Testing and Comparing Value-at-Risk Measures in the Bulgarian Stock Market

Abstract

The purpose of this thesis is to compare commonly used Value-at-Risk measures calculated through Historical and Monte Carlo Simulations and to answer the question whether these measures adequately capture market risk in EU new member country. Data set of daily returns price for ten years period from 24 October 2000 to 30 April 2010 was collected for the following market indices: SOFIX, S&P 500, NASDAQ, OMXS, FTSE 100 and DAX, to give representative overview of the developed world markets and compare them with the new EU member state Bulgaria. The behaviour of Value-at-Risk models with 99 % and 95 % confidence level using rolling data windows of 100 and 250 days is analyzed with the help of a range of backtesting procedures.

Employed tests revealed that the distribution of daily returns of SOFIX index differs significantly from the Normal distribution, with high kutosis and large negative skewness. Highest Value-at-Risk violation levels were observed during periods with steep volatility jumps, which indicate that the measure reacts poorly to volatility changes and underestimate risk in turbulent market conditions. Based on the backtesting results it can be derived that VaR models that are commonly used in developed stock markets are not well suited for measuring market risk in EU new member states.