The author has been described by News Ltd as an "iconoclast", "Svengali", a pollie's "economist muse", and "pungently accurate". Fairfax says he is a "Renaissance man" and "one of Australia’s most respected analysts." Stephen Koukoulas concludes that he is "85% right", and "would make a great Opposition leader." Terry McCrann claims the author thinks "‘nuance’ is a trendy village in the south of France", but can be "scintillating" when he thinks "clearly". The ACTU reckons he’s "an enigma wrapped in a Bloomberg terminal, wrapped in some apparently well-honed abs."

Wednesday, August 31, 2011

Steve Keen wrong on house prices (again)

In my column last week, I argued that it was reasonable to believe that over the next decade credit and house price growth would be driven by disposable incomes, all things being equal.

My regular sparring partner Steve Keen responded with a detailed critique in which he rejected my arguments, and sought to re-cast his infamous September 2008 claim that “there's no point in paying a mortgage on an asset that is going to fall by 40 per cent or so in the next few years.”

Like most of Keen’s predictions, such as the ‘best case scenario’ during the GFC being 11 per cent unemployment and a recession “more severe than 1990 and lasting 1.5 times as long” (unemployment peaked at 5.8 per cent while there was no recession), his 2008 projection proved way wide of the mark.

For the record, Australian dwelling prices are today 13.3 per cent higher than when Keen put his reputation on the chopping block, and 89 per cent higher than the level at which Keen expected them to be.

Keen now tells us that he thinks Australian house prices will plummet by 40 per cent over the next decade (or 13 years since his original declaration). That actually represents a forecast downgrade given a two-fifths decline in prices from today’s level is equivalent to a 32 per cent fall in September 2008 terms.

Keen further dismisses my proposition that the growth in Australian housing costs over the last 20-30 years can be explained by simple reference to incomes and interest rates.

Instead, he points to a 50 per cent ‘correlation’ between a much more nebulous conception, which he calls the ‘mortgage accelerator’ (the rate at which housing credit accelerates or decelerates), and changes in house prices over time.

Setting aside the fact that Keen offers no evidence of any causality between these two variables, and rises in house prices could just as easily cause an acceleration in credit (rather than the other way around), Keen’s ‘correlation’ still fails to explain half of the times-series change in Australian dwelling values.

Indeed, when I look at the relationship between Keen’s ‘mortgage accelerator’ variable, which he supplied to me, and a more precisely-measured estimate of changes in dwelling values—RP Data-Rismark’s Hedonic Index—I get a much lower 12 per cent correlation.

More importantly for the Keen critique, I will show that by indexing up median Australian dwelling prices by per capita disposable incomes and changes in borrowing capacity (as determined by mortgage rates) one can account for around 90 per cent of the rise in Australian housing costs over the last two and a half decades.

If we accept that the big structural changes that took place over the last 20-30 years are behind us, we can then project with confidence that future dwelling prices will be determined by household purchasing power, ceteris paribus.

The structural changes I am talking about here include the circa 40 per cent plus reduction in variable mortgage rates from the 12.6 per cent average that prevailed between 1980 and 1995 to the 7.3 per cent level that has held since (see the grey line in the first chart below).

This was brought about by the RBA’s adoption of a so-called ‘inflation-targeting’ approach to monetary policy in the early 1990s, which has helped keep inflation, on average, within the central bank’s target 2-3 per cent band and allowed nominal interest rates to remain lower than had previously been the case.*

Consigning double-digit inflation and commensurately high mortgage rates to history has in turn empowered households and businesses to undertake a once-off upward adjustment to their borrowing capacity (or leverage) in recognition of the 40 per cent reduction in the cost of debt.

In the chart below, this can be seen by the rise in the RBA’s household debt-to-disposable income ratio from the early 1990s onwards, which coincided with the time that the RBA started using its 2-3 per cent inflation target (refer to the red and blue lines).


You can see that in 2005-06 (ie, several years prior to the GFC) Australia’s household debt-to-disposable income ratio started to flat-line.

The RBA’s (and my) central case is that this will continue: that is, the ‘new normal’ will be one where credit growth is bounded by incomes. This brings me to one of Keen’s factual errors.

In his article, Keen asserts, “My data comes from the RBA; Chris’s comparable figure clearly uses different data to conclude that “Australia’s household debt-to-disposable income ratio has, in fact, flat-lined since a number of years prior to the GFC”, since RBA data shows it rising from 115 per cent to 135 per cent from 2005 until now.”

In private discussions with Keen I have confirmed that his statement is, in fact, wrong. That is, I was using the correct RBA ratios, and he has arrived at very different (and presumably specious) numbers through his own independent calculations.

According to the RBA’s official measure, Australia’s household debt-to-disposable income ratio hit 155 per cent in 2006, which is where it remains today. It has not, as Keen maintains, expanded from 115 per cent to 135 per cent over this period. The striking deceleration in Australian housing credit growth since its peak in late 2004 can be vividly seen in the next chart (refer to the downward-sloping black line).


A third major structural change I highlighted was the extra growth in disposable household incomes over the last couple of decades resulting from the emergence of multi-income families (and the rise in the female participation rate) and the decline in Australia’s ‘non-inflationary’ unemployment rate, which is thought to be around 5 per cent today (see the blue and red lines in the chart below).


If, as discussed, these structural shifts have run their course, household earnings will be the main driver of future dwelling price changes. This in turn implies a long-term house price growth rate of about 4-5 per cent per annum, assuming normalised productivity growth and more investments in the supply-side (the absence of which could produce higher growth). This is notably less than the 7.8 per cent per annum pace that has applied since the mid 1980s

We tested this position by surveying the top 21 market economists, including the major bulls and bears, on their own outlook. As it happened, the average and median of the forecasts for Australian house price growth over the next 10 years was 4.4 per cent and 5.0 per cent per annum, respectively.

Another way to resolve this debate is by working out what proportion of Australian house price growth can be explained by incomes and interest rates. To address this question, Rismark’s research group took the median Australian dwelling price in June 1985 and indexed it up by:

(1) the change ABS disposable incomes per household before interest repayments; and

(2) the change in household borrowing capacity that arises from movements in mortgage rates (assuming a fixed, 80 per cent loan-to-value ratio and a constant disposable income-to-mortgage repayments ratio).

This ‘income- and interest rate-adjusted’ median dwelling price can then be compared to actual median prices over time. The next chart presents the results of the analysis.


Observe how the confluence of rising household incomes and declining mortgage rates over the 1990s meant that ‘purchasing power-adjusted’ median house prices (black line) rose to be well above ‘actual’ median prices (red line) for the entirety of the period between 1993 and 2003.

It was only in the years after 2003 that the median price implied by changes in purchasing power and actual prices started to diverge in the opposite direction, a trend that was brought to a halt by the 2007-08 crisis.

Today we find that around 88 per cent of the change in Australian housing costs since 1985 can be explained by mortgage rates and household incomes.

Here it should be noted that these are two ‘demand-side’ variables, and asset prices are ultimately determined by the intersection between demand and supply.

Demand- and supply-side factors not accounted for in this work include shifts in household preferences—there is, for example, evidence that families are willing to spend more of their incomes on buying housing services because it is a ‘superior good’—and, of course, the availability of housing assets (ie, supply) to satisfy demand-side needs. The latter refers to both the quantum of homes listed for sale (ie, effective supply) and the number of dwelling required to practically accommodate the resident population (underlying supply).

ANZ Bank’s economists recently completed a similar exercise using capital city median prices (in contrast to our ‘all regions’ price) and a slightly different treatment of mortgage rates. In particular, ANZ ‘smoothed out’ movements in mortgage rates over time to capture the underlying structural change.

ANZ’s results are nevertheless very similar. They find that the median house prices imputed by increases in purchasing power (via incomes and interest rates) sat well above transacted prices between 1993 and 2003. As at 2011, ANZ’s analysis could explain 93 per cent of the total change in dwelling prices since 1986 (see the next chart below).


While the powerful long-run relationship between house prices and household purchasing power will continue to hold, there is inevitably significant year-on-year variability.

In 2007, Australian capital city dwelling prices rose by 13.6 per cent. In 2008 they fell by 2.5 per cent. In 2009 and 2010 they appreciated again by 12.1 per cent and 5.3 per cent, respectively. Yet if we look over the span of the last 7-8 years, we find that the average growth in housing costs and incomes has been near-identical.

This year dwelling prices look like they will be down a bit, unless the RBA starts cutting rates, as Westpac, Goldman Sachs and Deutsche Bank expect they will.

In addition to the fact that in November 2010 the RBA boosted rates to a level above their historical average, the principal driver of the soft 2011 conditions has been the exceedingly ‘hawkish’ Australian consumer.

In February Australians expected, on average, more than three further rate hikes (ie, an 8.5 per cent plus mortgage rate), according to Westpac-Melbourne Institute analysis.

Even in the midst of the recent financial market turbulence, consumers were still banking on two more rate hikes when surveyed in August. A remarkable 29 per cent of all respondents thought there would be more than four future hikes.

When Australians eventually buy into the idea that rates are not heading higher, housing market conditions will likely improve.

*There was a cap on lending rates in Australia until the early 1980s, which makes like-for-like comparisons prior to this period difficult.