RDP 2015-06: Credit Losses at Australian Banks: 1980–2013 3. Descriptive Analysis

3.1 Credit Losses over Recent Decades

The Australian bank credit loss experience since 1980 is dominated by the very high rate of losses before, during, and after the early 1990s recession, as well as the smaller losses during and after the global financial crisis (Figure 3). Losses around the early 1980s recession were much lower. Relative to lending, credit losses during the early 1990s far exceeded those incurred by banks during and after the global financial crisis. Current losses between September 1989 and September 1994 totalled around 8½ per cent of the average value of banks' lending during this period. In comparison, current losses during September 2007 to September 2012 were equivalent to around 2½ per cent of average lending over this period.

The average sample aggregate CLR during 1980–2013 was 56 basis points. The median, less influenced by the high levels in the early 1990s, was 34 basis points, which was also the 2013 level.

Credit losses have strongly influenced the profitability of the Australian banking system during the sample period. This can be seen by decomposing changes in aggregate return on equity, a common measure of bank profitability (Figure 4).^[9] Credit losses were the largest contributor to the cycles in profitability during the early 1990s and global financial crisis episodes.^[10]

3.1.1 The early 1990s

The partial portfolio-level data that are available for the early 1990s episode indicate that the bulk of credit losses were incurred on lending to businesses rather than households. Two major banks published usable portfolio breakdowns of their net write-offs in their annual reports for some or all of the early 1990s, but the categories used in this data were not well defined (Figure 5).^[11] They show losses on non-construction housing loans were minimal (these fall within the ‘Real estate – mortgage’ category). Loans to individuals for construction of housing probably fell within the ‘Real estate – construction’ category, but this category also contains lending for commercial property. Losses on this category were significant, but only make up around 13 per cent of reported losses for these two banks. The key point is that most of the losses reported by these two banks fall in the ‘Other business’ category. Losses on personal lending, such as credit cards and non-housing term loans, were non-negligible, but appear to be less cyclical than losses on business lending.

Portfolio-level data are available on all banks' non-performing assets from mid 1990 to mid 1994, and these support the conclusion that losses were incurred mainly on business lending (Figure 6).^[12] It shows that the share of banks' lending to businesses that was non-performing far exceeded the share of their lending to households (including non-mortgage personal lending) that was non-performing.

Contemporary accounts of the period also indicate that credit losses were primarily on lending to businesses. Trevor Sykes' (1994) classic account of corporate and banking collapses during this period, The Bold Riders, is one example. Edna Carew's (1997) account of Westpac's experience during the period indicates its losses were concentrated in business lending, and more specifically, in property development lending. The dominant role of business lending is also suggested by contemporary accounts from industry participants (Phelps 1989; Lee 1991).

Various authors have set out potential reasons why credit losses were so large in the early 1990s (Battellino and McMillan 1989; Fraser 1994; Sykes 1994; Carew 1997; Conroy 1997; Ullmer 1997; Gizycki and Lowe 2000). There was a recession during 1990–91, and downturns in financial and property markets, but losses were many times greater than those seen in earlier (and later) downturns, suggesting other factors at play. A short version is that deregulation of the banking sector in the 1980s was accompanied by very fast business lending growth and declining lending standards, all during a period of strong economic and financial conditions.^[13] When conditions eventually worsened, a sharp rise in credit losses was the result. In more detail:

1. Deregulation allowed banks to extend credit to meet demand from borrowers (Battellino and McMillan 1989). The rates and terms at which banks could offer deposits were liberalised over the 1970s and first half of the 1980s. Prior to this, banks passively accepted deposit flows and restricted lending during periods of deposit outflow. The change allowed banks to actively manage their funding to match the demand for credit, and was accompanied by the removal of interest rate caps on lending products and requirements for banks to lend to certain borrowers. In addition, in 1985 foreign banks were allowed to enter the Australian banking market as retail deposit-takers for the first time in over 40 years (Fraser 1994). The net result of these changes was a market where banks competed intensely to grow their loan books and maintain market share. Annual growth in nominal business credit rose above 20 per cent in September 1984, and didn't fall below this level again until June 1989 (Figure 7).

2. In part due to the competitive pressures unleashed by deregulation, bank lending standards loosened considerably over the 1980s (Macfarlane 1991; Sykes 1994; Conroy 1997; Ullmer 1997). From the late 1970s, banks departed from the practices of earlier decades and began lending to large companies on an unsecured basis, and accepting riskier forms of collateral (such as equity in subsidiaries and mortgages over unfinished developments). Banks also relaxed covenants around the use of borrowed funds, loan-to-valuation and interest-coverage ratios. Another major driver of the losses over this period was a lack of transparency on borrowers' total use of debt finance. When borrowers entered financial difficulty, banks would sometimes discover total debt was higher than thought, and even their own group exposures were higher than thought, due to lending by subsidiary finance companies and merchant banks. Remuneration was one potential driver of the fall in lending standards: corporate lending officers in banks were frequently remunerated on the basis of volume, with little consideration of long-run asset performance. Arguably, this loosening of lending standards occurred because banks, emerging from an era of tight regulation, lacked the proper corporate governance and sophisticated credit risk management frameworks that have come to be seen as necessary for prudent banking in a deregulated financial system.
3. Macroeconomic and financial conditions facilitated these developments. Real GDP grew at an average rate of about 4¼ per cent over the five years to September 1989. Equity prices rose by almost 50 per cent per annum from late 1984 until the crash in October 1987. Commercial property price growth rose above 10 per cent per annum at the start of 1986, and accelerated in subsequent years. This price growth was accompanied by an exceptional amount of non-residential construction, particularly of offices (Figure 8; Kent and Scott 1991). Commercial property was a key form of collateral for the business loans that were secured.

4. Immediate triggers for the rise in credit losses are easier to discern than the underlying reasons why they were so large. Business interest rates rose from around 13 per cent at the start of 1988 to over 20 per cent by the end of 1989, due to rises in official rates. Together with slowing business profits growth and the significant growth in business debt, this meant that the aggregate business sector interest burden was very high (Figure 9). By early 1990, large highly geared companies across a range of industries were unable to meet their increased loan repayments and defaulted on their debts (Sykes 1994). This, together with a weakening in the commercial property market, exposed banks to a first round of credit losses (Gizycki and Lowe 2000). These losses broadened as business profits began to fall and Australia entered a recession around the end of the year. By September 1991, large additions to the supply of office property had combined with flat or falling demand to sharply raise vacancy rates and drive prices down by over 20 per cent on an Australia-wide basis; some banks were forced to recognize significant credit losses on commercial property lending (Carew 1997).

Of the banks in the long-run dataset, the one that incurred the highest rate of credit losses during the early 1990s was a small foreign-owned bank (Figure 10). These losses were equivalent to a significant proportion of this bank's capital, but it was recapitalised by its parent entity. Of the groups that make up larger portions of the sample, state government-owned banks experienced the highest credit loss rates over this period. Two, the State Bank of South Australia (SBSA) and the State Bank of Victoria (SBV), effectively failed, in that they had to rely on extraordinary financial support from their state government owners (Fitz-Gibbon and Gizycki 2001). The major banks also experienced large credit losses over this period. Two major banks – Westpac and ANZ – reported large overall losses in their annual reports for 1991. The other banks in the sample, primarily smaller Australian-owned banks, incurred significantly lower losses over the period. These banks' portfolios were generally more concentrated in lending to households.

Even during a period in which system-wide lending standards loosened, there are indications lending standards at state government-owned banks, particularly SBSA and SBV, were below average:

• These banks grew their lending very quickly over the late 1980s. SBSA and SBV grew their lending at rates of 43 and 27 per cent per year between 1985 and 1990, versus growth in total credit of around 18 per cent per year over this period. This fast growth was driven by business lending – the share of these banks' portfolios made up by business lending increased by over 20 percentage points over the same period. State government owners encouraged fast lending growth, both to support state economies and to provide a new source of revenue for state coffers, and installed aggressive managers (Sykes 1994).

• There is some direct evidence on lending standards at these institutions. The Auditor-General of South Australia's report into SBSA stated:

  … the Bank's corporate lending … was poorly organised, badly managed and badly executed. Credit risk evaluation was shoddy. Corporate lending policies and procedures were not even compended into a credit policy manual until 1988, and even then contained serious omissions. The ultimate loan approval authority – the Board of Directors – lacked the necessary skills and experience to perform its function adequately. Senior management's emphasis was on doing the deal, and doing it quickly. MacPherson 1993, p 1–24

• State government-owned banks were not formally subject to prudential supervision by the Reserve Bank, though they had given undertakings to comply with the Reserve Bank's prudential regulations. Despite this, there were instances where they did not do so.^[14]

Despite large credit losses, there were no disorderly bank failures during the early 1990s (Gizycki and Lowe 2000). The liabilities of state government-owned banks were always explicitly guaranteed by their owners. The banking system as a whole remained well-capitalised; partly due to some banks raising equity, the aggregate capital ratio actually rose over this period (Fraser 1994). Both ANZ and Westpac maintained capital ratios above regulatory minima, despite their losses in 1991. There were short-lived deposit outflows at some small banks, but these were quickly ended by Reserve Bank assurances about their solvency.

3.1.2 The global financial crisis

The elevated credit losses experienced during and after the global financial crisis were due to business lending; the better data available for this period make this clear (Figure 11). Losses on household lending barely rose over the period. Losses on the business loan portfolio were much lower than those incurred during the early 1990s: annual net write-off rates on business lending averaged 0.8 per cent over the four years beginning in March 2009, well-below average total write-offs rates during the early 1990s (and, presumably, even higher business loan write-off rates at that time).

The low loss rate on lending to households over this period was driven by very low losses on housing loans, which made up around 90 per cent of bank lending to households over this period. The net write-off ratio on housing lending averaged 3 basis points per year during 2008–13.^[15] Most of the losses on lending to households during this period arose from personal lending (credit card and other personal lending) (Figure 12). Though personal lending has a relatively high loss rate, it appears to be significantly less cyclical than business lending, and anyway only makes up around 5 per cent of bank lending in Australia.

Around one-fifth of Australian-owned banks' consolidated assets are offshore, so the consolidated Pillar 3 data used in Figure 11 (left panel only) and Figure 12 reflect overseas credit risk to some extent. Australian banks' credit losses on offshore lending were significant during the GFC (see, for example, RBA (2010)), but domestic credit risk is the focus of this paper.

One part of the explanation for the lower credit losses experienced during the global financial crisis is the less severe nature of this episode: GDP fell for only a single quarter and office property prices fell by around a quarter, compared with a peak-to-trough decline of around one-half in the early 1990s (see Figure 8). Bank lending to businesses grew at around 15 per cent per annum over the five years up to mid 2008; this was around 8 percentage points below its growth rate over the five years up to mid 1989 (higher inflation in the earlier period only accounts for around half of this gap). This smaller rise in debt, together with structurally lower interest rates that fell quickly in response to large cuts to the cash rate, meant the business sector's aggregate interest burden peaked at around 17 per cent of profits during the global financial crisis, well below its level in the early 1990s (see Figure 9).

There is also evidence that more conservative business lending standards were a key contributor to the better credit loss experience during this episode. Partly in response to the problems in the early 1990s, and partly in response to the imposition of risk-based capital requirements and other regulatory pressures, banks had improved their management of credit risk by the start of the global financial crisis according to many observers (Eales 1997; Ullmer 1997; Gray 1998; APRA 1999; Laker 2007). Better IT systems were put in place to assess and monitor credit risk, and the governance of credit risk decisions within banks had improved.

3.2 Other Aspects of Credit Losses

This section explores the timing of credit losses with respect to the economic cycle, and relationships between credit risk measures. If credit losses peak quickly after troughs in output, this means the financial strength of the banking sector may start to improve soon afterwards – a key consideration for economic policymakers after the global financial crisis. Likewise, if credit losses peak before non-performing assets, they might provide an early signal of future improvement in the financial strength of the banking sector.

3.2.1 Timing

The temporal relationship between credit losses and output was reasonably similar during the early 1990s and global financial crisis episodes. The peak in current losses in the early 1990s, as measured by the long-run dataset (which provides annual losses as at September of each year), was in 1991. The trough in annual GDP during this episode was in the December quarter of 1991. APRA's quarterly credit loss data for all banks (available from 2003), allow more precise measurement of timing. Quarterly credit losses, a volatile series, peaked in the same quarter as the trough in quarterly GDP during the global financial crisis episode (Figure 13). Losses rose noticeably three years before their peak in the early 1990s, while they were only slightly elevated a year before their global financial crisis peak.

3.2.2 Relationships between credit risk measures

The relationships between different measures of credit losses differed somewhat across the two main episodes (see Figure 3). The BDR exceeded the CLR in the years immediately prior to both the downturns, indicating that banks were increasing collective provisions in anticipation of a deterioration in loan performance. During the global financial crisis, banks continued to increase collective provisions during the downturn itself, perhaps owing to an overly pessimistic view of future developments. The profile of credit losses was a relatively symmetric hump in the early 1990s, but credit losses generally declined more slowly in the years following the global financial crisis. This may reflect economic conditions over this period, or banks adjusting their behaviour in recognising and disposing of troubled loans. This difference makes comparing the delay between initial losses and final write-offs between the two episodes difficult; but, in aggregate, the net write-off ratio peaked two years after the other two ratios in the early 1990s, and a year after in the global financial crisis episode.

Credit losses in Australian banking have generally peaked before non-performing assets (NPA) and impaired assets (IA), though these measures have risen in tandem at the start of downturns. APRA's quarterly credit loss data for all banks show a lead of three quarters between the peak in annual credit losses and that in NPAs during the global financial crisis episode (Figure 14); the lead is five quarters between the peak in quarterly credit losses and that in NPAs. For IAs, these leads are arguably zero and two quarters (respectively), given the June quarter 2009 value for this variable is very close to its peak in the March quarter of 2010.

Footnotes

The data used in this exercise differ somewhat from the credit losses dataset: it is consolidated data for Australian-owned banks only. [9]

Decomposing changes in a ratio requires choices as to the ordering of the decomposition. I have used the ordering that minimises the contribution of credit losses to the change. [10]

These two banks, CBA and NAB, accounted for 33 per cent of bank lending at September 1991. CBA's write-offs include those made within the State Bank of Victoria's loan book after its acquisition in November 1990. [11]

No similar data were collected before June 1990, and the regulatory collection from September 1994 onwards did not have a portfolio breakdown. These rates are slightly downward biased. The numerator uses non-performing assets data from the Australian operations of all banks' consolidated groups. In contrast, the denominator includes all lending done by financial intermediaries in Australia, including lending done by non-bank financial companies not owned by banks. [12]

I use a broad definition of lending standards in this paper: non-price differences in borrower characteristics and loan terms that are ex ante observable by a bank. I expand on this definition below, but it is important to note that I do not include changes in portfolio composition between business, housing, and personal loans within my definition. [13]

Sykes (1994) provides the examples of a large exposure and a related-party transaction that were undertaken by SBSA contrary to Reserve Bank advice. The SBV failed to meet the Reserve Bank's capital adequacy standards during the late 1980s (Victoria 1991). [14]

This loss rate is after the effect of lenders mortgage insurance (LMI), which Australian banks hold on a significant portion of their housing loans (estimates suggest LMI covers roughly one-quarter of housing loans). Reserve Bank estimates suggest the annual loss rate faced by lenders mortgage insurers averaged 3 basis points over 1984 to 2012. [15]