Deviations from Zipf’s Law for American cities: an empirical examination

This work presents a simple method for calculating deviations regarding city size and the size which would correspond to it with a Pareto exponent equal to one unit (Zipf’s Law). Recent works show that when considering the entire sample without size restrictions, the estimated Pareto exponent tends to be much lower than one. Our aim is to analyse the distribution element by element, taking data from all American cities in 2000, and explain the deviation of the size predicted by Zipf’s Law and the real size of each city, using variables for each city of per capita income, distribution of employment among sectors, individuals by level of education, etc.; explicative variables which attempt to capture the influence of local externalities. To do this a Multinomial Logit Model is used.