RDP 2001-06: The Effect of Macroeconomic Conditions on Banks' Risk and Profitability 2. Sources of Variation in Banking Risk and Performance

2.1 Variation Between Banks

Differences between banks in risk exposure and profitability arise from two broad sources: the different operating strategies that banks choose to adopt and the efficiency achieved by each institution.

Each bank's operating strategy has three main elements: its scale, the scope of its operations and its risk appetite. A significant body of empirical work suggests that the scale economies in banking are not large, thus performance is not strongly related to the scale of each bank's operations (Calomiris and Karceski 1998). Nevertheless, differences across banks in the scale of their operations will result in different levels of diversification and thus different levels of riskiness.

Economies of scope present a trade-off between the benefits of specialisation (for example, a simple product range allows for ease of administration) and the benefits from diversification and cross-selling. In seeking to maximise efficiency in making loans and related customer-servicing activities, banks tend to focus on areas where they believe they have a comparative advantage (Allen 1997). This leads to concentrations by geography, industry, demographics and other market characteristics. Such differences in the scope of banks' activities will lead to differences in bank riskiness and profitability as economic conditions vary across different regions and industrial sectors.

While the scale and scope of operations of each bank will influence its overall risk, banks may also separately manage their risk profile. The development of financial techniques such as securitisation, credit derivatives and other financial derivatives has enhanced banks' capacity to determine their risk profile independent of their underlying business activities (Basel Committee on Banking Supervision 2000).

The ability of managers to achieve efficient use of inputs will influence the dispersion in risk and performance across banks. In particular, inefficient institutions have a tendency to carry higher risk (Eisenbeis, Ferrier and Kwan 1999). This may be because poor management quality results in both poor resource and risk management. Alternatively, individual institutions may attempt to compensate for their inefficiency by taking on greater risk.

There is an extensive literature that seeks to identify which individual-bank factors best predict bank performance. Since the bulk of this research has been carried out in the US (with its large population of banks and long history of bank failures), the work mostly focuses on modelling the likelihood of bank failure. In this work, capital adequacy, earnings and impaired assets are found to be the most useful indicators of the probability of failure (Demirgüc-Kunt 1989). In addition, rapid expansion of lending activities tends to increase risk (Keeton 1999). More elaborate modelling, which distinguishes between the likelihood of failure and time to failure, finds that basic indicators of a bank's condition such as capital, net income and impaired assets are also important determinants of the timing of bank failure (Cole and Gunther 1995).

A number of studies compare the effects of individual-bank risk-taking and macroeconomic conditions on the likelihood of bank failure. Emmons (1993), when considering US banking failures, concludes that increased risk-taking at individual banks alone does not fully account for the observed pattern of bank failures. Local economic conditions are also important predictors of bank failure. It is the coincidence of risky bank portfolios and difficult economic conditions that makes bank failure most likely.

González-Hermosillo, Pazarbaşioğlu and Billings (1997), in their study of the 1994 Mexican financial crisis, refine the distinction between the effect of bank-specific and economy-wide factors on the likelihood of bank failure. They find that factors determining the likelihood of failure differ from those determining the timing of failure. Bank-specific variables, in combination with aggregate banking sector factors help to explain the likelihood of bank failure, while macroeconomic factors play a pivotal role in influencing the time of failure. In Mexico, high real interest rates, exchange rate depreciation and an increase in the overall gearing of the economy triggered bank failures.

2.2 System-wide Variation

While each bank, on its own, can choose to take on more or less risk by changing its own behaviour, its riskiness and performance will also be influenced by developments in the aggregate supply and demand for loan finance. On the demand side, banks may become more risky because either borrowers put more risky projects forward for bank finance or the amplitude of the economic cycle may, unexpectedly, increase. On the supply side, the behaviour of individual banks may become more risky due to industry-wide developments, such as changes in regulation and the level of competition in the banking market.

One widely used measure of credit risk is the ratio of impaired assets to total on-balance sheet assets.^[1] Figures 1 and 2 compare the movement in the industry-average impaired assets ratio during the 1990s with movements in a number of macroeconomic variables that influence loan supply and demand.^[2] Following the unravelling of the credit boom of the late 1980s the overall level of Australian banks' impaired assets increased sharply in 1990 and 1991, reaching a peak in 1992. As banks wrote down the value of their bad loans the impaired assets ratio declined sharply. Subsequently, as the underlying condition of borrowers recovered, the ratio fell more gradually up until late 1997. Over the last few years, the industry-average impaired assets ratio has remained roughly constant at 0.7 per cent.

2.2.1 The demand for loan finance

Many studies show that bank performance is correlated with the business cycle (Lowe and Rohling 1993; Kaufman 1998). Firms' ability to service debt is most directly determined by their income, their gearing and the level of interest rates. In the first instance, the ability of firms to meet their debt obligations depends upon the share of their income taken up by interest payments. In Australia, the share of interest payments in corporate income and household income fell steadily over the first few years of the 1990s. In the latter half of the decade growth in corporate debt has been offset by lower interest rates, leaving the share of interest in income roughly unchanged. In contrast, the gearing of the household sector has grown more strongly since 1994. The peaks in the share of interest payments in household income coincided with peaks in interest rates in 1989 and late 1994.

All else equal, growth in aggregate income and output will strengthen firms' ability to meet their debt obligations. Calomiris, Orphanides and Sharpe (1997) outline three factors that exacerbate non-financial firms' sensitivity to cycles in aggregate activity: information asymmetries between borrowers and financiers; the advantages of investing during periods of rapid growth; and firms' tendency towards excessive optimism. These factors suggest that non-financial firms' performance (and banks' credit risk, in turn) will depend upon both the level and the rate of growth of aggregate activity.

Firstly, firms that rely on debt to expand their operations aggressively during periods of rapid economic growth are likely to be the least creditworthy when recession strikes. In the presence of information asymmetries between the managers and financiers of firms, debt contracts can go some way towards aligning managers' incentives with financiers' interests. This has the potential to reduce moral-hazard driven behaviour by firm management, adverse selection and monitoring costs. The firms that rely on debt (rather than equity) finance, therefore, are likely to be those for which asymmetric information problems are the most pronounced. These firms are likely to be most susceptible to slowdowns in economic growth.

Secondly, theoretical models of optimal investment strategies suggest that there are advantages in expanding rapidly. For example, Murphy, Shleifer and Vishny (1989) posit that growth in one sector of the economy has spillover effects by increasing demand for other sectors' output. Such considerations emphasise the advantages of investing during periods of rapid economic growth.

Thirdly, Calomiris et al (1997) argue that firms may not properly anticipate how aggregate economic circumstances may affect the value and liquidity of their assets. As a result, firms may have a tendency to be excessively optimistic regarding their ability to avoid financial distress and therefore, take on excessive leverage during periods of economic expansion. This view is in line with Minsky (1995) who characterises economic cycles as being driven by euphoric over-expansion of credit. Over a run of good times (characterised by minor cycles in economic activity) firms' and households' balance sheets change so that ever-larger proportions of their gross cash flows are committed to debt service. That is, preferences for leverage follow a cyclical pattern.

These three arguments suggest that, in the short term, stronger output growth will reduce banks' impaired assets, although over longer horizons this relationship may work in the opposite direction, with an acceleration in output growth leading to higher impaired assets. The sharp contraction in economic output in 1990–1991 coincided with the rise in impaired assets, whilst the sustained real growth of around 4 per cent since 1992 has been associated with the steady decline in impaired assets.

In Australia, there is a long history of slumps in building activity leading to banking problems (Kent and Lowe 1997). As the share of activity taken up by construction grows, therefore, the economy's overall credit quality is likely to decline. Particularly during the early 1990s a large proportion of banks' problem loans were associated with the financing of commercial property. The sharp peak in the share of construction in GDP in early 1990 led the peak in impaired assets. The spike in construction activity in 1998 was not, however, reflected in an increase in impaired assets. This reflected, in part, the fall in the share of commercial property finance provided by banks as listed property trusts took on a greater role.

Diamond (1991) suggests that real interest rates influence companies' choice between risky and safe projects. Low real interest rates increase the present value of firms' future profits. Since choosing more risky projects would put that future return at risk, as the expected future value of the company rises, the incentive to adopt low-risk projects increases. Thus, lower real interest rates are predicted to reduce the likelihood of default. This, in turn, reduces the riskiness of banks' loan portfolios. While the peak in the real interest rate preceded the rise in impaired assets in 1991 by more than a year, the increase in the real interest rate in 1994 was not associated with a commensurate rise in impaired assets.

Several recent studies (which are surveyed in Laker (1999)) have sought to identify those macroeconomic variables that best predict system-wide banking crises. Consistent with the arguments discussed above, the variables most often found to be associated with a high probability of banking crisis are low output growth, high real interest rates and strong credit growth.

2.2.2 The supply of loan finance

Supply-side developments also influence banks' riskiness and performance. Chief among these factors are agency costs, the regulatory environment, the competitiveness of the banking market and incentives for herding behaviour amongst banks.

The credit risk on any individual loan can be broken down into two components: the probability that the borrower will default, and the losses incurred in the event of default. The principal determinant of the losses incurred in the event of default is the value of security held as collateral against bank debt. Asset price deflation, by eroding the collateral against which banks lend, heightens financial institutions' vulnerability to borrowers' defaults. Reduced collateral values increase adverse selection problems as banks try to distinguish between sound and unsound borrowers. Increases in asset prices, by increasing the perceived collateral of potential borrowers, make financial institutions willing to supply a greater volume of funds at any given interest rate (Kiyotaki and Moore 1997). In Australia, the collateral for most loans is real estate. The collapse in commercial property prices in 1990–1991 was associated with the sharp increase in impaired assets. The subsequent recovery in commercial property prices has coincided with the steady improvement in the banks' impaired assets position.

Regulatory constraints, both prudential controls and those aimed at influencing macroeconomic activity, may directly constrain banks' risk-taking. Financial deregulation in Australia during the 1980s saw banks expand into areas that they previously would not have entered. There was both a rapid expansion of credit and a lowering of credit standards applied by banks (Macfarlane 1991). The Australian experience mirrored that of other countries that undertook similar programs of deregulation, most notably the Scandinavian countries, but also the US and UK. During the 1990s, however, the changes in the Australian banking industry's operating environment were less marked.

Changes in the level of competition within the banking market (particularly when driven by deregulation) may also generate system-wide movements in riskiness and performance. For instance, a more competitive environment may prompt individual institutions to seek to capture greater market share. While such expansion may be viable for one institution acting in isolation, when all banks behave in the same way such expansion sees increased lending to more marginal, risky borrowers (Drake and Llewellyn 1997). More generally, increased competition may also erode super-normal profits thus making banks' profits more sensitive to the underlying riskiness of their loan portfolios.

There are a number of models that suggest that it may be optimal for banks to adopt herding-type behaviour. Rajan (1994) provides one such model that links herding behaviour with cycles in credit growth and credit quality. This model is based on two stylised facts. Firstly, the market is seen to regard an individual bank's poor performance more leniently when the entire banking sector has been hit by an adverse shock. Thus there is an incentive for each to adjust its credit policy in line with other banks in the market. Secondly, a liberal credit policy, on the part of an individual bank, will boost current earnings at the expense of future earnings. Poor quality borrowers will meet their repayment obligations for at least a short time before becoming unable to service their loan obligations. Since in the short term expanding lending boosts earnings, the banks have an incentive to ease their credit standards in times of rapid credit growth (and likewise, to tighten standards when credit growth is slowing).

In the short term, increased credit growth, by adding to banks' total assets without immediately increasing impaired assets, would be expected to reduce the impaired assets ratio. Over the longer term, however, it would be expected that rapid credit growth, by increasing lending to more marginal borrowers, would lead to increases in the impaired assets ratio. The rapid credit growth of the late 1980s preceded the Australian banks' loan loss problems of 1990–1992. Since then, however, strong credit growth has accompanied the steady improvement in the impaired assets ratio.

Footnotes

Following the definition specified by the Australian Prudential Regulation Authority (APRA), impaired assets are taken to be the sum of non-accrual items, restructured items and assets acquired through security enforcement (for more details see Reserve Bank of Australia (1995)). The Reserve Bank revised asset quality measurement and reporting arrangements in September 1994. Prior to that, impaired assets were taken to be the sum of non-accrual items and accrual items in arrears 90 days and longer. Tests for a structural break in the data associated with the change in reporting arrangements find that the break is not statistically significant. [1]

Definitions of these macroeconomic variables are presented in Appendix A. [2]