Volatility Estimation with Financial Data

Modeling and estimating volatility plays a crucial role in financial practice. Devoted efforts are made to investigate this topic using both low-frequency and high-frequency financial data.;Traditionally, volatility modeling and analysis are based on either historical price data or option data. Finance theory shows that option prices heavily depend on the underlying stocks' prices, and thus the two kinds of data are related. This thesis explores the approach that combines both stock price data and option data to perform the statistical analysis of volatility. We investigate the Black-Scholes model and an exponential GARCH model and derive the relationship among the Fisher information for volatility estimation based on stock price data alone or option data alone as well as joint volatility estimation for combining stock price data and option data. Under the Block-Scholes model, asymptotic theory for the joint estimation is established, and a simulation study was conducted to check finite sample performances of the proposed joint estimator.;Being more accessible than ever, high-frequency data have provided researchers and practitioners with incredible tools to investigate assets pricing and market dynamics. Non-synchronous observations, microstructure noise, and complex pricing models are challenges coming along with high-frequency data. Moreover, large volatility matrix estimation is involved in many finance practices and encounters "curse of dimensionality". Although it is widely used in large covariance estimation, imposing sparsity assumption on the entire volatility matrix is not reasonable in financial practice. In fact, due to the existence of common factors, assets are widely correlated with each other and their volatility matrix is not sparse. In this thesis, we focus on incorporating the factor influence in asset price modeling and volatility matrix estimation. We propose to model asset price using a factor-based diffusion process. The idea is that assets' prices are governed by a common factor, and that assets with similar characteristics share the same association with the factor. Under the proposed factor-based model, we developed an estimation scheme called "Blocking and Regularizing", which deals with all of the four changeless. The asymptotic properties of the proposed estimator are studied, while its finite sample performance is tested via extensive numerical studies to support theoretical results.