A PRACTITIONER'S PERSPECTIVE ON THE INDEX OF DISSIMILARITY.

The article presents four major objections to the indexes of dissimilarity as a measure of residential or geographic segregation. The property that gives rise to their first objection is that the zero point of no segregation is an even distribution of blacks and whites among subareas of the city. Where the size of subareas is small and the proportion of blacks in the city is also small, the minimum attainable value of the index may be quite high. A corresponding argument could have been made concerning the maximum attainable value of the index of dissimilarity, but the conditions contributing to the severity of this constraint are different. For a perfectly segregated distribution, the minority must completely fill some set of subareas. The larger the size of the subareas, the less likely that an exact match is possible between the number of minority households and the total number of households in any set of subareas. A random residential distribution process allows additional deviations from strict evenness in addition to those caused by the constraint of working with integer units.