Testing the weak form of the Property Market Efficiency

• Heteroscedasticity in the Repeat Sales Regression (Cont...)
  

The following returns per quarter and the resulting property price index was obtained in table 2.






But through another verification it appears that the heteroscedasticity with the holding period still exists follow the corrections described by Case and Shiller (1989). This persistence of heteroscedasticity may be caused by the correlation of the holding period with another variable in which the transformations did not account for.

In a comprehensive paper, heteroscedasticity in repeat house or property price indexes was studied by Goodman and Thibodeau in 1998. They identify that both age and the holding period is a major cause of heteroscedasticity in the estimators. They conclude that the holding period is positively correlated to the building age, based on the premise that a new building will sell much faster than an older building (Goodman and Thibodeau, 1998).

They propose the following relationship in the squared residuals.


Although the data provided only included the decade in which the original structure was built, the median date for that decade in which it was constructed was used as a proxy. Then the approximate age, based on years, was determined by using the first purchase date of the property less the approximate construction date for every observation. In the process of estimating the age of properties, there were a few observations in the sample that did not include the decade of construction and were removed from the final sample. With the age proxy and holding period known for each observation, the index was run again. The significance of the relationship as predicted by Goodman and Thibodeau (1998) is shown below.


Cont...