Abstract

Correct market risk assessment is of great importance for the investors and hence can facilitate the investment process in developing economies like these of Balkan countries.

The Basel Committee on banking supervision at the Bank for International Settlements requires financial institutions to meet capital requirements on base Value at Risk (VaR) estimates, which has made the VaR methodology a fundamental market risk management tool employed by the financial institutions.

Although it is widely used, the practicability of VaR was questioned and the traditional approaches to VaR computations – the variance-covariance method, historical simulation, Monte Carlo simulation (based on ARIMA models only), and stress-testing – were claimed to provide a non-satisfactory evaluation of possible losses for stock markets with long memory in returns.

The main research question of this paper is: is there any underestimation of the maximum probable loss earned on the next trading day assessed by traditional approaches for VaR measure when applied to capital markets showing long memory in returns? The study also brings a localized flavor by exploring Balkans Capital Markets and modifications of the classical approach that give a more precise measure could be suggested.

In this paper, we have analyzed four Balkan’s stock markets (Bulgarian Stock Exchange (BSE) – Sofia, Zagreb Stock Exchange, Belgrade Stock Exchange and Istanbul Stock Exchange (ISE) in Turkey) and one of the major European stock market (Frankfurt Stock Exchange) for possible underestimation of the maximum probable loss earned on the next trading day assessed by traditional and modified approaches for risk estimation using VaR measure.

Our results show that incorporation of long-range dependence in VaR estimation gives more-precise Risk prediction for Balkan markets where it is present.