![Sample our Urban Studies journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/110826/TOC-Subject-Sample-2021-Urban-Studies.jpg"})
Abstract
The aim of this study is two-fold. Firstly, this study examines the size distribution of cities by using Zipf's law. The second objective of this study is to investigate the effects of determinants of urban growth in Turkey by using the data for the 1980–2007 time period. The main findings of the study show that there is some evidence that Zipf's law holds in Turkey. Moreover, according to the rank-minus-half rule, the results suggest stronger support for Zipf's law in size distribution of the cities. Furthermore, the regression results indicate that fertility rate, location of the city, migration, agglomeration in services and specialization in manufacturing industry have positive impact, whereas schooling rate has a negative effect on growth of the urban population regarding Turkey.
Acknowledgements
This paper is based on the project supported by TUBITAK. The project number is 106K 283.
Notes
†The previous version of this study was also presented at IASK international conference 13–15 October 2008, Porto, Portugal.
City population indicates the population of municipial areas of the province and the district centres in Turkey (see TURKSTAT, 1980–2008). The population of the cities indicates urban population and, therefore, there are 81 provinces and 81 cities with regard to the data provided by TURKSTAT.
The share of the agriculural sector in GDP was around 40% in 1968, and it was around 9% in 2008, and the sectoral breakdown of labour force in the agricultural sector was around 55% in 1980, and it was around 33% in 2003 (see TURKSTAT, Turkey in Statistics, 1980–2008).
Nota and Song (Citation2008) used both OLS and the rank-minus-half rule methods to estimate the Pareto exponent for the same sample, and they found that the estimated coefficients of cities by using the rank-minus-half rule are bigger than the estimated coefficients of cities by using the OLS method.
A test for nonlinearity is to include a quadratic term to equation (2) as in Black and Henderson (Citation2003). We have also estimated the equation with the quadratic term and found some evidence on the significance of this term. However, Gabaix and Ioannides (Citation2004) show that the estimated coefficient of this quadratic term will turn out to be statistically significant in situations where Zipf's law perfectly holds. Therefore, the results are not presented here.
We have also used the Hill approach to estimate the Pareto exponent, but we have found that the Hill estimator is always smaller than the OLS estimator as in Soo (Citation2007), indicating that the Hill estimator is unable to correct the downward bias of the OLS estimate. The results are available upon request.
Deliktaş (Citation2008) finds less evidence of supporting Zipf's law using the standard OLS method. As Gabaix and Ioannides (Citation2004) argue, the standard estimation of OLS may cause biased coefficients. Therefore, we have used more recent techniques as suggested by Gabaix and Ioannides and found stronger evidence for Zipf's law.
The indices are calculated by authors in the study.
These indices are computed for the manufacturing industry and the service sector separately by using the same formula.
We could not calculate the competition index for the service sector due to lack of data for the firm level in Turkey.
We conducted multicollinearity diagnostic through calculation of the variance inflation factor (VIF). Although the results show that the relationship between the two variables are negative, the VIF number which is smaller than 3 does not sign a multicollinearity problem between the variables in the regression.
