A Flexible Model for Spatial Volatility with an Application to Chicago Housing Market (Job Market Paper)

The existing volatility models normally emphasize the behavior of prices in a temporal sense and comparatively few studies have explicitly analyzed the spatial variation of volatility. By focusing on nonlinear behavior of the spatial dependence in returns using their squared terms with a Box-Cox transformation, this paper proposes a flexible spatial volatility model which encompasses both the linear and log-linear forms. Maximum likelihood method is used to estimate this model and Monte Carlo simulations are conducted to investigate the finite sample performance of the maximum likelihood estimator. Use of the model is also illustrated by a substantive application to the housing price data in the city of Chicago. The estimation results suggest that the log-linear form is more appropriate to identify the spatial dependence in volatility.