Australian Financial Review – 12 August, 2014

Christopher Joye

http://www.afr.com/p/blogs/christopher_joye/are_we_being_misled_over_house_prices_SKNQGsIVCJwAILOuohbtbM

On Tuesday the Australian Bureau of Statistics published its quarterly house price index results, which show Australian house prices have surged 10.1 per cent over the 12 months to June 30, with values across Sydney rocketing more than 15 per cent.

But we already knew this because the most-timely indices had reported their more comprehensive data through to the end of July more than a month ago.

So, do we get too much house price information? Are monthly house price index releases excessively noisy and liable to mislead consumers? And how volatile is this asset class compared to other investments?

Recently Australia’s largest real estate information business – RP Data, which spends $15 million a year collecting property data – has come in for criticism over its monthly house-price benchmarks.

A key source of the dispute was RP’s estimate of 3.7 per cent price growth in Melbourne over the month of July, because it annualises to capital gains of 44 per cent.

“Anyone watching this space over the past few months will know that according to RP Data, the nation’s property values are bounding around like a yo-yo,” one journalist said.

This debate requires several clarifications. The first is that RP Data’s hedonic home value index, which is probably the most sophisticated measure of house-price movements in the world (and the Reserve Bank of Australia’s index of choice), is not nearly as volatile as it seems.

On an annualised basis, the index volatility is only about 3 per cent to 4 per cent, which is about half the variability of the less precise stratified median price indices produced by the ABS or Australian Property Monitors (APM), which is owned by Fairfax Media, publishers of The Australian Financial Review.

Indeed, the capital cities index published by RP Data has about half the volatility of Australian government bonds and a fifth the variability of the Australian sharemarket. In the equities market, prices can jump 2 per cent to 4 per cent in a single day.

APM produces a quarterly as opposed to monthly index that competes with RP Data. It’s fair to say that RP Data’s hedonic indices have over the years taken market share from APM, which employs a simpler, but still valuable, stratified median price methodology.

The next thing to note is that the risk to which an individual home buyer is exposed is about five times greater than the indices RP Data publishes, as I have previously gone to great lengths to explain.

Put differently, when you buy a home you get one asset in one street in contrast to the about 8 million homes that underpin RP Data’s national capital city benchmark. So whereas RP Data’s index suggested prices across the nation slumped 8 per cent over 2011 and 2012, the losses within some suburbs, and on individual homes, were much steeper than that.

Super safe investment a fiction

One challenge here is that we are just not used to synthesising high-frequency housing data. Prior to RP Data launching its monthly indices after prompting from the RBA, which had for years criticised the timeliness of the old quarterly proxies published by the ABS, APM and CBA/HIA, we only received updates on price changes every three months.

This was a problem. Home buyers have historically been duped into thinking that bricks and mortar is a super-safe investment, which is a complete fiction. The volatility of an individual home, which, as I mentioned earlier, is five times greater than the national indices you see reported, is actually similar to the riskiness of the sharemarket. Few people understand this.

Yet this statement also heroically assumes you are buying a property outright and have zero leverage (or debt). Once you factor in the fact that most new home buyers are geared, on average, 70 per cent to 80 per cent, with a mortgage you find that empirically a single property is significantly more risky than shares. This would likely come as a shock to many pundits.

In principle, we want as much timely real estate information as we can get – subject to the obvious caveat that this data contains more signal than noise. This brings us to the accuracy of the house price information we have in Australia, which the RBA thinks is as good as anything you can find anywhere in the world. This is in no small part because of the efforts the RBA made over the years to improve the quality of the indices the industry releases through numerous bulletin articles, research discussion papers, and private seminars with the index providers.

RP Data graduated its monthly index to a daily model in early 2012 after close consultation with the RBA. This significant innovation allowed us to quickly determine in May 2012 that the Australian housing market was starting to recover after a record decline in prices. The daily numbers, which, averaged over a month, are referenced in the RBA’s board minutes and its Statement on Monetary Policy, also enabled us to promptly call double-digit house-price growth in the second half of 2013, and work out that the boom was persisting into 2014.

Reliably tracking daily house prices

Yet how can RP Data reliably track daily house price changes, one might reasonably ask? Across all metro and non-metro regions, RP Data captures about 50,000 sales every month. Over time it has the benefit of including close to 100 per cent of all transactions in its indices, which overcomes the sample selection biases that plague measures in the US and UK.

To produce the daily index RP Data started statistically revaluing a portfolio representing most of Australia’s housing stock, or more than 8 million homes in total, rather than relying exclusively on the homes that sell each month, which can be very different in their characteristics to the wider market. This is the same approach we use with equities where the All Ordinaries Index tracks variations in the value of all companies listed on the Australian Stock Exchange.

More technically, RP Data employs computer-generated automated property valuation models (AVMs), which suck in vast amounts of historical and current information on the portfolio of 8 million homes, including past sales values, current features (eg, number of bedrooms, bathrooms, car spaces, land size, and the property type), granular sales listings information, and recent sales for homes in the proximate region, to re-price the portfolio every day.

This portfolio approach is also very accurate. If you compare the AVM estimates of value with the sales price of, say, the last 500,000 to 600,000 homes that sold in Australia over the past year you will find that the total difference between the two is typically less than 0.2 per cent.

{rokcomments}