Detecting ethnic residential clusters using an optimisation clustering method

Abstract

To understand residential clustering of contemporary immigrants and other ethnic minorities in urban areas, it is important to first identify where they are clustered. In recent years, increasing attention has been given to the use of local statistics as a tool for finding the location of racial/ethnic residential clusters. However, since many existing local statistics are primarily developed for epidemiological studies where clustering is associated with relatively rare events, its application in studies of residential segregation may not always yield satisfactory results. This article proposes an optimisation clustering method for delineating the boundaries of ethnic residential clusters. The proposed approach uses a modified greedy algorithm to find the most likely extent of clusters and employs total within-group absolute deviations as a clustering criterion. To demonstrate the effectiveness of the method, we applied it to a set of synthetic landscapes and to two empirical data sets in Auckland, New Zealand. The results show that the proposed method can detect ethnic residential clusters effectively and that it has potential for use in other disciplines as it offers an ability to detect large, arbitrarily shaped clusters.

Keywords:

• segregation
• residential clustering
• optimisation
• heuristic approach

Notes

1. Residential segregation is a multidimensional phenomenon that represents several distinctive spatial patterns (Massey and Denton 1988, Reardon and O'Sullivan 2004). The index of dissimilarity addresses one aspect of segregation, evenness, and there are other indices that can assess, for example, the degree of interaction between the majority population groups and ethnic minorities. While such exposure/interaction measures can provide useful insights into the ethnic residential segregation (St John and Clymer 2000, Spivak et al. 2011), they were not discussed in this article as it is less relevant to the clustering method proposed here.

2. In April 2009, the New Zealand government made a decision to combine the seven local authorities into one unitary Auckland Council. The new council is composed of 13 wards, each of which is subdivided into a number of local boards and subdivisions. The study area based on the district boundaries is approximately equivalent to the area consisting of the following wards: Albany (excl. Hibiscus Coast Subdivision), North Shore, Waitakere (excl. Waitakere Ranges Local Board), Albert–Eden–Roskill, Waitemata and Gulf (excl. Waiheke Local Board and Great Barrier Local Board), Ōrākei, Maungakiekie–Tāmaki, Manukau, Howick, and Manurewa–Papakura (excl. Papakura Local Board).