Lies, Damned Lies, and Housing Statistics

            Get your facts first, then you can distort them as you please.
                -  Mark Twain

The Great Affordability Swindle

House price to income ratios may seem a bit dry for a first post.  Nevertheless I would ask that you bear with me as I discuss one of the cornerstones of understanding Australia's housing market.  After much discussion over the last few years, the Australian housing market bubble debate took a seriously weird turn when attention turned to debating the merits of competing 'price to income' ratio methodologies.  Even if you are not an egg-head interested in economic data, this analysis is interesting...and scary!

For many years the debate about Australia's unaffordable housing market raged.   We will spend lots of time going through these various arguments in future posts (you know the drill: housing shortage, high population growth, cultural affinity to property, strong banking sector, etc etc).  The argument about 'bubbles' is largely about affordability though. 

Those who are convinced there is a property bubble in Australia will often point to our house price to income ratio as evidence that prices are reaching an unsustainable levels and are due to correct (i.e. reduce).  In simple terms this ratio is defined as:

Price to Income Ratio = Median House Price / Median Disposable Household Income

One way of interpreting this is how many years it would take (with no other expenses) to save for a 'middle of the road' house with a 'middle of the road' income.  Sounds like a sensible test for affordability, right?

Demographia Data

One of the most helpful sources for this information in recent years has been Demographia (http://www.demographia.com/dhi.pdf).  Their annual survey publishes 'median multiples' across urban areas in the US, UK, Ireland, Australia, Canada and New Zealand.  To be clear, the survey looks at both median 'price' and 'income' levels for major urban areas in all of these six comparable developed nations, and generates a ratio of median price/income for each.  The higher the 'median multiple' in a given city, the more unaffordable it is.

Over recent years the preponderance of Australian cities in this survey among the most 'severely unaffordable' has been given considerable media coverage (if for no reason than the Australian media's unending ability to sell papers when anyone discusses the question of whether house prices are over-valued!).  The 2010 Demographia survey was the most scathing yet of the level of affordability in Australia, and was used by a number of commentators and investors to justify their highly publicized views that Australia's housing market is overvalued.  The top ten most unaffordable cities according to the 2010 survey were:

To rub salt into the wound of Australia's housing market supporters, the survey also summarized the results of its survey by country as follows:

The conclusion of this survey is that only one of twenty-three major urban real estate markets in Australia is not 'severely unaffordable' according to Demographia's methodology, and that market is still 'seriously unaffordable'.  This level of unaffordability massively exceeds any of the other nations in the survey.

Other Median Multiple Data

Other commentators, such as GMO founder and chief investment officer Jeremy Grantham, have made a lot of headlines quoting price-income ratios which, while lower than Demographia, are still much higher than historical levels and much higher than the levels seen in other nations.

Gratham has been quoted as saying: "The price of housing typically trades about 3.5 times of family income and in bubble it goes to 6 or.. 7.5 (times)”... "you are at near 7.5 times family income... which suggests you are twice the size that you should be." (The Australian, 16 June 2010)

The main difference between Grantham's numbers and Demogaphia's appear to be Grantham's use of median price for all dwellings as the numerator in the ratio, wheras Demographia appear to use house prices only,  which is a material difference. 

Nevertheless the result of both Demographia and Grantham's numbers are the same: a clear indication of the historically stretched valuations being ascribed to Australian houses.

The Empire Fights Back

Unhappy with these turn of events, and even more unhappy with a simple and straightforward comparison to international affordability metrics, the Australian property lobby fought back.  For many years the mouthpieces of the property lobby were content in relying on the 'stock vs flow' argument for why such damning data was irrelevant.  In other words, the idea that any comparison of a stock (house prices) with a flow (income) was bound to be inaccurate.  A related argument against such 'crude' measures of affordability was the idea that the level of mortgage repayments (driven by interest rates) are more important than the level of price and income themselves.  I will address these propositions in a future post, but to preview, this argument is only reasonable if one considers long term interest rates, not current interest rates - after all it is a 30 year mortgage!  Suffice it to say that many people could see through the chimera.  Thus, the property lobby had to attack the affordability ratio itself and find away to make it more pliable to their tale.

Bullish property commentators such as Christopher Joye realized that Demographia and virtually everyone that was tracking affordability was using Australian Bureau of Statistics survey data for the 'median household income' portion of the affordability ratio discussed above.  So they uncovered a 'better' source for household income: the national accounts used to calculate GDP.  As described in this post, Joye claims the national accounts defnitition of household income is better for two reasons: 1) the household survey data is "dated" being a 2007 study whereas National Accounts data is released in a much more timely fasion; and 2) "earnings on savings and investments" should be included because "people buy homes out of disposable incomes rather than wages".

As I note below these are somewhat spurious reasons to use the National Accounts data as 1) it is very easy to adjust the 2007 numbers for subsequent wage/income growth (as I do below) and 2) there are a number of sources of 'disposable income' in the national accounts which contratary to Joye's post are not actually available for people to "buy homes" (such as superannuation contributions and 'imputed' rental income).  More on that below too.

To cut a long story short, the numbers that Joye used imply that the ratio of house prices to incomes is actually closer to 4 to 5x which - surprise, surprise - is in line with other developed nations.  The most recent incarnation of this logic has been recently updated by Joye here and here, which provides a useful snapshot of the way in which this 'modified ratio' is being used as a debating weapon.  I examine the different numbers below, but the short answer on how a ~50% reduction in the affordability ratio is conjured up is that they 'find' additional disposable household income of over $20,000 compared to the ABS survey data.  Oh, and they use the average household income rather than the median which is worth about another $12,000 in annual income (because the highest income earners 'drag up' the average).  The 4-5x ratio implies that the disposable household income that should be compared with the median house price, across Australia, is ~$95,000.  This is the after tax level of household income.  With 25% effective tax rate this implies a gross income of $125,000 is what the 'median' household earns in income that can be spent on bills, food and mortgage payments. 

These numbers have been picked up by all the usual suspects wishing to argue that the affordability problem is, err, not really a problem.  Once the RBA started using this data it was on for young and old, including the big four banks who have begun parroting these same ratios in their presentations, shamelessly using them to 'alter' the unfavorable international comparisons applied to the asset class that chokes their balance sheets and income statements - residential property lending.

Westpac's effort here (pages 8 and 9, and chart 7, in particular).

CBA's effort here.

The CBA's alteration of the international affordability comparisons was perhaps the most clumsy and crude, but you have to give them marks for being brazen!  Note that my comments to the right were not in the CBA's charts released to investors, showing that they adjusted the Australian numbers but not the international comparisons.  For more background on this cheeky effort, see this Steve Keen post which picked up on Kris Sayce's initial discovery of this chimera in his Money Morning blog.

Decoding the Numbers

All of this begs the obvious question: how can two official measures of household income be so different?  Before providing a quick reconciliation of the numbers, here's what they are.

a) The ABS median household income survey

The most recent ABS household income survey was conducted in 2007/2008 and can be downloaded here.  Details of the methodology employed by the ABS in this survey can be found here.

In summary the ABS data involves a survey of individuals living in dwellings, and counts income that as a general rule is "available for consu mption" - in other words cash or cash like in nature and available to meet cash obligations.   

b) The national accounts household income data

The national accounts data is essentially designed to count all economic production and activity in every given period in order to calculate (among many other things) GNP and GDP.  These sources of economic activity are in many cases non cash sources of income to household, or otherwise unavailable to be actually spent.

In using this data to come up with a measure of 'household income' the banks have divided the aggregate proportion of Australia's GDP apportioned to the Household Sector in the national accounts by the total number of households according to the ABS.

c) Bridging the gap

The ABS published the following statement as part of its 2007/2008 household income survey:

The concepts of income used in SIH have many similarities to the household income definition used in the Australian System of National Accounts (ASNA), but also differ in some respects. A detailed comparison of 1997–98 SIH and ASNA estimates was published as an appendix to the 1997–98 issue of Income Distribution, Australia, 1997–98 (cat. no. 6523.0). Comparison of SIH data from 1994–95 to 2007–08 with ASNA data indicated that the relationship between the two estimates had not changed significantly over that period. [Link Added - from page 55]

I have used this 1997 reconciliation between the national accounts data and the household income survey to provide a bridge between the two sets of numbers.  I happily accept these are somewhat 'rounded' numbers, but the ABS has specifically noted that the relationship "had not changed significantly over the period", so this should provide a reasonable estimate.  My approach has been to take the 1997 reconciliation and scale the adjustments to the 2008 household income numbers discussed above. 

Before diving into the details, it should be noted up front that the National Accounts data showed 57% higher household (gross) income in 1997 than what was indicated by the ABS household survey.  So to be clear the order of magnitude of the difference seems to be about the same.

One additional adjustment that must be noted is that the national accounts numbers implicitly use an average household income measure, rather than a median.  This is clearly inappropriate in the context of a median house price to income ratio, and so the scale of this adjustment is estimated also using the difference between average and median incomes in the ABS household survey (which should be highly comparable).  I have also grossed up the 2008 figures to 2010 to also aid the comparability (i.e. helping the case of the housing bulls and addressing one of the original reasonse the banks etc gave for using national income data rather than household survey data).

Here is the summary of my analysis:

So what does this analysis tell us?  All of the differences between the higher income number used by the banking sector to support a lower price to income ratio are, in my opinion, very difficult to justify as being useful in an affordability metric.  For example how does the imputed rent of owner occupied homes provide any capacity to afford more expensive house prices?  The inclusion of implied income received from higher housing costs itself seems very ponzi-like, without even looking deeply into the calculation of imputed rent.  Or consider superannuation contributions (the other major difference).  How do superannuation payments, locked up until retirement to fund consumption, facilitate paying higher prices and mortgage payments for a home before retirement?  Perhaps the most egregious difference is the first one highlighted above - the national accounts definition of 'households' include private non-profit organisations like churches and clubs which have nothing to do with actual household's ability to affod their homes and mortgages!!

Finally, the whole purpose of calculating a price-income ratio is to provide some long term benchmark to assess how affordable housing is relative to historical trends, and to enable international comparisons.  The argument made by the banks in lowering the ratio by using the highest possible income figure does nothing to refute the unaffordable nature of current prices relative to historical measures of the Australian price-income ratio.  While the banks argue there are structural reasons for this (esp a lower interest rate environment), this is a very different argument to saying that the ratio is not currently high.  As for international comparisons, this blog has not conducted an in-depth study of the income measures used in surveys such as the Demographia study.  It seems unlikely, however, that relatively obtuse national accounts data would be used to estimate income in the US, UK, Ireland, NZ and Canada, when all of these nations also publish timely household income surveys (I have asked Demographia the question but no response yet). 

Don't Believe the Hype

Based on the above, this blogger thinks the following statements are irrefutable:

- Australian house prices are considerably higher (relative to our incomes) than other developed countries

- Australian house prices are at an historically high multiple of income (however measured), and this increase has particularly occurred over the last 15 years.

- The use of the modified affordability multiple by the banks and several prominent housing industry advocates is at best inaccurate, and appears to paint the rosiest picture possible of the state of the market.   This is not only because of logical flaws in the use of national accounts income as the denominator in an affordability ratio, but equally importantly because of the modified ratio's inappropriate use in international comparisons.

Padding the numbers in this way might make for pretty slides and graphs in bank presentations, but it exposes a degree of delusion and desperation that is truly scary.

And I believe it is misleading.

Note: Here are the excerpts of the key data sources I used for this analysis (source data linked).

ABS Household Survey vs National Accounts Household Income Reconciliation from 1997:

ABS Household Survey 2008 Income Data - Cat. 65230.0 (note that I have used the 'raw' income data from the ABS Household Survey in Appendix 2, not the 'equivalised' income data that is used in most of the survey analysis  - equivalised income is normalised for household size and not appropriate for this analysis):