John Laker, Chairman - Australian Business Economists, Sydney
I am pleased to join the Australian Business Economists today and renew my acquaintance with an organisation for which, in an earlier life, I was one of the founding Reserve Bank of Australia sponsors.
My talk carries the title "The Australian Banking System Under Stress?" The question-mark is deliberate. Was our banking system under stress during the depths of the global financial crisis? Is it now? APRA and the Reserve Bank of Australia comment regularly on those questions. I want to pose a different question "Would our banking system be under stress if …?"
For authorised deposit-taking institutions (ADIs) in Australia (banks, building societies and credit unions), the global financial crisis has not been the "near death" experience that many banks abroad have confronted, and to which some succumbed.
In October 2008, however, it would have been a brave person to have forecast this outcome. Our larger banks were facing a potential seizing-up of global funding markets and a collapse of investor confidence in banks, whatever their origin, while all ADIs were facing the prospect of a domestic economic contraction of unknown severity.
In the event, our ADIs have emerged from the global financial crisis in generally sound condition. Obviously, souring loans put a dent — significant in some cases — in profitability. But core earnings have held up, the latest readings suggest that the rise in bad debts may have peaked and profitability is rebounding. Analysts now are even beginning to berate banks for achieving returns on equity that are "only" in the mid-teens! This gathering of economists will be familiar with the main factors — including the unprecedented policy responses — that have underpinned this "good news" story and helped to ensure that the Australian economy suffered only a relatively mild downturn.
Throughout the global financial crisis, APRA has stated on a number of occasions that the Australian banking system was well capitalised. Capital is the bedrock of a resilient banking system, enabling it to absorb shocks and continue its vital role in intermediating between savers and investors. APRA's assurances on capital have been important to community and investor confidence, but they have not been given lightly.
Our assurances have been based on some objective facts:
- the Australian banking system has operated over the past decade with a Tier 1 capital ratio in the order of seven to eight per cent, and a total capital ratio at or a little above 10 per cent;
- the predominant form (80 per cent) of Tier 1 capital is its highest quality form — ordinary shares and retained earnings — following APRA's tightening of its requirements in 2006, and
- APRA has taken a more conservative approach than some other jurisdictions to Tier 1 capital calculations to remove items such as intangible assets with uncertain values in liquidation.
Our assurances have also been based on work that has been undertaken by ADIs and by APRA to stress-test the adequacy of capital in the Australian banking system. The global financial crisis, of course, has been a "live" stress-test but not, as I have noted, as demanding a test in Australia as in many other jurisdictions. Stress-testing has therefore been critical to our understanding of what "near death" could imply for our ADIs and whether capital buffers would be sufficient.
Stress-testing is the subject of my talk today. In particular, I want to share with you the results of a macroeconomic stress-test conducted by APRA during the crisis that has reinforced our confidence in the resilience of our banking system.
II. The role of stress-testing
Stress-testing is a valuable tool used by financial and other institutions to manage their risks. In its simplest form, a stress test is an evaluation of an institution's financial position under a "severe but plausible" scenario, to assist in decision-making. Stress-testing also describes a broader process of determining the likely impact of a range of possible scenarios, assessing the willingness of the institution to accept those impacts within its "risk appetite" and considering the actions that need to be taken to reduce risks to an acceptable level.
There is nothing new about the concept of stress-testing. It is centuries old. The first lender that took collateral to cover credit risk would have applied what we recognise today as stress-test thought processes. That lender would have asked the simple but powerful question "what if?" After taking that step, the lender would have thought through the consequences of the loan going bad and made a decision about how to mitigate that risk or whether that risk was simply intolerable, no matter what.
At its essence, stress-testing is the process of considering:
- what may go wrong;
- the resulting outcomes; and
- the actions that could be taken to minimise adverse outcomes or avoid those outside the institution's risk appetite.
Risk managers in banking institutions use these processes for day-to-day decision-making — for example, testing a borrower's ability to repay at different interest rates. However, stress-testing has much broader application and it is now becoming an important tool for managing risks on an enterprise-wide basis.
There is a range of techniques in stress-testing which fit into three main categories:
- sensitivity analysis — where a single risk parameter is moved, holding all other risk parameters constant, to understand the impact on an exposure or risk position;
- scenario analysis — where multiple risk parameters are moved simultaneously under a well-defined stress event. This may involve applying adverse historical experience or hypothetical "worst case" shocks to existing portfolios or positions; or
- reverse stress-testing — where the losses that would render an institution unviable or subject to material regulatory intervention are identified and attention then focused on the types of scenarios that would generate these losses. This type of stress-testing, colourfully described as "testing to destruction", is gaining prominence as a means of engaging boards and senior management in considering the risks that would be terminal to the health of their institution and how they could be addressed.
For an institution, stress-testing can be either top-down or bottom-up and there is a place for both. Top-down testing at the enterprise-wide level can provide a timely and high-level perspective for senior management, but without richness of detail. Bottom-up testing is more granular and takes the characteristics of individual portfolios into account in building up overall results, but it can be time-and resource-intensive.
Like all risk management activities, the real benefits of stress-testing can only be realised in the right operating environment. Boards, business line management and risk management staff must be willing to discuss possible scenarios and resultant risks openly and frankly and to test the assumptions upon which business decisions are made. The effort put into formulating and challenging "what if" scenarios can be more important than the mathematical elegance of the stress-testing methodologies themselves. Stress-testing should not be a back-room exercise conducted solely by quant specialists in the risk management area.
So much for the theory! The reality is that, prior to and during the global financial crisis, the stress-testing practices of many global banks proved inadequate. The crisis itself was far more severe than the scenarios that were tested and stress-testing practices were replete with weaknesses.
In an unflattering review last year, the Basel Committee on Banking Supervision highlighted these weaknesses in four broad areas. It noted, firstly, that stress- -testing was not well integrated in risk governance. Stress-testing generally did not foster internal debate nor challenge prior assumptions; it was often undertaken as an isolated exercise in the risk management area and was not viewed as credible by the business areas; and it was often conducted mechanically and inflexibly.
Secondly, stress-testing methodologies were inadequate. A common failing was the use of historical statistical relationships that were established in benign periods and did not allow for market feedback effects in turbulent times. Stress-tests did not give a comprehensive enterprise-wide perspective across risks and different activities.
The third area of weakness was the selection of scenarios. Most bank stress-tests were not designed to capture the extreme market events that were experienced. As the Basel Committee noted, prior to the crisis "severe" stress scenarios typically resulted in estimates of losses that were no more than a quarter's worth of earnings, and typically much less. Scenarios based on previous episodes of stress proved to be inadequate and hypothetical scenarios involved only moderate stress; anything more severe was not deemed plausible by boards and senior management.
The notion that institutions are reluctant to contemplate their own mortality is described as "disaster myopia". It is seen as an important psychological factor behind various banking crises over recent decades and, most recently, behind the poor performance of narrow Value-at-Risk (VAR) modelling during the crisis. In simple terms, disaster myopia refers to the propensity of economic agents to underestimate the probability of adverse outcomes after long periods of stability. Memories coloured by the "Golden Decade" proved far too short!
The final area of weakness was that a range of particular risks — especially those associated with complex structured products and leveraged lending — were not subject to sufficiently severe stress scenarios. Stress-testing by individual institutions did not capture the systemic nature of the crisis nor the reality that funding markets for innovative financial instruments, and even traditional instruments, could dry up.
The Basel Committee‟s carefully worded critique of stress-testing takes four full pages. The Bank of England cut to the chase in just over four lines. Responding to claims by U.K. banks that proposed new stress-tests would take months to conduct, a senior Bank official said:
"…If even the most obvious stress-test took many weeks to prepare and assess, how could these tests meaningfully be used to manage risk? The short answer, I think, is that stress-testing was not being meaningfully used to manage risk. Rather, it was being used to manage regulation. Stress-testing was not so much regulatory arbitrage as regulatory camouflage."
During the global financial crisis, the growing recognition of the importance of robust stress-testing and the need for supervisors in a number of jurisdictions to provide assurances about the strength of their banking systems, coincided to put a much stronger spotlight on stress-testing activities. This has had two immediate consequences.
Firstly, banking institutions around the globe are seeking to enhance and better resource their stress-testing frameworks and practices. The aim is to improve the capture of relevant risks and their aggregation so as to produce more effective enterprise-wide stress-testing. To support these efforts, the Basel Committee has released a set of guiding principles. These are intended to apply to large, complex banks but they offer prudent guidance to all ADI boards and management asking that basic question "what if?"
Secondly, a number of supervisors have conducted their own industry-wide stress tests. Though varying considerably in detail, the general approach has been the application of a common set of macroeconomic parameters to assess risks and predict capital requirements at the individual institution and system level. The most well-known is the Supervisory Capital Assessment Program (SCAP) conducted in 2009 by the Federal Reserve, the OCC and the FDIC to assess the financial condition of the 19 largest bank holding companies in the United States. The SCAP was intended to be a „deliberately stringent‟ test to identify how much of an additional capital buffer, if any, these institutions would need if the U.S. economy were to weaken more than expected. The release of individual SCAP results, which naturally attracted intense media and analyst interest, was an unusual step that most supervisors, including APRA, would be unwilling or unable to take. However, in the face of considerable uncertainty about the viability of major U.S. banking institutions around that time, the U.S. authorities thought it essential to release individual SCAP findings to reduce that uncertainty and restore confidence.
Critically, too, the U.S. Government stood ready to provide whatever additional capital was necessary to keep these 19 bank holding companies strongly capitalised and lending.
Other supervisors that have conducted macroeconomic stress-tests have generally treated individual results as confidential, but some have published aggregate results.
Let me turn now to how these two recent developments — enhanced stress-testing at the institution level and industry-wide stress-testing — have been paralleled in Australia.
III. Stress-testing by ADIs in Australia
ADIs have been making increasing and more sophisticated use of stress-testing for planning and risk management purposes and their stress-testing capabilities have continued to improve. In a number of areas, APRA‟s prudential requirements have been the driver. In the area of market risk, for example, banks approved to use their internal models must have a comprehensive stress-testing program embodied within the models; in the area of liquidity risk, the "name crisis" scenario to which larger ADIs are subject implicitly involves stress-testing.
The Basel II Framework further promotes the use of stress-testing. Banks approved to use the advanced internal ratings-based (IRB) approaches must have sufficient capital held against credit risk to be able to withstand severe changes in economic conditions. More generally, the use and quality of stress-testing form part of APRA‟s evaluation of the Internal Capital Adequacy Assessment Process (ICAAP) for all ADIs. We acknowledge that stress-testing should be proportional to the nature and size of the ADI but we expect that any comprehensive risk management and capital planning process would include reasonably frequent simulations of a range of unlikely but plausible adverse scenarios. The "what if" question is relevant to institutions of all shapes and sizes.
As I mentioned earlier, the Basel Committee released its Principles for sound stress testing practices and supervision in May 2009. The underlying themes of these Principles are integrating stress-testing into the overall governance and risk management culture of the banking institution, from the board down; ensuring stress-testing provides a complementary risk perspective to other risk management tools; continual updating of the stress-testing framework; and coverage of a range of risks and business areas, including those likely to show the greatest vulnerability. These Principles, of course, have applicability beyond the banking industry.
After the Principles were released, APRA requested that the five "advanced" banks conduct a self-assessment of their compliance. This was followed-up by discussions with each of the banks to help identify better practice in stress-testing and areas for improvement.
Based on these reviews, areas of better practice include:
- stress-testing scenarios applied at an enterprise-wide level are typically well communicated within the institution, including at the board level;
- stress-testing scenarios are updated on a regular basis and are responsive to the changing economic environment; and
- at the enterprise-wide level, a range of risks other than credit risks are considered and potential correlations between risk classes are recognised in the aggregation process.
This work could be extended to incorporate reputation risks (including from subsidiaries and related entities) and liquidity impacts.
Areas where further development is planned include:
- extension of regular stress-testing regimes to the more detailed portfolio and product level, where stress-testing has tended to be more ad hoc;
- upgrading information and IT systems to support stress-testing exercises and reduce pressure on resources. Enterprise-wide stress-testing has tended to rely on only a limited number of staff who dedicate significant time to these exercises but who have come under increased pressure as institutions have sought to run a larger number of scenarios;
- ensuring that documentation keeps pace with developments in stress-testing practices; and
- constructing "severe but plausible‟ stress tests that would provide boards with a perspective on the fundamental risks to their institution‟s business model. In many cases, the stress-testing scenarios applied were not severe enough to threaten viability.
This last point raises some particular challenges in the Australian environment. Australia‟s experience of a prolonged period of benign economic conditions before the crisis, and its resilience during the crisis, may tempt institutions to underestimate the probability and severity of more remote risks and complacency may set in. On the other hand, boards and management may find it hard to engage with scenarios perceived to be too remote or unrealistically severe. No institution, however, is immune from "disaster myopia‟. This points to the need to develop better reverse stress-testing capabilities. To recall, reverse stress-testing is not simply layering scenario upon scenario until sufficient losses are built up but, rather, looking at the sorts of events that could break the business model.
IV. Stress-testing in APRA
Stress-testing forms part of APRA's supervisory armory that goes well beyond the prudential requirements for stress-testing I outlined earlier. APRA conducts its own stress-testing to better understand the vulnerabilities facing individual institutions and industries and the potential for systemic threats. Much of this work is institution-specific and is conducted behind closed doors. However, APRA has also conducted or participated in three comprehensive stress-tests of the Australian banking system.
The results of the first two stress-tests have been well published and I refer to them today mainly to illustrate the evolution of APRA's industry-wide stress-testing and the increasing complexity of the scenarios.
APRA's first stress-test, in 2002/03, focused on the resilience of ADI capital positions in the face of a substantial housing market correction, defined as a 30 per cent fall in property prices. This took a bottom-up approach, which estimated housing loan default and loss rates based on individual loan characteristics. It was not a macroeconomic stress-test because it did not focus on the potential causes of that downturn nor on its second-round effects on other ADI lending or on other profit sources.
The stress-test generated loss rates of around one per cent of the value of housing portfolios, very manageable but well above earlier ADI estimates. No ADIs would have failed and only a small number would have breached minimum capital requirements.
Looking back, though industry-wide stress-testing by supervisors was very much in its infancy at that time, APRA's first stress-test proved pivotal to the subsequent strengthening of APRA's prudential framework for ADI housing lending and lenders mortgage insurance and in confirming that ADI housing lending is a sound asset class, though not without downside risks. It has been one of APRA's most valuable exercises, still paying dividends going into the global financial crisis.
The second stress-test was a macroeconomic stress-test of the Australian banking system conducted in 2005/06 under the International Monetary Fund's (IMF) Financial Sector Assessment Program. This three-year stress scenario, developed by the IMF in conjunction with the Reserve Bank of Australia, APRA and Treasury, was driven by a large fall in house prices (30 per cent again) contributing to a short but sizeable recession and, quite prophetically, domestic banks having difficulty rolling over their foreign liabilities, resulting in higher funding costs and a significant depreciation of the exchange rate. The scenario envisaged a fall in real GDP by one per cent in the first year before a recovery in response to the lower exchange rate, and an increase in the unemployment rate from around five per cent to around nine per cent.
The stress-test was undertaken by the five largest Australian banks at that time. Though the results varied considerably between banks, the results did not reveal any near-term financial stability issues. The banks showed considerable resilience, incurring no losses but suffering a decline in profitability (by 40 per cent at its lows) because of higher bad debt expenses and lower net interest income. The increase in bad debt expenses came mainly from commercial loan portfolios as households cut back consumption; housing loan portfolios were more resilient.
In its assessment of the FSAP stress-test, the Reserve Bank of Australia expressed a caveat about the design of the macroeconomic scenario, which involved a domestic recession but ongoing expansion of the global economy. All previous recessions in Australia have been associated with a global downturn, and incorporating a weaker world economy in the FSAP scenario would have made for a significantly more challenging environment for the Australian banking system. This was, of course, exactly the prospect that confronted our banking system in October 2008.
When the global financial crisis first hit, APRA conducted a range of top-down stress tests on individual ADIs to determine their resilience in the face of a sharp decline in access to, and the cost of, offshore wholesale funding (a significant source of funding for the larger banks) and the continued absence of securitisation markets (an important source of funding for many smaller ADIs). Once the potential severity of the crisis became clearer, APRA focused on identifying potential weaknesses in the capital position of individual ADIs as well as capital adequacy and availability for the Australian banking system as a whole. APRA undertook a further series of top-down stress-tests of ADIs based on the information it had on their lending portfolios. This involved a number of internally generated scenarios based on expected default rates, underlying profit generation, dividend payout expectations and capital availability. Where weaknesses were identified, APRA addressed these with the ADIs concerned.
APRA has also undertaken a more comprehensive stress-test of the larger ADIs based on a specified macroeconomic scenario. This is the third of the industry-wide stress-tests to which I referred earlier.
(I should add that, during the crisis, APRA also undertook stress-testing in the other industries it regulates. In particular, it stress-tested the impact of declining equity and asset prices, widening credit spreads and increased market volatility on the capital position of the life and general insurance industries.)
The three-year macroeconomic scenario was developed in conjunction with the Reserve Bank of Australia and the Reserve Bank of New Zealand; the latter's involvement was important because the stress-test covered the New Zealand operations of the major banks. The scenario was built around a continued deterioration of global economic conditions. World growth and industrial output were assumed to be weaker than consensus expectations as at June 2009, particularly in China, while the fragility of North Atlantic banks acted as a drag on recovery and contributed to wider spreads in global funding markets. An important driver was the assumption that China‟s growth rate fell by at least as much as other countries, resulting in a substantial drop in commodity prices and a depreciation in the Australian dollar exchange rate.
This global scenario generated an economic downturn in Australia significantly worse than that experienced in the early 1990s and that modelled in the FSAP stress-test. The disproportionate impact of China's slowdown meant that the Australian economy contracted more than its trading partners, with the contraction hitting the business sector harder than the household sector. Some of the specific macroeconomic parameters for Australia used as the basis for the stress-test were:
- a sharp (three per cent) contraction in real GDP in the first year followed by a V-shaped recovery;
- a rise in the unemployment rate to 11 per cent;
- a peak-to-trough fall in housing prices of 25 per cent; and
- a peak-to-trough fall in commercial property prices of 45 per cent.
Given the relatively strong position of the Australian economy going into the crisis, the severity of this downturn macroeconomic scenario - evidenced particularly by the pronounced rise in the unemployment rate - was expected to be greater than scenarios used in some other jurisdictions. Remember, this was a "severe but plausible" scenario, not a forecast!
The stress-test was conducted in two phases. In Phase 1, the "advanced" Basel II banks were asked to apply the macroeconomic scenario in their own models and provide their assessment of the impact on the migration of credit ratings, default behaviour, profitability and capital. Granular loss and risk information was collected around five key portfolios - commercial property, income-producing real estate, corporate, small and medium enterprises and residential mortgages.
After analysing the results from Phase 1, which varied considerably across the banks, APRA then determined a common set of portfolio-specific risk estimates. Estimates of defaults were generally within the range provided by the banks but estimates of credit migration were more severe, drawing on APRA's expert judgment of the outcomes that would, on average, apply in the downturn scenario. APRA's risk estimates were structured so that the corporate and commercial property portfolios were the first to experience credit migration and defaults, followed by the small and medium enterprise and income-producing real estate sectors. Residential mortgage lending portfolios experienced their greatest losses towards the end of the three-year scenario as unemployment peaked.
In Phase 2, the advanced banks were asked to apply APRA's risk estimates to their loan portfolios. In addition, APRA supervisors applied the estimates, in a more simplified form, to a select group of "standardised" Basel II ADIs by way of a top-down approach. Phase 2 thus extended the stress-test to cover the 20 largest locally incorporated ADIs by asset size, accounting for around 98 per cent of ADI assets (excluding foreign bank branches).
The main results of the stress test for the 20 ADIs, taken as a group, are as follows:
- none of the ADIs would have failed under the downturn macroeconomic scenario;
- none of the ADIs would have breached the four per cent minimum Tier 1 capital requirement of the Basel II Framework;
- the weighted average reduction in Tier 1 capital ratios from the beginning to the end of the three-year stress period was 3.1 percentage points.
This reduction in Tier 1 capital ratios is calculated gross. That is, it does not allow for any actions by ADIs to mitigate the impact of the stress. Such actions would be expected to include repricing of risk on existing businesses, tighter underwriting standards for new business, improving the security positions on loans and more intense management of higher-risk accounts; it would also include reducing dividends and issuance of new capital. These were, in fact, the actions taken by banking institutions in Australia and other jurisdictions to protect profitability and capital positions through the crisis.
Part of the reduction in Tier 1 capital ratios reflected losses arising from higher bad debt charges, particularly in corporate and commercial property lending. For the advanced banks, however, the primary driver was downwards ratings migration, which pushed up the measure of risk-weighted assets and, hence, capital requirements. In effect, APRA's stress-test imposed a procyclicality effect for which the Basel II Framework has been criticised, although I would add that global studies have thus far found it difficult to identify a significant impact from ratings migration during the crisis.
To put APRA's stress-test into perspective, the reduction in Tier 1 capital ratios are at the higher end of the range of macroeconomic stress-test results in other jurisdictions with which we are familiar. Comparisons in this area are particularly difficult because the stress-tests differ considerably in their construction, severity and treatment of risk-weighted assets. That said, APRA's results are indicative of a tough stress-test at work.
Considerable effort — intellectual, IT and otherwise — has gone into APRA's most recent industry-wide stress test, on the part of ADIs and supervisors alike. For all that, stress-testing is necessarily judgmental and outcomes are critically dependent on the assumptions and judgments made. And one thing can be assured — the scenario on which the stress-test was based will not play out as specified. It is just one of a myriad of future possible outcomes.
Those caveats aside, APRA‟s recent stress-test has provided important evidence that the Australian banking system has the capital resources to weather an economic contraction much worse than that expected during the depth of the global financial crisis.
Our banking institutions are on a journey to improve their stress-testing capabilities. So is APRA! We welcome indications that the larger institutions, in particular, are resourcing themselves to undertake this work and are engaging their boards and senior management more fully. An active and well-informed board can provide the „bucket of cold water‟ response to soft stress-tests, reluctant challenging and undue optimism. For our part, top-down stress-testing is becoming increasingly embedded as an important tool in our supervisory activities. As recent events in Europe remind us, supervisors also need to refresh stress scenarios since the direction of threats to the viability of individual ADIs, and to the stability of the banking system as a whole, is ever-changing.
- Basel Committee on Banking Supervision, Principles for sound stress testing practices and supervision, Bank for International Settlements, May 2009.
- Andrew G. Haldane, Why Banks Failed the Stress Test, Bank of England, 13 February 2009, p13.