Testing the Semi Strong form of Market Efficiency
 

In the second part of this paper the semi strong form of market efficiency is to be tested, by testing whether future excess returns can be identified and exploitable from all publicly available information. In the recent paper by Jane Londerville (1998), she used an intuitive trading rule adapted from Linneman (1986) to test this form of market efficiency. In that paper she identifies potentially underpriced apartment buildings from past and current information in actual sale prices, and tests whether those underpriced properties exhibit higher future excess returns. A modified version of this method is adopted to test the hypothesis of a semi strong form of market efficiency in the Auckland residential property market.

• The Trading rule Method

To identify underpriced properties Londerville (1998) proposed the follow hedonic pricing model:

The log taken in the dependant variable is common in hedonic pricing applied to housing, Londerville (1998). This regression is run on repeated actual sales data in half yearly periods so that subsequent returns can be known, the resulting parameters of each period is then used to price the apartments in the following period. The explanatory variables derived from sales data in say the first half of 1971 is run on the subsequent properties put up for sale for the second half of 1971, so the expected selling prices of different properties in following periods can be known.

A resulting negative residual of the actual selling price less the expected price implies that the particular property maybe underpriced. The reasoning behind the application of this trading rule is based on the assumption that the market is inefficient in the semi strong form, that the market has wrongly priced properties and that prices fail to refect all market information at that period in time.

The model that is used by Londerville (1998) is for apartment buildings that don’t exhibit major differences in quality and specific location based differences in pricing. But for residential properties, quality and location highly influence the pricing, in that each property is much more unique. Thus the pricing model she proposes is not appropriate for use in a residential property market setting.

An alternative hedonic pricing model is used for the purpose of estimating expected residential property prices in Auckland.


This pricing model is adapted from Dotzour et al. (1998); in that particular paper they conducted a study on how real estate auctions in Christchurch, New Zealand impact residential property selling prices. For that purpose they constructed a hedonic pricing model that is applicable for a study in Auckland residential properties, since the data was sourced from the same company and contained fields that were also the same. The auctioning dummy variable and interaction variables were removed from the original model. And after some trial regressions, explanatory variables such as age and wall condition that did not improve on the explanatory power in Auckland property prices were also removed, resulting in the model above.

A further modification was also made to Londerville’s trading rule; the periods in which the regressions were run and estimated were shortened to year quarters for increased accuracy in the timeliness of market prices. Since in a long time period of half a year, the real estate market prices maybe more volatile and the pricing model may fail to make accurate estimates for the different properties in the next half year.