Abstract
Executive Summary. This paper examines the gain of housing portfolio efficiency obtainable through a mixed portfolio by combining geographic characteristics and high-tech industry activities across 40 metropolitan areas. A Bayesian stochastic search is conducted for model restriction selection to compute the efficient covariance matrix of the high-dimensional panel-data model. Quadratic programming of Fortran/IMSL subroutines is applied to simulate the risk-return efficient frontier of various diversification strategies. The evidence shows that the mixed diversification strategy outperforms the geography-based strategy. The gain is superior and can reach as high as 50% in relative risk reduction during high-tech cycle growth periods.