Dealing with the bias of the dissimilarity index of segregation

The dissimilarity index is widely used to evaluate the extent of segregation in the allocation of a minority group in two or more spatial units. There is a widespread awareness that due to its sensitivity to random allocation, it is inherently subject to an upward bias. This bias may be irrelevant when the sets concerned are large, but if the area population is small or the minority proportion is very low, the index can be highly misleading. Common strategies used in literature to deal with the index bias rely on the use of informal rules of thumb, which at least have the side effect of restricting the scope of segregation studies. In time, solutions for dealing directly with the index bias have been proposed. These solutions are based on adjusting the index scores downward, and require the computation of the sampling distribution of the index. Most of the methods proposed use computation-intensive techniques that have the drawback of introducing complexity and substantial computational burdens. The lack of userfriendly computer programs implementing these methods has strongly affected their wider adoption.