Wednesday, June 24, 2015

Where Do People Meet? Urban Segregation in Cellular Data

Where Do Ethnic Groups Meet?
Ott Toomet, Siiri Silm, Rein Ahas, Erki Saluveer, Tiit Tammaru

Immigrants and ethnic minorities often live more or less separately from the majority population. This fact is sometimes touted as a major problem that hampers social integration. But does the place where we live actually matter? Anecdotal evidence suggests that we do not interact much with our neighbors. Why should we worry then who they are?

To shed more light on this issue we analyzed the spatial behavior of Estonian speaking and Russian speaking population in Tallinn, Estonia. It is an interesting city as it is almost 50-50 split between the corresponding ethnic groups. We used cellphone data for the analysis. Mobile operators always record which antenna talks to our phones, and such data essentially form a spatial track of our activities where one can see our approximate location and time. It is relatively straightforward to analyze the tracks and deduce who were close to each other, and when and where it happened. Looking at the repeated patterns of calls we can also guess where people live and work. In this way we can see who can potentially meet each other, and where such meetings might occur.

So, are Estonian speakers close to Russian speakers in Tallinn? Well, it depends on when. We find that when at home and at work, both groups are substantially segregated. However, this is much less true when people are elsewhere and busy with other tasks, such as shopping, but also during various leisure-time activities. Even more, such free-time segregation is not closely related to the environment in the place of residence and place of work (see the figures). Even those who live in almost completely Estonian or Russian neighborhoods experience a rather mixed free-time environment.

Fig 1. Relationship between free-time meetings and composition of residential neighborhood. Circles and triangles represent the averages across different neighborhoods. For Russians (red triangles), the percentage of Russian speakers ranges between less than 10% to almost 80%, for Estonians (black circles) between 20% and almost 100% (horizontal axis). Despite of living in such different neighborhoods, during free-time both groups meet own-group members for approximately 50-60% of the time.
Fig 2. Relationship between free-time meetings and composition of work neighborhood. The message is similar to that in Figure 1: work neighborhoods vary a lot, but it is not closely associated with our free-time environment.

What do these results tell us? There are several interesting conclusions.

  • residential segregation may be less of a concern than often suggested. We spend much of our active time elsewhere, and much of the time we are at home we sleep.
  • As a typical European city, Tallinn has a dense urban core where a substantial part of these meeting occur. It suggests that a dense vibrant downtown is favorable for bringing together people of different background. The results for Los Angeles may well be different.
  • Finally, we do not know what is behind these meetings. Most of these are probably related to just being close to each other in a crowded city. But so are our relations with our neighbors: most of them we would not even recognize on street. There is need for more analysis on what are the "meaningful places" in terms of where do we actually socialize with people.

The full article is available at