Getting accurate measurements of the changing value of Australia $3-4 trillion housing sector is crucial for households, policymakers, investors and developers. And it is a task that we have been engaged in thinking about since 2001. Today many economists and regulators believe Australia has amongst the best housing data in the world. Our crucial advantage rests in the fact that government agencies collect 100% of all home sales—there is no ‘sample selectivity bias’ in the technical jargon. Australia’s largest property information provider, RP Data Ltd (ASX: RPX), then spends up to $10 million a year augmenting that data with a rich range of supplementary statistics. By combining world-class data with best-in-breed academic research we are able to produce some of the most accurate and advanced house price indices in the world.
Notwithstanding all of the above, there will always be those who seek to put a subjective spin on the facts. In my previous article I referred to the colourful "non-expert commentary" that pervades discussions of the housing market, which is likely due to the non-institutionalised nature of the asset-class. One striking example of this is the occasional scribblings of the sharemarket and CFD promoter Kris Sayce, who often attacks RP Data in his contributions to an online newsletter, which is associated with Bill Bonner and Dan Denning's Daily Reckoning. In a January 4 missive, Sayce makes at least four highly misleading and deceptive statements, which is legal jargon for outright lies, regarding the November RP Data-Rismark Index results. In the public’s interest, these must be immediately rectified. And I would not be surprised if his claims attract a legal response from RP Data.
RP Data reports capital growth rates based on the market-leading ‘hedonic’ index (known as the RP Data-Rismark Index), which is a regression-based methodology that statistically controls for changes in the types of homes transacted over time. In the academic research literature there is a general consensus that hedonic indices are the best approach to measuring house prices so long as you have access to high quality data (most countries do not, although hedonic indices dominate in the United Kingdom). Many nations also use hedonic techniques to track changes in inflation over time.
In the old days, Australia relied on simple ‘median’ price indices. A median is merely the 'middle' observation in a sample of transactions. The problem with this method is that in any given month the median is highly volatile for reasons completely unrelated to the changing value of the housing market. If more first time buyers are in the market buying cheaper homes, the index may artificially fall. If, on the other hand, there is a surge of upgraders purchasing more expensive properties, the index may artificially rise. You can also find these 'compositional biases' arising from more homes trading in specific geographic regions (eg, West Sydney), or changes in the composition of the housing stock over time (eg, a shift towards higher density accommodation).
Two final sources of bias for a median index are the fact that people renovate their homes over time, which may increase the median even though the underlying capital growth rate has not changed, and because the types of properties built can vary over time. In the past, we have, for example, tended to produce larger houses, which can artificially inflate the median.
Decades worth of academic research has demonstrated that a hedonic index is the best way to control for all of these biases (conditional on having highly granular data on the ‘attributes’ of individual homes—ie, number of bedrooms, bathrooms, land size etc).
Following calls from the RBA and the Treasury in the early 2000s for the private sector to develop more advanced house price indices to overcome the median price problem, RP Data and Rismark combined to produce our benchmark hedonic measure, which has become the most closely followed index in the country. We also produce a range of other 'repeat-sales' and 'stratified median price' techniques (the same as those published by Case-Shiller in the US and the ABS and APM here in Australia), which are not publicly reported but are available to anyone on request. As a minimum, every sale included in the hedonic index has high resolution data on the home’s land size, street address, longitude, latitude, number of bedrooms and number of bathrooms.
Returning to the irrepressible stock promoter, Kris Sayce, the first mistake he makes is the claim that the 17% capital growth observed in Melbourne in 2009 is wrong. Sayce tries to refute this by simply comparing the median price of Melbourne homes from RP Data’s November press release with the hedonic index value in the October release. Yes, I know this sounds odd: these two numbers represent entirely different things. If Sayce had taken the time to read the notes at the end of the November press release, which he subsequently appears to have done, he would appreciate the error he had made.
For Sayce’s benefit, the 17% capital growth rate has nothing to do with ‘median prices’. As any review of the media material or RP Data’s website shows, this derives from RP Data’s hedonic index. The comparison Sayce makes between the November median price and the hedonic index value is absolutely meaningless. He was probably not helped by the fact that the hedonic index value was originally based at inception on a median estimate, so it looks superficially similar to a median price. But the footnotes to RP Data’s November media release explain the difference between these two variables.
Now the reason there are ‘median prices’ in the press release accompanying the November hedonic index results is because the media likes to talk about the ‘cost of housing’ at any given in point in time. The median simply tells you what the middle or 50th percentile transaction is for the sales observed in the relevant period. Journalists like to talk about Sydney or Melbourne houses 'costing' a certain amount of money, and to facilitate this RP Data has committed to releasing medians based on sales over the previous three months. The reason three months is selected is because if you took the median in any single month it will move around significantly due to the compositional biases I outlined above. Again, the medians tell you little about true growth rates (particularly in the short-term), which can only be determined by referencing the hedonic index results.
The second mistake Sayce makes is he informs his readers that “the median figure received a nice little boost thanks to the three Melbourne properties that each sold for more than $20 million in November.” All this shows is that Sayce does not know the difference between an 'average' and a 'median', which is, to be frank, prep school maths. If he did, he would not have made this error. The median is, by construction, completely unaffected by the value of homes selling at the top or the bottom of the market. As the RP Data media release explains, it simply reflects the middle or 50th percentile transaction. An average, by way of contrast, is influenced by all transactions including the outliers. This is precisely why statisticians like to use medians. The fact that Sayce cannot distinguish between the two raises questions about his financial literacy, which I will return to shortly.
In this regard, Sayce further states, “[W]e’ll be intrigued to see how the December median house price numbers for Perth are reported and whether any mention is made of the $57.5 million sale of iron ore magnate Angela Bennett’s mansion.” Once again, Sayce confuses medians with averages—Angela Bennett’s sale will have no impact at all on the median price of Perth property. If Sayce had contacted us and asked what a ‘median’ means, we could have helped him avoid making this blunder.
The third mistake Sayce makes is that he compares the median price of properties in Melbourne for the month of November, which was reported by The Age newspaper, with the median of all sales over the three months to end November. Understandably, these two figures are quite different given the inherent volatility associated with medians. Furthermore, these medians tell us little about capital growth rates, which is what the hedonic indices report. They are simply published to facilitate the media’s need to be able to say that the cost of a Melbourne home during this period was a certain value. (The Age opted for the medians attributable to the month of November as opposed to the three months to end November.)
The final mistake Sayce makes is that he claims that because the RP Data press release referenced ‘median prices’, RP Data is no longer publishing its hedonic index. As before, this statement is categorically incorrect.
If Sayce had actually read the media release he would know that the index reported by RP Data remains its market-leading hedonic benchmark. For the avoidance of doubt, here is one quote from the release:
“The RP Data-Rismark Hedonic Indices benefit from exclusive access to the most comprehensive property database in Australian and NZ, which is owned by RP Data Limited (ASX: RPX). RP Data spends over $9 million annually collecting new property information and has amassed a database comprising over 130 million property data records.”
In the past, median prices were not published in the release. To accommodate journalists’ demands, and to avoid some confusion between the hedonic index value, which looked similar to a median price (and was unhelpfully labeled), and the actual medians from all observed sales, RP Data now publishes both.
Sayce’s difficulties with basic maths are also evident in a previous article he published in which he tried to argue that the 'net' return to investing in residential property was low even if one assumes reasonable price growth. Sayce supposed that a property owner had a $10,000 deposit, which he used to purchase a $160,000 home given a $150,000 loan. He tells us that while this property appreciates to $363,200 over a 10 year period, the “return on that median Australian house…is a not so impressive sounding 30%. Or 2.6% per annum…When you’re leveraged up to that extent, house prices have to double just for you to break even on your ‘investment.” The fatal flaw in this analysis is that Sayce is not calculating the home owner’s actual returns. He is looking at the after-cost returns assuming that the borrower has no leverage (but still factoring in the cost of the leverage—I know, odd). As he contradictorily points out, this borrower actually had a lot of leverage—his initial equity contribution was only $10,000; hence he used $150,000 worth of debt to purchase the property. So Sayce’s return estimates are wrong by a factor of more than 10.
Another damning indictment of Sayce's claims is that despite using very different methodologies, and in some respects different datasets, all of Australia's property indices are broadly saying the same thing: house prices recovered strongly in 2009. And while the indices can vary markedly on a month-to-month basis, the 2009 year-to-date results attributable to the ABS, APM and RP Data-Rismark are strikingly similar.
As I noted at the start of this article, Australia benefits from unusually high quality residential property data. Because of our stamp duty system, government offices in every state and territory collect 100% of all home sales that are executed across the country. This means that in contrast to, say, the US where the Case-Shiller index is based on only a ‘sample’ of sales, which miss around 30-40% of all US states, RP Data records every single sale.
If you care for what genuine experts think, enclosed below are some recent conclusions from economists and independent research groups.
In May 2009, CommSec’s chief economist, Craig James, commented, “The RP Data-Rismark index has emerged as Australia’s authoritative source on home price trends. The property database is Australia’s largest and, unlike the Bureau of Statistics, all properties are counted, not just free-standing homes.”
Macquarie Bank’s interest rate strategist, Rory Robertson, has also observed, “RP Data-Rismark’s monthly estimates are more timely and reliable than the ABS’s quarterly readings.”
And Paul Braddick, Head of Property & Financial System Research at ANZ Bank, has described the monthly research report RP Data and Rismark produce on the basis of the index results as follows:
“The RP Data-Rismark Monthly is Australia’s leading monthly residential property research publication. The analysis is first-rate and backed by one of the nation’s top research teams that focus exclusively on this area. It is essential reading for individual and institutional investors that want the latest insights into the asset-class.”
The RP Data-Rismark indices have been independently audited by the Securities Industry Research Centre of Asia-Pacific ("SIRCA"), which is a not-for-profit financial services research organisation involving twenty-six collaborating universities across Australia and New Zealand. In its review of the RP Data-Rismark indices, SIRCA concluded:
"...the results of our analysis indicate that the stratified median price, hedonic and repeat-sales property price indices developed by RP Data-Rismark represent a material improvement over the simple median price series that have historically predominated in the Australian market. In this regard, it is pleasing to see private sector organisations that are committed to undertaking sophisticated residential real estate research and advancing our otherwise crude understanding of this complex market."
In addition to the SIRCA audit, Moody’s Economy.com, which is a leading global economics research organisation, has also undertaken an extensive review of the methodology, data and procedures used by RP Data-Rismark to construct house price indices. The Moody’s report’s summary conclusions, which were authored by two economics PhDs, are as follows:
“The suite of indexes calculated by RP Data-Rismark represents a significant improvement in the quality of housing price statistics available in Australia.”
“We look forward to seeing these indexes as they are released and believe that they will quickly take a central place in the macroeconomic data framework of Australia.”
“These data are more sophisticated, detailed and have better coverage than that used in the construction of existing housing price indexes in Australia. The high quality of the data makes it possible to implement not only median price and repeat sales indexes but also hedonic indexes, which up to this point had proved difficult to construct in Australia due to data constraints”
Other findings from the Moody’s report include:
“Perhaps the most exciting methodological development is the introduction of hedonic price indexes to the Australian market. This approach to price index construction controls for compositional change by obtaining information on housing characteristics (e.g. bedrooms, bathrooms, land size, suburb, etc.)...”
“Through their Strategic Alliance with RP Data, Rismark has access to arguably one of the best databases available anywhere for the construction of hedonic indexes.”
“At present no hedonic price indexes are regularly produced for Australia, a void which RP Data-Rismark aims to fill. The adoption of the hedonic approach means that a better balance can be achieved between sample coverage and adequate control for compositional change. Hedonic methods potentially offer an improvement over the house price measures currently available in Australia.”
“..the way in which the real-time data is collected, via real estate agents (over and above the traditional valuer-general and land titles sources), has the potential to solve many of the timeliness problems that has plagued measurement in this area.”
