Mainstream media, social media, and the ethnic reputations of neighborhoods
Author(s): Benjamin Jarvis, Miriam Hurtado Bodell, Selcan Mutgan
Thursday 15 | 11:20-11:40
Room: TP52
Session: Frontiers in segregation research
One of the fundamental challenges for empirical scholarship on residential segregation is to find evidence for why ethnic segregation emerges as a function of residential choices in a population. One dominant idea is that the perceived ethnic composition of a neighborhood matters when people decide where to live. For example, Schelling (1971) theorizes that people look around their neighborhood and decide to move or stay based on the perceived (dis)similarities of their neighbors. However, most empirical studies rely on census or register data on the true distribution of ethnicities. The degree to which people’s beliefs about the ethnic composition of their neighborhood align with reality is largely unknown. In this paper, we seek to investigate the alignment between the ethnic reputation of neighborhoods and true ethnic compositions in Sweden and explore how certain events impact the ethnic reputation of neighborhoods.
We collect and analyze texts from one of Sweden’s largest national newspapers, Dagens Nyheter, and posts written in the housing section on Sweden’s largest online discussion forum, Flashback, to study the ethnic reputation of neighborhoods. We extract measures of the ethnic reputation of neighborhoods using a dynamic word embedding model. This model allows us to compare how closely different ethnic groups and neighborhoods are related at different times by calculating the cosine similarity between word vectors of ethnicities and neighborhoods. To study the alignment between the ethnic reputation and the true ethnic composition, we compare the cosine similarities with the true ethnic composition of a neighborhood identified through Swedish register data. Finally, we test if the decision to publish a list of "vulnerable areas" — an area at risk of being impacted by criminals — by the Swedish Police in 2015 resulted in an increased racialized perception of areas on the list.
We find that the ethnic reputation of neighborhoods and their real ethnic compositions, in general, do not correlate much in either traditional media or social media. However, the ethnic reputation and true compositions align more for neighborhoods characterized by high numbers of refugees and asylum-seekers — especially from Middle Eastern and Sub-Saharan African countries — much more than neighborhoods with a majority of ethnic Swedes in both traditional and social media. We find support for the territorial stigmatization hypothesis, indicating that the Swedish police impacted the ethnic reputation of neighborhoods in Stockholm. Moreover, the low correlation between neighborhood reputation and real-world characteristics of neighborhoods suggests that objective measuresof the ethnic composition of neighborhoods may not be optimal to explain how individuals' residential choices drive city-level ethnic residential segregation.