The duration of Turkish exports and its determinants

The standard theory of international trade almost implies that trade patterns are highly static and persistent. Therefore, the issue of trade duration is generally ignored in these standard trade models. The literature on trade duration has been popular since the seminal works by Besedes and Prusa (2006a-b). These papers reveal that trade relationships are often very short-lived contrary to previously thoughts. In line with this unexpected result, this thesis provides a thorough statistical description and regression analysis of the duration of Turkish exports. The aim of this thesis is two-fold. Firstly, it attempts to identify the duration of Turkish exports by performing a highly detailed analysis of descriptive statistics and the Kaplan-Meier method. Secondly, this thesis explores the factors that affect on the hazard rates of export flows. To this end, two different regression analysis are performed by using Cox proportional hazard model and discrete-time Probit, Logit and Cloglog hazard models. The detailed trade data reported by CEPII-BACI are employed to investigate Turkey's export to 245 partners from 1998 to 2013 according to the 6-digit Harmonized system. Results obtained from the analysis of descriptive statistics suggest that the duration of Turkish exports is very short-lived. The median and mean duration of Turkish exports are merely one year and 3,41, respectively. 51% of all trade spells, however, cease during the first year. The Kaplan-Meier estimates of survival functions show that all survival curves are downward sloping with decreasing rate and about 42% of export relationships is likely to fail in the first year. The results of regression analysis indicate that discrete-time Logit hazard model is more suitable hazard model for estimation. Empirical evidences of discrete-time Logit hazard model demonstrate that common language, common border, importer GDP, relative real exchange rate, initial export value, lagged duration, total export value, number of export products and number of export markets variables have a strong negative impact on the hazard rates of export flows. Whereas distance, difference in GDP per capita and EU-27 variables have a positive effect on the hazard rates of export flows.