The Dynamics of Ethnic Hierarchies: Evidence from the Age of Mass Migration
P4-01
Presented by: Alessandra Stampi-Bombelli
Recent decades have witnessed the relocation of immigrant groups. In destination countries, new minorities come into contact with both the host population as well as more established extant immigrant communities. An open empirical question is how the arrival of new immigrants affects the perceived socio-cultural distance between the extant immigrant groups and host population.
We address this question by studying an important period in U.S. history, the Age of Mass Migration (1860-1920), in which a sizeable and diverse group of migrants arrived. Applying advanced computational linguistics techniques to a 11 million pages of daily newspapers, representative of the local context, we present a novel text-based measure of perceived distance between each immigrant group and American-born natives.
On this corpus, we train a word embedding model that learns meanings and semantic relations between words. For each individual mention of an immigrant group, we compute “à-la-carte” context-specific vectors (Khodak et al. 2018) and project the vectors onto a dimension pointing from “American” at one pole to “Immigrant” at the other pole. Higher values in our measure indicate that the mention more closely resembles contexts used when portraying immigrants, rather than natives.
At the end of this process, we use this outcome measure to analyse the dynamic response of local perceptions of ethnic hierarchies to shifts in immigrant exposure. In particular, we causally estimate the effect of an increase of an immigrant group size on distance towards that group, as well as second-order effects on local attitudes towards other immigrant groups.
We address this question by studying an important period in U.S. history, the Age of Mass Migration (1860-1920), in which a sizeable and diverse group of migrants arrived. Applying advanced computational linguistics techniques to a 11 million pages of daily newspapers, representative of the local context, we present a novel text-based measure of perceived distance between each immigrant group and American-born natives.
On this corpus, we train a word embedding model that learns meanings and semantic relations between words. For each individual mention of an immigrant group, we compute “à-la-carte” context-specific vectors (Khodak et al. 2018) and project the vectors onto a dimension pointing from “American” at one pole to “Immigrant” at the other pole. Higher values in our measure indicate that the mention more closely resembles contexts used when portraying immigrants, rather than natives.
At the end of this process, we use this outcome measure to analyse the dynamic response of local perceptions of ethnic hierarchies to shifts in immigrant exposure. In particular, we causally estimate the effect of an increase of an immigrant group size on distance towards that group, as well as second-order effects on local attitudes towards other immigrant groups.