In emerging markets, prices of financial assets might display immediate and high changes. High volatility in returns leads investors to take nonlinear decisions on risk-return balance by disrupting their perceptions. Traditional Value-at-Risk (VaR) models might have difficulties in capturing the fat tails emerging from high volatilities in the asset prices. This study by employing daily closing values of the Istanbul Stock Exchange (ISE) National 100 Index between January 2003 and February 2007 estimates value at risk using expected shortfall with conditional threshold. Performance of the model is compared to those of Generalized Autoregressive Conditional Heteroskedasticity (GARCH) with normal distribution and Generalized Pareto Distribution (GPD). The results reveal that expected shortfall with conditional threshold has better estimation performance in the middle and long run. The results of the backtests imply that expected shortfall with conditional threshold is more proper for the emerging markets because of their ability to capture the value-at-risk from a higher level. This study has originality in that it is the first research applying the expected threshold with conditional threshold effect into the Turkish equity markets.

The value-at-risk (VaR) calculated by parametric methods that are based on the assumption that the return on financial assets has a normal distribution results in deviations in foresight performance in financial markets like Turkey which have high volatility. The assumption that the returns on financial assets in the developing countries have non-linear behaviors usually causes to the emergence of misperception in risk appettite.

The first criticism of calculation of risks based on the assumption that financial assets have normal distribution came from Mandelbrot (1963), who claimed that return on shares displayed thick tail behaviors and that the assumption that their distribution is normal would mean the undervaluation of risk. Contrary to the assumptions of the classical financial theory, neither the investors display rational behaviors nor the risk-return balance has a consequent linear distribution in the financial market.Normal distribution and likelihood theory forecasts the likelihood of a shift of 5% or more on return on shares a day as almost one in a hundred thousand. However, when we take a look at the behavior of the ISE, we can be sure that price shifts of 5% or more have been experienced more than once in a year. Therefore, a need for reference to the use of methods offering modeling opportunities without the normal distribution of returns arises with good reason.

When one approaches the issue from another point of view, the need for the use of models alternative to normal distribution in risk prediction arises. To the effect that VaR calculations done with normal distribution assumptions provide risk managers with the maximum value losses that may occur within 99% confidence interval. However, prediction of the expected probable value that might occur below the normal distribution should not be the aim of the risk manager. A risk manager (or the portfolio manager) is expected to predict the unexpected losses and those immediate and high losses that might occur outside the 99% confidence interval.

The IUP Journal of Applied Finance, Risk Forecasting, Emerging Markets, Financial Assets, Traditional Value-at-Risk Models, Istanbul Stock Exchange, ISE, Generalized Autoregressive Conditional Heteroskedasticity, GARCH, Financial Markets, Financial Theory, Empirical Research, Risk Management Techniques, Parametrical Methods, Derivatives Strategy.