Forecasting SASX-10 Index Using Multiple Regression Based on Principal Component Analysis

Abstract

In this paper we forecast SASX-10 Index (SArajevo Stock Exchange Index 10) by using multiple regression based on Principal Component Analysis scores (PCAS). In order to forecast stock market index SASX-10, as dependent variable, we use multiple regression and various macroeconomic indicators as independent variables to investigate indicators that significantly affect the performance of stocks actively traded on the Bosnia and Herzegovina (B & H) financial market.
Initially, the sample of study covered 17 macroeconomic factors as independent variables but we chosen in our model 9 statistically significant factors as independent variables (p < 0.05). After that, we have used multiple regression based on PCA scores to establish a meaningful relationship among various explanatory variables identified through the empirical analysis considering the available research studies. This paper provides an econometric analysis of the valuation SASX-10 Index.
Principal Component Analysis was used to reduce large number of explanatory variables and we have taken into consideration the multicollinearity problem among different independent variables. The main objective of this study was to forecast the value for SASX-10 Index using a multivariate statistical approach, Principal Component Analysis, to classify predictor variables according to interrelationships and to predict SASX-10 Index. For this purpose, PCA scores of 9 macroeconomic indicators were used as independent variables in multiple linear regression model for prediction of SASX-10 Index.
We have got some relationships of macroeconomic indicators with the SASX-10 market index. The result shows that the empirical characteristics of the SASX-10 Index are determined by the CPI, BIRS Index, SASX-10t-1 Index, CROBX10 Index, ATX Index, FTSE Italian STAR Index, SBITOP Index, KM/HRK and M1. Finally, we create four models with their loss function. After that, we compare loss function of all created forecasting models and the model Forecast 1 has a minimum of all loss function.
As it can be seen, 81.10% of variation in SASX-10 can be explained by explanatory variables. Accordingly, we forecast SASX-10 Index closed price for the period 01/12/2014 through 31/12/2014 by using four models.
Key words: Forecasting; SASX-10 index; Multiple regression analysis; Principal component analysis