The objective of this Master's thesis is to find out if the calibration of a univariate stochastic volatility model using plain vanilla options leads to a perfect result with the model parameters being correctly recovered. To this end, prices of plain vanilla options with various maturities and strikes are computed for different sets of known model parameters (so called true parameters) using the univariate stochastic volatility model of Heston. Given the sets of true parameters and the respective true option prices, the Heston model is calibrated, as usual, within a least square framework. Since artificially created calibration instruments instead of market observed option data are used for the calibration, an optimal solution exists (i.e. the global minimum of the associated minimization problem) for the calibration problem at which the calibrated parameters take the values of their true counterparts. Using a local optimization algorithm it is shown that the performance of the calibration process is highly dependent on three factors: The number of calibration instruments available for the calibration, the error measure chosen to compute the distance between the model and the market prices and the choice of the initial parameters for the optimization algorithm. In the empirical part of the thesis, the impact of the three mentioned factors on the calibration performance is systematically evaluated. Several conclusions, useful for practitioners and academics who apply the concept of calibration to determine the parameters of option pricing models, can be drawn from the empirical results. The more calibration instruments are available for a calibration, the higher the probability that the actual model parameters are recovered. It also turns out that the preferable error measure computes the relative rather than the absolute distance between the model and true prices and that the model should be calibrated, using Black-Scholes implied volatilities derived from option prices instead of the option prices themselves. Based on the outcome of the empirical analysis, it is, in addition, found that the closer the initial parameters are to the true parameters the better are, on average, the calibration results.
«
The objective of this Master's thesis is to find out if the calibration of a univariate stochastic volatility model using plain vanilla options leads to a perfect result with the model parameters being correctly recovered. To this end, prices of plain vanilla options with various maturities and strikes are computed for different sets of known model parameters (so called true parameters) using the univariate stochastic volatility model of Heston. Given the sets of true parameters and the respecti...
»