Investigation of the effects of financial earthquakes with wavelet transformation time series modelling: Case study for BIST 30

In 1900s, in several countries financial and economic crisis are occurred in consequences of the globalization, emerging financial markets, money and capital markets, international trade and increasing competitions in the international markets, investing and developing new derivatives products and insufficient legal infrastructures. Depending on the incremental conjuncture, used financial modelling techniques supported to develop new analyzing methods. The aim of this study is to analyze and forecast the effects of the domestic chaos in Turkey financial market as of 2013 using the wavelet theorem. Within this scope, giving information to the investors related with the predicted magnitudes of cycles and expected duration of the crisis period and also rewarding return forecasts, has been purposed. In addition, the wavelet transformation is a new technique in the literature, this is the first academic study which inspects Turkey financial market in between 2013-2017 terms by applying wavelet transformation with GARCH and ARMA models to the BIST 30 index. It is made a significant contribution to the literature with this study. In accordance with this purpose, firstly the stationarity of BIST 30 index prices were examined and index prices made stationary, then dataset were decomposed into approximation and detail series. After reconstruction of the decomposed series, the autocorrelation and heteroscedasticity were detected in the residuals. By using logarithmic likelihood, Akaike and Bayes Information Criteria's, the best fitted ARMA and GARCH models were determined. By applying inverse wavelet transformation, modelled reconstructed series were composed and forecasted then compared with the real index prices. Calculated MAE and RMSE percentages showed that the forecasted prices transformed by wavelet theorem were more accurate than general heteroscedasticity models.