House of cards?

The Reserve Bank announced last month its decision to require banks to classify all loans secured on residential investment properties separately from other residential mortgage loans. This applies not just to large commercial operators, but to borrowers with just a handful of investment properties. The Reserve Bank will now require banks to use higher risk weights (ie have more capital) in respect of the former than in respect of the latter.

This has been quite a saga. The Bank went through a couple of rounds of consultation on earlier proposals last year (then focused on larger operators), and then came back earlier this year with a revised proposal. I made a brief submission on that consultative document, as no doubt did a variety of other people (although we don’t know who, as the Reserve Bank – unlike parliamentary select committees – does not routinely publish the submissions it received). The Reserve Bank’s summary response to the submissions can be found half way down the page here (various specific links on the RB website don’t appear to be working today),

The proposal that the Reserve Bank consulted on in March/April, and which it recently adopted, had a strong feel of being reverse-engineered. The Governor had apparently decided that he wanted to be able to impose additional direct controls on lending for residential property investment, and to do that he needed banks to have systems in place which would clearly delineate between investment property loans and owner-occupied loans. To support that prior policy conclusion, the Bank has sought to argue that loans on residential investment property are, all else equal, riskier than other residential mortgage loans.   To be clear, the Reserve Bank is asserting that a loan is riskier because it is secured on an investment property, even if the initial LVR, the initial date at which the loan was taken out, the nature of the house itself, the borrowers’ income etc were all exactly the same as those for an owner-occupied loan.

What has always been a bit surprising is how little in-depth effort the Bank has put into demonstrating that its argument is correct. It has run a variety of arguments in principle about why investment property loans might be riskier than those to owner-occupiers. Most of those have never seemed overly compelling, especially not in a New Zealand context.   Indeed, there are some reasons why the result could be reversed (for example, unemployment is probably the largest single risk, all else equal, in respect of an owner-occupier mortgage, but rental income flows – which help service investment property loans – tend to be less discontinuous).

But the issue should ultimately, be an empirical one. All else equal, have investor property loans proved to be riskier than owner-occupier loans? Getting good comparable data isn’t always easy.  Material loan losses tend to arise only when nominal house prices fall, and although real house prices fell sharply in the late 1970s, large nationwide falls in nominal house prices haven’t happened in New Zealand since the 1930s. Data from that period aren’t available – although perhaps it is an opportunity for an economic history PhD project working in bank archives. But even more recently, nominal house prices have fallen materially in a number of regions, and I have encouraged the Bank to ask banks for data on the loan loss experience (investor vs owner-occupier) in places like Gisborne, Wanganui, or Invercargill.

In fact, the Bank has tended to rely on a handful of overseas studies, about a handful of overseas experiences. This isn’t one of those areas where there are dozens of studies about dozens of episodes. That makes it all the more important that what studies exist are read carefully and applied and interpreted to New Zealand very carefully. That appears not to have been done. Worse, even when some weaknesses in the way the Bank interpreted and applied such papers were pointed out to them (in submissions on the consultative document), they largely just repeated their assertions and interpretations.

I’ve worked my way through some of the papers, and had had concerns about how the Bank had interpreted and applied the results. My former colleague, Ian Harrison, who consults as Tailrisk Economics, and is much more expert in the specialist risk aspects than I am, has worked his way carefully through each of the empirical papers the Bank has cited, and several that they should have cited, but did not. He has sent me a forthcoming paper “A House of Cards”, in which he has worked his way carefully through each of the Bank’s arguments and pieces of evidence. Cumulatively, it is a pretty damning read. Ian has given me permission to run some excerpts here, and I hope that when his paper is published it will get the attention it deserves.

On the international experience, Ian summarises as follows:

The international literature does not provide support  for the Bank’s contention that investor loans are riskier and owner-occupier loans. Four of the four studies that controlled for other loan attributes found that investor status had no impact, or only a trivial impact, on default rates. A European Banking Authority survey of 41 advanced modelling banks found that none identified investor status as a risk driver in their retail housing mortgage lending models.

A good example of what appears to have gone on is how the Bank has represented an important paper on the Irish experience

Lydon and McCarthy 2011 “What lies beneath? Understanding recent trends in Irish Mortgage arrears”

The graph presented in paragraph 11 of the March 2015 Consultation document presents data from the Lydon and McCarthy paper, which addressed the question of whether BTL [buy to let] status was, in itself, a default driver, or whether the higher default experience could be explained by differences in other loan characteristics.

It was found that after controlling for differences in LVR and servicing costs, BTL status had no impact on default rates.  The higher increase in observed BTL default rates was due to the fact that a larger share of BTL loans were made in the lead up to the GFC when underwriting standards were at their lowest point and house prices at a peak.

Naturally subsequent default rates were higher for investors who bought at the wrong time and who offered scant protection to the lender, but default rates would also have been higher than average for owner occupiers with the same characteristics.

The results of their analysis are presented in table 7 of the paper which shows that the coefficient  for the marginal impact of BTL status is 0.00.  This estimate is significant at the 1% level.

In a subsequent presentation (“The Irish Mortgage market in Context – Central Bank of Ireland 2011) the authors said:

“Controlling for LTV & MRTI…

Relative to next-time-buyers (NTB), FTB borrowers are 2% less likely to be in arrears

–whereas, no relative difference for BTL”(our emphasis)

The data presented in the Consultation document does not provide evidence that Irish BTL loans are a riskier asset class. It is misleading to represent the paper, as the Bank does in several documents, that it provides evidence that BTL loans are riskier.’

Or, in respect of a US study:

Palmer C. (2014) ‘Why do so many subprime borrowers default during the crisis: Loose credit or plummeting prices’

The Bank made the following statement:

“Palmer (2014) reports that default rates increased in a multivariate regression with loan to value ratio and for loans that were declared non-owner occupiers.”

In his paper Palmer uses comprehensive loan-level data to decompose sub-prime loan loss defaults amongst three default drivers. His conclusion is as follows.

Decomposing the observed deterioration in subprime loan performance, I find that the differential impact of the price cycle on later cohorts explains 60% of the rapid rise in default rates across subprime borrower cohorts. Loan characteristics, especially whether the mortgage had an interest-only period or was not fully amortizing, are important as well and explain 30% of the observed default rate differences across cohorts. Changing borrower characteristics, on the other hand, had little detectable effect on cohort outcomes. While quite predictive of individual default, borrower characteristics simply did not change enough across cohorts to explain the increase in defaults.”

There is no marker for investment property as such in the study, just a marker for whether the dwelling was  to be owner occupied or not. It is not clear whether holiday or other second homes would fall. Regardless, the non-occupier marker fell into the borrower characteristic category, which in total provided little independent explanation of deliquency. There was no result that investor status increased defaults. The Bank’s statement was false.

I suggest that you read the entire document when it is available. As far as I can tell, none of the studies the Bank cites appears to have been fairly represented, or applied to New Zealand.

If Ian’s reading of the papers is accurate (and I have no reason to doubt it) it is a very disconcerting commentary on the processes in the Reserve Bank.   The Bank has plenty of able people, who would have been well able to pick up on each of the weaknesses Ian Harrison has identified.  And yet not just once, but again in the response to submissions, and in the new consultative documents, after the Bank had had time to consider the criticisms that submitters have made, the studies have continued to be explained or applied in ways that are, at best, misleading.

Reasonable people can reach different views on appropriate policy measures. I think the Governor of the Reserve Bank has far too much power in this area, and I disagree with the proposed restrictions.   But if citizens cannot trust the Bank to cite evidence in a balanced and accurate way, confidence in the entire policymaking process is likely to be severely eroded.

As I have noted, in such areas the Governor is effectively prosecutor, judge, and jury in his own case. Worse, he is also responsible for the investigative work that is presented in support of the case that he will himself decide. Of course, he has staff to do the work for him, but the staff (and their managers) are hired, rewarded, and potentially fired, by the Governor. A strong Governor will want to know the weakest points in his own case, and to ensure that those weaknesses are appropriately aired, balanced presumably by other strong evidence or arguments for the sorts of regulatory initiatives he is proposing. But the Wheeler Reserve Bank appears to be one in which either no one is willing to stand up and point out the weaknesses or, if someone did point them out, where the Governor and his senior managers said, in effect, “oh just ignore that, continue to repeat the same lines”. One would hope there is a better explanation, but it isn’t obvious. The Bank’s Deputy Governor, Grant Spencer, for example, has spent decades in senior roles in the Bank, and many thought he was a strong candidate to become Governor in 2012. He has direct line responsibility for the two departments dealing with these banking regulatory issues. How did he let documents this weak go out, not just once, but repeatedly? Anyone can make a mistake citing a single paper, but the breadth and repeated nature of what Ian highlights has the feel of something more deliberate.

Even if the Bank could show, with some degree of confidence, that investor property loans were riskier than those to owner-occupiers, other characteristics held equal, the case for the proposed ban on lending in excess of a 70 per cent LVR for residential investment properties in Auckland has serious weaknesses. I elaborated on those in my recent submission (and have also requested copies of all the submissions the Bank has received).

I’ve been critical of the Governor’s conduct of monetary policy over the last couple of years. But reasonable people will, at times, reach quite different views on what monetary policy stance is required. His turned out to be wrong although, as I noted this morning, he had plenty of company for too long.   But repeated misrepresentation of data to support a controversial regulatory initiative strikes me as much more serious. It might do less damage to the economy, but it strikes at the heart of the integrity of the institution, and raises serious questions about the extent to which the public can have confidence in the (unelected) Governor’s ability and willingness to carry out his statutory duties in the public interest, in an objective and dispassionate manner. Cynics might expect such standards from politicians. We certainly shouldn’t tolerate them from officials.

I hope that when Ian Harrison’s full paper is published, the Bank’s Board will start asking some pretty searching questions.  The Board is charged to, inter alia,

    • keep under constant review the performance of the Bank in carrying out—
      • (i) its primary function; and
      • (ii) its functions relating to promoting the maintenance of a sound and efficient financial system; and
      • (iii) its other functions under this Act or any other enactment:
    • (b) keep under constant review the performance of the Governor in discharging the responsibilities of that office:

Perhaps the Minister of Finance might refer the issue to Rod Carr, chair of the Board, for his views.

The Shadow Board on tomorrow’s OCR

The NZIER this morning released the results of its Shadow Board exercise. They survey nine people (currently three market economists, three business people, and three with academic affiliations) and ask them to assign probabilities for the “most appropriate level of the OCR for the economy”. In principle, I suppose “the most appropriate level for the economy” could be different from “the most appropriate level to be consistent with the requirements of the Policy Targets Agreement”, although I suspect that respondents will typically be treating the two as the same. Note that, in principle, the information in the Shadow Board responses is different from the information in financial market prices (which are close to a direct view on what the Reserve Bank will do – as distinct from what people think it should do) or from ipredict, which runs direct contracts allowing people to bet anonymously on what they think the Reserve Bank will do. When I looked just now, the prices reflected an 84 per cent chance of a 25 basis point cut tomorrow

Launching the Shadow Board was a modest but useful initiative by NZIER. It helps spark a little more debate, and a little more scrutiny, about what the Reserve Bank is doing, and puts the results in a useable (and reportable) format. It was inspired by a similar exercise in Australia (which is slightly more (too?) ambitious in that it also asks respondents for probabilities for the right rate six and twelve months ahead).

But what the Shadow Board doesn’t really do is provide any additional information on what the Reserve Bank should do. As everyone recognises, there is a great deal of uncertainty around monetary policy. Central banks talk of trying to target inflation a couple of years out, and yet have no great certainty even as to what is going on right now, let alone what will be going on 12-18 months hence.

Here is a chart showing the actual OCR following the relevant review and the median view of the Shadow Board (thanks to Kirdan Lees at NZIER for sending me the historical data).

shadow board 1

They are all but identical, at least over this relatively short period. And yet the Reserve Bank has subsequently acknowledged that, with the benefit of hindsight they would have had the OCR lower in 2011 and 2012.   And most observers would now agree that the OCR did not need to have been as high as it was over the last year.

Perhaps the information is in the distribution of probabilities rather than in the median view?  Here is the mean of the views of the Shadow Board members. It does deviate a little from the actual OCR, and perhaps during last year the Shadow Board’s views were a little more cautious than the Reserve Bank was. The Shadow Board’s mean view was a little below the actual OCR, while the Reserve Bank itself was still stressing upside risks and the probable need for further rate increases.

shadow board 2

And here are the 25th and 75th percentiles. Respondents collectively put at least a 25 per cent chance on something at or below the 25th percentile being appropriate, and at least a 25 per cent chance on something at or above the 75th percentile.

shadow board 3

It is striking just how tight these ranges are. I noted back in June that most individual respondents’ views seemed excessively tightly bunched, given the huge historical uncertainty about the appropriate OCR. This time around there is a little more dispersion. The Shadow Board exercise has now been running for 27 reviews, and this is one of only three in which respondents collectively put a less than 50 per cent weight on one particular OCR (the other two were January last year, when the Reserve Bank was just about to commence raising the OCR, July last year which proved to be the last of the OCR increases).  I doubt, if I’d been assigning my own probabilities, if I would ever have put even a 40 per cent weight on any particular OCR in any particular review.

One other way of looking at the scale of the uncertainty around the OCR is the fan charts that the Reserve Bank published in the June MPS last year.  These were somewhat controversial, and are hedged around with lots of caveats in the technical notes, but the Governor presumably regarded them as a sufficiently useful device to run prominently in the main policy analysis chapter of a Monetary Policy Statement.   On the subset of shocks and uncertainties considered in that exercise, the 90 per cent confidence interval for the 90 day rate (proxy for the OCR) two years ahead was some 400 points wide. Perhaps a little embarrassingly for the Bank, that range did not even encompass an OCR of 3 per cent or less by July 2015.

fan chart

What do I take from all this?   I’d probably make only two points:

  • There is a huge amount of uncertainty in running discretionary monetary policy.  Some would argue that it is a mug’s game and only likely to introduce additional volatility.  That isn’t my view, but the uncertainty (across a range of different dimensions) is large enough that in general everyone should be a little cautious in taking a stand on a particular OCR (of course, under the current regime, the Reserve Bank must take a view, in actually setting the OCR). Mistakes won’t be uncommon –  whether by commentators or central banks – and that should be recognised, with appropriate humility, by all involved.  Of course, Reserve Bank mistakes matter more because they are charged with the power to take decisions that affects all of us in one way or another.
  • That very uncertainty highlights just how important it is that there is robust debate around a range of perspectives.  In this post, I haven’t looked at the diversity in the individual respondents’ views (partly because the panel of respondents has kept changing, and the sample is quite short), but looking through that data there hasn’t been much diversity of view across respondents either (with the creditable exception earlier in the period of Shamubeel Eaqub).    It is very easy for consensus views to form – both within the Reserve Bank – and in the wider New Zealand debate more generally.  And yet those consensus views will often be wrong.  Sometimes those looking at New Zealand from the outside have had a better take on things, but I doubt that has been consistently true (in the last year, HSBC in Sydney has been running the “rockstar economy” story, while other offshore players were rather more sceptical of the Reserve Bank’s continuing tightening cycle).  Encouraging that diversity of perspective is particularly important within the Reserve Bank, and yet it can be hard to maintain.  That is probably true in all central banks, but is a particular risk in our system, in which the Bank’s chief executive controls resources and rewards and is also single monetary policy decision-maker.  A very good single decision-maker would probably want as much debate and challenge as possible, recognising just how uncertain the game is.  A less-good one would find it too easy to discourage debate and challenge –  while never explicitly saying so, or perhaps even meaning to – preferring material that supports the decision-maker’s own priors and predilections.