A latent class model of residential choice behaviour and ethnic segregation preferences

The nature of ethnic residential clustering involves different population segments which through their location decisions influence the spatial patterns of ethnic settlements. Understanding the differences in the residential behaviour of a heterogeneous population and, in particular, the tastes dissimilarities for ethnic composition of neighbourhoods becomes essential for analysing the dynamics of ethnic concentrations. However, the residential location choice (RLC) behaviour and especially the preferences for ethnic description of the neighbourhood are subject to heterogeneity in tastes that quite often depend on attitudes and other elements not directly observable by researchers. Employing a latent class choice modelling approach, we aim to examine the observed and unobserved heterogeneity in RLC behaviour across households of different ethnic and socio-economic background. Combining the results from the choice and class-membership model components and interpreting the sensitivities and probabilistic composition of the latent classes allows us to evaluate the impact that each attribute exercises for each typology of respondents. For the empirical analysis we use a dataset collected from a specifically designed Stated Preference Experiment of Neighbourhood Choice, conducted in the Swiss city of Lugano. Such experiment permits us to uncover the impact of preferences for ethnic neighbourhood composition, free from the constraints component usually existing in the real housing markets (such as access barriers to some urban areas, shortage of accommodation options or discrimination effects), by implying a hypothetically free choice among alternative neighbourhoods. Pivoted design of the choice experiment, i.e. design based on the present values of attributes describing the residential location of each respondent, ensures the matching between the real housing situation of individuals and the choice situations they face throughout the experiment. The results support the hypothesis of existence of three latent classes which differ in their housing choice behaviour and tastes for ethnic residential environment. In particular, different ethnic attributes are considered as important choice drivers by households belonging to two latent clusters. Swiss citizens and foreigners belonging to disadvantaged ethnic minorities share similar tastes being mainly concerned about the foreigners? share in the neighbourhood, both showing disutility associated to this ethnic attribute. However the presence of conational neighbours does not seem to play an important role in shaping their residential location choices. Such results could indicate the propensity of the native population to a grouping behaviour, being the underlying motivation the aversion to high presence of foreign inhabitants rather than self-segregation preferences. Similarly, the results for the disadvantaged foreign communities suggest that the clustering among own community of origin, often observed in the reality, is not driven by the voluntary self-segregation preferences. On contrary, advantaged foreign communities value the residential proximity to their co-nationals, but result indifferent to concentrations of other foreign groups.