The Unconditional Blog

The subject of property statistics attracts a large amount of comment and in turn opinion. This is in large part due to the importance to which people nowadays ascribe to these stats as indicators of the health of the market and in some ways as a surrogate of consumer sentiment: house prices rising – consumer sentiment rises; house price falls – consumer sentiment falls.

When it comes to the total number of property sales and total number of listings of property on the market – there is no debate, the numbers are clear unequivocal and objective; however comment on a median price or an average price and the waters will part and the two opposing sides will commence their passionate debate as to the merits of each. In putting together the recent first edition of the NZ Property Report we thought long and hard about the judgement of describing the asking price as a median or average.

Whilst the definitions of each are well documented they do veer towards the academic, so in simple terms the merits of each can be described as such.

• The average or mean price is just that the total of all sales divided by the number of sales. However its major failing is inherent in the diversity of property for sale in NZ – from $150,000 leasehold units to $15m luxury properties – each can when the sample size is of the order of 5,000 to 10,000 properties, cause significant swings in the final average.
• The median price is the mid-point price of a range ranked from smallest to largest – this has the benefits of effectively ignoring the extreme variances of sales highlighted above. However there is a flaw in median pricing which has been highlighted most conspicuously when it comes to the asking price statistic and that is something called “lumpy data”. The fact is property marketed is usually priced increments of between $10,000 and $50,000 – typically you will find an asking price of $320,000 or $350,000 seldom will an asking price be $332,567! – this lumpiness can and does lead to the occurrence of the median moving by as much as $20,000 for the sake of 1 extra listing.

This dilemma has lead us to seek the advice from academics and economists who are as we have found most interested in this recent lead indicator data of asking price. From these discussions has emerged a recommendation for a new measure of asking prices for properties be adopted. This new measure is called a ‘truncated mean‘.

The logic of switching to a truncated mean is that in our judgment it affords us the best of both worlds – a more accurate measure of price with less influence of extreme values. A truncated mean is calculated by eliminating a percentage from the tails of the price distribution curve. The best method of explanation is to look at the data for March as shown in the chart below:

The chart details the distribution of the asking price of the 13,284 listings which were added to the website in March 2009. The application of a truncated mean at 10% results in properties priced with an asking price of less than $200,000 and property with an asking price of over $1m are excluded ( the actual cut off point is not so clear cut but for explanation these points are used). The important point is that the 10% exclusion point at the ends of the data ensure that as prices moves so the sample is based on the most representative 80% of the market.

The choice of 10% as opposed to 5% or 15% both of which were investigated is as a result of detailed analysis of the data set going back to 2007. The rationale is that the more the extremes are removed the more the truncated mean approximates to the true average asking price for the majority of property however as ever there are diminishing returns. At 5% there are still too many extreme priced properties, equally at 15% the data set is significantly reduced by 30%. That is why 10% is judged to be appropriate given the scale of the data set and the makeup of the listings.

To provide some insight into the impact of this new measure as opposed to the existing measures the chart below provides some clarity.

A key test of this new measure is shown when hypothetically some changes are made to the base data as a function of extreme sales results.

• Scenario 1 – within the data sample of 13,284 listings just 2 new listings are added both of which are priced at $5m. The result – no change in the median price – no change in any of the truncated prices, however the average price would increase by $1,000 – all for the sake of just 2 listings
• Scenario 2 – within the data sample of 13,284 listings just 5 new listings are added, all of which are priced below $200,000. The result is that the median price falls from $399,000 to $395,000 – no change in the truncated prices and the average price falls very slightly by $200.

These scenarios highlight the vulnerabilty of the data of asking price to movements in listing data sets that are in many way unique to asking price data, that is why with the new NZ Property Report to be released on the 1st of May and all subsequent NZ Property Reports will utilise this new truncated mean to present the asking price data, both current and retrospectively.