Abstract
A new descriptive model of the frequency distribution of city size, termed the constant-Gini model, is introduced. The motivation behind this new model is threefold. First, it is believed that studies of the population structure of urban systems are most rewarding when systems are defined in terms of urban umlands instead of entire national territories. Second, it is noted that many large cities exhibit moderate or strong degrees of urban primacy within their own umlands. Third, the traditional rank-size rule generally is unable to describe a primate structure without significant error. Hence a new model is needed for application at the regional (i.e., umland-based) scale. The proposed constant-Gini model appears to satisfy this need. The performance of the new model is compared to that of the rank-size rule when applied to 34 regional urban systems in three different countries, using various levels of truncation of the ranked lists of city populations. Overall, the constant-Gini model reproduces the observed population data more accurately than the rank-size model in two-thirds of all cases examined. Brief comments are offered on the factors that appear to determine which of the two models provides the better fit in any particular case.