Diversification effect of real estate investment trusts: comparing copula functions with kernel methods

摘要

Value at Risk estimated with joint distribution methodologies demonstrates that risk is lower for portfolios of real estate investment trusts (REITs) and small-business equities compared with a single-asset holding. Benefits from diversification were largest in 2001-2003 and the smallest from 2006-2008. Previous research using Value at Risk points out the importance of model selection. Various estimation approaches affected results modestly over the entire period (1989-mid 2008). The Value at Risk is -3.1% for two copula models and -3.2% for a nonparametric empirical joint density, at a 1% probability level for weekly returns. After June 1996, the nonparametric copula model consistently returned the lowest risk estimate among the three joint distribution methods. Time-varying risk is a more important driver in the results than model specification. The highest portfolio risk was found for the period after August 2006 (weekly losses of 4.4% to 5%). The distribution-based model results were closer to the undiversified model results than in the earlier time periods, which supports the premise that contagion across asset classes characterises the post-2006 real estate bust, but is not a strong characteristic of the market over a longer investment horizon that includes growth phases of the business cycle.