Graph 1

Housing Credit Growth and Housing Loan Defaults

Over the past seven years, it would seem, housing loan default rates have become uncoupled from growth in housing credit.  The likely explanation is that the rapid increase in housing prices over the period has enabled any troubled borrowers to exit the market without incurring losses.  The decoupling might continue for a time whilever the belief persists – on the part of both borrowers and lenders – that housing prices can continue to rise without respite.  When reality inevitably intrudes, however, the earlier linkages between housing credit growth and loan defaults might reassert themselves.  This is not a prediction on APRA’s part, but it is the kind of “worst case” scenario which has motivated our stress test of ADI loan portfolios.

APRA’s stress test model

Let me turn now to the stress test itself.  APRA’s stress test model is a microeconomic model, which estimates housing loan default rates and losses based on the characteristics of individual loans – in particular, the LVR at origination and the age of the loan.  The model does not attempt to forecast or simulate the future path of any of the macroeconomic factors that could affect house prices and default rates and losses.  As a result, the APRA stress test does not model the potential causes of a housing price correction, and this is quite deliberate.  Rather, it focuses on the impact that a potential correction could have on the capital position of ADIs.

The stress to which ADI housing loan portfolios have been subjected is a 30 per cent real reduction in housing prices over a one‑year period and a substantial increase in mortgage defaults.  Remember, these are assumptions, not predictions.  At first glance, the assumption on housing prices appears a severe one.  However, in the late 1980s/early 1990s, other countries have experienced peak‑to‑trough falls in real property prices of greater than 25 per cent, including the United Kingdom, Finland, Norway and Sweden.  Sharper falls have been observed more recently in some South‑east Asian capitals.  APRA’s assumption corresponds to a situation where property prices return to around mid‑2001 levels, which could not be called  a draconian result.

In APRA’s stress test, the impact of a housing price correction of this magnitude on housing loan portfolios depends on three key considerations:

• the probability that a loan will default; 
• the extent of losses if loans do default (loss‑given‑default);  and
• any mortgage insurance recoveries that may be available.

Probability of default

For any given loan, the probability of default is a function of its LVR, age, size and type (owner‑occupied or investment).  In general, high LVR loans are more likely than low LVR loans to default since the borrowers tend to have a higher proportion of their incomes committed to debt servicing, and a lower equity buffer to withstand a fall in housing prices.

The relationship between the probability of default and age of a loan tends to be non‑linear.  Default rates are low in the year after the loan was originated because borrowers under stress are often able to access other forms of credit or sell assets to maintain repayments for a short period.  Similarly, borrowers that have been making loan repayments beyond four years have generally passed the most difficult debt servicing period, and are less likely to default.  Such borrowers will have built up their equity in the property concerned due to a combination of the historical experience of rising housing prices and loan repayments (often ahead of schedule).  Borrowers in the middle years (two to four years since origination) tend to have the highest default rates under normal market conditions.

Based on the housing loan default data provided, the APRA model assumes that, for a given LVR and loan age, larger loans are likely to have a higher debt servicing burden than smaller loans and therefore a higher probability of default.  Moreover, larger loans are more likely to be secured against properties with a higher luxury component and/or in geographical areas that have experienced the greatest price appreciation.  Such loans, in turn, are likely to be more vulnerable to market corrections.

Finally, and also based on the default data provided, the APRA model assumes that investment loans are likely to be more at risk of default than loans for owner‑occupation.  Some may challenge this assumption, but the underlying reasoning is that owner‑occupiers are likely to place more importance on retaining their primary place of residence.  Moreover, investment loans often have a speculative element and face the additional risk of rental vacancies.

The APRA stress test has a range of default rates, depending mainly on the LVR and age of loans, which are calibrated to an median probability of default on a housing loan portfolio of five per cent.  Based on the data provided to APRA, this equates to a probability of default on housing loans for ADIs as a group of around 3.5 per cent.  How tough is this scenario?  The answer is that it is tough, judged against Australian experience over the past twenty years, when housing loan default rates – measured by claims under mortgage insurance arrangements – have averaged only around 0.12 per cent a year.  At the same time, it is not out of line with experience in Queensland in the mid 1980s and Victoria in the early 1990s, when default rates reached five to six per cent (on an underwriting year basis).  Episodes of housing market stress in the United Kingdom and some regions of the United States over that same period generated higher default rates.

Loss-given-default

Losses on defaulted loans arise mainly because of a shortfall between the proceeds from the sale of property held as security and the amount of the loan outstanding.  Losses can also increase significantly because of legal costs, real estate agent fees, marketing expenses and other costs involved in the collections process.

In the APRA model, the loss‑given‑default is determined by the LVR and age of loans.  The higher the LVR, the lower the security coverage is likely to be in the event of default and hence the higher the potential losses.  The younger the loan, the higher are debt levels likely to be and the less likely that security against such loans would have benefited from housing price appreciation.  Hence, loss‑given‑default is expected to decline as the age of a loan increases.

APRA has calculated loss‑given‑default ratios based on historical data supplied by the mortgage insurance industry in Australia and data provided by a major international mortgage insurer to the Basel Committee on Banking Supervision.  In APRA’s stress test, the average loss‑given‑default equates to around 25 per cent.

Mortgage insurance recoveries

As I noted earlier, around one‑fifth of ADI housing loans are covered by lenders’ mortgage insurance.  Where defaulted loans have such insurance, the ultimate loss to an ADI will depend on the level of mortgage insurance coverage held, and on the payout ratio.

Modelling the determinants of payout ratios is a difficult task, hampered by the lack of available data.  Based on APRA’s analysis, payout ratios in a stress scenario are likely to be determined by two general factors.  One is whether the insurance was written directly by the mortgage insurer or by the ADI under “open policy”, where ADI staff manage the insurance origination process.  The other factor is the age of the loan.

Payout ratios are likely to be lower for open policy contracts because the mortgage insurer’s terms and conditions are less likely to have been followed precisely.  Payout ratios are likely to be higher for older loans.  If a loan defaults shortly after origination, it is more likely that an error has been made in the origination process and that mortgage insurance policies have not been followed correctly.  This would require some adjustment to the claim made by the lender.  Loans that default well after origination, however, are more likely to be the result of changes in the borrower’s circumstances rather than poor loan origination procedures, in which case no claims adjustment is likely to be required.

Drawing on overseas experience, the APRA stress test allows for the possibility that claims under mortgage insurance policies will be subject to greater adjustment in the event of a substantial housing market correction, particularly where this is preceded by a surge in new housing lending.

The stress test results

Drawing these various elements together, the APRA stress test model involves determining a common set of assumptions about the probability of default and losses‑given‑default on ADI housing loan portfolios, in the face of a “shock” of a substantial housing market correction.  These assumptions were then applied to the actual housing loan portfolio of each ADI, taking into account the LVR, age, size and purpose of its loans and its mortgage insurance arrangements.  By applying a common set of assumptions in this way, APRA has prepared an easily comparable projection of defaults, losses and resultant impact on each ADI’s capital position.

The key results of the stress test, for the ADI sector as a whole, are worth emphasising.

Firstly, ADIs as a group enjoy strong capital positions today and this strength – though reduced – would not be materially affected under the stress scenario modelled by APRA. 

Secondly, the estimated default rate for ADIs as a group would be around 3.5 per cent.  This default rate leads to a loss rate of around one per cent of the value of housing loan portfolios for ADIs as a group, equivalent to around 66 basis points of current regulatory capital ratios.

Thirdly, over 90 per cent of ADIs would survive such losses, taken on their own, without breaching minimum regulatory capital requirements.  For a small number of ADIs, the losses, again taken on their own, would not be covered by surplus capital, but the breaches of minimum capital requirements would not be large.  No ADI would fail or come uncomfortably close to failing under the stress scenario.

Finally, ADIs as a group would have net claims on lenders’ mortgage insurers of a little below 3 per cent of current capital, a modest exposure in the context of ADI capital positions overall.

These results are, as I mentioned, averages for the ADI sector and there is obviously variation across institutions.  Those institutions with older, lower LVR loans mainly for owner‑occupiers would be expected to perform better in the stress test, and generally did so.

Broadly speaking, the results are reassuring.  They demonstrate that the ADI sector – even though heavily exposed to Australia’s very buoyant housing market at present – remains well capitalised and could withstand a substantial housing market correction, if one were to eventuate, without putting depositors at undue risk.

It could be argued, of course, that the impact of a housing market correction should not be assessed in isolation.  Such a correction would create pressures on the household sector, and on economic conditions more generally, that could threaten the quality of other ADI lending and other profit sources.  The APRA stress test makes no allowance for this, but we acknowledge that a significant increase in loan default rates would be associated with a deterioration in economic conditions.  On the other hand, the stress test has not factored in any ADI response to signs that the credit environment was becoming tougher.  If conditions were to turn, ADIs would be taking steps to bolster their capital positions in anticipation.

Lessons learned

Having completed the stress test, APRA’s supervisors are following up on the results with individual institutions.  We are providing ADIs with an analysis of their performance, including benchmarking to relevant peer groups to help ADIs understand where they sit in the industry.

APRA’s stress test has drawn attention to some particular issues in ADI housing lending, which APRA will be taking up with the institutions concerned.  One issue is the quality of management information systems.  Some institutions required a number of attempts to provide accurate data to APRA;  this was particularly the case for institutions that have been through mergers, and had to collect consistent data across the organisation.  More surprisingly, many ADIs – both large and small – are not keeping vital information such as LVRs, loan purpose or insurance status in an electronically accessible format.

Over the past year, APRA has been emphasising the importance of investment in risk management systems, in which information systems play a crucial part.  We acknowledge the constant pressure on ADI boards and management for earnings increases, but these increases should not be achieved through short‑term cost cutting measures that “hollow out” risk management capacity.

A second issue concerns mortgage insurance.  The stress test results showed that the net exposure of the ADI sector to mortgage insurers is only a modest proportion of ADI capital.  For some individual ADIs, however, the ability to claim effectively on their mortgage insurance policies has a material effect on the profitability of their housing lending activities.  Such an exposure requires considerable attention to detail by the ADI.  A mortgage insurance policy is not a guarantee.  Just as with any other insurance policy, the holder must justify a claim, and an ADI engaging in poor underwriting, loan management or collection processes may well find that its mortgage insurance claims require adjustment or, in the worst case, are denied.  This is not a reflection on the capacity or willingness of mortgage insurers to pay claims;  APRA supervises these institutions closely and acknowledges their commercial imperatives.  It is a simple statement of the operational risk facing ADIs that find themselves unable to claim properly.

A third issue concerns the ability of ADIs to handle any significant housing market correction from a management and resourcing viewpoint.  We are asking ADIs to discuss with us their management plans for dealing with an increase in problem loans and loan defaults.

Summing up

APRA has subjected ADI housing loans portfolios to the “what if” scenario of a housing market correction – involving a substantial fall in housing prices and rise in loan defaults – that is far worse than Australia’s experience over the past twenty years and beyond.  In that sense, the assumptions underlying the stress test could be viewed as conservative and searching.  However, the assumptions are not out of line with the experience of other industrial countries in the 1980s and 1990s and, though we would not wish to have this experience visited on our shores, the assumptions could not be called unreasonable.

The core finding is that a substantial housing market correction, if one were to eventuate, would not of itself be cause for undue alarm as far as the strength and stability of Australia’s ADI sector is concerned.  This finding gives considerable comfort that the ADI sector remains able to protect depositor safety over the medium to longer‑term.  For a few institutions, nonetheless, APRA has identified specific improvements that are required to risk management policies and capital positions, and APRA will be working with these institutions to achieve improvements.  Where necessary, APRA would raise minimum capital ratios for ADIs whose housing lending practices are not up to the mark.

The conclusions from APRA’s stress test fully support the view of the Reserve Bank in its 2003 Annual Report.  Having identified a substantial housing market correction as the main potential source of risk to financial stability in Australia, the Reserve Bank offered the assessment that the increase in lending to households “… does not pose a significant danger of a financial crisis, such as occurred in the early 1990s after the build‑up in corporate debt”.  The Reserve Bank’s assessment was based on its readings of a range of aggregate financial soundness indicators and its understanding of ADI risk management procedures, while APRA’s conclusions are drawn from data on the characteristics of housing loan portfolios of individual ADIs.  The two approaches obviously complement each other.

In our view, the stress test has been a highly worthwhile exercise.  It has helped APRA and industry participants alike to gain a fuller understanding of the risks in housing loan portfolios in Australia.  We also believe it will sharpen the focus of ADI boards and management on these risks, and provide a counterweight to the current exuberance in housing credit growth and housing prices.  The housing cycle will inevitably turn at some point.  For this reason – and notwithstanding the positive results for the ADI sector as a whole – APRA’s message to individual ADI boards and management is “Proceed with Caution”.