Risk Modelling Pakistan Stock Market within the Context of Shanghai Stock Exchange

Abstract

This study presents a comparative predictive performance of different VaR models used by Securities and Exchange Commission of Pakistan (SECP) and literature recommended simple VaR models for Chinese Stock market for Chinese Securities Regulatory Commission (CSRC). In this study, we have used Historical Simulation as Non-Parametric Method, GARCH, Variance-Covariance and EWMA as Parametric and Filtered Historical Simulation as Semi-Parametric Method to calculate VaR. Another aspect focused in this thesis is China Pakistan Economic Corridor (CPEC). This study has tried to check wither CPEC create impact on risk modelling in KSE and SSE. So, we have examined the more accurate predictive model of Value at Risk for Pakistan and China before the time of CPEC and after starting these mega projects. For this purpose, we have used the data of Karachi stock Exchange 100 Index of Pakistan and Shanghai Composite Index of China from 1 st Jan 2001 to 31 st Dec 2016. Furthermore, this data is divided by using a breakpoint of CPEC 20 April 2015. The VaR is calculated by different approaches before and after CPEC for both stock markets.The major findings of the study showed that VaR models of GARCH family like ARMA-GARCH, FHS-GARCH and EWMA provides more accurate results for both stock markets. Finding shows that VaR models used for KSE 100 index and SSE composite index have been changed before and after starting of CPEC. Despite all of this the risk modelling of both countries have not been changed completely. So it does not mean that CPEC “Game Changer” have no impact on the financial markets of Pakistan and China. Because the main limitation of my study is non-availability of complete data of both stock markets after CPEC, due to the in-progress projects. When all of these projects of infrastructure, energy and communication will be complete then this change will be more prominent. At that time, this study will be a foundation for the researchers. Furthermore, models like EVT, Capula, Monte-Carlo should be applied to have more sophisticated models.