Wednesday, 30 October 2013

Household debt and house prices

Steve Keen has an article on household debt and house prices in multiple countries, using BIS data. Since Iceland is not on the list - and we arguably had a housing bubble - I sat down and did the same (or at least similar) analysis on Icelandic data. Sources are the Central Bank of Iceland, Registers Iceland and Statistics Iceland.

First, the real (CPI deflated) house prices.

Real house prices in Iceland, monthly data.

Now let's see what happens when we add the household debt data.

Keen uses the acceleration in household debt and the change in house prices. The idea is that acceleration of household debt, which is mainly mortgages, should be like a jolt of energy into the housing market. More mortgages (increased acceleration) means more purchasing power and more purchasing power means more demand which should push up house prices. So the acceleration in the level of household debt should be correlated with the change in house prices. If the correlation is high it is a sign that changes in house prices are primarily driven by new mortgages rather than "fundamentals" of any kind.

In the case of Iceland we've got data back to 1997, limited by quarterly national accounts. We can stretch the data back to 1994 when the index for house prices is first published but I'm using quarterly data here.

Icelandic household debt acceleration and change in house prices

The second graph here above makes perfect sense until about 1Q 2008. Then we can see a great divergence between the two series and after October 2008 there is all sorts of nonsense going on.

The likely reason for the apparent non-correlation after October 2008 is that the data is all messed up. When the new banks were established they got a free book-value haircut on the existing debt of up to 30% or even more. The great profits of the banks since the collapse is to a large extent due to the banks clawing back this haircut in their books: a credit contract worth 10,000,000 ISK (nominal value) was booked, when the banks were established, at 6,000,000 ISK and then revalued, in steps, up to e.g. 9,000,000 ISK, creating a nice 3,000,000 ISK profit in the books.

It's the October 2008 haircut which is the reason for the massive drop in debt acceleration around that time. The second drop in 2010 is probably due to the fact that FX-indexed loans were deemed illegal and the banks had to write them down to a large extent. Meanwhile, the divergence in early 2008 is at least partially explained with the fact that by that time the ISK was collapsing and FX-indexed loans, the same which were later found to be illegal, were rocketing up in nominal value.

So data problems are probably the reason why the correlation doesn't make any sense after 2008. Until 3Q 2008 the correlation is as we would have expected it from Keen's idea. And the value: 0.56. If we want to be selective and skip the last three quarters of 2008 when the ISK was collapsing at the fastest pace the correlation goes up to 0.74. That's high but still lower than some of the coefficients that Keen has in the case of other countries.

And today? Well, the banks began offering non-indexed mortgages (majority of mortgages in Iceland is indexed to the CPI) in 2009 and they really took off in 4Q 2009 according to data from the CBI. A large share of new mortgages is now non-indexed but their share is fluctuating somewhat. As far as I'm aware of no official data series exists which shows their share at all times.

We could check out the correlation between household debt acceleration and house prices from e.g. October 2008 or December 2009 but we would still have to deal with the rather bizarre drop in the data in 2010, probably caused by the FX-indexed being found illegal. If we however check the correlation only from 4Q 2010 we get this:

Household debt acceleration and change in house prices since 4Q 2010. Correlation: 0.85

Now, this is guesswork and should be taken with a pinch of salt! I do not know when exactly the banks book their losses due to the illegal FX loans, I'm guessing, from the data itself, that it's before 4Q 2010. Furthermore, this is less than three years of data so we should be careful in trusting that the correlation will stay there forever. Also, since the CPI includes house prices and the CPI affects the nominal value of indexed mortgages we should expect to see correlation through that link as well. Finally, correlation is not causation and all that - and this applies to the previous graph as well.

All the same, this feels intuitively too comfortable to outright deny it. It at least seems that the current housing market in Iceland is driven first and foremost by new credit being created by the banking system. I wonder how important fundamentals, such as wages, really are.

Oh, well! Back to work!