I was late getting round to reading the whole of the Reserve Bank’s consultation document, that backs its bid to persuade the Minister of Finance to agree to authorise them (at some future time) to impose debt to income limits on banks’ mortgage lending. I’d heard from some people who’d read it that it wasn’t very good, but even so I was surprised how weak the document making the Bank’s case is. This post isn’t a substantive response to the body of the document, which will probably come in a few posts over the month or so until submissions close. Today I wanted to focus on just one assumption they make.
The Minister of Finance insisted that the Reserve Bank include a cost-benefit analysis in the consultation document, and one that was a bit more than the usual Reserve Bank effort (an unweighted list of unquantified pros and cons). It is hard to do so when they aren’t wanting to impose the control right now, but they made a valiant effort. The value in these things is not in the precise bottom line number (inevitably wrong), but in forcing regulators to spell out their assumptions.
In their cost-benefit analysis, the Reserve Bank assumes that a DTI type instrument can reduce – by a third – the risk of a financial crisis. And they assume that (a) financial crises are really expensive (lost GDP) and (b) that in addition to reducing the probability of a financial crises, a DTI instrument can reduce – by a quarter – the severity (again, lost GDP) of such a crisis. If all three assumptions aren’t correct – if, say, a DTI instrument could reduce the probability but not the cost, or vice versa, or if a plausible crisis wasn’t as costly as the Bank assumed – the expected net benefits shown in the paper would simply evaporate.
So how costly are financial crises (especially one concentrated in developments around housing) in moderately well-governed market economies which (a) have their own monetary policy, and (b) haven’t run up against hard fiscal constraints? The Reserve Bank assumes a cumulative loss of 20 per cent (of a single year’s GDP) – and they describe that as “conservative”, meaning towards the lower end of a plausible range.
The honest answer is that we don’t really know. The relevant historical sample (of such crises) is exceptionally small. And even when a financial crisis happens, it is hard to disentangle the contribution of the financial crisis itself from adjustments that would have happened anyway.
Of course, there is the United States in the last decade – the case that grabbed everyone’s attention at the time. Plenty of writers since have described it as ‘the worst financial crisis since the Great Depression” – in some respects (narrow financial system stresses) one could mount an argument that the recent episode was worse. The Reserve Bank constantly like to invoke Ireland, but while that case study might be useful for some purposes, it isn’t for this one. Ireland gave up its own monetary policy when it joined the euro, and so had little or no scope for any stabilising macro policy when the crisis hit.
So lets have a look at how things unfolded in the United States. They had a nasty recession but they weren’t alone in that. So one benchmark might be to look at how the US relative to, say, other moderately well-governed floating exchange rate countries, and especially ones that had lots of housing debt and house price inflation but didn’t have a domestic financial crisis. Australia, New Zealand, Canada, and Norway seemed like a nice subset of such countries.
This chart uses IMF WEO annual data. It shows real GDP per capita for the US normalised to 100 in 2007, the last year before the recession (and before the financial crisis itself intensified). And it shows the average for the four rising house price non-financial crisis countries on the same basis.
Sure enough, the US recession was deeper than that in the average of these other four floating exchange rate countries which – despite the debt and run-up in house prices – avoided both housing busts and financial crises. But the cumulative gap between the two lines (ie adding up the differences across the nine years) is just under 10 per cent, which isn’t even quite half of the “conservative” assumption the Reserve Bank is using.
Of course, even among these four countries there are some quite different experiences: Australia didn’t have a real GDP recession at all, and Norway still hasn’t regained the level of per capita income they had in 2007. That is why it helps to average across a range of non-crisis countries.
Is it a fair test? If anything, I think the simple difference between the two lines errs towards overstating the costs of the US financial crisis. After all, the US ran into the effective lower bound on nominal interest rates. Standard Taylor-rule prescriptions would have had the Fed cut interest rates a lot more than the 500 basis points they did cut by (a nice chart I have in front of me from the Boston Fed illustrates that in the previous six easing cycles the Fed had cut by an average of more like 800 basis points). And the US went into the crisis with much less fiscal leeway than our fairly unindebted comparative sample. And, as it happens, each of the four comparators benefited from average terms of trade in the years since 2007 that were higher than those in the previous half decade or so. By contrast, the terms of trade for the US have been weaker than they were in the pre-crisis years.
Of course, if I compared Iceland with the four non-crisis countries, I could come out with a number that exceeds the Reserve Bank’s 20 per cent loss estimate. But the Icelandic crisis (a) wasn’t concentrated on housing, (b) was an order of magnitude more severe (in its own financial system) than the US one, and (c) the Icelandic government ran into severe policy constraints, including exhausting their capacity to borrow. It is an important case study, but it isn’t the sort of crisis we should be thinking about in contemplating the possible use of DTI controls here. Arguably, even the US experience is only somewhat enlightening given that an oversupply of houses was a significant element in the US experience. An oversupply of houses might be fine thing here one day, but it seems unlikely to be an issue here or in other Anglo countries while tight land-use restrictions are in place. But that is an issue – not touched on in the Reserve Bank paper – for another day.
If a reasonable “cost of crisis” were, say, a third lower than then Reserve Bank assumes then, on their assumptions about everything else, there are no net benefits from a DTI instrument.