Abstract
This study examines the degree of persistence in the prices, returns, excess returns, and risk-adjusted excess returns in the condominium market for five U.S. metropolitan areas: Boston, Chicago, Los Angeles, New York, and San Francisco, using fractional integration techniques that account for structural breaks and nonlinearity. The results show that the prices, returns, excess returns, and risk-adjusted returns for condominiums across the five metropolitan areas are highly persistent, with orders of integration much greater than one in a vast majority of the cases. Such results challenge the conventional view of market efficiency. The degrees of persistence in West Coast metropolitan areas of Los Angeles and San Francisco are much greater than those of Chicago and the East Coast metropolitan areas of Boston and New York City.
Notes
1 This is particularly important for mortgage default modeling, which in general assumes market efficiency (Lean & Smyth, Citation2013).
2 Following Jin et al. (Citation2014) and the suggestion of a referee, we define annual returns as the log of condominium prices at time t minus the log of condominium prices at time t - 12. Excess annual returns are given as the annual returns minus the 3-month U.S. Treasury bill rate. Risk-adjusted excess returns are excess annual returns divided by the moving average standard deviation for the period t - 1 to t - 12.
3 Other time and frequency domain approaches (Sowell, Citation1992; Beran, Citation1995) produced essentially the same results.
4 Results for no intercept and intercept and trend are available upon request. However, when including an intercept and a linear time trend, the time trend coefficient was found to be statistically insignificant in all cases. Hence, we report only the results based on an intercept.
5 A detailed description of these polynomials can be found in Hamming (Citation1973) and Smyth (Citation1998).
6 Performing tests for the equality in the orders of integration (Robinson and Yajima, Citation2002, and Hualde, Citation2013), we obtain evidence against this hypothesis in all cases except in those reported in Panel A with autocorrelated errors.