Just before Christmas the Reserve Bank released a consultative document on the Governor’s idiosyncratic proposal to increase required bank capital ratios to levels unknown anywhere else in the world.    I will have some fairly extensive commentary on aspects of that (unconvincing) document over the next few weeks, but today I wanted to focus in on stress tests –  something the Reserve Bank would prefer you paid little or no attention to in thinking about the appropriateness of their proposal.

Over the last decade or so, bank stress tests have come to play an important role in assessments of the soundness of banks, and banking systems, in many countries.   Devise a sufficiently demanding shock (or set of shocks) and then require banks to test their individual loan portfolios on those assumptions and see what losses would be thrown up.     Sometimes there has been a sense of the system being gamed – the shocks and associated assumptions deliberately set in such a way that banks the supervisors want to protect don’t emerge too badly.  There were suspicions of such issues in the US in 2009, and in the euro-area stress tests more recently (I heard a nice story about the clever way one set of tests were set up to minimise the adverse results for some Greek banks).    When you are in the middle of a crisis, that sort of thing is always a bit of risk: supervisors and their political masters have rather mixed motivations in those circumstances.

But there haven’t any credible suspicions of this sort of “rigging the game” in the stress tests conducted in New Zealand (and Australia) this decade.  That is no real surprise.  Our banking systems have appeared to be in good shape, and it wasn’t obvious that there was anything the supervisors and regulators would want to hide.  If anything, with both APRA and the RBNZ champing at the bit to interfere more in banks’ choices (especially around housing finance), the incentives ran the other way (if you could show more vulnerability, your case for intervention was stronger).  I was still at the Reserve Bank when the first results came in for the stress tests published in late 2014, and I vividly call a seminar in which various sceptics (me included) pushed and prodded, unconvinced that the results could possibly be as good as they appeared to be.  But, various iterations later, the broad picture of the results stood up to scrutiny.

There have been several stress test results published in the last few years (nota bene, however, that unlike the Bank of England, the Reserve Bank has not published results for individual banks.  The Bank of England approach should be adopted here –  publishing individual bank results should be a key component of disclosure and transparency.)  One of those was a dairy-specific stress test, about which I’m not going to say anything more  here (I had a few sceptical comments here).

The other two stress tests  are more useful in thinking about the overall soundness resilience of the banking system, in the face of severe adverse shocks.

The first set was published in late 2014.   This is how they described the main scenario

In scenario A, a sharp slowdown in economic growth in China triggers a severe double-dip recession. Real GDP declines by around 4 percent, and unemployment peaks at just over 13 percent. House prices decline by 40 percent nationally, with a more marked fall in Auckland. The agricultural sector is also impacted by a combination of a 40 percent fall in land prices and a 33 percent fall in commodity prices. The decline in commodity prices results in Fonterra payouts of just over $5 per kilogram of milk solids (kg/MS) throughout the scenario.

Auckland house prices were assumed to fall by 50 or 55 per cent (as large as the biggest falls seen anywhere).   In a 2015 commentary on these stress tests I pointed out just how demanding this stress test was, especially as regards the increase in the unemployment rate (around 8 percentage points).

My point is simply to highlight that the Reserve Bank’s stress tests were very stringent, using an increase in the unemployment rate larger than any seen in any floating exchange rate country in at least 30 years.  It is right that stress tests are stringent (the point is to test whether the system is robust to pretty extreme shocks)  but these ones certainly were.  And yet not a single one of big banks lost money in a single year.  That might seem a bit optimistic –  it did to me when I first saw the results –  but they are the Reserve Bank’s own numbers.

No bank lost money in a single year, and –  this is the Bank’s own chart –  none of them even had to raise any new capital (none would otherwise have fallen below minimum required capital ratios).

This should have been a bit of a problem for the Reserve Bank, as they published these results –  sold at the time as an indication of a sound and resilient system – just a few months before the then-Governor launched a new wave of LVR controls on housing lending.  I wrote various commentaries on this point back in 2015, and occasionally the Governor and his deputy seemed to squirm a little (one example here), but not ones to let rigorously done stress tests get in a way of a favoured intervention, they went on their merry way.   Ever since then, they’ve been trying to convince us that their interventions further reduced the risks associated with the New Zealand financial system.

In 2017, the results of another set of stress tests were published.     Here was how they described the main scenario in that set of stress tests.

The four largest New Zealand banks have recently completed the 2017 stress testing exercise, which featured two scenarios.¹ In the first scenario, a sharp slowdown in New Zealand’s major trading partner economies triggered a downturn in the domestic economy. The scenario featured a 35 percent fall in house prices, a 40 percent fall in commercial and rural property prices, an 11 percent peak in the unemployment rate, and a Fonterra payout averaging $4.90 per kgMS. Banks were required to grow their lending book in line with prescribed assumptions, and also faced funding cost pressures associated with a temporary closure of offshore funding markets and a two notch reduction in their credit rating.

(By then, the unemployment rate was starting from a slightly lower level).

If this test was less demanding regarding the fall in house prices, it not only explicitly assumes huge losses in asset values across the full range of types of collateral banks take in their lending, but also imposed material increasses in funding costs (rather than allowing any such pressures to emerge endogenously), and required banks to keep on growing their lending through a savage recession (in which demand for credit is in any case likely to be very subdued).   If there are one or two areas where this stress test could have been made a bit more demanding, overall the test is likely to materially overstate the potential loan losses in an economic downturn of this sort, because large dairy losses and large housing/commercial losses are highly unlikely to occur at the same time.  In any serious adverse economic shock, both the OCR and the New Zealand exchange rate are likely to fall –  typically a long way.   A fall in the exchange rate acts as a huge buffer to the dairy payout, even if global dairy prices fall a long way in an international recession.   These are details –  perhaps important ones –  but they go to the point that overall the 2017 stress test was a pretty demanding one (which is what one wants –  there is no value in soft tests, especially in good times).

And again, no bank made losses, and no bank fell below the minimum capital requirements. Here is some of the Bank’s text.

Credit losses: Due to the deteriorating macroeconomic environment in the scenario, cumulative credit losses associated with defaulting loans were around 5.5 percent of gross loans. Losses were spread across most portfolios, with residential mortgages and farm lending together accounting for 50 percent of total losses. Credit losses reduced CET1 ratios by 600 basis points.

RWA growth: The key driver of RWA outcomes were (i) risk weights increasing in line with deterioration in the average credit quality of nondefaulted customers and (ii) the requirement that banks’ lending grows on average by 6 percent over the course of the scenario. RWA growth reduced CET1 ratios by approximately 160 basis points.

Underlying profit: The banking system’s net interest margin declined by approximately 50 basis points per annum in the scenario. Banks only gradually passed on higher funding costs to customers, reflecting a desire to maintain long-term customer relationships and that some customers are on fixed rates. Underlying profits remained sufficient to provide a substantial buffer of earnings that accumulate to around 550 basis points of additional capital for the average bank.

The banking system survived, quite comfortably, the very demanding test thrown at it, based on bank loan books as they stood in early 2017.  As the Bank goes on to note, the results are sensitive to the assumptions used, but the Reserve Bank had no incentive whatever to understate the potential scale of the losses –  after all, these stress test results were released when they already had their capital review project underway.

Of course, we had one more “stress test”; the actual events of 2008/09.   Going into that recession, the Reserve Bank had been becoming increasingly uneasy about bank balance sheets.  There had been several years of rapid growth in housing lending, but there had also been very rapid growth in commercial property and other business lending, and in farm lending, and a sense that not all of this lending had been done with anything like the discipline that might have been prudent.    The 2008/09 recession was pretty severe, and quite a bit of poor-quality lending was revealed (especially in the dairy sector).  And yet, of course, the banking system came through that shock substantially unscathed.   One could argue that the test really wasn’t that demanding, since asset prices didn’t stay down for long, but in a sense that was the point: even with a severe international recession, and lending standards that did seem to have become quite relaxed, we experienced nothing like the sort of asset price or unemployment adjustments that the stress tests assume.   Capital ratios then were lower than they are now –  the latter now regarded by the Bank as totally inadequate.    Really severe adverse events don’t arise out of the blue, they are typically a reflection (as in Ireland or Iceland) of severe misallocations and reckless lending in the years leading up to the reckoning.   This latter point is one that seems lost on the Reserve Bank.

You’d have thought the Reserve Bank couldn’t have it both ways.  Sure, the most recent stress test results are now two years old, but they’ve spent the last few years telling us that they are pretty comfortable with lending standards (especially after imposing their LVR controls).  What used to be a focus of particular concern –  Auckland housing –  has largely gone sideways since then, and overall credit growth has been pretty subdued.  There is no credible story they can tell (and they haven’t even tried) as to how robust balance sheets in 2017 are now such as to make it imperative –  using the coercive power of the state –  for banks to have much higher capital ratios again.

Stress tests do get a (brief) mention in the capital consultation document.  They acknowledge the results

64. Recent stress tests have found that the banking system can maintain significant capital buffers above current minimum requirements during a severe downturn. During the 2017 stress test, the capital ratios of major banks fell to around 125 basis points above minimum requirements, while earlier tests had a trough buffer ratio of around 200 basis points. However, stress test results are sensitive to assumptions on the scale and timing of credit losses, and on the ability of banks to generate underlying profit under stress.

To which one might reasonably respond with “well, sure, but that is the sort of things you are supposed to test”.

But where they get rather desperate is in the next paragraph.

And here we have come full circle, back to rather strained reliance on the US, Spain, and Ireland in the 2008/09 crisis.

This chart attempts to imply that there is something wrong with the stress test results, rather than drawing the more obvious conclusion that they say something good about the health of the New Zealand and Australian banking systems, and about the macro environments within which those systems are operating.

Take the first panel of experiences, those from the late 1980s and early 1990s.  All five countries had newly liberalised their financial systems.  Neither banks nor borrowers nor supervisors (to the extent that the latter even existed) new much about lending or borrowing in a market economy.  In New Zealand and Australia there was wild corporate exuberance.  And in the three countries where there were systemic banking crises, all that was compounded by fixed exchange rate regimes –  that misallocated resources during the boom phase and then compounded the adjustment difficulties in the bust.

And what of the second panel?   In both Spain and Ireland there was a fixed exchange rate (membership of a common currency, but it amounted to much the same thing for practical purposes), and in the United States there was a deeply government-distorted housing finance market.   None of those cases bear any resemblance whatever to the situation of the New Zealand economy (and banks) in 2019.   We haven’t had the reckless lending, and although a future severe recession will involve losses, there is no reason to think that the 2017 stress test results materially misrepresent the health of the system.  (As the Bank has noted in the past, research suggests that in most serious banking crisis it isn’t residential lending that is the problem, but corporate and property development lending. The Bank has also previously been on record highlighting the importance of the floating exchange rate in providing a buffer in severe shocks, but now they seem to wilfully downplay or ignore that.)

The consultative document is now out for….consultation (although I think few believe the Governor is serious about taking on board other perspectives, and being open to changing his view).    But a story out the other day suggests they aren’t content to wait calmly for submissions to come in, and that the Governor has returned to the fray already in a letter to a single media outlet

Orr was responding to a BusinessDesk story questioning whether the central bank’s proposed new capital requirements for the major banks amount to gold-plating.

(I’ve asked them for a copy of this letter –  clearly already in the public domain –  but even though the Official Information Act requires them to respond as soon as reasonably practicable, I’m still waiting).

And what does the Governor have to say?

Reserve Bank governor Adrian Orr says stress tests of banks have inherent limitations, suggesting they shouldn’t be relied on.

“We emphasise in our public articles that stress testing results should not be read at face value,” Orr says in a letter.

“Both the significant modelling uncertainties, and the fact that the banks know how/when the stress situation ends, limits the value of stress tests,” Orr says.

“Further, passing a stress test covering only dairy portfolios is not a meaningful indication of overall capital strength, given it is only approximately 10 percent of banks’ exposures.”

As Newsroom notes, the final point is simply irrelevant –  the Governor attempting to play distraction –  when as the Governor and anyone else interested knows the Bank has done economywide stress tests, across the entire loan books of the banks (see above).

And of course stress tests have inherent limitations.   That is one of the reasons to have as much transparency as possible about the tests so that users can evaluate for themselves just how demanding the Reserve Bank has been.   And while the Governor seems to want to imply that the limitation he highlights could understate the potential loan losses etc, there are alternative perspectives.  For example, knowing at the start of the stress test just how severe and lengthy the eventual shakeout (asset price falls, high and enduring unemployment) will prove to be may lead to more loans being called in earlier, perhaps at larger losses.    One thing we saw in the US in 2007/08 was that banks were able to raise fresh capital early on, at a time when few appreciated just how bad things were going to get.

I don’t, for example, recall anyone suggesting (for example) that the stress test results should result in a reduction in minimum capital ratios (despite the ample margins).  They are one input, but they should be something the Governor engages with a great deal more seriously, particularly when he proposes to go out on a limb and adopt capital requirements out of step with those anywhere else in the world.   And if he wants to run these arguments, it might be better form to do so in the published consultative document, than in knee-jerk responses to individual people casting doubt on his preferred option.  I can think of one or two half-decent counter arguments –  and I’ll come back to them in a later post –  but they aren’t ones the Bank has ever advanced.

Reality is that in thinking about the consultative document, on the one hand we have detailed and specific results from repeated stress tests (and the aftermath of a period of rapid lending growth and loose lending standards not many years ago), using the specifics of banks’ loan books as they stand.   A stake in the ground as it were.   And on the other hand, we have numbers that –  despite pages and pages of the document –  are really just plucked out of the air –  both the proposed requirements themselves, and the economic and financial consequences if those whims eventually form policy.

But more on some of those issues over the next few weeks.