Testing the weak form of the Property Market Efficiency

• Data(Cont...)
  

The results of the regression in year quarters are shown in table1, and the corresponding property price index is shown in graph 1.

• Heteroscedasticity in the Repeat Sales Regression

A refinement by Case and Shiller of the BNM repeat sales regression is the acknowledgement of heteroscedasticity in model. They claim that the noise term from the log of change in the property price increases with respect to the related holding period (Case & Shiller 1989). In accordance to Case and Shiller’s method, following their corrections made for this presence of heteroscedasticity, each observation in the original model was weighted by the estimated standard deviation using the following formula.


So Case and Shiller (1989) believe that the higher the holding period of the property the higher the noise created. This model of squared residuals was run to verify the relationship predicted by Case and Shiller



The holding period in this case is the approximate months of holding for each property with a repeat sale. With a t-value of 10.6, the coefficient on the holding period can be concluded to be highly significant and indeed a presence of heteroscedasticity.

Following the steps taken by Case and Shiller to correct for this heteroscedasticity, all observations of the property transactions was divided by their respective standard deviation, estimated from its holding period. This transformation is applied to the log of the change in price and the relative rows in the dummy matrix. Then ordinary least squares were run with these weighted observations, and by following this transformation the weightings should correct for this bias.

Cont...