Testing the Semi Strong form of Market Efficiency

• Data (Cont...)
  

In this next section the excess returns will be examined by dividing the properties into portfolio’s, assigned by the sign of their estimation errors, a result from the “actual” selling price less the estimated price derived from the model.

Following the steps taken by Londerville (1998), appreciation returns for each repeated sale was defined by the implied annualised internal rate of return, so that returns of property from different holding periods between properties are comparable. Then the relative annual Treasury bill rate at that time of purchase was subtracted from these annualised appreciation returns, in order to derive the excess annual returns. No rent was included, following suit from Londerville (1998); and all property returns are based on capital appreciation.

The properties are divided into portfolios according to the sign of the residuals, the table below shows summary statistics for the excess annual returns of the total sample, and the under priced and over priced properties, as determined by the valuation models.




According to the table above the trading rule can clearly identify properties with higher returns, with the logic that under priced properties will yield a higher appreciation as the market corrects for this over sight in the future sale. And conversely over priced or correctly priced properties do not appreciate as much, since they are priced closer to the actual market value.

The results also show that the estimated under priced portfolio is riskier than overpriced or correctly priced properties. This increase in volatility is described by Londerville (1998) as originating from property specific idiosyncratic risk, that some of those under priced properties are in fact correctly priced due to the inherent relative risks that are not captured in the pricing model.

To correct for these differences in risk of the two property portfolios, so that valid comparisons can be made, the Sharpe ratios are calculated for each of the portfolios. and if the semi strong efficient market hypothesis does not hold, then the differences in the Sharpe ratios should be statistically significant.