Testing the Semi Strong form of Market Efficiency

• Data
  
  

One major advantage of using a method similar to Londerville’s, is that the same sample data of property that was used from the construction of the house price index can be utilised. Since the data only includes repeated sales, the subsequent returns can be known for under priced or over priced properties that were determined by applying the trading rule (Londerville 1998). Similar data cleaning steps were made to the property samples, repeat sales data were selected only if they had no change in improvements, land, and floor area, and had a holding period of more than 90 days. Transfers were also omitted but no step was taken to restrict returns of properties that performed above or below an implied IRR per quarter of 50% or -50%.

Observations that had missing fields that were required for the regression were also omitted, the final sample consisted of a total 16,209 transactions and 8,739 repeated sales from the 5th of January 1993 to the 30th of July 2003, which can be used to run the pricing regression. For the location based aspect of pricing, the suburb of each property was assigned to designated areas based on the distance from the city centre and the relevant suburb class, the suburb of Mission bay is denoted as upper class, Royal Oak as middle class, and Avondale as lower class. The following table shows the division of suburbs into the relevant areas.



With a total of 41 regressions and 11 explanatory variables, the results for all the regressions that were made are not shown, instead an example of the results obtained by a particular regression is shown below.




The coefficients of the independent variables are of the correct signs and are mainly significant, and given the significance of the t-tests and f test throughout the periods in study. We can conclude that the model is sufficient enough for the purposes of identifying underpriced properties in the fore coming periods. It is interesting to note that for every allocated area’s, as predicted, area one has the highest premium being located near the city central, and as we go further down the area rankings the explanatory variable gives lower property prices.

These pricing models, determined by the current period were run on the next period’s properties to estimate those market values, and if residuals of the actual selling price of the property less the estimated price is negative, then that property is deemed underpriced, Londerville (1998).

By applying current and past information to identify future property prices, essentially, the semi-strong efficient market theory is tested, that all available public information should be reflected in prices (Londerville, 1998). So if the semi-strong form of market efficiency holds then there should be no underpriced properties when these regressions are run, and those that are identified below the market value, implies that those properties possess idiosyncratic risks that justifies for the lower price. And subsequently when those under valued properties are adjusted for risk, they should not exhibit relatively higher excess returns when compared to the properties that are correctly priced or over priced.

Cont...