Project Topics
Get Complete Project Material File(s) Now! »
DATA AND STATISTICAL METHOD
This section is devoted to inform the reader about the sources of collected data, the difficulties encountered during data collection and modifications that have been done before running the regression. Moreover, I will give a short introduction on statistical methods that will be used for empirical analysis.
Data Collection
wider variety than international databases such as World Bank, IMF or Eurostat. It is possible to reach various statistics databases at city-level through the web pages of different governmental or semi-governmental organizations in Turkey. The web pages that I benefit from are Turkish Exporters’ Assembly (TIM), Republic of Turkey Prime Ministry – Undersecretariat of Treasury, Automotive Manufacturers Association (OSD), Republic of Turkey Ministry of Transport and Communication – General Directorate of Highways, and Turkish Statistical Institute (TurkStat) – Regional Statistics Unit. The organizations I benefited from and the kind of data I retrieved from those sources are presented in the appendix 2.
Data Modification and Presentation of Variables
The main difficulty that I encountered during data collection was to decide on what data to use as dependent variable. In literature, a common way to control location decision choices is using employment statistics in automotive industry or using a kind of census of local industry and business units in automotive industry. A census by TurkStat -General Census of Industry and Business Local Units Provinces- is available, however, this census is conducted in 2002 and their results are presented at regional level instead of city-level. After investigating most of the resources related to topic, I could come up with “export of motor vehicles and related industry” data from 2007 which is released by TIM to substitute General Census of Industry and Business Local Units Provinces of TurkStat. At first, this data might seem inadequate or inapplicable to test the location decisions of automotive producers in Turkey. Nevertheless, this data becomes plausible when we take into consideration that 73,6% of all automotive production in 2007 is exported according Automotive Manufacturers Association Bulletin.
Thus, city-level export of motor vehicles and related industry data by TIM can be used as dependent variable which actually demonstrates the location-choice decisions of automotive producers in Turkey.
Another issue regarding data is the modification of the existing data. Accessibility to ports is an important element of the study since 94% of automotive export is achieved by sea transportation according to Aksam newspaper which bases its news on Automotive Strategy Document by Republic of Turkey – Ministry of Industry and Trade. Automotive Logistic Plan for Marmara Region which was prepared by Koc University for Automotive Manufacturers Association in 2008 is a research paper that gives units of auto handling for the seaports in Marmara region (in cities of Bursa and Kocaeli) where almost all motor vehicle export is handled in Turkey in 2007.
As it can be seen from Table 1, “export” represents the figure of export of motor vehicles and related industry products at city level in USD, thus it also demonstrates the location-choice decisions of automotive producers in Turkey. Number of companies operating in automotive industry with foreign capital at city level is represented with “mnf”. Market potential -“mp” represents the accessibility of cities to seaports and expressed in units of vehicles. Dummy variable demonstrates whether there is motorway structure in the city or not and it takes the value of 1 for yes, and 0 for no. Distist shows the distance to the industrial and commercial core of Turkey – Istanbul in kilometres. Number of motor vehicles per thousand residents (mvehicle) is used to test the importance of domestic market. Initially GDP per capita is planned to be used for measuring domestic market strength. However, GDP per capita at city level has not been released since 2001. Moreover, I think number of motor vehicles per thousand residents is a better measurement since it gives a better picture of domestic demand. According to a working paper by MIT’s Industrial Performance Center, market penetration (people/car) for Turkey in 1995 is much higher than western industrialized countries, respectively 21 for Turkey and less than 5 for western countries. This displays that Turkish domestic automotive market has a room to expand, thus I believe that the number of motor vehicle per thousand residents is an important measure to find out market strength.
Statistical Method
A multiple regression will be run by using Ordinary Least Square (OLS) estimation method which bases on minimizing the sum of the squares of the residuals. Since it is an analysis based on cross-sectional data, heterogeneity might be a problem. Thus, “the user of the cross-sectional data should be aware of the risks generated by size and scale effect” (Gujarati, 2003). To eliminate those effects, I derive benefit from log transformations by taking natural log (ln) of some variables with great size. By doing so, I reach results that are easy to interpret and get a regression line which has a better fit without including extra independent variables. In regression analysis, the log transformation choice allowed me to refrain from heterogeneity problem which might be generated by huge export figures of motor vehicles in USD and market potential figures in units of vehicle.
After discussing OLS results, the Tobit model will be used to control the robustness of the results of OLS regression. The Tobit model is used when dependent variable is not observed in all data set. If the non-observed dependent variables are ignored, the results based on observed part of dependent variable can lead us to be biased and inconsistent (Gujarati, 2003). Thus, the Tobit model in which the left-hand side is censored to 0 is applied in order to check the robustness of the results.
EMPIRICAL ANALYSIS
In this section I firstly will present some results that indicate geographical concentration of automotive industry in Turkey before proceeding to regression analysis. Afterwards, regression analysis will be conducted by testing them according to the different econometric models mentioned in the previous section. The analysis will continue with interpretation of the results, control for common problems encountered at the regression analysis and finally end up with evaluating the results of the regression.
1.INTRODUCTION
2.THEORETICAL DISCUSSION
2.1 Background
2.2 Theoretical Framework
2.3 Hypothese
3.DATA AND STATISTICAL METHOD
3.1 Data Collection
3.2 Data Modification and Presentation of Variables
3.3 Statistical Method
4.EMPIRICAL ANALYSIS
4.1 Evidence for geographical concentration of automotive industry
4.2 Regression Model and Results
5.CONCLUDING REMARKS
6.REFERENCES
GET THE COMPLETE PROJECT
The determinants of location choices of automotive producers in Turkey
