Effective risk management is a very important issue for financial institutions, their investment decisions and regulation. The necessity of accurate risk measures has also been stressed by the recent global financial crisis. The exposure to market risk of a financial institution is measured by value at risk (VaR). It is a standard tool, which represents the maximum potential loss of the asset or portfolio value within a given time period at a given confidence level (usually 1 % or 5 %). In other words, VaR is a particular quantile of future portfolio values conditional on current information. Although it is widely used due to its conceptual simplicity, it is necessary to investigate value at risk dynamically as the volatility of assets exhibits strong dynamics. The nature of risks and also the distribution of portfolio returns change over time; therefore there is a need for models accounting for time-varying conditional quantiles. The aim of my thesis will be to examine whether dynamic modeling of VaR on Central European stock market provides more satisfactory results. To do this, I would like to apply conditional autoregressive value at risk (CAViaR) proposed by Engle and Manganelli (2004) and evaluate its predictive performance in comparison with other value at risk models. To my best knowledge, there has not been published any academic work focusing on applying this model on CEE market so far.