"The VIX was introduced in 1993 by the CBOE and has been commonly referred to as the fear gauge due to decreases in market sentiment leading market participants to purchase protection from declining asset prices. As market sentiment improves, declines in the VIX are generally observed. In reality the VIX measures the markets expectations about future volatility with asset prices either rising or falling in value. With the VIX gaining popularity in the marketplace a proliferation of derivative products has emerged allowing investors to trade volatility. In observance of the behavior of the VIX we attempt to model the derivative VXX as a mean reverting process via the Ornstein-Uhlenbeck stochastic differential equation. We extend this analysis by calibrating VIX options with observed market prices in order to extract the market density function. Using these parameters as the diffusion process in our Ornstein-Uhlenbeck model we derive futures prices on the VIX which serves to value our target derivative VXX."

The focus of this dissertation is modeling volatility in option pricing by the Black-Scholes formula. A major drawback of the formula is that the returns from assets are assumed to have constant volatility over time. The empirical evidence is overwhelmingly against it. In this dissertation, we allow random volatility for estimating call option prices by Black-Scholes formula and by Monte Carlo simulation. The Black-Scholes formula follows from an assumption that assets evolve according to a Geometric Brownian Motion with constant volatility. This dissertation allows time-varying random volatility in the Geometric Brownian Motion to outline a proof of the formula, thus addressing this drawback. To estimate option prices with the Black-Scholes, the dissertation considers its expectation with respect to two potential probability models of random volatility. Unfortunately, a closed form expression of the expectation of the formula for computing the option prices is intractable. Then the dissertation settles with using an approximation which to its credit incorporates in it the kurtosis of the probability model of random volatility. To our knowledge, option pricing methods in literature do not incorporate kurtosis information. The option pricing with random volatility is pursued for two stochastic volatility models. One model is a member of generalized auto regressive conditional heteroscedasticity (GARCH). The second is a member of Stochastic Volatility models. For each model, estimation of their parameters is outlined. Two real financial series data are then used to illustrate estimation of the option prices, and compared them with those from the Black-Scholes formula with constant volatility. Motivated by a Monte Carlo procedure in the literature for option pricing when the volatility follows a GARCH model, this dissertation lays a foundation for future research to simulate option prices when the random volatility is assumed to follow a Stochastic Volatility model instead of GARCH. / Statistics