Foreceasting BIST-100 price index

---

Thesis Type: Postgraduate

Institution Of The Thesis: Middle East Technical University, Faculty of Economic and Administrative Sciences, Department of Economics, Turkey

Approval Date: 2017

Thesis Language: English

Student: Buğra Yetginer

Supervisor: DİLEM YILDIRIM KASAP

Abstract:

The ultimate goal of this study is to forecast the BIST-100 Price Index using its mostly significative macroeconomic and financial determinants. For this aim, we have adopted an exhaustive search algorithm which takes the advantage of theoretical candidate variables to find the possible effects of these variables on the BIST-100 Price Index. The algorithm, which is built in the form of linear ARIMAX models, is to exploit every possible combination of explanatory variables to capture the behaviour of the index over the time period from 2002 to 2013 using monthly based data. To this end, best models have been obtained out of a huge number of models with regard to Akaike and Bayesian information criteria. The model with minimum AIC value outperforms the model with minimum BIC value with respect to root mean square error measure. Moreover, the 2Y Turkish bond interest rate, the DAX and Bovespa Indices are the best explanatory variables found to estimate the index. Besides, out-of-sample testing has been implemented over the 2014-2015 time period.