demographic characteristics. Second, it is difficult to reach a definitive answer to why refugees interact with locals, although the data granularity allows me to classify calls among broad categories (such as private versus professional). Third, I likely miss many communications on alternative free messaging services, such as WhatsApp or Messenger. Crucially, my analysis relies on the assumption that refugees are not less likely to use these free substitutes to calling and texting when they live in enclaves.
The analysis of the dataset reveals a stark pattern. Refugees who live in localities where the refugee community's size is relatively small are socially isolated: they have few interactions with other refugees and few interactions with non-refugees. As the size of the refugee community living in a locality increases, individual refugees start to interact more with other refugees, but also with the local population (Figure 1). Moreover, the proportion of calls placed to non-refugees increases.
Strikingly, these correlations are stronger in localities where the refugee community's size is relatively small. This is especially the case for calls to locals: an increase in the size of refugee community is strongly correlated with more calls to locals only when refugees' share is below 10%. This finding points to a potential detrimental effect of social isolation on refugees' ability to interact with locals.
Refugees having more interactions with locals in places that host a larger refugee community doesn't necessarily mean there is a causal relationship between refugee network size and integration. Broadly speaking, we can make a distinction between two main sources of endogeneity in this context. First, my results could come from the unobserved characteristics of localities that host more refugees. Second, refugees might decide to self-select into some localities based on their own preferences for integration with Turks. I argue that these two alternative explanations are unlikely to explain my results.
Let's start with the first hypothesis: refugees might be more likely to self-select into localities where they anticipate their social integration will be easier. To account for this, I take advantage of the fact that localities on the way from Syria to a major economic center in Turkey host more sizeable refugee communities than other localities, despite being similar in many other predetermined dimensions. I use this intuition to build an instrument variable strategy, which corroborates the positive relationship between refugees’ community size and social integration.
Moreover, refugees might decide to self-select into some localities based on their own preferences for integration with Turks. For example, refugees who don’t speak Turkish might be more likely to live in ethnic enclaves. However, this type of bias is likely to lead to an underestimate of the positive relationship between the size of a refugee community and social integration. In addition, such biases induced by selective migration should be weaker when using a more aggregated unit of observation than villages or neighborhoods. I show that my main results hold even when the refugee community’s size is defined at a more aggregated level.
Why do co-refugee networks help their members interact with locals?
I examine two important mechanisms that could explain the main results.
1. Professional activities. If refugees who have access to a larger co-refugee network are more likely to be employed, they could have more professional interactions with non-refugees. However, I show that the papers’ main results remain unchanged when focusing on calls and texts occurring outside of working hours (between 6 pm and 8 am from Monday to Friday, or during the weekend). Moreover, by analyzing refugees' mobility patterns, I do not find any strong evidence supporting the hypothesis that co-refugee networks increase employment of network members.
2. Networks make it easier to contact locals. Social scientists have claimed that co-refugee networks can also provide invaluable help to their members by sharing information on local norms and by creating new opportunities to interact with locals. I show this mechanism is stronger when the cultural distance between refugees and locals is wider. To do this, I build upon Bozcaga et al. (2019) and construct a locality-level measure of religious distance which compares the average number of calls for both communities during Friday prayers and at the same time on Thursdays. I confirm that co-refugee networks are especially useful for the social integration of their members in localities where refugees and locals are culturally distant.
Overall, co-refugee networks play an important role in breaking the social isolation of refugees. While many countries use dispersal policies of asylum seekers and refugees at arrival, my job market paper highlights an important lesson. Excessive efforts at dispersing refugees across different geographic regions could be counter-productive by making it more difficult for new arrivals to interact with locals.
Thomas Gautier is a PhD student at Boston University.