The study uses Social Identity Theory as a framework to explain how language acts as a source of social identity and motivates individuals to sort themselves by residential location. To assess the validity of the framework, the study tests the hypotheses that group size, geography, and institutions matter in the preservation of language identity, using the 1991, 1996 and 2001 census data for urban Montreal, Toronto and Vancouver, Canada. The study models a system of three simultaneous equations that describe changes in property values and mobility of language groups, accounting for the presence of spatial lag and spatial error. The study estimates the model by generalized spatial two-stage least squares (Kelejian and Prucha, 1998). The results of the study show that, while residential segregation by language could be a cognitive behaviour, people’s search for language identity within a social group is influenced by economic opportunities in terms of capital gains in properties; it is also affected by proximity to peers and by government policies favouring language-based activities.
