Zipf’s law of city-size distributions can be expressed by three types of mathematical
models: one-parameter form, two-parameter form, and three-parameter form. The one-parameter
and one of the two-parameter models are familiar to urban scientists. However, the
three-parameter model and another type of two-parameter model have not attracted attention. This
paper is devoted to exploring the conditions and scopes of application of this Zipf models. By
mathematical reasoning and empirical analysis, new discoveries are made as follows. First, if the
size distribution of cities in a geographical region cannot be described with the one- or
two-parameter model, maybe it can be characterized by the three-parameter model with a scaling
factor and a scale-translational factor. Second, all these Zipf models can be unified by hierarchical
scaling laws based on cascade structure. Third, the patterns of city-size distributions seems to
evolve from three-parameter mode to two-parameter mode, and then to one-parameter mode.
Four-year census data of China’s cities are employed to verify the three-parameter Zipf’s law and
the corresponding hierarchical structure of rank-size distributions. This study is revealing for
people to understand the scientific laws of social systems and the property of urban development.