A New Approach to Econometric Modelling of Monthly Total Air Passengers: A Case Study for Atatürk Airport

Creative Commons License

Çelik R. , Karaboğa H. A. , Demir İ.

y_BIS2019, İstanbul, Turkey, 25 - 28 September 2019, pp.84

  • Publication Type: Conference Paper / Summary Text
  • City: İstanbul
  • Country: Turkey
  • Page Numbers: pp.84
  • Yıldız Technical University Affiliated: Yes


Air transportation is a major contributor to the development of countries. To achieve that, the

airports should have crucial features such as sufficient capacity, accessibility and meeting the

needs of passengers and airways, etc. In 2003, air transport sector has developed tremendously

with opening of the domestic lines to the competition. This conversion leads decreasing in ticket

prices. Also, the convenience of airway transportation and the speed of transportation have led to

large increases in the number of passengers. With the effect of the building new airports and

increasing in the domestic connecting flights, international airports, especially Atatürk Airport,

started to provide services beyond capacity. Hence, to prevent overcrowding, the increase in the

number of passengers has to be considered. This is only possible with correct modeling. In this

study, Atatürk Airport's total number of passengers was modeled by multiple linear regression

analysis. According to analysis results, some of the basic regression assumptions like

homoscedastic residuals was not provided. Also, it can be seen that, the residuals show a special

type of the heteroskedastic distribution named butterfly type distribution. The heteroscedasticity

detected by RCEV test which is developed in 2017. It is obvious that modeling errors will cause

many economic and social losses The analysis was repeated with the Weighted Least Squares

Regression method (WCEV Regression) due to many economic and social losses caused by

modeling errors in the heteroscedasticity examination detected using RCEV test. End of the

analysis, it can be concluded that all assumptions are achieved and butterfly distributed residual

problem had solved. Adjusted R 2 of the weighted model is found 92.1%. Finally, total air

passenger of the Atatürk Airport successfully modelled with a new weighted regression model.

As a future direction, we are planning to model new Istanbul Airport’s service and capacity

schedule in the context of econometric framework.