Modeling and Forecasting Selected Indicators of Dhaka Stock Exchange

Abstract

This study was an attempt to build a suitable time series model to forecast selected indicators of Dhaka Stock Exchange (DSE). Time series data of STR, IMC, TEC during 1990-2012 in annual scale; DGI, GDP, GNI, GS, GI, DIR and GFI during 2005-2012 in annual scale; capital, volume, value, trade, DGI during 2004- 2013 in monthly scale; and DSEX, DSES, and DSE30 indices during 2014-2018 in monthly scale were used for modeling and forecasting purposes. The data were collected from the World Bank and DSE websites. Exploratory Data Analysis (EDA) was used to uncover the hidden information carried out through the observed data. The time series plot showed that DSE indicators had a rightly upward trend over time but non-seasonality was present in the series. Cobb-Douglas (CD) functional regression of STR on IMC and TEC was estimated. There was a negative STR trend during the period from 1990 to 2012. There was no multicollinearity problem among the regressors in the CD regression model. The estimated residuals of the CD model satisfied that the model was free from the problem of outliers and also confirmed the normality condition. To investigate the indirect and long-run impact on the portfolios of DSE prices, the multiple log-linear regression model was estimated considering the DGI as the dependent variable and the macroeconomic indicators like GDP, GNI, GS, GI, DIR, and GFI, respectively as the independent variables. A negative DGI trend (α = -44.936) was found during the period 2005 to 2012. Multicollinearity, normality, and outliers were checked for this model too. For the multicollinearity problem, the multiple linear regression model was re-estimated by dropping GNI due to very severe multicollinearity and for severe/moderate multicollinearity, standardized GDP, standardized GS, and standardized GFI were used as the explanatory variables. A positive DGI trend during the period 2005 to 2012 was found.