Methods and data considerations related to the size distribution of settlements

Since Zipf, research on rank-size distribution of settlements has generally dealt with the upper tail of the distribution. Previously Turkish rank-size distribution was examined only for the upper part by Dökmeci and others. Parr had pointed on the similarity between the log-normal distribution and the rank size distribution at the upper tail. Within this paper the whole dataset of Turkish settlements (over 36.000 units) is employed for the period 1980-2000, thus, an analysis with no threshold population is undertaken. The results reveal a lognormal distribution. Furthermore the rank-size distribution from 1927 until present is also examined using a threshold of 1.000 population. With every successive census, the convex distribution of 1927 gradually evolves into a linear distribution reaching the slope of unity at the 1990’s. For recent censuses there is a point of refraction at population size 2.000 which is the threshold population for municipal incorporation and it seems that the unnatural refraction is a sign of fraud in census caused by localities over-enumerating with the aim of municipal incorporation. Another focus is on proper delimitation of boundaries of settlements in order to eliminate the possible distortion caused by administrative boundaries irrelevant to the unity of settlements. Methods applying international standards of urban area delimitation to the Turkish context, urban-rural differentiation and proper selection of threshold figures are discussed. Also definitions of administrative units (provinces, districts) and settlement units (cities, towns, villages) are given for the Turkish legal system compared with other countries.