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

This paper presents a simple method for calculating deviations between actual city size and the size which would correspond to it with a Pareto exponent equal to one (Zipf’s Law). The results show two differentiated behaviours: most cities (80.25%) present a greater size than that which would fulfil Zipf’s Law, while small cities (19.75%) tend to be too small. Our aim is to analyse the distribution element by element, using data about city characteristics from all American cities in 2000, and to explain the deviation between the size predicted by Zipf’s Law and the actual size of each city. To do this a Multinomial Logit Model is used. The most important variables affecting the probability of a city presenting a negative or positive deviation are per capita income, human capital levels and the percentage of the population employed in some sectors